article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
although the topic of spacetime wormholes was popularized in modern times by the seminal works of morris and thorne and visser @xcite , the original idea traces back to the bridge of einstein and rosen @xcite and even to the embedding diagram of flamm @xcite . in brief
, it is a hypothetical shortcut spacetime tunnel that connects vastly distant points belonging either to the same or different universes .
classical physics prohibits such travels due to instability unless an exotic matter source is taken for granted to support the tunnel .
similar to black holes wormholes are also exact solutions to einstein s field equations .
the idea of time travel through a wormhole , however , transcends classical considerations . to draw a rough analogy
we may refer to the art of escher @xcite , where in the same picture birds transmute into fishes etc . while this transmutation takes place in our minds , for the sake of art , physical theory of wormholes
demands far more than this kind of visualization . in brief
an observer can travel from one universe into the other through a traversable wormhole can also connect distant parts of the same universe . yet in this analogy we can say that in the realm of wormholes einstein meets escher .
wormholes demand physical transition between vastly separated points in warped spacetime in which curvature of spacetime plays the principal role in einstein s relativity and given the suitable energy - momentum such a travel becomes possible according to the laws of physics .
another interesting development took place recently in connection with wormholes : the einstein - rosen ( er ) bridge and the spooky interaction of quantum particles known as einstein - podolski and rosen ( epr ) pair may be related .
symbolically this situation has been summarized by @xmath2 @xcite , which may serve to connect wormholes with the realm of quantum theory . for these reasons we took wormhole physics seriously and attempted to construct these objects on physical , i.e. , non - exotic matter @xcite . to certain extent we obtained results that employ non - circular / non - spherical throat topology in the wormholes @xcite .
in addition to the energy matters recently we have also revised the well - known flare - out conditions @xcite . in this paper
we resort to the non - linear electromagnetism to provide a possible source for our traversable wormhole in @xmath0dimensions .
this is the square - root lagrangian of the maxwell invariant which breaks the scale invariance .
being a square - root expression our electromagnetic field is automatically pure magnetic , i.e. , we have @xmath3 as the only non - zero electromagnetic field component .
the energy density turns out to be exotic and under this condition we present exact wormhole solutions . in mt
the idea of a traversable wormhole is introduced and the flare - out conditions which every traversable wormhole must satisfy are also found . in accordance with @xcite the general line element of a traversable , circularly symmetric wormhole in @xmath0dimensions
is written as@xmath4 in which @xmath5 is called the redshift function and @xmath6 stands for the shape function of the wormhole . if we consider the location of the throat which connects two distant spacetimes , at @xmath7 , the flare - out conditions state that : i ) @xmath8 and ii ) @xmath9 where prime means @xmath10for @xmath11 although , in @xcite a throat is a gate between two asymptotically flat spacetimes ( i.e. , @xmath12 and @xmath13 ) this condition is not necessary due to the existence of non - asymptotically flat spacetimes as the solutions of the einstein s gravity coupled to different matter fields such as dilaton 2,3,4,5 .
therefore , the only constraint on @xmath14 is to be finite in the domain of @xmath15 having the einstein s equations and flare - out conditions all satisfied yields a negative energy density @xmath16 this can be seen from the @xmath17 component of the einstein s equation where @xmath18 ( @xmath19)@xmath20 from the line element ( 1 ) and the fact that @xmath21 one finds@xmath22from ( 2 ) one easily observes that with the flare - out condition fulfilled i.e. , @xmath23 the energy density becomes negative .
hence , the traversable wormholes are supported by exotic matter which violates the null energy condition
. wormholes in @xmath0dimensions , relatively , received less attention than the @xmath24dimensional counterparts @xcite .
it worths to mention that the first work on @xmath0dimensional wormholes was studied by perry and mann in @xcite . in this paper
we consider traversable wormholes in @xmath0dimensions supported by a nonlinear electrodynamic ( ned ) matter source .
the nonlinear maxwell s lagrangian which is employed in this study , namely the square root of the maxwell lagrangian is of the form given in @xcite which was developed further in @xcite .
versus @xmath25 and @xmath26 in cylindrical coordinates ( see eq .
we note that at @xmath7 , @xmath27 is where the throat lies . at the location of the throat the magnitude of the curvature
scalar i.e. @xmath28 is maximum while at large @xmath29 it goes to zero .
the negative energy density gets its maximum value also at the throat while at large @xmath30 it vanishes.,width=340 ] let s start with the line element@xmath31which we wish to call the morris - thorne type wormhole ( mttw ) in @xmath0dimensions .
note that in 3 + 1-dimensions such a wormhole was introduced by ellis @xcite .
herein , @xmath32 is a real parameter , @xmath33 @xmath34 and @xmath35 the ricci scalar of mttw@xmath36is clearly negative and the geometry is regular everywhere .
the ricci scalar admits an absolute / relative minimum located at @xmath37 while for @xmath38 it vanishes .
furthermore , the only nonzero component of the einstein s tensor is given by@xmath39which yields@xmath40where @xmath41 is the energy density of the matter , supporting the mttw .
it can easily be seen that @xmath41 behaves the same as @xmath42 such that a minimum occurs at @xmath43 upon taking a time slice of the spacetime ( 3 ) and embedding the result in cylindrical coordinates as@xmath44yields @xmath45 and @xmath46 these clearly imply that @xmath47 and @xmath48which is the same paraboloid of revolution as in the @xmath49dimensional mttw @xcite .
we note that @xmath50 is equivalent to @xmath51 in fig .
1 we plot @xmath52 in terms of @xmath53 and @xmath54 this figure supports the idea of having a throat located at @xmath27 corresponding to @xmath7 and therefore @xmath55 where @xmath56 to complete this section we add that a transformation of the form we introduced above , i.e. , @xmath57 helps us to find the more familiar form of the line element of the mttw as@xmath58which suggests that @xmath59 and @xmath60 in ( 1 ) .
once more we stress that in ( 9 ) , @xmath61 so that @xmath7 corresponds to @xmath37 which defines the location of the throat
. there is no need to state also that the flare - out conditions are perfectly satisfied .
let s start with the line element of a static and circularly symmetric spacetime given in ( 1 ) in which @xmath5 and @xmath62 depend only on @xmath30 .
we note that @xmath63 @xmath64 and @xmath65 .$ ] the action for gravity coupled to ned in @xmath0dimensions is given by ( @xmath66 ) @xmath67 in which @xmath42 is the ricci scalar , @xmath68 the cosmological constant , and @xmath69 stands for the nonlinear maxwell s lagrangian .
note that @xmath70 is a coupling constant and @xmath71 with @xmath72 is the maxwell s invariant .
let us remind that historically it was born and infeld @xcite who considered a non - linear version of electromagnetic lagrangian that survived to present time . in their approach
one could obtain the linear maxwell lagrangian as a limiting procedure . in our choice of square - root maxwell lagrangian , however , we shall have no such a limit .
once we let @xmath73 with @xmath74 we arrive at the btz @xcite black hole solution . for the choice @xmath75 ( with @xmath76 ) in 2 + 1-dimensions we recover nothing but the flat spacetime .
it should also be added that the original motivation of ned was to eliminate the divergences in electromagnetic field due to the point charges .
we comment that recently einstein s gravity coupled minimally to the nonlinear maxwell s lagrangian of the form @xmath77 , received attentions from different aspects @xcite . here
, we consider @xmath78 i.e. , @xmath79 with a pure magnetic field .
let us add that the particular power @xmath80 corresponds to the scale invariant case , i.e. invariance under @xmath81 and @xmath82 for @xmath83constant , in @xmath0dimensions .
our choice @xmath84 therefore breaks the scale invariance with physical consequences .
it should also be remarked that @xmath79 in flat spacetime had been studied long ago by nielsen and olesen @xcite in string theory while t hooft @xcite highlighted a linear potential term to be effective toward confinement .
our choice of the maxwell s 2-form is just a magnetic field of the form@xmath85 in which @xmath86 is a function of @xmath30 to be found .
breaking the scale invariance the lagrangian @xmath87 has the interesting property that it confines geodesics @xcite .
the source - free nonlinear - maxwell s equation for the specific lagrangian chosen , is given by@xmath88 in which@xmath89with its dual 1-form@xmath90hence , ( 12 ) yields@xmath91and consequently@xmath92 in which @xmath93 is an integration constant .
we note that , @xmath94 can be easily absorbed in time @xmath95 and therefore without loss of generality we set @xmath96 next , the einstein s equations with a cosmological constant @xmath68 are given by@xmath97 in which @xmath98upon ( 11 ) and ( 18 ) , one finds @xmath99 and @xmath100furthermore , with @xmath101 the only nonzero component of the einstein s tensor is @xmath102 in which a prime stands for the derivative with respect to @xmath103 we obtain as a result , @xmath75 in order to have the @xmath104 and @xmath105 components of the einstein - maxwell s equations satisfied .
next , we consider the @xmath17 component of the einstein - maxwell s equation which reads @xmath106this equation gives a relation between the magnetic field @xmath86 and the shape function @xmath107 in other words , a general class of solutions is determined by ( 21 ) such that the redshift function is zero while the shape function and the magnetic field satisfy the constraint @xmath108from this expression one finds that a possible throat is located at @xmath7 .
we note that ( 22 ) may or may not result in a wormhole .
for instance @xmath109 yields @xmath110 and the spacetime becomes flat .
hence , to have a traversable wormhole one should find specific function for @xmath111 such that the morris - thorne s flare - out conditions are satisfied . in the following sections we give two specific wormhole solutions . in the first example we consider the shape function to be of the form @xmath112 in which @xmath32 is the location of the throat . this shape function has been found above in mttw in @xmath0dimensions ( see eq .
having @xmath113 one finds the form of the magnetic field which is determined as@xmath114this is a singular function of @xmath30 such that at the location of the throat it diverges .
the maxwell invariant , however , @xmath115 is finite at @xmath116 in addition to that , at large @xmath30 the magnetic field vanishes to give an asymptotically flat limit for the wormhole .
the ricci and kritchmann scalars , respectively , are @xmath117and@xmath118which are clearly regular at the throat .
we wish to proceed now with the investigation of geodesic completeness @xcite in order to verify that divergence of the magnetic field at the throat is of no significance .
the geodesics lagrangian is @xmath119where a dot stands for derivative with respect to the proper time .
the first integrals of @xmath95 and @xmath120 equations are @xmath121the timelike geodesics ( @xmath122)give@xmath123the radial geodesics ( @xmath124 ) yields the hyperbolic curve @xmath125 in which @xmath126 it is observed that for @xmath127 we have @xmath128 which reflects the completeness of geodesics in the wormhole spacetime .
next , for @xmath129 we arrive at the expression@xmath130with the initial time constant @xmath131 this can be reduced to an elliptic integral form and naturally the geodesic completeness is valid here as well . finally , the tidal forces at the throat can be analyzed through the geodesics deviation equation@xmath132where @xmath133 , ( @xmath134 ) are displacements along the radial and angular directions .
one obtains@xmath135and@xmath136which indicate the finiteness of tidal forces in the vicinity of the wormhole throat . as a second example we consider the shape function to be of the form @xmath137 in which @xmath138 is a free , real parameter .
note that in order to have the flare - out conditions satisfied we must impose @xmath139 the case @xmath140 has already been considered in our example a. among other possibilities , we consider @xmath141 which yields @xmath142 consequently , the magnetic field and the energy density become@xmath143and @xmath144it should be stressed here also that the maxwell invariant in the present case is @xmath145 which is regular at the throat .
one finds that the proper distance takes the form@xmath146\]]and the shape function is@xmath147the ricci and kretchmann scalars become now@xmath148and@xmath149which imply that the throat is a regular hypersurfacet .
given the analysis of the previous section it is not difficult to anticipate that the tidal forces / accelerations are finite in this generalized mttw model as well .
due to the power @xmath138 however the integrals of geodesics will not be any simpler .
we constructed a class of traversable wormhole solutions in the theory of gravity coupled to nonlinear electrodynamics in @xmath0dimensions . a similar model of wormhole with an anisotropic fluid source
was considered in culetu . for specific choice of the shape
function the solution is the morris - thorne type wormhole in @xmath0dimensions which shares most of its properties with its @xmath49dimensional version .
the matter source which supports our wormhole solution is a pure magnetic field of the form given in ( 22 ) .
the square - root of pure magnetic maxwell lagrangian provides automatic satisfaction of the non - linear maxwell equation in @xmath0dimensions .
confining of geodesics is another interesting property of such a square - root lagrangian @xcite .
we comment that the magnetic field diverges at the throat and vanishes fast with @xmath150 a particularly simple example with @xmath151 is considered . in this example also the magnetic field diverges at the throat while the maxwell invariant @xmath152 is finite at the throat .
next , we consider a more general ansatz which involves an arbitrary parameter @xmath153 we note that divergence of the magnetic field at the throat was used in lobo as a counter argument against existence of such 2 + 1-dimensional wormholes .
the only singularity of the problem lies at @xmath154 which is a naked spacetime singularity but since the wormhole condition stipulates that @xmath155 the singularity at @xmath154 where the scalar curvature invariants diverge remains ineffective for particle geodesics .
finally , we should add that the class of solutions found in this paper consists of a large number of solutions which only depends on the form of @xmath111 .
any choice of @xmath111 satisfying the flare - out conditions acts as a generator and gives rise to a new morris - thorne type wormhole .
the fact that we work in the reduced @xmath0dimensions simplifies the problem to a great extend . in @xmath49dimensions
obviously wormholes ca nt be generated from a single throat function . | in the context of @xmath0dimensional gravity coupled to a particular nonlinear electrodynamics ( ned ) , we obtain a class of traversable / morris - thorne type wormhole solutions .
the problem is reduced to a single function dependence in which the shape function acts as generator to the wormholes .
the field ansatz is pure magnetic and the nonlinear lagrangian is @xmath1 i.e. the square root of the maxwell lagrangian . in @xmath0dimensions the source - free pure magnetic non - linear maxwell equation with square - root lagrangian is trivially satisfied .
the exotic energy density is found explicitly and the flare - out conditions are emphasized . |
after the cmb photons decoupled at redshift about @xmath0 , the intergalactic medium ( igm ) remained neutral until the first generation of luminous sources produced ionizing photons .
recent measurements of the spectra of high redshift quasars in the sloan digital sky survey ( @xcite ) and of polarization anisotropies in the cosmic microwave background ( cmb ) by the _ wilkinson microwave anisotropy probe ( wmap ) _ ( @xcite ) indicate that the igm was reionized during redshifts @xmath1 .
the history of cosmic reionization contains abundant information about the formation of the first cosmic structures , and can be probed in various future observations , such as redshifted 21 cm signatures ( @xcite ) , the kinetic sunyaev
zeldovich effect ( @xcite ) , improved measurements of the large angle cmb polarization fluctuations ( @xcite ) , the thermal state of the intergalactic medium ( @xcite ) , and lyman @xmath2 galaxy populations ( @xcite ) .
being directly related to many observables , the distribution of the hi / hii regions during reionization has been studied extensively using both semi - analytic models and hydrodynamic simulations ( see , e.g. , the recent review by choudhury & ferrara ( 2006 ) , and references therein ) .
if the spectrum of the first ionizing sources is soft ( such as a normal stellar spectrum ) , then the distribution of the ionized regions can be described by discrete hii bubbles around high density regions ( @xcite ) . on the other hand ,
if the source spectrum is hard and extends to x ray energies ( such as for miniquasars ) , the photons can more readily escape into the igm , and may ionize the low density regions first ( @xcite ) .
the reionization topology in the above two limiting cases are often referred to as `` inside out '' and `` outside in '' respectively .
we will adopt this distinctive terminology here , and we will use it to describe the ionization topology on arbitrarily large scales .
existing analytic / semi - analytic models have generally described either one or the other of these two limiting scenarios ( i.e. by following the filling factor of ionized bubbles , or by assuming a uniformly rising ionizing background ) .
the previous methods used to implement such models can not be easily generalized to derive statistics of the ionization topology , as a continuous function of the hardness of the source spectra . _
the goal of the present paper is to construct a model of inhomogeneous reionization in which we can quantify properties of the ionization topology for sources with different spectral hardness and clustering properties .
the hope is to derive from first principles , rather than assume , the large scale ionization topology and statistics . _
more generally , the nature the spectrum , spatial clustering , and evolution of the ionizing source population is poorly known at present , and future observations ( as mentioned above ) promise to shed much light on these quantities .
parameterized models , such as the one presented here , will be useful for interpreting these future observations . in principle ,
hydrodynamic simulations offer an alternative and more reliable way of studying the topology of reionization .
however , it is difficult for the state - of - art simulations to cover large scales ( @xmath3mpc ) while resolving the sources and the gas distribution on small scales .
recently , several groups have attempted to study the large scale properties of reionization using large scale simulations ( @xcite ) .
kohler et al .
( 2005 ) for instance adopt a hybrid approach which combines small scale and large scale simulations . as will become clear below
, our analytic treatment is in some sense similar : we use linear perturbation theory to address large scale questions while taking into account the effects of small scale clumping in the background evolution . even if in the future simulations that simultaneously resolve the necessary small scale structure and span over large scales were to become available
, we would still need an analytic framework to better understand the physics of reionization .
furthermore , the ionizing sources have to be inserted into the simulations , and their properties specified , essentially by hand . in practice , it is likely that the intepretation of future data on the ionization topology will require an exploration of various parameters of the sources that can not be computed ab
this , in turn , will require a semi analytic , computationally less expensive model . in a hierarchical universe such as our own ,
one expects linear perturbation theory to work on sufficiently large scales .
the spatial fluctuations in the ionized / neutral hydrogen and in the ionizing radiation can be related to the dark matter distribution via bias factors which are scale dependent in general . for a given source distribution ,
these bias factors are determined by the radiative transfer equations and the equation of the photo - ionization balance .
the approach followed in this paper is to solve the linearized versions of these equations in fourier space .
given the source distributions and spectra , this approach allows us to calculate the linear biases of the ionized / neutral hydrogen and of the ionizing radiation , and to follow their evolution .
the price paid is that our predictions for the spatial fluctuations are invalid on small scales .
our calculation takes into account all of the relevant physical processes , including photo - ionization , recombination , the diffusion of photons , the peculiar velocity of the baryons , and the redshifting of photons due to the expansion of the universe .
it is worth noting that in the existing semi - analytic models , typically only photo - ionization and recombination are treated in an exact manner .
most of the other processes are either missing or treated approximately .
our formalism accounts for all the relevant processes ( albeit in a perturbative manner ) and allows a comprehensive study of the dynamical evolution of the hi / hii regions and the radiation field with a general source spectra .
it is our hope that the approach taken here can be developed further in future work , and will ultimately make it possible to constrain the high redshift source properties using future 21 cm and cmb observations .
our approach is also useful for addressing the question of what robust cosmological information one can obtain from future 21 cm and cmb observations ( for instance , to what extent are the various linear biases scale independent ) .
the rest of this paper is organized as follows . in
[ formalism ] , we introduce our formalism , and present techniques for solving the equations . in [ sources ] , we discuss our assumptions about the ionizing source population , based on the extended press - schechter model and assuming different source spectra . in
[ results ] , we present our main results , i.e. the evolution of the ionization fluctuations on different scales , and for different source spectra . in [ discussion ] , we discuss various caveats , and possible extensions of the present paper to future work . finally , in [ conclude ] , we summarize our conclusions and the implications of this work . appendix a presents an analytic solution ( up to the numerical solution of a simple integral equation ) to the first order radiative transfer and ionization equations .
appendix b contains a preliminary discussion of the validity of linear perturbation theory . throughout this paper ,
we adopt a standard @xmath4cdm cosmological model , with the parameters @xmath5 , @xmath6 , @xmath7 , @xmath8 , @xmath9 , @xmath10 , and @xmath11 , favored by the combination of the three - year wmap data and the weak lensing data of canada - france - hawaii telescope legacy survey ( cfhtls ) ( see @xcite ) .
we begin with definitions of the physical quantities and descriptions of relevant equations in [ definition ] .
the equations are solved by splitting into two pieces : a spatially averaged piece in [ homo ] and a ( first order ) fluctuating piece in [ linear ] .
let us first define the relevant physical quantities : @xmath12 where @xmath13 and @xmath14 are the comoving number densities of the ionized and total hydrogen atoms ( ionized+neutral ) respectively .
they are defined as functions of the comoving coordinate @xmath15 and the conformal time @xmath16 . here
, @xmath17 refers to the comoving photon number density per unit solid angle @xmath18 around the propagation direction @xmath19 and per unit frequency parameter @xmath20 .
the frequency parameter @xmath20 is defined as : @xmath21 where @xmath22 is the photon frequency , @xmath23 is the ionization threshold of hydrogen , and @xmath24 is the planck constant .
the frequency parameter will turn out to be more convenient to use than the frequency itself .
the quantity @xmath25 is the differential ionizing emissivity , which gives the number of photons emitted by sources per unit comoving volume , per unit conformal time , per unit frequency parameter @xmath20 and per unit solid angle .
it is useful to relate @xmath26 and @xmath27 here to perhaps more familiar quantities .
the proper specific intensity @xmath28 of the ionizing radiation is related to @xmath26 via @xmath29 where @xmath30 is the scale factor .
note that we set the speed of light to unity throughout this paper .
the quantity @xmath27 is related to the proper emissivity @xmath31 as usually defined via @xmath32 throughout this paper , for simplicity , we will ignore the presence of helium atoms , with @xmath33 the hydrogen mass fraction . ] .
this could affect our results somewhat for the hard
spectrum cases considered below ( since most of the @xmath34 ev photons will be absorbed by hei , rather than hi ) .
we postpone the study of helium reionization to future work .
taking into account peculiar velocities , photo ionization and recombination , the equation for ionization equilibrium is given by @xmath35 where @xmath36 is the cosmological scale factor ; @xmath37 is the comoving velocity of the ionized hydrogen atoms ; @xmath38 is the photo - ionization cross section ( osterbrock et al .
2005 ) ; and @xmath39 is the case b recombination coefficient at temperature equal to @xmath40 .
typical of the photoionized intergalactic medium ( where recombination is relevant ) , the recombination coefficient scales roughly as @xmath41 .
the temperature is in turn related to gas density to some power , with the power index ranging from about @xmath42 to @xmath43 ( @xcite ) .
overall , the spatial fluctuation of the recombination coefficient in ionized regions due to the fluctuation in density ( and therefore temperature ) is rather weak , and is ignored here .
] we have implicitly assumed electric neutrality everywhere in the universe , and we are ignoring the electrons that would result from the ionization of helium , therefore the recombination term is proportional to @xmath44 .
the factor @xmath45 is included to account for multiple ionizations by an x - ray photon through secondary ionizations by the fast photoelectrons . here
@xmath46 is the local ionization fraction , and we adopt the values of @xmath47 , @xmath48 , and @xmath49 in the fitting formula above ( according to @xcitekev ) . ] ) .
the evolution of the radiation background is affected by the sources , the photo - ionization process , the diffusion of photons in eq.[[mainfgama ] ] , and does not imply scattering .
this is a very loose usage because this is not the usual term in the diffusion equation . ] and the redshifting due to the expansion of the universe , all of which are reflected in the following radiative transfer equation ( e.g. @xcite ) , @xmath50 where @xmath51 is the hubble parameter .
our main task is to solve eq.[[mainfe ] ] and eq.[[mainfgama ] ] . for our purpose ,
it is useful to rewrite the definitions in eq.[[define1 ] ] in terms of the spatial averages and the perturbations : @xmath52\\ \nonumber & = & \bnh [ \fhii(\tau)+\dhii ( \xtau)]\\ \nonumber \\ \nonumber \nh & = & \bnh [ 1+\delta(\xtau)]\\ \nonumber \\
\nonumber \ngam & = & \bnh\fgam ( \tau,\mu)[1+\dgam ( \xtaumuo)]\\ \nonumber & = & \bnh [ \fgam ( \tau,\mu)+\dgam ( \xtaumuo)]\\ \nonumber \\ \nonumber s&=&\bnh f_s(\tau,\mu)[1+\delta_s(\xtaumuo)]\\ \nonumber & = & \bnh [ f_s(\tau,\mu)+\delta_s(\xtaumuo)]\end{aligned}\ ] ] where @xmath53 and @xmath54 are the mean number densities of ionized hydrogen atoms and photons respectively , and @xmath55 is the mean source emissivity all normalized by the mean comoving total ( i.e. ionized plus neutral ) hydrogen number density @xmath56 , which is a constant in time .
@xmath57 , @xmath58 , @xmath59 and @xmath60 are the corresponding overdensities .
@xmath61 , @xmath62 and @xmath63 are introduced because they greatly simplify the following calculations .
we will assume below that the ( total ) hydrogen fluctuations faithfully trace the dark matter fluctuations ( with no bias ) , which is justified on scales well above the jeans scale ( e.g @xcite ) .
note that according to the above definitions the neutral hydrogen density is given by @xmath64 \nonumber\end{aligned}\ ] ] and @xmath65 taking the spatial ( and angular ) averages of eq.[[mainfe ] ] and eq.[[mainfgama ] ] , we find : @xmath66 @xmath67 where @xmath68 and @xmath69 are the clumping factors for photo - ionization , defined as : @xmath70 @xmath71 is the clumping factor for recombination . to solve for @xmath53 and @xmath54
, one needs to know the evolutions of @xmath68 , @xmath69 , and @xmath72 . since the clumping factors are likely dominated by non
linear density variations on small scales ( e.g. haiman , abel & madau 2001 ) their values can not be computed reliably in our present framework .
we assume , for simplicity , that the clumping factors are known from small scale hydrodynamic simulations .
note that recent large scale radiative transfer simulations effectively adopt the same assumption ( e.g. @xcite ) . instead of @xmath16 ,
let us use @xmath73 as the time variable . to further simplify the notation
, we define @xmath74 and @xmath75 , which can be interpreted as follows : @xmath76 is the probability that a photon of energy @xmath20 experiences a direct photo - ionization in hubble time if the universe is neutral ; @xmath77 is the average number of recombinations a proton would experience in a hubble time if the universe is completely ionized .
eq.[[homofe ] ] and eq.[[homofgama ] ] can then be rewritten as : @xmath78 @xmath79 eq.[[homofenew ] ] and eq.[[homofgamanew ] ] can be solved iteratively .
to do so , we need to use a trick which also appears in [ linear ] .
first , let us change the variables from @xmath80 to @xmath81 which are defined as : @xmath82 therefore @xmath83 eq.[[homofgamanew ] ] can be transformed into : @xmath84 or in a simpler form as : @xmath85 where @xmath86 and @xmath87 .
eq.[[fgamma2 ] ] is a standard first order differential equation .
its solution reads : @xmath88\ ] ] or in terms of the variables @xmath80 as : @xmath89\end{aligned}\ ] ] the above solution assumes that there are no ionizing photons at arbitrarily early times and/or arbitrarily high energies . using eq.[[fgamma5 ] ] , we can obtain @xmath90 for an initial @xmath91 .
@xmath54 can then be used to calculate a new @xmath53 using eq.[[homofenew ] ] . in practice
, we find that if this procedure is repeated , @xmath54 and @xmath53 converge to their true values after around ten iterations .
eq.[[fgamma5 ] ] presents a useful picture of physics : it shows that at the early stage of reionization ( when the igm is not highly ionized ) , the low energy photons ( but above @xmath92ev ) have a short memory of their past history because they are quickly absorbed by neutral hydrogen .
this can be seen from the largeness of the exponent in eq.[[fgamma5 ] ] when the cross section is large .
in contrast , the hard photons retain a long memory . another way of saying the same thing is that the soft photons have a short mean free path compared to the hard photons .
this fact leads to a much simpler way of solving eq.[[homofenew ] ] and eq.[[homofgamanew ] ] when the average ionized fraction @xmath53 is not close to one .
first of all , we notice that in the low energy limit , eq.[[fgamma5 ] ] reduces to a simple form : @xmath93 secondly , we assume there is a critical frequency parameter @xmath94 , above which the photons hardly ionize any neutral hydrogen during reionization , and below which eq.[[smallmu ] ] is valid . therefore , eq.[[homofenew ] ] can be rewritten as : @xmath95 eq . [ [ smallmu ] ] essentially describes an emission - absorption equilibrium , i.e. all emitted photons with @xmath96 are consumed by ionization .
it should be noted that if the source spectrum is hard ( high percentage of photo - ionization is caused by high energy photons . ) , eq .
[ [ smallmu ] ] and [ [ homofe_approx ] ] are not a good approximation anymore .
this will be discussed further in [ linear ] and [ results ] .
note that throughout this paper , we compute the exact solutions rather than employ the approximations outlined above , though we will compare the two in [ results ] .
let us denote the fourier transforms of @xmath57 , @xmath63 , @xmath61 and @xmath62 as @xmath97 , @xmath98 , @xmath99 and @xmath100 respectively .
the fourier transforms of eq.[[mainfe ] ] and eq.[[mainfgama ] ] to the first order are : @xmath101 @xmath102 where @xmath103\\ \nonumber g&=&\frac{d\ln d}{d\omega}\fhii\\ \nonumber & + & 4\pi \int_0^{\infty}d\mu\tsigma\fgam\left[\langle\kappa\rangle-(1-\fhii)\fhii\left .
\frac{\partial\kappa}{\partial\phi}\right|_{\phi=\fhii}\right]\\ \nonumber b&=&(1-\fhii)\tsigma\\ \nonumber m&=&(1-\fhii)\tsigma-\frac{i\vec{k}\cdot\vec{\omega}}{ha}\\ \nonumber n&=&(4\pi ha)^{-1}\\ \nonumber r&=&\tsigma\fgam\end{aligned}\ ] ] and @xmath104 is the wave vector . by @xmath105
] for simplicity , we do not show the @xmath106 dependence explicitly for the fourier modes . in deriving eq.[[flinearfe ] ] ,
we have used the fact that to the first order , on large scales , the dark matter overdensity grows linearly with a growth factor defined by @xmath107 , and the peculiar velocity is proportional to the gradient of the gravitational potential .
the great advantage of the fourier transformation is that it allows us to solve eq.[[flinearfgama ] ] using tricks similar to those introduced in the previous section , : @xmath108\}\\ \nonumber \times \exp[-\int_{\omega'}^{\omega}d\omega''m(\omega'',\mu+\omega-\omega'',\vec{\omega})]\end{aligned}\ ] ] the first line above describes the contributions to fluctuations in radiation from fluctuations in the spatial distribution of the ionizing sources ; the second line describes the contributions from fluctuations in absorbers ; the third line accounts for the modulation by optical depth , i.e. it tells us the distance to which one needs to integrate . by integrating eq.[[solutionfgama ] ] over @xmath19
, one can obtain the monopole perturbation of the radiation field , which is what is needed in eq.[[flinearfe ] ] : @xmath109\}\\ \nonumber & \times&\exp[-\int_{\omega'}^{\omega}d\omega''b(\omega'',\mu+\omega-\omega'')]\\ \nonumber & \times&\frac{\sin[p(\omega,\omega')k]}{p(\omega,\omega')k}\end{aligned}\ ] ] where @xmath110 in writing down the above expression , we have assumed that the dominant contribution is from the monopole of @xmath98 ( note that we have dropped the argument @xmath19 to signify the fact that this is the monopole component ) .
note that we are _ not _ assuming @xmath111 has no higher multipoles ; rather , we are assuming that , as far as the source contribution is concerned , the monopole of @xmath111 is dominated by the monopole of the source emissivity .
this assumption can be motivated in two different ways .
first , in the low @xmath106 limit ( which is the regime where perturbation theory probably works best ) , one can show that the monopole dominates .
second , on large scales , since one is averaging over many sources , the resulting smoothed emissivity is probably close to isotropic even if the individual sources are not .
[ [ angleavefgama ] ] , together with eq .
[ [ flinearfe ] ] , allows for an iterative solution for @xmath99 and @xmath100 given the source distribution @xmath98 and the dark matter overdensity @xmath97 . in appendix
a , we present a further improved scheme of finding the numerical solution . similar to what we have done in [ homo ] , when the universe is not close to being fully ionized , eq.[[angleavefgama ] ] for soft photons can be greatly simplified .
first , we notice that the factor @xmath112/[p(\omega,\omega')k]$ ] in eq.[[angleavefgama ] ] should be very close to unity for soft photons .
there are two reasons for this : one is because the wave number @xmath106 of interest is small ; the other reason is that the large exponent @xmath113 limits the magnitude of @xmath114 to be very small in the integration . taking these into account , eq.[[angleavefgama ] ] becomes much simpler : @xmath115\}\end{aligned}\ ] ] the above expression is the analog of eq . [ [ smallmu ] ] , representing essentially emission - absorption equilibrium . using the same critical frequency parameter @xmath94 to isolate the contributions from the soft photons as in eq.[[homofe_approx ] ] , and assuming eq.[[angleavefgama2 ] ] is correct for such soft photons , we obtain : @xmath116 where @xmath117 again , eq.[[flinearfe2 ] ] is a good approximation when the source spectrum is not too hard . in [ results ] , we quantify this statement and discuss the choice of @xmath94 with realistic examples .
we reiterate that all of our conclusions in this paper are based on the exact solutions rather than the approximations outlined above , though we do compare the two in [ results ] . at this point
, the reader might wonder : since reionization is likely a complicated process with large fluctuations on small scales , could these fluctuations invalidate linear perturbation theory on large scales ?
in other words , how should one justify the use of perturbation theory on large scales ?
this is actually a difficult question .
the same question arises in the context of large scale structure formation : how do we justify the use of linear perturbation theory on large scales today , knowing that there are highly nonlinear structures on small scales , such as galaxies , clusters and so on ? some plausibility arguments exist ( @xcite ) , and we will discuss the analogs of these arguments for reionization in appendix b. however , the only rigorous method of valdiation that we know of is to appeal to numerical simulations .
we will perform comparisons of our calculations with numerical radiative transfer simulations in another paper .
at least one difference between reionization and large scale structure is , however , worth emphasizing . in the case of reionization ,
the spatial averages are undoubtedly affected by small scale clumping hence the need to introduce clumping factors in eq .
[ [ homofe ] ] and [ [ homofgama ] ] . in some sense
, our treatment here is quite similar to some of the recent numerical simulations ( @xcite ) which incorporate small scale clumping by hand while focusing on the large scale fluctuations . in the case of large scale structure ,
the spatially averaged equations are those that govern the global expansion of the universe i.e. the friedmann equation and energy - momentum conservation . in that case
, small scale clumping appears not to affect significantly the evolution of the spatial averages ( e.g. @xcite ; see however @xcite for a different view ) .
solving the equations described in the previous section requires specifying the source properties , including the emissivity as a function of redshift , and the spatial distribution and spectrum of the ionizing sources . in this paper
, we use the extended press - schechter theory ( @xcite ) to model the abundance and spatial distribution of the ionizing sources .
it is important to emphasize that this is for illustration only our formalism as laid out in the previous section is certainly not wedded to this particular model of the ionizing source population .
the minimum halo that can host luminous sources should have a virial temperature above @xmath118k to allow efficient hydrogen line cooling ( see , @xcite ) .
this leads to a minimum halo mass given by : @xmath119 where @xmath120 is the mean molecular weight , which is chosen to be @xmath43 ( appropriate for the ionized gas in halos with a virial temperature above @xmath118k ) in the following calculations .
according to the extended press - schechter model , on mass scale @xmath121 , the fraction of mass collapsed in halos with masses larger than @xmath122 is : @xmath123}}\right]\ ] ] where @xmath124 is the critical overdensity in the spherical collapse model ; @xmath125 and @xmath126 are the overdensity and the variance of the density fluctuations on mass scale @xmath121 respectively ; @xmath127 is the density variance corresponding to the mass scale @xmath122 . on large scales
, @xmath126 is much smaller than @xmath127 and therefore neglected in the following calculations . assuming on average each hydrogen atom in the collapsed objects emits @xmath128 ionizing photons per unit frequency parameter @xmath20 , the emissivity function smoothed over scale @xmath121 is given by : @xmath129\ ] ] the factor @xmath130 takes into account the mass overdensity .
this equation also implicitly assumes that the collapsed objects produce their photon output on a time - scale much shorter than @xmath131 $ ] ( which is justified for short lived , massive stars , or efficiently accreting black holes ) .
if we choose the scale @xmath121 to be very large ( the horizon size ) , @xmath125 can be neglected , and from eq.[[emissivity ] ] we obtain the spatially averaged emissivity function defined in [ formalism ] as : @xmath132 by taylor expanding eq.[[emissivity ] ] around @xmath133 and then performing a fourier transformation , we obtain the spatial fluctuation of the emissivity as : is dropped . ]
@xmath134\\ \nonumber & = & \gamma(\mu)\frac{\tilde{\delta}(\vec{k},\tau)}{d(\tau)}\frac{\partial}{\partial \tau}\left[r(\tau)d(\tau)\right]\end{aligned}\ ] ] where @xmath135 and @xmath107 is the linear growth factor . in our model
, we do not take into account the feedback effects discussed in recent papers ( @xcite , also see haiman & holder 2003 for a general reference for feedback effects during reionization ) .
for example , the ionized regions usually have relatively higher temperatures due to photo - heating , leading to a larger filtering scale ( @xcite ) that suppresses the formation of small scale structures .
this reduces the clumping factor and hence the recombination rate , and also results in an anti - correlation between the source overdensity and the local ionized fraction . in principle
, such a feedback effect can be included in the linear perturbation calculation by inserting into eq .
[ [ taylore ] ] a term that is proportional to @xmath99 .
we leave a more careful study of such effects to the future . in our calculation
, the normalization of the source emissivity is chosen to give a mean optical depth of thomson scattering equal to @xmath136 , which is suggested by the three - year wmap data .
the remaining freedom is to choose the source spectrum , which depends on the type of sources ( see , , @xcite ) . rather than exhausting spectra of general shapes ,
we focus on spectra of power - law forms with different spectral indices , which cover the possible range of effective spectra of stars and miniquasars . a power - law form in the frequency @xmath22
is an exponential form in the frequency parameter @xmath20 , : @xmath137d\mu\ ] ] where @xmath138 is the spectral index , @xmath139 is a normalization factor , and @xmath140 is the total number of ionizing photons generated per baryon in stars , and managing to escape into the igm .
for example , if 10 percent of the baryons turn into stars with a normal salpeter mass function , and 10 percent of their ionizing radiation escapes , then @xmath141 . the spectrum is smoothly cut off at @xmath142 ( @xmath143 kev ) to allow for a proper normalization even when @xmath144 .
the cutoff does not introduce any artificial effects because the mean free path of photons at this energy greatly exceeds the hubble distance .
it is worth noting that in a more complicated scenario , the shape of the source spectrum may vary with redshift .
for example , if quasars are the dominant ionizing sources at high redshifts ( @xmath145 ) , and followed by stars at @xmath146 , the hardness of the source spectrum varies with time .
such a case will be studied in a future paper .
as we have discussed in [ homo ] , the spatial averages of the ionized fraction and the radiation intensity can be calculated by solving eq.[[homofenew ] ] and eq.[[homofgama ] ] once the clumping factors @xmath68 , @xmath69 , and @xmath72 are provided .
according to the existing hydrodynamic simulations , @xmath72 is around order of ten during reionization for a uv dominated source spectrum , but can be less than one for a harder source spectrum . for simplicity , we ignore the time dependence of the clumping factors in this paper . ] .
the values of @xmath68 and @xmath69 vary with both the redshift and the photon energy , but are rarely far from unity ( @xcite ) . for simplicity , in this paper , we choose @xmath147 for @xmath148 or @xmath149 , @xmath150 for @xmath151 , and @xmath152 for all cases .
we caution that the precise values of the clumping factors are rather uncertain .
the important point to keep in mind is that the clumping factors show up explicitly only in the equations for the spatial averages but not for the fluctuations . in other words
, they affect the linear perturbations only indirectly through their effects on the background ( i.e. @xmath153 and @xmath154 ) . once the background is specified , our predictions for the linear perturbations are quite robust .
this is discussed further in
[ discussion ] below . to illustrate the results , we choose three types of source spectra : @xmath155 , @xmath156 and @xmath157 .
we show the evolution of the spatially averaged ionization fraction @xmath53 in fig .
[ fefig ] , the hii bias ( @xmath158 ) at @xmath159 in fig .
[ dfefig ] , and the hii bias as a function of the scale @xmath106 at redshift @xmath160 in fig .
[ dfe_kfig ] . from fig .
[ dfefig ] , we see that in the case of a soft source spectrum ( @xmath148 and @xmath149 ) , the bias of the hii regions remains at a high value during most time of reionization , and quickly drops to one when the hii regions percolate the igm .
this can be understood in the usual bubble picture , in which the hii regions are confined within the hii bubbles due to the short mean free path of the photons .
therefore , the hii regions do not merge on large scales until the mean ionization fraction of the igm reaches a very high level .
in contrast , for a harder source spectrum , the bias of the hii regions decreases in a more continuous fashion .
interestingly , this is not only because hard photons have a much longer mean free path , but also due to the fact that secondary ionizations are more intense in less ionized regions .
the hii bias is intimately related to the issue of inside - out versus outside - in ionization .
a reasonable definition of inside - out ionization is that @xmath161 , where @xmath162 is the fluctuation in the ionized fraction .
a positive correlation means higher density regions exhibit a higher ionized fraction .
conversely , a negative correlation can be taken as the definition of outside - in ionization . to the lowest order , @xmath163
therefore , the sign of @xmath164 is determined by whether the hii bias is greater or less than unity .
it is worth noting that whether reionization is inside - out or outside - in can be a scale dependent question : the sign of @xmath165 could depend on the scale of interest ( think of @xmath57 and @xmath162 as quantities smoothed on some scale , or think of their fourier counterparts ) . an important feature we learn from fig . [ dfefig ]
is that the large scale bias of the hii regions is larger than unity in all three cases , which therefore means the high density regions tend to be more ionized than the low density regions .
this fact shows that at least on large scales , the topology of the hii regions is inside - out , even for a hard source spectrum .
this seems to hold on all scales that we have examined ( see fig .
[ dfe_kfig ] ) , although one must note that our perturbative approach is expected to break down on sufficiently small scales . by choosing @xmath166 ( @xmath167ev )
, we find that eq.[[homofe_approx ] ] and eq.[[flinearfe2 ] ] are good approximations for @xmath53 and @xmath168 on large scales ( @xmath169 ) during most of the reionization process for the @xmath148 and @xmath149 cases , as shown in fig .
[ fefig ] and fig .
[ dfefig ] .
the reason for choosing this value of @xmath94 is not only that it provides a good fit for both @xmath53 and @xmath168 , but also because @xmath170 corresponds to a critical energy threshold , below which photons are significantly absorbed by neutral hydrogen , as shown in fig .
[ fgama_mu ] is in logarithmic scale .
we do not show labels on the y - axis because they are not important for our purpose . ] .
in other words , eq . [ [ smallmu ] ] works well for photons with energies below @xmath170 during the early stages of reionization . for smaller scales , the agreement becomes worse because the photon propagation suppresses the small scale fluctuations , thus reducing the amplitude of the ( hii to dark matter / baryon ) bias .
this is shown in fig .
[ dfe_kfig ] .
a useful way to see why the bias of hii decays faster for a harder source spectrum is to calculate the mean free path of the ionizing photons . for a specific frequency parameter @xmath20 ,
the comoving mean free path is simply @xmath171 $ ] .
the average mean free path can be defined as : @xmath172^{-1}\ ] ] using eq.[[smallmu ] ] for three different source spectral indices , we find : @xmath173 these numbers basically answer the question of why in fig .
[ dfefig ] the hii bias in the @xmath148 or @xmath149 case does not drop until the ionization fraction is very close to one .
this is because in both cases the wave length of @xmath159 is larger than the average photon mean free path during most time of reionization .
radiative transfer is unable to suppress perturbations on scales larger than the mean free path , and the hii bias more or less tracks the source bias .
in contrast , hii perturbations on smaller scales are more easily suppressed by the radiation , which can be seen in fig .
[ dfe_kfig ] as the suppression in @xmath174 for @xmath175 .
however , we caution that our linear theory becomes inaccurate on small scales .
we also find that other ways of estimating the average photon mean free path only change the results by a factor of a few , meaning that eq.[[meanfreepath ] ] provides a robust order of magnitude estimate .
for example , if secondary ionization is included in eq.[[meanfreepath ] ] , the average mean free path in all thre cases is increased by no more than a factor of two . in fig .
[ dhifig ] , we show the redshift dependence of the fluctuation of the neutral fraction with respect to the dark matter ( @xmath176 ) at @xmath159 , which is proportional to the 21 cm signal from the neutral hydrogen ( see @xcite ) . at the early epoch of reionization
, this quantity is close to unity since most of the igm is still neutral .
the non - monotonic behavior at lower redshifts may suggest a best window for detecting the 21 cm emission .
such a feature is more pronounced when the source spectrum is soft .
the bias of the radiation monopole ( @xmath177 , which is simply called @xmath178 in the figures ) on scales of @xmath159 and @xmath179 is shown in fig .
[ dfgama_mu ] as a function of the photon energy for different redshifts , and in fig .
[ dfgama_time ] as a function of redshift for different photon energies .
again , one can see a clear difference between the soft and the hard photons , which are divided by the critical line at @xmath167ev .
the bias of the soft photons remains high until the hii percolation , following closely the bias of hii itself , at least during the early stages of reionization . on the other hand ,
the high energy photons diffuse relatively freely into the space , leading to an ever decreasing bias .
an interesting feature in fig .
[ dfgama_time ] deserves a brief discussion : the bias of the radiation field ( especially for soft photons ) shoots up quickly right before percolation . in fig .
[ dfefig ] , we notice that for a soft source spectrum , the bias of hii also rises before the time of percolation ( but less dramatically ) .
this appears counter - intuitive : naively , one expects that both the hii abundance and the radiation roughly trace the ionizing sources , whose bias in our version of the extended press - schechter model always decays with time ( shown in fig .
[ bias_source ] ) .
while this intuition is correct in the early stages of reionization , the situation starts to change when the mean ionization fraction becomes significant .
this is because a high density region generates more photons than it can consume ( by ionization when the ionized fraction is already high ) , and this leads to run - away , causing the abrupt rise of the bias of the radiation field ( with a corresponding , but less dramatic , rise of the hii bias ) .
after percolation , the radiation bias of course drops precipitously because the photons are free to diffuse to large distances .
this effect is less pronounced for hard photons because percolation for hard photons is a more gradual process to begin with .
another interesting feature in fig .
[ dfgama_time ] is that the bias of the soft photons exhibits damped oscillations after reionization is complete .
this behavior can be traced back to the oscillatory kernel eq.[[angleavefgama ] ] . in the case of @xmath180 , the bias of the soft photons
right before the percolation of hii bubbles rises up to a high value , meaning that high density regions contain many more photons than low density regions .
such a difference between the high and the low density regions quickly decays away when the hii bubbles merge , when the soft photons are `` released '' and can freely travel to the neighbouring low density regions .
the relaxation of this process leads directly to the oscillations we see in fig .
[ dfgama_time ] .
the oscillations are most pronounced when the wavelength is small ( high @xmath106 ) and when the post - percolation drop in bias is most abrupt ( i.e. when the photons are soft ) .
note that these oscillations can be washed out by the stochastic bias of the ionizing sources , which is not considered in this paper . finally , we find that the biases of both hii / hi and the radiation field remain scale invariant on large scales ( @xmath181 ) .
this is not only because the source bias is scale invariant in our model , but also due to the fact that the diffusion of photons is negligible when the scale of interest is much larger than the mean free path .
using eq.[[angleavefgama ] ] again , we can see that at small @xmath106 the oscillatory kernel remains constant , therefore the @xmath106 dependence is essentially removed .
the scale invariant nature of the large scale hii / hi bias can be very useful , because it makes it possible to measure the shape of the primordial mass power spectrum using the hi power spectrum .
the evolution of the hi fluctuation shown in fig .
[ dhifig ] may be useful for this purpose in future 21 cm observations .
in this section , we briefly discuss several interesting issues related to our results above .
first , recall that the clumping factors @xmath182 and @xmath72 are set by hand in our calculations . in particular
, we have neglected their dependence on redshift and frequency ( the latter is relevant for @xmath182 ) .
the precise dependence is uncertain , and is only partially addressed by the most recent simulations ( see @xcite ) .
however , as emphasized before , the clumping factors appear explicitly only in the evolution equations for the spatially averaged quantities ( eq . [ [ homofe ] ] & [ [ homofgama ] ] ) , and not in the equations for the first order perturbations ( eq . [ [ flinearfe ] ] & [ [ flinearfgama ] ] ) . the clumping factors only influence the first order perturbations indirectly through their influence on the background quantities @xmath53 and @xmath54 .
while quantitative details regarding the perturbations do depend on the precise values of the clumping factors , the overall qualitative behavior , such as the evolution of the various biases , remain quite robust .
another issue concerns the topology of the hii distribution .
it is interesting to ask if the topology is still inside - out if the source spectrum is made up of only hard photons .
this , for example , is relevant to the scenario proposed by ricotti & ostriker ( 2004 ) , in which reionization at high redshift is caused by highly obscured miniquasars . to answer this question
, we consider three cases , in which we remove the soft photons by mutilating the @xmath151 spectrum and erasing all photons below @xmath183 , @xmath184 , @xmath185 , respectively ( the emissivity factor @xmath140 is raised to @xmath186 , @xmath187 , and @xmath188 , respectively to yield the same optical depth ) . in fig .
[ fe_x_ray ] , we see that the bias of hii always exceeds unity , meaning the topology is not outside - in , but the hii bias does decrease with increasing hardness of the ionizing photons .
one can also try to reduce the hii bias by lowering the source bias . in fig .
[ fe_sourcebias_one ] , we repeat the calculation for the case of ( @xmath189 ) except that the source bias is artificially set to unity , which is probably the lowest value one can expect , in any scenario in which the ionizing sources populate collapsed halos .
we find that the bias of hii is again always larger than one is due to the run - away effect introduced in [ results ] . ] . for @xmath149 or
@xmath151 , the evolution of the hii bias becomes more featureless and very close to unity at all redshifts and a source bias of unity , the hii bias right before the end of reionization does go slightly below unity .
this means that the ionization topology is marginally outside - in . ] .
therefore , our conclusion about the topology of the hii distribution is robust . )
( see , , @xcite ) . whether or not the non - linear terms can affect the large scale ionization topology is still an open question ( although the arguments in appendix b suggest that the high order effects can be neglected on sufficiently large scales ) . ] at the end of reionization , our conclusion about the ionization topology may seem counter - intuitive , especially for x - ray reionizations . assuming a homogeneous radiation background ( which is the limiting case for very hard spectra of the x - ray reionization scenarios considered here )
, one finds that in the limit of a low neutral fraction , @xmath190 , ionization equilibrium implies @xmath191 , i.e. more overdense regions are less ionized .
therefore , the ionization topology should be `` outside - in '' rather than inside - out , in the limit of late times and a very hard spectrum .
our calculations above do not reveal this limiting case , however , for the following reasons : ( i ) our sources do not have arbitrarily hard spectra , ( ii ) we include a clustering of the ionizing sources , which , together with the rapid evolution of the emissivity , causes the background radiation to be persistently non uniform .
we have , however , checked that our code can reproduce the outside - in limiting case when we set the source bias to be zero ( i.e. at late times , our code yields the expected behavior @xmath192 , which follows from @xmath191 in the limit of a small mean neutral fraction @xmath193 and small fluctuations ) .
we have made at least two important simplifications in our calculations .
one is ignoring helium .
the other is to assume the source bias is deterministic .
it is in principle straightforward to relax these assumptions / restrictions .
we hope to do so in a future paper . in practice , we still know little about the source properties at redshift above six . this work can serve as a link between the high redshift sources and the large scale distributions of the hii regions and the radiation fields .
for example , as we have shown , the bias of the source distributions is similar to that of the hii regions during the early period of reionization ; the evolution history of the bias of the hii regions can be used to constrain the hardness of the source spectrum .
it allows us to constrain the source properties with the upcoming 21 cm observations and the kinetic sz effect from the cmb .
furthermore , as we have found in our calculation that the distribution of hii traces the dark matter s on large scales with a scale independent bias , these observations may also be used to measure the shape of the linear matter power spectrum .
a more careful discussion of these issues will appear in another paper .
we have developed a perturbation theory of cosmic reionization by solving the linearized radiative transfer equation and the equation of ionization balance in fourier space .
the formalism can be used to predict the large scale fluctuations of the hii regions and the radiation fields for a given spatial distribution and spectrum of the ionizing sources .
the numerical solutions are straightforward to obtain . in the case of uv dominated source spectra ,
we have found an approximate analytic solution which works accurately in the early stages of reionization . to illustrate our formalism
, we use the extended press - schechter theory to model the source clustering and adopt three different power - law type source spectra .
we find that for uv dominated source spectra , the biases of the hii regions and the uv photons remain high during most of the reionization process . for hard source spectra ,
the hii regions tend to be ionized in a more homogeneous manner .
the hii bias decays faster with time due to a longer photon mean free path and due to secondary ionization .
the topology of the hii distribution is always inside - out , with overdense regions more highly ionized , at least on large scales .
our findings suggest that clustering measurements from future 21 cm and cmb observations can be used to put constraints on properties ( both clustering and spectra ) of the ionizing sources .
furthermore , on sufficiently large scales , both hii and hi have a scale independent bias with respect to dark matter this means the same observations might be used to measure the shape of the matter power spectrum . a direct comparison of our results with 3d simulations ( such as those by kohler et al .
2005 , iliev et al .
2005 , and zahn et al .
2006 ) would be valuable and interesting .
an accurate comparison requires detailed information on the source properties and the clumping factors from the simulations , and we hope to perform such a comparison in a future paper .
we thank greg bryan , steven furlanetto , ilian iliev , adam lidz , and jeremiah p. ostriker for useful comments on an earlier version of this manuscript . research for this work is supported in part by the doe , grant number de - fg02 - 92-er40699 .
zh acknowledges support by nasa through grants nng04gi88 g and nng05gf14 g , by nsf through grants ast0307291 , and by the hungarian ministry of education through a gyrgy bksy fellowship . arons j. & wingert d. 1972 , apj , 177 , 1 barkana r. & loeb a. , 2001 , phys . rep .
, 349 , 125 baugh c. , cole s. , frenk c. & lacey c. , 1998 , apj , 498 , 504 becker r. et al .
, 2001 , aj , 122 , 2850 benson a. , nusser a. , sugiyama n. & lacey c. , 2001 , mnras , 320 , 153 benson a. , lacey c. , baugh c. , cole s. & frenk c. , 2002 , mnras , 333 , 156 benson a. , sugiyama n. , nusser a. & lacey c. , 2005 , astro - ph/0512364 bond j. , cole s. , efstathiou g. & kaiser n. , 1991 , apj , 379 , 440 bruscoli m. , ferrara a. , fabbri r. & ciardi b. , 2000 , mnras , 318 , 1068 chiu w. & ostriker j. , 2000 , apj , 534 , 507 chiu w. , fan x. & ostriker j. , 2003 , apj , 599 , 759 choudhury t. & ferrara a. , 2006 , astro - ph/0603149 ciardi b. , ferrara a. , governato f. & jekins a. , 2000 , mnras , 314 , 611 ciardi b. , ferrara a. , marri s. & raimondo g. , 2001 , mnras , 324 , 381 ciardi b. , stoehr f. & white s. , 2003 , mnras , 343 , 1101 cole s. , lacey c. , baugh c. & frenk c. , 2000 , mnras , 319 , 168 dijkstra m. , haiman z. & scharf c. , 2005 , apj , 624 , 85 dijkstra m. , haiman z. , rees m. & weinberg d. , 2004 , apj , 601 , 666 djorgovski s. , castro s. , stern d. & mahabal a. , 2001 , apj , 560 , 5 fan x. et al . , 2002 ,
aj , 123 , 1247 fan x. , carilli c. & keating b. , 2006 , araa , 44 , 415 field g. , 1958 , proc .
ire , 46 , 240 field g. , 1959 , apj , 129 , 525 furlanetto s. , zaldarriaga m. & hernquist l. , 2004 , apj , 613 , 1 gnedin n. , 2000 , apj , 535 , 530 gnedin n. , 2000 , apj , 542 , 535 gnedin n. & hui l. , 1998 , mnras , 296 , 44 gnedin n. & ostriker j. , 1997 , apj , 486 , 581 gruzinov a. & hu w. , 1998 , apj , 508 , 435 gunn j. & peterson b. , 1965 , apj , 142 , 1633 haiman z. , 2002 , apjl , 576 , 1 haiman z. , abel t. & madau p. , 2001 ,
apj , 551 , 599 haiman z. & cen , r. , 2005 , apj , 623 , 627 haiman z. & holder g. , 2003 , apj , 595 , 1 haiman z. & loeb a. , 1997 , apj , 483 , 21 haiman z. & spaans m. , 1999 , apj , 518 , 138 hui l. , haiman z. , 2003 , apj , 596 , 9 hui l. , gnedin n. , 1997 , mnras , 292 , 27 hui l. , seljak u. , 1996 , proceedings of the 173rd symposium of the iau , edited by kochanek c. s. and hewitt j. n. , kluwer academic publishers iliev i. , mellema g. , pen u. , merz h. , shapiro p. & alvarez m. , 2005 , astro - ph/0512187 iliev i. , scannapieco e. & shapiro p , 2005 , apj , 624 , 491 iliev i. , pen u. , bond j. , mellema g. & shapiro p. , 2006
, astro - ph/0607209 kaplinghat m. , chu m. , haiman z. , holder g. , knox l. , skordis c. , 2003 , apj , 583 , 24 knox l. , scoccimarro r. & dodelson s. , 1998 , prl , 81 , 2004 kogut a. et al . , 2003 , apjs , 148 , 161 kohler k. , gnedin n. & hamilton a. , 2005 , astro - ph/0511627 kolb e. w. , matarrese s. & riotto a. , 2005 , astro - ph/0506534 kramer r. , haiman z. & oh s. , 2006 , astro - ph/0604218 lacey c. & cole s. , 1993 , mnras , 262 , 627 madau p. , meiksin a. & rees m. , 1997 , apj , 475 , 429 mellema g. , iliev i. , pen u. & shapiro p. , 2006
, astro - ph/0603518 mesinger a. , bryan g. & haiman z. , 2006 , astro - ph/0604148 miralda - escud ' e j. , haehnelt m. & rees m. , 2000 , apj , 530 , 1 oh s. , 2001 , apj , 553 , 499 oh s. & haiman z. , 2003 , mnras , 346 , 456 osterbrock d. e. , ferland g. j. , 2005 , astrophysics of gaseous nebulae and active galactic nuclei ( university science books ) page l. , hinshaw g. , komatsu e. , nolta m. , spergel d. , bennett c. , barnes c. , bean r. , dore o. , halpern m. , hill r. , jarosik n. , kogut a. , limon m. , meyer s. , odegard n. , peiris h. , tucker g. , verde l. , weiland j. , wollack e. , wright e. , 2006 , astro - ph/0603449 peebles p. , 1980
, large - scale structure of the universe ( princeton university press ) press w. & schechter p. , 1974
, apj , 187 , 425 malhotra , s. , rhoads j. e. , 2004 , apjl , 617 , 5 ricotti m. & ostriker j. , 2004 , mnras , 352 , 547 salvaterra r. , ciardi b. , ferrara a. & baccigalupi c. , 2005 , mnras , 360 , 1055 santos m. et al . , 2003 , apj , 598 , 756 santos m. et al . , 2004 , apj , 606 , 683 scott d. & rees m. , 1990 , mnras , 247 , 510 sokasian a. , abel t. , hernquist l. & springel v. , 2003 , mnras , 344 , 607 sokasian a. , yoshida n. , abel t. , hernquist l. & springel v. , 2003 , mnras , 350 , 47 spergel d. et al . , 2003 , apjs , 148 , 175 spergel d. , bean r. , dore o. , nolta m. , bennett c. , hinshaw g. , jarosik n. , komatsu e. , page l. , peiris h. , verde l. , barnes c. , halpern m. , hill r. , kogut a. , limon m. , meyer s. , odegard n. , tucker g. , weiland j. , wollack e. , wright e. , 2006 , astro - ph/0603449 shull j. & van steenberg m. , 1985 , apj , 298 , 268 sunyaev r. & zeldovich i. , 1980 , mnras , 190 , 413 theuns t. , schaye j. , zaroubi s. , kim t .-
, tzanavaris p. , carswell b. 2002 , apjl , 567 , 103 valageas p. , balbi a. & silk j. , 2001 , a&a , 367 , 1 valageas p. & silk j. , 1999 , a&a , 347 , 1 venkatesan , a. , giroux , m. l. , & shull , j. m. 2001 , apj , 563 , 1 white r. , becker r. , fan x. & strauss m. , 2003 , aj , 126 , 1 wyithe j. & loeb a. , 2004 , nature , 427 , 815 zahn o. , lidz a.,mcquinn m. , dutta s. , hernquist l. , zaldarriaga m. , furlanetto s. , 2006 , astro - ph/0604177 zaldarriaga m. , 1997 , prd , 55 , 1822 zaldarriaga m. , furlanetto s. & hernquist l. , 2004 , apj , 608 , 622
use eq.[[angleavefgama ] ] in eq.[[flinearfe ] ] , we find the following equation for @xmath99 : @xmath194 where @xmath195}{p(\omega,\omega')k}\\ \nonumber & \times&\int_0^{\infty}d\mu b(\omega,\mu)\langle\kappa(\mu)\rangle r(\omega',\mu+\omega-\omega')\\ \nonumber & \times&\exp[-\int_{\omega'}^{\omega}d\omega''b(\omega'',\mu+\omega-\omega'')]\end{aligned}\ ] ] and @xmath196}{p(\omega,\omega')k}n(\omega')\\ \nonumber & \times&\int_0^{\infty}d\mu b(\omega,\mu)\langle\kappa(\mu)\rangle\tilde{\delta}_s(\omega',\mu+\omega-\omega')\\ \nonumber & \times&\exp[-\int_{\omega'}^{\omega}d\omega''b(\omega'',\mu+\omega-\omega'')]\end{aligned}\ ] ] to solve eq.[[deltafe ] ] , we use an integrating factor @xmath197 which is defined as : @xmath198 and multiply both sides of eq.[[deltafe ] ] by @xmath197 : @xmath199 or @xmath200 defining the function @xmath201 : @xmath202 and changing the order of the integration , we can re - write eq.[[deltafe3 ] ] as : @xmath203 where @xmath204 eq.[[deltafe4 ] ] is a standard volterra integral equation of the second kind . its solution can be easily generated by inverting a triangular matrix . using @xmath201 , it is not hard to calculate @xmath99 .
one can then use eq.[[solutionfgama ] ] to get @xmath100 or eq.[[angleavefgama ] ] to get the monopole of @xmath100 .
in this appendix , we address the following question : is linear perturbation theory justified on large scales even when highly nonlinear structures exist on small scales ? this is a deep question that arises both in the present context of reionization and in the more familiar context of large scale structure formation theory .
we will make no attempt to provide a rigorous justification here .
intead , we will make some plausibility arguments , which are borrowed from the field of large scale structure ( peebles 1980 ) .
ultimately , numerical simulations are needed to rigorously justify the use of perturbation theory on large scales .
setting the last term of eq.[[linearnonlinear ] ] to be zero , we have @xmath208_{\rm linear } e^{i \vec{k } \cdot \vec{x}}\end{aligned}\ ] ] where we have introduced a mask @xmath209 which vanishes in the nonlinear regions , and equals unity in the linear regions , and we have linearized the ionization and recombination terms . ultimately , we are interested in the power spectrum of fluctuations in @xmath13 for instance .
eq.([linear2 ] ) tells us that its evolution can be regarded as linear as long as we are studying scales @xmath106 on which the mask @xmath210 , or its fourier transform , has a negligible effect on the power spectrum .
generally , the linear approximations will be valid only on scales much larger than the size of the nonlinear regions , and likely even larger than the clustering scale of the nonlinear regions .
for instance , in the early stages of reionization , we expect perturbation theory to work only on scales that encompass many hii bubbles .
note that the mask in question is more complex than the usual masks in galaxy surveys : here , the mask is correlated with the signal in non - trivial ways , and it is by no means obvious that the large scale power spectrum is unaffected by such masking . addressing
this important issue is beyond the scope of this paper .
how about the radiative transfer equation ? applying a similar split , we have @xmath211 e^{i \vec{k } \cdot \vec{x } } \\ \nonumber & & + \int_{\rm nonlinear } d^3 x [ { \,\rm source\ , } - { \,\rm sink\ , } ] e^{i \vec{k } \cdot \vec{x}}\end{aligned}\ ] ] here , the situation is similar to the ionization balance equation , in that in nonlinear regions , one expect the source and sink to roughly cancel , but in general , we do nt expect exact cancelation . for instance , one can think of a galaxy where the uv photons are propagating out of a thick medium .
most of the photons are consumed within the galaxy , but inevitably there will be some that escape .
one usually quantifies this by the escape fraction . here , we can account for this effect by renormalizing the source : @xmath212 e^{i \vec{k } \cdot \vec{x}}\end{aligned}\ ] ] the mask @xmath210 is the same as before , and the sink term can be linearized .
the source term can be thought of as a new effective source that accounts for the escape fraction from the nonlinear regions . | we develop an analytic approach to study inhomogeneous reionization on large scales by solving the equations of ionization balance and radiative transfer to first order in perturbations . given the spatial distribution and spectrum of the ionizing sources , our formalism can be used to predict the large scale power spectra of fluctuations in the abundances of hii , hi and radiation .
our approach avoids common approximations / assumptions in existing analytic methods for instance , we do not assume a specific ionization topology from the outset ; nor do we make a step - function bubble - like approximation to the hii distribution . applying our formalism to sources biased according to the press - schechter prescription ,
we find : 1 .
reionization always proceeds `` inside out '' , with dense regions more highly ionized , at least on large scales ; 2 . on sufficiently large scales ,
hii , hi and radiation exhibits a scale independent bias relative to dark matter ; 3 .
the bias is suppressed on scales comparable to or smaller than the mean free path of the ionizing photons ; 4 .
if the ionizing source spectrum is sufficiently soft , the hii bias closely tracks the source bias for most of the reionization process but drops precipitously after percolation ; 5 .
if the ionizing source spectrum is sufficiently hard , the hii bias drops in a more steady fashion throughout the reionization process .
the tools developed here will be useful for interpreting future 21 cm , cmb and lyman - alpha forest observations , both to learn about the reionization astrophysics ( such as the hardness of the source spectrum and therefore the nature of the ionizing sources ) and to possibly extract interesting cosmological information .
[ firstpage ] cosmology : theory - intergalactic medium - diffuse radiation |
after the initial boom in graphene research , recent years have seen a surge of interest in other two - dimensional ( 2d ) atomic crystals @xcite . among them , molybdenum disulfide ( @xmath0 ) , a prototypical transition metal dichalcogenide , has attracted great attention due to its excellent electronic and optical properties @xcite .
similar to graphite , bulk @xmath0 consists of vertically stacked layers that are loosely coupled via the van der waals interaction .
when shaped into monolayer , the energy gap changes from indirect to direct ( @xmath31.9ev ) , right in the visible frequency range @xcite , allowing applications such as transistors @xcite , photodetectors and electroluminescent devices .
in addition , due to the strong spin - orbit interaction and broken inversion symmetry , monolayer @xmath0 also exhibit novel valley and spin physics as demonstrated recently @xcite . to realize its application potential in multi - functional electronic devices ,
it is essential to understand the transport mechanisms in @xmath0 .
compared with graphene @xcite , the mobility of pristine @xmath0 is rather low , typically on the order of 10 @xmath1/(v@xmath2s ) at room temperature .
higher mobilities can be achieved by gate dielectric engineering to effectively screen the coulomb scattering on charged impurities and suppress electron - phonon scattering , values in the range from 200 to 700 @xmath1/(v@xmath2s ) have been reported @xcite . on the other hand , theoretical calculations of the phonon - limited mobility , using the deformation potential approximation or full band monte carlo simulation , have placed an intrinsic limit in the range of @xmath4 @xmath1/(v@xmath2s ) at room temperature @xcite , which are close to experimental values .
this suggests that further improvement of the mobility must come from better control of electron - phonon coupling .
strain engineering @xcite and electron - doping @xcite have been successfully used to improve the performance of @xmath0 , such as energy gap and superconductivity . here
, we have systematically investigated the electron - doping and strain dependence of the transport electron - phonon coupling constant @xmath5 and the phonon - limited mobility @xmath6 in monolayer @xmath0 based on first - principles calculations and the boltzmann transport theory .
it is found that @xmath5 increases with increasing doping concentration @xmath7 , reaches the maximum at @xmath8 @xmath9 before starts decreasing .
remarkably , @xmath5 can be significantly reduced by applying a small amount of strain .
this is due to the change of energy difference between the k and q valleys , which effectively suppresses the inter - valley scattering @xcite .
we show that even with 3% strain the mobility can be increased from 275 to 1150 @xmath1/(v@xmath2s ) .
our results provide an effective method to modulate the electron transport of @xmath0 by external strain .
our calculation is based on the semiclassical boltzmann transport theory .
details of theory can be found in ref . .
the key quantity is the transport electron - phonon coupling strength @xmath10for wavevector @xmath11 and mode index @xmath12 , defined by @xcite @xmath13 where @xmath14 is the electron - phonon interaction matrix element , @xmath15 and the efficiency factor , @xmath16 in eq .
( [ eq : epcq ] ) , @xmath17 is the density of state at the fermi surface , @xmath18 is the total number of @xmath19 points , @xmath20 and @xmath21 are the band energy and group velocity of the bloch electrons , respectively . in eq .
( [ eq : m ] ) , m is the atomic mass , and @xmath22 is the derivative of the self - consistent effective potential with respect to atomic displacement associated with the phonon from branch @xmath12 with the wave vector @xmath11 and frequency @xmath23 .
the efficiency factor @xmath24 shows that only backward scattering will contribute to the resistance .
similar to the relation between the eliashberg function and the superconducting electron - phonon coupling constant , the transport spectral function @xmath25and the transport electron - phonon coupling constant @xmath5 can be obtained by @xcite , @xmath26 @xmath27 where @xmath28 is the total number of @xmath29 points . finally , using transport spectral function @xmath25 , the relaxation time @xmath30 can be derived by solving the boltzmann equation in the lowest - order variational approximation ( lova ) as , @xmath31 @xmath32\\ \{[n(\omega ) + 1]\delta ( \epsilon -\epsilon'-\hbar \omega ) + n(\omega ) \delta ( \epsilon -\epsilon'+\hbar
\omega ) \ } \label{eq : sinhw } \end{aligned}\end{aligned}\ ] ] where @xmath33 .
@xmath34 and @xmath35 are the fermi - dirac and the bose - einstein distribution function , respectively . moreover
, the influence of impurities and electron - electron scattering are not included evidently in our calculations , which are subject to further investigations . utilizing the transport electron - phonon coupling ,
the temperature dependence of mobility @xmath36 can be obtained by @xmath37 where @xmath38 is the average square of the fermi velocity along the @xmath39 direction , @xmath40 is the area of unit cell . at doping concentration of @xmath41 @xmath9 .
the doping effect ( @xmath42 @xmath9 ) at @xmath43 ( c ) and the strain effect ( @xmath44 ) at @xmath41 @xmath9 ( d ) on the lowest conduction band along the @xmath45-@xmath46 direction ( labeled by red color in ( a ) ) . the band structure at @xmath45 and q points
are obviously modulated by doping and strain , respectively . the doping effect
( e ) and the strain effect ( f ) on the la phonon mode along the @xmath47-@xmath45 direction ( labeled by red color in ( b)).[fig:1],scaledwidth=50.0%,height=412 ] technical details of the calculations are as follows .
all calculations in this work were carried out in the framework of density functional theory ( dft ) with local - density approximation ( lda ) @xcite , as implemented in the quantum espresso package @xcite .
the ion and electron interactions are treated with the norm - conserving pseudopotentials @xcite .
the kinetic energy cutoff of 30 ry and monkhorst - pack @xmath48-mesh of @xmath49 were used in all calculations of electronic properties .
the atomic positions were relaxed fully with the energy convergence criteria of @xmath50 ry and the force convergence criteria of @xmath51 ry / a.u . in our slab model , a vacuum layer with 15 was set to avoid the interactions between the adjacent atomic layers .
the equilibrium lattice constant of the monolayer @xmath0 was found to be @xmath52 .
the electron doping was achieved by increasing the valence charge and at the same time introducing the same amount of uniform background charge .
the strain was introduced by adjusted the lattice constant @xmath53 of the monolayer @xmath0 with the strain capacity @xmath54 .
the phonon dispersion and electron - phonon coupling were calculated on a @xmath55 @xmath56-grid using the density functional perturbation theory ( dfpt ) @xcite . we have carefully checked the convergence of q - point sampling with the denser grid of @xmath57 , while the difference is only below 1% .
figure [ fig:1 ] shows the electronic band structure and phonon dispersion of monolayer @xmath0 with different electron doping concentrations and strains .
central to our discussion is the existence of a second energy minima in the conduction band , called q valley , which is about 100 mev higher than the conduction band minimum at the k valley , and approximately located at the halfway point of the @xmath45-k line . because of the close proximity in energy , the inter - valley scattering between the k and q valleys has a significant effect on transport @xcite .
we note that while the conduction band at the @xmath45 point displays a strong dependence on the doping concentration , with a large drop after @xmath58 @xmath9 , the band at the q point shows little change with various concentrations [ fig .
[ fig:1](c ) ] . on the other hand , the energy difference between the k and q valleys ( e@xmath59 ) can be modified significantly by the strain [ fig .
[ fig:1](d ) ] , consistent with previous calculations @xcite .
the change of phonon dispersion also shows the important roles of the electron doping and strain .
the acoustic phonon branch ( _ i.e. , _ longitudinal acoustic mode ) around the m point ( denoted by @xmath60 for simplicity ) appears phonon softening with increasing doping concentration up to @xmath61 @xmath9 .
however , for @xmath58 @xmath9 the phonon dispersion does not show any obvious change as @xmath7 is varied [ fig .
[ fig:1](e ) ] .
the @xmath60 phonons softening is also significantly suppressed by applying the strain [ fig .
[ fig:1](f ) ] .
the strong dependence of both the electronic band structure and phonon dispersion on the doping and strain indicates that the electron - phonon coupling should show similar trend . to quantify this statement ,
we have calculated the transport spectral function @xmath25 and transport electron - phonon coupling constant @xmath5 .
the map distribution of wave - vector - resolved transport electron - phonon coupling parameter @xmath62 as a function of doping concentrations in the absence of strain is shown in figs .
[ fig:2](a)-[fig:2](d ) .
one can see that the transport electron - phonon coupling around the m point is always the strongest , while those of k and @xmath45 points are rather small .
it indicates that @xmath60 phonons is the dominating contributor to the transport electron - phonon coupling .
the electronic transitions between the k and q valleys scattered by @xmath63 phonons @xcite is described as @xmath64 , where @xmath65 ( @xmath66 ) is the electronic state with the momentum k ( q ) . with the increase of doping concentration , the electrons occupying on the k and q valleys increase and the frequency of phonons around the m point decreases [ fig . [ fig:1](e ) ] .
consequently , the coupling to @xmath63 phonons increases sharply [ figs .
[ fig:2](a)-[fig:2](c ) ] .
however , it starts decreasing when @xmath58 @xmath9 [ fig .
[ fig:2](d ) ] .
one reason is the little dependence of the @xmath63 phonons on @xmath7 after it reaches @xmath67 @xmath9 [ fig .
[ fig:1](e ) ] .
the other is the reduction of electrons in the q valley , which arises from the lowering of the conduction band at the @xmath45 point [ fig .
[ fig:1](c ) ] .
furthermore , the lowering of conduction band at @xmath45 point also results in the appearance of two more channels of electronic transitions , @xmath68 and @xmath69 , which can be derived from the band structure and nesting function ( see appendix ) .
however , since the phonon frequencies at both k and q points are higher than that of m point [ fig .
[ fig:1](b ) ] , according to the eq . ( [ eq : epcq ] ) , these two additional channels have much weaker electron - phonon coupling comparing with the @xmath64 channel [ fig . [
fig:2](d ) ] .
more interestingly , @xmath70 can be greatly suppressed once a small amount of strain is introduced [ figs .
[ fig:2](e)-[fig:2](h ) ] .
this is because of the hardening of the @xmath60 phonons under strain condition and the increase of e@xmath59 [ fig .
[ fig:1](d ) ] . at the doping concentration of @xmath71 @xmath9 with @xmath72 ( black dash line ) and @xmath41 @xmath9 with @xmath73 ( black solid line ) and @xmath74 ( red solid line ) .
( b ) transport electron - phonon coupling constant @xmath5 as a function of the doping concentration @xmath75 at various strains of @xmath76 = 0% , 0.4% , 0.8% , 1.2% , and 3.0% .
[ fig:3],scaledwidth=50.0% ] as a function of the doping concentration @xmath75 at various strains of @xmath76 = 0% , 0.4% , 0.8% , 1.2% , and 3.0% . [ fig:4],scaledwidth=50.0% ]
figure [ fig:3](a ) shows the transport spectral function @xmath25 versus doping concentration under different strains .
@xmath25 spreads through a wide frequency range , but shows strong peaks at low frequencies , 80 @xmath3 150 @xmath77 , corresponding to the @xmath63 phonons .
the position of this main region shifts down with increase of the doping concentration , consistent with the behavior of the @xmath63 phonons softening , while position of the high frequency region is almost unchanged . after the application of strain , @xmath25
is then greatly suppressed .
the curves of @xmath5 with electron doping concentration under different strains are plotted in fig .
[ fig:3](b ) .
one can see that in the absence of strain , @xmath5 is strongly dependent on the doping concentration , in agreement with the results of raman spectroscopy and superconductivity researchs @xcite .
the maximum ( @xmath31.6 ) of @xmath5 appears at @xmath78 @xmath9 .
however , with added strain , the maximum of @xmath5 shifts down to 0.14 , and the overall @xmath5 versus @xmath7 curve becomes rather flat .
the doping and strain effects on @xmath5 are consistent with the discussion of @xmath70 as above illustration .
the mobility @xmath36 calculated at room temperature ( @xmath79 k ) are plotted in fig .
[ fig:4 ] . under the strain - free condition
, @xmath6 is inversely proportional to the doping concentration in a wide range of doping level and increases slightly in the high doping level due to the change of electron - phonon coupling . in the two previous theoretical studies
@xcite , the mobility has been estimated to be 130 ( 410 ) @xmath1/(v@xmath2s ) respectively , depending on with ( without ) the consideration of the scattering of q valley .
it agrees well that our estimation of 275 @xmath1/(v@xmath2s ) at @xmath80 @xmath9 falls somewhere between them .
next , we turn to discuss the stain effect on the mobility . at the low doping concentration of @xmath81 , @xmath76 = 3.0% strain can increase mobility from 275 to 1150 @xmath1/(v@xmath2s ) , for the strain can greatly suppress the electron - phonon coupling .
in addition , the change of mobility by strain from @xmath76 = 1.2% to @xmath76 = 3.0% is inconspicuous in the low doping level ( @xmath82 ) .
it can be ascribed to that the influence of q valley scattering to the mobility is rather weak such that can be almost totally removed by a small strain ( @xmath83 ) . at the high doping concentration ( @xmath84 ) ,
the modulation of mobility by strain becomes more significant , from 4.9 to 66.0 @xmath1/(v@xmath2s ) by a factor of 10 [ fig .
[ fig:4 ] ] .
although the present estimation of 275 @xmath1/(v@xmath2s ) is coincide with theoretical predictions and some experiments as well , a significant difference was observed compared with the high value of 700 @xmath1/(v@xmath2s ) @xcite .
considering the lattice mismatching with the substrates in the different experimental preparations , the strain appears inevitably in @xmath0 .
nevertheless , it is provided in our results that the mobility is exquisitely sensitive to the external strain , @xmath3 1000 @xmath1/(v@xmath2s ) improved by the strain of 1.2% .
clearly , the question about disparity between theoretical predictions and experimental results can be reasonably illustrated in this work .
in summary , we have studied the doping and strain effects on the transport electron - phonon coupling and intrinsic mobility in monolayer @xmath0 based on first - principles methods with boltzmann transport theory .
we estimate that under the strain - free condition the mobility has a maximum 275 @xmath1/(v@xmath2s ) . based on the analysis of electron - phonon coupling
, it is found that the inter - valley scattering between the q and k valleys assisted by @xmath85 phonons are the main source of electric resistance .
however , applying a small amount of strain can significantly increase the mobility by several times for the modulation effects on the lowest conduction band .
our results illustrate the effective modulation of strain on mobility in @xmath86 .
this work was supported by the most project of china ( grants nos .
2014cb920903 and 2011cba00100 ) , the nsfc ( grant nos.11225418 , 11174337 and 11374033 ) , the specialized research fund for the doctoral program of higher education of china ( grant no .
20121101110046 and 20131101120052 ) , the excellent young scholars research fund ( grant no .
2013cx04004 ) and the basic research fund ( grant no .
20121842009 ) of beijing institute of technology .
dx is supported by the u.s .
department of energy , office of basic energy sciences , materials sciences and engineering division .
in this appendix , we illustrate two additional channels of electronic transitions , @xmath87 , resulting from the lowering of conduction band at the @xmath45 point when the doping concentration increases from 1.7 to 1.9 @xmath88 @xmath9 . for clear observation , we calculated the nesting function @xmath89 , which describes the geometrical property of the fermi surface . figure [ fig:5 ] shows the change of @xmath91 at different doping concentrations . in the doping level of @xmath92 1.7 @xmath88 @xmath9 ,
@xmath89 around the q ( k ) point has no obvious change [ figs .
[ fig:5](a - c ) ] .
but one can clearly see that @xmath89 around the q ( k ) point enlarges significantly when @xmath75 increases to 1.9 @xmath88 @xmath9 as shown in fig .
[ fig:5](d ) . combining with the evolution of band structure along with increasing @xmath75 from 1.7 to 1.9 @xmath88 @xmath9 [ fig .
[ fig:1](c ) ] , one can deduce that there appears two additional channels of electronic transitions , i.e. , @xmath93 .
however , due to the weak electron - phonon coupling of these two additional channels , they can not be clearly seen in fig .
[ fig:2](d ) .
therefore , the dominated channel of electronic transition is @xmath64 , as discussed in the main text .
a. k. geim and i. v. grigorieva , nature * 499 * , 419 ( 2013 ) . q. h. wang , k. kalantar - zadeh , a. kis , j. n. coleman and m. s. strano , nature nanotechnol .
* 7 * , 699 ( 2012 ) .
f. zahid , l. liu , y. zhu , j. wang and h. guo , aip advances * 3 * , 052111 ( 2013 ) . j. feng , x. f. qian , c. w. huang and j. li , nat .
photonics * 6 * , 866 ( 2012 ) . h. j. conley , b. wang , j. i. ziegler , r. f. haglund , jr , s. t. pantelides and k. i. bolotin , nano lett . * 13 * , 3626 ( 2013 ) .
castellanos - gomez , r. rold@xmath98n , e. cappelluti , m. buscema , f. guinea , h. s. j. van der zant , and g. a. steele , nano lett .
* 13 * , 5361 ( 2013 ) .
a. splendiani , l. sun , y. b. zhang , t. s. li , j. kim , c. y. chim , g. galli and f. wang , nano lett .
* 10 * , 1271 ( 2010 ) .
k. f. mak , c. lee , j. hone , j. shan and t. f. heinz , phys .
105 * , 136805 ( 2010 ) .
b. radisavljevic , a. radenovic , j. brivio , v. giacometti and a. kis , nat . nanotechnol . * 6*,147 ( 2011 ) .
d. xiao , g. b. liu , w. x. feng , x. d. xu and w. yao , phys .
lett . * 108 * , 196802 ( 2012 ) .
h. l. zeng , j. f. dai , w. yao , d. xiao and x. d. cui , nat .
* 7 * , 490 ( 2012 ) .
k. f. mak , k. he , j. shan and t. f. heinz , nat . nanotechnol . * 7 * , 494 ( 2012 )
. t. cao , g. wang , w. p. han , h. q. ye , c. r. zhu , j. r. shi , q. niu , p. h. tan , e. g. wang , b. l. liu and j. feng , nat .
commun . * 3 * , 887 ( 2012 ) .
a. k. geim and k. s. novoselov , nat . mater . * 6 * , 183 ( 2007 ) .
j. h. chen , c. jang , s. d. xiao , m. ishigami and m. s. fuhrer , nat .
* 3 * , 206 ( 2008 ) .
h. wang , l. l. yu , y. h. lee , y. m. shi , a. hsu , m. l. chin , l. j. li , m. dubey , j. kong and t. palacios , nano lett .
* 12 * , 4674 ( 2012 ) .
w. z. bao , x. h. cai , d. kim , k. sridhara and m. s. fuhrer , appl .
. lett . * 102 * , 042104 ( 2013 ) .
s. das , h. y. chen , a. v. penumatcha and j. appenzeller , nano lett .
* 13 * , 100 ( 2013 ) .
k. kaasbjerg , k. s. thygesen and k. w. jacobsen , phys .
b * 85 * , 115317 ( 2012 ) .
x. d. li , j. t. mullen , z. h. jin , k. m. borysenko , m. b. nardelli and k. w. kim , phys .
b * 87 * , 115418 ( 2013 ) . j. t. ye , y. j. zhang , r. akashi , m. s. bahramy , r. arita and y. iwasa , science * 338 * , 1193 ( 2012 ) . k. taniguchi , a. matsumoto , h. shimotani and h. takagi , appl .
phys . lett .
* 101 * , 042603 ( 2012 ) .
b. radisavljevic and a. kis , nat . mater .
* 12 * , 815 ( 2013 ) .
m. m. perera , m. w. lin , h. j. chuang , b. p. chamlagain , c. y. wang , x. b. tan , m. m. c. cheng , d. tom@xmath98nek and z. x. zhou , acs nano * 7 * , 4449 ( 2013 ) .
y. song and h. dery , phys .
lett . * 111 * , 026601 ( 2013 ) .
p. b. allen , phys .
b * 17 * , 3725 ( 1978 ) .
j. p. perdew and a. zunger , phys .
b * 23 * , 5048 ( 1981 ) .
p. giannozzi , et al . j. phys . : condens .
matter * 21 * , 395502 ( 2009 ) . n. troullier and j. l. martins , phys .
b * 43 * , 1993 ( 1991 )
. s. baroni , s. d. gironcoli , a. d. corso and p. giannozzi , rev .
phys . * 73 * , 515 ( 2001 ) . c. h. chang , x. f. fan , s. h. lin and j. l. kuo , phys .
b * 88 * , 195420 ( 2013 )
. b. chakraborty , a. bera , d. v. s. muthu , s. bhowmick , u. v. waghmare and a. k. sood , phys .
b * 85 * , 161403 ( 2012 ) . | the doping and strain effects on the electron transport of monolayer @xmath0 are systematically investigated using the first - principles calculations with boltzmann transport theory .
we estimate the mobility has a maximum 275 @xmath1/(v@xmath2s ) in the low doping level under the strain - free condition .
the applying a small strain ( @xmath33% ) can improve the maximum mobility to 1150 @xmath1/(v@xmath2s ) and the strain effect is more significant in the high doping level . we demonstrate that the electric resistance mainly due to the electron transition between k and q valleys scattered by the m momentum phonons .
however , the strain can effectively suppress this type of electron - phonon coupling by changing the energy difference between the k and q valleys .
this sensitivity of mobility to the external strain may direct the improving electron transport of @xmath0 .
these authors contributed equally to this work ] |
final state interaction ( fsi ) effects have been observed in @xmath3 collisions .
they are responsible of strangeness enhancement , @xmath4 supression , fixed @xmath0 supression , azimuthal asymmetry , ... is it the manifestation of the formation of a new state of matter or can it be described in a fsi model with no reference to an equation of state , thermalization , hydrodynamics , ... ?
we take the latter view and try to describe all these obseervables within a unique formalism : the well known gain and loss differential equations .
we assume @xcite that , at least for particles with @xmath0 larger than @xmath5 , the interaction with the hot medium produces a @xmath0-shift @xmath6 towards lower values and thus the yield at a given @xmath0 is reduced .
there is also a gain term due to particles produced at @xmath7 . due to the strong decrease of the @xmath0-distributions with increasing @xmath0 ,
the loss is much larger than the gain .
asuming boost invariance and dilution of the densities in @xmath8 due to longitudinal expansion , we obtain [ 1e ] dn_^0 ( b , s , p _ ) d= - n(b , s ) here @xmath9 is the transverse density of the medium and @xmath10 the corresponding one of the @xmath1 @xcite .
this has to be integrated between initial time @xmath11 and freeze - out time @xmath12 .
the solution deepnds only on @xmath13 .
we use @xmath14 mb at both energies and @xmath15 for @xmath16 gev and @xmath17 for @xmath18 gev @xcite .
( [ 1e ] ) at small @xmath19 describes an interaction at the partonic level
. indeed , here the densities are very large and the hadrons not yet formed . at later times
the interaction is hadronic .
most of the effect takes place in the partonic phase .
we use a single ( effective ) value of @xmath20 for all values of the proper time @xmath19 .
the results at rhic and lhc are given in fig . 1 .
at lhc only shadowing @xcite has been included in the initial state .
the suppression is given by the dashed line .
it coincides with @xmath21 for @xmath0 large enough when shadowing and cronin efffects are no longer present .
the lhc suppression is thus a factor of two larger than at rhic .
= 7.5 cm = 7.45 cm -0.25 cm -7.45 cm = 7.5 cm = 7.45 cm -0.25 cm -0.3 cm
final state interaction in our approach gives rise to a positive @xmath2 @xcite ( no need for an equation of state or hydro ) .
indeed , when the @xmath1 is emitted at @xmath22 its path length is maximal ( maximal absorption ) . in order to compute it
we assume that the density of the hot medium is proportional to the path length @xmath23 of the @xmath1 inside the interaction region determined by its transverse position @xmath24 and its azimuthal angle @xmath25 .
hence , we replace @xmath26 by @xmath27@xmath28 where @xmath29 is the @xmath1 path length and @xmath30 denotes its average over @xmath25 .
( in this way the averaged transverse density @xmath26 is unchanged ) .
the suppression @xmath31 depends now on @xmath25 and @xmath2 is given by [ 2e ] v_2(b , p _ ) = d_r s_^0 ( b , p _ , _ r ) 2 _ r d_r s_^0 ( b , p _ , _ r ) the results at rhic and lhc are presented in fig .
99 a. capella , e. g. ferreiro , a. b. kaidalov and d. sousa , eur .
j. c * 40 * , 129 ( 2005 ) [ arxiv : hep - ph/0403081 ] .
a. capella and e. g. ferreiro , arxiv:0706.1880 [ hep - ph ] .
a. capella and e. g. ferreiro , phys .
c * 75 * , 024905 ( 2007 ) [ arxiv : hep - ph/0602196 ] . | using a final state interaction model which describes the data on these two observables , at rhic , we make predictions at the lhc using the same cross - section and @xmath0-shift .
the increase in the medium density between these two energies ( by a factor close to three ) produces an increase of the fixed @xmath0 @xmath1 suppression by a factor 2 at large @xmath0 and of @xmath2 by a factor 1.5 .
-0.5 cm |
in a previous paper @xcite statistical models were presented starting from a class of multiparameter braid matrices of odd dimensions ( ref .
1 cites previous sources ) .
this class of @xmath10 braid matrices for @xmath11 ( @xmath12 ) depends on @xmath13 parameters and has a basis of a _ nested sequence _ of projectors introduced before @xcite .
recently this class has been extended to include _ even _
dimensions @xcite .
the even dimensional @xmath0 braid matrices depend , apart from the spectral parameter @xmath4 , on @xmath1 free parameters .
one such parameter can , as usual , be absorbed by a suitable choice of an overall normalization factor .
such a quadratic increase of free parameters with @xmath14 ( even or odd ) is the most salient , unique feature of our models with their _ nested sequence _ basis .
we start with a detailed study of the @xmath3 case .
this is our _ new _ eight vertex model .
it should be compared to _
eight vertex model based on the s3 bialgebra @xcite .
there negative boltzmann weights were involved and were commented upon . here , for suitable choices of parameters , one has all weights positive or zero for @xmath15 or for an alternative choice for @xmath16 .
this @xmath3 case has another major distinguishing feature .
it finds its place as the first step in an explicitly constructed multiparameter hierarchy .
this is to be contrasted with the multistate generalization of six vertex model ( ref . 5 and the original sources cited there ) in which the parametrization remains restricted to the six vertex level as the dimensions increases .
we will present some remarkable general features conserved in the entire hierarchy .
but here our study will be brief - a beginning .
we hope to explore further elsewhere .
we will also briefly present spin chain hamiltonians for our class of solutions .
some aspects are better discussed after the constructions are presented .
such comments are reserved for the end .
we briefly recapitulate the construction of ref .
3 , closely related to that for odd dimensions @xcite . the notation of ref . 3 are maintained .
define the _ nested sequence _ of projectors @xmath17\right\},\ ] ] where @xmath18 , @xmath19 , @xmath20 , @xmath21 and @xmath22 denotes the matrix with an unique non - zero element 1 on row @xmath23 and column @xmath24 .
they provide a complete basis satisfying @xmath25 where @xmath26 is obtained by changing @xmath24 with @xmath27 in @xmath28 .
the @xmath29 braid matrices @xmath30 satisfy @xmath31 where @xmath32 and @xmath33 . here , as for odd dimensions @xcite the free parameters enter as exponents @xmath34 and the basic constraint @xcite @xmath35 has already been incorporated in ( 2.3 ) .
this leaves @xmath1 free parameters .
the simplest examples are : @xmath36 : @xmath37 where @xmath38 @xmath36 : @xmath39 with @xmath40 and @xmath41 we will often write @xmath42 and so on . generalization for @xmath43 is entirely straightforward .
to simplify computations , taking out an overall factor @xmath46 we write @xmath47 of ( 2.5 ) as @xmath48 where @xmath49 with @xmath50 . note that @xmath51 we consider these two domains separately , assuring nonnegative , real boltzmann weights . for purely imaginary parameters @xmath52 ( @xmath53 real ) @xmath54 with a normalization factor @xmath55 , one obtains a unitary braid matrix @xmath56
this has been pointed out in sec .
3 citing ref . 6 .
we continue to study here real matrices .
the basic @xmath57 equation can now be written as @xmath58 where @xmath59 we denote @xmath60 , @xmath61 , @xmath62 and the 4 blocks of @xmath63 as @xmath64 defining the @xmath65 matrices @xmath66 ( 3.6 ) is now ( with @xmath67 , @xmath68 , etc . )
@xmath69 a detailed study of ( 3.10 ) is given in appendix .
recursion relations will be extracted from it and implemented to construct the spectrum of the eigenvalues of the transfer matrix in sec .
the transfer matrix @xmath70 is obtained by starting with fundamental @xmath65 blocks ( for @xmath45 ) and constructing @xmath71 blocks using standard prescriptions ( coproduct rules - see sec .
the starting point , the @xmath72 blocks for @xmath73 , is provided by the yang - baxter matrix corresponding to ( 3.1 ) , namely @xmath74 where @xmath75 is the standard @xmath3 permutation matrix . then one uses the coproduct rule @xmath76 such a construction guarantees the commutativity of the transfer matrix , i.e. @xmath77=0\ ] ] quite generally and reducing for our present case to @xmath78=0\ ] ] for all @xmath79 .
this is the basic ingredient of exactly solvable statistical models @xcite .
the trace and the highest eigenvalue of @xmath80 provide significant features of the corresponding models . we will obtain these quite simply and generally for our class . for @xmath81 certain essential features will be briefly presented in sec .
the next essential step is the construction of eigenstates and eigenvalues of the transfer matrix . in this section
we display the results for @xmath82 .
they provide explicit examples of the iterative structure of the transfer matrices @xmath83 to be derived in the next section .
moreover they illustrate how multiplets involving @xmath79-th roots of unity and possible multiplicities of them combine to provide a complete basis of mutually orthogonal eigenstates of @xmath83 spanning the base space . referring back to them one grasps better the full content of the general formalism of sec .
we start with the notations of sec .
@xmath36 : @xmath84 the eigenstates are evidently @xmath85 with @xmath86 and @xmath87 @xmath36 : @xmath88 where @xmath89 with @xmath90 and @xmath91 .
one has @xmath92 we have anticipated the iterative structure of sec .
5 . that @xmath93 obtained by a straightforward use of coproduct rules as @xmath94 can be expressed as ( 4.5 ) on a basis satisfying ( 4.7 )
is the central lesson .
denote @xmath95 for @xmath96 the order of the indices @xmath97 indicates the structure of the tensor product .
for example , @xmath98 .
one obtains , with upper or lower signs , @xmath99 and @xmath100 the eigenfunctions of @xmath101 ( resp .
@xmath102 ) are annihilated by @xmath102 ( resp .
@xmath101 ) consistently with ( 4.7 ) . finally , again anticipating sec . 5 , @xmath103 @xmath36
: from this stage onwards one can better appreciate the role of multiplets involving roots of unity in constructing eigenstates .
one obtains by implementing coproducts in step @xmath104 , @xmath105 where the @xmath106 matrices satisfy @xmath107 and they are @xmath108-independent .
they are easy to obtain but will not , for brevity , be presented explicitly .
the base space splits up into 4-dim .
ones , closed under the action of @xmath109 and are characterized by even ( odd ) number of the index 1 ( considering zero as even ) respectively . define @xmath110 where @xmath111 @xmath112 .
one has @xmath113 , @xmath114 , where @xmath115 denotes the transform of @xmath116 with conjugated coefficients .
denote also @xmath117 the states @xmath118 form a complete basis of orthogonal states .
consistently with ( 4.13 ) , ( 4.14 ) one obtains @xmath119 and , finally , @xmath120
@xmath36 : now @xmath121 where the @xmath122 constant matrices satisfy @xmath123 as consequence of recursion relations involved in @xmath124 .
the @xmath125 s are obtained fairly easily .
now the even and odd subspaces are 8-dim .
we display the eigenstates explicitly to illustrate a new feature .
now @xmath126 is not a prime number and 4-th roots and square roots of unity ( corresponding to @xmath127 ) both contribute multiplets involving respective coefficients @xmath128 and @xmath129 ( for @xmath130 one would have thus 2-plets , 3-plets and 6-plets ) .
the @xmath131 and @xmath132 spaces have the orthogonal bases @xmath133 and @xmath134 define @xmath135 then one obtains , in order , @xmath136 and @xmath137
we consider below exclusively the @xmath3 case .
coproduct rules lead , for @xmath139 , to the recursion relations @xmath140 where @xmath141 thus @xmath142 denote @xmath143 , @xmath144 and for @xmath145 factors @xmath146 , @xmath147 . starting from @xmath148 , @xmath149 , @xmath150 and iterating one obtains @xmath151
using @xmath152 one can express @xmath153 in term of any one of them .
in particular , @xmath154 where @xmath155 thus @xmath156 thus exchanging members of successive pairs of rows @xmath157_{\hbox{(rows)}}$ ] one obtains @xmath158 from @xmath159 . the following recursion relations are implied ( with the definition ( 5.6 ) of @xmath160 ) @xmath161 the signification of these relations concerning eigenvalues and why we display also @xmath162 will be explained below .
let us introduce at this point the general possibilities @xmath163 since the sequence for odd and even @xmath79 have some distinct typical features a _ two - step _ iteration can be of interest for @xmath164 .
one has ( in evident notations ) @xmath165 leading to @xmath166 from ( 5.12 ) one obtains an analogous result for @xmath167 .
we now extract one fundamental consequence of the recursion relations ( 5.8 - 13 ) . starting with @xmath168 and
@xmath169 these relations imply ( with @xmath108-independent matrices @xmath125 ) @xmath170 for even @xmath79 and @xmath171 for odd @xmath79 , powers of @xmath172 being always even in both cases . correspondingly ( again with @xmath108-independent @xmath71 matrices @xmath173 )
@xmath174 where @xmath175 .
there being here only odd powers of @xmath172 . when these series are inserted in ( 5.8 ) and ( 5.9 ) the features above are conserved . from the general constraints ( see appendix ) @xmath176 where @xmath67 , @xmath68 , etc .
it already follows that the @xmath125 ( resp .
@xmath173 ) must commute .
thus with @xmath177 , @xmath178 one must have , for all @xmath179 @xmath180= \left[\mathbf{y}_{(p)},\mathbf{y}_{(q)}\right]=0.\ ] ] but our recursion relations imply stronger constraints . from ( 5.8 ) @xmath181 note that @xmath182 along with recursion relations , systematically exploiting the constraints ( a.2 - 12 ) of appendix , one obtains that not only do @xmath183 , @xmath184 commute for each @xmath79 but @xmath185 analogously one can show @xmath186 these are indeed sufficient and necessary conditions for the eigenvalue spectrum derived for @xmath82 in sec .
4 . let @xmath187 denote an eigenstate of @xmath183 with the eigenvalue @xmath188 @xmath189 .
then for @xmath190 , @xmath191 hence @xmath192 but there is still one more class of constraints . starting with ( see ( 5.2 ) ) @xmath193 and noting that ( see ( 5.8 ) ) @xmath194 one obtains @xmath195 hence @xmath196 how is this constraint implemented for each @xmath79 ? in our examples ( sec . 4 , @xmath82 )
we saw that 1 .
@xmath197 has a multiplicity 2 , saturating ( 5.27 ) .
@xmath198 @xmath199 comes with multiplicity , each case providing a subset of zero sum . for each @xmath145 one
has , one or more , subsets @xmath200 3 . here
@xmath79-th roots of unity play a crucial role .
typically in ( 5.29 ) one has @xmath201 4 . for @xmath202
one has only cube roots of unity . for @xmath126 one
has both 2-plets and 4-plets , ( square roots of unity being also fourth roots ) .
let us now consider possible submultiplets from a more general point of view .
the even and odd subspaces introduced in sec .
4 ( even and odd multiplicities of the index 1 distinguishing them ) can be generalized to all @xmath79 , each one closed under the action of @xmath203 and of dimension @xmath204 . in each one
there is exactly 1 state with eigenvalue @xmath205 saturating ( 5.27 ) .
hence one can now consider separately two base spaces of dimension @xmath206 .
when @xmath79 is a prime number ( say @xmath207 ) there is a relative simplicity concerning the multiplet structure . a theorem of fermat ( see ref .
1 , appendix b : encounter with a theorem of fermat ) adapted to our case assures @xmath208 where @xmath209 is an integer .
thus for @xmath210 , etc . , @xmath211 , etc .. hence an integer number of @xmath207-plets can span adequately the @xmath212 dimensional space with @xmath213 .
when @xmath79 is not a prime number each prime factor of @xmath79 ( @xmath214 , @xmath215 , etc . )
can lead to submultiplets with zero sum .
finally , if a singlet occurs in @xmath216 the even subspace ( i.e. apart from @xmath205 ) it must occur in @xmath217 the odd one with an opposite sign ( ex : @xmath218 in @xmath216 and @xmath219 in @xmath217 for @xmath126 ) .
the number of possibilities increase rapidly with @xmath79 .
our study remains incomplete concerning the precise number of multiplets and the multiplicity for each higher @xmath79 .
we have however delineated completely , for all @xmath79 the dependence of the eigenvalue spectrum on @xmath108 ( or the spectral parameter @xmath4 ) .
leu us note one point .
we changed over from @xmath220 to @xmath221 in sec .
3 . thus gives conveniently a single parameter @xmath108 on the anti - diagonal . with the original normalization @xmath222 finally , starting with @xmath223 or @xmath93 ( of ( 4.8 ) ) and implementing recursions
one can show that the sum of the elements in each row ( and each column ) of @xmath224 is @xmath225 .
the sum of basic components of @xmath216 and @xmath217 ( @xmath226 , @xmath227 for @xmath228 in sec .
4 and their direct generalizations ) thus each corresponds to the eigenvalue @xmath225 .
the sum of these two furnishes the total trace of @xmath83 , namely @xmath229 . all the other states together contribute zero trace .
moreover , since for our choice of domains ( sec .
3 ) always @xmath230 , @xmath231 , @xmath232 assuming roots of unity in ( 5.24 ) for @xmath198 .
@xmath205 is the largest eigenvalue .
this is significant in statistical models . in this context
one should note that the special status of @xmath233 in the iterative structure . from ( 5.19 ) one obtains the matrices structure @xmath234 thus , in each subspace , one can construct iteratively exclusively @xmath233 staring from @xmath235 and the corresponding @xmath204 mutually orthogonal eigenstates : 1 .
one with eigenvalue @xmath205 .
@xmath236 with eigenvalue zero .
the latter provide non - zero eigenvalues for @xmath183 with @xmath145 non zero .
one thus obtains the complete basis of eigenstates .
one has for all @xmath14 @xmath238\right\},\nonumber\\ & & \mathrm{p}p_{i\bar{j}}^{(\epsilon)}=\frac 12\left\{\left(\bar{j}i\right)\otimes \left(i\bar{j}\right)+\left(j\bar{i}\right)\otimes \left(\bar{i}j\right)+\epsilon\left[\left(\bar{j}\bar{i}\right)\otimes \left(ij\right)+\left(ji\right)\otimes \left(\bar{i}\bar{j}\right)\right]\right\}.\end{aligned}\ ] ] these lead ( for fundamental blocks or @xmath73 ) to @xmath239 where @xmath240 . for @xmath241 ( resp .
@xmath242 ) all boltzmann weights are nonnegative for @xmath15 ( resp .
@xmath16 ) . note that the fundamental blocks are now @xmath243 matrices with only two non - zero elements each .
this number does not change with @xmath14 . in a compact notation with indices @xmath244 and @xmath245 correspondingly , one can write , for @xmath73 , @xmath246 with ( not only @xmath247 , but ) @xmath248 the coproduct rules gives the iterative structure @xmath249 the fact that , as in ( 6.3 ) , only diagonal blocks have diagonal elements can be easily shown to lead to @xmath250 this is the direct multiparametric generalization of the @xmath251 case @xmath252 it is instructive to study the case @xmath253 explicitly . denoting ( as in ref . 1 with @xmath254 , @xmath255 )
@xmath256 for @xmath73 ( with now @xmath257 , @xmath258 below - a change of notation convenient for displaying symmetries ) one has @xmath259 @xmath260 the remaining 8 blocks are given by @xmath261 one has @xmath262 the recursion @xmath263 leads to @xmath264 a particular case of ( 6.6 ) .
for @xmath253 the diagonal blocks , for example , have the iterative structures @xmath265 @xmath266 and @xmath267 are now obtained by setting respectively in @xmath268 and @xmath269 @xmath270 for @xmath271 and so on , systematically in an evident fashion .
their sum gives the transfer matrix of order @xmath272 exhibiting the iterative structure .
for @xmath73 the transfer matrix is directly diagonal for all @xmath14 giving directly the eigenvalues . for @xmath253 , for example ( consistently with ( 6.13 ) )
@xmath273 with evident generalization for @xmath43 .
for @xmath45 we have systematically explored the remarkable structure of the transfer matrix for all @xmath79 ( see ( 5.14 - 18 ) ) and consequences for eigenstates .
a parallel study for @xmath274 is beyond the scope of this paper .
our results in the section already indicate how the multiparametric aspects start playing on essential role . in sec .
5 , starting with the @xmath65 matrix @xmath275 ( 5.3 - 7 ) at the level of @xmath73 and implementing tensor products , powerful recursion relations were obtained .
we started by relating @xmath138 among themselves .
for @xmath276 one can similarly relate ( for a given pair of indices @xmath277 ) the quartet @xmath278 given , for @xmath73 , by ( 6.2 ) .
evidently one can relate through constant matrices only blocks involving the same pair of parameters @xmath279 . for
this one introduces the matrix @xmath280 generalizing @xmath281 .
thus for @xmath253 , one has @xmath282 the generalization is evident . from ( 6.8 ) @xmath283 with exactly analogous results for other subsets . for each @xmath14 one
has , in evident notations , for fixed @xmath284 @xmath285 we will not attempt to explore in the present paper the applications of such relations generalizing our results of sec .
our construction of odd dimensional hamiltonians ( sec . 4 , ref .
1 ) can be adapted to the present even dimensional cases as follows .
one has , taking derivatives and setting @xmath286 , for @xmath45 , @xmath287 where @xmath288 and @xmath289 is obtained by setting @xmath286 in @xmath290 . for @xmath253 , setting ( starting from ( 6.8 ) ) @xmath291 from ( 2.7 ) and @xmath292 where @xmath293 the extension of our formalism for @xmath43 is straightforward . for @xmath79 sites the standard result for the hamiltonian is ( see sources cited in ref .
1 ) @xmath294 where for circular boundary conditions @xmath295 .
we intend to present a more complete study of our spin chain elsewhere . but here one already sees how the two aspects , multistate ( higher spins at each site ) and multiparameter @xmath279 get directly associated for our hierarchy . in our previous papers @xcite and here again ( see ( 3.4 ) , ( 3.5 ) ) we showed how the passage to imaginary parameters can lead to unitary @xmath47 .
the corresponding @xmath289 has only an overall factor @xmath296 which can be extracted from the sum ( 7.5 ) .
[ [ i .- status - of - eigenstates - for - n1 ] ] * i. status of eigenstates for @xmath45 : * + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + for the simplest @xmath3 braid matrix in our hierarchy the construction of the eigenvalues and eigenfuntions of transfer matrices of successive orders @xmath297 has attained the following stage : ( 1 ) : : the transfer matrix at the level @xmath79 has been expressed in the form of a @xmath71 matrix @xmath298 where @xmath299 and @xmath300 ( resp .
@xmath301 ) for @xmath79 even ( resp .
odd ) ( @xmath302 ) .
@xmath4 being the spectral parameter , @xmath53 two free parameters ( see ( 3.3 ) ) .
the matrices @xmath183 are constant ones ( @xmath108-independent ) and satisfy @xmath303 they can be computed systematically via the recursion relations ( for @xmath304 ) .
( 2 ) : : for @xmath305 and @xmath306 denote phase factors which come in multiplets of zero sum formed by roots of unity corresponding to @xmath79 and its prime factors ( see examples in sec .
the index `` @xmath23 '' denotes such possible multiplicity of each @xmath145 .
the exception of the zero sum rule corresponds to @xmath307 .
one obtains for each @xmath79 , twice @xmath308 giving @xmath309 a general constraint obtained via recursions .
this multiplicity 2 corresponds to two @xmath204 dimensional subspaces ( see `` even '' , `` odd '' subspaces defined in sec .
4 ) each providing just one eigenstate of @xmath310 with non - zero eigenvalue @xmath205 . ( 3 ) : : thus the problem has been reduced to construction of eigenstates of each @xmath183 separately , reducing the dimension by considering each ( even , odd ) subspace by turn .
this involves solving sets of linear constraints with only positive and negative integers as coefficients .
one finally keeps only the non - zero eigenvalues for each @xmath145 , they being associated with zero eigenvalues for @xmath184 , @xmath190 .
in fact as noted below ( 5.33 ) , it suffices to construct the full sect of mutually orthogonal eigenstates of @xmath233 , all but one in each subspace having eigenvalue zero .
( 4 ) : : as already stated ( sec .
5 ) our results remain incomplete concerning the pattern of possible multiplets and submultiplets corresponding to roots of unity provided by @xmath79 and its prime factors - and multiplicities of such multiplets . a canonical enumeration
when @xmath79 has a very large number of prime factors seems to be an unlikely possibility . nor
have we established rigorously that @xmath306 in ( 8.3 ) are always @xmath311 or higher roots of unity phases factors .
this what happens in examples @xmath312 of sec . 4 and directly leads to the following obligatory constraint ( 8.4 ) .
( 4 ) : : we have completely , and for all @xmath79 , extracted the @xmath4-dependence of eigenvalues in ( 8.1 ) . [ [ ii .- comparisons - with - standard - six - vertex - and - eight - vertex - models ] ] * ii . comparisons with standard six vertex and eight vertex models :
* + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + from ( 2.5 ) and ( 3.1 ) our yang - baxter matrix ( for @xmath45 ) with corresponding normalizations , is @xmath313 where @xmath314 and , equivalently , @xmath315 where @xmath316 .
let us now compare this to the very well known @xmath3 six vertex and eight vertex models - concerning which it is sufficient to a cite a standard text book @xcite and review articles @xcite which cite basic sources . all such cases are of the form @xmath317 for ( 8.6 ) @xmath318 , @xmath319 and for ( 8.5 ) @xmath320 , @xmath321 . for six
vertex , crucially , @xmath322 and @xmath323 is being given , according to the regime , by circular or hyperbolic functions .
for eight vertex , famously , elliptic functions appear and @xmath324 . for our case @xmath324 but arguably , one has maximal simplicity and symmetry compatible with non - trivial solution for a @xmath3 yang - baxter ( or braid ) matrix .
we have all eight vertices but ( with , say ( 8.5 ) ) ( see fig .
1 ) fig . 1 corresponding to @xmath320 and ( see fig .
2 ) fig . 2 corresponding to @xmath321 . in six
vertex the last two vertices are excluded @xmath325 .
we will not study , in this paper , the implications of the results below ( 8.7 ) concerning various properties of our model ( compare the relevant detailed study of eight vertex in ref .
but we would like to contrast our approach to the construction of eigenstates and extraction of eigenvalues of @xmath83 with that via bethe ansatz in standard six vertex models @xcite .
the systematic study of @xmath326 constraints are particularly relevant ( see appendix a ) . in six vertex the bethe ansatz construction involves pushing @xmath327 through the product @xmath328 acting on one single state @xmath329 and eliminating unwanted terms to obtain a complete set of eigenstates corresponding to the sets of resulting constraints .
this involves solving nonlinear equations .
our resorts to programs and numerical studies for higher @xmath79 s . in our case
recursion relations for @xmath330 ( for @xmath304 ) are sufficient to attain the stage systematically presented in part ( i ) of this section .
one solves , at each stage , linear equations with integer coefficients .
in fact since each @xmath79 ( see ( 5.5 - 7 ) ) @xmath331 , where @xmath160 in the @xmath71 matrix with @xmath204-times @xmath332 on diagonal , the actions of @xmath333 follows immediately from those of @xmath334 .
they need hardly be studied separately .
however that @xmath160 and hence @xmath335 connect the even and odd subspaces . on the other hand @xmath336 can not be pushed through a product of @xmath337 s displayed above .
the nearest approaches are typically our ( a.10 ) and ( a.12 ) ( changes of signs in @xmath164 and @xmath338 are to be noted as @xmath336 is pushed through ) .
subtle analytic properties , unlike six vertex and particularly eight vertex models , play no role in our case .
one has only to look at the @xmath339 ( or @xmath4 ) dependence in ( 8.1 ) . in a simple situation
the contrast between our model and the standard eight vertex one shows up very clearly : for the asymptotic case @xmath4 tending to infinity one sets @xmath340 in ( 8.6 ) to obtain @xmath341 whereas , choosing adequately the normalizing factor ( see sec .
7 of @xcite and in particular eq .
( 7.13 ) ) for the standard case the corresponding limiting form is @xmath342 when diagonalized ( or block diagonalized ) our classes of matrices , in general , lose braid ( or yang - baxter ) property unless the diagonalizer has a corrected tensor structure ( sec . 4 and addendum of ref .
[ [ iii .- passage - to - higher - dimensions - ngeq-2 ] ] * iii .
passage to higher dimensions @xmath5 : * + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + as already emphasized before , a major interest of our simple model for @xmath45 ( the @xmath3 braid matrix ) is that it is the first one in a hierarchy of @xmath29 braid matrices with @xmath1 free parameters at each level .
some of the simplicity of the @xmath45 case is inevitably lost as @xmath14 increases .
but we have pointed out in sections 6 and 7 how certain basic features vary in a simple , canonical fashion .
thus for example , ( 1 ) : : @xmath343 ( 2 ) : : each blocks @xmath344 of the @xmath345 matrix has just two non - zero elements ( see ( 6.2 ) and ( 6.9 ) ) out of @xmath346 , namely @xmath347 for all @xmath14 . ( 3 ) : : for a fixed pair of indices @xmath284 the blocks @xmath278 can be quite simply related among themselves ( see ( 6.18 ) ) via direct generalizations of the matrix @xmath348 for the @xmath3 case . for @xmath45
such relations led to recursion relations yielding ( 8.1 - 3 ) .
( 4 ) : : spin chain hamiltonians present ( see eqs .
( 7.1 - 5 ) ) a simple canonical sequence as @xmath14 increases .
we hope to study the higher dimensional cases more fully elsewhere .
0.5 cm * acknowledgments : * _ one of us ( ba ) wants to thank pierre collet and paul sorba for precious help . _
from ( 3.10 ) one obtains with @xmath275 from ( 3.9 ) @xmath350 the last step follows from @xmath351 on both sides of the first two expressions since under this interchange ( see ( 3.7 ) ) @xmath352 .
the consistency of the last two implies that each element of the total matrix on the right ( apart from the factor @xmath353 ) must be symmetric in @xmath354 .
this can indeed be verified starting from @xmath355 using the standard construction in sec .
3 . from the last two steps
( the factor @xmath353 cancelling ) one obtains , with only upper or lower signs , @xmath356 one gets 8 relations of the type @xmath357 with @xmath358 in sec . 5 ( see ( 5.4 - 7 ) ) we obtained , for any @xmath108 and all @xmath79 , @xmath359 where ( @xmath108-independent ) matrix @xmath160 is given .
hence multiplying in ( a.3 ) by @xmath160 on the left one gets another set with @xmath360 replaced by @xmath361 , where thus @xmath362 and so on .
for another class of relations we introduce @xmath363 from ( 3.7 ) @xmath364 one can show that ( with upper or lower signs ) @xmath365 again , as for ( a.5 ) , ( a.6 ) , multiplying the above from the left by @xmath160 one gets another set of relations such that @xmath366 and so on . for `` two - steep '' relations redefine @xmath367 with indices @xmath368 and @xmath369 , @xmath370-cosech@xmath371 correspondingly .
one obtains , for example , for arguments @xmath372 @xmath373 see the relevant remarks in sec . 8 . | we study statistical models , specifically transfer matrices corresponding to a multiparameter hierarchy of braid matrices of @xmath0 dimensions with @xmath1 free parameters @xmath2 . the simplest
, @xmath3 case is treated in detail .
powerful recursion relations are constructed giving the dependence on the spectral parameter @xmath4 of the eigenvalues of the transfer matrix explicitly at each level of coproduct sequence .
a brief study of higher dimensional cases ( @xmath5 ) is presented pointing out features of particular interest .
spin chain hamiltonians are also briefly presented for the hierarchy . in a long final section
basic results are recapitulated with systematic analysis of their contents .
our eight vertex @xmath3 case is compared to standard six vertex and eight vertex models .
0.2 cm * b. abdesselam@xmath6 and a. chakrabarti@xmath7 * _ @xmath8 laboratoire de physique thorique , universit doran es - snia , 31100-oran , algrie + and + laboratoire de physique quantique de la matire et de modlisations mathmatiques , centre universitaire de mascara , 29000-mascara , algrie _ + _
@xmath9 centre de physique thorique , ecole polytechnique , 91128 palaiseau cedex , france . _ |
correlated systems with competing on - site and intersite coulomb interactions @xcite and fillings away from one electron per site ( @xmath11 , half filling ) are presently a subject of intensive investigation due to the appearance of complex phases such as unconventional charge and magnetic orders .
these systems become even more challenging when novel lattice structures emerge out of the original lattice in the region of strong correlations @xcite .
this phenomenon is often found in geometrically frustrated systems , such as triangular and kagome lattices . on the triangular lattice , for instance ,
large on - site @xmath0 and nearest - neighbor @xmath1 coulomb interactions generate effective honeycomb and enlarged triangular lattices at @xmath12 filling ( @xmath13 ) by inducing charge disproportionation @xcite . when @xmath14 the system tends to create a honeycomb lattice of empty sites and an enlarged triangular lattice of doubly occupied sites , while at smaller ratios of @xmath15 the system evolves into a honeycomb lattice of singly occupied sites with long - range antiferromagnetic order .
a similar charge ordered state with noncollinear magnetic order has also been proposed in the kondo lattice system @xcite .
while these states are insulating , such exotic charge and magnetic orders become metallic away from the commensurate filling @xcite . furthermore , at quarter filling ( @xmath16 ) , metallic states , named pinball liquids , have been also recently proposed @xcite .
they are characterized by a three - sublattice structure , in which the carriers of one sublattice are essentially localized ( pins ) , with the remaining charges ( balls ) building an itinerant liquid on the interstitials . recently , other mechanisms than direct charge disproportionation have been also proposed to generate new lattice structures such as the emergence of a kagome lattice via spontaneous ferrimagnetic order coexisting with a @xmath17 charge order pattern in a triangular kondo lattice @xcite .
similarly , on the kagome lattice , large values of @xmath0 and @xmath1 have been discussed to induce nearly isolated six - site rings and an enlarged kagome lattice at @xmath12 filling ( @xmath13 ) @xcite . specifically , when @xmath18 and @xmath19 , each corner - sharing triangle possesses charge order characterized by two singly occupied sites and an empty site .
the empty site randomly sits on one of the three vertices of a triangle , which gives macroscopic charge degeneracy .
nonzero hopping @xmath9 lifts the charge degeneracy and appears to stabilize a @xmath17 charge pattern , whose unit cell contains nine sites @xcite .
recently , by mapping the system into a hard - core boson hamiltonian , a topological liquid was also proposed @xcite . reported realizations of such emergent lattices are for instance the generation of a honeycomb structure through charge disproportion in the metallic magnet @xmath20
@xcite or the appearance of effective spin-@xmath21 chains in the heavy - fermion spinel @xmath22 @xcite . actually , even when the lattice structures themselves are not geometrically frustrated , such a formation of new lattices is possible due to strong electron correlations .
for example , in the cubic lattice , a staggered @xmath23 charge order generates doubled face - centered - cubic lattices @xcite
. moreover , effective spin and charge interactions in such systems may acquire additional geometrical frustrations .
indeed , unconventional noncoplanar magnetic orders have been proposed in the periodic anderson model on the cubic lattice at @xmath3 filling @xcite .
one of the most discussed bipartite lattices in two - dimensional systems is the honeycomb lattice . while correlation effects on the honeycomb lattice have been intensively discussed in the past , many studies focused mostly at half filling @xcite .
et al . _
@xcite and grushin _ et al . _
@xcite studied extensively the phase diagram in honeycomb systems for arbitrary filling ; however , they considered only the spinless hubbard model . in this work
, we investigate the possible emergence of correlation - induced new lattice structures both in a bipartite and in a geometrically frustrated lattice at fillings that , to our knowledge , have not been investigated before .
in particular , we perform an extensive analysis of the spinful extended hubbard model on honeycomb and triangular lattices away from half filling , and focus on the interplay between on - site @xmath0 and intersite nearest- and next - nearest coulomb interactions @xmath1 and @xmath4 , respectively . by using the hartree - fock approximation , as well as perturbation theory and the variational monte carlo method
, we find that a triangular structure emerges from charge order on the honeycomb lattice for large values of @xmath0 and @xmath1 at @xmath2 filling ( @xmath3 , three electrons per two sites on average ) .
charge - poor sites possess spin degrees of freedom , and their spin correlations are found to be antiferromagnetic in most of the phase diagram , while they become ferromagnetic when @xmath0 is much larger than @xmath1 . on the other hand , for the triangular lattice with @xmath0 , @xmath1 , and @xmath4 interactions at @xmath5 filling ( @xmath6 , three electrons per four sites on average ) , considering large values of @xmath0 and a finite @xmath4 ( we set @xmath24 ) , we find that the system shows rich charge orders : a kagome structure emerges for intermediate values of @xmath1 , while a one - dimensional structure is stabilized for large values of @xmath1 .
both examples show an enhancement of geometrical frustration , from the honeycomb to the triangular lattice in the former case , and from the triangular to the kagome in the latter one . the paper is organized as follows . in sec .
[ sec : honeycomb ] , we present the extended hubbard model on the honeycomb lattice and show the possible phases of the model as a function of @xmath0 and @xmath1 obtained with the hartree - fock approximation and with variational monte carlo . we also discuss how @xmath0 and @xmath1 determine magnetic patterns of the emergent charge ordered states by means of perturbation theory . in sec .
[ sec : triangular ] , we present variational monte carlo results for the extended hubbard model on the triangular lattice and discuss possible phases of the model as a function of @xmath1 for large @xmath0 and @xmath25 . finally , in sec .
[ sec : conclusions ] , we draw our conclusions .
, on - site coulomb interaction @xmath0 , and nearest - neighbor coulomb interaction @xmath1 , eq .
( [ eq : hubbard ] ) and eq .
( [ eq : hubbard_2orbital ] ) , on ( a ) the honeycomb lattice and ( b ) its equivalent two - band representation , eq .
( [ eq : hubbard_2orbital ] ) .
blue and red circles denote orbitals @xmath26 and @xmath27 , respectively . ]
we consider an extended hubbard model on the isotropic honeycomb lattice [ see fig .
[ fig : lattice](a ) ] where the hamiltonian is given as @xmath28 @xmath9 denotes the hopping parameter , and @xmath0 and @xmath1 are the on - site and nearest - neighbor coulomb interaction , respectively .
hereafter , we investigate repulsive coulomb interactions ( @xmath29 ) , and focus on @xmath2 filling ( @xmath3 ) .
we note that on the honeycomb lattice @xmath2 and @xmath30 fillings are equivalent via the particle - hole transformation .
this model , being defined on a lattice with two sites per unit cell , can be also regarded as a two - band ( two - orbital ) hubbard model whose hoppings connect only different orbitals [ see fig .
[ fig : lattice](b ) ] , @xmath31 both representations of the hamiltonian are equivalent and we will make use of the latter representation for computational purposes . in order to investigate the interplay between charge and magnetic order , we start with a mean - field analysis of the above presented hamiltonian . figure [ fig : mf_phase_diag ] ( a ) shows the @xmath0-@xmath1 phase diagram of the honeycomb model of eq .
( [ eq : hubbard_2orbital ] ) at @xmath2 filling obtained with the restricted hartree - fock method ( as explained in appendix [ sec : hartree - fock ] ) . for simplicity ,
we have restricted ourselves to coplanar magnetic order patterns . .
antiferromagnetic ( up - up - down - down ) insulator without charge order . ] filling for @xmath32 obtained by the hartree - fock method .
( a ) energy of each state as a function of @xmath7 .
( b ) magnetization of the metallic ground state ( normal and fm ) .
( c ) density of states for the ground - state fm metal at each value of @xmath7 .
the fermi level is set to @xmath33 .
when the up - spin band is completely below the fermi level , the state becomes semimetallic . ] filling for @xmath34 obtained by the hartree - fock method .
( a ) energy of metallic states as a function of @xmath8 .
( b ) the number of electrons per each sublattice for the metallic ground states ( with and without charge order ) .
( c ) density of states for the ground - state metallic states .
the fermi level is set to @xmath33 . ] in the absence of nearest - neighbor coulomb interaction @xmath1 ( along @xmath32 ) we find four ground - state candidates : normal metal , ferromagnetic metal [ fig .
[ fig : mf_phase_diag](b ) ] , and antiferromagnetic insulator with and without charge order [ see fig . [
fig : mf_phase_diag](d ) and fig .
[ fig : v0_states ] ] . as shown in fig .
[ fig : v0_mf](a ) , the energies of antiferromagnetic and charge ordered states are always higher than those of normal and ferromagnetic metal states .
a continuous phase transition from normal to ferromagnetic metal occurs at @xmath35 , as shown in fig .
[ fig : v0_mf](b ) .
this ferromagnetic metal [ fig .
[ fig : mf_phase_diag ] ( b ) ] is consistent with the result obtained by hanisch _
et al _ @xcite . when @xmath36 , spins are fully polarized . in the ferromagnetic state at @xmath2
filling , the up - spin lower band is fully occupied while the down - spin upper band is half filled and the density of states ( dos ) is zero at the fermi energy indicating a semimetallic behavior .
figure [ fig : v0_mf](c ) shows the dos as a function of @xmath0 . .
( a ) normal metal without charge and magnetic order .
( b ) charge ordered stripe antiferromagnetic insulator .
( c ) charge ordered 120@xmath37 antiferromagnetic insulator .
the energy of the stripe antiferromagnetic state is found to be the lowest one . ] filling for @xmath38 obtained by the hartree - fock method .
( a ) energy of each state as a function of @xmath38 .
( b ) the number of electrons per each sublattice for ground states ( normal metal and charge ordered collinear af state ) .
( c ) magnetization of ground states .
charge - poor ( charge - rich ) sites show large ( small ) magnetization .
( d ) density of states for the ground state .
the fermi level is set to @xmath33 . ] on the other hand , in the absence of on - site coulomb interaction @xmath0 ( along @xmath34 ) and for finite @xmath1 , within the @xmath39 sublattice structure , we obtain a staggered charge order state , where @xmath26-orbital sites are charge - rich ( @xmath40 ) while @xmath27-orbital sites are charge - poor ( @xmath41 ) , as shown in fig .
[ fig : mf_phase_diag](c ) . in the absence of on - site
coulomb interaction ( @xmath34 ) , this charge ordered state does not have any magnetic order . as shown in fig .
[ fig : u0_mf ] , we find a phase transition from the nonmagnetic metal to the charge ordered metal at @xmath42 which is stabilized by splitting the upper and lower bands .
this state is metallic since the upper band is always half filled for @xmath3 .
we consider now the case of large @xmath0 and @xmath1 values , where charge order is expected to generate complex magnetic orders . as shown in fig .
[ fig : mf_phase_diag ] ( a ) , when both @xmath0 and @xmath1 are large , we find a charge ordered antiferromagnetic insulator .
it has a rich - poor - rich - poor type charge pattern , and the charge - poor sites form an emergent triangular structure .
magnetic order is found to be collinear and shows stripe order [ fig .
[ fig : mf_phase_diag ] ( d ) ] . on the other hand ,
when @xmath0 is much larger than @xmath1 , a charge ordered ferromagnetic insulating phase [ fig .
[ fig : mf_phase_diag ] ( e ) ] appears .
it also has triangular - like charge order , and charge - poor sites show dominant ferromagnetic order . in order to investigate the possible antiferromagnetic patterns on the emergent triangular structure
, we consider the collinear and the commensurate spiral state with @xmath43 order of fig .
[ fig : uv_states ] along the @xmath38 line of the phase diagram .
we note that , in general , magnetic states may show incommensurate coplanar spiral order or noncoplanar order in doped hubbard models @xcite
. however , here we restrict ourselves to the coplanar case . as shown in fig .
[ fig : uv_mf ] , the stripe antiferromagnetic charge ordered state is found to have a lower energy than the 120@xmath37 nel ordered state , although the energy difference becomes extremely small as @xmath0 and @xmath1 increase .
the stability of the collinear phase should be induced by second - order processes where a doubly occupied site is formed in the charge - poor sublattice , after the hopping of one electron from the charge - rich sublattice .
indeed , the hopping of one electron from a doubly occupied site to a singly occupied neighboring one is favored when collinearity holds , even for large values of @xmath0 and @xmath1 .
since the intermediate state costs an energy @xmath44 , see for instance the first part of the process in fig .
[ fig : uv_jeff_jnn ] , the energy of the second - order process scales as @xmath45 , as confirmed by the hartree - fock calculations for both stripe and 120@xmath37 nel antiferromagnetic charge ordered states in the insulating phase .
moreover , as discussed in appendix [ sec : perturbation ] , effective next - neighbor exchange couplings on the emergent triangular lattice are generated for moderately large values of @xmath0 and @xmath1 , favoring collinear orderings @xcite .
all these contributions may then break down the 120 nel spin state , and induce the observed collinear pattern . ) and ( [ eq : h_perm ] ) and ( c ) a four - site spin ring exchange . ] in the previous section , we showed that when correlations generate staggered charge order patterns on the honeycomb lattice , charge - rich and charge - poor sites form triangular lattices , respectively . at @xmath2 filling , charge - rich sites contain two electrons on average , while charge - poor sites contain one electron on average with spin degrees of freedom . in order to investigate how magnetic order appears in this limit , we apply perturbation theory to obtain an effective spin hamiltonian . at the lowest order , the effective low - energy spin hamiltonian on the triangular lattice ( see fig . [
fig : uv_2_3_4_interaction ] ) contains a spin exchange interaction and a three - particle permutation term , namely , @xmath46 and @xmath47 ( further details are given in appendix [ sec : perturbation ] . ) here , the sum is taken over all nearest - neighbor sites denoted by @xmath48 for @xmath49 , while it is taken over all triangles , which connect charge - poor sites , for @xmath50 , as illustrated in fig .
[ fig : uv_2_3_4_interaction ] .
the symbol @xmath51 denotes a cyclic permutation operator . by considering virtual hopping processes , as shown in fig .
[ fig : uv_jeff_jnn ] , the exchange interaction @xmath52 can be evaluated as a function of @xmath9 , @xmath0 , and @xmath1 , namely , @xmath53 with a positive constant @xmath54 .
this results in antiferromagnetic spin correlations .
similarly , the coefficient @xmath55 in the permutation terms can be evaluated by considering six cyclic processes for right - pointing triangles that lie inside the hexagons ( @xmath56 ) and for left - pointing triangles that connect three hexagons ( @xmath57 ) . in appendix
[ sec : perturbation ] we show in fig .
[ fig : uv_jeff_jr3 ] one of the virtual processes generating @xmath56 , which does not require the formation of intermediate double - occupied sites .
this coefficient survives even for @xmath58 , namely , @xmath59 with a positive constant @xmath60 .
when @xmath61 , the formation of intermediate double - occupied sites leads to other six cyclic process in @xmath57 and in @xmath56 . since @xmath62 can be mapped to two - spin exchange operators @xcite , these permutation terms finally result in ferromagnetic exchange interactions .
the effective spin hamiltonian is given by @xmath63 with @xmath64 .
in the limit @xmath65 , antiferromagnetic spin correlations become relevant ( @xmath66 ) , and hence @xmath67 becomes antiferromagnetic . on the other hand , when @xmath68 , antiferromagnetic spin correlations are suppressed ( @xmath69 ) , leading to a ferromagnetic @xmath67
this is consistent with the results in the previous section .
furthermore , when @xmath70 , ferromagnetic @xmath55 and antiferromagnetic @xmath52 nearly cancel out . in this case , higher order processes in the perturbation theory become relevant .
one of the dominant terms is a four - spin ring - exchange interaction @xmath71 on the effective triangular lattice , as shown in figs .
[ fig : uv_2_3_4_interaction ] ( c ) and [ fig : uv_jeff_jr4 ] .
this may induce exotic spin liquid states @xcite or chiral magnetic order @xcite .
moreover , the effective energy scale is extremely small ( @xmath72 ) , which induces highly degenerate low - energy states . filling for @xmath38
obtained using hartree - fock ( mf , top ) and variational monte carlo ( vmc , bottom ) . ]
the presence of charge order is a necessary condition for the validity of the perturbation expansion discussed in sec .
[ sec : spin_correlations ] .
the mean - field approach tends however to overestimate the stability range of ordered phases and to underestimate the stability of nonordered metallic phases stabilized in turn by quantum fluctuations . in what follows
, we investigate the stability of the ordered state performing variational monte carlo simulations .
we prepare the initial states by choosing the charge ordered states found in the restricted hartree - fock method and then optimize the variational parameters .
the details of the method are presented in appendix [ sec : vmc ] .
( phase separated state ) for the honeycomb lattice with @xmath73 sites , as obtained by vmc .
( a ) total charge structure factor @xmath74 .
( b ) charge - rich orbital ( @xmath26-orbital-@xmath26-orbital ) charge structure factor @xmath75 . ( c )
charge structure factor between two different orbitals ( @xmath26-orbital and @xmath27-orbital ) @xmath76 .
( d ) charge - poor orbital ( @xmath27-orbital-@xmath27-orbital ) charge structure factor @xmath77 . the total charge structure factor @xmath74 shows sharp peaks at the achievable smallest wave vector @xmath78 , suggesting phase separation .
since @xmath79 and @xmath74 is similar to the @xmath27-orbital ( charge - poor orbital ) charge structure factor @xmath77 , phase separation mainly occurs in the @xmath27-orbital sector . ]
the hartree - fock calculation suggests that a triangular - like charge order appears at large @xmath1 .
we first check with variational monte carlo ( vmc ) the stability when @xmath0 as well as @xmath1 are large .
we confirm the presence of the insulating state with charge order and stripe antiferromagnetic order at @xmath80 . both the number of electrons per orbital and the magnetization are nearly saturated as shown in figs .
[ fig : uv_vmc](d)[fig : uv_vmc](f ) .
the momentum distribution @xmath81 is a smooth function of @xmath82 [ fig .
[ fig : uv_vmc](f ) ] , suggesting the state to be insulating .
note that our variational wave function also finds a metallic state without charge and magnetic orders at @xmath83 [ see figs . [
fig : uv_vmc](a)[fig : uv_vmc](c ) ] .
figure [ fig : uv_pd ] presents the schematic phase diagram for @xmath84 obtained with the various approaches considered here . besides
, we do not find any indication of phase separation , which can be detected by the divergence of the charge structure factor @xmath74 at the smallest achievable wave vector @xmath85 @xcite .
we now investigate the case of @xmath34 and large @xmath1 , namely , @xmath86 , @xmath87 , @xmath88 , with the vmc method .
note that perturbation theory is not applicable in this case since @xmath0 is not large enough .
when @xmath86 , the charge ordered metallic state found in the mean - field calculation is replaced by a metal without any charge order .
the total charge structure factor @xmath74 , see eq .
( [ eq : def_structure_factor_nq ] ) , shows @xmath78-linear behavior near @xmath89 , suggesting the state to be metallic . on the other hand , when @xmath90 and @xmath88 , we find a charge disproportionate state , where the average number of @xmath26 electrons is larger than that of @xmath27 electrons . as shown in fig .
[ fig : u0_vmc_nq ] , @xmath74 is found to have sharp peaks near @xmath85 , suggesting phase separation @xcite .
the peaks in @xmath74 appear to be dominated by that of the charge - poor @xmath27-orbital charge structure factor @xmath77 ; see eq .
( [ eq : def_structure_factor_nabq ] ) .
this means that phase separation is mainly activated in the @xmath27-orbital sector . in order to clarify the mechanism of phase separation
, we also take a snapshot of this state . as shown in fig .
[ fig : u0_vmc_snap_shot](a ) , phase separation is characterized by the charge ordered insulating state with a 2020@xmath91 structure ( doubly occupied empty doubly occupied empty @xmath91 sites ) and the metallic state with a mixture of doubly occupied and singly occupied sites . following the conventions @xcite , we denote single - occupied sites with spin as `` spinons '' while doubly occupied ( empty ) sites with no spin as `` doublons '' ( `` holons '' ) . for the charge - rich @xmath26
orbital , each site is nearly doubly occupied and there are no holon sites ( empty sites ) [ see fig . [ fig : u0_vmc_snap_shot](b ) ] . on the other hand , for the charge - poor @xmath27 orbital
there are two islands : one formed by holons and the other one that is a mixture of doublons and spinons [ see fig .
[ fig : u0_vmc_snap_shot](c ) ] . in this doublon - spinon mixture region
, each spinon can hop through a doublon sea of the @xmath26 and @xmath27 orbitals .
this does not cost an additional energy if two spinons are not next to each other on the original honeycomb lattice .
each spinon is always surrounded by three doublons , which reside on the nearest - neighbor sites of the honeycomb lattice .
it can be assumed that one spinon and at least one doublon are bound together , and this new quasiparticle freely moves inside the doublon sea .
the total kinetic energy gain is determined by the size of the doublon sea and the effective filling of the new quasiparticles .
we note that the concept of spin - charge separation has been only rigorously defined in one spatial dimension ; however , such a possibility has been also discussed in higher dimensions in the presence of geometrical frustrations @xcite .
our numerical results imply that the system wants to generate a larger doublon sea with spinons to gain kinetic energy .
this mechanism is similar to what has been found in the doped extended hubbard model on a one - dimensional chain @xcite and a two - leg ladder @xcite .
summarizing , a nearest - neighbor coulomb interaction @xmath1 at @xmath34 stabilizes a charge ordered metal at the hartree - fock level .
however inclusion of quantum fluctuations via the gutzwiller approximation ( ga ) , see appendix [ sec : gutzwiller ] , and via finite - size vmc calculations suggests that the charge - ordered metal is replaced by phase separation , see fig .
[ fig : u0_pd ] , although the energies of these two states are found to be very close , as shown in appendix [ sec : u=0 ] .
figure [ fig : u0_pd ] shows the schematic phase diagram for @xmath34 and finite @xmath1 .
we note that the critical @xmath0 is shifted to a larger value in the vmc result . filling and @xmath34 obtained using restricted hartree - fock ( mf , top ) , the gutzwiller approximation ( ga , middle ) and variational monte carlo ( vmc , bottom ) . co denotes a charge ordered phase , while ps denotes a phase separated phase . ]
on the triangular lattice .
the unit cell contains four sites and is denoted by a green parallelogram .
small empty circles denote empty sites , while full red circles denote single - occupied sites .
( b ) effective 1d chains generated by charge order at @xmath6 on the triangular lattice . small empty circles denote empty sites , full small red circles denote single - occupied sites , while full large red circles denote double - occupied sites .
, title="fig : " ] + on the triangular lattice .
the unit cell contains four sites and is denoted by a green parallelogram .
small empty circles denote empty sites , while full red circles denote single - occupied sites .
( b ) effective 1d chains generated by charge order at @xmath6 on the triangular lattice .
small empty circles denote empty sites , full small red circles denote single - occupied sites , while full large red circles denote double - occupied sites .
, title="fig : " ] in this section we investigate the extended hubbard model on the isotropic triangular lattice , as defined by the hamiltonian : @xmath92 where @xmath9 denotes the hopping parameter , @xmath0 is the on - site coulomb repulsion , @xmath1 is the nearest - neighbor coulomb interaction , and @xmath4 is the next - nearest - neighbor one . as in the previous section
, we investigate repulsive coulomb interactions , focusing on the appearance of charge - ordered states induced by a nonlocal potential . here
, we focus on 3/8 filling ( @xmath6 ) , where emergent kagome and one - dimensional structures may be generated by the appearance of charge order .
both effective lattices are shown in fig .
[ fig : effective_lattices ] .
when @xmath93 , double occupancies are prohibited and the charge ordered ground state has a kagome - like structure , with three sites of the unit cell singly occupied and one site empty . by increasing the ratio @xmath15 , the number of empty sites increases in order to avoid the energy loss from the @xmath1 term , thus inducing a one - dimensional ( 1d ) charge structure .
we remark that the presence of a nearest - neighbor coulomb repulsion is not sufficient to stabilize the aforementioned charge orders and one additional next - nearest neighbor @xmath4 is necessary in the triangular lattice case . indeed , if we consider for example the kagome - like order of fig .
[ fig : effective_lattices ] ( a ) , we can describe it as alternate rows that are fully occupied and rows where only half of the sites are occupied .
if interaction is restricted to nearest neighbors , the reciprocal positions of the empty sites between different rows can be changed without any further energy cost , implying the absence of charge order in the system . in the @xmath94 limit
, the energy of the kagome and the 1d phases can be easily computed being equal to @xmath95 for the kagome substructure and to @xmath96 for the 1d one .
the 1d phase is then more favorable than the kagome one when @xmath97 . here
we set @xmath98 , in analogy with our former investigation of charge ordered phases on the triangular lattice @xcite , and @xmath99 . as a function of @xmath8 obtained from vmc calculations for each of the four sublattices @xmath100 , @xmath101 , @xmath102 , and @xmath103 , as present in the effective lattices emerging from the charge ordered @xmath6 triangular lattice , as illustrated in fig .
[ fig : effective_lattices ] .
the data are for @xmath104 , @xmath99 , and a lattice size @xmath105 . ] the model of eq .
( [ eq : hubbard_triang ] ) is studied by means of the variational monte carlo method , the details being presented in appendix [ sec : vmc ] . in order to distinguish the different kinds of charge ordering in the model , we plot in fig .
[ fig : density ] the average electronic density per sublattice @xmath106 , with @xmath107 for each of the four sublattices that build up the unit cell ( see fig . [
fig : effective_lattices ] ) .
our results show that for @xmath108 , the charge is uniformly distributed in the lattice , while for @xmath109 one sublattice depletes , with the electrons forming an effective kagome lattice . in this case
the frustration of the original lattice is effectively enhanced . finally , as expected from the coulomb energy argument , the 1d substructure of fig .
[ fig : effective_lattices ] is stabilized for @xmath110 . as discussed also in the honeycomb lattice section ,
the static structure factor @xmath111 is a good indicator for metallic behavior .
the metallic phase is characterized by @xmath112 for @xmath113 , which implies a vanishing gap for particle - hole excitations . on the contrary ,
@xmath114 for @xmath113 , implies a finite charge gap and insulating behavior @xcite .
the results shown in fig . [ fig : n_q ] indicate that the system is metallic in the absence of charge order ( @xmath115 ) , while the charge ordered state with an effective kagome lattice exhibits an insulating behavior ( @xmath116 ) .
@xmath74 is shown along the path in the brillouin zone connecting the point @xmath117 to the point @xmath118 but a similar behavior can be obtained also along other directions .
the results for the 1d charge ordered phase at @xmath119 indicate also an insulating behavior although we observe a dependence on the path chosen in the brillouin zone , with strong finite - size effects . by increasing the lattice size up to @xmath120
we find , however , an insulating behavior along all the selected paths ( not shown ) . as a function of @xmath121 for different values of @xmath8 .
the data are for @xmath6 and @xmath104 for the triangular lattice and for momenta @xmath78 connecting @xmath117 and @xmath118 . the results for lattice sizes @xmath105 and @xmath122 are superimposed . ] in a similar way , one can consider the small-@xmath78 behavior of the spin - spin correlations @xmath123 to discriminate between a spin gapped and a spin gapless behavior .
our results indicate that the effective kagome lattice , induced by charge order , is characterized by gapless spin excitations , since @xmath124 for @xmath113 ; see fig .
[ fig : s_q ] .
moreover , no peak can be observed in the spin - spin correlations , implying the absence of magnetic correlations , even at the short - range scale .
we point out that gapless spin excitations have been also proposed for the heisenberg model on the kagome lattice , by a similar variational approach @xcite , while the density matrix renormalization group approach suggests a finite gap in the spin excitations @xcite . as a function of @xmath121 for different values of @xmath8 , within the region where the effective kagome lattice is stabilized .
data are shown along the line between @xmath117 and @xmath118 in the brillouin zone on the @xmath105 and the @xmath122 lattice sizes . ]
we finally summarize the vmc phase diagram of the model of eq .
( [ eq : hubbard_triang ] ) at 3/8 filling , as a function of @xmath8 , in fig .
[ fig : pd_triang ] . ) as a function of @xmath8 at 3/8 filling , where we set @xmath104 and @xmath99 .
for @xmath108 we observe a metallic phase with a uniform charge distribution .
for @xmath125 we stabilize the charge ordered insulator with an effective kagome lattice of fig .
[ fig : effective_lattices ] ( a ) .
finally , for @xmath110 , the charge ordered insulator with effective 1d chains of fig .
[ fig : effective_lattices ] ( b ) occurs . ]
in conclusion , by using a combination of hartree - fock , perturbation theory , and variational monte carlo , we have investigated the possibility of novel lattice structures emerging from charge disproportionation in doped systems via strong correlations . in particular , we find an emergent geometrical frustration on bipartite honeycomb lattices , and an enhancement of the underlying geometrical frustration on a triangular lattice when coulomb interactions beyond on - site are considered . concerning the honeycomb lattice , we have found that in the presence of both on - site @xmath0 and nearest - neighbor @xmath1 coulomb interactions , charge order converts the original honeycomb structure at @xmath2 filling into an effective _ half - filled _ triangular lattice where the charge ordered state is characterized by a @xmath126 ordered pattern , while the singly occupied sites have a macroscopic spin degeneracy .
a nonzero hopping @xmath9 lifts the spin degeneracy by forming magnetic order , which can be controlled by the coulomb interactions @xmath0 and @xmath1 .
our analysis via hartree - fock of charge order and spin correlations shows that most of the @xmath127 phase diagram at large values of @xmath0 and @xmath1 is characterized by a charge ordered antiferromagnetic insulator .
this result is corroborated by vmc calculations for selected values of the parameter space .
the emergent antiferromagnetic spin correlations are consistent with the effective antiferromagnetic heisenberg model predicted by our perturbation theory analysis .
when @xmath0 is much larger than @xmath1 , a charge ordered ferromagnetic insulating state appears instead , which is consistent with the results from perturbation theory . by further decreasing @xmath1 ,
charge order completely disappears , and eventually a nagaoka ferromagnetic semimetal @xcite appears for @xmath128 . for @xmath34 and finite @xmath1 ,
we find a charge ordered metal as the ground - state candidate .
inclusion of quantum fluctuations via the vmc method , as well as the gutzwiller approximation , suggests however that this state may be unstable towards phase separation by forming a @xmath129 charge ordered insulating region and a metallic one .
concerning the triangular lattice , we find that effective kagome and one - dimensional lattices are generated at @xmath5 filling ( @xmath6 ) because of the presence of charge order .
we consider a large value of the on - site coulomb repulsion @xmath0 and a small , but finite , value of the next - nearest - neighbor coulomb interaction @xmath99 . by increasing the ratio @xmath15 above @xmath130
, the uniform metallic phase evolves into an insulating state , where the electronic charges form a kagome structure , each site being singly occupied .
the emergence of a kagome lattice out of the original triangular one effectively enhances the frustration of the original lattice .
the behavior of the spin - spin correlations @xmath131 shows that the effective kagome lattice generated by charge order is nonmagnetic , with gapless spin excitations . by further increasing the ratio @xmath15 above @xmath132 ,
the number of empty sites increases in order to avoid the energy loss due to the @xmath1 term , thus generating another charge ordered insulator , where electrons form a one - dimensional charge structure .
note added : recently we became aware of a paper @xcite by sugita and motome that reports the emergence of kagome and one - dimensional charge orders on a triangular extended hubbard model in the presence of spin - orbit coupling .
the authors would like to thank f. becca , a. kim , and s. m. winter for fruitful discussions .
, r.v . , and c.g
. acknowledge the support of the german science foundation through grant no .
sfb / trr49 . l.f.t . acknowledges the support of the italian ministry of education , university , and research through grant no .
prin 2010 2010llkjbx .
the variational monte carlo code , which is used in the honeycomb lattice , is based on a code first developed by tahara @xcite .
for the honeycomb lattice , we have analyzed the extended hubbard model on the honeycomb lattice by ( 1 ) the restricted hartree - fock method , ( 2 ) the gutzwiller approximation , ( 3 ) the variational monte carlo ( vmc ) method , and ( 4 ) perturbation theory .
for the extended hubbard model ( up to second neighbors ) on the triangular lattice we have restricted ourselves to the vmc calculations . in this appendix
we show the details of the methods .
we consider a system size of @xmath133 .
the restricted hartree - fock approximation consists of a mean - field decoupling of the hamiltonian of eq .
( [ eq : hubbard_2orbital ] ) .
the mean - field hamiltonian up to constant terms is given as @xmath134 for a collinear state , we assume @xmath135 which contains @xmath136 independent parameters . here
, @xmath137 with the momentum @xmath138 for a @xmath39 sublattice
. we can rewrite the hamiltonian as @xmath139 with @xmath140 , \\
y_{\bm{k},\sigma } & = & - v [ \delta\chi_{\sigma } + \delta\eta_{\sigma } ( e^{-ik_x}+e^{-ik_y})].\end{aligned}\ ] ] we further diagonalize the hamiltonian and determine the above parameters self - consistently . the self - consistent equations are given by @xmath141 , \quad\quad \\
n_{\sigma}^{d } & = & \frac{1}{n } \sum_{k}^{\mathrm{rbz } } \left [ \left < d_{k,\sigma}^{\dagger } d_{k,\sigma}\right > + \left < d_{k+q,\sigma}^{\dagger } d_{k+q,\sigma}\right > \right ] , \\
\delta n_{\sigma}^{c } & = & \frac{1}{n } \sum_{k}^{\mathrm{rbz } } \left [ \left < c_{k,\sigma}^{\dagger } c_{k+q,\sigma}\right > + \left < c_{k+q,\sigma}^{\dagger } c_{k,\sigma}\right > \right ] , \\ \delta n_{\sigma}^{d } & = & \frac{1}{n } \sum_{k}^{\mathrm{rbz } } \left [ \left < d_{k,\sigma}^{\dagger } d_{k+q,\sigma}\right > + \left < d_{k+q,\sigma}^{\dagger } d_{k,\sigma}\right > \right ] , \\ \chi_{\sigma } & = & \frac{1}{n } \sum_{k}^{\mathrm{rbz } } \left [ \left < d_{k,\sigma}^{\dagger } c_{k,\sigma}\right > + \left < d_{k+q,\sigma}^{\dagger } c_{k+q,\sigma}\right > \right ] , \\ \delta\chi_{\sigma } & = & \frac{1}{n } \sum_{k}^{\mathrm{rbz } } \left [ \left < d_{k,\sigma}^{\dagger } c_{k+q,\sigma}\right > + \left < d_{k+q,\sigma}^{\dagger } c_{k,\sigma}\right > \right ] , \\ \eta_{\sigma } & = & \frac{1}{n } \sum_{k}^{\mathrm{rbz } } \left [ e^{ik_x } \left < d_{k,\sigma}^{\dagger } c_{k,\sigma}\right>\right.\\ & - & \left .
e^{ik_x } \left < d_{k+q,\sigma}^{\dagger } c_{k+q,\sigma}\right > \right ] , \\
\delta\eta_{\sigma } & = & \frac{1}{n } \sum_{k}^{\mathrm{rbz } } \left [ e^{ik_x } \left < d_{k+q,\sigma}^{\dagger } c_{k,\sigma}\right > \right.\\ & - & \left .
e^{ik_x } \left < d_{k,\sigma}^{\dagger } c_{k+q,\sigma}\right > \right],\end{aligned}\ ] ] with @xmath142 being the number of @xmath82 points in the brillouin zone .
the ground states are obtained by filling the lowest - energy orbitals up to the number of electrons .
we employ an antiperiodic - periodic boundary condition , and set the number of @xmath82 points as @xmath143 .
each state is characterized by the magnetization and the number of electrons for orbitals @xmath26 and @xmath27 , which are defined as @xmath144 similarly , for spiral states , we consider momentum @xmath145 for a @xmath146-site sublattice , namely , the order parameters are given as @xmath147 , @xmath148 , @xmath149 with @xmath150 ( for charge - rich sites ) , @xmath151 with @xmath152 ( for charge - poor sites ) , @xmath153 , and @xmath154 , and the corresponding reduced brillouin zone .
we also calculate the gap in the density of states at the fermi level to check whether the state is metallic or insulating .
we applied the gutzwiller approximation to the extended hubbard model on the honeycomb lattice . in the gutzwiller approximation , the nearest - neighbor correlations @xmath155 in the correlated wave function @xmath156
are evaluated via those in the uncorrelated wave function @xmath157 through the renormalization factor @xmath158 @xcite : @xmath159 the matrix elements of @xmath160 in the uncorrelated wave function @xmath157 are proportional to @xmath161 here , @xmath162 at @xmath2 filling .
generalizing this to the matrix elements of the correlated state , we find @xmath163 for the gutzwiller approximation of the kinetic energy term @xmath164 .
we define occupancies for empty ( @xmath165 ) , singly ( @xmath166 ) , and doubly ( @xmath27 ) occupied sites for orbital @xmath26 ( similarly for orbital @xmath27 ) : @xmath167 the expectation value of the nearest - neighbor coulomb interaction is given as @xmath168 the energy per bond in the gutzwiller approximation is finally given as @xmath169 the @xmath12 factor in the @xmath0 term comes from the number of nearest - neighbor bonds per site . here , the expectation value of the hopping is a function of @xmath170 , and is defined as @xmath171 with @xmath172 . besides , the renormalization factor @xmath158 is a function of @xmath173 , @xmath174 , and @xmath175 .
one can eliminate the spinon @xmath176 and the holon @xmath177 using @xmath178 therefore , the energy becomes a function of the parameters @xmath179 , @xmath180 , @xmath181 , and @xmath182 . naively , the energy minimum can be obtained by the condition @xmath183 in the absence of on - site coulomb interaction ( @xmath34 ) , this yields @xmath184 the expectation value of the doublon is a simple product of the number of up and down spins .
this simplifies the renormalization factor : @xmath185 now , the energy is a function of a few parameters , namely , @xmath179 and @xmath180 .
the energy minimum can be searched analytically .
for @xmath186 , however , the stationary condition does not give us simple conditions .
we , instead , numerically find the energy minimum by controlling parameters @xmath179 , @xmath180 , @xmath181 , and @xmath182 . honeycomb lattice .
compare eq .
( [ eq_a_h_spin ] ) . ] honeycomb lattice .
compare eq .
( [ eq_a_h_spin ] ) . ] as a third method which includes the effects of quantum fluctuations beyond mean field , we consider the variational monte carlo ( vmc ) technique .
we use the jastrow - slater wave functions which allow metallic and insulating states with charge and antiferromagnetic orders . for the honeycomb lattice eq .
( [ eq : hubbard_2orbital ] ) , we define @xmath187^{n_{\mathrm{e}}/2 } \left|0\right > , \\
f_{ij } & = & \left\ { \begin{array}{cc } f^{a}(\bm{r}_j-\bm{r}_i ) & i \in \mathrm{a~sublattice } , \\
f^{b}(\bm{r}_j-\bm{r}_i ) & i \in \mathrm{b~sublattice } , \end{array } \right . \\
p_{\mathrm{cj } } & = & \exp\biggl [ \frac{1}{2 } \sum_{ij } \bigl ( v^{\mathrm{cj},cc}_{ij } n_{i}^{c } n_{j}^{c } \nonumber \\ & & + v^{\mathrm{cj},cd}_{ij } n_{i}^{c } n_{j}^{d } + v^{\mathrm{cj},dd}_{ij } n_{i}^{d } n_{j}^{d } \bigr ) \biggr ] , \\
p_{\mathrm{sj } } & = & \exp\biggl [ 2 \sideset{}{'}\sum_{ij } \bigl ( v^{\mathrm{sj},cc}_{ij } s_{i}^{z , c } s_{j}^{z , c } \nonumber \\ & & + v^{\mathrm{sj},cd}_{ij } s_{i}^{z , c } s_{j}^{z , d } + v^{\mathrm{sj},dd}_{ij } s_{i}^{z , d } s_{j}^{z , d } \bigr ) \biggr ] , \\ v^{\mathrm{sj}}_{ij } & = & v^{\mathrm{sj}}(|\bm{r}_j-\bm{r}_i| ) , \\ v^{\mathrm{cj}}_{ij } & = & v^{\mathrm{cj}}(|\bm{r}_j-\bm{r}_i|).\end{aligned}\ ] ] here , @xmath188 denotes a sum over @xmath189 .
we prepare the slater part @xmath190 by taking the hartree - fock solutions as initial states , and optimize the variational parameters @xmath191 , @xmath192 , and @xmath193 .
( surviving even in the limit @xmath194 ) between the triangular lattice sites emerging from the charge ordered @xmath3 honeycomb lattice .
compare eq .
( [ eq_a_h_spin ] ) . ]
we use eqs .
( [ eq : def_order_magnetization ] ) and ( [ eq : def_order_charge ] ) to characterize each phase .
in order to see whether the phase is metallic or insulating , we calculate the total momentum distribution : @xmath195 and the density - density structure factors for two orbitals : @xmath196 metallic states are detected by the jump of the momentum distribution @xmath81 and @xmath78-linear behavior of the total charge structure factor @xmath197 near @xmath89 . on the other hand , @xmath81 is smooth and @xmath198 ( @xmath89 ) for insulating states .
analogously , in order to simulate the triangular lattice model of eq .
( [ eq : hubbard_triang ] ) , we have used the variational monte carlo method based on the variational ansatz @xmath199 @xcite , where @xmath200 is the noninteracting filled fermi sea , to which a finite small superconductive term is added in order to regularize the wave function , i.e. , to separate the highest occupied and the lowest unoccupied states by a gap .
the term @xmath201 is the density - density jastrow factor @xcite , where the @xmath202 s are optimized with variational monte carlo calculations for every independent distance @xmath203 ( including on site ) . in order to investigate the formation of charge - ordered phases , we include four different chemical potentials in @xmath200 , as variational parameters , one for each site of the unit cell , similarly to what has been done in ref . .
we tested that inclusion of backflow correlations to further improve the correlated state @xmath204 @xcite is not crucial to describe charge - ordered states .
all results presented here are obtained by optimizing individually @xcite every variational parameter in the wave function and then performing a monte carlo sampling of the observables over the optimal state .
the error bars are not shown since they are always smaller than the symbol size . honeycomb lattice .
compare eq .
( [ eq_a_h_spin ] ) . ] on the honeycomb lattice , when @xmath205 , we expect a triangular charge order at @xmath2 filling ( @xmath3 ) .
charge - rich sites contain two electrons per site , and do not have any left spin degrees of freedom . on the other hand ,
charge - poor sites contain one electron per site , and possess macroscopic spin degeneracy .
the ground - state degeneracy is lifted in the presence of hopping @xmath9 . from perturbation theory ,
the effective hamiltonian can be obtained as a sum of the heisenberg spin exchange and permutation terms : @xmath206 for the spin exchange terms the sum is taken over all nearest - neighbor ( next - nearest - neighbor ) sites on an effective triangular lattice for @xmath207 ( @xmath208 ) . on the other hand , for the permutation terms , the sum is taken over all right - pointing ( left - pointing ) triangles which are located inside ( outside ) of hexagons for @xmath209 ( @xmath210 ) and all squares for @xmath211 .
the symbol @xmath51 denotes a cyclic permutation operator .
honeycomb lattice , see eq .
( [ eq_a_j_eff ] ) , as a function of @xmath7 and @xmath8 . ] unlike the perturbation expansion in the triangular and kagome systems where electrons can hop odd times in local triangles , spin exchange interactions appear through only an even number of electron hoppings .
besides , the spin degeneracy is first lifted by a four virtual hopping process rather than a conventional two hopping process . for example , as shown in fig . [
fig : uv_jeff_jnn ] , nearest - neighbor exchange on the triangular lattice appears through four hopping processes .
the intermediate states have energy @xmath44 , @xmath0 , and @xmath44 , respectively
. there are four different ways of exchanging spins .
this gives antiferromagnetic exchange @xmath212 = t^{4}/(v^2u)$ ] .
similarly , next - nearest - neighbor exchange on the triangular lattice is given as @xmath213 with @xmath214 , @xmath215 being positive constants ( see fig .
[ fig : uv_jeff_jnnn ] ) . these exchange interactions are always antiferromagnetic : @xmath216 which results in strong geometrical frustration .
they are similar to what has been found in the quarter - filled extended hubbard models on a two - leg ladder @xcite and a square lattice @xcite .
analogously to spin exchange interactions , permutation terms in the present system always appear through an even number of virtual processes . as shown in fig .
[ fig : uv_jeff_jr3 ] , three - site permutation terms appear through six cyclic processes , namely , @xmath217 [ @xmath218 for triangles in ( out of ) hexagons with @xmath60 , @xmath219 , and @xmath220 being positive constants .
note that @xmath221 survives even when @xmath58 , which gives ferromagnetic interactions .
moreover , four - site permutation terms appear through eight cyclic processes as shown in fig . [
fig : uv_jeff_jr4 ] , namely , @xmath222 with @xmath223 , @xmath224 , @xmath225 , and @xmath226 being constants .
since the three - spin permutation operator can be written as a product of two exchange operators @xmath227 these terms become nearest - neighbor exchange interactions on the effective triangular lattice @xcite @xmath228 @xmath56 and @xmath57 will be renormalized into @xmath52 , and the hamiltonian up to the constant term is rewritten as @xmath229 here , @xmath230 is a linear combination of @xmath52 , @xmath56 , and @xmath57 , namely , @xmath231 .
the effective nearest - neighbor interaction @xmath230 can be both ferromagnetic and antiferromagnetic depending on the size of @xmath7 and @xmath8 , as shown in fig .
[ fig : uv_jeff_pd ] . when @xmath0 is moderately large and @xmath1 is extremely large [ @xmath232 , @xmath233 and hence @xmath234 is antiferromagnetic . since @xmath235 ,
the hamiltonian effectively becomes an antiferromagnetic @xmath230-@xmath236 heisenberg model with four - spin ring exchange interaction @xmath237 .
when @xmath236 is large enough , collinear antiferromagnetic order overcomes 120@xmath37 order @xcite .
when @xmath0 is extremely large and @xmath1 is moderately large [ @xmath68 ] , @xmath239 and hence @xmath240 is ferromagnetic .
since @xmath241 , the hamiltonian effectively becomes a ferromagnetic heisenberg model with small four - spin ring - exchange @xmath237 . finally , when @xmath10 , @xmath242 terms in interactions @xmath243 , @xmath244 , and @xmath245 nearly cancel out , and @xmath230 becomes extremely small @xmath246 . in this case , @xmath247 is much smaller than @xmath230 , while @xmath248 is comparable to @xmath230 .
the hamiltonian effectively becomes a four - spin ring - exchange model .
in this appendix we present details on the calculations of the phase diagram for the extended hubbard model on the honeycomb lattice at @xmath3 by employing the gutzwiller approximation . as we have discussed in the main text , by using first the restricted hartree - fock method
, we find the transition from a normal metal to a ferromagnetic metal at @xmath35 .
when @xmath36 , spins are fully polarized .
an up - spin band becomes fully occupied , while a down - spin band becomes half occupied .
since the up- and down - spin bands are similar to the original honeycomb band , the fermi level is located at the dirac node of the down - spin band ( semimetallic ) . in order to assert the stability of the ferromagnetic state against quantum fluctuations beyond the mean - field treatment , we apply the gutzwiller approximation . by assuming that the @xmath26 and @xmath27 orbitals are equivalent
, we obtain the energy per bond as where @xmath251 is the number of electrons , @xmath252 is the magnetization , and @xmath253 is the number of doublons . here , @xmath254 is a function of @xmath255 . by numerically searching the energy minimum for @xmath256
$ ] and @xmath257 $ ] , we find a first - order transition from a normal metal to a ferromagnetic metal at @xmath258 , as shown in fig .
[ fig : v0_ga ] .
inclusion of quantum fluctuations as done in the gutzwiller approximation shifts the critical @xmath0 to larger values than in the hartree - fock approximation ( see fig . [
fig : v0_pd ] ) . quantum fluctuations seem to favor a metallic state without ferromagnetism . obtained for the honeycomb lattice at @xmath2 filling by hartree - fock and by the gutzwiller approximation .
the energy is saturated when the fm metal becomes fully polarized .
( b ) schematic phase diagrams obtained by each method . ] filling for @xmath34 obtained by the gutzwiller approximation . ( a ) charge order parameter @xmath259 as a function of @xmath8 , from the uniform to the charge - ordered state .
the number of electrons of charge - rich sites is given as @xmath260 .
( b ) energy difference between charge ordered and phase separated states .
they are nearly degenerate . ] as we have discussed in the main text , the restricted hartree - fock method finds the transition from a normal metal to a charge ordered metal at @xmath42 .
charge - poor sites form an emergent triangular structure as shown in fig .
[ fig : mf_phase_diag ] ( c ) ; however , we find it to be nonmagnetic .
we now proceed with the gutzwiller approximation .
let us first focus on the uniform charge ordered state .
we assume the absence of magnetization : @xmath262 since the total number of electrons is conserved ( @xmath263 ) , the energy is given as a function of a single parameter @xmath179 : @xmath264 the expectation value of the hopping @xmath254 is a constant since @xmath265 is a constant .
hereafter , we abbreviate @xmath266 . by minimizing the energy for @xmath267
$ ] , we find a phase transition from a normal metal to a charge ordered metal at @xmath268 .
the charge order parameter for @xmath269 is given as @xmath270^{-1/2}}.\ ] ] in the metallic phase the energy is @xmath271 , while in the charge ordered phase the energy is @xmath272 ^ 2}\}$ ] .
thanks to the kinetic energy gain , the latter energy is lower than that of the fully charge ordered ( 2121@xmath91 ) insulating state ( @xmath273 ) .
next , we consider the possibility of phase separation consisting of charge ordered insulator and metal phases .
we separate the system into two regions : in the region @xmath274 a fully charge ordered insulating state ( @xmath275 and @xmath276 ) is realized , while in the region ( @xmath277 ) a metallic state ( the average number of electrons is @xmath278 ) is realized . in the former region
, there is no kinetic energy gain ( @xmath279 ) and no intersite coulomb energy loss ( @xmath280 since @xmath276 ) .
one only has to consider energy in the latter region , which is given as @xmath281.\ ] ] the conservation of charge yields @xmath282 which simplifies the energy @xmath283.\ ] ] here , @xmath254 is a function of @xmath278 .
if the energy is minimized for @xmath284 , phase separation takes place . by numerically searching the energy minimum for @xmath285
$ ] ( @xmath286 $ ] ) , we find a phase separated state for @xmath287 .
the charge ordered and phase separated states are found to be nearly degenerate , as shown in fig .
[ fig : u0_ga ] , with the energy of the phase separated state being slightly lower . 99 m. imada , a. fujimori , and y. tokura , rev .
phys . * 70 * , 1039 ( 1998 ) . g. baskaran , supercond .
* 29 * , 124002 ( 2016 ) .
h. watanabe and m. ogata , j. phys .
. jpn . * 74 * , 2901 ( 2005 ) .
tocchio , c. gros , x .- f .
zhang , and s. eggert , phys .
lett . * 113 * , 246405 ( 2014 ) .
r. kaneko , l.f .
tocchio , r. valent , f. becca , and c. gros , phys .
b * 93 * , 125127 ( 2016 ) .
s. reja , r. ray , j. van den brink , and s. kumar , phys .
b * 91 * , 140403(r ) ( 2015 ) . c. hotta and n. furukawa , phys . rev .
b * 74 * , 193107 ( 2006 )
. m. miyazaki , c. hotta , s. miyahara , k. matsuda , and n. furukawa , j. phys .
jpn . * 78 * , 014707 ( 2009 ) .
l. cano - cortes , a. ralko , c. fevrier , j. merino , and s. fratini , phys .
b * 84 * , 155115 ( 2011 ) .
j. merino , a. ralko , and s. fratini , phys .
lett . * 111 * , 126403 ( 2013 ) .
y. akagi and y. motome , phys .
b * 91 * , 155132 ( 2015 ) .
j. wen , a. ruegg , c .- c .
j. wang , and g. a. fiete , phys .
b * 82 * , 075125 ( 2010 ) .
k. ferhat and a. ralko , phys .
b * 89 * , 155141 ( 2014 ) .
f. pollmann , k. roychowdhury , c. hotta , and k. penc , phys .
b * 90 * , 035118 ( 2014 ) .
k. roychowdhury , s. bhattacharjee , and f. pollmann , phys .
b * 92 * , 075141 ( 2015 ) .
e. wawrzyska , r. coldea , e. m. wheeler , i. i. mazin , m. d. johannes , t. srgel , m. jansen , r. m. ibberson , and p. g. radaelli , phys .
lett . * 99 * , 157204 ( 2007 ) .
e. wawrzyska , r. coldea , e. m. wheeler , t. srgel , m. jansen , r. m. ibberson , p. g. radaelli , and m. m. koza , phys . rev .
b * 77 * , 094439 ( 2008 ) .
p. fulde , a. n. yaresko , a. a. zvyagin , and y. grin , europhys . lett . * 54 * , 779 ( 2001 ) .
p. fulde , k. penc , and n. shannon , ann . phys . * 11 * , 892 ( 2002 ) .
s. hayami , t. misawa , y. motome , jps conf . proc . * 3 * , 016016 ( 2014 ) .
z. y. meng , t. c. lang , s. wessel , f. f. assaad , and a. muramatsu , nature * 464 * , 847 ( 2010 ) .
s. sorella , y. otsuka , and s. yunoki , sci . rep . * 2 * , 992 ( 2012 ) .
s. raghu , x .-
qi , c. honerkamp , and s .- c . zhang , phys .
lett . * 100 * , 156401 ( 2008 ) .
c. weeks and m. franz , phys .
b * 81 * , 085105 ( 2010 ) .
s. capponi and a. m. luchli , phys .
b * 92 * , 085146 ( 2015 ) .
j. motruk , a.g .
grushin , f. de juan , and f. pollmann , phys . rev .
b * 92 * , 085147 ( 2015 ) .
scherer , m.m .
scherer , and c. honerkamp , phys .
b * 92 * , 155137 ( 2015 ) .
m. kurita , y. yamaji , m. imada , phys .
b * 94 * , 125131 ( 2016 ) . e.v .
castro , a.g .
grushin , b. valenzuela , m.a.h .
vozmediano , a. cortijo , and f. de juan , phys .
* 107 * , 106402 ( 2011 ) .
grushin , e.v .
castro , a. cortijo , f. de juan , m.a.h .
vozmediano , and b. valenzuela , phys .
b * 87 * , 085136 ( 2013 ) .
t. hanisch , g.s .
uhrig , and e. mller - hartmann , phys .
b * 56 * , 13960 ( 1997 ) .
k. pasrija and s. kumar , phys .
b * 93 * , 195110 ( 2016 ) .
jolicoeur , e. dagotto , e. gagliano , and s. bacci , phys . rev .
b 42 , 4800(r ) ( 1990 ) .
a. v. chubukov and th .
jolicoeur , phys .
b * 46 * , 11137 ( 1992 ) .
m. roger , j.h .
hetherington , and j.m .
delrieu , rev .
55 , * 1 * ( 1983 ) .
motrunich , phys .
b * 72 * , 045105 ( 2005 ) .
t. grover , n. trivedi , t. senthil , and p. a. lee , phys .
b * 81 * , 245121 ( 2010 ) .
m. holt , b.j .
powell , and j. merino , phys .
b * 89 * , 174415 ( 2014 ) s.e .
korshunov , phys .
b * 47 * , 6165(r ) ( 1993 ) .
k. kubo , t. momoi , z. phys .
* b * 103 , 485 ( 1997 ) . f. becca , m. capone , and s. sorella , phys .
b * 62 * , 12700 ( 2000 )
anderson , g. baskaran , z. zou , j. wheatley , t. hsu , b.s .
shastry , b. doucot , s. liang , physica c * 153 * , 527 ( 1988 ) .
l. balents , nature * 464 * , 199 ( 2010 ) .
d. poilblanc and h. tsunetsugu , _ mobile holes in frustrated quantum magnets and itinerant fermions on frustrated geometries _ , springer series in solid - state sciences , edited by claudine lacroix , philippe mendels , and frederic mila , vol .
164 ( springer , berlin , heidelberg , 2011 ) , p. 563 .
f. mila and x. zotos , europhys .
* 24 * , 133 ( 1993 ) .
k. penc and f. mila , , 9670 ( 1994 ) .
m. vojta , a. hbsch , and r.m .
noack , phys .
b * 63 * , 045105 ( 2001 ) . l.f .
tocchio , f. becca , and c. gros , phys . rev .
b * 83 * , 195138 ( 2011 ) .
y. iqbal , f. becca , s. sorella , and d. poilblanc , phys .
b * 87 * , 060405(r ) ( 2013 ) . s. depenbrock , i.p . mcculloch , and u. schollwck , phys .
109 * , 067201 ( 2012 ) .
y. nagaoka , phys . rev . * 147 * , 392 ( 1966 ) .
y. sugita and y. motome , j. phys .
* 85 * , 073709 ( 2016 ) .
d. tahara and m. imada , j. phys .
* 77 * , 114701 ( 2008 ) . t. ogawa , k. kanda , and t. matsubara , prog .
* 53 * , 614 ( 1975 ) .
zhang , c. gros , t. m. rice , and h. shiba , supercond .
* 1 * 36 ( 1988 ) . c. gros , phys
b * 38 * , 931(r ) ( 1988 ) . m. capello , f. becca , m. fabrizio , s. sorella , and e. tosatti , phys .
94 * , 026406 ( 2005 ) .
tocchio , f. becca , a. parola , and s. sorella , phys .
b * 78 * , 041101(r ) ( 2008 ) . s. yunoki and s. sorella , phys . rev .
b * 74 * , 014408 ( 2006 ) . r.h .
mckenzie , j. merino , j.b .
marston , and o.p .
sushkov , phys .
b * 64 * , 085109 ( 2001 ) . | spontaneous charge ordering occurring in correlated systems may be considered as a possible route to generate effective lattice structures with unconventional couplings . for this purpose
we investigate the phase diagram of doped extended hubbard models on two lattices : ( i ) the honeycomb lattice with on - site @xmath0 and nearest - neighbor @xmath1 coulomb interactions at @xmath2 filling ( @xmath3 ) and ( ii ) the triangular lattice with on - site @xmath0 , nearest - neighbor @xmath1 , and next - nearest - neighbor @xmath4 coulomb interactions at @xmath5 filling ( @xmath6 ) .
we consider various approaches including mean - field approximations , perturbation theory , and variational monte carlo .
for the honeycomb case ( i ) , charge order induces an effective triangular lattice at large values of @xmath7 and @xmath8 , where @xmath9 is the nearest - neighbor hopping integral .
the nearest - neighbor spin exchange interactions on this effective triangular lattice are antiferromagnetic in most of the phase diagram , while they become ferromagnetic when @xmath0 is much larger than @xmath1 . at @xmath10 ,
ferromagnetic and antiferromagnetic exchange interactions nearly cancel out , leading to a system with four - spin ring - exchange interactions . on the other hand , for the triangular case ( ii ) at large @xmath0 and finite @xmath4
, we find no charge order for small @xmath1 , an effective kagome lattice for intermediate @xmath1 , and one - dimensional charge order for large @xmath1 .
these results indicate that coulomb interactions induce [ case ( i ) ] or enhance [ case(ii ) ] emergent geometrical frustration of the spin degrees of freedom in the system , by forming charge order . |
the study of isotopic @xmath2he-@xmath0he mixed clusters is attracting a growing interest in recent years . from the experimental viewpoint ,
the diffraction of clusters from a transmission grating @xcite has opened new perspectives in the detection and identification of small clusters .
there is no evidence for the existence of the dimer @xmath2he@xmath0he , but clusters @xmath2he@xmath3@xmath0he with @xmath4 have been definitely detected @xcite .
it seems possible at present to resolve clearly the clusters of mass up to about 25 amu @xcite .
as the weakly van der waals he - he interaction is isotope independent , the properties of such mixed clusters are determined solely by quantal effects , namely the different zero - point motion and the different statistics of the two isotopes .
it turns out that helium clusters are weakly bound systems , and the lighter ones are challenging for microscopic theoretical methods . the stability of small mixed clusters has been the object of several recent microscopic studies .
guardiola and navarro have investigated clusters containing up to eight @xmath2he atoms and up to 20 @xmath0he atoms , based on both a variational monte carlo ( vmc ) wave function @xcite and the diffusion monte carlo ( dmc ) method @xcite in the fixed - node approximation .
bressanini and collaborators @xcite have considered clusters with up to 17 @xmath2he and up to three @xmath0he atoms , by means of the dmc method , also in the fixed node approximation . the dmc description is based in an importance sampling wave function which plays a triple role : it controls the variance of the ground - state energy , it carries on the quantum numbers and other properties of the considered cluster , and it specifies the nodal ( or set of nodal ) surfaces . in particular , in refs .
@xcite the antisymmetry required for fermions has been taken into account by means of two slater determinants , one for each spin orientation , which have been built up in terms of harmonic polynomials in the cartesian coordinates of each fermion . moreover , a harmonic - oscillator ( ho ) ordering of the fermionic shells has been assumed .
although this seems a reasonable hypothesis , supported by the findings of density - functional calculations in medium sized mixed droplets @xcite , from a microscopic point of view there are no conclusive _ a priori _ arguments in favor of such an ordering .
for instance , to describe mixed systems with three @xmath0he atoms , a configuration with total angular momentum @xmath5 has been assumed in refs .
@xcite , which corresponds to the filling of the single - particle levels @xmath6 .
in contrast , the @xmath7 ho ordering , which has been assumed in refs .
@xcite , results in an angular momentum @xmath8 .
the comparison of respective binding energies indicates that the @xmath8 state has a lower energy than the @xmath5 one . in conclusion ,
a more general criterion to select the shell ordering needs to be specified .
the aim of this paper is to determine the excitation spectrum of a single @xmath0he atom bound to a @xmath2he@xmath3 cluster .
the ordering of these single - particle levels will be relevant to describe mixed clusters with a higher number of @xmath0he atoms .
it is also worth noticing that the knowledge of the one - fermion spectra in terms of the number of bosons is also relevant to determine the constants entering the rate equations which establish the formation chemical process @xcite , and thus the abundances in the production experiments .
our calculations are based in the dmc method , using an importance sampling wave function which carries out the orbital angular momentum of the @xmath0he relative to the @xmath2he@xmath3 cluster .
the dmc procedure is thus adapted so as to determine the lowest energy state of the subspace of orbital angular momentum @xmath9 . in order to obtain the excited states within each subspace of angular momentum
@xmath9 we use an optimized form of the upper bounds provided by the sum rules method .
the paper is organized as follows . in section
ii we briefly review previous investigations of @xmath2he@xmath3@xmath0he clusters . in section
iii we present a detailed description of the method used to study the ground and excited states of these clusters .
our results are presented and discussed in section iv .
some final comments are given in section v.
a pure @xmath2he@xmath10 cluster is described by the hamiltonian @xmath11 where @xmath12 is the mass of @xmath2he , and @xmath13 is the interaction potential .
recent forms @xcite of this interaction are quite similar and we will use along the paper the one known as hfd - b @xcite potential . given that the he - he interaction is a consequence of the interaction between the electrons in the atoms , it is independent of the mass or the spin of the nucleus .
to convert the @xmath14-th atom into an @xmath0he atom , thus dealing with the cluster @xmath2he@xmath15he , corresponds to a simple change in the hamiltonian @xmath16 where @xmath17 is the mass of @xmath0he .
the corresponding many - body problem is then not much different from that of a pure @xmath2he cluster . in the rest of the paper
we will use the subindex @xmath18 instead of @xmath19 , to alleviate the notation .
mixed @xmath2he@xmath3@xmath0he clusters containing a single fermion have been investigated using several theoretical methods .
the first systematic study of the excitation spectrum of the @xmath0he atom was made by dalfovo @xcite , based on a zero - range density functional .
the use of a non - local finite - range density functional @xcite results in small quantitative differences , related to the fact that finite - range functionals are more repulsive than zero - range ones . as the size on the drop increases ,
the @xmath0he atom is pushed to the surface region , due to its large zero - point motion , and for large enough clusters , the centrifugal term @xmath20 entering the schrdinger equation for the @xmath0he atom can be treated as a perturbation .
actually @xcite , the general trend of the spectrum is a series of tight rotational bands on top of radial excitations , related to the number of nodes in the radial wave function . in the limit of very large @xmath1 ,
the spectrum of the @xmath0he atom becomes independent of the quantum angular momentum , forming the analogous of the two - dimensional andreev states in bulk helium .
the so - called lekner approximation was used in ref .
@xcite , where a vmc calculation was performed , as well as in ref .
@xcite , based on the hypernetted - chain method in its optimized version .
such approximation , used by lekner @xcite to analyze the andreev states in bulk liquid , assumes that the pair correlations between @xmath0he and @xmath2he atoms are the same as those between pairs of @xmath2he atoms , so that the cluster @xmath2he@xmath3@xmath0he can be considered as a perturbation of the cluster @xmath2he@xmath10 , the perturbation being given by @xmath21 further elaboration of the perturbation scheme results in a single - particle schrdinger equation describing the @xmath0he atom with an effective potential given by @xmath22 where @xmath23 and @xmath24 are respectively the @xmath2he particle and kinetic energy densities , both defined in the unperturbed system with @xmath19 bosons .
notice that the laplacian operator acting on the square root of @xmath25 produces a strongly attractive force peaked at the surface of the cluster . in order to classify the resulting spectra , krotscheck and zillich @xcite defined an effective wave number @xmath26 for each excitation characterized by the orbital angular momentum @xmath9 , where @xmath27 is the equivalent hard sphere radius @xmath28 , defined in terms of the root mean square radius @xmath29 of the droplet . by plotting all excitation energies as a function of @xmath30 for a large number of clusters , krotscheck and zillich found that all results fall reasonably well on a universal quadratic line , in nice agreement with the density - functional results .
both sets of results may be approximately pictured as molecular rigid rotor , in which the @xmath2he@xmath15he cluster is viewed as a two body system , the cluster formed by the @xmath1 bosons plus the single fermion , tied by a spring with a rather large rigidity constant .
various angular momentum @xmath9 states are associated to each vibrational - like state of the spring , obeying the law @xmath31 this equation defines the rotational constant @xmath32 ( units of energy ) in terms of the momentum of inertia @xmath33 , the later being proportional to @xmath34 , where @xmath27 is the average distance of the fermion to the center - of - mass of the bosonic cluster . in the case of light clusters ( @xmath35 ) the calculations of ref .
@xcite find appreciable deviations from the universal behavior eq .
( [ constants ] ) .
small @xmath2he@xmath3@xmath0he clusters have been studied in ref .
@xcite based on the dmc method , and in refs .
@xcite without the use of the lekner approximation . to this respect
it is worth stressing that the lekner approximation is basically a _ weak coupling _
description of the interaction of a @xmath0he atom with a @xmath2he cluster , because it does not include the perturbation that the outer @xmath0he atom should generate on the binding cluster .
one may expect this picture to be satisfactory for large bosonic clusters , but inappropriate for small @xmath1 . as a consequence ,
the interesting area to explore corresponds to the case of a small @xmath2he cluster , which may be appreciably modified by the @xmath0he atom .
to carry out the diffusion monte carlo calculation requires an importance sampling or guiding wave function , which incorporates as much as possible the characteristics of the system to be described .
in particular , due to the strong short - range repulsion of the atom - atom interaction , it is advisable to introduce at least two - body jastrow correlations . moreover , the guiding wave function must confine the system and , finally , it has to include the bosonic symmetry related to the @xmath2he atoms and the desired quantum numbers for the @xmath0he atom . in this section
we shall describe a variational wave function which will be used to calculate the lowest - energy states for a given value of the angular momentum @xmath9 of the system .
excited states corresponding to radial excitations will be estimated by means of sume - rule techniques . in order to describe the system @xmath2he@xmath3@xmath0he in a state where the @xmath0he atom is in an orbital angular momentum @xmath9 with respect to the @xmath2he@xmath3 system ,
a simple but nevertheless complete wave function is given by @xmath36 where the subindexes @xmath37 and @xmath18 stand for bosons and fermions , respectively , whereas @xmath38 refers to the mixed boson - fermion part of the wave function .
the bosonic coordinates run from 1 to @xmath1 and the coordinate of the fermion is labelled by @xmath18 .
finally , @xmath39 represents the center - of - mass coordinate of the bosonic subsystem .
this model wave function includes an internal bosonic part ( @xmath40 ) and the coupling of the fermion to individual bosons ( @xmath41 ) as well as to the bosonic cluster ( @xmath42 ) .
we have taken @xmath40 and @xmath41 to be of the jastrow form @xmath43 with @xmath44 the two - body correlation terms include a short - range part , associated with the parameters @xmath45 and @xmath46 , and a long - range confining part associated with the parameters @xmath47 and @xmath48 .
the short - range part is mainly related to the small - distance behavior of the relative two - body wave function .
consequently , the parameters @xmath49 and @xmath50 have been taken to be the same for all systems studied , and in our calculations they have been kept fixed to the values @xmath51 , @xmath52 and @xmath53 , as obtained in our previous calculations for pure @xmath2he and @xmath0he clusters @xcite , by direct minimization of the expectation value of the energy . on the other hand ,
the long - range confining parameters @xmath54 have been determined by means of the vmc method . in the absence of the last term @xmath42 of the importance sampling wave function ( [ importance_sampling ] ) we would describe a state of null angular momentum , explicitly translational invariant and including the bosonic symmetry of the indistinguishable @xmath2he atoms .
the role of the last term @xmath42 of eq .
( [ importance_sampling ] ) , describing the motion of the fermion , is to determine the value of the orbital angular momentum @xmath9 .
it has been taken as a long - range wave function depending on the relative coordinate of the fermion with respect to the center - of - mass of the bosons , @xmath55 .
the explicit form used is the harmonic polynomial @xmath56 this function is particularly simple and corresponds to a state with orbital angular momentum @xmath9 and null third component .
one could have taken a more sophisticated form , by putting a radial dependence different from the simple @xmath57 , but it is reasonable to expect the dmc algorithm to be able to improve this simple and computationally convenient form . by making this part of the trial wave function to depend on the relative distance of the fermion to the center - of - mass of the bosons the translational invariance of the importance sampling wave function is not spoiled .
notice that if we had considered the function to depend on the distance of the fermion to the center - of - mass of the full system , the only difference would have been a trivial scale factor . with the structure of the importance sampling wave functions
one may describe the lowest - energy states for each angular momentum @xmath9 .
it should be mentioned that apart from the case @xmath5 for which @xmath58 , all other cases correspond to functions with a nodal surface , with nodes depending only on the angular variables .
this fact has to be taken into account when using the dmc algorithm .
when considering a subspace of angular momentum @xmath9 the dmc algorithm gives only the energy of the ground state of that subspace , and there is no information about the excited states of the same angular momentum .
a way to have an estimate , actually an upper bound , of the first excited state is to use the sum rules method @xcite .
consider the exact ground state for a given angular momentum @xmath9 , represented here by @xmath59 , and the full set of eigenstates of this subspace ordered by increasing energy and represented by @xmath60 .
let @xmath61 be an arbitrary hermitian operator which may depend on all atomic coordinates , which is assumed to be scalar under rotations , i.e. , to commute with @xmath62 and @xmath63 .
let us consider the sum rule of order @xmath64 @xmath65 = \sum_{(n,\ell)\neq ( 0,l ) } ( e_{n\ell } - e_{0l})^p |\langle\psi_{n\ell}|q|\psi_{0l}\rangle|^2 \ , , \ ] ] where the sum extends to all eigenstates of the hamiltonian but the lowest energy state of angular momentum @xmath9 .
this is important because in order to obtain easily computable properties it will convenient to use the completeness relation .
because of the assumed properties of @xmath66 only states with angular momentum @xmath67 will contribute to the sum .
the @xmath68 rule fulfills the property @xmath69 & \equiv & \sum_{n\neq 0 } ( e_{nl}-e_{0l } ) |\langle\psi_{nl}|q|\psi_{0l}\rangle|^2 \nonumber \\ & \geq & ( e_{1l}-e_{0l } ) m_l^{(0)}[q ] \ , , \end{aligned}\ ] ] from which one obtains an upper bound to the energy of the first excited state of the subspace @xmath9 @xmath70}{m_l^{(0)}[q ] } \ , .\ ] ] the evaluation of the sum rules is simpler than seem , because of the relations @xmath71 & = & \langle\psi_{0l } | q^2 | \psi_{0l } \rangle - |\langle\psi_{0l } | q | \psi_{0l } \rangle |^2 \\ \label{eme1 } m_l^{(1)}[q ] & = & \frac{1}{2}\langle\psi_{0l } | [ q,[h , q ] ] | \psi_{0l } \rangle \ , .\end{aligned}\ ] ] the double commutator may be simplified for a general hamiltonian of the form given in eq .
( [ he3 ] ) , obtaining @xmath72 note that to compute these expressions one only requires knowledge of the ground state wave function of the angular momentum @xmath9 subspace .
this method was used in ref .
@xcite to obtain upper bounds to the first @xmath5 excitation , as well as to the low - lying even - l states .
given that we are obtaining the @xmath73 excitations directly from the dmc procedure , the sum rules method will be used here to obtain the energies of the first excited states in each @xmath9-subspace . in the appendix we use the sum rule method to also estimate @xmath73 excitations based in the knowledge of @xmath74 by relaxing the scalar character of @xmath66 as an alternative to the direct dmc calculations . the upper bound given by eq .
( [ upper_bound ] ) is a functional of the operator @xmath66 , so it may be variationally optimized by equating to zero its functional derivative with respect to @xmath66 .
unrestricted minimization will give rise to the unpractical relation @xmath75 , its solution being equivalent to the solution of the many - body schrdinger equation for the excited state .
an alternative is to optimize the operator inside a restricted subspace , which is the approach followed by chin and krotscheck @xcite and is closely related to the procedure of krisna and whaley @xcite .
here we have followed a simpler procedure , based in the linear expansion of the operator in a basis of easily computable operators .
to determine the basis we have assumed a single - particle - like form for the operator , by considering that it depends only on @xmath76 , i.e. , on the coordinate of the fermion referred to the bosonic center - of - mass @xmath77 this simple form preserves the translational invariance and does not spoil the boson symmetry of the @xmath2he subsystem . for this type of general operator eq .
( [ simple_m1 ] ) can be further simplified , since the following relation @xmath78 holds .
the resulting sum rule @xmath79 becomes @xmath80 where @xmath81 is the reduced mass of the @xmath0he atom and the @xmath2he cluster . in the calculations to be described below the monopole operator
has been optimized by using a simple functional form depending on few parameters , @xmath82 with @xmath83 the minimization of the upper bound of eq .
( [ upper_bound ] ) with respect to @xmath84 , for angular momentum @xmath9 , gives rise to a generalized eigenvalue problem , @xmath85 with a hamiltonian - like matrix @xmath86 and a normalization matrix @xmath87 the lowest eigenvalue of eq .
( [ valorpropio ] ) provides an optimized upper bound . by inserting eq .
( [ q_operator ] ) into eqs .
( [ h_matrix ] ) and ( [ m_matrix ] ) the matrices are further simplified to @xmath88 and @xmath89 it is worth keeping in mind that in our dmc calculations the matrix elements @xmath90 are based in a mixed estimate , so the strict variational character of eq .
( [ upper_bound ] ) may be lost . some insight on the structure of these excitation
may be drawn by considering the leading @xmath91 terms of both the hamiltonian and norm matrices .
for this one - dimensional subspace the upper bound to the radial excitation energy is given by @xmath92 the size of this bound depends on the difference between the mean square radius and the squared mean radius of the fermion with respect to the boson center - of - mass .
this difference will be small if the distribution of the fermion with respect to the boson center - of - mass is sharply peaked near a given value say @xmath93 , corresponding to a rather rigid spring .
then the denominator will be small and the radial excitation will have large energy .
the excited states considered in this section result from excitations related to the distance between the fermion and the center - of - mass of the cluster of bosons .
the underlying optimal wave function ( not determined , however , from the dmc procedure ) should have at least a node along this radial coordinate , in order to be orthogonal to the angular momentum @xmath9 ground state , and may be properly termed as a _
radial excitation_. the dmc algorithm @xcite is nowadays a well - known and used technology .
it is based in integrating the imaginary - time schrdinger equation for an auxiliary function @xmath94 which is the product of a trial wave function @xmath95 and the true ground - state wave function @xmath96 .
the solution is given in terms of an approximate small - time green funtion @xmath97 @xmath98 by means of a series of small time steps @xmath99 .
we have used the @xmath100 approximate green function @xcite which provides an @xmath101 approximation for the energy . in our calculations
we have used a set of 1000 walkers , on the average , a value of @xmath102 , 8000 iterations to settle down the system and 80000 iterations to compute the averages .
for @xmath103 we have used a smaller time step @xmath104 and the number of iterations has been doubled .
the wave function is not definite positive when @xmath105 , and this leads to the known and irritating _ sign problem _ of the dmc algorithm .
we have used the so - called _ fixed node approximation_. the auxiliary function @xmath106 will remain positive if both functions @xmath95 and @xmath107 have ( at any time ) the same nodal surfaces .
the fixed node approximation consists in killing any walker which attempts to cross a nodal surface .
it has been shown @xcite that this procedure leads to an upper bound to the ground state .
we have first performed a vmc calculation to determine the free parameters @xmath47 and @xmath48 of the importance sampling wave function . as mentioned in the previous section ,
the parameters @xmath45 , @xmath46 and @xmath50 are fixed by the atom - atom interaction at very short distances , and do not depend on the size of the cluster . on the contrary , the long - range parameters @xmath54 are very sensitive to the number of bosons .
their values , determined by minimizing the ground state expectation value of the hamiltonian , are reported in table [ params ] .
.optimal values of the parameters ( in @xmath108 ) for the variational description of the system @xmath0he@xmath2he@xmath3 , as a function of the number of bosons @xmath1 .
the last two columns contain the ground state vmc and dmc energies ( in k ) .
figures in parenthesis are the standard deviations of the monte carlo calculations . [
cols=">,>,>,>,>",options="header " , ] in table [ uppbound ] are given the values obtained for these upper bounds , after solving the generalized eigenvalue problem , and are quite close to the dmc excitation energies displayed in table [ spectrum ] .
as in the case of radial excitations , the sum rules have been calculated by means of mixed matrix elements , so that they are not strictly variational .
99 w. schllkopf and j.p .
toennies , science * 266 * , 1345 ( 1994 ) . w. schllkopf and j.p .
toennies , j. chem .
* 104 * , 1155 ( 1996 ) .
o. kornilov and j.p .
toennies ( private communication ) .
r. guardiola and j. navarro , phys .
lett . * 89 * , 193401 ( 2002 ) .
r. guardiola and j. navarro , few - body syst . suppl .
* 14 * , 223 ( 2003 ) .
r. guardiola and j. navarro , phys .
a * 68 * , 055201 ( 2003 ) .
d. bressanini , m. zavaglia , m. mella and g. morosi , j. chem . phys .
* 112 * , 717 ( 2000 ) .
d. bressanini , g. morosi , l. bertini , m. mella , few - body syst .
* 31 * , 199 ( 2002 ) .
d. bressanini and g. morosi , phys .
* 90 * , 133401 ( 2003 ) .
m. barranco , m. pi , s.m .
gatica , e.s .
hernndez and j. navarro , phys .
b * 56 * , 8997 ( 1997 ) . l.w .
bruch , w. schllkopf and j.p .
toennies , j. chem .
* 117 * , 1544 ( 2002 ) .
j. navarro , m. barranco , m. pi , and a. poves , phys .
a * 69 * , in press ( 2004 ) .
aziz , f.r .
mccourt and c.c.k .
wong , mol .
61 * , 1487 ( 1987 ) .
r.a . aziz and m.j .
slaman , jour .
phys . * 94 * , 8047 ( 1991 ) .
tang , j.p .
toennies and c.l .
yiu , phys .
lett . * 74 * , 1546 ( 1995 ) .
f. dalfovo , z. phys .
d * 14 * , 263 ( 1989 ) .
a. belic , f. dalfovo , s. fantoni and s. stringari , phys .
b * 49 * , 15253 ( 1994 ) .
e. krotscheck and r. zillich , j. chem .
phys . * 115 * , 10161 ( 2001 ) .
j. lekner , philos . mag . * 22 * , 669 ( 1970 ) .
r. guardiola , m. portesi and j. navarro , phys .
b * 60 * , 6288 ( 1999 ) .
r. guardiola and j. navarro , phys .
lett * 84 * , 1144 ( 2000 ) .
r. guardiola , phys .
b * 62 * , 3416 ( 2000 ) .
r. guardiola , j. navarro and m. portesi , phys .
b * 63 * , 224519 ( 2001 ) .
s.a . chin and e. krotscheck , phys . rev
. lett . * 65 * , 2658 ( 1990 ) .
s.a . chin and e. krotscheck , phys . rev .
b * 45 * , 852 ( 1992 ) .
rama krisna and k.b .
waley , j. chem .
phys . * 93 * , 6738 ( 1990 ) .
reynolds , d.m .
ceperley , b.j . alder and w.a .
lester jr . , j. chem .
* 77 * , 5593 ( 1982 ) .
moskowitz , k.e .
schmidt , m.a .
lee and h.m .
kalos , j. chem . phys . * 77 * , 349 ( 1982 ) .
j. vrbik and s.m .
rothstein , j. comput .
phys . * 63 * , 130 ( 1986 ) .
siu a. chin , phys . rev .
a * 42 * , 6991 ( 1990 ) .
ceperley , in _ recent progress in many - body theories _ , edited by j.g .
zabolitzky , m. de llano , m. fortes and j.w .
clark , lect . not .
phys . * 142 * , 262 ( springer verlag , berlin 1981 ) | the diffusion monte carlo technique is used to calculate and analyze the excitation spectrum of a single @xmath0he atom bound to a cluster with @xmath1 @xmath2he atoms , with the aim of establishing the most adequate filling ordering of single - fermion orbits to the mixed clusters with a large number of @xmath0he atoms . the resulting ordering looks like the rotational spectrum of a diatomic molecule , being classified only by the angular momentum of the level , although vibrational - like excitations appear at higher energies for sufficiently large @xmath1 . |
99 falter c , bauer t and schnetgke f , 2006 , phys .
* 73 * 224502 bauer t and falter c , 2008 , phys . rev .
b * 77 * 144503 svane a 1992 _ phys . rev .
lett _ * 68 * 1900 vojta m , arxiv : 1202.1913 kohsaka y _ et al _ 2007 science * 315 * 1380 pintschovius l 2005 _ phys .
status solidi b _ * 242 * 30 uchiyama h et al 2004 _ phys . rev .
lett _ * 92 * 197005 graf j _ et al _ 2008 _ phys .
lett _ * 100 * 227002 pintschovius l and reichhardt w 1998 _ neutron scattering in layered copper - oxide superconductors ( physics and chemistry of materials with low dimensional structures vol 20 ) _ ed a furrer ( dordrecht : kluwer - academic ) pintschovius l and braden m 1999 _ phys . rev .
b _ * 60 * r15039 reichardt w 1996 _ j. low temp . phys . _
* 105 * 807 fukuda t
_ et al _ 2005 _ phys .
b _ * 71 * r060501 pintschovius l _ et al _ 2006 _ phys . rev .
b _ * 74 * 174514 mcqueeney r j _ et al _ 2001 _ phys .
lett . _ * 87 * 077001 dastuto _ et al _ 2002 _ phys .
167002 dastuto _ et al _ 2003 _ int . j. mod .
b _ * 17 * 484 braden m _ et al _ 2005 _ phys . rev .
b _ * 72 * 184517 bauer t and falter c 2010 _ j. phys . : condens . matter _ * 22 * 142201 bauer t and falter c 2009 _ phys . rev .
b _ * 80 * 094525 falter c 2005 _ phys .
sol ( b ) _ * 242 * 78 hcker m
_ et al _ 2011 _ phys .
b _ * 83 * 104506 tao wu _ et al _ 2011 _ nature _ * 477 * 191 le boeuf d _
et al _ 2011 _ phys .
rev b _ * 83 * 054506 sakurai y _ et al _ 2011 science * 323 * 698 harrison n and sebastian s e 2011 _ phys .
lett . _ * 106 * 226402 yao h , lee d - h and kivelson s 2011 phys . rev .
b * 84 * 012507 | we calculate for la@xmath0cuo@xmath1 the phonon - induced redistribution of the electronic charge density in the insulating , the underdoped pseudogap and the more conventional metallic state as obtained for optimal and overdoping , respectively .
the investigation is performed for the anomalous high - frequency - oxygen - bond stretching modes ( obsm ) which experimentally are known to display a strong softening of the frequencies upon doping in the cuprates .
this most likely generic anomalous behaviour of the obsm has been shown to be due to a strong nonlocal electron - phonon interaction ( epi ) mediated by charge fluctuations on the ions .
the modeling of the competing electronic states of the cuprates is achieved in terms of consecutive orbital selective incompressibility - compressibility transitions for the charge response .
we demonstrate that the softening of the obsm in these states is due to nonlocally induced dynamic charge inhomogenities in form of charge stripes along the cuo bonds with different orbital character .
thus , a multi - orbital approach is essential for the cuo plane .
the dynamic charge inhomogeneities may in turn be considered as precursors of static charge stripe order as recently observed in la@xmath2ba@xmath3cuo@xmath1 in a broad range of doping around @xmath4 .
the latter may trigger a reconstruction of the fermi surface into small pockets with reduced doping .
we argue that the incompressibility of the cu@xmath5 orbital and simultaneously the compressibility of the o@xmath6 orbital in the pseudogap state seems to be required to nucleate dynamic stripes . in recent work @xcite
we have developed a description of the electronic charge response and phonon dynamics for @xmath7-doped and @xmath8-doped cuprate high - temperature superconductors ( htsc s ) .
this approach is based upon a microscopic modeling of the change of the charge response across the insulator - metal transition in terms of the compressibility of the electronic system as a primary tool to characterize the corresponding competing ground states depending on doping .
this is achieved starting from the insulating state via a metallic pseudogap state to a more conventional metallic state by consecutive orbital selective incompressibility - compressibility transitions in terms of strict sum rules of the charge response given in ref .
@xcite and listed below .
these sum rules are a proper instrument to describe the competition between electronic localisation and itinerancy which is at the heart of the physics in the cuprates . in the conventional metallic state the electronic partial density of states ( pdos ) at the fermi level @xmath9
is related to the polarizability matrix in the longwavelength limit @xmath10 according to @xmath11 and the total density of single particle states at energy @xmath12 is given by @xmath13 @xmath14 , @xmath15 are the orbital degrees of freedom in the elementary cell of the crystal . in our calculation for lacuo , cu@xmath5 , cu@xmath16 and o@xmath6 orbitals in the cuo plane
are taken into account .
@xmath17 is a wavevector from the first brillouin zone ( bz ) . on the other hand , for the insulating state
we obtain the sum rules @xmath18 and @xmath19 the sum @xmath20 is equal to @xmath21 with @xmath22 the average density and @xmath23 the compressibility of the electronic system .
the latter provides a measure of the charge gap in the electronic spectrum because @xmath23 vanishes as a function of the chemical potential in the gap region .
moreover , @xmath24 provides a measure of the change of the chemical potential with particle number .
approaching the delocalization - localization transition from the conventional metallic region representative for optimal and overdoping , when @xmath7-doping is decreased the cu@xmath5 component of the wave function is admitted to become incompressible , insulator - like in the underdoped state according to our modeling .
this means that the correlated 3d electrons are allowed to become more localized as compared with the conventional metallic state and disappear from the fermi surface ( fs ) . in terms of the sum rule we then have @xmath25 accordingly , the partial density of states is suppressed at the fermi level for the correlated , localized cu@xmath5 orbitals ( orbital intrinsic incompressibility due to correlation ) but not for the more delocalized o@xmath6 orbitals at the o@xmath26 sublattices in the cuo plane , where the holes are predominantly injected in the @xmath7-type cuprates .
so , we have an orbital selective compressible , metallic charge response only for the o@xmath6 states with a renormalized pdos @xmath9 and a loss in the total density of state at @xmath27 ( _ pseudogap _ ) because of the incompressibility of the cu@xmath5 states according to equation ( [ 5 ] ) .
the loss in spectral weight also is never recovered on entering the superconducting state .
the incompressible regions in the pseudogap state compete with overall metallic behaviour and with superconductivity .
our modeling of the pseudogap state in terms of compressible , metallic o@xmath6 states and cu@xmath5 states treated as incompressible and disappearing from the fs because of strong correlation is supported by calculations performed within the self - interaction - corrected ( sic ) local spin density ( lsd ) approximation for la@xmath0cuo@xmath1 @xcite .
here it is shown that the self - interaction pulls the occupied localized cu@xmath5 states below the o@xmath6 band , only little cu@xmath5 admixture is found in the o@xmath6 band .
furthermore , our modeling of the pseudogap state corresponds with a smaller fermi volume as compared with the conventional metallic fermi - liquid state because the fermi volume is determined by the density of the compressible metallic charge carriers ( holes ) of dominantly o@xmath6-type alone while the cu3d states are incompressible as in the mott insulator and can not contribute to the fermi volume quite in contrast to the conventional metallic state where both the o@xmath6 and the cu@xmath5 states are compressible metallic leading to a large fermi volume .
the pseudogap state is separated from the conventional overall compressible fermi - liquid state by an orbital selective compressibility - incompressibility transition of the cu@xmath5 orbital . in the context of our modeling of the pseudogap state
it is interesting to point to another modeling of this state within the framework of the so called fractionalized fermi liquid .
for a recent discussion see e.g. ref . @xcite and references therein . in this approach
an idea is important which is based on the concept of an orbital - selective mott transition . for the latter ,
e.g. in a two band model , we have one metallic band with fermi - liquid like quasiparticles while the other band , e.g. related to localized d electrons , undergoes a mott transition to an insulating state due to strong electronic correlations .
such a transition is analogous to the compressibility - incompressibilty transition of the cu3d orbital in our approach . because particle density fluctuations scale with the compressibility of the system the sum rules introduce varying fluctuations of the particle number in the different metallic states of the htsc according to our modeling .
the tuning parameter for these fluctuations of the particle number is doping and the pseudogap state may be considered as a state with reduced particle density fluctuations as compared with the optimal and overdoped state . moreover ,
the partial incompressibility of the cu@xmath5 orbitals competes with superconductivity because the phase fluctuations @xmath28 of the order parameter grow with decreasing particle fluctuations @xmath29 due to the uncertainty relation @xmath30 .
qualitatively , the incompressibility of the cu@xmath5 states also corresponds to a compartmentalisation of configuration space .
this means that some parts of direct space can not be approached or are hardly accessible to the charge carriers . on the other hand , this should be accompanied by a reciprocal compartmentalisation of momentum space , according to the uncertainty relationship @xmath31 . as a consequence
the momentum should have corresponding fluctuations and the fs of the conventional metallic state becomes fuzzy when passing to the underdoped state and possibly reconstructs .
arpes studies and quantum - oscillation measurements seem to reflect this in form of a reconstruction of the large fs in the optimal and overdoped state into small fermi surface pockets with fermi - liquid like charge carriers in the underdoped pseudogap state characterized by a loss of density of states at the fermi level consistent with our modeling . in our orbital based local picture
the pseudogap state of the cuprates looks like a `` two component '' electronic structure where a real space organisation of the low lying charge excitations is achieved in form of a metallic charge response by mobile holes on the oxygen network in the cuo plane that is blocked at the incompressible insulator - like cu sites .
inspection of the sum rule in equation ( [ 5 ] ) shows that the incompressible cu@xmath5 states and the compressible o@xmath6 states are entangled by the off - diagonal elements of the polarizability matrix .
so our modelling of the pseudogap state is consistent with a strange not overall compressible metallic state coexisting with atom - like , incompressible cu@xmath5 orbitals similar as in the undoped , insulating state favoring short ranged magnetic spin correlations and corresponding excitations , typically spin fluctuations .
the latter should be damped by electron - hole charge excitations of the compressible metallic o@xmath6 states . in @xmath32-space language
such a strange mixed locally compressible - incompressible state should result in a heavily renormalized electronic bandstructure and fermi surface ( fs ) as compared with the more conventional overall compressible fermi liquid state with a large fs at higher doping .
the pseudogap state displaying properties characteristic of both , insulators and correlated metals evolves from the incompressible insulating state with no low energy charge excitations and seems to promote the tendency to charge and spin ordering that breaks rotational and translational symmetry , e.g. in form of stripes .
the latter appear to be quite common in underdoped cuprates . in this context
it is interesting to point out that the stripe - like modulation observed in @xcite is strongest on the oxygen orbitals .
so , the latter are very essential to understand the physics of charge order in the cuprates and also the charge response as shown below . finally , the enhanced phase fluctuations due to the incompressibility of the cu@xmath5 orbital favour pairing correlations in the normal state , possibly incoherent preformed pairs as discussed in the literature . and @xmath33 branch of la@xmath2sr@xmath3cuo@xmath1 for various doping levels ( ref .
4 ) .
@xmath34 : @xmath35 , @xmath36 : @xmath37 , @xmath38 : @xmath39 , @xmath40 : @xmath41 .
the lines are a guide to the eye .
( b ) calculated results of the phonon branches shown in ( a ) as obtained with our model approach for the electronic state of the htsc s @xcite .
@xmath42 : insulating state , @xmath43 : underdoped state , @xmath44 : optimally doped state , @xmath45 : overdoped state . for comparison ,
the results for the rim are shown @xmath46 in order to demonstrate the crucial influence of the _ nonlocal _ epi effects in form of cf s on the dispersion . ]
our modeling of the density response is strongly supported by corresponding calculations of phonon dynamics @xcite , which compare well with the experimental data .
this is shown for @xmath7-doped lacuo in case of the strongly doping dependent obsm phonon anomalies , see , where the experimental results for the highest @xmath47 and @xmath33 branches of la@xmath2sr@xmath3cuo@xmath1 are shown for various doping levels @xcite and compared with the calculated results .
the high - frequency obsm are generally known to display an anomalous softening upon doping in the cuprates @xcite .
such a behaviour of the experimental phonon dispersion is clearly visible for lacuo in a and is very well accounted for by the calculation in our model for the electronic state in terms of incompressibility - compressibility transitions , see b. we also have included in b the phonon dispersion of a rigid - ion model ( rim ) for lacuo .
it seems appropriate to explain at least the ideas of our modeling of phonon dynamics and charge response of the cuprates .
a quantitative treatment of the theory and the modeling would be too extensive and can be found in refs .
@xcite .
the rigid part of the electronic charge response and the epi is approximated by an ab initio rigid - ion model ( rim ) , taking into account covalent ion softening in terms of ( static ) effective ionic charges calculated from a tight - binding analysis .
in addition , scaling of the short range part of certain pair potentials between the ions is performed to simulate further covalence effects in the calculation in such a way that the energy - minimized structure is as close as possible to the experimental one .
the rim with the corrections just mentioned then serves as an unbiased reference system for the description of the cuprate superconductors and can be considered as a first approximation for the insulating state of these compounds because of the strong ionic nature of bonding in the cuprates .
starting with such a rigid reference system nonlocal , electronic polarisation processes are introduced in the form of more or less localized electronic charge fluctuations ( cf s ) at the outer shells of the ions . especially in the metallic state of the cuprates the latter
dominate the nonlocal contribution of the charge response and the epi and are particularly important in the cuo planes .
in addition , anisotropic dipole fluctuations are admitted in our approach .
they prove to be specifically of interest for the ions in the ionic layers . comparing in b the results of the phonon dispersion of the rim , where only local epi effects are considered , with the full calculation we can extract the crucial influence of the nonlocal epi effects generated by the charge fluctuations . for example
we find a very strong softening of the @xmath48 bond - stretching mode ( half - breathing mode ) increasing with doping up to about 5 thz .
comparing the calculated results with the experiments in a we find a good agreement . in order to achieve such a realistic description of the phonon modes , both
, nonlocal epi effects and multi - orbital effects are required .
the comparison with the results from the rim demonstrates that nonlocality is needed and from our detailed calculations we find that the cf s of the more delocalized o@xmath6 and cu@xmath16 orbital contribute significantly to the softening of the modes , besides the more localized cu@xmath5 orbital .
the strong softening of the phonon anomalies is accompanied by corresponding dynamic charge inhomogenities in form of charge stripes along the cuo bonds .
the modes are not really soft but one may consider these dynamic stripes as precursors of static charge stripe order with ordering vectors corresponding to the wavevectors of the phonon anomalies . finally , it should be remarked that a constructive interplay of electron - ion interaction with short - ranged antiferromagnetic order is important for the cu sites in the underdoped and insulating state , because electrons are transferred from the cu ion where the cu bond is compressed to the cu ion where the bond is stretched , see e.g. .
both ions have opposite spin in the antiferromagnetic case .
in contrast if a ferromagnetic order would be present the pauli blocking would suppress the transport of electrons with the same spin and consequently also the charge fluctuations and the corresponding softening of the phonon modes .
modelling of the electronic polarizability @xmath49 for the insulating and pseudogap state is needed because a quantitative calculation of @xmath49 is not available due to the missing of a rigorous knowledge of this state . on the other hand , for the more conventional metallic state @xmath49
can be calculated within the local density approximation ( lda ) , however , with some important corrections concerning the electronic @xmath50-axis response which is overestimated in typical lda - based calculations @xcite . concerning this case
it should be remarked that a comparison of the calculated obsm anomalies from the model in b with corresponding results obtained within the framework of the corrected lda @xcite , taking into account the full @xmath17 dependence of the polarizability matrix , leads to a good agreement .
this also speaks in favour of our modeling of the charge response in terms of the compressibility sum rules , in which only the longwavelength limit of the polarizability is considered . for the @xmath48 mode within the rigid ion model ( rim ) for la@xmath0cuo@xmath1 .
( b ) same as ( a ) , but for the nonlocal part of the charge response in the optimally doped state due to nonlocal epi effects in the half - breathing mode @xmath48 .
the displaced ions are indicated by a dot with an arrow .
full lines ( @xmath51 mean that electrons are accumulated in the corresponding region of space and the broken dotted lines ( @xmath52 ) indicate regions where the electrons are pushed away . @xmath53 is given in units of @xmath54 . ] in we display the calculated phonon induced charge redistribution of the @xmath48 bond - stretching mode for the optimal doped state .
figure [ fig02]a illustrates the effect of the _ local _ epi within the rim connected with the displacement of the @xmath55 ions in the cuo plane , i.e. a stretching or compression of the cuo bond , respectively , ( half - breathing mode ( hbm ) ) .
the local epi reveals a charge redistribution of dipole character localized only at the ions as can be expected from the rigid shift of the ionic densities in the rim during the phonon mode . from a we also obtain the message that no dynamical charge stripes are excited by local epi . on the other hand ,
b that displays the nonlocal part of the charge response due to nonlocal epi effects of charge fluctuation type demonstrates , how the strong nonlocal epi generates dynamic charge stripes in the half - breathing mode .
thus , the moving @xmath55 ions in the hbm which are linked up with strong cf s on the silent cu and @xmath56 ions induce dynamical stripes in the @xmath57 direction with a recurrence periond of two lattice constants along the orthogonal @xmath58-direction .
note , that there are no changes of the transfer integral between @xmath59 and @xmath7 orbitals for the silent cu and @xmath56 ions , nevertheless there is a charge transfer . in case of the full breathing mode @xmath60 , at the the @xmath61 point of the bz where the @xmath55 and @xmath56 ions vibrate in phase along the bonds against the silent cu ion @xcite nonlocal epi generates cf s at the cu and we obtain an electronic charge transfer from that cu ion where
the bonds are compressed to the cu where the bonds are stretched @xcite . while in case of the hbm the charge stripes point along the @xmath58- or @xmath57-direction , respectively , for @xmath60 the stripes point along the diagonals in the cuo plane .
the self - consistent changes of the potential felt by an electron due to nonlocal epi also have been calculated for the obsm and yield a strong coupling of the order 100 mev @xcite .
it should be remarked that such a strong nonlocal epi associated with the obsm , which can reliably be calculated in the _ adiabatic _ approximation , is further enhanced by an order of magnitude because of the poorly screened long range coulomb interaction and the slow charge dynamics for the polar modes from the _ nonadiabatic _ sector of phase space around the @xmath50-axis where phonon - plasmon mixing is relevant @xcite . for @xmath48 in the insulating state ( @xmath62 ) .
( b ) same as figure 2 but for the nonlocal part of the charge response @xmath53 for @xmath48 in the pseudogap state ( @xmath63 ) . ] in a and b we present the calculated results for the nonlocally induced charge redistribution of the hbm for the insulating and the pseudogap state , which should be compared to the corresponding results for the optimal doped state in b. first of all we find for _ all _ electronic states dynamical stripes along the bond direction excited by the nonlocal epi being simultaneously responsible for the corresponding phonon softening in .
this is quite interesting because dynamic charge inhomogenities of this type may be considered as precursors of static charge stripe order as recently observed in la@xmath2 ba@xmath3 cuo@xmath1 , in a broad range of doping around @xmath4 with @xmath64 @xcite or in yba@xmath65cu@xmath66o@xmath67 , respectively , @xcite . to the best of our knowledge in overdoped samples
static stripe order has not been observed so far .
this would mean that the pseudogap in the underdoped state may be required to nucleate dynamic stripes . in our approach
this implies that for the emergence of static stripe order the incompressibility of the cu@xmath5 orbitals seems to be essential .
the latter brings about a separation of mobile holes populating the compressible o@xmath6 states and regions with local mott- and spin correlations related to the insulator like cu@xmath5 states .
such stripes producing the onset of a new periodicity that breaks translational and rotational symmetry also may help to trigger a reconstruction of the large hole fermi surface ( fs ) into small pockets as detected by quantum oscillation measurements in underdoped yba@xmath0
cu@xmath68 o@xmath69 , see e.g. @xcite and references therein . also from an energetic point of view our modeling of the pseudogap state is susceptible for the tendency to establish broken symmetry states with hole - rich compressible and hole - poor incompressible regions . in general terms lowering of the kinetic energy
occurs in the compressible metallic regions and lowering of the interaction energy is possible by antiferromagnetic spin correlations and the coulomb repulsion between the electrons in the region of the incompressible insulator - like cu@xmath5 states .
altogether , in the pseudogap state the doped holes and the superexchange - coupled spins arrange themselves in a cooperative way to promote both charge mobility and locally antiferromagnetic correlations . comparing the stripe patterns of the insulating and underdoped state with that of the more conventional metallic optimal and overdoped state ( not shown ) with a large fs as the locus of the crucial low energy excitations , we find a relatively small contribution of the o@xmath6 orbitals degrees of freedom to the density redistribution in the latter case .
the weight of the o@xmath6 contribution is significantly enhanced in the insulating and pseudogap state , see a and b reflecting in particular the importance of the o@xmath6 orbital in the pseudogap state .
holes in the pseudogap state primarily populate the o@xmath6 orbitals , while the holes in the more conventional metallic state gain more and more cu@xmath5 character by an increased hybridization as can be seen from a large pdos for cu@xmath5 leading to a dominating contribution of the charge redistribution at the cu ion where the charge response is now metallic , too .
see also @xcite where it is investigated how the nature of the hole state evolves with doping .
overall , the stripe pattern is more delocalized in the metallic state and localization increases towards the insulating state via the pseudogap state . in the overdoped state a growing importance of the cu@xmath16 orbital and accordingly
an increase of the cu@xmath16 element of the polarizability matrix leads to an enhanced softening of the anomalous obsm , see .
the corresponding calculated dynamical stripe pattern is very similar to that of the optimally doped state in b but somewhat more delocalized because of the increased contribution of the more delocalized cu@xmath16 contribution . from our calculations
it becomes obvious that a multi - orbital approach is needed that includes besides cu@xmath5 at least o@xmath6 orbitals and for the optimally doped and overdoped state in addition cu@xmath16 orbitals in order to give a reliable representation of phonon dynamics and charge response in the cuprates .
single - band models for the cuo plane believed by many people to describe the cuprates are not sufficient , in particular in context with the discussion of stripes . altogether , in the optimal and overdoped state increasing of doping enhances the hybridization between cu@xmath5 , cu@xmath16 and o@xmath6 orbitals , respectively , with a dominant weight for cu@xmath5 .
this state becomes compressible metallic with a large pdos at @xmath27 .
so , the particle number fluctuations that characterize the different competing ground states in our model are enhanced as compared to the ground state in the pseudogap regime . here
the hybridization is weaker due to the incompressible insulator - like cu@xmath5 orbital with a vanishing pdos at @xmath27 . for @xmath70 in the optimally doped state .
] translational symmetry breaking models of the electronic structure of the cuo plane generated by charge density wave order of a certain wavevector are discussed in the literature and used to characterize fermi surface reconstructions of a large fs into small pockets , for recent work see @xcite . from scanning tunneling microscopy ( stm ) , neutron diffraction experiments and nuclear magnetic resonance measurements
the ordering in many cases occurs with a period of about 4 lattice constants @xcite .
this corresponds to the wavevector @xmath71 of the @xmath70 mode .
therefore it is meaningful to calculate the nonlocally induced charge redistribution also for this mode .
the result for the optimal doped state is shown in .
we again obtain a stripe like charge pattern with translational symmetry in the @xmath72 direction but a four unit cell repeat distance along the orthogonal direction ( 1,0,0 ) .
contrasting this result with the charge stripe pattern as found for the hbm in b the stripes for @xmath70 are expanded over the whole unit cell , because the cf s do not vanish at the sites of the moving oxygens , unlike to @xmath48 .
however , the amplitudes of the cf s are considerably smaller . according to our calculation of the eigenvectors
this is caused by the strongly reduced bond - stretching amplitude of the oxygen ions in the plane for @xmath70 , while the displacement of the apex oxygen ions is strongly enhanced in the @xmath70 mode as compared with the hbm . in summary , we have applied our modeling of the doping dependent electronic states of the cuprates to the calculation of the phonon induced charge inhomogeneities nonlocally induced by the @xmath48 and @xmath70 mode in lacuo .
the calculations have been performed for the insulating- , pseudogap and the more conventional metallic state .
the compressibility is used as a primary term to characterize the competing ground states controlled by doping .
the strong nonlocal epi of cf type excites in all cases investigated dynamical charge stripes along the cuo bonds .
differences in orbital character of the charge response in the various states are discussed at hand of the stripe patterns . experimentally , charge stripe order has been observed in a broad range of doping in la@xmath2ba@xmath3cuo@xmath1 and other cuprates which supports the point of view that dynamical charge stripes generated by certain strongly coupling obsm may be considered as precursors of static stripe order at least outside the overdoped regime .
such a charge order breaks translational symmetry and may enforce a reconstruction of a larger hole fs into small pockets . the interplay between the reduced particle density fluctuations related to the insulator - like behaviour of the localized atomic - like cu@xmath5 orbital that remains incompressible for low enough @xmath7-doping and on the other hand the compressible , metallic o@xmath6 orbitals appear to play an important role for the pseudogap state and the tendency to charge ordering in the cuprates . |
let @xmath6 be the class of normalized analytic univalent functions @xmath7 on the unit disk @xmath8 we sometimes write @xmath9 to indicate the function @xmath10 the bieberbach conjecture asserts that @xmath11 for @xmath12 with equality holding only for the koebe function @xmath13 and its rotation @xmath14 it had been a long - standing problem in geometric function theory and was finally proved by de branges @xcite .
at least , soon after the conjecture was posed , it was recognized that @xmath15 holds for @xmath12 with an absolute constant @xmath16 ( see , for example , @xcite ) .
therefore , it is somewhat surprising that the difference @xmath17 is bounded for @xmath18 indeed , hayman proved in his 1963 paper @xcite that @xmath19 in @xmath20 where @xmath21 is an absolute constant .
note that implies that @xmath22 unfortunately , it is known that the constant @xmath23 must be greater than 1 .
in fact , the sharp inequalities @xmath24 hold for @xmath25 in @xmath20 where @xmath26 is the solution @xmath27 in @xmath28 to the equation @xmath29 ( see ( * ? ? ? * theorem 3.11 ) ) . schaeffer and spencer
@xcite showed even that for each @xmath30 there is an _ odd _
univalent function @xmath31 with real coefficients such that @xmath32 ( it is well known that @xmath33 for every odd univalent function @xmath34 see @xcite . ) the problem to find the minimal value for the constant @xmath23 in is not solved yet .
the best result so far is the estimate @xmath35 due to grinspan @xcite .
a function @xmath36 is called _ starlike _ ( resp .
_ convex _ ) if the image @xmath37 is starlike with respect to the origin ( resp .
the class of starlike functions is denoted by @xmath38 and the class of convex functions is denoted by @xmath39 in 1978 leung @xcite proved that holds with @xmath40 for @xmath41 ( see also @xcite ) . for every @xmath42 , the following inequalities hold : @xmath43 for each @xmath44 equality occurs in the left - hand side if and only if @xmath3 is @xmath45 or its rotation @xmath46 with @xmath47 for some integer @xmath48 with @xmath49 likewise , equality occurs in the right - hand side if and only if @xmath3 is @xmath45 or its rotation with @xmath50 for some integer @xmath48 with @xmath51 here , @xmath52 note that one of @xmath53 and @xmath54 vanishes when equality holds in the theorem
. the reader may consult ( * ? ? ?
* , 5.9 and 5.10 ) for more information about the difference of succesive coefficients . in this note
, we will look for a counterpart of theorem a for convex functions . since the result seems to be asymmetric in this case , to clarify the assertion
, we consider the two quantities @xmath55 for @xmath56 note that the suprema can be replaced by maxima in because of compactness of the class @xmath39 it is well known that the sharp inequalities @xmath57 hold for a convex function @xmath58 so that one easily gets @xmath59 and @xmath60 when @xmath61 we easily have @xmath62 and @xmath63 from now on , we thus assume that @xmath64 we will show that @xmath65 and @xmath66 are much smaller than 1 . before presenting our main results , we recall a related result due to robertson @xcite .
let @xmath7 be a convex function .
then for each @xmath30 the following inequality holds : @xmath67 the factor @xmath68 can not be replaced by any smaller number independent of @xmath10 the sharpness of the factor was confirmed by the fact that @xmath69 as @xmath70 where @xmath71 is the convex function given by @xmath72 for @xmath73 here , we should take a suitable limit when @xmath74 for instance , @xmath75 note the relation @xmath76 namely , @xmath71 is a natural counterpart of @xmath45 for convex functions . [
thm : main ] let @xmath77 and @xmath78 be given in .
then the following hold . 1 .
@xmath79 for @xmath80 and the function @xmath81 is extremal for @xmath82 2 .
@xmath83 and @xmath84 is extremal for @xmath85 3 .
@xmath86 and @xmath87 is extremal for @xmath88 in view of ( ii ) and ( iii ) in the theorem , one might expect that @xmath89 and that the function @xmath90 would be extremal for @xmath78 when @xmath91 as well .
it is , however , not true .
we will , in fact , prove the following .
[ thm : main2 ] @xmath92 for each @xmath93 it is an open problem to find the value of @xmath78 for @xmath94 as the triangle inequality implies @xmath95 one may think that the study of the functional @xmath2 would be helpful to our problem .
however , we immediately see that the sharp bound of @xmath96 for convex functions is 2 as the function @xmath97 serves as an extremal one . on the other hand , it is indeed helpful to consider the functional @xmath96 for refined subclasses of @xmath39 for a given number @xmath98 with @xmath99 let @xmath100 note that the union @xmath101 is smaller than the whole class @xmath39 this sort of refined subclasses of @xmath102 were considered , for instance , in @xcite .
on the other hand , each function @xmath3 in @xmath102 is a suitable rotation of a function in @xmath103 for @xmath104 since the functional @xmath17 is rotationally invariant , one can replace @xmath102 by @xmath105 in the definition of @xmath77 and @xmath106 we should note that @xmath96 is not necessarily invariant under rotations
. therefore , it might be meaningful to consider the extremal problem maximizing @xmath96 among the class @xmath103 for a given @xmath5.$ ] noting the relation @xmath107 for @xmath108 robertson s theorem ( theorem b ) implies @xmath109 for @xmath110 in @xmath111 these inequalities can be improved in the following way .
[ thm:2 ] let @xmath112 suppose that @xmath110 is a function in @xmath111 then the following sharp inequalities hold : @xmath113 furthermore , @xmath114 where the supremum is attained by @xmath115 with @xmath116.$ ] likewise , @xmath117 where the supremum is attained by @xmath115 with @xmath118.$ ] we will prove this theorem in section 3 . in section 4 , we will prove theorems [ thm : main ] and [ thm : main2 ]
. the next section will be devoted to preparations for the proofs .
let @xmath119 denote the class of analytic functions @xmath120 with positive real part on @xmath121 which has the form @xmath122 a member of @xmath119 is called a carathodory function .
some preliminary lemmas are needed for the proof of our results .
the first one is known as carathodory s lemma ( see @xcite for example ) .
[ lem : c ] for a function @xmath123 the sharp inequality @xmath124 holds for each @xmath125 the sharpness can be observed through the example @xmath126 we will use also the following result due to carathodory and toeplitz ( see @xcite or @xcite ) . [
lem : ct ] let @xmath127 be a formal power series with complex coefficients . then @xmath120 represents a carathodory function if and only if @xmath128 is non - negative for each @xmath129 where @xmath130 for @xmath131 moreover , if @xmath132 and if @xmath133 then @xmath134 is of the following form : @xmath135 since @xmath136 the numbers @xmath137 must satisfy @xmath138 as a special case with @xmath139 one can deduce the following useful assertion from lemma [ lem : ct ] .
[ lem : d2 ] let @xmath140 be a carathodory function with @xmath141 and @xmath142 then @xmath120 must be of the form @xmath143 moreover , the functions @xmath41 and @xmath144 determined by @xmath145 have the forms @xmath146 and @xmath147 where @xmath148.$ ] we first note that @xmath149 $ ] by lemma [ lem : c ] .
we next observe that @xmath150 by assumption , where @xmath151 is given in with @xmath152 when @xmath153 which is equivalent to the condition @xmath154 the assertion holds clearly with @xmath155 or @xmath156 thus , we may assume that @xmath157 lemma [ lem : ct ] now implies that @xmath120 has the form @xmath158 for @xmath159 and @xmath160 with @xmath161 and @xmath162 we write @xmath163 for @xmath164.$ ] by comparing the coefficients of @xmath165 and @xmath166 in the above formula , we obtain the relations @xmath167 if one of @xmath168 is real , by , the other must be real , too .
this occurs only when @xmath169 so that implies @xmath170 which contraditcs @xmath171 hence , @xmath172 and @xmath173 are both non - real ; i.e. , @xmath174 taking the imaginary part of and , we have @xmath175 since @xmath176 is a non - zero vector , one has @xmath177 which implies @xmath178 hence , we conclude that @xmath179 and that @xmath180 put @xmath181 then , we take the real part of to obtain @xmath182 thus we have seen that @xmath120 has the required form . the remaining part can be easily shown by solving the differential equations @xmath183 the proof is now complete . under the conditions @xmath184
as was seen in the proof , @xmath185 if and only if either @xmath186 or @xmath187 the latter case occurs precisely when @xmath188 for some constant @xmath189.$ ] by making use of the carathodory - toeplitz theorem , libera and zotkiewicz @xcite showed the following lemma .
[ lem : lz ] let @xmath190 and @xmath191 there exists a function @xmath192 with @xmath193 if and only if @xmath194 and @xmath195 for some @xmath196 with @xmath197 and @xmath198 [ rem:-1 ] it is worth observing the following simple fact .
when holds with @xmath199 one has the relation @xmath200 so that the assumption in lemma [ lem : d2 ] is satisfied .
therefore , the form of @xmath120 can be described by lemma [ lem : d2 ] in this case . for given real numbers
@xmath201 the quantity @xmath202 was used in @xcite .
it is also helpful in the current study .
[ lem : y ] let @xmath203 with @xmath204 and @xmath205 then @xmath206 the maximum in the definition of @xmath207 is attained at @xmath208 in the first case according as @xmath209 when @xmath210 we set @xmath211 to be 0 in the above .
the lemma can easily be verified by the fact that @xmath212 in this case .
see the proof of proposition 6 in @xcite for details .
the next elementary result is helpful to show lemma [ lem : f ] below .
we recall that the discriminant @xmath213 of the real quadratic polynomial @xmath214 is defined to be @xmath215 we will allow a degenerated case such as @xmath216 to define it . note that @xmath217 for all @xmath218 if @xmath219 and if @xmath220 [ lem : diff ] let @xmath221 and @xmath222 be ( possibly degenerated ) real quadratic polynomials .
suppose that @xmath223 and @xmath224 on an interval @xmath225 and that @xmath226 if there is a positive constant @xmath227 such that 1 .
@xmath228 and 2 .
@xmath229 for @xmath230 then the function @xmath231 is convex on @xmath232 by ( i ) , we have @xmath233.\ ] ] thus the condition ( ii ) implies that @xmath234 for @xmath235 the following technical result will be used in the proof of theorem [ thm : main ] . [
lem : f ] let @xmath236 and @xmath237 for @xmath238 and consider the function @xmath239 then @xmath240 for @xmath241 fix an @xmath242.$ ] a standard computation yields the expression @xmath243 where @xmath244 @xmath245 and @xmath246 to apply lemma [ lem : diff ] , we put @xmath247 and @xmath248 and we will show the two inequalities @xmath249 and @xmath250 we first note that @xmath251 ^ 2}{48 ^ 2(4-p^2 ) } \ge \frac{(100-p^2)p^4}{64(4-p^2)}.\ ] ] since the last quantity is increasing in @xmath252 we have @xmath253 for @xmath254 on the other hand , @xmath255 and thus @xmath256 the proof of is now completed .
next we consider the quadratic polynomial @xmath257}{4(4-p^2)^2 } + \frac{5p^2r^3}2 t + \frac{6p^4r^2}{4-p^2}t^2.\end{aligned}\ ] ] then its discriminant is @xmath258 where @xmath259 since the partial derivative @xmath260 is negative for @xmath261 and @xmath262 one gets @xmath263 for @xmath264 and @xmath265 it is easy to verify that @xmath266 is increasing in @xmath267 hence , @xmath268 for @xmath254 therefore , we have proved the inequality @xmath269 and therefore .
convexity of @xmath270 implies that @xmath271 for @xmath272.$ ] we will show that @xmath273 to this end , we first observe that the polynomial @xmath274 has two roots @xmath275 and @xmath276 with @xmath277 because @xmath278 moreover , the inequality @xmath279 implies @xmath280 we note also that for @xmath281 @xmath282 if and only if @xmath283 in view of @xmath284 we now obtain @xmath285-\big(|q(-r)|-|q(r)|\big ) \\ & = -4r-|q(-r)|+|q(r)| \\ & = \begin{cases } -4r+2br & \text{if}~ r<|x_-| , \\
-4r+2a-2cr^2 & \text{if}~|x_-|\le r. \end{cases}\end{aligned}\ ] ] here , we see that @xmath286 we also have @xmath287/(4-p^2).$ ] by monotonicity in @xmath288 we observe that @xmath289 therefore , we conclude now that @xmath290 at any event , as required .
we recall the well - known fact that for an analytic function @xmath3 on @xmath121 with @xmath291 @xmath292 if and only if @xmath293>0 $ ] on @xmath294 therefore , for @xmath292 , there is a function @xmath192 such that @xmath295 we write @xmath296 equating the coefficients of @xmath297 in both sides of for @xmath298 we obtain @xmath299 for @xmath108 we have @xmath300 therefore , by lemma [ lem : lz ] , for some @xmath301 we have @xmath302 here , equality occurs if @xmath303 thus has been shown .
since @xmath5 $ ] , we have @xmath304 which proves . here , equalities hold simultaneously precisely when @xmath305 and @xmath306 now lemma [ lem : d2 ] together with remark [ rem:-1 ] gives the required form of the extremal function .
next we show . by substituting and into
, we have @xmath307 where @xmath207 is given in and @xmath308 note here that equalities hold simultaneously in the above for a suitable choice of @xmath309 and @xmath310 a simple computation tells us that for @xmath5,$ ] @xmath311
if and only if @xmath312 thus lemma [ lem : y ] yields @xmath313 hence , we have the estimate @xmath314 for @xmath108 where @xmath315 since @xmath316 is convex in @xmath317 we see that @xmath318 on the other hand , we see that @xmath316 takes its maximum value in @xmath319 at @xmath320 which is @xmath321 therefore , the maximum of @xmath316 is taken at @xmath322 with the choice @xmath305 so that the required form of the extremal function follows from lemma [ lem : d2 ] and remark [ rem:-1 ] .
we begin with the proof of theorem [ thm : main ] . by alexander s theorem ( see ( * ? ? ?
* theorem 2.12 ) ) , @xmath7 is convex if and only if @xmath323is starlike .
thus , by theorem a , we have @xmath324 for a convex function @xmath3 and equalities occur only when @xmath3 is a rotation of @xmath71 given in for some @xmath73 in particular , @xmath325 satisfies @xmath326 thus we have @xmath327 which proves part ( i ) . as we noted in introduction , to compute @xmath328 we can restrict the range of @xmath3 to @xmath329 for @xmath110 in @xmath103 with @xmath99 by and , we have @xmath330 we have thus obtained @xmath331 and equality is attained when @xmath332 , @xmath199 which corresponds to the function @xmath71 with @xmath333 = \pi/3 $ ] by lemma [ lem : d2 ] and remark [ rem:-1 ] . now
the proof of part ( ii ) is complete . in the proof of theorem [ thm:2 ]
, we saw that @xmath334 when @xmath335 on the other hand , because @xmath336 we see that @xmath337 only if @xmath338 hence , by virtue of theorem [ thm:2 ] , we can harmlessly assume that @xmath238 to show @xmath339 for @xmath340 let @xmath343 be the unique positive root of the quadratic polynomial @xmath344 in @xmath345 since @xmath346 we have @xmath347 when @xmath348 , @xmath349 since the right - hand side is increasing in @xmath350 we have @xmath351 for @xmath352 when @xmath353 , @xmath354 the right - hand side is decreasing in @xmath355 so that @xmath351 for @xmath356 hence , we obtain @xmath357 the inequality @xmath358 is equivalent to @xmath359 this can indeed be verified as @xmath360 here , equality holds when @xmath361 by tracing back the above proof , we see that all equalities hold , in addition , when @xmath362 again , by lemma [ lem : d2 ] and remark [ rem:-1 ] , we see that an extremal function is given as @xmath71 with @xmath363=\pi/4.$ ] thus the proof of part ( iii ) has been complete .
first , we recall the fact that @xmath57 for a convex function @xmath25 ( see @xcite ) .
thus , the first inequality in together with this yields @xmath364 here , we note that equality never holds above in view of the equality cases in theorem a. thus the right - hand inequality in the theorem has been proved . as we noted in introduction
, the function @xmath71 given in belongs to the class @xmath39 therefore , we have @xmath365 where @xmath366 we certainly have @xmath367 which implies @xmath368 here , @xmath369 when @xmath370 or @xmath371 @xmath372 is an extremal value for @xmath78 as we saw before .
however , this is not the case when @xmath94 indeed , we will show that the left derivative @xmath373 is negative for @xmath91 which implies that @xmath374 for a small enough @xmath375 for @xmath376 we have the expression @xmath377 therefore , we compute @xmath378 where @xmath379 we note that @xmath380 is strictly convex in @xmath381 because @xmath382 there . since @xmath383 the inequality @xmath384 holds for @xmath385 hence , @xmath386 for @xmath387 as required .
the above proof showed that @xmath388 it seems difficult to find the exact value of this maximum .
for instance , a numerical computation tells us that the maximum of @xmath389 is attained at @xmath390 and its value is approximately @xmath391 see figure 1 , which was generated by mathematica ver .
10.2 , for the graph of the function @xmath392 the first peak is located slightly left to the bending point @xmath393 it might be meaningful to ask whether equality holds or not in for @xmath394 * acknowledgements . *
the present note grew out of an unpublished paper on successive coefficients of convex functions by professor derek thomas and discussions with him .
the authors would like to express their sincere thanks to him for kind suggestions .
a. z. grinspan , _ improved bounds for the difference of adjacent coefficients of univalent functions ( russian ) _ , questions in the modern theory of functions ( novosibirsk ) , sib .
inst . mat . , 1976 , | in this note , we investigate the supremum and the infimum of the functional @xmath0 for functions , convex and analytic on the unit disk , of the form @xmath1 we also consider the related problem to maximize the functional @xmath2 for convex functions @xmath3 with @xmath4 for a prescribed @xmath5.$ ] |
nonequilibrium transitions from stuck to flowing phases underlie the physics of a wide range of physical phenomena . in a first class of systems
the onset of a stuck or frozen state occurs as a result of intrinsic dynamical constraints , due to interactions or crowding , and is usually referred to as _ jamming _ @xcite .
familiar examples are supercooled liquids that become a glasses upon lowering the temperature , colloidal suspensions that undergo a glass transition due to crowding upon increasing the density or the pressure , foams and granular materials that jam under shear , arrays of dislocations in solids that jam under an applied load . in a second class of systems
the transition to a stuck state is due to external constraints , such as the coupling to quenched disorder ( pinning centers from material defects in vortex lattices , optical traps in colloids , etc . ) , and is denoted as _ pinning _ @xcite .
both classes of systems can be driven in and out of glassy states by tuning not only temperature , density or disorder strength , but also an applied external force .
the external drive may be a shear stress in conventional glasses or simply a uniform applied force in systems with extrinsic quenched disorder , where even a uniform translation of the system relative to the fixed impurities represents a nontrivial perturbation .
vortex lattices in superconductors @xcite and charge density waves ( cdws ) in metals @xcite can be driven in and out of stuck glassy states by a uniform external current or electric field , respectively . as recognized recently in the context of jamming ,
the external drive plays a role not unlike that of temperature in driving the system to explore metastable configurations and should be included as an axis in a complete phase diagram . in this lectures i will focus on zero - temperature depinning transitions of interacting condensed matter systems that spontaneously order in periodic structures and are driven over quenched disorder .
the prototype examples are vortex lattices in type - ii superconductors @xcite and charge density waves in anisotropic metals @xcite .
other examples include wigner crystals of two dimensional electrons in a magnetic field moving under an applied voltage @xcite , lattices of magnetic bubbles moving under an applied magnetic field gradient @xcite , and many others .
in general these systems form a lattice inside a solid matrix , provided by the superconducting or conducting material and are subject to pinning by random impurities .
the statics of such disordered lattices have been studied extensively @xcite .
one crucial feature that distinguishes the problem from that of disordered interfaces is that the pinning force experienced by the periodic structure is itself periodic , although with random amplitude and phase . as a result , although disorder always destroys true long - range translational order and yields glassy phases with many metastable states and diverging energy barriers between these states , the precise nature of the glassy state depends crucially on disorder strength . at weak disorder the system , although glassy , retains topological order ( the resulting phase has been named bragg glass in the context of vortex lattices ) .
topological defects proliferate only above some characteristic disorder strength , where a topologically disordered glass is formed .
the driven dynamics of disordered periodic structures have been studied extensively by modeling the system as an overdamped elastic medium that can be deformed by disorder , but is not allowed to tear , that is by neglecting the possible formation of topological defects due to the competition of elasticity , disorder and drive .
this model , first studied in the context of charge density waves , exhibits a nonequilibrium phase transition from a pinned to a _
unique _ sliding state at a critical value @xmath0 of the driving force .
this nonequilibrium transition displays universal critical behavior as in equilibrium _ continuous _ transitions , with the medium s mean velocity @xmath1 acting as an order parameter @xcite . while the overdamped elastic medium model may seem adequate to describe the dynamics of driven bragg glasses , many experiments and simulations of driven systems have shown clearly that topological defects proliferate in the driven medium even for moderate disorder strengths @xcite .
the dynamics near depinning becomes spatially and temporally inhomogeneous , with coexisting moving and pinned degrees of freedom .
this regime has been referred to as plastic flow and may be associated with memory effects and even hysteresis in the macroscopic response .
the goal of the present lectures is to describe coarse - grained models of driven extended systems that can lead to history - dependent dynamics .
such models can be grouped in two classes . in the first class
the displacement of the driven medium from some undeformed reference configuration remains single - valued , as appropriate for systems without topological defects , but the interactions are modified to incorporate non - elastic restoring forces @xcite .
in the second class of models topological defects are explicitly allowed by removing the constraint of single - valued displacements . here
we will focus on the first class and specifically consider driven periodic media with a linear stress - strain relation , where the stress transfer between displacements of different parts of the medium is nonmonotonic in time and describes viscous - type slip of neighboring degrees of freedom .
a general model of this type that encompasses many of the models discussed in the literature was proposed recently by us @xcite . here slips between neighboring degrees of freedom
are described as viscous force , that allows a moving portion of the medium to overshoot a static configuration before relaxing back to it .
it is shown below that such viscous coupling can be considered an effective way of incorporating the presence of topological defects in the driven medium .
related models have also been used to incorporate the effect of inertia or elastic stress overshoot in crack propagation in solids @xcite .
the precise connection between the two classes of models has been discussed in ref .
@xcite .
in section [ sec : singleparticle ] we review the simplest example of depinning transition , obtained when non - interacting particles are driven through a periodic pinning potential . by contrasting the case of periodic and non - periodic pinning
, we stress that care must be used in the definition of the mean velocity of the system . in section [ sec : extendedmedium ] , we first describe the generic coarse - grained model of a driven elastic medium that exhibits a _ continuous _
depinning transition as a function of the driving force from a static to a _
unique _ sliding state .
next we introduce an anisotropic visco - elastic model as a generic model of a periodic system driven through strong disorder .
the model considers coarse - grained degrees of freedom that can slip relative to each other in the directions transverse to the mean motion , due to the presence of small scale defects ( phase slips , dislocations , grain boundaries ) at their boundaries , but remain elastically coupled in the longitudinal directions .
the slip interactions are modeled as viscous couplings and a detailed physical motivation for this choice is given in section [ subsec : viscous_couplings ] .
most of our current results for these type of models are for the mean - field limit and are presented in section [ sec : mean_field ] .
the studies carried out so far for finite - range interactions suggest that the mean - field theory described here may give the correct topology for the phase diagram , although there will of course be corrections to the critical behavior in finite dimensions @xcite .
finally , we conclude in section [ sec : other_models ] by discussing the relation to other models described in the literature and the connection to experiments .
it is instructive to begin with the problem of a single particle driven through a _ periodic _ pinning potential as the simplest illustration of driven depinning . assuming overdamped dynamics , the equation of motion for the position @xmath2 of the particle is @xmath3 where @xmath4 is a friction coefficient ( in the following we choose our units of time so that @xmath5 ) , @xmath6 is the external drive and @xmath7 , with @xmath8 an integer , is a periodic function of period @xmath9 . for simplicity
we choose a piecewise linear pinning force , corresponding to @xmath10 , for @xmath11 . in this case a periodic solution of eq .
( [ single_per ] ) is obtained immediately in terms of the time @xmath12 needed to traverse a potential well , or period . introducing an arbitrary time @xmath13 such that if @xmath14 , then @xmath15 , the particle position for @xmath16 is @xmath17 where @xmath12 is given by @xmath18 for @xmath19 and diverges for @xmath20 .
in other words if @xmath20 the particle never leaves the initial well , i.e. , it is pinned .
the threshold force for depinning is then @xmath21 . in the sliding state the mean
velocity is defined as the average of the instantaneous velocity @xmath22 over the arbitrary initial time @xmath13 .
this gives @xmath23 this definition naturally identifies the mean velocity of the particle with the inverse of the period .
the logarithmic behavior of @xmath24 near threshold , @xmath25 , is peculiar to a discontinuous pinning force . for an arbitrary pinning force
@xmath26 the period @xmath12 is @xmath27 and can be evaluated analytically for various forces .
for instance , for a sinusoidal pinning force , @xmath28 , one finds @xmath29 , which gives @xmath30 near threshold , a generic behavior for continuous pinning forces
. the main focus of the remainder of this paper will be on the modeling of extended driven systems as collections of interacting degrees of freedom .
it will then be important to distinguish two cases . for extended systems that are periodic , such as charge density waves and vortex lattices , the pinning potential is itself periodic as each degrees of freedom sees the same disorder after advancing one lattice constant . for non - periodic systems , such as interfaces ,
each degree of freedom moves through a random array of defects .
when interactions are neglected , an extended _ periodic _ system moving through a periodic random pinning potential can be modeled as a collection of @xmath31 non - interacting particles , where each particle sees its own periodic pinning potential .
the pinning potentials seen by different particles may differ in height and be randomly shifted relative to each other , as sketched in fig .
the equation of motion for the @xmath32-th particle at position @xmath33 is then @xmath34 where @xmath35 are random phases uniformly distributed in @xmath36 and the pinning strengths @xmath37 are drawn independently from a distribution @xmath38 .
since the displacements @xmath33 are decoupled , they can be indexed by their disorder parameters @xmath39 and @xmath40 instead of their spatial label @xmath32 , i.e. , @xmath41 . the mean velocity of the many - particle system can then be written as an average over the random phases and pinning strengths , @xmath42 where @xmath43 .
the average over the random phase of each degree of freedom is equivalent to the average over the random time shift @xmath13 described for the single - particle case and yields @xmath44 , with @xmath45 the period of each particle given in eq .
( [ period ] ) .
the mean velocity is then @xmath46 where @xmath47 denotes the average over the barrier height distribution . for distributions @xmath48 that
have support at @xmath49 , a system of _ noninteracting _ particles with periodic pinning depins at @xmath50 , as there are always some particles experiencing zero pinning force .
( a ) sketch of noninteracting degrees of freedom driven over a random periodic pinning potential in one dimension .
spatial coordinates have been discretized so that degrees of freedom are labelled by an index @xmath32 . in ( b ) the case where each degree of freedom interacts elastically with its neighbors
this is a discretized one - dimensional realization of the elastic medium model described by eq .
( [ elastic_model ] ) below . ]
a different single - particle problem that has been discussed in the literature is that of a particle moving through a _
random ( non - periodic ) _ array of defects @xcite .
the defects can be described as pinning potential wells centered at random positions and/or with random well heights . to make contact with the periodic case we consider a particle moving through a succession of evenly spaced pinning potential wells of random heights .
the equation of motion is @xmath51 where @xmath52 is the total number of pinning centers and the pinning strengths @xmath53 are drawn independently from a distribution @xmath54 .
choosing again the piecewise - linear pinning force , the time to traverse the @xmath55-th well is simply @xmath56 , with @xmath12 given by eq .
( [ period ] ) .
the mean velocity of the particle is defined as the total distance travelled divided by the total time and is given by @xmath57 in this case , unless the distribution @xmath48 is bounded from above , there is always a finite probability that the particle will encounter a sufficiently deep potential well to get pinned .
therefore for unbounded @xmath48 the particle is always pinned in the thermodynamic limit .
if @xmath48 is bounded from above by a maximum pinning strength @xmath58 , this value also represents the depinning threshold .
finally , the case of many noninteracting particles driven through a random array of defects is equivalent to that of a single particle , as the mean velocity of each particle can be calculated independently .
the mean velocity of the system is then again given by eq .
( [ vmean_nonper ] ) .
we consider a @xmath59-dimensional periodic structure driven along one of its symmetry directions , chosen as the @xmath2 direction .
the continuum equations for such a driven lattice within the elastic approximation were derived by various authors by a rigorous coarse - graining procedure of the microscopic dynamics @xcite .
assuming overdamped microscopic dynamics , the equation for the local deformation @xmath60 of the medium ( in the laboratory frame ) from an undeformed reference state is written by balancing all the forces acting on each portion of the system as @xcite @xmath61 where @xmath62 is the stress tensor due to interactions among neighboring degrees of freedom , @xmath6 is the driving force and @xmath63 is the periodic pinning force .
the periodicity of the pinning force , which contains fourier components at all the reciprocal lattice vectors of the lattice , arises from the coupling to the density of the driven lattice . for conventional short - ranged elasticity
the stress tensor is @xmath64 where @xmath65 and @xmath66 are the compressional and shear moduli of the driven lattice , respectively , and @xmath67 is the strain tensor .
it was shown in ref .
@xcite that deformations of the driven lattice along the direction of the driving force grow without bound due to large transverse shear stresses that generate unbounded strains responsible for dislocation unbinding .
for this reason , we focus here on the dynamics of a scalar field @xmath68 , with @xmath69 , describing deformations of the driven lattice along the direction of mean motion .
the @xmath70-dimensional vector @xmath71 denotes the coordinates transverse to the direction of motion .
assuming @xmath72 , we obtain a scalar model for the driven elastic medium , given by @xmath73 where @xmath74 denotes the @xmath2 component of the pinning force . for simplicity
we also consider a model that only retains the component of the pinning force at the smallest reciprocal lattice vector and choose our units of lengths so that the corresponding period is 1 .
the pinning force is then taken of the form @xmath75 where @xmath76 is a periodic function .
the random pinning strengths @xmath40 are drawn independently at every spatial point from a distribution with zero mean and short - ranged correlations to be prescribed below .
the random phases @xmath39 are spatially uncorrelated and distributed uniformly in @xmath36 . the model of a driven overdamped elastic medium embodied by eq .
( [ elastic_model ] ) has been studied extensively both analytically and numerically @xcite .
it exhibits a depinning transition at a critical value @xmath0 of the applied force from a static to a _
unique _ sliding state @xcite .
the depinning can be described as a continuous equilibrium transition , with the mean velocity @xmath77 playing the role of the order parameter , and universal critical behavior .
the velocity vanishes as @xmath0 is approached from above as @xmath78 .
the critical exponent @xmath79 depends only on the system dimensionality and was found to be @xmath80 using a functional rg expansion in @xmath81 @xcite .
strong disorder can yield topological defects in the driven lattice , making the elastic model inapplicable . in this case
the dynamics becomes inhomogeneous , with coexisting pinned and moving regions @xcite
. the depinning transition may be discontinuous ( first order ) , possibly with macroscopic hysteresis .
several mean - field models of driven extended systems have been proposed to describe this inhomogeneous dynamics . here
we focus on a class of models that retains a single - valued displacement field and a linear stress - strain relation , but assumes that the presence of topological defects can be effectively incorporated at large scales by a non - instantaneous stress transfer that couples to gradients of the local velocity ( rather than displacement ) .
more precisely , we consider an anisotropic model of coarse - grained degrees of freedom that can slip relative to each other in at least one of the directions transverse to the mean motion , due to the presence of small scale defects ( phase slips , dislocations , grain boundaries ) at their boundaries , but remain elastically coupled in the longitudinal directions @xcite .
this model incorporates the anisotropy of the sliding state in the plastic flow region that results either from flow along coupled channels oriented in the direction of the drive ( e.g. , as in the moving smectic phase @xcite ) or in layered materials such as the high-@xmath82 cuprate superconductors .
it also encompasses several of the models discussed in the literature . for generality , consider a @xmath83-dimensional medium composed of degrees of freedom that are coupled elastically in @xmath84 direction and can slip relative to each other in the remaining @xmath85 directions .
the axis @xmath2 along which the driving force is applied is along one of the @xmath84 directions .
the equation of motion for the displacement @xmath86 is given by @xmath87 with @xmath88 the local velocity .
this model will be referred to as the visco - elastic ( ve ) model as it incorporates elastic couplings of strength @xmath89 in @xmath84 directions and viscous couplings of strength controlled by a shear viscosity @xmath90 in the remaining @xmath85 directions . a two - dimensional cartoon of this anisotropic model is shown in fig .
[ fig : cartoon ] . a two - dimensional realization of the anisotropic driven medium described in the text .
spatial coordinates have been discretized in the figure so that degrees of freedom are labelled by indices @xmath91 and @xmath92 , respectively transverse and longitudinal to the direction of the driving force , @xmath6 .
each degree of freedom interacts with its neighbors via elastic couplings in the longitudinal direction and via viscous or similar slip couplings in the transverse direction . ] for @xmath93 ( or @xmath94 ) the ve model reduces to the elastic model ( but with isotropic elasticity ) of eq .
( [ elastic_model ] ) .
conversely , for @xmath95 ( or @xmath96 ) eq .
( [ ve_model ] ) reduces to the purely viscous model studied earlier by us @xcite . for any distribution of pinning strengths with support at @xmath49 ,
the purely viscous model has zero threshold for depinning , but it does exhibit a critical point separating regions of unique and multivalued solutions for the mean velocity . in the ve model ( @xmath97 and @xmath98 ) even when fluid - like shear takes place ,
particle conservation gives a sharp depinning transition in flow along the channels .
furthermore , as shown below , the model has a sharp mean - field tricritical point separating a region of parameters where depinning is continuous , in the universality class of elastic depinning , from one where depinning become discontinuous and hysteretic .
it is important to stress that the ve model still assumes _ overdamped microscopic dynamics_. velocity or viscous couplings can appear generically in the large - scale equations of motion upon coarse - graining the microscopic dynamics of a dissipative medium .
in fact , next we show that viscous couplings indeed represent an effective way of incorporating the local dissipation due to the presence of topological defects .
the goal of this section is to provide some justification to the anisotropic ve model as an effective description of topological defects in a driven lattice . to this purpose
we consider a two dimensional medium and take advantage of the continuum equations developed many years ago by zippelius et al .
@xcite to describe the time - dependent properties of two - dimensional solids near melting .
these authors combined the equations of free dislocation motion with solid hydrodynamics to construct a semimicroscopic dynamical model of a solid with free dislocations .
they further showed that the dynamics of such of a `` heavily dislocated solid '' ( an elastic medium with an equilibrium concentration of free dislocations ) is identical to that of the hexatic phase obtained when a two - dimensional solid melts via the unbinding of dislocations @xcite .
more recently we @xcite reconsidered the dynamical equations for the `` heavily dislocated solid '' of ref .
@xcite and showed that they can be recast in the form of the phenomenological equations of a viscoelastic fluid ( with hexatic order ) introduced many years ago by maxwell @xcite . in the presence of free dislocations
the local stresses in the medium have contributions from both elastic stresses and defect motion . the latter couple again to the the local strains which control the defect dynamics . by eliminating the defect degrees of freedom ,
one obtains a linear , although nonlocal , relation between strain and stress , given by @xcite @xmath99\;,\end{aligned}\ ] ] where @xmath100 and the velocity @xmath101 is defined here in terms of the momentum density @xmath102 as @xmath103 , with @xmath104 the equilibrium mass density of the medium .
also in eq .
( [ sigmave ] ) @xmath105 is the compressional modulus of the liquid and @xmath106 and @xmath107 are the compressional and shear relaxation times , with @xmath108 the dislocation glide and climb mobility , respectively .
of course in the presence of dislocations the displacement @xmath109 is no longer single - valued ( although the strain @xmath110 remains single - valued and continuous ) and @xmath111 due to both the motion of vacancy / interstitial defects and of dislocations .
the phenomenological maxwell model of viscoelasticity is obtained by assuming that @xmath112 even in the presence of dislocations .
then for @xmath113 the viscoelastic stress @xmath114 reduces to the familiar elastic stress tensor given in eq .
( [ stress ] ) , @xmath115 conversely for @xmath116 one obtains @xmath117 which describes stresses in a viscous fluid of shear viscosity @xmath118 and bulk viscosity @xmath119 . the first term on the right hand side of eq .
( [ sigma_fluid ] ) is the pressure and incorporates the fact that even a liquid has a nonzero long - wavelength compressional elasticity , which is associated with density conservation . as we will see below this terms plays a crucial role in controlling the physics of depinning of a viscoelastic medium .
the maxwell viscoelastic fluid has solid - like shear rigidity at high frequency , but flows like a fluid at low frequency . since the relaxation times
@xmath120 and @xmath121 are inversely proportional to the density @xmath122 of free dislocations , the maxwell model behaves as a continuum elastic medium on all time scales when @xmath123 and as a viscous fluid when @xmath124 .
dislocation climb is much slower than dislocation glide ( @xmath125 ) , resulting in @xmath126 .
we therefore assume that the response to compressional deformations is instantaneous on all time scales , but retain a viscoelastic response to shear deformations . letting @xmath127 , we find @xmath128\;.\end{aligned}\ ] ] we now turn to the case of interest here , where topological defects are generated in a an extended medium driven through quenched disorder . in this case
the medium has no low frequency shear modulus , but particle conservation still requires long wavelength elastic restoring forces to _ compressional _ deformations . on the other hand ,
the number of topological defects is not fixed as dislocations are continuously generated and annihilated by the interplay of elasticity , disorder and drive @xcite .
furthermore , unbound dislocations can be pinned by disorder and do not equilibrate with the lattice . in the plastic region near depinning the dynamics remains very inhomogeneous and fluid - like and the pinning of dislocations by quenched disorder is not sufficient to restore the long wavelength shear - stiffness of the medium .
for this reason we propose to describe the effect of topological defects near depinning by replacing elastic _ shear
_ stresses by viscoelastic ones , while retaining elastic _
compressional _ forces .
of course the resulting model that assumes a fixed density of dislocations becomes inapplicable at large driving forces where dislocations heal as the lattice reorders .
for the case of interest here of a scalar model describing only deformations along the direction of motion , the viscoelastic model of a driven disordered medium is @xmath129 with @xmath88 .
this model naturally incorporates the anisotropy and channel - like structure of the driven medium , where _ shear _ deformations due to gradients in the displacement in the directions transverse to the mean motion ( @xmath130 ) are most effective at generating the large stresses responsible for the unbinding of topological defects .
it is instructive to note that due to the exponential form of stress relaxation the integro - differential equation ( [ full_ve ] ) is equivalent to a second order differential equation for the displacement , @xmath131 with @xmath132 an effective friction @xcite . in other words
the effect of a finite density of dislocations in the driven lattice yields `` inertial effects '' on a scale controlled by the time @xmath133 . the purely viscous model obtained from eq .
( [ utwodots ] ) with @xmath134 was analyzed in detail in ref .
@xcite where it was shown that if @xmath120 and @xmath118 are tuned independently , then @xmath120 is a strongly irrelevant parameter in the rg sense .
this allows us to consider a simplified form of the equation for the driven medium obtained from eq .
( [ utwodots ] ) with @xmath135 , but @xmath118 finite , leading to the general anisotropic viscoelastic model introduced in eq .
( [ ve_model ] ) .
the mean - field approximation for the ve model is obtained in the limit of infinite - range elastic and viscous interactions . to set up the mean field theory , it is convenient to discretize space in both the transverse and longitudinal directions , using integer vectors @xmath92 for the @xmath84-dimensional intra - layer index and @xmath91 for the @xmath85-dimensional layer index .
the local displacement along the direction of motion is @xmath136 .
its dynamics is governed by the equation , @xmath137 + f+h_{{\ell}}^{{i}}y(u_{{\ell}}^{{i}}-\gamma_{{\ell}}^{{i}})\;,\ ] ] where the dot denotes a time derivative and @xmath138 ( @xmath139 ) ranges over sites @xmath140 ( @xmath141 ) that are nearest neighbor to @xmath92 ( @xmath142 ) .
the random pinning strengths @xmath143 are chosen independently with probability distribution @xmath144 and the @xmath145 are distributed uniformly and independently in @xmath36 . for a system with @xmath146 sites , _ one _ mean field theory is obtained by assuming that all sites are coupled with uniform strength , both within a given channel and with other channels . each discrete displacement then couples to all others only through the mean velocity , @xmath147 , and the mean displacement , @xmath148 .
we assume that the disorder is isotropic and the system is self averaging and look for solutions moving with stationary velocity : @xmath149 .
since all displacements @xmath109 are coupled , they can now be indexed by their disorder parameters @xmath39 and @xmath40 , rather than the spatial indices @xmath92 and @xmath91 .
the mean - field dynamics is governed by the equation @xmath150 the cases @xmath151 and @xmath98 need to be discussed separately .
when @xmath95 , the mean field equation becomes identical to that of a single particle discussed in section [ sec : singleparticle ] driven by an effective force @xmath152 ( with friction @xmath153 ) . in this case
different degrees of freedom move at different velocities even in the mean field limit .
the mean field velocity is determined by the self - consistency condition @xmath154 , where the average over the random phases is equivalent to the average over the random times shifts @xmath13 given in eq .
( [ vmean_single ] ) . for the case of a piecewise linear pinning force using eq .
( [ vmean_per ] ) we find @xmath155 with @xmath156 given by eq .
( [ period ] ) .
the mean velocity obtained by self - consistent solution of eq .
( [ vmean_viscous ] ) is shown in figs .
[ fig : viscous_fixed_h ] and [ fig : viscous_broad_h ] for two distributions of pinning strengths .
( a ) velocity versus driving force for the purely viscous model ( @xmath95 , @xmath97 ) with a narrow distribution of pinning strength , @xmath157 , for @xmath158 .
there is a finite depinning threshold at @xmath159 . in @xmath160
the depinning and repinning forces @xmath161 and @xmath162 are shown as functions of @xmath90 . ] ( a ) velocity versus driving force for the purely viscous model ( @xmath95 , @xmath97 ) with a broad distribution of pinning strength , @xmath163 for @xmath164 . in this case there are no stable static ( pinned states ) .
the velocity is single valued for @xmath165 and multi - valued for @xmath166 . in this case when @xmath6 is ramped up from zero , the velocity jumps discontinuously at @xmath161 where the system goes from the `` slow - moving '' to the `` fast - moving '' state . here and below `` coexistence '' refers to multistability of the solutions to the equations of motion .
when @xmath6 is then ramped down from within the fast - moving state the jump in @xmath24 occurs at the lower value @xmath162 .
the forces @xmath162 and @xmath167 become equal at the critical point , as shown in frame ( b ) . ] for a narrow distribution , @xmath157 , there is a finite threshold @xmath159 , independent of @xmath90 .
the velocity is multivalued for any finite @xmath90 .
when the force is ramped up adiabatically from the static state the system depins at @xmath168 .
when the force is ramped down from the sliding state , the system repins at the lower value @xmath169 .
the depinning and repinning forces are shown in fig .
[ fig : viscous_fixed_h](b ) .
the region where unique and multivalued velocity solutions coexist extend to @xmath93 . for a broad distribution with support at @xmath49 ,
e.g. , @xmath163 , the threshold for depinning is zero as some of the degrees of freedom always experience zero pinning and start moving as soon as a force is applied .
there is a critical point at @xmath170 . for @xmath166
the analytical solution for @xmath171 is multivalued , as shown in fig .
[ fig : viscous_broad_h ] .
if the force is ramped up adiabatically from zero at a fixed @xmath166 , the system depins discontinuously at @xmath172 , while when the force is ramped down it repins at the lower value @xmath169 , as shown in fig .
[ fig : viscous_broad_h ] .
the viscous model has also been studied in finite dimensions by mapping it onto the nonequilibrium random field ising model ( rfim ) @xcite . in the mapping ,
the local velocities correspond to spin degrees of freedom , the driving force is the applied magnetic field and the mean velocity maps onto the magnetization .
the rfim has a critical point separating a region where the magnetization versus applied field curve displays hysteresis with a discontinuous jump to a region where there is no jump in the hysteresis curve @xcite . in the viscous model
the critical point separates a region where the velocity curve is smooth and continuous from the region where the `` depinning '' ( from `` slow - moving '' to `` fast - moving '' states ) is discontinuous and hysteretic .
the critical point is in the ising universality class , with an upper critical dimension @xmath173 . when @xmath174 , all degrees of freedom are coupled by a spring - like interaction ( the first term on the right hand side of eq .
( [ mft_ve ] ) ) to the mean field @xmath175 and can not lag much behind each other .
this forces all the periods to be the same , independent of @xmath40 , and yields a nonvanishing threshold for depinning . in this case
the mean field velocity is determined by imposing @xmath176 .
it is useful to first review the case where @xmath174 and @xmath93 . in this limit ,
( [ mft_ve ] ) reduces to the mean field theory of a driven elastic medium worked out by fisher and collaborators @xcite .
no moving solution exists above a finite threshold force @xmath0 . for the piecewise linear pinning force
this is given by @xmath177 for @xmath178 there is a _
unique _ moving solution that has a universal dependence on @xmath6 near @xmath0 , where it vanishes as @xmath179 . in mean - field
the critical exponent @xmath79 depends on the shape of the pinning force : @xmath180 for the discontinuous piecewise linear force and @xmath181 for generic smooth forces . using a functional rg expansion in @xmath81 ,
narayan and fisher @xcite showed that the discontinuous force captures a crucial intrinsic discontinuity of the large scale , low - frequency dynamics , giving the general result @xmath80 , in reasonable agreement with numerical simulations in two and three dimensions @xcite . for simplicity and to reflect the `` jerkiness '' of the motion in finite - dimensional systems at low velocities , we use piecewise linear pinning below .
when @xmath182 the nature of the depinning differs qualitatively from the @xmath93 case , in that hysteresis in the dynamics can take place .
again , no self - consistent moving solution exists for @xmath183 , with @xmath0 independent of @xmath90 . above threshold ,
both unique and multi - valued moving solutions exist , depending on the values of the parameters : @xmath90 , @xmath89 , and the shape of the disorder distribution , @xmath48 . to obtain the mean field solution in the sliding state
, we examine the motion during one period @xmath184 during which the displacement advances by 1 .
( [ mft_ve ] ) is is easily solved for @xmath185 and @xmath186 , with the result , @xmath187 where @xmath188 . at long times , regardless of the initial condition , @xmath189 approaches a periodic function of period @xmath184 with jumps in its time derivative at times @xmath190 , with @xmath8 an integer .
the constant @xmath191 is determined by requiring that if @xmath192 , then @xmath193 .
writing @xmath194 , it is easy to see that for an arbitrary value of @xmath39 , the solution @xmath195 will have the form @xmath196 .
the mean velocity is then obtained from @xmath197 .
averaging @xmath195 over @xmath39 is equivalent to averaging @xmath195 for a fixed @xmath39 over a time period , @xmath12 , with the result , @xmath198 finally , averaging over @xmath40 and using the consistency condition , we obtain @xmath199 + \big\langle\frac{h^2}{k(k+h)}\frac{1}{e^{(k+h)/(1+\eta)\overline{v}}-1 } \big\rangle_h\;,\end{aligned}\ ] ] with @xmath0 the threshold force given in eq . ( [ ft ] ) and @xmath200 given by @xmath201 ( a ) velocity versus driving force for the ve model with @xmath202 and a broad distribution of pinning strength , @xmath163 .
the velocity is continuous and single - valued for @xmath165 and becomes multivalued for @xmath166 .
the dashed line on he curve for @xmath203 indicates the value @xmath162 where the system repins when the drive is ramped down from the sliding state .
frame @xmath160 shows the depinning and repinning forces @xmath161 and @xmath162 as functions of @xmath90 .
the tricritical point at @xmath170 separates continuous from hysteretic depinning .
pinned and sliding states coexist in the region @xmath204 . ] as in the purely elastic case ( @xmath93 ) only static solutions exist for @xmath183 .
for @xmath178 there is a _
unique _ sliding solution , provided @xmath205 , with mean velocity near threshold given by @xmath206 giving @xmath180 , as in the purely elastic case . the critical line @xmath207 separating unique from multivalued sliding solutions is determined by @xmath208 , @xmath209 the velocity - force curves and a phase diagram are shown in fig .
[ fig : ve_versus_eta ] for @xmath163 .
there is a _ tricritical point _ at @xmath210 .
in contrast to the purely viscous model with @xmath95 , for finite long - time elasticity ( @xmath211 ) the behavior is _ independent of the shape of the pinning force distribution , _
@xmath48 . for @xmath165 , a continuous depinning transition at @xmath0
separates a pinned state from a sliding state with _
unique _ velocity . in finite dimensions ,
this transition is likely to remain in the same universality class as the depinning of an elastic medium ( @xmath93 ) . in our mean - field example
, the linear response diverges at @xmath212 , @xmath213 . for @xmath166
there is hysteresis with coexistence of stuck and sliding states .
numerical simulations of the ve model in two dimensions ( @xmath214 ) indicate a strong crossover ( possibly a tricritical point ) at a critical value of @xmath212 from continuous to hysteretic depinning @xcite .
although it is always difficult to establish conclusively on the basis of numerics that hysteresis survives in the limit of infinite systems , the size of the hysteresis loop evaluated numerically does appear to saturate to a finite value at large system sizes , indicating that the mf approximation may indeed capture the correct finite - dimensional physics .
other models of driven systems with inertial - type couplings have been proposed in the literature .
it is useful to discuss in some detail their relationship to the viscoelastic model considered here . in the context of charge density waves , littlewood @xcite and levy and collaborators
@xcite modified the fukuyama - lee - rice model @xcite that describes the phase of the cdw electrons as an overdamped elastic manifold driven through quenched disorder by incorporating the coupling of the cdw electrons to normal carriers .
this was realized via a global coupling in the equation of motion for the phase to the mean velocity of the cdw , not unlike what obtained by a mean - field approximation of our viscous coupling .
the model was argued to account for the switching and non - switching behavior observed in experiments .
schwarz and fisher @xcite recently considered a model of crack propagation in heterogeneous solids that incorporates stress overshoot , that is the fact that a moving segment of the crack can sometimes overshoot one or more potential static configurations before settling in a new one , inducing motion of neighboring segments .
these effects may arise from elastic waves that can carry stress from one region to another of the driven medium .
stress overshoots , just like topological defects in a driven disordered lattice , have an effect similar to that of local inertia and were modeled by fisher and schwarz by adding couplings to gradients in the local crack velocity in the equation of motion for a driven elastic crack .
these authors considered an automaton model where time is discrete .
it is straightforward to define an automaton version of our ve model , where both the displacement @xmath215 and time are discrete , as shown in ref .
it is then apparent that the automaton version of the viscoeleastic model given in ref .
@xcite is identical in its dynamics to the model of crack propagation with stress overshoot studied by schwarz and fisher , provided the strength @xmath200 of the stress overshoot is identified with the combination @xmath216 .
the two models differ in the type of pinning considered as the random force used in by schwarz and fisher is not periodic .
we find , however , that the two models have identical mean - field behavior , with a mean - field tricritical point separating continuous from hysteretic dynamical transitions .
the connection between the viscoelastic and the stress - overshoot model is important because it stresses that distinct physical mechanisms ( inertia , nonlocal stress propagation , unbound topological defects ) at play in different physical systems can be described generically by a coarse - grained model that includes a coupling to local velocities of the driven manifold .
finally , in a very recent paper , maimon and schwarz suggested that out of equilibrium a new type of generic hysteresis is possible even when the phase transition remains continuous @xcite .
driven models with both elastic and dissipative velocity couplings may therefore belong to a novel universality class that exhibits features of both first and second order equilibrium phase transitions .
they clearly deserve further study .
we now turn briefly to simulations and experiments . for comparison with experiments
it is useful to point out that the tricritical point of the viscoelastic model can also be obtained by tuning the applied force and the disorder strength , rather than the applied force and the viscosity .
since the phase diagram does not depend on the form of the disorder distribution , @xmath48 , we choose for convenience a sharp distribution , @xmath217 .
the phase diagram in the @xmath218 plane is shown in fig . [
fig : vephased_f_vs_h ] . for weak disorder
the depinning is continuous , while for strong disorder it becomes hysteretic , with a region of coexistence of pinned and moving degrees of freedom .
the _ tricritical point _ is at @xmath219 , with @xmath220 .
mean - field solution of the ve model with a piecewise parabolic pinning potential , @xmath217 and @xmath221 . left frame : phase diagram .
right frame : velocity versus drive for @xmath222 ( blue ) , @xmath223 ( red ) and @xmath224 ( black ) . also shown for @xmath224
are the discontinuous hysteretic jumps of the velocity obtained when @xmath6 is ramped up and down adiabatically . ]
simulations of two - dimensional driven vortex lattices clearly show a crossover as a function of disorder strength from an elastic regime to a regime where the dynamics near depinning is spatially inhomogeneous and plastic , with coexistence of pinned and moving degrees of freedom @xcite .
in fact a bimodal distribution of local velocity was identified in ref .
@xcite as the signature of plastic depinning .
this local plasticity does not , however , lead to hysteresis in the macroscopic dc response in two dimension : the mean velocity remains continuous and single - valued , although it acquires a characteristic concave - up ward form near depinning that can not be described by the exponent @xmath225 predicted by elastic models in all dimensions .
hysteresis is , however , observed in simulations of three - dimensional layered vortex arrays where the couplings across layers are weaker than the in - layer ones @xcite . in this case
the phase diagram is qualitatively similar to that obtained for the viscoelastic model .
recent experiments in @xmath226 have argued that memory effects originally attributed in this system to `` plasticity of the driven vortex lattice @xciteare actually due edge contamination effects @xcite . in the experiments
a metastable disordered vortex phase is injected in a stable ordered bulk vortex lattice .
memory effects may then arise in the macroscopic dynamics during the annealing of the injected disordered phase .
edge contamination does not , however , explain the plasticity seen in simulations , where periodic boundary conditions are used @xcite
. a possible scenario may be that while in the experiments the vortex lattice in the bulk is always in the ordered phase , in the simulations the vortex lattice in the bulk of the sample may be strongly disordered even in the absence of drive .
such a disordered vortex lattice would then naturally respond plastically to an external drive .
finally , it is worth mentioning one experimental situation where hysteresis of the type obtained in our model is indeed observed in the macroscopic response .
this occurs in the context of charge density waves , driven by both a dc and an ac field . in this case
the dc response exhibits mode - locking steps .
the ' ' depinning " from such mode - locked steps was found to be hysteretic @xcite .
several colleagues and students have contributed to various aspects of this work : alan middleton , bety rodriguez - milla , karl saunders , and jen schwarz .
i am also grateful to jan meinke for help with some of the figures .
the work was supported by the national science foundation via grants dmr-0305407 and dmr-0219292 .
equation ( [ eqmotion ] ) should include a convective derivative that arises from the transformation to the laboratory frame @xcite .
this is negligible near a continuous depinning transition where the mean velocity is very small and it drops out in the mean field limit considered here .
it will therefore be neglected , although it does play a crucial role on the dynamics well into the sliding state @xcite . | invited lecture presented at the xix sitges conference on _ jamming , yielding , and irreversible deformations in condensed matter _ , sitges , barcelona , spain , june 14 - 18 , 2004 . |
we initially consider the full time dependent hamiltonian @xmath97 where @xmath98 and @xmath99 , which is the usual coupling times a phase factor due to the presence of a vector potential .
the vectors @xmath100 are the unit cell vectors @xmath101 , @xmath102 , @xmath103 , and the distance to the nearest neighbors @xmath104 are given by @xmath105 , @xmath106 , and @xmath107 .
importantly , the interaction terms does not couple to the ac field , which can be understood as follows : the interaction terms depend on the momentum difference between the electrons , and then for a spatially homogeneous fields this is time independent . alternatively , as the coupling with the electric field can be also writen as @xmath108 , where the electric field @xmath109 , the field will only couple to those terms which does not commute with the position operator .
therefore , on site interactions are proportional to the possition operator and will not become time dependent .
+ following ref .
@xcite the floquet operator can be calculated in sambe space , leading to @xmath110 where the hoppings are @xmath111 with @xmath112 , being @xmath113 and @xmath83 the fourier components of the fourier expansion of each floquet state .
the arguments of the bessel functions and phase factors are @xmath114 @xmath115 to obtain the effective hamiltonian let us consider a stroboscopic evolution operator over a period @xmath116 @xmath117 next , due to its time periodicity we consider the fourier decomposition of the hamiltonian @xmath118 where we have considered just the first harmonic contribution , and @xmath119 .
expanding the matrix exponential in taylor series we obtain : @xmath120 then , in terms of the previous expansion , the stroboscopic evolution operator is given by @xmath121,\end{aligned}\ ] ] where in the last line we have applied the time ordering operator @xmath122 . using eq .
we can perform the time integrals , and reordering terms compare with the expansion of the effective time evolution operator @xmath123-\left[h_{0}^{0},h_{0}^{1}\right]+\left[h_{0}^{-1},h_{0}^{1}\right]\right)\right\ } \\ & \simeq & 1-ih_{\text{eff}}t-\frac{1}{2}h_{\text{eff}}^{2}t^{2}+\ldots\end{aligned}\ ] ] the effective hamiltonian is finally given by @xmath124-\left[h_{0}^{0},h_{0}^{1}\right]+\left[h_{0}^{-1},h_{0}^{1}\right]\right),\label{eq : heff}\ ] ] which is eq . of the main text .
explicitly , the single particle hamiltonian in @xmath125 domain @xmath126 is given by @xmath127\\ & + j_{0}\left(a_{-}\right)\left[j_{1}\left(a_{-}\right)\cos\left(\psi_{-}\right)-j_{1}\left(a_{+}\right)\cos\left(\sqrt{3}k_{x}\right)\cos\left(\psi_{+}\right)+j_{1}\left(a_{y}\right)\cos\left(\phi\right)\cos\left(\frac{\sqrt{3}}{2}k_{x}-\frac{3}{2}k_{y}\right)\right]\\ & + j_{0}\left(a_{y}\right)\left[j_{1}\left(a_{-}\right)\cos\left(\psi_{-}\right)\cos\left(\frac{\sqrt{3}}{2}k_{x}-\frac{3}{2}k_{y}\right)-j_{1}\left(a_{+}\right)\cos\left(\psi_{+}\right)\cos\left(\frac{\sqrt{3}}{2}k_{x}+\frac{3}{2}k_{y}\right)+j_{1}\left(a_{y}\right)\cos\left(\phi\right)\right]\\ & -j_{1}\left(a_{y}\right)\left[j_{1}\left(a_{-}\right)\sin\left(\frac{\sqrt{3}}{2}k_{x}-\frac{3}{2}k_{y}\right)\sin\left(\psi_{-}+\phi\right)+j_{1}\left(a_{+}\right)\sin\left(\frac{\sqrt{3}}{2}k_{x}+\frac{3}{2}k_{y}\right)\sin\left(\phi-\psi_{+}\right)\right]\\ & + j_{1}\left(a_{-}\right)j_{1}\left(a_{+}\right)\sin\left(\sqrt{3}k_{x}\right)\sin\left(\psi_{-}+\psi_{+}\right)\}.\end{aligned}\ ] ] eq .
is the single particle hamiltonian considered all throughout this work . from it it is possible to extract a topological phase diagram for the chern number @xmath128 as a function of the external parameters of the electric field shown in fig .
[ fig : chern number phase diagram ] .
the band structure for the point @xmath4 , and @xmath5 focus of this work is shown in fig .
[ fig : schematic ] .
a ) many - body spectrum , b ) flux insertion , and c ) tower - of - states for the same parameters for a @xmath129 lattice that evidence the presence of a ferromagnetically ordered ffci state .
d ) we show also the local density profile @xmath130 for the case of a @xmath131 lattice defined in for one representative many - body groundstate .
the small density variation is symptomatic of the fci state.,title="fig : " ] a ) many - body spectrum , b ) flux insertion , and c ) tower - of - states for the same parameters for a @xmath129 lattice that evidence the presence of a ferromagnetically ordered ffci state .
d ) we show also the local density profile @xmath130 for the case of a @xmath131 lattice defined in for one representative many - body groundstate .
the small density variation is symptomatic of the fci state.,title="fig : " ]
in this section , additional numerical results mentioned in the main text are presented as supporting evidence for the ferromagnetic floquet fractional chern insulator ( ffci ) . first , we focus on a different lattice size , @xmath129 with @xmath70 particles , to compare with @xmath132 with @xmath71 particles discussed in the main text . the spectrum , flux insertion and tower - of - states for this lattice size is presented in fig . [ fig : band - structure ] ( a)-(c ) .
the figures show the same qualitative features as for the @xmath132 lattice case discussed in the main text .
+ firstly , as shown in fig .
[ fig : band - structure ] ( a ) the groundstate is three - fold degenerate for each @xmath79 sector .
the groundstates appear at the total momenta @xmath86 predicted by the counting rule of ref . @xcite . under an adiabatic flux insertion in the @xmath88 direction , the states interchange signalling the topological degeneracy and charge fractionalization
[ see fig .
[ fig : band - structure ] ( b ) ] .
furthermore , all groundstates have the maximum spin @xmath82 which is evidence for the ferromagnetic nature of the ffci state .
this picture is further supported by the tower - of - states of fig .
[ fig : band - structure ] ( c ) that shows that the higher values of @xmath91 have the lowest energy , a finite size signature of the thermodynamic spontaneous symmetry breaking of the su(2 ) symmetry , as discussed in the main text . to ascertain the topological nature of the floquet fractional chern insulator state
it is important to rule out the possibility of the appearance of a charge density wave ( cdw ) order , since both states can present the same momentum counting @xcite . a way to distinguish between both phases
was put forward in ref .
@xcite by calculating the real space density profile , a procedure that we follow here . to this extent ,
we consider the local density operator defined by the next step is to obtain a set of @xmath113 maps of the local fermion density in the ground - state manifold .
labelling @xmath137 the set of orthonormal eigenvectors of @xmath138 at some arbitrarily chosen site @xmath139 we evaluate the @xmath113 real functions @xmath140 the functions @xmath130 are density maps that show the variation of the local fermion density in position space . featureless or flat density maps would be strong evidence in favour an fci .
therefore , finding strongly patterned density maps is a signature of a ground - state manifold that supports a cdw . + a typical charge density map for the three - fold ground state manifold of the ffci state is shown in fig .
[ fig : band - structure](d ) for the particular point @xmath4 @xmath5 discussed in the main text .
the density variation in position space is small , of about @xmath141 which is indication of a fci state .
we attribute the small deviation to the lattice anisotropy , already observed in ref .
@xcite .
the ffci state described in the main text appears for circular polarization @xmath37 . although we have focused on @xmath4 , a similar behaviour is observed in the vicinity of such point . in this section
we show that the appearance of the ffci state with this particular polarization ( circular ) and for the honeycomb lattice is not a fine tuned unique situation .
+ we observe that this state also appears for elliptical polarization . in fig . [
fig : chersupp ] ( b ) we show the phase diagram for the case where @xmath142 on the @xmath143 phase space .
as in the circular polarization , there are regions in the phase diagram where the effective bands have a non trivial chern number @xmath64 . for concreteness
we consider the @xmath144 sector and @xmath142 with @xmath145 for a @xmath146 lattice with @xmath71 particles taking @xmath147 , without single particle dispersion ( unlike in the main text ) .
we note that adding dispersion only constrains the region of the phase diagram where the ffci phase appears but does not eliminate it from the phase diagram . as in the main text , we compute the many - body spectrum via exact - diagonalization . the results are shown in fig .
[ fig : hc_lattice_elliptic ] where ( a ) shows the many body energy spectrum ( b ) the flux insertion and ( c ) a representative local density profile .
( a ) the energy spectrum for the driven honeycomb lattice model with @xmath142 and @xmath145 for a @xmath146 lattice with @xmath71 particles against the total momentum @xmath86 .
the three fci states appear at momenta predicted by the counting rule of ref .
@xcite ( b ) density profile @xmath130 defined in for one representative many - body groundstate .
( c ) spectral evolution of the energy spectrum of ( a ) in the sector upon inserting a flux @xmath88 into the system .
the zoom shows the interchange of the three fractional chern insulator groundstates under one flux period and thus their topological degeneracy and charge fractionalization . ]
all three figures point to the realization of the fci state since the ground state fall at the total momentum sectors @xmath86 expected from the counting rule of ref .
@xcite , they mix and interchange with each other under adiabatic flux insertion and the density profile is relatively uniform , as discussed in the previous section and in the main text .
we nevertheless note that the relative density change is close to @xmath148 , approximately twice the value for the circular polarization .
we have also checked that all these three features also appear for different elliptical polarization such as @xmath149 .
+ the above results strongly indicate that the ffci state described in the main text does not depend strongly on light polarization but rather on that the chern number is finite .
we have also studied whether this still holds for different lattices and find that indeed the honeycomb lattice is not a fine - tuned case . to support this claim
we hereby provide numerical evidence of the appearance of the same @xmath0 ffci state for the @xmath150flux lattice at half - filling .
because of the historical relevance of the model in condensed matter these results are interesting on their own right and therefore here we will sum up the most relevant results and further detail will be presented elsewhere .
+ following the procedure outlined in the first section of this supplementary material , we have calculated the effective hamiltonian @xmath151 corresponding to the @xmath150flux model @xcite with the conventions of ref .
the non - interacting phase diagram for this case , analogous to fig .
[ fig : chern number phase diagram ] in the main text for the honeycomb lattice is shown in fig . [
fig : chersupp ] ( a ) .
as in the honeycomb case , there are regions in the parameter space @xmath94 where the chern number of the effective bands is non - trivial and takes the values @xmath53 .
+ again , we consider only the fully polarized spin sector @xmath144 with a @xmath146 lattice with @xmath71 particles and include band dispersion and @xmath147 . choosing the particular point @xmath152 and @xmath153 corresponding to the non - trivial case with @xmath64 we find a three fold degenerate ground state as shown in the spectrum of fig .
[ fig : sq_lattice ] ( a ) together with the flux insertion in ( b ) and a representative local density profile ( c ) .
all three figures point to the realization of the fci state : ( i ) the momentum sectors for the groundstate manifold coincide with those predicted by the counting rule of ref .
@xcite , ( ii ) the flux insertion reveals that the groundstates evolve independently and interchange upon one flux insertion and ( iii ) the uniformity of the density profile discards the cdw state as the many body ground state .
therefore , we conclude that the floquet fractional chern insulator state is also realized in the @xmath150flux lattice . furthermore , the symmetries of the square lattice make the sampling of the berry curvature bz particularly uniform favouring the use of the simple formula of ref .
@xcite . with this
we have checked that the hall conductivity of this system is closely quantized to @xmath154 as expected for the fci state .
we found that for this lattice the fci state is defined in an even wider region of the phase diagram as compared to the honeycomb case details of which will be reported elsewhere .
floquet fractional chern insulator state for the @xmath150flux model with @xmath155,@xmath153 , @xmath144 and @xmath146 , @xmath71 .
( a ) the energy spectrum for the @xmath150flux model against the total momentum @xmath86 .
the three fci states appear at momenta predicted by the counting rule of ref .
@xcite ( b ) spectral evolution of the energy spectrum of ( a ) in the sector upon inserting a flux @xmath156 into the system .
the evolution of the three fractional chern insulator ground states signals their topological degeneracy and charge fractionalization .
c ) density profile @xmath130 defined in for one representative many - body groundstate . ] | fractional chern insulators are theoretically predicted states of electronic matter with emergent topological order .
they exhibit the same universal properties as the fractional quantum hall effect , but dispose of the need to apply a strong magnetic field . however , despite intense theoretical work , an experimental realization for these exotic states of matter is still lacking . here
we show that doped graphene turns into a fractional chern insulator , when irradiated with high - intensity circularly polarized light .
we derive the effective steady state band structure of light - driven graphene using floquet theory and subsequently study the interacting system with exact numerical diagonalization .
the fractional chern insulator state equivalent to the @xmath0 laughlin state appears at @xmath1 total filling of the honeycomb lattice ( @xmath2 filling of the upper band ) .
the state also features spontaneous ferromagnetism and is thus an example of the spontaneous breaking of a continuous symmetry along with a topological phase transition .
sketch of the proposed experiment .
a graphene - flake is irradiated with light of frequency @xmath3 while a gate voltage is applied via a backgate ( yellow ) to change the band filling . in the high - frequency regime ,
the incident light changes the single - particle band structure into an effective floquet band structure that acquired a gap at the dirac points ( shown are the floquet bands for the electric field configuration @xmath4 , @xmath5 , and @xmath6 ) . ]
fractional chern insulators ( fcis ) @xcite have been discovered numerically in lattice models of two - dimensional electronic systems .
they emerge as the ground state of repulsively interacting electrons that partially fill bloch bands with a nontrivial topological attribute , namely a nonvanishing chern number @xcite .
the time - reversal symmetry ( trs ) breaking electronic hopping integrals on the lattice , that are responsible for the band topology , take the role played by the strong magnetic field in the factional quantum hall effect .
fcis show that fractional quantum hall states can appear more generically than previously assumed , and do not rely on the specific energetical or analytical properties of landau levels . for a system to support a fci ground state , the energetics have to satisfy specific conditions .
for example , if the topological band is spectrally flat @xcite or the energy scale of the repulsive interaction exceeds the energy scales of the band @xcite , fcis are favored .
this is why the experimental discovery of fcis is still a formidable experimental challenge , despite the recent experimental realization of its `` noninteracting '' parent band structure , the chern insulator or anomalous quantum hall effect @xcite .
needed are systems with a large amount of tunability , to meet both the topological and energetical requirements .
ultracold atomic gases in optical lattices @xcite , artificial graphene @xcite , photonic crystals @xcite , and light - driven solid state systems @xcite are such tunable platforms , all of which have been shown to potentially host topological band structures . in this study , we shall focus on light - driven graphene [ see fig . [ fig : schematic ] ] , for which several works have proved that circularly polarized light allows to open a gap in the dirac cones , leading to a topologically nontrivial state , characterized by chiral edge states @xcite .
the resulting periodically driven steady state is theoretically described using floquet theory and is thus called a floquet chern insulator .
the key ingredients for its emergence are ( i ) the critical nature of the dirac electrons in graphene and ( ii ) the time - reversal symmetry breaking provided by the non - linearly polarized light @xcite . here , we show by means of numerical exact diagonalization that graphene at @xmath1 total filling of the @xmath7-bands , when irradiated with high - intensity circularly polarized light , realizes a ferromagnetic fci steady state , which we call a floquet fractional chern insulator ( ffci ) .
the ffci state is characterized by a three - fold topological groundstate degeneracy and a contribution to the hall conductivity of @xmath8 .
furthermore we prove that the full su(2 ) spin - rotation symmetry of the model hamiltonian ( the light - field does not couple to the spin and the spin - orbit coupling is negligible ) is spontaneously broken for a ferromagnetic steady state , and gapless magnon excitations emerge , coexisting with the ffci groundstate .
we emphasize that we do not rely on a mean - field approximation to obtain this result .
[ [ floquet - approach - for - ac - driven - graphene ] ] floquet approach for ac driven graphene + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we model irradiated monolayer graphene by considering spinful electrons that populate a honeycomb lattice @xmath9 and interact repulsively via their on - site ( @xmath10 ) and nearest neighbor ( @xmath11 ) electronic densities @xmath12 , for @xmath13 and @xmath14 , @xmath15 for the single particle hamiltonian @xmath16 , we adopt the convention used in @xcite .
the time dependence is induced via the electromagnetic vector potential @xmath17 of the external light field , where @xmath3 is the frequency of the light , @xmath18 is the phase difference , @xmath19 , @xmath20 is the electron charge , @xmath21 its mass , @xmath22 the lattice spacing , and @xmath23 the @xmath24th component of the electric field . only the single particle hamiltonian @xmath16 is affected by this time dependence , and not the interaction hamiltonian . the latter remains as in the undriven case since it is expressed in terms of the density operator .
floquet theory provides a powerful formalism to study periodically driven systems .
it allows to easily obtain effective time evolution operators , especially in the regime where the driving frequency is the dominant energy scale @xcite .
the single particle hamiltonian , which enters the time dependent schrdinger equation , can be obtained following ref . @xcite and the supplementary material , and has the fourier decomposition @xmath25 where @xmath26 labels the fourier component in frequency space , and the hopping integrals are given by @xmath27 , @xmath28 , and @xmath29 . here , @xmath30 is the nearest - neighbor hopping integral for electrons on the honeycomb lattice and @xmath31 denotes the bessel functions of the first kind .
the arguments of the bessel functions contain the explicit electric field configuration @xmath32 , and the phase factors are given by @xmath33 .
finally , we define @xmath34 , @xmath35 , and @xmath36 as the unit cell vectors of the honeycomb lattice . topological phase diagram of periodically driven graphene at high frequency regime ( @xmath6 ) for circularly polarized light .
the phase difference @xmath18 and the field amplitude @xmath37 allow to tune the chern number of the lower spin - degenerate bands of the effective floquet - bloch hamiltonian between @xmath38 and @xmath39 .
the black dot indicates the parameter values for which exact diagonalization calculations are presented in figs .
[ fig : spectrum ] and [ fig : flux tower ] . ] in this work we focus on the high frequency regime ( @xmath40 ) , where the system dynamics can be accurately described by a static effective hamiltonian , which can be expanded order by order in @xmath41 .
to zeroth order in @xmath41 , the hopping integrals are renormalized by zeroth order bessel functions only , with no breaking of trs .
to first order in @xmath41 , corrections to the effective hamiltonian must be considered due to hybridization with the nearest floquet bands .
importantly , for non linear field polarization , trs is broken by these corrections , a prerequisite to realize a floquet chern insulator .
the effective hamiltonian for long - time dynamics is then defined by including higher fourier components of the hamiltonian , leading to the @xmath42 floquet - bloch hamiltonian ( for details see the supplementary material ) : @xmath43\\ & -\left[h_{0,{\bf k}}^{0},h_{0,{\bf k}}^{1}\right]+\left[h_{0,{\bf k}}^{-1},h_{0,{\bf k}}^{1}\right]\bigr ) .
\end{split } \label{eq : floquet - bloch}\ ] ] then , the effective single particle hamiltonian expressed using the second - quantized operators @xmath44 , that create an electron with momentum @xmath45 and spin @xmath46 in sublattice @xmath47 and @xmath48 , respectively , reads @xmath49 the hamiltonian has two pairs of spin - degenerate bands that touch in two dirac points for @xmath50 .
the correction to first order in @xmath41 in the @xmath42 floquet - bloch hamiltonian @xmath51 is proportional to the third pauli matrix @xmath52 . due to the trs breaking it can thus potentially open a haldane - type gap @xcite in the spectrum , so that the resulting spin - degenerate bands can acquire a chern number @xmath53 for each spin species .
the phase diagram for different externally tunable parameters is shown in fig . [
fig : chern number phase diagram ] .
it is worth emphasizing that in the high frequency limit relevant for this work , the hamiltonian is a time independent effective hamiltonian that governs stroboscopic evolution .
therefore , it allows to fill the bands as in the case of time independent systems .
[ [ exact - diagonalization - results ] ] exact diagonalization results + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we are now going to show that the ground state of the floquet hamiltonian @xmath54 , and with this the steady state of the driven hamiltonian , can be tuned into a ffci by controlling ( i ) the filling of the system with electrons and ( ii ) the amplitude and phase of the light field . the emergence of a ffci depends crucially on the ratios between the energy scales of the single - particle band gap @xmath55 , the single particle band width @xmath56 and the repulsive electron - electron interactions , where mathematical band flatness is not always the optimal choice @xcite . in graphene ,
the interaction parameters are given by @xmath57 and @xmath58 @xcite .
choosing incident light with amplitude @xmath4 and the phase shift @xmath5 , for example , results in an @xmath59 with @xmath60 .
we study the system above half - filling , so that the upper spin - degenerate band of @xmath59 is partially filled .
given the size of the single particle gap @xmath55 , it is reasonable to approximate the states in the lower band to be occupied with probability one , even in the interacting many - body ground state @xcite .
we thus ignore those single - particle states and project the degrees of freedom of the system to the upper spin - degenerate band of @xmath59 by means of the projector @xmath61 and to study the hamiltonian @xmath62 previous studies have shown that fci ground states equivalent to the @xmath63 laughlin state of the fractional quantum hall effect in landau levels emerge quite generically if a flat band with chern number @xmath64 is populated with spineless fermions at a filling @xmath65 @xcite .
in contrast , we are considering a _ dispersionful _ band with chern number @xmath64 that is partially filled with _
spinful _
electrons in such a way that the hamiltonian is su(2 ) spin - rotation symmetric . in anticipation of a spontaneous breaking of the su(2 ) symmetry by the many - body ground state
, we therefore study the system at @xmath66 filling to obtain an ffci that is equivalent to a @xmath67 laughlin state .
we have performed numerical exact diagonalization of the hamiltonian on lattices with @xmath68 and @xmath69 unit cells with @xmath70 and @xmath71 electrons , respectively , with periodic boundary conditions in place .
good quantum numbers of the many - body states are the total spin @xmath72 , the total spin-@xmath73 component @xmath74 and the center of mass momentum @xmath75 $ ] .
all results and conclusions presented below extend to both lattice sizes and we focus on the @xmath69 lattice here , delegating the consistency check with @xmath68 to the supplementary material
. + low - lying portion of the energy spectrum of hamiltonian on a @xmath69 with @xmath71 particles .
encircled is the ground state manifold with 3-fold topological ( quasi- ) degeneracy of the fractional chern insulator and @xmath76 fold degeneracy as a precursor of spontaneous symmetry breaking toward a ferromagnetic phase in the thermodynamic limit .
the good quantum numbers total momentum @xmath77 , total spin @xmath78 and total spin in @xmath73 direction @xmath79 are indicated .
] we observe that the low - energy states have both exact and approximate degeneracies : in each of the @xmath80 sectors of @xmath79 , three low - lying states are found which are not exactly , but approximately , degenerate .
each of these three states has an exactly degenerate partner in every other @xmath79 sector ( see fig .
[ fig : spectrum ] ) .
the total ground state degeneracy that we anticipate in the thermodynamic limit is thus @xmath81 .
as all ground states have the maximum spin @xmath82 , we interpret the exact @xmath80-fold degeneracy as a finite - size precursor of a spontaneous breaking of the su(2 ) symmetry towards a ferromagnetic ground state in the thermodynamic limit . both the tower - of - states structure @xcite in the spectrum as a function of @xmath78 [ fig .
[ fig : flux tower ] ( b ) ] and the exactness of the degeneracy support this conclusion ( the order parameter @xmath79 of ferromagnetism commutes with the hamiltonian thus rendering the ground state degeneracy exact already for finite systems ) .
in contrast , the 3-fold approximate degeneracy in each @xmath79 sector is of topological origin .
it is the @xmath83-fold topological ground state degeneracy of a @xmath63 laughlin state on the torus in the case @xmath84 .
the non - local wilson - loop order parameter does not commute with the hamiltonian , rendering the degeneracy approximate in the finite system . further supporting arguments that the three low - lying states in each @xmath79 sector are indeed topologically ordered ffci states are : ( i ) by inserting a flux in the torus ( which is equivalent to changing the boundary conditions from periodic to twisted @xcite ) , the three states permute and return to their original order after three flux quanta [ see fig . [
fig : flux tower ] ( a ) ] .
this evidences charge fractionalization with quasiparticles of charge @xmath85 and indicates that the topological groundstates survive in the thermodynamic limit .
( ii ) the three states occur at the momentum sectors @xmath86 that are predicted by the counting rule of ref
. @xcite that is based on a @xmath0 laughlin state .
( iii ) any superposition of the three ground states has a nearly constant charge density in position space , which excludes that these states would form a charge - density wave in the thermodynamic limit ( see supplementary material ) .
a ) spectral evolution of the energy spectrum of fig .
[ fig : spectrum ] in the sector with @xmath87 upon inserting a flux @xmath88 into the system , which is synonymous to twisting the boundary conditions in @xmath89 direction with a complex phase @xmath90 .
the three fractional chern insulator ground states evolve independently of the rest of the spectrum , and trade places , which signals their topological degeneracy and the charge fractionalization .
b ) the energy spectrum of fig .
[ fig : spectrum ] plotted against the total spin @xmath91 reveals the tower of states which evidencing the ferromagnetic nature of the ground state . ] from the presented results we conclude that the steady state of graphene at @xmath1 total filling of the @xmath7 bands ( or @xmath2 filling of the band above the dirac cone ) is a ferromagnetic ffci .
this distinct driven topological state has gapped charged excitations but supports gapless spin excitations , namely ferromagnetic magnons .
the state features both conventional order and gapless topological order .
experimentally , its signature is a fractional contribution of @xmath92 ( in addition to the integer contribution of the lower band ) defined for the driven system @xcite . to certify the robustness of the ffci state we have also tested its stability under changing some of the conditions discussed above .
firstly , we have investigated its fate upon changing to different light polarizations ( using @xmath93 as two examples ) .
the ffci is still the groundstate of the system as long the chern number of the non interacting band is @xmath53 .
secondly , we have investigated its appearance on a different lattice system .
in particular , we find that for spinless electrons on the @xmath7-flux square lattice with nearest neighbor repulsive interaction the ground state is also the ffci whenever @xmath94 are such that the non - interacting bands have @xmath53 ( see supplementary material ) .
both of these results evidence the ubiquitousness and robustness of the ffci state .
[ [ experimental - realization ] ] experimental realization : + + + + + + + + + + + + + + + + + + + + + + + + + the practical realization of this novel state in graphene possess two experimental challenges .
the first is to reach incident field amplitudes @xmath95 .
although it is in principle possible to reach such a regime , today experiments have only explored amplitudes of one order of magnitude lower @xcite .
however , given the robustness and tuneability of the effect it is conceivable that lower frequencies and field amplitudes can in fact be sufficient to access the ffci state in the thermodynamic limit .
the second experimental issue is reaching the necessary filling factor of graphene s band structure .
in particular , the electron density at @xmath2 filling of the upper band is of the order of @xmath96 , still below the van hove singularity in graphene .
even higher values up to the van hove singularity have already been reached by chemical doping @xcite and there are promising other routes by using for example polymer electrolytes @xcite .
finally , other 2d materials with dirac electrons such as silicene can potentially be used to tune the band structure parameters and host similar phases .
+ to summarize , we have found that doped graphene , irradiated with polarized light , undergoes an interaction driven spontaneous topological phase transition to a floquet fractional chern insulator ( ffci ) state that features both topological and spontaneous ferromagnetic order .
the robustness and tuneability of the ffci state , appearing for different parameters , polarizations and lattices , evidences that this phase can be experimentally discovered in graphene .
more generally , our work opens up a promising route to achieve tuneable realizations of elusive interacting fermionic and bosonic phases by periodically driving interacting systems .
[ [ acknowledgements ] ] acknowledgements : + + + + + + + + + + + + + + + + + we acknowledge e. berg , d. podolsky , g. platero , m.a.h .
vozmediano , for useful discussions .
financial support from pib2010bz-00512 ( a.g.g . ) , jae program , mat 2011 - 24331 and itn , grant 234970 ( eu ) ( a.g - l . ) and swiss national science foundation ( t.n . ) is greatly acknowledged .
32 natexlab#1#1bibnamefont # 1#1bibfnamefont # 1#1citenamefont # 1#1url # 1`#1`urlprefix[2]#2 [ 2][]#2 , , , , * * , ( ) . , , , , * * , ( ) . , * * , ( ) . , * * , ( ) . , , ( ) , . ,
* * , ( ) . , , , * * , ( ) .
, , , , * * , ( ) . , , , * * , ( ) . , , , , , , , , , , ,
* * , ( ) . , , , , , , , ( ) .
, , , , , * * , ( ) . , , , , , , , , , * * , ( ) . , , ,
* * , ( ) .
, , , , , * * , ( ) . , , , ( ) .
, , , , * * , ( ) . , * * , ( ) . , , , ( ) . , , , * * , ( )
.
, * * , ( ) . ,
, , , * * , ( ) . , , , , , ,
* * , ( ) . , , , _
_ , vol .
( , ) . , , , * * , ( ) . , * * , ( ) . , , ,
* * , ( ) .
, , , , , , , * * , ( ) . , , , , , * * , ( ) . , , , , * * , ( ) . , , , * * , ( ) . , , , ,
* * , ( ) . |
propagating relativistic shock waves are the leading candidates for the origin of the emission from gamma - ray bursts ( grbs ) and their afterglows ( e.g. , piran 1999 , 2000 ; mszros 2000 ) as well as of the spectral flares and apparent superluminal motions exhibited by the blazar class of active galactic nuclei ( agns ; e.g. , ulrich , maraschi , & urry 1997 ) and their `` miniquasar '' galactic counterparts ( e.g. , mirabel & rodrguez 1999 ; @xcite ) .
the shocks are thought to be associated with relativistic , jet - like outflows that are driven from the vicinity of a compact object : a massive black hole in blazars , a stellar - mass black hole in miniquasars , and a stellar - mass black hole or a rapidly rotating neutron star in grbs . in all cases except grb afterglows ,
the shocks are believed to be _ internal _ , possibly resulting from the collision of `` shells '' ( the products of nonsteady ejection episodes ) inside the jet ( e.g. , rees 1978 ; @xcite ; @xcite ) . in the case of grb afterglows ,
the most common interpretation is in terms of shocks that form at the interface between the relativistically outflowing material and the surrounding medium ( with the bulk of the observed emission arising in the `` forward '' shock that propagates into the ambient medium , whereas the `` reverse '' shock that is driven into the ejecta gives rise to an early `` optical flash '' and associated `` radio flare '' ) .
the radiation in all three types of sources is inferred to be nonthermal , with the dominant emission mechanisms most commonly invoked being synchrotron and inverse compton . in particular , both the @xmath0-ray burst emission and the radio through x - ray afterglow radiation in grbs , as well as the radio through x - ray emission in blazars and the radio through
optical radiation in microquasars , has been attributed to synchrotron radiation ( or , in certain models of grbs , a closely related mechanism ; e.g. , @xcite , @xcite , @xcite ) . in the case of blazars ,
the measurement of high ( @xmath1 ) optical linear polarization in some objects directly supports the synchrotron interpretation ( e.g. , angel & stockman 1980 ) . in the case of grbs
there are already two confirmed detections of optical polarization ( covino et al . 1999 ; wijers et al .
1999 ; rol et al .
2000 ) , which , despite the relatively low measured values ( @xmath2 ) , appear to be consistent with a synchrotron origin . in both blazars and microquasars the case for synchrotron radiation is supported also by measurements of radio polarization .
synchrotron radiation requires the presence of charged particles ( typically electrons or electron - positron pairs ) with relativistic random velocities that move in a magnetic field .
the magnitude of the energy densities of the random relativistic electron motions and of the magnetic field are parametrized in terms of their ratios to the total thermal energy density in the emission region by @xmath3 and @xmath4 , respectively . in most cases ,
the values of these two parameters can not be accurately determined from the observational data . however , in the case of blazars there are indications that the emitting gas is in approximate equipartition between the particle and magnetic energy densities ( i.e. , @xmath5 e.g. , readhead 1994 ; bower & backer 1998 ) .
this conclusion is also consistent with models of the nonthermal flares in microquasars ( e.g. , kaiser et al . 2000 ) . in the case of grbs ,
if the observed @xmath0-ray emission indeed represents synchrotron radiation , then considerations of radiative efficiency ( based on estimates of source energetics ) imply that @xmath3 , and likely also @xmath4 , are not much smaller than 1 ( e.g. , @xcite ; piran 1999 ) . in the case of grb afterglows , there have been a few instances in which data were available for deriving more direct estimates of these parameters .
in particular , theoretical fitting of the broadband spectrum of the afterglow of grb 970508 has yielded @xmath6 and @xmath7 (; @xcite ; ) .
the inference that @xmath3 and @xmath4 are not much smaller than 1 in grbs and at least some afterglows has important implications to the nature of the emitting regions in these sources .
a comparatively large value of @xmath3 implies a high efficiency of shock acceleration of relativistic electrons in an electron - proton plasma or else a high electron - positron ( @xmath8 ) pair content in the postshock region ( reflecting the composition of the preshock gas or arising from copious pair production in the shock ) .
a comparatively large value of @xmath4 implies either a highly magnetized preshock medium or else an efficient mechanism of field amplification in ( or behind ) the shock .
if @xmath9 and the radiative cooling time is much shorter than the dynamical time ( which , for synchrotron radiation , would require also @xmath4 not to be too small ) , then the shock will be _ radiative _ : it will contain a radiative cooling layer , where most of the internal energy is radiated away , that will directly influence its dynamics .
nonrelativistic , thermally emitting shocks of this type are prevalent in the interstellar medium ( e.g. , ) , but their properties in relativistic , nonthermal sources have not yet received much attention in the literature .
a radiative shock wave generally consists of a `` shock transition , '' where the bulk of the kinetic energy of the preshock gas is dissipated , and a `` radiative zone , '' where most of the dissipated energy is radiated away .
the energy dissipation scale is typically much smaller than all other relevant length scales , so the shock transition can be treated as an infinitesimal front at the upstream end of the shock wave .
the radiative zone , on the other hand , has a finite width , corresponding to the cooling length over which the postshock gas radiates away a significant fraction of the dissipated energy . after passing through the shock transition ,
the gas is usually compressed by a factor of a few ( for newtonian or mildly relativistic shocks ) , and in the absence of significant radiation this is the density enhancement that also characterizes the far - downstream ( or postshock ) region .
however , it is a well - established result in the theory of interstellar shocks that the asymptotic compression can be far larger if the shock is radiative .
this is a consequence of the fact that the total ( thermal plus magnetic ) pressure behind the shock transition remains roughly constant , so that , as long as magnetic effects are not important , the particle density @xmath10 increases inversely with decreasing temperature @xmath11 .
a related important result for radiative interstellar shocks is that , if the gas has a `` frozen in '' magnetic field @xmath12 with a significant component ( @xmath13 ) perpendicular to the shock normal , then magnetic stresses become progressively more important as @xmath11 decreases behind the shock , a reflection of the fact that the magnetic pressure , associated with @xmath13 , scales as @xmath14 .
if the preshock magnetic field is sufficiently strong , then the magnetic pressure in the cooling layer will reach equipartition with the thermal pressure at some point within the radiative zone . beyond this point
the gas will undergo little additional compression , and the cooling will proceed at nearly constant density rather than at constant thermal pressure ( e.g. , hollenbach & mckee 1979 ) . since the cooling time is proportional to @xmath15 ( a consequence of the fact that the cooling function @xmath16 , the energy radiated per unit volume and per unit time , scales as @xmath14 for a thermally emitting gas ) , the inhibition of further compression by the magnetic stress results in a lower cooling efficiency in this part of the radiative zone .
one can anticipate that the basic properties of radiative hydromagnetic shocks in the interstellar medium ( ism ) would carry over to the relativistic , nonthermal regime .
in particular , one can expect a significant enhancement of the compression as well as of the downstream magnetic field in a radiative shock .
the two cases are , however , expected to differ in detail .
in particular , the cooling function of a synchrotron - emitting shocked gas satisfies @xmath17 ( assuming a predominantly transverse magnetic field behind the shock ) , where @xmath18 is the number density of electrons ( and positrons ) . the explicit density dependence of @xmath16 is thus stronger than in a thermally emitting gas , which suggests that the dynamical influence of the magnetic field on the structure of the shock would be even more pronounced in the nonthermal case .
this could have important implications to astrophysical sources in which such shocks occur .
for one thing , it may provide a mechanism for amplifying the magnetic field up to equipartition strengths in cases where it is initially comparatively weak .
furthermore , as the processes of field amplification and synchrotron emission are mutually reinforcing in the radiative zone , this mechanism may naturally lead to synchrotron emission regions that are characterized by a high value of @xmath4 .
there is even the possibility , if inverse - compton cooling also plays a role , that the bulk of the emitted synchrotron radiation would come from regions where the magnetic field is already at ( or near ) equipartition . in this paper
we examine the aforementioned effects in a quantitative manner by studying the properties of relativistic , hydromagnetic , nonthermally emitting shocks . in [ shocktran ] we describe the framework of our analysis and calculate the shock jump conditions .
the structure of the radiative zone is considered in [ radzone ] . in [ b1 ]
we discuss the circumstances under which radiative shocks of this type are likely to occur in astrophysical sources , focusing on the case of grbs and their afterglows .
our conclusions are presented in
[ conclusions ] .
we analyze the steady - state propagation of relativistic , planar , radiative shocks in a medium that contains a large - scale , ordered magnetic field .
as we noted in 1 , radiative shocks can be divided into a very thin `` shock transition '' and a `` radiative zone '' of finite thickness .
we assume , for simplicity , that all the energy dissipation occurs in the shock transition and all the cooling in the radiative zone .
the assumption of a stationary flow throughout the shock is justified by the fact that , by the definition of radiative shocks , the radiative cooling time @xmath19 of the gas that passes the shock transition is much shorter than the shock dynamical time @xmath20 .
the assumption of a plane - parallel flow is easily justified for the shock transition , and we now argue that it is also appropriate for the radiative zone .
the global shape of the shock is generally not planar .
for example , if the shock is triggered by a nearly isotropic energy release at the source , as in supernova remnants and certain grb scenarios , then it may assume an approximately spherical shape .
the instantaneous distance of the shock front from the central source , @xmath21 , is then given by the radius . if the geometry is not exactly spherical , then one needs to distinguish between the distance @xmath21 and the local radius of curvature @xmath22 .
typically , @xmath23 and both are much larger than the width of the radiative zone .
for example , in a relativistic blast wave of lorentz factor @xmath24 , most of the energy and swept - up matter are concentrated within a thin shell of width @xmath25 in the observer frame , or @xmath26 in the shock frame ( e.g. , @xcite ) .
the width of the radiative zone is @xmath27 in the observer frame , or @xmath28 in the shock frame ( where @xmath29 is the local radius of curvature in the shock frame ) . in any reasonable scenario , @xmath30 , and
even for @xmath31 , the width of the radiative zone would still be smaller than @xmath32 by a factor @xmath33 . given also that @xmath22 is not expected to change considerably on time scales shorter than the dynamical time , it is thus well justified to assume that the shock is planar when calculating the properties of the radiative zone .
we distinguish between the shock rest frame , in which we measure the time @xmath34 , the distance behind the shock front @xmath35 , and the fluid speed @xmath36 , and the fluid rest frame ( the frame comoving with the local mean speed of the flow ) , in which we measure the thermodynamic quantities ( the particle number density @xmath10 , the rest - mass density @xmath37 , the internal energy density @xmath38 , the thermal pressure @xmath39 , and the enthalpy density @xmath40 ) as well as the magnetic field amplitude @xmath12 .
we simplify the treatment by assuming that the field is _ purely transverse _
, @xmath41 , i.e. , it only has a component that is perpendicular to the shock velocity and parallel to the shock front .
this is justified by the fact that , upon traversing the shock , @xmath13 is amplified by a factor equal to the downstream - to - upstream density ratio ( which can be quite large in a relativistic shock ) whereas the parallel component remains unchanged , so that the postshock magnetic field is typically almost completely transverse .
we denote quantities in the unshocked upstream region ( @xmath42 ) by the subscript 1 , whereas quantities immediately behind the shock transition are denoted by the subscript 2 .
quantities without numerical subscripts refer to the radiative zone . the shock is assumed to propagate in an electrically neutral plasma consisting of protons , electrons , and possibly also positrons .
we define the parameter @xmath43 where @xmath44 , @xmath45 and @xmath46 are the number densities of protons , electrons , and positrons , respectively .
we assume that the number of protons is unchanged by the passage through the shock transition , but we allow the number of @xmath8 pairs to possibly increase in the shock transition ( although not during the subsequent transit through the radiative zone ) .
thus we have @xmath47 where @xmath48 is the combined number density of electrons and positrons , @xmath49 is the proper speed ( in units of the speed of light @xmath50 ) , and the parameter @xmath51 accounts for the possible production of pairs in the shock .
the constraint @xmath52 implies that @xmath53 from equations ( [ chi ] ) and ( [ n ] ) we infer @xmath54 with this parameterization it is straightforward to take the following limits : ( i ) no pair production at the shock corresponds to @xmath55 ; ( ii ) no protons corresponds to @xmath56 ; ( iii ) no pairs upstream corresponds to @xmath57 . the pressure and energy density of the magnetic field are given by @xmath58 , and hence the total pressure and enthalpy density are @xmath59 and @xmath60 , respectively . in this paper
we focus on the case in which the fluid is `` cold '' upstream and `` hot '' downstream .
the former assumption means that the thermal pressure and internal energy density of the upstream fluid can be neglected ( @xmath61 ) , so that @xmath62 where @xmath63 and @xmath64 are the proton and electron masses , respectively .
the latter assumption means that the shock lorentz factor is sufficiently large for the random motions of the particles of the shocked fluid to be highly relativistic , so that , everywhere behind the shock transition , @xmath65 where @xmath66 , @xmath67 , @xmath68 , and @xmath69 . in the absence of a detailed shock model ,
it is necessary to specify how the postshock internal energy is divided between the proton and electron / positron components .
we do this by parameterizing the ratio of the pressure of the electrons and positrons to that of the total thermal pressure just behind the shock transition , @xmath70 in view of the assumed postshock equation of state ( eq . [ [ w2 ] ] ) , @xmath71 is just the value of the parameter @xmath3 ( see 1 ) immediately behind the shock transition ( @xmath72 ) . under the assumption of ideal mhd
, the magnetic flux is frozen into the fluid and remains constant in the shock frame .. ] for a purely transverse field , this implies @xmath73 the momentum and energy conservation relations across the shock transition take the form @xmath74 since our main interest is in relativistic shock waves , we henceforth set @xmath75 and @xmath76 and neglect the second term on the left - hand side of equation ( [ momentum_cons ] ) .
substitution of equations ( [ w1 ] ) and ( [ w2 ] ) into equations ( [ momentum_cons ] ) and ( [ energy_cons ] ) then yields , after some algebra , an equation for @xmath77 , @xmath78 and an equation for @xmath79 , @xmath80 where @xmath81 is the ratio of the magnetic pressure to the thermal pressure just behind the shock transition , and @xmath82 is the square of the alfvn proper speed ( so @xmath83 ) . after the gas enters the radiative zone it cools and gets compressed , resulting in an increase in the magnetic - to - thermal pressure ratio .
in fact , as we show in [ results ] , @xmath84 is a monotonically increasing function of @xmath35 .
the solution of equation ( [ eq_beta_2 ] ) is @xmath85 in the limit @xmath86 the magnetic effects near the shock transition are weak , and one recovers from equations ( [ beta_2 ] ) and ( [ p_2 ] ) the expressions for a nonmagnetic , ultrarelativistic ( @xmath87 ) shock , @xmath88 and @xmath89 . as @xmath90 is increased , @xmath77 becomes larger while @xmath79 becomes smaller .
the dependence of @xmath77 , @xmath91 , and @xmath92 on @xmath90 is shown in figure [ beta_u_delta_2 ] .
note that , for a given shock speed , @xmath90 has an upper bound , namely @xmath93 .
this follows from the fact that , in order for a shock wave to exist , the relativistic alfvn mach number , given by @xmath94 ( e.g. , @xcite ) , must exceed 1 .
we take the radiation mechanism behind the shock to be synchrotron emission and inverse - compton scattering ( ic hereafter ) of seed photons by the relativistic electron - positron component
. the seed photons may consist of internally emitted synchrotron photons , giving rise to synchrotron self - compton radiation ( ssc ) , or of an externally produced population of soft photons , in which case the ic emission is referred to as external - radiation compton ( erc ) .
although we treat the postshock protons as a relativistic fluid , we neglect their ( much weaker ) nonthermal emission.[(1-\chi_2)/(1+\chi_2)](m_p / m_e)^4 $ ] , which is typically @xmath95 . for a discussion of possible observable consequences in grb sources of nonthermal emission by protons see , e.g. , zhang & mszros ( 2001 ) and references therein . ]
furthermore , we neglect any possible transfer of energy between the proton and electron / positron components in the radiative zone . under these assumptions ,
the protons evolve adiabatically , and the fraction of the energy dissipated in the shock transition that can be radiated away is determined uniquely by the parameter @xmath71 defined in equation ( [ p_ep2 ] ) .
the radiating particles are assumed to have a monoenergetic energy distribution that can be characterized by a random ( or `` thermal '' ) lorentz factor @xmath96 ( the same for electrons and positrons ) that is a function of the distance @xmath35 behind the shock .
this approximation to the local energy distribution is appropriate for electrons and positrons that have undergone significant cooling .
we note in this connection that observations and particle acceleration models of relativistic shock sources suggest that the energy distribution immediately behind the shock transition has the form of a power law . in this case
the radiative zone is expected to consist of a thin layer just behind the shock transition where most of the electrons and positrons have not yet cooled significantly ( and are thus not well approximated by a monoenergetic distribution ) , followed by a much wider layer where substantial cooling has already occurred and where the monoenergetic representation is adequate ( see @xcite ) .
although the region where the monoenergetic approximation does not apply is relevant to the spectral modeling of the source , it can be safely neglected in a calculation of the cooling - zone dynamics .
the synchrotron emissivity ( energy per unit volume per unit time measured in the rest frame of the fluid ) from a population of relativistic electrons and positrons of total number density @xmath18 that have an isotropic velocity distribution is given by @xmath97 ( e.g. , ) , where @xmath98 is the thomson cross section . to obtain the total emissivity
, one must add the ic component .
the ic power of a single electron is @xmath99 , where @xmath100 is the energy density of photons , assuming an isotropic photon distribution .
this is also the average ic power of an electron for an isotropic distribution of electron velocities , even when the photon distribution is not isotropic .
therefore , the total emissivity is @xmath101 where @xmath102 and @xmath103 are , respectively , the energy densities of the synchrotron and the external photons in the local frame .
we assume that the thomson optical depth is much smaller than 1 , a situation that often applies in astrophysical relativistic - jet sources .
we further assume that the emitting region is optically thin to synchrotron self - absorption near the characteristic synchrotron frequency , which is typical in grbs and their afterglows . under these conditions ,
most of the radiated energy can freely escape the system .
is increased by opacity effects and the shock may not even be radiative . in any case , if the self - absorption frequency exceeds the characteristic frequency , the shock will become spectrally indistinguishable from a nonradiative shock ( e.g. , granot et al . 2000 ) . ] the average fractional energy gain , in a single electron scattering , of seed photons that propagate through the radiative zone is measured by the compton @xmath104 parameter . in the thomson limit ( @xmath105 , where @xmath106 is the incident photon frequency in the local frame , i.e. the fluid rest frame ) ,
the average fractional energy gain of a seed photon moving at an angle @xmath107 ( measured in the shock frame ) with respect to the @xmath35 direction is give by @xmath108 where @xmath109 is the differential optical depth for compton scattering and @xmath110 is the differential of the distance traveled by the photon in the local frame .
one can obtain @xmath104 by averaging the expression ( [ y_mu ] ) over the angular distribution of the seed photons in the shock frame . in the case of an infinite planar shock
, the term @xmath111 that appears in equation ( [ y_mu ] ) will cause the integral to diverge logarithmically for any angular distribution that does not vanish at @xmath112 .
however , under more realistic circumstances , one may expect the derived value of @xmath104 to remain finite .
we approximate the integral in equation ( [ y_mu ] ) by setting @xmath113 and then multiplying the value of the integrand just behind the shock transition by the characteristic size @xmath114 of the radiative zone ( see eq .
[ [ x_c ] ] below ) .
this leads to @xmath115 where we have introduced a numerical factor @xmath116 ( expected to be of order unity ) that incorporates the various uncertainties involved in our approximation . for @xmath117
only single scatterings are important , whereas for @xmath118 multiple scatterings can significantly affect the energy and spectrum of the scattered photons . however , a typical seed photon with initial energy @xmath119 is upscattered on the average to an energy @xmath120 in a single scattering .
usually , the typical electron lorentz factor @xmath96 is sufficiently high for the thomson limit to no longer apply even in the second scattering ( @xmath121 ) .
this causes the energy gain in each successive scattering to be significantly reduced on account of the klein - nishina decrease in the cross section and of electron recoil .
we therefore neglect multiple scatterings in our analysis .
the energy density @xmath102 of the synchrotron photons may be calculated in a manner analogous to that of the compton @xmath104 parameter and is subject to similar uncertainties .
we treat it in the same manner as above by introducing a new parameter of order unity , @xmath122 : @xmath123 where @xmath124 is the length of the trajectory of a photon in the @xmath35 direction ( @xmath125 ) within the radiative zone , as measured in the comoving frame .
we also consider an external photon distribution that is isotropic and has an energy density @xmath126 in the frame of the preshock fluid . in the local rest frame of the shocked fluid
, the energy density of the external photons is given by @xmath127 ( e.g. , ) , where @xmath128 is the relative lorentz factor between the shocked and unshocked fluids .
the cooling length @xmath114 is the distance behind the shock transition over which the electrons / positrons lose the bulk of their internal energy ; it is thus the characteristic size of the radiative zone .
since a significant fraction of the internal energy must be radiated away before the hydrodynamic quantities start to deviate significantly from their values just behind the shock transition , a substantial portion of the radiation would be emitted under conditions that prevail at the upstream end of the radiative zone .
one can therefore evaluate @xmath114 by using the parameter values at @xmath129 , @xmath130 where @xmath131 and @xmath132 , is also approximately equal to @xmath133 , as both energy densities increase by a factor @xmath134 across the shock transition ( see eqs .
[ [ b ] ] and [ [ e_ext ] ] ) . ] and where we have used equations ( [ lambda ] ) and ( [ e_syn ] ) . from equations ( [ x_c ] ) and ( [ y ] )
we obtain an equation for @xmath104 : @xmath135 which yields @xmath136 in the first two limits the external photons are not important , and we obtain the familiar expressions for emission dominated by ssc and synchrotron , respectively ( @xcite ) .
the physical meaning of each of these limits can be understood by comparing the relative magnitudes of the different components of the energy density , keeping in mind that @xmath137 , @xmath138 , and @xmath139 . in the first limit @xmath140 and @xmath141 , implying that @xmath142 , which according to equation ( [ lambda ] ) means that most of the radiated power goes into ssc .
one also obtains @xmath143 , which is consistent with this interpretation , since , in a steady - state radiative shock , the incoming flux of internal energy ( @xmath144 ) equals the outgoing radiation flux , which in this limit is dominated by ssc and is therefore approximately equal to @xmath145 . in the second limit @xmath146 and @xmath147 , implying that @xmath148 , which means that synchrotron radiation is dominant over ic . in this case
we also have @xmath149 , which implies that all the internal energy goes into synchrotron radiation . in the third limit @xmath150 and @xmath151 , implying @xmath152 , which means that most of the emitted power goes into ic - scattered external photons ( erc ) .
this conclusion is consistent with the fact that @xmath153 in this case , which means that all the internal energy is going into erc .
note that the value of @xmath104 ceases to be pertinent in this limit as it no longer gives the ratio of the ic and synchrotron emissivities ( in particular , @xmath104 could be @xmath154 even as ic remains dominant ) .
the photon energy densities @xmath103 and @xmath155 are approximately constant throughout the radiative zone , whereas , as we demonstrate in [ results ] , the magnetic energy density @xmath156 may be considerably amplified if @xmath157 .
thus , even if synchrotron emission is not dominant immediately behind the shock transition ( as in limits 1 and 3 above ) , it may become so further downstream .
however , by the time this happens , only a small fraction of the initial internal energy would typically be left in the electron / positron component .
therefore , the total emission would still be dominated by ic ( either ssc , as in the first limit , or erc , as in the third limit ) .
nevertheless , given that the characteristic ic frequency is usually much higher than that of the synchrotron emission ( by a factor @xmath158 in the case of ssc ) , it is likely even in this case that only the synchrotron component is detected when the observation frequency is not too high .
the steady - state momentum and energy conservation equations in the shock frame can be written as @xmath159 the terms on the right - hand sides of these equations represent the contribution of the radiation field to the momentum and energy fluxes , respectively .
their forms are appropriate when the emitted or scattered radiation exerts no net force on the fluid ; in particular , they apply when the radiation field is isotropic in the fluid frame .
if the magnetic field is ordered , as we have assumed , then the synchrotron emission of any given particle is not isotropic in the fluid rest frame ; however , so long as the electron / positron rest - frame velocity distribution is isotropic , the total synchrotron radiation will carry no net momentum flux in this frame and will , therefore , exert no net force on the fluid . the resulting synchrotron photon energy density ( eq . [ [ e_syn ] ] ) , when scattered by the isotropic electron / positron distribution , gives rise to an ssc emissivity that is isotropic in the fluid frame .
the same applies to an erc component that arises from the scattering ( by electrons / positrons that are isotropic in the local rest frame ) of an isotropic ( in the rest frame of the central source ) external radiation field ( where now the photon energy density that appears in the isotropic - emission expression , in the local rest frame , is given by eq .
[ [ e_ext ] ] ; see ) .. ] from equations ( [ momentum_evol ] ) and ( [ energy_evol ] ) one obtains an equation for the entropy : @xmath160 the magnetic field evolves according to equation ( [ b ] ) and the protons evolve adiabatically , @xmath161 so the terms that involve them in equation ( [ entropy ] ) cancel out .
one is left with @xmath162 eliminating the radiative terms from equations ( [ momentum_evol ] ) and ( [ energy_evol ] ) results in @xmath163 or , in terms of the dimensionless pressure @xmath164 , @xmath165 equation ( [ dptd_du ] ) can be integrated to give @xmath166\over ( 1+u^2)^2}\ .\ ] ] using equations ( [ p_tilde ] ) , ( [ protons ] ) , and ( [ p_ep2 ] ) , we deduce @xmath167 @xmath168\over \eta(1+u^2)^2 } -{1-\eta\over\eta}\left({u_2\over u}\right)^{4/3}\ , \ ] ] where @xmath169 and @xmath170 .
we already have explicit expressions for @xmath171 , and @xmath172 as functions of @xmath173 . to obtain their values as a function of the distance behind the shock ,
we need to know @xmath174 . introducing the dimensionless variable @xmath175 and using equations ( [ p_2 ] ) , ( [ lambda ] ) , ( [ x_c ] ) , ( [ entropy_e ] ) , and ( [ dp_du ] )
, we get @xmath176\over 4(1 - 2u^2 ) \left[(1-\eta)u_2^{4/3}u^{2/3}+\eta u^2\tilde{p}(u)_e\right ] + 6\delta_2 u_2 ^ 2}\ , \ ] ] where the function @xmath177 represents the ratio of the local ic ( including ssc and erc ) and synchrotron emissivities , and @xmath178 is the value of @xmath179 just behind the shock transition .
note that the only dependence of the function @xmath180 on the compton @xmath104 parameter ( see eq . [ [ y ] ] ) is through the combination @xmath181 , which , in turn , is given ( in terms of the parameters @xmath182 and @xmath183 ) by equation ( [ y3 ] ) .
as both @xmath91 and @xmath92 are functions of @xmath90 ( see eqs .
[ [ delta_2 ] ] and [ [ beta_2 ] ] ) , @xmath182 is determined by the parameters @xmath122 , @xmath90 , and @xmath71 . for simplicity , we set @xmath184 in the rest of this paper
. therefore , @xmath180 and the right - hand side of equation ( [ du_dxi ] ) are specified by the parameters @xmath71 , @xmath90 , and @xmath183 , and @xmath185 can be obtained by solving equation ( [ du_dxi ] ) numerically , using the boundary condition @xmath186 . before proceeding to present the solutions of the above set of equations ,
we digress briefly to check on the validity of the ideal - mhd approximation that underlies our analysis . under the assumption of ideal mhd ,
the electric field vanishes in the rest frame of the fluid .
the magnetic field in the shock frame is then given by @xmath187 , where we have chosen the @xmath104 axis to point along the direction of the magnetic field .
this implies a current density @xmath188 since in our case @xmath189 is perpendicular to the direction of the velocity , it has the same magnitude in the shock frame . in order for the ideal - mhd approximation to be valid
, the current density in the fluid rest frame can not exceed @xmath190 , where @xmath191 is the total number density of charged particles ( e.g. , ) .
furthermore , so long as @xmath192 , the effect of anomalous resistivity associated with current - driven plasma instabilities is expected to be small ( e.g. , @xcite ) .
the condition @xmath193 translates into @xmath194 in the radiative zone @xmath195 , so equation ( [ check_1 ] ) may be written as @xmath196 note that this inequality can be rewritten in terms of the postshock electron larmor radius @xmath197 as @xmath198 .
this shows that , if ideal mhd holds in the radiative zone , then the condition for the fluid approximation to be valid there ( essentially @xmath199 ) will also be satisfied .
we now check whether the inequality ( [ check_2 ] ) holds under the most unfavorable choice of parameters .
the latter is obtained by taking @xmath200 and @xmath201 , which implies @xmath202 , @xmath86 , @xmath203 , and @xmath204 , resulting in @xmath205 the minimal value of @xmath206 is @xmath207 , which is obtained for @xmath57 .
this leads to @xmath208 where here and below numerical subscripts ( such as the subscript 0 in @xmath209 ) refer to powers of 10 in c.g.s units . as an illustration we apply this condition to grbs , considering both the prompt high - energy emission ( resulting from internal shocks within the flow ) and the lower - energy afterglow ( arising from the interaction of the flow with the surrounding medium ) .
we begin with internal shocks .
we use the conventional parameterization @xmath210 ( where we identify @xmath211 ) , and write ( e.g. , piran 1999 ) @xmath212 , @xmath213 , and @xmath214 , where @xmath215 is the total energy deposited in the flow , @xmath0 is the typical lorentz factor of the flow with respect to the central object ( and the observer ) , @xmath216 is the relative lorentz factor between the shocked and the unshocked fluids , @xmath217 is the duration of the burst , and @xmath218 is a measure of the variability of the burst , with @xmath219 being the smallest time scale for significant variation during the burst .
thus we obtain @xmath220^{3/2}}\approx { 3.6\times 10 ^ 3 t_{\rm dur,1}^{3/2}\gamma_2 ^ 3\zeta_{-2}\over(1+y_2 ) \epsilon_{b2}^{3/2}e_{52}^{1/2}}\ , \ ] ] where we used @xmath221 on the right - hand side ( consistent with the expectation that @xmath216 is of the order of 1 ) .
turning now to the afterglow stage of grbs , we evaluate the inequality ( [ check_3 ] ) by using @xmath222 in the parametrized expression for @xmath223 .
this yields @xmath224 where we scaled @xmath0 and @xmath225 by their rough upper limits for typical afterglows .
( the lorentz factor @xmath0 decreases during the afterglow evolution , as does @xmath225 if the shock propagates into a stellar - wind environment . )
since the factor @xmath226 in equations ( [ check_is ] ) and ( [ check_es ] ) never exceeds 1 in the absence of a strong external radiation field ( being approximately equal to @xmath227 ; see eq . [ [ y3 ] ] ) , we conclude that the use of the mhd approximation for studying the cooling layers behind radiative shocks should be well justified throughout the evolution of typical grbs and their afterglows . at large distances behind the shock ( @xmath228 )
the electrons and positrons are left with only a small fraction of their internal energy , and all hydrodynamic quantities approach some constant asymptotic value : @xmath229 , @xmath230 . by equating the radiative cooling time with the flow time ,
one infers that @xmath231 in that region ( e.g. , granot et al . 2000 ) .
also , @xmath232 , and hence @xmath233 for @xmath228 .
the fact that @xmath234 for any finite value of @xmath235 implies that @xmath236 , so @xmath173 becomes equal to @xmath237 only asymptotically .
one can determine @xmath237 by setting equation ( [ p_e_tilde ] ) equal to zero and ( numerically ) solving for @xmath173 : this yields @xmath237 as a function of @xmath71 and @xmath90 . in figures
[ u_min_eta ] and [ u_min_m ] we fix one of these variables and show @xmath237 , normalized by @xmath91 , as a function of the other variable . in the limit @xmath238
there are no radiation losses , so the speed of the shocked fluid does not change : @xmath239 . in the limit
@xmath240 the electrons and positrons contain all the postshock internal energy and radiate it away , and the attendant compression eventually renders the magnetic pressure dominant . since the magnetic field is amplified by a factor @xmath241 in the radiative zone ( see eq . [ [ b ] ] ) , a smaller initial magnetic field ( smaller @xmath90 ) results in a lower value of @xmath242 .
furthermore , the total pressure , @xmath243 , is approximately constant throughout the radiative zone , and for @xmath86 it becomes independent of @xmath90 ( @xmath244 ) . since in the limit
@xmath240 the magnetic pressure eventually becomes dominant ( @xmath245 ) , @xmath246 also becomes independent of @xmath90 for @xmath86 , and we obtain @xmath247 , as can be seen in figure [ u_min_m ] .
figure [ u_xi_m ] displays @xmath185 , normalized by @xmath91 , for @xmath248 , @xmath249 , and several values of @xmath90 . for large values of @xmath90 ( @xmath250 ) the asymptotic value @xmath251 ( shown by a dashed line )
is approached at smaller values of @xmath235 ( traced by the asterisks ) as @xmath90 decreases , whereas the converse occurs for smaller values of @xmath90 .
this can be understood as follows . for large values of @xmath90
the initial magnetic field @xmath223 is strong and @xmath252 is close to @xmath253 , so synchrotron cooling is important early on . as the internal energy is radiated , the fluid becomes denser and the magnetic field grows stronger , which enhances the synchrotron emission . since the cooling length is calculated by using the initial synchrotron emissivity , the value of @xmath114 is overestimated in this case and hence the asymptotic value of @xmath254 is approached at lower values of @xmath235 .
this effect is stronger the larger the value of @xmath255 , which for a fixed @xmath71 is obtained for lower values of @xmath90 .
however , for sufficiently small values of @xmath90 , @xmath252 comes to exceed 1 and ic becomes the dominant cooling mechanism . for @xmath86 synchrotron radiation hardly affects the dynamics , and @xmath114 is determined by the ic emissivity just behind the shock transition . since , unlike the magnetic energy density , the photon energy density that determines the ic cooling rate does not increase with @xmath235 , the value of @xmath114 better approximates the actual cooling length , and consequently @xmath237 is approached at larger values of @xmath235 for smaller @xmath90 .
figure [ u_xi_eta ] shows @xmath185 , normalized by @xmath91 , for @xmath256 , @xmath249 , and several values of @xmath71 . in this case the asymptotic value of @xmath173 is approached at smaller values of @xmath235 for larger values of @xmath71 , which correspond to larger values of @xmath257 and a stronger enhancement of the synchrotron cooling .
it is straightforward to verify from equations ( [ p_e_tilde ] ) and ( [ du_dxi ] ) that @xmath258 is a monotonically increasing function of @xmath173 and that @xmath173 is a monotonically decreasing function of @xmath35 .
therefore , as expected , both @xmath258 and @xmath173 decrease the further one gets from the shock front . to demonstrate the evolution of the various components of the pressure , we show in figure [ p_u ] the dependence of @xmath39 , @xmath258 , @xmath259 , @xmath260 and @xmath243 ( each normalized by @xmath79 ) on @xmath173 for @xmath248 and @xmath261 .
equipartition occurs when the total thermal pressure equals the magnetic pressure , @xmath262 , with @xmath263 denoting the value of @xmath173 at equipartition .
however , this condition is not satisfied in all cases .
there is a critical value of @xmath90 , which we denote by @xmath264 , for which @xmath265 .
we find that @xmath266 .
for @xmath267 the magnetic field just behind the shock transition is already above equipartition , @xmath268 , where @xmath269 is given by @xmath270 . on the other hand , @xmath271 for @xmath272 , and
only if @xmath71 is above some critical value @xmath273 will the asymptotic magnetic field be above equipartition .
since @xmath91 , and therefore also @xmath92 , are functions of @xmath90 , @xmath274 is a function only of @xmath90 and @xmath173 ( see eq . [ [ p_tilde ] ] ) .
therefore , @xmath273 may be found by ( numerically ) solving the equation @xmath275=\delta_2(u_{a1 } ) \left[{u_2(u_{a1})\over u_{\rm min}(u_{a1})}\right]^2\ , \ ] ] where @xmath276 $ ] .
figures [ eta_min ] and [ eta_min_delta2 ] show @xmath277 as a function of @xmath90 and @xmath92 , respectively .
the critical value of @xmath71 is close to 1 for small values of @xmath90 ( or @xmath92 ) since , in this case , the initial magnetic field is small , and almost all of the internal energy needs to be lost in order for @xmath278 ( @xmath279 ) to be large enough for @xmath257 to reach equipartition . as we noted in 1 ,
there are direct observational indications that at least in some relativistic shock sources the magnetic field is not far below equipartition . using the calculated radiative shock structure
, we can address the question of how far from equipartition the magnetic field is at the location where the bulk of the observed radiation is emitted .
a fluid element starts radiating just behind the shock transition , where @xmath280 and @xmath281 , and as it flows downstream the magnetic field increases , and so does the magnetic - to - thermal pressure ratio @xmath282 .
thus , even if initially the magnetic field were much below equipartition ( @xmath201 ) , it would be amplified by a large factor ( @xmath283 ) in the radiative zone and might eventually even exceed equipartition ( @xmath284 ) .
it is therefore conceivable that a considerable fraction of the radiation could be emitted from a region where the magnetic field is much larger than the immediate - postshock value .
we now examine this possibility in a quantitative manner .
since the internal energy of the electrons and positrons is characterized by their thermal lorentz factor @xmath285 , their fractional remaining energy is given by @xmath286 .
figure [ pu_delta2_combined ] shows this quantity as a function of @xmath92 at two different instants : ( i ) at equipartition , when @xmath287 , and ( ii ) when @xmath288 . for @xmath71 close to 1 a significant fraction of the internal energy of the electrons and positrons is still left when the field reaches equipartition even if @xmath289 . in this case a significant portion of the radiation originates in a region where the magnetic field is close to equipartition .
we find that , so long as the energy dissipated in the shock transition is deposited mainly in the @xmath8 component ( i.e. , @xmath157 ) , then , even if @xmath92 is as small as @xmath290 , a considerable fraction of the radiation is still emitted at @xmath291 .
note that , for @xmath157 and @xmath201 , @xmath292 ( see eq . [ [ y2 ] ] ) , so that , in the absence of a strong external photon field ( @xmath293 ) , @xmath294 ( see eq . [ [ y3 ] ] ) . this implies that ssc dominates the synchrotron emission , with the synchrotron component contributing only a fraction @xmath295 of the total radiation .
however , as we noted at the end of [ radiation ] ( see also [ b1 ] ) , the synchrotron emission may dominate the ssc contribution in the observed frequency range , in which case a significant fraction of the observed radiation could , in fact , originate from a region where @xmath283 . in figure [ syn_b ]
we show the fraction of the total synchrotron emission that is emitted above a given value of @xmath296 , as well as the fraction of the total emission contributed by the synchrotron process above that value of @xmath296 , as a function of the magnetic field strength behind the shock .
three sets of parameters are shown , labeled a through c , which correspond to the three limits ( from top to bottom , respectively ) of equation ( [ y3 ] ) . in case
a , ssc is dominant near the shock front , whereas further downstream the magnetic field is strongly amplified by compression , and synchrotron emission becomes dominant .
altogether , synchrotron emission constitutes 45% of the total emission , whereas near the maximal field amplitude ( @xmath297 ) the synchrotron process contributes 94% of the total emission .
a large fraction of the synchrotron radiation comes from a region where the magnetic field is close to equipartition even though the magnetic field just behind the shock transition is a factor of 10 below the equipartition value ( @xmath298 ) . in case b
, synchrotron emission is dominant and contributes 84% of the integrated total emission .
the compression behind the shock is small and the magnetic field is hardly amplified , so even near @xmath257 the synchrotron fraction of the total emission ( 86% ) is only slightly higher than the overall fraction . in case c ,
the values of @xmath71 and @xmath90 are the same as in case a , and therefore the initial and final values of the magnetic field are the same . however , in case c the emission is dominated by ic scattering of external photons ( @xmath299 ) , which are absent in case a. synchrotron emission now constitutes only 13% of the integrated total emission , but near @xmath257 it contributes 59% of the total .
most of the synchrotron emission comes from a region where the magnetic field is close to equipartition a larger fraction than in case a. this is a reflection of the fact that the ratio of synchrotron to total emission near @xmath257 relative to that near @xmath223 is larger in case c because of the added cooling by ic scattering of the external photons .
to assess the relevance of cooling - induced magnetic field amplification to relativistic shock sources , one needs to address the following questions : what are the likely values of the normalized energy densities of the electrons / positrons and the magnetic field ( @xmath3 and @xmath4 , respectively ; see 1 ) in the emission region ?
what are the expected values of @xmath3 and @xmath4 immediately behind the shock transition ( which can be written , in terms of the shock parameters defined in 2 , as @xmath300 and @xmath301 ) ? is the shock radiative ? if the shock is radiative but @xmath302 is lower than the observationally estimated value of @xmath303 , can the compressional amplification of the magnetic field in the radiative zone make up the difference ? a strong relativistic shock ( @xmath304 , @xmath305 ) with a transverse magnetic field satisfies @xmath306 and @xmath307 ( see 2 ) , implying that the ratio of the magnetic energy density to the total matter energy density is approximately constant across the shock and hence that @xmath302 can be expressed in terms of the enthalpy density @xmath308 ( @xmath309 for a `` cold '' upstream medium ) and the magnetic field amplitude @xmath310 ahead of the shock : @xmath311 as a concrete example of how the above considerations can be applied , we again focus on grbs and their afterglows .
however , at the end of this section we comment briefly on agns . starting with the @xmath0-ray bursts themselves , we have already observed ( see 1 ) that efficiency considerations imply that @xmath3 ( in particular ) and @xmath4 can not be much smaller than 1 in the emission region . the value of @xmath4 may be constrained by observations through the requirement that the cooling time does not exceed the variability time scale @xmath219 .
this translates into the condition that the lorentz factor of an electron that cools on the time @xmath219 , @xmath312 , must remain smaller than the characteristic synchrotron lorentz factor , @xmath313 , that corresponds to the observed frequency @xmath314 . following the same steps as in the derivation of equation ( [ check_is ] ) and again setting @xmath315
, we deduce @xmath316 ( cf .
@xcite ) . when the energy density of external photons is negligible , one has @xmath317 ( see eq . [ [ y ] ] with @xmath249 ) , so that , if @xmath318 ( in which case @xmath319 ; see eqs . [ [ y2 ] ] and [ [ y3 ] ] ) , the constraint ( [ epsilon_b_1 ] ) on @xmath4 would be relaxed .
however , @xmath318 means that ic is dominant , with the observed synchrotron radiation being only a fraction @xmath320 of the total radiated energy .
the requirement of a reasonable radiative efficiency would then constrain @xmath4 to be not much smaller than @xmath3 ( e.g. , piran 1999 ) .
the latter restriction would not arise if the observed @xmath0-ray emission represented the ssc rather than the synchrotron component , but the synchrotron model appears to account quite well for the observed spectra ( e.g. , ) , whereas an ssc interpretation could entail a prohibitively large energy input at the source .- ray spectrum is identified with a singly scattered ssc component .
the characteristic thermal lorentz factor of the emitting electrons in this picture is @xmath321 , which implies that the peak synchrotron frequency is not very far from the optical regime .
there should thus be an optical pulse that coincides with the prompt @xmath0-ray emission , and the fact that such a component has not been detected ( e.g. , @xcite ; @xcite ) indicates that the compton @xmath104 parameter is very large ( panaitescu and mszros 2000 estimate @xmath322 ) .
now , for the postulated thermal lorentz factor , a typical seed photon undergoes two scatterings before the klein - nishina limit is reached .
since the emissivity ratios of the synchrotron , singly scattered ssc , and doubly scattered ssc components are @xmath323 , it follows that the observed @xmath0-ray pulse may constitute only a fraction @xmath324 of the total burst emission ( and hence that the energy requirement on the source is a factor of @xmath104 larger ) . ] as we have already noted , the prompt @xmath0-ray emission in grbs is believed to be the result of internal shocks within the flow .
the relevant preshock field is then the relic magnetic field advected from the origin , and , to the extent that flux freezing is maintained , @xmath325 keeps the value that was imprinted on the ejecta at the source . ) shows quite generally that @xmath302 will remain constant with time if the medium into which the shock propagates is either a supersonic outflow with a predominantly transverse magnetic field that is frozen into the matter ( in which case both @xmath310 and @xmath326 scale inversely with distance from the center ) or else is uniform ( in which case the values of @xmath310 and @xmath327 are fixed ) .
the former case applies to internal shocks and to the reverse shock ( which propagate in the ejecta ) as well as to a forward shock that propagates in a magnetized stellar - wind environment .
the latter case corresponds to a forward shock in a uniform ism .
this shows that the predictions of conventional emission models of grbs and their afterglows , which simply postulate a constant postshock value of @xmath4 , will by and large continue to apply when this parameter is calculated self - consistently . ] spruit et al .
( 2001 ) pointed out that , under these circumstances , values of @xmath302 as high as @xmath328 may be expected in the commonly invoked grb models that envision large - scale , ordered magnetic fields actively driving the relativistic outflow by tapping the rotational energy of the central source . in particular , @xmath302 could well be @xmath329 in collimated ( jet - like ) outflows of this type .
spruit et al .
further estimated that , if the field is `` passive '' ( in that it does not drive the flow but is merely advected outward from the source ) , then @xmath302 would lie in the range @xmath330 . in the latter case
, compressional magnetic field amplification in the radiative zone could increase the fraction of radiation going into the synchrotron component ( see figs .
[ pu_delta2_combined ] and [ syn_b ] ) . for this mechanism to operate ,
it is necessary that the radiative cooling time be much shorter than the dynamical time and that @xmath3 be @xmath329 .
the first requirement should in general be satisfied in internal shocks that produce grbs ( e.g. , piran 1999 ) .
the conditions under which @xmath3 could be close to 1 are somewhat uncertain , but there are several conceivable mechanisms for depositing a significant fraction of the postshock internal energy in the electron component even if the outflow consists predominantly of an electron - proton plasma ( e.g. , )
. it should be even easier to attain high values of @xmath3 if the plasma is dominated by @xmath8 pairs , a situation that in fact arises naturally in some grb scenarios that involve magnetically driven outflows ( e.g. , usov 1994 ; @xcite ) .
the long - lived afterglow emission of grbs is most likely associated with the forward shock that propagates into the ambient medium .
we have already noted in 1 that comparatively high values of @xmath3 ( @xmath328 ) and @xmath4 ( @xmath331 ) have been directly inferred from observations of the grb 970508 afterglow .
these parameters have so far been less well constrained in other sources , but in a recent modeling of four other afterglows ( @xcite ) values in the range @xmath332 and @xmath333 have been inferred .
it is also worth noting in this connection the suggestion by galama et al .
( 1999 ) that comparatively low values of @xmath4 may characterize afterglow sources that exhibit weak radio emission ( attributed by them to low values of the characteristic frequency and peak flux of the synchrotron spectrum , both of which scale as @xmath334 ) .
the expected values of @xmath335 and @xmath302 in grb afterglows depend on the grb environment ( which directly affects such factors as the pair content , the electron / positron acceleration efficiency , and the degree of magnetization in the shocked gas ) .
the uniform - density model fittings of the grb 970508 afterglow (;
@xcite ) yielded upstream densities in the range @xmath336 , which are typical of a diffuse ism . in the case of grb 980329 , a preshock density
@xmath337 , which is typical of a molecular cloud , has been deduced ( @xcite ) . in the galactic ism
, the inferred magnitude of the alfvn speed does not vary strongly from diffuse to dense environments and corresponds to values of @xmath302 in the range @xmath338 .
if these values are also typical of the host galaxies of grbs , then the synchrotron emission from their associated afterglows would in most cases be too weak to be detected . however , all the afterglow sources observed so far correspond to `` long '' bursts ( @xmath339 ) , for which a plausible origin is the collapse of a massive ( e.g. , wolf - rayet ) star ( e.g. , woosley 2000 ; mszros 2000 ) . in this case , the medium into which the forward shock expands could correspond to the preburst wind from the central star ( e.g. , ; @xcite ) .
it has been suggested ( e.g. , @xcite ) that winds from wolf - rayet stars are magnetically driven and have spatially constant alfvn speeds corresponding to values of @xmath302 that can be as high as @xmath340 .
if such winds constituted the environment into which the forward shock expands , then they could account for afterglows with the lowest values of @xmath4 inferred to date even if there were no additional magnetic field amplification beyond that induced by the passage of the ambient field through the shock transition .
radiative cooling effects could increase the magnitude of @xmath4 in the afterglow emission region above its value immediately behind the forward shock .
the cooling time behind the forward shock is typically shorter than the dynamical time during the early phase of the afterglow , but subsequently this inequality is reversed . for a shock propagating into a uniform - density ism ,
the time of transition from radiative to adiabatic evolution was calculated by sari et al .
( 1998 ) and is given by @xmath341 where @xmath342 is the initial lorentz factor of the ejecta . in the absence of a strong coupling between protons and electrons behind the shock transition
, @xmath343 is required for the shock to be radiative .
equation ( [ t_ad_ism ] ) indicates that , even if this condition is satisfied , @xmath344 can not be much smaller than 1 if the radiative phase is to last through a significant fraction of the early evolution of a typical afterglow ( i.e. , a few days or so ) . taking @xmath345 as a representative ism value , we infer @xmath346 , implying that @xmath347 is only a few seconds !
in fact , in this case , even if the radiative phase lasted longer , the expected magnetic field amplification in the radiative zone would not be sufficient to raise @xmath4 to a detectable level ( see [ results ] ) .
in the case of a stellar - wind environment , the ambient density scales with distance @xmath348 from the source as @xmath349 , where the constant @xmath350 is defined in terms of the wind mass - loss rate @xmath351 and speed @xmath352 .
we are interested in the evolutionary phase during which the forward shock has already decelerated significantly but is still relativistic .
assuming that the mass is concentrated in a thin , spherical shell of radius @xmath21 and lorentz factor @xmath0 , we can write @xmath353 , where @xmath354 ( @xmath355 ) is the initial mass of the ejecta and @xmath356 is the rest mass of the swept - up matter ( @xcite ; ) . in our case @xmath357 and therefore @xmath358 , where @xmath359 is the distance from the center where the flow becomes nonrelativistic . using @xmath360
, we obtain @xmath361 the lorentz factor of an electron that cools on a time scale @xmath34 is given by @xmath362 whereas the typical lorentz factor of the electrons can be approximated by @xmath363 ( e.g. , sari et al . 1998 ) . assuming that the preshock field @xmath364 scales as @xmath365 ( e.g. , @xcite ) , we find @xmath366 , so the ratio of the radiative cooling time to the dynamical time ( @xmath367 ) is @xmath368 .
this shows explicitly that at early times @xmath369 and the evolution is radiative .
the transition from the radiative to the adiabatic phase occurs approximately when @xmath370 increases to 1 . using @xmath371 and @xmath372
, we obtain @xmath373 where @xmath374 corresponds to a stellar wind characterized by a speed of @xmath375 and a mass loss rate of @xmath376 ( ) . as can be seen from a comparison of equations ( [ t_ad_ism ] ) and ( [ t_ad_wind ] ) , the radiative phase of the forward shock will typically last longer in a wind environment than in a uniform - density medium , which can be attributed ( ) to the higher ambient density encountered at early times by a shock that propagates into a wind . using equations ( [ y2 ] ) and ( [ y3 ] ) to approximate @xmath377 and setting @xmath378 ( the expected value in a strongly magnetized wolf - rayet outflow )
, we infer from equation ( [ t_ad_wind ] ) that @xmath379 . during the radiative phase
, a typical electron could radiate up to @xmath380 of its postshock random energy in a region with @xmath381 ( see fig .
[ pu_delta2_combined ] ) .
however , the estimated value of @xmath382 is too small for the radiative field - amplification mechanism to apply to the observations of a source like grb 970508 , in which values of @xmath383 have been inferred over substantially longer time scales .
furthermore , even if the model were applicable , this mechanism could not account for a value of @xmath4 as high as 0.1 ( the best - fit value for a stellar - wind model ; ) if @xmath302 is not larger than @xmath384 .
another potential source of detectable emission in grb sources is the reverse shock , which runs into the ejecta shell when the latter starts to be decelerated by the inertia of the swept - up ambient gas .
in fact , a shock of this type is believed to be the source of the prompt optical emission ( or `` optical flash '' ) that was observed in grb 990123 ( @xcite ; sari & piran 1999a , b ; @xcite ) as well as of a rapidly decaying `` radio flare '' detected about a day after the burst in this source ( and evidently also present in grb 990123 ; kulkarni et al .
so far , however , these observations have not provided strong constraints on the parameters @xmath3 and @xmath4 in the emission region .
it may , however , be expected that , since the reverse shock propagates into the ejecta , it is characterized by values of @xmath335 and @xmath302 that are similar to those in internal shocks .
grb 990123 was a comparatively long @xmath0-ray burst : such sources likely possess a relatively thick ejecta shell ( of thickness @xmath385 as measured by a stationary local observer ) , and their prompt optical emission should peak at @xmath386 , the shock crossing time of the shell ( which is approximately equal to the duration of the burst ) , by which time the reverse shock would be relativistic ( e.g. , @xcite ) .
this is a natural time scale to use in a comparison with the cooling time for determining whether the shock is radiative .
we can repeat the procedure applied above to the forward shock by comparing the characteristic lorentz factors @xmath387 and @xmath388 , except that in this case we substitute for @xmath0 in equation ( [ gamma_c ] ) for @xmath388 , by the lorentz factor @xmath389 of the shocked shell material , which is given in terms of the lorentz factor @xmath390 of the un - shocked portion of the ejecta shell ( both measured in the stationary frame ) by @xmath391 ( e.g. , @xcite ) , where @xmath392 is the ratio of the proper mass densities in the shell and the ambient medium ( assumed to be dominated by protons ) and @xmath393 is the radius where the reverse shock crosses the shell .
similarly , we replace @xmath0 in equation ( [ gamma_m ] ) for @xmath387 by the relative lorentz factor between the shocked and the unshocked shell material , @xmath394 . in the case of a uniform ambient medium
, the condition @xmath395 implies @xmath396 where we have normalized @xmath397 and @xmath390 by their estimated values in grb 990123 ( @xcite ) .
is probably a bit conservative . for one thing ,
as we have demonstrated in 3 , the cooling induces compression in the radiative layer , which has the effect of enhancing the radiative losses .
furthermore , the radiative efficiency decreases only gradually after @xmath388 comes to exceed @xmath387 : in fact , the fraction of the internal electron / positron energy that is radiated by the synchrotron process after that time is given by @xmath398 , where @xmath39 ( which has a canonical value of 2.5 ) is the power - law index of the radiating particle distribution ( e.g. , @xcite ) .
thus , the right - hand sides of the inequalities ( [ epsilon_b_epsilon_e_ism])([epsilon_b2_wind ] ) are probably only upper limits . ] assuming @xmath399 [ which implies , in the absence of an external radiation field , @xmath377 ] , we obtain a lower bound on @xmath302 , @xmath401 the inequality ( [ epsilon_b2 ] ) could in principle be satisfied if the grb outflow is driven magnetically , in which case the ejecta could be characterized by @xmath402 . however , even if @xmath302 is near its estimated lower bound , compressional amplification could boost the magnetic field to equipartition values in the radiative zone ( see figs .
[ eta_min_delta2 ] and [ pu_delta2_combined ] ) . the corresponding condition for a stellar wind environment ( assuming strong hydrogen depletion as appropriate for a wolf - rayet progenitor ) is @xmath403 ( ) . for @xmath404 and a negligible external radiation field , equation ( [ epsilon_b_epsilon_e_wind ] ) implies @xmath405 which demonstrates that the reverse shock is always radiative in this case for plausible parameter values .
it is worth noting , however , that if the reverse shock is radiative at the time that it crosses the shell , then it will rapidly cool and become invisible at later times ( e.g. , @xcite ) . since this was not the case for the optical flash in grb 990123 , we can conclude , on the basis of equations ( [ epsilon_b_epsilon_e_wind ] ) and ( [ epsilon_b_epsilon_e_ism ] ) , that the outflow in this source does not expand into a wind ( as was already inferred by ) and
that the value of @xmath406 behind the reverse shock is @xmath407 .
we can summarize our discussion of grb sources as follows .
the field amplification mechanism discussed in this paper could in principle operate in shocks that are driven into the ejecta : either the internal shocks that give rise to the prompt @xmath0-ray emission , or the reverse shock that produces the optical flash and radio flare emission . except for reverse shocks in sources with a uniform ambient medium in the case when the condition ( [ epsilon_b_epsilon_e_ism ] ) is not satisfied
, these shocks will be radiative if most of the energy dissipated in the shock transition is deposited into the electron / positron component .
compressional field amplification in the radiative zone of these shocks could enable a significant fraction ( @xmath408 ) of the radiated energy to be emitted from regions where the field is substantially stronger ( by up to @xmath409 orders of magnitude in @xmath4 ) than immediately behind the shock transition .
if the preshock magnetization is comparatively weak ( @xmath410 ) , the radiative amplification could significantly enhance the synchrotron emission and may in certain cases help to bring it above the detection limit .
if the ejecta is already moderately strongly magnetized ( @xmath411 ) , this process could increase the field in the emission region to equipartition values .
we have also concluded that the compressional amplification mechanism will not significantly affect the afterglow emission of grb sources , both because the forward shock typically undergoes a transition to adiabatic evolution early on , and because , even during the radiative phase , the expected amplification falls short of the required enhancement for standard ism environments .
we noted that magnetized stellar winds from wolf - rayet stars could provide an adequate environment in which afterglows with values of @xmath4 that lie near the low end of the observationally inferred range would be produced in adiabatic shocks without requiring further field amplification .
however , the comparatively high ( @xmath412 ) values of @xmath303 indicated in a source like grb 970508 can not be accounted for by the basic shock model and typical ism parameters .
we do not pursue this topic further in this paper , but we note that several suggestions that bear on the need to have high values of @xmath4 and @xmath3 in such shocks have already been made in the literature ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) .
the observational data on spectral flares in agns are not as extensive as in grb sources , and , correspondingly , their theoretical study has so far been less developed . however , as we noted in 1 , the blazar flares have been similarly interpreted in terms of nonthermal emission from relativistic , and likely highly collimated , outflows from the vicinity of a central compact object . in blazars vs. @xmath413 in grbs . ] in particular , the variable emission has been modeled in terms of internal shocks in a magnetized relativistic outflow , possibly formed from overtaking collisions of disturbances ( `` shells '' ) in the flow ( e.g. , @xcite ; @xcite ; @xcite ; levinson 1998 , 1999 ) . in these models ,
the high - energy emission is attributed to an ssc or an erc component .
the quasi - isotropic external radiation field invoked in the erc interpretations can arise from the nuclear accretion disk ( e.g. , ) , or , perhaps more likely , from scattering by the broad emission - line region ( belr ) `` clouds '' of the nuclear continuum radiation ( e.g. , @xcite ; @xcite ) or of the beamed jet emission itself ( @xcite ) , as well as from the near - infrared emission of warm dust outside the belr ( @xcite ) . as we discussed in [ results
] , if the emission from a radiative shock is dominated by ic scattering of external photons , then the synchrotron radiation component of the shock will have a relatively stronger contribution from regions of high values of @xmath296 ( see fig .
[ syn_b]).-ray emission arises in the course of the compton - drag deceleration of a relativistic outflow as it propagates through the debris of a progenitor massive star ; in this scenario , magnetic fields play no role in the emission process . on larger scales , madau , blandford , & rees ( 2000 )
proposed that backscattering of the grb radiation by the comparatively dense massive - star environment could lead to a delayed mev emission , again through the compton - drag effect on the bulk flow . in the latter case ,
erc emission by a `` hot '' ( i.e. , relativistic ) electron / positron component of the flow could in principle also contribute to the observed radiation ( cf . the analogous proposal for blazars by @xcite ) .
so far , however , there has been no discussion in the literature of the possible appearance of an erc emission component in grb sources . ]
most of the blazar emission models considered to date have , however , taken @xmath387 to lie below @xmath388 ( e.g. , @xcite ; sikora et al . 1994 ; but see @xcite ) and thus correspond to nonradiative shocks .
it is , nonetheless , likely that the cooling times could be quite short in these sources , as suggested , for example , by very rapid flares in blazars like 3c 279 ( e.g. , @xcite ) , mrk 421 ( e.g. , @xcite ) , and ao 0235 + 164 ( e.g. , @xcite ) .
furthermore , there are indications that agn jets may contain a significant , or even dominant , electron / positron component ( e.g. , @xcite ; @xcite ; @xcite ) , which would naturally lead to high values of @xmath3 in the shocks . given that the outflows are likely magnetically driven ( e.g. , @xcite ; @xcite ) , comparatively high values of @xmath4
may also be expected .
it is thus conceivable that at least some of the synchrotron emission from these sources originates in radiative shocks in which the magnetic field is amplified by cooling - induced compression .
gamma - ray bursts and their afterglows , as well as energetic flares from blazars and miniquasars , are commonly interpreted in terms of shocks in a relativistic outflow from a compact object that emit nonthermal ( synchrotron and inverse compton ) radiation . in this paper
we have investigated the structure of radiative shocks in such sources .
a shock is `` radiative '' if the emission mechanism can tap into the bulk of the postshock internal energy and if the radiative cooling time behind the shock transition is shorter than the relevant dynamical time ( the characteristic time for adiabatic energy losses ) .
such a shock is characterized by a radiative zone of finite width where most of the energy dissipated in the shock transition front is radiated away .
the structure of radiative shocks has been previously studied in the nonrelativistic regime in the context of the interstellar medium , where the dominant emission processes are thermal .
one particularly interesting finding of these studies has been the strong effect that such a shock could have on the normal component of a magnetic field that is frozen into the gas , and , in turn , the potentially important role that the field could play in limiting the compression in the shock .
when the shock is nonradiative , the compression ( @xmath414 ) and resultant magnetic field amplification ( @xmath415 ) have a value of at most a few in the nonrelativistic regime ( generalizing to a few times the shock lorentz factor @xmath416 in the relativistic case , with @xmath10 and @xmath12 measured in the fluid rest frame ) .
efficient cooling can lead to a strong enhancement of the density and the field amplitude in the radiative zone , but when the magnetic pressure attains equipartition with the thermal pressure , further compression is inhibited and subsequent cooling proceeds at nearly constant density . we have shown that , when synchrotron radiation dominates the cooling , the interplay between the cooling and the compression becomes more pronounced on account of the feedback effect between the field amplification and the emission process : a stronger magnetic field increases the emissivity , which in turn induces a larger compression that further amplifies the field .
the cooling - induced compressional field amplification may have potentially significant consequences in sources that harbor radiative shocks : if the preshock magnetic field is low , this mechanism could in some cases increase it to a level where the synchrotron radiation becomes detectable , and if the preshock field is already moderately strong ( so that it is within an order of magnitude of equipartition in the shocked but uncooled gas ) , this process could result in a considerable fraction ( @xmath417 ) of the radiation being emitted from regions where @xmath418 .
when the shock is radiative but the field immediately behind the shock transition is below equipartition , the initial cooling would be dominated by the inverse - compton process : either synchrotron self - compton ( ssc ) or external - radiation compton ( erc ) , depending on which seed - photon energy density is higher in the shock frame . in this case ,
a larger fraction of the synchrotron emission comes from regions with a stronger magnetic field if erc dominates ssc . for standard scenarios of grb outflows and their environments
, this field amplification mechanism may be relevant to internal shocks that produce the prompt @xmath0-ray emission and possibly also to the reverse shock responsible for the optical `` flash '' and radio `` flare '' in sources like grb 990123 .
it will likely not be of much relevance to the forward shock that gives rise to the grb afterglow , although we pointed out that if the shock expands into a magnetized wind from a wolf - rayet star then no further field amplification would be required to account for the lowest @xmath419 values that have so far been inferred from afterglow observations .
we also suggested that radiative shocks in which this mechanism operates could occur in blazars .
we are grateful to tsvi piran for many valuable conversations .
jg acknowledges a priscilla and steven kersten fellowship at the university of chicago and thanks the department of astronomy and astrophysics at the university of chicago for its hospitality .
ak acknowledges a forchheimer fellowship and thanks the racah institute at the hebrew university for hospitality . this research was supported in part by the us
israel bsf grant bsf-9800225 ( jg ) and by nasa grant nag 5 - 9063 ( ak ) . | we calculate the structure of a relativistic shock wave in which the internal energy of the shocked fluid is radiated away on a time scale much shorter than the characteristic shock propagation time . the shock is assumed to move through a uniformly magnetized , neutral plasma consisting of protons , electrons , and positrons , and allowance is made for the possible production of electron / positron pairs in the shock itself .
the radiation mechanism is taken to be synchrotron and inverse - compton emission ( the latter involving both synchrotron - produced and externally supplied seed photons ) by the electrons and positrons .
we simplify the discussion by considering a shock in which the magnetic field is transverse to the direction of propagation and focus attention on the properties of the radiative zone that forms behind the shock transition .
in particular , we investigate the possibility that the compression induced by the cooling of the gas amplifies the magnetic pressure until it reaches ( and ultimately exceeds ) equipartition with the thermal pressure ( which , in turn , limits the overall compression ) .
we show that , if a significant fraction of the postshock thermal energy is deposited in the electron / positron component , then a considerable portion of the emitted radiation will come from regions of strong magnetic field even if the field immediately behind the shock transition is well below equipartition .
this field amplification mechanism may be relevant to the production of synchrotron flares in blazars , miniquasars , and gamma - ray burst sources .
we consider the latter application in some detail and show that this process may play a role in the prompt @xmath0-ray and possibly also the optical `` flash '' and radio `` flare '' emission , but probably not in the afterglow . |
it is well known that shannon s source - channel separation result for point - to - point communication @xcite does not hold for in general for multi - terminal systems , and thus joint source - channel coding may be required to achieve the optimum .
one simple yet intriguing scenario where source - channel separation is known to be suboptimal is broadcasting gaussian sources on gaussian channels .
when a single gaussian source is at the encoder , the achievable distortion region is known when the source bandwidth and the channel bandwidth are matched @xcite , for which a simple analog scheme is optimal .
however when the source and channel bandwidths are not matched , exact characterization of the achievable distortion region is not known .
the best known coding schemes are based on joint source - channel using hybrid signaling @xcite , and approximate characterizations were given in @xcite ; see references therein for related works . as a simple extension to this problem of single gaussian source broadcasting , when the source is a bandwidth - matched bivariate gaussian and
each decoder is interested in one source component , only partial characterization is known when uncoded scheme is shown to be optimal under certain signal - to - noise ratio conditions @xcite . in this work ,
we provide a complete characterization of the achievable distortion region for broadcasting bivariate gaussian sources over gaussian channels when each receiver is interested in only one component , where the source bandwidth and the channel bandwidth are matched .
we further show that in this joint source channel coding setting , the gaussian problem is the worst scenario among the sources and channel noises with the same covariances , in the sense that any distortion pair that is achievable in the gaussian setting is also achievable for other sources and channel noises .
our work is built on the outer bounds given in @xcite and we show that a hybrid coding scheme ( different from the one given in @xcite proposed for the same problem ) can achieve the outer bounds .
our main contribution in this work is this new coding scheme and a detailed and systematic analysis of the inner and outer bounds , which result in a complete solution . to the best of our knowledge ,
this is the first case in the literature that a hybrid scheme is shown to be optimal for a joint source - channel problem , whereas neither an uncoded scheme alone nor a separation - based scheme alone is optimal .
let @xmath0 be a memoryless and stationary bivariate gaussian source with zero mean and covariance matrix @xmath1 where without loss of generality , we can assume @xmath2 .
the vector @xmath3 will be written as @xmath4 for @xmath5 .
we use @xmath6 to denote the domain of reals .
the gaussian memoryless broadcast channel is given by the model @xmath7 where @xmath8 is the channel output observed by the @xmath9-th receiver , and @xmath10 is the zero - mean independent additive gaussian noise in the channel .
thus the channel is memoryless in the sense that @xmath11 is a memoryless and stationary process .
the variance of @xmath10 is denoted as @xmath12 , and without loss of generality , we shall assume @xmath13 . throughout the paper ,
we use natural logarithm . , width=453 ]
the mean squared error distortion measure is used , which is given by @xmath14 for @xmath5 .
the encoder maps a source sample block @xmath15 into a channel input block @xmath16 ; the decoder observing channel output block @xmath17 reconstruct the source @xmath4 within certain distortion ; see fig .
[ fig : systemdiag ] .
the channel input @xmath18 is subject to an average power constraint .
more formally , the problem is defined as follows .
an @xmath19 bivariate gaussian source - channel broadcast code is given by an encoding function @xmath20 such that @xmath21 and two decoding functions @xmath22 and their induced distortions @xmath23 where @xmath24 is the expectation operation .
in the definition , @xmath25 in the expression @xmath26 is understood as the length-@xmath27 vector addition .
it is clear that the performance of any gaussian joint source - channel code depends only on the marginal distribution of @xmath28 , but not the joint distribution @xmath29 .
note that this implies that physical degradedness does not differ from statistical degradedness in terms of the system performance .
since the gaussian broadcast channel is always statistically degraded , we can assume physical degradedness without loss of generality .
a distortion pair @xmath30 is achievable under power constraint @xmath31 , if for any @xmath32 and sufficiently large @xmath27 there exists an @xmath19 bivariate gaussian source - channel broadcast code such that @xmath33 the collection of all the achievable distortion pairs under power constraint @xmath31 for a given bivariate source is denoted by @xmath34 , and this is the region we shall characterize in this work . in fact , we shall determine the following function which clearly provides a characterization of the achievable distortion region .
@xmath35 note that @xmath34 is a closed set , and thus the minimization above is meaningful . since the minimum @xmath36 that is achievable is given by @xmath37 when the second receiver is completely ignored , the function @xmath38 is thus only meaningfully defined on the domain @xmath39 $ ] .
when the source is not gaussian but with the same covariance structure , and the channel noises are not gaussian , but has the same variances , we shall denote the achievable distortion region as @xmath40 .
we shall show that the gaussian setting has the worst case property , and thus the gaussian scheme is a `` robust '' scheme .
the main result of this paper is summarized in the following theorem .
[ theorem : maintheorem ] if @xmath41 , where @xmath42 , then @xmath43 if @xmath44 , then in the range @xmath45 $ ] , @xmath46 where @xmath47}.\end{aligned}\ ] ] on the other hand if @xmath48 , then @xmath49\\ d_2^h(p,\sigma^2,\rho , n_1,n_2,d_1)&d_1\in [ d^-_1,d^+_1 ] \end{array}\end{aligned}\ ] ] where @xmath50}\end{aligned}\ ] ] and @xmath51 _ remark : _ depending on the power constraint , the achievable distortion region may have two operating regimes . in the regime where @xmath52 , the uncoded scheme given in @xcite is optimal , whereas in the regime @xmath53 , a hybrid scheme given in the next section is optimal , but the uncoded scheme is not .
typical achievable distortion regions for these two cases will be illustrated in the next section after the hybrid coding scheme is given , where more observations regarding these schemes can be discussed .
consider a source pair @xmath54 whose covariance is given by ( [ eqn : covariance ] ) , and channel noise pair @xmath55 whose variances are given by @xmath56 .
we have the following theorem .
[ theorem : worstcase ] if @xmath57 , then @xmath58 .
this theorem essentially says the gaussian setting has the worst sources and channels among those having the same covariance structure , a result similar to the well - known ones that the gaussian source is the worst source @xcite ( ex .
13.8 ) and the gaussian channel is the worst channel @xcite ( ex .
10.1 ) .
the proofs of theorem [ theorem : maintheorem ] and theorem [ theorem : worstcase ] are given in the next section .
in this section , we shall first review some previous results which provide partial characterization of the achievable distortion region , then give a new hybrid coding scheme , which provides the missing portion of the characterization .
finally , we provide a proof for the worst - case property of the gaussian setting .
it is straightforward to show that a simple analog scheme by sending @xmath59 directly ( after certain scaling ) achieves the distortion pair ( see also @xcite ) @xmath60 for which @xmath61 can not be reduced even when the first receiver is not present .
thus we trivially have @xmath62 thus we only need to characterize the function @xmath38 when @xmath63 $ ] .
the uncoded scheme was investigated thoroughly in @xcite , which was shown to be optimal under certain conditions .
more precisely , the uncoded scheme uses the single letter mapping @xmath64 where @xmath65 $ ] and @xmath66 .
the reconstruction in this uncoded scheme is thus naturally the single letter mapping given by @xmath67 $ ] for @xmath5 .
the distortion pair is thus given by @xmath68\\ & \hat{d}_2(\alpha,\beta)=\sigma^2\left[\frac{p\alpha^2(1-\rho^2)}{(p+n_2)(\alpha^2 + 2\alpha\beta\rho+\beta^2)}+\frac{n_2}{p+n_2}\right].\end{aligned}\ ] ] the main result of @xcite is the following theorem .
[ theorem : bross ] for any @xmath57 , and @xmath69 there exist @xmath70 such that @xmath71 where the threshold @xmath72 is given by @xmath73 this result partially characterizes the achievable distortion region @xmath34 , or equivalently the function @xmath38 .
later we shall translate it into a more convenient form , such that by combining all the portions , the function @xmath38 can be characterized completely .
also inherent in the proof of @xcite is a genie - aided outer bound , where the source @xmath59 is given directly to the decoder observing @xmath74 ; see fig .
[ fig : genie ] .
this outer bound , which is a special case of the more general outer bound given in @xcite , was rederived in a simpler manner in @xcite ; the general outer bound in @xcite is more involved , and it usually requires optimization over several parameters . let us follow the notation in @xcite and denote the achievable distortion @xmath75 in the genie - aided setting as @xmath76 , and the simplified characterization of the achievable distortion region in @xcite is as follows .
[ theorem : genie ] the achievable @xmath76 when @xmath59 is present at the first receiver is all the pairs @xmath77 for some @xmath65 $ ] .
as we shall show in the next subsection , this genie - aided outer bound is in fact tight for a certain regime . intuitively speaking ,
since the first receiver is stronger than the second receiver , and @xmath59 is not required at the first receiver , the information @xmath59 is redundant at the first receiver in a certain sense , and thus we can expect the genie - aided outer bound to be reasonably good . ,
width=453 ] the coding scheme we propose is a hybrid one , where the channel input is given by @xmath78 where @xmath79 is ( roughly ) the quantized version of source sequence @xmath80 after some proper scaling ; @xmath81 is the digital portion of the channel input , and @xmath82 are two scaling parameters to be specified later .
more precisely , consider the single - letter distribution @xmath83 where @xmath84 is a zero mean gaussian random variable independent of everything else with variance @xmath85 , and @xmath86 is again a scaling parameter to be specified later . since @xmath87 is a markov string , the joint distribution of @xmath88 is uniquely determined , and they are jointly gaussian .
we also need to define the coefficients @xmath89 and @xmath90 @xmath91=a_kx_d+b_k(\tilde{\alpha } s_1+\tilde{\beta } s_2+z_k ) , \qquad k=1,2.\end{aligned}\ ] ] this proposed hybrid scheme in this work is somewhat similar to the scheme given in @xcite for joint source channel coding on the multiple access channel . in what follows ,
we only outline the coding scheme and some important analysis steps , but omit the rather technical detailed proof ( a rigorous proof can be straightforwardly adapted from that given in @xcite ) . * codebook generation : generate @xmath92 codewords single - letter wise according to the marginal distribution of @xmath93 ; this codebook is revealed to both the encoder and decoders . *
encoding : find a sequence @xmath79 in the codebook that is jointly typical with the source sequence @xmath80 ; if successful , the transmitter sends @xmath94 . * digital decoding : the @xmath9-th decoder tries to find a unique @xmath81 codeword in the codebook that is jointly typical with @xmath17 ; the decoder also recovers the sequence @xmath95 after removing the digital codeword . *
estimation : if the digital decoding succeeds , then the decoder reconstructs the respective source sequence as @xmath96 .
an error occurs in the above scheme if the encoder fails to find a codeword that is jointly typical with @xmath80 , or one of the decoders fails to find the correct digital codeword .
note that due to the markov string @xmath87 , we indeed have that the chosen @xmath79 is jointly typical with @xmath97 with high probability in the above scheme . because the second receiver is a degraded version of the first receiver , the error probability can be made arbitrarily small if the following condition holds ( after ignoring the @xmath98 s often seen in the typicality argument ) @xmath99 furthermore , to ensure the power constraint is not violated , we need @xmath100 it is evident that as long as @xmath101 we can find @xmath102 such that ( [ eqn : powerconstraint ] ) holds with equality , because the left hand side of ( [ eqn : powerconstraint ] ) is monotonically increasing in @xmath103 in the range @xmath104 $ ] , and we shall choose precisely this value in the scheme . with this choice of @xmath103 , ( [ eqn :
rateconstraint ] ) can be simplified .
note that @xmath105 moreover , we have also @xmath106 putting ( [ eqn : leftside ] ) and ( [ eqn : rightside ] ) together , ( [ eqn : rateconstraint ] ) finally reduces to the following expression through some algebraic calculation @xmath107 we shall thus choose @xmath108 in the scheme ( again ignoring some asymptotically small @xmath98 terms ) . by the joint typicality of the sequences when digital decoding succeeds , the distortions can be computed , which is summarized in the following theorem . for @xmath82 such that @xmath109 the proposed hybrid scheme can achieve the distortion pair @xmath110 @xmath111+n_1n_2}{(1-\rho^2)\tilde{\alpha}^2(p+n_1)\sigma^2+pn_1+n_1n_2+(\tilde{\alpha}^2+\tilde{\beta}^2 + 2\tilde{\alpha}\tilde{\beta}\rho)(n_2-n_1)\sigma^2},\nonumber\\ & \tilde{d}^h_2=\sigma^2\frac{\tilde{\alpha}^2(1-\rho^2)\sigma^2+n_2}{p+n_2}.\label{eqn : innerbound}\end{aligned}\ ] ]
it is not clear a priori why in the analog part , the source @xmath59 needs to be included .
however , this is rather critical , as we shall discuss shortly in the next section .
the proof of the main result is now organized into three propositions .
we start by rewriting theorem [ theorem : bross ] in the following form .
we have * if @xmath44 , then the uncoded scheme is optimal in the range @xmath45 $ ] and @xmath52 ; * if @xmath48 , then the uncoded scheme is optimal over the range @xmath112\cup [ d^+_1,d^{\max}_1]$ ] , and in this range @xmath52 .
this proposition is a direct consequence of theorem [ theorem : bross ] .
it is straightforward to show that when @xmath44 , the condition in ( [ eqn : snrbound ] ) holds for any choice of @xmath36 by analyzing the quadratic inequality ( see @xcite corollary 1 ) ; on the other hand , when @xmath48 , ( [ eqn : snrbound ] ) only holds in the given range . given this alternative form , we only need to focus on the case that @xmath48 , and moreover , in the range @xmath113 $ ] .
it is worth noting that it is always true that @xmath114 .
next we use theorem [ theorem : genie ] to write an lower bound for the function @xmath38 .
[ prop : lowerbound ] for any @xmath63 $ ] , @xmath115 .
this proposition is a direct consequence of the fact that the achievable @xmath76 region given in theorem [ theorem : genie ] is an outer bound for @xmath34 ; the proof is obtained by simple algebraic manipulation to eliminate the parameter @xmath116 in theorem [ theorem : genie ] , and we thus omitted the details . the following proposition is the final piece for the proof of the main result .
the proposed hybrid scheme achieves @xmath117 when @xmath113 $ ] by choosing @xmath118 in order to prove this proposition , we need to show firstly that the given choice of @xmath119 does not violate the power constraint , i.e. , the condition ( [ eqn : reducedpowercondition ] ) is satisfied ; secondly , the given choice of @xmath119 reduces the distortion pairs in ( [ eqn : innerbound ] ) to those given in ( [ eqn : matchingbound ] ) . notice that @xmath120 and @xmath121 given in ( [ eqn : alphachoice ] ) and ( [ eqn : betachoice ] ) are well - defined and non - negative when @xmath122 $ ] , since @xmath123 in this range . for ( [ eqn : reducedpowercondition ] ) to hold , we need to have @xmath124 solving the inequality gives that the necessary and sufficient condition that @xmath125 substituting ( [ eqn : alphachoice ] ) into ( [ eqn : middlealpha ] ) and after certain algebra , we arrive at the conditions under which ( [ eqn : middlealpha ] ) is true , which is exactly @xmath113 $ ] . in order to show that the lower bound as stated in proposition [ prop : lowerbound ]
can be achieved , we first simplify the expression of @xmath126 given in ( [ eqn : innerbound ] ) in terms of @xmath120 , which ( after quite some algebra ) eventually gives @xmath127 substituting our choice of @xmath120 given in ( [ eqn : alphachoice ] ) into ( [ eqn : almostfinald1 ] ) leads to @xmath128 ; again substituting ( [ eqn : alphachoice ] ) into the expression of @xmath129 in ( [ eqn : innerbound ] ) gives the expression stated in the proposition , which completes the proof .
readers may wonder how the magic value of @xmath121 was found , which optimizes @xmath126 in the hybrid scheme . indeed , directly optimizing the distortion @xmath126 is extremely cumbersome this way . ] . to circumvent this difficulty
, we instead solve for @xmath121 such that the inner bound matches the outer bound , which gives the given expression .
this approach is less intuitive , since it is possible that neither the genie - aided outer bound nor the hybrid scheme inner bound is tight , however extensive numerical comparison indeed suggested that these bounds match , which motivated us to take such an approach . in fig .
[ fig : regime1 ] and fig .
[ fig : regime2 ] we give two typical achievable distortion regions , where for comparison we also include the performance of a simple separation - based scheme where the digital broadcast messages encoding @xmath59 and @xmath130 ( conditioned on the reconstructed @xmath59 ) , respectively . in both figures ,
each horizontal red line is the performance of the hybrid scheme by varying @xmath121 while keeping @xmath120 fixed .
note that the hybrid scheme includes the uncoded scheme as a special case when the digital portion is allocated no power .
[ fig : regime1 ] is plotted with the choice of source and channel satisfying the condition @xmath44 , and thus the uncoded scheme is always optimal . for this case , adding digital code in the hybrid scheme is always inferior .
in contrast , fig [ fig : regime2 ] is plotted under the condition @xmath48 , and thus uncoded scheme is only optimal at high and low @xmath36 regimes . in the regime
that uncoded scheme is not optimal , it can be seen that even when analog portion does not include @xmath59 , the hybrid scheme can sometimes outperform the uncoded scheme , however , by optimizing @xmath59 in the analog portion , the inner and outer bounds indeed match .
moreover , observe that the distortion of @xmath126 achieved by the hybrid scheme is not monotonic in @xmath121 when @xmath120 is fixed ( each red line ) , where the two extreme values of @xmath121 give the uncoded scheme and the hybrid scheme without analog @xmath59 , respectively . .
here @xmath131 , @xmath132 , @xmath133 and @xmath134.[fig : regime1],width=642 ] .
here @xmath131 , @xmath132 , @xmath133 and @xmath135.[fig : regime2],width=642 ] next we prove theorem [ theorem : worstcase ] , i.e. , the worst case property of the gaussian setting .
we have shown an optimal scheme in the gaussian setting is the proposed hybrid scheme , and thus we can limit ourselves to the distortion pairs achievable by this scheme .
in fact we shall continue to use this scheme and the associated parameters when the sources and channel noises are not gaussian .
more precisely , we shall now use @xmath136 instead of @xmath93 to construct the digital source codewords @xmath137 where @xmath84 is still a gaussian random variable with variance of @xmath138 , independent of everything else .
the overall covariance structure of the scheme remains intact as in the gaussian case , and thus the same ( mse ) distortion pairs can be achieved , as long as the digital codewords can be correctly decoded at both the decoders , i.e. , @xmath139 with our choices of the parameters , where @xmath140 is the channel output in this non - gaussian setting .
note that unlike in the gaussian case , here the broadcast channel is not necessarily degraded , and thus we also need to make sure that the codeword can be correctly decoded at the first decoder .
to show the second decoder can succeed ( with high probability ) , we only need to observe that @xmath141 where in the second inequality we substitute @xmath93 of the gaussian version of the problem , because the terms have the same covariance structure , and gaussian distribution maximizes the differential entropy ; in the last but one equality , we add and subtract the same term @xmath142 , and the last equality is due to our specific choice of the parameters in the gaussian problem . similarly , we can write @xmath143 where the second inequality is guaranteed by the relation in the gaussian case , which is indeed a degraded broadcast channel .
this completes the proof .
we provide a complete solution for the joint source - channel coding problem of sending bivariate gaussian sources over gaussian broadcast channels when the source bandwidth and channel bandwidth are matched .
thus this problem joins a limited list of joint source - channel coding problems for which complete solutions are known .
possible extension of this work includes the case with more than two users or more than two sources , and approximate characterization for bandwidth mismatched case , which are part of our on - going work . c. tian , s. n. diggavi and s. shamai , `` approximate characterizations for the gaussian broadcasting distortion region , '' in _ proc .
ieee international symposium on information theory _ , seoul , korea , jul .
2009 , pp 2477 - 2482 .
r. soundararajan and s. vishwanath , `` hybrid coding for gaussian broadcast channels with gaussian sources , '' in _ proc .
ieee international symposium on information theory _ , seoul , korea , jul .
2009 , pp .
2790 - 2794 . | we provide a complete characterization of the achievable distortion region for the problem of sending a bivariate gaussian source over bandwidth - matched gaussian broadcast channels , where each receiver is interested in only one component of the source .
this setting naturally generalizes the simple single gaussian source bandwidth - matched broadcast problem for which the uncoded scheme is known to be optimal .
we show that a hybrid scheme can achieve the optimum for the bivariate case , but neither an uncoded scheme alone nor a separation - based scheme alone is sufficient .
we further show that in this joint source channel coding setting , the gaussian setting is the worst scenario among the sources and channel noises with the same covariances .
gaussian source , joint source - channel coding , squared error distortion measure . |
more than 50 years ago john von neumann traced first parallels between architecture of the computer and the brain @xcite . since that time computers became an unavoidable element of the modern society forming a computer network connected by the world wide web ( www ) .
the www demonstrates a continuous growth approaching to @xmath2 web pages spread all over the world ( see e.g. http://www.worldwidewebsize.com/ ) .
this number starts to become even larger than @xmath3 neurons in the brain .
each neuron can be viewed as an independent processing unit connected with about @xmath4 other neurons by synaptic links ( see e.g. @xcite ) .
about 20% of these links are unidirectional @xcite and hence the brain can be viewed as a directed network of neuron links . at present , more and more experimental information about neurons and their links becomes available and the investigation of properties of neuronal networks attracts an active interest of many groups ( see e.g. @xcite .
the www is also a directed network where a site @xmath5 points to a site @xmath6 but no necessary vice versa .
the classification of web sites and information retrieval from such an enormous data base as the www becomes a formidable challenge of modern society where the search engines like google are used by internet users in everyday life . an efficient way to classify and extract the information from www
is based on the pagerank algorithm ( pra ) , proposed by brin and page in 1998 @xcite , which forms the basis of the google search engine .
the pra is based on the construction of the google matrix which can be written as ( see e.g. @xcite for details ) : @xmath7 here the matrix @xmath8 is constructed from the adjacency matrix @xmath9 of directed network links between @xmath10 nodes so that @xmath11 and the elements of columns with only zero elements are replaced by @xmath12 .
the second term in r.h.s . of ( [ eq1 ] ) describes a finite probability @xmath13 for www surfer to jump at random to any node so that the matrix elements @xmath14 .
this term with the google damping factor @xmath1 stabilizes the convergence of pra introducing a gap between the maximal eigenvalue @xmath15 and other eigenvalues @xmath16 . as a result
the first eigenvalue has @xmath17 and the second one has @xmath18 .
usually the google search uses the value @xmath19 @xcite . by the construction @xmath20 so that the asymmetric matrix @xmath0 belongs to the class of perron - frobenium operators which naturally appear in the ergodic theory @xcite and dynamical systems with hamiltonian or dissipative dynamics @xcite .
the right eigenvector at @xmath15 is the pagerank vector with positive elements @xmath21 and @xmath22 .
the classification of nodes in the decreasing order of @xmath21 values is used to classify importance of www nodes as it is described in more detail in @xcite .
the pagerank can be efficiently obtained by a multiplication of a random vector by @xmath0 which is of low cost since in average there are only about ten nonzero elements in a typical line of @xmath0 of www .
this procedure converges rapidly to the pagerank .
fundamental investigations of the pagerank properties of the www have been performed in the computer science community ( see e.g. @xcite ; involvement of physicists is visible , e.g. @xcite , but less pronounced ) .
it was established that the pagerank is satisfactory characterized by an algebraic decay @xmath23 with @xmath5 being the ordering index and @xmath24 ; the number of nodes with the pagerank @xmath25 scales as @xmath26 with the numerical value of the exponent @xmath27 @xcite .
it is known that such type of algebraic dependencies appear in various types of scale - free networks @xcite .
the pagerank classification finds its applications not only for the www but also for the network of article citations in physical review as it is described in @xcite .
this shows that the approach based on the google matrix can be applied to vary different types of networks . in this work
we construct the google matrix @xmath0 for a model of brain analyzed in @xcite .
the properties of the spectrum and the eigenstates of @xmath0 are described in the next section ii .
the results are discussed in section iii .
= 8.5 cm -1.0 cm = 8.5 cm -0.3 cm to construct the google matrix of brain we use a directed network of links between @xmath28 neurons @xcite generated from the brain model @xcite . in total there are @xmath29 links in the network .
they form @xmath30 _ outgoing _ links and @xmath31 _ ingoing _ links ( @xmath32 ) , so that there are about 200 outlinks ( or ingoing ) per neuron .
these numbers include multiple links between certain pairs of neurons ; certain neurons have also links to themselves ( there is one neuron linked only to itself ) .
the number of weighted symmetric links is approximately @xmath33% . due to existence of multiple links between the same neurons we constructed two @xmath0 matrices based on unweighted and weighted counting of links . in the first case
all links from neuron @xmath5 to neuron @xmath6 are counted as one link , in the second case the weight of the link is proportional to the number of links from @xmath5 to @xmath6 . in both cases
the sum of elements in one column is normalized to unity .
the distributions of ingoing ( @xmath34 ) and outgoing ( @xmath35 ) links are shown in fig .
[ fig1 ] . the weighted distribution of ingoing links have a pronounced peaked structure corresponding to different regions of brain model considered in @xcite .
we note that the distribution of links is not of free - scale type .
= 4.5 cm = 4.5 cm -0.2 cm = 4.5 cm = 4.5 cm -0.3 cm the dependence of the pagerank on @xmath1 is shown in fig . [ fig2 ] . for @xmath36 almost all probability @xmath21 is concentrated on one neuron .
this is the only one neuron which is linked only to itself . with the increase of @xmath1 up to @xmath37 the main part of probability
is concentrated mainly on about 10 neurons that approximately corresponds to the number of peaks in the distribution of weighted ingoing links in fig .
[ fig1 ] ( bottom left panel ) . at the same time
the pagerank has a long tail at large @xmath5 where the probability @xmath21 is practically homogeneous .
for @xmath38 the peak of probability at @xmath39 is washed out and the pagerank becomes completely delocalized .
we note that a delocalization of the pagerank with @xmath1 appears in the ulam networks describing dynamical systems with dissipation @xcite . at the same time
the www networks remain stable in respect to variation of @xmath1 as it is discussed in @xcite .
recently , for the studies of procedure call network of the linux kernel @xcite it was proposed to study the properties of the importance - pagerank @xmath40 which is given by the eigenvector at @xmath15 for the google matrix constructed from the inverted links of the original adjacency matrix . it was argued that @xmath40 can give an additional information about certain important nodes .
our results for @xmath40 are shown in panels ( c , d ) of fig .
they show that @xmath40 is practically delocalized and flat for all used values of @xmath1 .
this indicates that all nodes have practically equal importance .
the popularity - importance correlator introduced in @xcite and defined as @xmath41 is rather small ( @xmath42 at @xmath38 and @xmath43 at @xmath19 for unweighted , weighted links respectively ) .
this shows that there are no correlations between @xmath25 and @xmath44 in our neuronal network that is similar to the linux kernel case .
= 4.5 cm = 4.5 cm -1.2 cm = 4.5 cm = 4.5 cm -0.3 cm the spectrum @xmath16 and the right eigenvectors @xmath45 of the google matrix of brain are defined by the equation @xmath46 the spectrum of @xmath47 is complex and is shown in fig .
the color of points is chosen to be proportional to the participation ratio ( par ) defined as @xmath48 .
this quantity determines an effective number of sites populated by an eigenstate @xmath45 , it is often used to characterize localization - delocalization transition in quantum solid - state systems with disorder ( see e.g. @xcite ) .
the spectrum has eigenvalues with @xmath49 being close to unity so that there is no gap in the spectrum of @xmath47 in the vicinity of @xmath15 ( we remind that the second term in the r.h.s . of ( [ eq1 ] ) transfers
@xmath16 to @xmath50 keeping only one @xmath17 @xcite ) .
this is different from the spectrum of random scale - free networks characterized by a large gap in the spectrum of @xmath47 @xcite .
= 7.5 cm -0.3 cm compared to the spectra of the university www networks studied in @xcite the spectrum of @xmath0 in fig .
[ fig3 ] is more flat being significantly compressed to the real axis . in this respect
our neuronal network has certain similarity with the spectra of vocabulary networks analyzed in @xcite ( see fig . 1 there ) . at the same time the spectrum of @xmath0 matrix of brain
has visible structures in the eigenvalues distribution in the complex plane of @xmath47 while the vocabulary networks are characterized by structureless spectrum .
the spectrum of fig .
[ fig3 ] has global properties being similar to those of the ulam networks considered in @xcite .
it is interesting to note that the spectra of unweighted and weighted networks of brain have similar structure .
this supports the view of structural stability of the spectrum of @xmath0 matrix . it is useful to determine the relaxation rate of eigestates by the relation @xmath51 .
the dependence of density of states @xmath52 on @xmath53 is shown in fig .
[ fig4 ] ( the density is normalized to unity so that @xmath54 corresponds to @xmath28 states ) .
the distribution in @xmath53 has a pronounced peak at @xmath55 , the density of states at small @xmath56 is relatively small ( this is also seeing in fig . [ fig3 ] ) .
the comparison of unweighted and weighted links shows the stability of the density distribution in respect to such modification of links .
= 4.5 cm = 4.5 cm -0.3 cm the dependence of the par @xmath57 on @xmath53 is shown in fig .
[ fig5 ] ( we note that except of the pagerank @xmath57 is independent of @xmath1 due to the unity rank of matrix @xmath58 , see e.g. @xcite ) .
the pagerank value of @xmath57 at @xmath59 is very large being more than half of the total number of neurons @xmath28 .
it is clear that this corresponds to a delocalized state .
the eigenstates with @xmath60 have relatively small @xmath61 being close to a localized domain while eigenstates with @xmath62 have @xmath63 being delocalized on the main part of the network ; the states with @xmath64 enter in the localized domain .
for @xmath65 the par is close to @xmath66 .
taking as a criterion that the delocalization takes place when @xmath67 we obtain that the pagerank becomes delocalized at @xmath68 ( see data of figs .
[ fig2],[fig6 ] ) .
the global dependence of the par @xmath57 of the pagerank on parameter @xmath1 is shown in fig .
[ fig6 ] with a sharp delocalization of @xmath57 for @xmath69 .
of course , the above analysis should be considered as an approximate one since the localization properties should be studied in dependence on the system size @xmath10 while we consider only one size of @xmath10 .
= 4.5 cm = 4.5 cm -0.3 cm = 4.5 cm = 4.5 cm -0.3 cm finally , following the approach proposed in @xcite , we show in fig . [ fig7 ] the distribution of pagerank values @xmath25 and @xmath44 for all sites
. such kind of distributions can be rather useful in determining sites which have maximal values of @xmath25 and @xmath44 at the same time .
however , a detailed analysis of the properties of this distribution would require networks with a larger size @xmath10 where statistical fluctuations are smaller .
in this work we studied the properties of the google matrix of a neuronal network of the brain model discussed in @xcite . for this network of @xmath4 neurons we found that the spectrum of the google matrix has a gapless spectrum at @xmath70 demonstrating certain similarities with the spectra of university www networks and vocabulary networks studied in @xcite . at the same time our neuronal network shows signs of delocalization transition of the pagerank at the google damping factor @xmath68 which was absent in the networks studied in @xcite .
a similar transition in @xmath1 was detected in the ulam networks generated by dissipative dynamical maps @xcite .
we attribute the appearance of such delocalization transition to a large number of links per neuron ( 200 ) which is by factor 10 larger than in the www networks ( 20 ) . of course , our studies have certain limitations since we considered only a fixed size neuronal network and since this network is taken from a model system of brain analysed in @xcite .
another weak point is that we do not consider the dynamical properties of the network which are probably more important for practical applications .
nevertheless , the spectral properties of @xmath0 matrix can be rather useful .
indeed , the gapless spectrum of @xmath47 shows that long living excitations can exist in our neuronal network .
such relaxation modes with small rates @xmath53 can be the origin of long living oscillations found in numerical simulations @xcite .
it is quite possible that the properties of spectra of @xmath0 can help to understand in a better way rapid relaxation processes and those with long relaxation times .
we conjecture that the rapid relaxation modes correspond to relaxation of local groups of neurons while long living modes can represent relaxation of collective modes representing dynamics of human thoughts .
the dynamics of such collective modes can contain significant elements of chaotic dynamics as it was discussed in the frame of the concept of creating chaos in @xcite .
it is possible that the brain effectively implements dynamics described by the evolution equation @xmath71 which without perturbations converges to the steady - state described by the pagerank ( which may be linked with a sleeping phase ) .
external perturbations give excitations of other eigenmodes of @xmath0 discussed here .
the evolution of these excitations will be significantly affected by the spectrum of @xmath0 .
further development of the google matrix approach to the brain looks to us to be rather promising .
for example , a detection of isolated communities and personalized pagerank , represented by other types of matrix @xmath58 in ( [ eq1 ] ) , is under active investigation in the computer science community ( see e.g. @xcite ) .
such type of problems can find their applications for detection of specific quasi - isolated neuronal networks of brain .
the usage of real neuronal networks , similar to those studied in @xcite , in combination with the google matrix approach can allow to discover new properties of processes in the brain .
the development of parallels between the www and neuronal networks will give new progress of the ideas of john von neumann .
99 j. von neumann , _ the computer and the brain _ , new haven , ct , yale uiv . press ( 1958 ) .
hoppensteadt and e.m .
izhikevich , _ weakly connected neural networks _ , springer - verlag , n.y .
izhikevich , _ dynamical systems in neuroscience : the geometry of excitability and bursting _ , the mit press , cambridge , ma ( 2007 ) .
o. sporns , _ brain connectivity _ , scholarpedia 2(10 ) ( 2007 ) 4695 . d.j . felleman and d.c .
van essen , cereb .
cortex 1 ( 1991 ) 1 .
laughlin and t.j .
sejnowski , science 301 ( 2003 ) 1870 .
o. sporns , d.r .
chialvo , m. kaiser , and c.c .
hilgetag , trends cognitive sci .
8 , ( 2004 ) 418 .
honey , r. ktter , m. breakspear , and o.sporns , pnas 104 ( 2007 ) 10240 .
m. kaiser , phil .
r. soc . a 365 ( 2007 ) 3033 .
p. hagmann , l. cammoun , x. gigandet , r. meuli , c.j .
honey , v.j .
weeden , and o.sporns , plos biology 6 ( 2008 ) 1479 .
izhikevich and g.m .
edelman , pnas 105 ( 2008 ) 3593 .
q. wen , a. stepanyants , g.n .
elston , a.y .
grosberg , and d.b .
chklovskii , pnas 106 ( 2009 ) 12536 .
s. cocco , s. leibler , and r. monasson , pnas 106 ( 2009 ) 14058 . s. brin and l. page , computer networks and isdn systems * 30 * ( 1998 ) 107 .
a. m. langville and c. d. meyer , _ google s pagerank and beyond : the science of search engine rankings _ , princeton university press ( princeton , 2006 ) ; d. austin , ams feature columns ( 2008 ) available at www.ams.org/featurecolumn/archive/pagerank.html i.p .
cornfeld , s.v .
fomin , and y. g. sinai , _ ergodic theory _ , springer ,
m. brin and g. stuck , _ introduction to dynamical systems _
, cambridge univ . press , cambridge ,
uk ( 2002 ) .
d. donato , l. laura , s. leonardi and s. millozzi , eur .
j. b 38 ( 2004 ) 239 ; g. pandurangan , p. raghavan and e. upfal , internet math . 3 ( 2005 ) 1 .
p. boldi , m. santini , and s. vigna , in _ proceedings of the 14th international conference on world wide web _ , a. ellis and t. hagino ( eds . ) , acm press , new york p.557 ( 2005 ) ; s. vigna , * ibid . * p.976 .
k. avrachenkov and d. lebedev , internet math . 3 ( 2006 ) 207 .
k. avrachenkov , n. litvak , and k.s .
pham , in _ algorithms and models for the web - graph : 5th international workshop , waw 2007 san diego , ca , proceedings _ , a. bonato and f.r.k .
chung ( eds . ) , springer - verlag , berlin , lecture notes computer sci . 4863 ( 2007 ) 16 .
n. litvak , w. r. w. scheinhardt , and y. volkovich , internet math . 4 ( 2007 ) 175 .
k. avrachenkov , d. donato and n. litvak ( eds . ) , _ algorithms and models for the web - graph : 6th international workshop , waw 2009 barcelona , proceedings _ , springer - verlag , berlin , lecture notes computer sci .
5427 , springer , berlin ( 2009 ) .
n. perra and s. fortunato , phys . rev .
e 78 ( 2008 ) 036107 .
dorogovtsev and j.f.f .
mendes , _ evolution of networks _ , oxford univ .
press , oxford ( 2003 ) .
p. chen , h. xie , s. maslov , and s. redner , j. infometrics 1 ( 2007 ) 8 .
f. radicchi , s. fortunato , b. markines , and a. vespignani , phys .
e * 80 * ( 2009 ) 056103 .
e.m izhikevich , private communication , august ( 2009 ) : the links between neurons have been generated by e.m .
izhikevich on the basis of the brain model described in @xcite ; the links are available at quantware library , k. frahm and d.shepelyansky ( eds . ) section qnr15 at www.quantware.ups-tlse.fr/qwlib d.l .
shepelyansky and o.v.zhirov , phys . rev .
e 81 ( 2010 ) 036213 .
l. ermann and d.l .
shepelyansky , phys .
e 81 ( 2010 ) 036221 .
b. georgeot , o. giraud and d.l .
shepelyansky , preprint arxiv:1002.3342[cs.ir ] ( 2010 ) .
chepelianskii , arxiv:1003.5455[cs.se ] ( 2010 ) .
f. evers and a.d .
mirlin , rev .
mod . phys .
80 ( 2008 ) 1355 .
o. giraud , b. georgeot and d. l. shepelyansky , phys .
e 80 ( 2009 ) 026107 .
b.v.chirikov , _ creating chaos and the life _ , preprint arxiv : physics/0503072 ( 2005 ) . | we apply the approach of the google matrix , used in computer science and world wide web , to description of properties of neuronal networks .
the google matrix @xmath0 is constructed on the basis of neuronal network of a brain model discussed in pnas * 105 * , 3593 ( 2008 ) .
we show that the spectrum of eigenvalues of @xmath0 has a gapless structure with long living relaxation modes .
the pagerank of the network becomes delocalized for certain values of the google damping factor @xmath1 .
the properties of other eigenstates are also analyzed .
we discuss further parallels and similarities between the world wide web and neuronal networks .
* keywords : * neuronal networks , world wide web ; google matrix ; pagerank |
the theory of weyl semimetals semimetals where conduction and valence bands have non - degenerate touching points has a long and intriguing history , bringing ideas originally developed in the context of particle physics into the realm of condensed matter and materials physics.@xcite recently the experimental discovery of a weyl semimetal was reported by various groups,@xcite soon followed by transport measurements demonstrating the chiral anomaly.@xcite ) ; the right panel shows a dispersion tilted in the negative @xmath0 direction with tilt parameter @xmath1.,title="fig:",scaledwidth=23.0% ] ) ; the right panel shows a dispersion tilted in the negative @xmath0 direction with tilt parameter @xmath1.,title="fig:",scaledwidth=23.0% ] the conventional understanding of a weyl semimetal is a system that , in the vicinity of the band touching points , is accurately described by the two - band weyl hamiltonian @xmath2 whose elegant form is dictated by lorentz invariance and the requirement of a linear dispersion of the crossing bands . here ,
momentum and energy are measured relative to the momentum and energy of the band crossing point , and @xmath3 is the pauli matrix .
however , as was only appreciated recently , symmetry breaking terms such as an anisotropic velocity and a `` tilt '' of the weyl dispersion ( see fig . [
fig : illustration ] ) occur generically in weyl materials and may have profound consequences for thermodynamic and transport properties.@xcite explicitly , a generic linear band crossing is described by @xmath4 where the @xmath5 now describe an anisotropic velocity , the @xmath6 represent a uniform linear tilting of both bands , and @xmath7 is the @xmath8 unit matrix . while the tilt looks most innocent being proportional to the unit matrix @xmath7 it does not alter the eigenstates of the model it turns out to be the more interesting symmetry - breaking term . whereas tilting terms have been considered early on in two - dimensional systems as consequences of perturbations,@xcite the topological stability of weyl points,@xcite or , viewed alternatively , the generic nature of non - degenerate band crossings@xcite in three dimensions allows for much stronger tilting .
in fact , it was first shown in simple toy models,@xcite and later in more realistic materials simulations@xcite and in models of superfluids,@xcite that the weyl dispersion can easily be tilted over to the extent that a finite fermi surface is formed , with hole and electron pockets touching at the weyl point .
these systems have been called `` type - ii weyl semimetals '' and have , following their prediction in wte@xmath9 , now been observed in a number of materials.@xcite the effect of smaller , sub - critical tilts , which preserve the point - like nature of the fermi surface , is more of a quantitative than of a qualitative nature , although even small tilts influence transport properties of weyl semimetals in a unique way . when the fermi level lies exactly at the nodal point ,
a weyl semimetal of size @xmath10 has a vanishing conductivity @xmath11 , but a non - vanishing @xmath12-independent conductance @xmath13.@xcite the finite conductance is reminiscent of the universal minimal conductance @xmath14 predicted and observed in graphene,@xcite with the important distinction that in two dimensions a finite minimal conductance corresponds to a non - zero conductivity due to its different scaling with system size ( @xmath15 in @xmath16 dimensions ) . as shown in ref .
, the value of the minimal conductance @xmath17 in a weyl semimetal depends on the tilt .
moreover , tilt was also found to affect the value of the `` fano factor '' @xmath18 at the weyl point ( the ratio of shot noise power @xmath19 to current @xmath20 @xcite ) , which was otherwise found to be a universal number @xmath21,@xcite independent of weak disorder , sample geometry , or anisotropy in the weyl dispersion . in this work
we consider the combined effect of a ( sub - critical ) tilt of the weyl dispersion and the presence of potential disorder , which is the second important addition to the weyl hamiltonian ( [ h ] ) required for the description of realistic condensed - matter realizations of a weyl semimetal . for isotropic weyl cones the remarkable properties of the ideal system in addition to the anomalous transport properties described above these include the vanishing density of states at the weyl point are known to persist up to a finite disorder strength , beyond which a diffusive phase with nonzero density of states at the weyl point and nonzero conductivity sets in.@xcite the fano factor @xmath22 takes a different , but again universal value @xmath23 in the diffusive phase .
our main finding , which we support with a combination of analytical and numerical arguments , is that the main effect of tilting the weyl dispersion is to reduce the critical disorder strength @xmath24 for the transition between the `` pseudo - ballistic '' weak - disorder phase and the diffusive strong - disorder phase .
such a result can be expected intuitively , based on the consideration that in the weak - disorder phase there is a vanishing density of states at the weyl point , and disorder acts by virtual excitations to states of higher energy , where the range of reachable energies is determined by the disorder strength.@xcite tilting the weyl dispersion lowers the energy of some states while raising the energy of others . the net effect , however , is that there are more states available in a given energy range than in the isotropic cone , see fig . [ fig : illustration ] .
explicitly , the density of states @xmath25 depends on the tilt parameter @xmath26 as @xmath27 where @xmath28 is the density of states without tilt .
this implies that the disorder of the same strength @xmath29 can be expected to have a stronger influence on a tilted weyl dispersion , and it is natural to expect a lower critical disorder strength @xmath30 for cones with larger tilt . upon approaching the critical tilt @xmath31
the critical disorder strength @xmath24 goes to zero again , not a surprise given the diverging density of states for @xmath32 .
this implies , however , that in the presence of disorder the tilt - driven transition between type - i and type - ii weyl semimetals will be preceded ( and , hence , masked ) by the tilt - driven transition from the pseudo - ballistic weak - disorder phase into the diffusive strong - disorder phase .
the remainder of this work is organized as follows : in sec .
[ sec : setup ] we introduce to the specific model employed throughout this work , which contains a tilted dispersion as well as a random potential . in sec .
[ sec : scba ] we employ the self - consistent born approximation to show how a tilted dispersion affects the disorder - induced phase transition .
we then corroborate these results by two distinct numerical approaches : we use the kernel polynomial method@xcite to compute the density of states in a tight - binding model using the same disorder type as in the scba , see sec .
[ sec : kpm ] , and we study transport properties in finite size systems using a scattering matrix approach , see sec . [ sec : transport ] .
the use of numerical methods to confirm our conclusions is necessary , since the self - consistent born approximation is known to be an uncontrolled approximation for the disorder - induced phase transition in weyl semimetals.@xcite we conclude in sec .
[ sec : discussion ] .
we consider a single disordered weyl node described by the hamiltonian @xmath33 where without loss of generality we have chosen the tilt to be in the @xmath34 direction .
momentum and energy are measured with respect to the weyl node .
we have chosen units such that @xmath35 .
the critical tilt is @xmath31 , and we will consider sub - critical tilts @xmath36 only . the disorder potential @xmath37 is taken to be a gaussian random potential with zero average and with two - point correlation function @xmath38 for a finite system of size @xmath10 . here
@xmath39 is the correlation length of the disorder potential and @xmath29 is the dimensionless disorder strength .
the restriction to a single weyl node requires that the correlation length @xmath39 is much larger than the inverse distance between the weyl nodes .
we have chosen not to include anisotropy of the weyl dispersion in to the hamiltonian ( [ loweh ] ) [ compare with eq .
( [ general_h ] ) ] .
although anisotropies are as ubiquitous in real materials as the tilts , they can largely be understood by simple means of rescaling of the coordinate axes , rendering the weyl dispersion isotropic but the disorder anisotropic.@xcite through such a rescaling procedure , anisotropies were found to affect the conductance @xmath17 ( since @xmath17 depends on sample geometry ) , but not the fano factor @xmath22 .
tilts , on the other hand , can not be removed by the rescaling procedure , and were found to affect both @xmath17 and @xmath22 in the absence of disorder.@xcite in the following three sections we calculate the critical disorder strength @xmath24 for the disorder - induced transition to a diffusive phase using three different methods .
section [ sec : scba ] employs the self - consistent born approximation and addresses the density of states , using the model described above .
section [ sec : kpm ] uses the kernel - polynomial method to calculate the density of states for a tight - binding model for which the continuum model ( [ loweh ] ) is the low - energy limit .
section [ sec : transport ] considers transport properties of the continuum model ( [ loweh ] ) , but for a system of size @xmath40 , where the sample width @xmath41 ( transverse to the transport direction ) is chosen much larger than its length @xmath12 ( in the transport direction ) to ensure that the transport properties do not depend on the choice of the boundary conditions in the direction transverse to the current flow .
is shown as errorbars ( see appendix for a discussion ) .
, scaledwidth=48.0% ] to study the density of states of the model ( [ loweh ] ) we employ the self - consistent born approximation ( scba ) .
we closely follow a similar calculation of ominato and koshino for a weyl dispersion without tilt.@xcite before we turn to a description of our calculations , we note that the scba relies on a diagrammatic expansion that is known to neglect important contributions . for a weyl dispersion without tilt , the scba is known to yield critical disorder strengths @xmath42 , which are larger by a factor @xmath43 than the ones obtained by more precise numerical simulations.@xcite we nevertheless see the scba as a useful approximation for a qualitative understanding of the way in which a tilted weyl dispersion affects the critical disorder strength .
furthermore , the scba allows us us to access the disorder - induced renormalizations of tilt and fermi velocity .
the disorder - averaged density of states @xmath25 is expressed in terms of the ( @xmath8 matrix ) green function as @xmath44 where the brackets @xmath45 indicate the disorder average . in the scba the disorder - averaged green function is expressed as @xmath46 with the scba approximation for the self - energy @xmath47 where the disorder average @xmath48 is given by .
the self - consistent integral equations and are solved numerically .
details on the numerical procedure are given in appendix [ scba - appendix ] .
the critical disorder strength @xmath24 is found as that disorder strength for which the density of states @xmath49 at the weyl point becomes finite , whereas @xmath50 for @xmath51 .
our main results are summarized in fig .
[ fig : weyl - tilted ] . vs. dimensionless disorder strength @xmath29 in the weak - disorder phase @xmath51.,scaledwidth=48.0% ] the scba phase diagram of fig .
[ fig : weyl - tilted ] confirms the intuitive picture of the introduction : the critical disorder strength decreases with increasing tilt and approaches @xmath52 when the tilt strength @xmath26 approaches the critical value @xmath31 .
this is consistent , since as a finite density of states develops when the cone tips over ( @xmath53 ) , we expect a diffusive phase at arbitrarily small disorder strengths for super - critical tilt .
the scba not only allows us to find the tilt - dependence of the critical disorder strength @xmath24 , the expansion of the self energy @xmath54 for small momenta at zero energy also allows us to find a disorder - renormalized fermi velocity @xmath55 and tilt @xmath56 . taking into account rotation invariance around the @xmath34 axis
, the small-@xmath57 expansion of the self energy reads @xmath58 from which we obtain @xmath59 our result for the renormalized fermi energy @xmath55 is consistent with that of ref . and will not be discussed further here .
the renormalized dimensionless tilt @xmath56 is shown in fig .
[ fig : effective - tilt ] , as a function of disorder strength @xmath29 .
figure [ fig : effective - tilt ] suggests that there is not only the single fixed point of the clean and isotropic weyl cone ( in the renormalization - group sense ) , but rather a continuous family of fixed points at no disorder which are distinguished by their tilt .
also note that while disorder @xmath51 leads to an increased dimensionless tilt , in the scba approximation sub - critical disorder can not lead to an `` over - tilting '' of the weyl dispersion . in other words , the disorder - induced phase transition to a diffusive phase with a finite density of states at the fermi level always takes place _ before _ the tilt - induced transition to a type - ii weyl semimetal , which also has a finite density of states at the nodal point .
( this is consistent with the tilt - dependence of the critical disorder strength shown in fig .
[ fig : weyl - tilted ] , which shows that the critical disorder strength approaches zero in the limit @xmath32 . )
this observation remains valid if the disorder - induced renormalization of the tilt is included , see fig . [
fig : effective - tilt ] .
to complement the scba analysis we now report a numerical calculation of the density of states , using the kernel polynomial method ( kpm).@xcite we consider the two - band lattice model@xcite @xmath60 \label{eq : h0}\end{aligned}\ ] ] where @xmath61 is the lattice constant .
the model ( [ eq : h0 ] ) features eight weyl points at momenta @xmath62 for @xmath63 with @xmath64 , @xmath65 , @xmath66 .
we add a disorder potential as in eq .
, with correlation length @xmath67 , which ensures that ( i ) the smooth disorder correlations are well represented on the discrete lattice and ( ii ) the inter - node scattering rate is suppressed as compared to the intra - node rate by a factor of order @xmath68<10^{-19}$ ] , such that the physics can be regarded as effectively single - node .
top : density of states @xmath25 as a function of energy @xmath69 without tilt ( left , @xmath70 ) and with tilt ( right , @xmath71 ) , as calculated by the kernel polynomial method for disorder strengths @xmath72 ( bottom to top ) .
the dashed lines denote the scba results , eq . .
[ for @xmath73 the scba coincides with the zero - disorder density of states of eq . .
] bottom : density of states @xmath49 at the nodal point versus disorder strength for different values of the tilt , @xmath74 ( right to left ) .
the calculations were carried out for an average over @xmath75 realizations of the random potential , for a cubic lattice of size @xmath76 , with @xmath77 random vectors for calculating the trace in the kpm and an expansion order of roughly 1000 .
the density of states is normalized to a single weyl node .
, scaledwidth=48.0% ] in fig .
[ fig : kpm ] , top panel , we show the kpm results for the density of states @xmath78 without tilt ( left panel , solid curves ) and with dimensionless tilt @xmath79 ( right ) .
we also show the predictions of the scba ( dashed curves ) . for weak disorder and away from zero energy , the kpm and scba results are in good agreement , but the comparison with the numerical results also shows that the scba overestimates the critical disorder strength @xmath24 above which the zero - energy density of states becomes finite .
the figure suggests , however , that this overestimation happens equally for zero tilt and for finite tilt , so that the trend predicted by the scba is indeed confirmed by the kpm . the bottom panel of fig .
[ fig : kpm ] shows the density of states @xmath49 at the nodal point as a function of the disorder strength for different values of the tilt parameter .
again , these results confirm the trend predicted by the scba , that larger tilt corresponds to a lower critical disorder strength .
however , a precise determination of @xmath80 is beyond the capabilities of the kpm method.@xcite
the disorder - induced phase transition is not only characterized by its effect on the density of states at the nodal point @xmath25 , which is zero for @xmath51 and finite for @xmath81 , the transition also strongly affects transport properties , such as the conductance or shot noise power . in this section
we show that the characteristic transport properties at the nodal point found for a weyl dispersion without tilt persist for a tilted dispersion , and we provide more evidence in support of the result of the previous sections , that the critical disorder strength is reduced in the presence of a tilted weyl dispersion .
we calculate the transmission matrix @xmath82 of a finite sample of length @xmath12 in the transport direction ( which we take to be the @xmath0-direction ) and width @xmath41 in the transversal ( @xmath83 and @xmath34 ) directions .
the method is explained in detail in ref . .
with a width @xmath41 the two - point correlation function ( [ correlator - finite - size ] ) of the disorder potential is replaced by @xmath84the conductance @xmath85 per weyl node and the fano factor @xmath22 are expressed in terms of @xmath82 as @xmath86 where the trace is taken with respect to the transverse momenta . to approach bulk results as closely as possible
, we choose the sample width @xmath87 and verify that the conductance is proportional to @xmath88 and is independent of the boundary conditions chosen in the transverse direction .
we present results for the dimensionless conductance rescaled for a cubic sample of size @xmath10 , @xmath89 .
the conductivity @xmath11 is calculated from the standard relation @xmath90 . the conductance @xmath17 and fano factor @xmath22 without disorder , but with a tilted dispersion , were calculated in ref . .
, rescaled for a system of size @xmath10 ( top ) and fano factor @xmath22 for sub - critical disorder strength @xmath91 and for tilt parameter @xmath92 , @xmath93 , and @xmath94 ( bottom to top ) .
the data points show results from numerical calculations , as described in the text .
the solid and dashed line indicate the values expected without disorder and for a clean system with disorder - renormalized tilt @xmath56 , respectively . we have not shown data for @xmath95 , which are dominated by finite - size effects .
the numerical curves are based on an average over at least @xmath96 disorder realizations .
the error bars show the residual statistical error .
, scaledwidth=50.0% ] figure [ fig : smatrix - result ] shows the dimensionless conductance @xmath17 and the fano factor versus system size @xmath12 for disorder strength @xmath91 and for the tilt parameter @xmath92 , @xmath93 , and @xmath94 .
the disorder strength @xmath91 is sub - critical for all three values of @xmath26 considered .
the figure also shows the conductance and fano factor expected for a clean weyl semimetal and for a clean weyl semimetal with scba - renormalized tilt @xmath56 .
especially the conductance data still show a considerable size dependence , even for the largest system sizes we could reach .
nevertheless , the data leave no doubt that the conductance remains bounded as a function of @xmath12 , indicating that the conductivity @xmath90 is zero in the thermodynamic limit , which implies that the transport characteristics of the quasi - ballistic weak - disorder phase persist in the present of a tilted weyl dispersion .
moreover , the finite - size data also show that the conductance increases with increasing tilt , consistent with the analysis of the clean limit in ref . and with the expectation that tilt drives the system closer to the disorder - induced phase transition .
however , the finite - size effects are too large to permit a more quantitative analysis . in particular ,
the finite - size effects are too large to quantitatively confirm or disprove the scba expectation for the conductance . as a function of system size @xmath12 for disorder strength @xmath97 and tilt parameters @xmath92 , @xmath93 , @xmath94 , and @xmath98 ( bottom to top ) .
the data points show results from numerical calculations after averaging over at least @xmath96 disorder realizations .
the error bars show the residual statistical error .
we have not shown data for @xmath95 , which are dominated by finite - size effects .
the conductance is bounded for @xmath70 and @xmath99 ( bottom two data sets ) , which is characteristic of the quasi - ballistic weak - disorder phase . for @xmath71 and @xmath100
the conductance is proportional to the system size @xmath12 for large @xmath12 ( top two data sets ) , characteristic of the diffusive strong - disorder phase .
the dashed lines indicate a linear increase with @xmath12 and are shown as a guide to the eye .
, scaledwidth=31.25% ] figure [ fig : smatrix - result - medium ] shows the dimensionless conductance @xmath17 versus system size for disorder strength @xmath97 , chosen such that the disorder is sub - critical without tilt , but above critical with large tilt .
the progression of the curves shown in the figure corresponds to the horizontal arrow in fig .
[ fig : weyl - tilted ] ( although the value of the disorder strength @xmath29 is less than in fig .
[ fig : weyl - tilted ] , reflecting the overestimation of the critical disorder strength in the scba ) . as in the previous figure , finite - size effects are considerable , nevertheless the asymptotic dependences characteristic of the weak - disorder phase ( @xmath17 vs. @xmath12 bounded ) and of the strong - disorder phase ( @xmath101 for large @xmath12 ) are clearly visible for small and large tilt , respectively .
figure [ fig : smatrix - result - medium ] thus illustrates how variation of the tilt can be used to scan through the disorder - induced phase transition between the quasi - ballistic weak - disorder phase and the diffusive strong - disorder phase .
the fano factor @xmath22 of the finite samples is subject to large fluctuations near the critical disorder strength and data is not shown . and
fano factor @xmath22 as a function of system size @xmath12 for disorder strength @xmath102 and tilt parameters @xmath92 , @xmath93 , and @xmath94 ( red , blue , green , i.e. top to bottom in the left panel ) .
the data points show results from numerical calculations after averaging over at least @xmath96 disorder realizations .
the error bars show the residual statistical error .
data for @xmath95 are dominated by finite - size effects and are not shown .
the dashed black line in the right panel indicates the fano factor @xmath103 expected in the diffusive regime.,scaledwidth=50.0% ] the case of very strong disorder is shown in fig .
[ fig : smatrix - result - strong ] , which shows conductance and fano factor for @xmath104 , which is well above the critical value regardless of the tilt @xmath26 .
for all values of the tilt the conductance increases linearly with @xmath12 , consistent with a finite conductivity @xmath90 in the thermodynamic limit .
up to small deviations ( which we can not explain , but which are ubiquitous in transport calculations ) the fano factor is at the value @xmath103 appropriate for the diffusive phase . in any case ,
our numerical calculations show that both conductance and fano factor can be used to distinguish the weak - disorder and strong - disorder phases . in fact , in the presence of tilt the difference between the fano factor for the weak - disorder and strong - disorder phases is even larger than without tilt , see fig .
[ fig : smatrix - result ] .
in this work we have studied how the combination of potential disorder and a tilted weyl dispersion affects the zero - energy density of states and transport properties of a weyl semimetal .
our main conclusions , supported via calculations of the density of states using the self - consistent born approximation ( scba ) , calculations of the density of states for a tight - binding model using the kernel polynomial method ( kpm ) , and calculations of the conductance and fano factor using the scattering approach , are that ( 1 ) the existence of a disorder - induced phase transition between a quasi - ballistic weak - disorder phase with zero density of states and finite conductance at the nodal point and a diffusive strong - disorder phase characterized by a finite conductivity and a finite density of states at the nodal point is unaffected by the presence of a tilt of the weyl dispersion , and ( 2 ) the critical disorder strength for the transition decreases upon increasing the tilt . importantly , the critical disorder strength approaches zero as the tilt approaches the critical tilt , so that in the presence of disorder the tilt - induced phase transition from a `` type - i '' weyl semimetal ( with a point - like fermi surface ) to a `` type - ii '' weyl semimetal ( with finite particle and hole pockets at the nodal point ) is always preceded by a disorder - induced transition to a diffusive phase with a finite density of states at the nodal point .
another conclusion of our findings is that there is a family of weak - disorder pseudoballistic fixed points , all of them being effectively disorder - free , but distinguished by their effective tilt .
the value of the tilt _ i.e. _ which fixed point one is in can be deduced from the fano factor , which is a nontrivial function of the tilt parameter @xmath26.@xcite ( in principle , the dimensionless conductance @xmath17 also depends on the tilt but , unlike the fano factor , @xmath17 also depends on a possible anisotropy of the weyl dispersion , which the fano factor @xmath22 does not . ) in principle tilted spectra can also occur in two - dimensional systems with a dirac dispersion , such as graphene . in that context one
expects that tilt , too , will quantitatively influence transport properties and the density of states .
however , in two dimensions there is no disorder - induced phase transition between a quasi - ballistic and a diffusive regime , and the effect of tilt will be mainly quantitative , and not qualitative .
moreover , in graphene , certainly the two dimensional dirac material studied the most , the dirac points are at high symmetry points in the brillouin zone , at which anisotropies and tilts are prohibited by crystalline symmetries .
while our findings suggest that the tilt - induced transition between type - i and type - ii weyl semimetals may be obscured by the presence of disorder , they also suggest that tilt can be used as a parameter that drives the system through the disorder - induced transition between quasi - ballistic and diffusive phases . in general strain or pressure
induces a change of the lattice geometry , which in turn influences the tilt .
while the magnitude of the tilt is directly influenced by the hopping parameters of a lattice model,@xcite the experimental feasibility depends on the specific material and the corresponding tilt s susceptibility to strain or pressure .
such limitations may make it necessary to start the tilt - induced phase transition already at a close - to - critical disorder strength , so that only a relatively minor change in the band structure is enough to drive the system through the quasi - ballistic - to - diffusive phase transition .
( we note that pressure or strain may also result in a change of fermi velocity , which would also influence @xmath42 , and may be an additional factor that helps / obstructs the observation of the phase transition . )
_ acknowledgments. _ this work is supported by dfgs emmy noether program ( be 5233/1 - 1 ) and crc tr 183 of the deutsche forschungsgemeinschaft , the helmholtz vi `` new states of matter and their excitations '' , the swedish research council and the wallenberg academy fellows program of the kaw foundation .
in this appendix we provide further details on the calculation of the density of states using the self - consistent born approximation ( scba ) , see section [ sec : scba ] . since there is rotational symmetry around the tilt direction ( the @xmath34 axis ) ,
the parameter dependence of the self energy @xmath106 can be restricted , and @xmath106 can be parameterized as @xmath107 where @xmath108 is the angle between @xmath57 and the @xmath34 axis and @xmath109 is the azimuthal angle corresponding to @xmath57 . using the short - hand notations ( with dependence on @xmath110 , @xmath108 , and @xmath69 left implicit ) @xmath111 we can write the self - consistency condition for the green function and self energy as with @xmath114 , @xmath115 , @xmath116 and @xmath117 where @xmath118 and @xmath119 are modified bessel functions .
self - consistency equations for @xmath120 , @xmath121 , and @xmath122 then immediately follow from the definitions ( [ xy ] ) . for the numerical computation we have to compute @xmath85 at small but finite imaginary values for the energy @xmath69 .
combined with the need to use a finite grid in momentum energy space to perform the integrations this leads to a slight rounding of the onset of a finite density of states upon entering the diffusive phase from the weak - disorder regime .
this obstructs a direct determination of the critical disorder strength from the density - of - states calculations in the scba .
we have estimated the precise location of the critical disorder strength from a linear fit of density of states for disorder strengths just above the critical disorder strength .
an example of such a linear fit is shown in fig . [
fig : determine - kcrit ] .
the fitting procedure leaves a small residual error for the value of @xmath24 , which is shown by the error bars in fig .
[ fig : weyl - tilted ] .
h. weyl , _ elektron und gravitation .
i _ , http://dx.doi.org/10.1007/bf01339504[z .
physik * 56 * , 330 - 352 ( 1929 ) ] .
g. e. volovik , _ the universe in a helium droplet _ , http://dx.doi.org/10.1093/acprof:oso/9780199564842.001.0001[oxford university press ( 2009 ) ] .
s. murakami , _ phase transition between the quantum spin hall and insulator phases in 3d : emergence of a topological gapless phase _ ,
http://iopscience.iop.org/1367-2630/9/9/356[new journal of physics * 9 * , 356 ( 2007 ) ] .
x. wan , a. m. turner , a. vishwanath and s. y. savrasov , _ topological semimetal and fermi - arc surface states in the electronic structure of pyrochlore iridates _ , http://link.aps.org/doi/10.1103/physrevb.83.205101[phys .
b * 83 * , 205101 ( 2011 ) ] .
a. a. burkov and l. balents , _ weyl semimetal in a topological insulator multilayer _ , http://link.aps.org/doi/10.1103/physrevlett.107.127205[phys .
lett . * 107 * , 127205 ( 2011 ) ] . s. xu , i. belopolski , n. alidoust , m. neupane , g. bian , c. zhang , r. sankar , g. chang , z. yuan , c. lee , s. huang , h. zheng , j. ma , d. s. sanchez , b. wang , a. bansil , f. chou , p. p.
shibayev , h. lin , s. jia and m. z. hasan , _ discovery of a weyl fermion semimetal and topological fermi arcs _ ,
http://science.sciencemag.org/content/349/6248/613[science * 349 * , 613617 ( 2015 ) ] .
b. q. lv , h. m. weng , b. b. fu , x. p. wang , h. miao , j. ma , p. richard , x. c. huang , l. x. zhao , g. f. chen , z. fang , x. dai , t. qian and h. ding , _ experimental discovery of weyl semimetal taas _ , http://link.aps.org/doi/10.1103/physrevx.5.031013[phys .
x * 5 * , 031013 ( 2015 ) ] .
l. lu , z. wang , d. ye , l. ran , l. fu , j. d. joannopoulos and m. soljai , _ experimental observation of weyl points _ , http://dx.doi.org/10.1126/science.aaa9273[science * 349 * , 622 - 624 ( 2015 ) ] .
x. huang , l. zhao , y. long , p. wang , d. chen , z. yang , h. liang , m. xue , h. weng , z. fang , x. dai and genfu chen _ observation of the chiral - anomaly - induced negative magnetoresistance in 3d weyl semimetal taas _ , http://dx.doi.org/10.1103/physrevx.5.031023[phys .
x * 5 * , 031023 ( 2015 ) ] . c. zhang , s. xu , i. belopolski , z. yuan , z. lin , b. tong , g. bian , n. alidoust , c. lee , s. huang and et al .
, c. zhang , s. xu , i. belopolski , z. yuan , z. lin , b. tong , g. bian , n. alidoust , c. lee , s. huang , t. chang , g. chang , c. hsu , h. jeng , m. neupane , d. s. sanchez , h. zheng , j. wang , h. lin , c. zhang , h. lu , s. shen , t. neupert , m. z. hasan and s. jia , _ signatures of the adler - bell - jackiw chiral anomaly in a weyl fermion semimetal _ , http://dx.doi.org/10.1038/ncomms10735[nat . commun .
* 7 * , 10735 ( 2016 ) ] .
e. j. bergholtz , z. liu , m. trescher , r. moessner and m. udagawa , _ topology and interactions in a frustrated slab : tuning from weyl semimetals to @xmath123 fractional chern insulators _ , http://link.aps.org/doi/10.1103/physrevlett.114.016806[phys .
lett . * 114 * , 016806 ( 2015 ) ] . m. trescher , b. sbierski , p. w. brouwer and e. j. bergholtz , _ quantum transport in dirac materials : signatures of tilted and anisotropic dirac and weyl cones _ , http://dx.doi.org/10.1103/physrevb.91.115135[phys .
b * 91 * , 115135 ( 2015 ) ] .
a. a. soluyanov , d. gresch , z. wang , q. wu , m. troyer , x. dai and b. a. bernevig , _ type - ii weyl semimetals _ , http://dx.doi.org/10.1038/nature15768[nature * 527 * , 495 - 498 ( 2015 ) ] . y. xu , f. zhang and c. zhang , _ structured weyl points in spin - orbit coupled fermionic superfluids _ , http://dx.doi.org/10.1103/physrevlett.115.265304[phys .
* 115 * ( 2015 ) ] .
y. i. rodionov , k. i. kugel and f. nori , _ effects of anisotropy and disorder on the conductivity of weyl semimetals _ , http://dx.doi.org/10.1103/physrevb.92.195117[phys .
b * 92 * , 195117 ( 2015 ) ] . c. w. j. beenakker ,
_ tipping the weyl cone _ , http://www.condmatjournalclub.org/?p=2644[journal club for condensed matter physics * august 2015 * ( 2015 ) ] .
k. deng , g. wan , p. deng , k. zhang , s. ding , e. wang , m. yan , h. huang , h. zhang , z. xu , j. denlinger , a. fedorov , h. yang , w. duan , h. yao , y. wu , y. s. fan , h. zhang , x. chen and s. zhou , _ experimental observation of topological fermi arcs in type - ii weyl semimetal mote2 _ , http://arxiv.org/abs/1603.08508v1[arxiv:1603.08508v1 ( 2016 ) ] .
n. xu , z. j. wang , a. p. weber , a. magrez , p. bugnon , h. berger , c. e. matt , j. z. ma , b. b. fu , b. q. lv , n. c. plumb , m. radovic , e. pomjakushina , k. conder , t. qian , j. h. dil , j. mesot , h. ding and m. shi , _ discovery of weyl semimetal state violating lorentz invariance in mote2 _ , http://arxiv.org/abs/1604.02116v1[arxiv:1604.02116v1 ( 2016 ) ] . l. huang , t. m. mccormick , m. ochi , z. zhao , m. suzuki , r. arita , y. wu , d. mou , h. cao , j. yan , n. trivedi and a. kaminski , _ spectroscopic evidence for type ii weyl semimetal state in mote2 _ , http://arxiv.org/abs/1603.06482v1[arxiv:1603.06482v1 ( 2016 ) ] . c. wang , y. zhang , j. huang , s. nie , g. liu , a. liang , y. zhang , b. shen , j. liu , c. hu , y. ding , d. liu , y. hu , s. he , l. zhao , l. yu , j. hu , j. wei , z. mao , y. shi , x. jia , f. zhang , s. zhang , f. yang , z. wang , q. peng , h. weng , x. dai , z. fang , z. xu , c. chen and x. j. zhou , _
spectroscopic evidence of type ii weyl
semimetal state in wte2 _ , http://arxiv.org/abs/1604.04218v1[arxiv:1604.04218v1 ( 2016 ) ] .
i. belopolski , s. xu , y. ishida , x. pan , p. yu , d. s. sanchez , m. neupane , n. alidoust , g. chang , t. chang , y. wu , g. bian , h. zheng , s. huang , c. lee , d. mou , l. huang , y. song , b. wang , g. wang , y. yeh , n. yao , j. e. rault , p. l. fvre , f. bertran , h. jeng , t. kondo , a. kaminski , h. lin , z. liu , f. song , s. shin and m. z. hasan , _ measuring chern numbers above the fermi level in the type ii weyl semimetal mo@xmath124w@xmath125te@xmath9 _ , http://arxiv.org/abs/1604.07079v1[arxiv:1604.07079v1 ( 2016 ) ] k. koepernik , d. kasinathan , d. v. efremov , s. khim , s. borisenko , b. bchner and j. v. d. brink , _ tairte4 a ternary type - ii weyl semi - metal _ , http://arxiv.org/abs/1603.04323v2[arxiv:1603.04323v2 ( 2016 ) ] .
goerbig , j .-
fuchs , g. montambaux and f. piechon , phys .
b 78 , 045415 ( 2008 ) .
_ tilted anisotropic dirac cones in quinoid - type graphene and @xmath126-(bedt - ttf)@xmath9i@xmath127 _ , http://journals.aps.org/prb/abstract/10.1103/physrevb.78.045415[arxiv:1603.08508v1 ( 2016 ) ] .
p. baireuther , j. m. edge , i. c. fulga , c. w. j. beenakker and j. tworzydo , _ quantum phase transitions of a disordered antiferromagnetic topological insulator _ , http://link.aps.org/doi/10.1103/physrevb.89.035410[phys .
b * 89 * , 035410 ( 2014 ) ] .
b. sbierski , g. pohl , e. j. bergholtz and p. w. brouwer , _ quantum transport of disordered weyl semimetals at the nodal point _ , http://link.aps.org/doi/10.1103/physrevlett.113.026602[phys .
lett . * 113 * , 026602 ( 2014 ) ] .
j. tworzydo , b. trauzettel , m. titov , a. rycerz and c. w. j. beenakker , _ sub - poissonian shot noise in graphene _ , http://link.aps.org/doi/10.1103/physrevlett.96.246802[phys .
96 * , 246802 ( 2006 ) ] .
m. i. katsnelson , _ zitterbewegung , chirality and minimal conductivity in graphene _ , http://link.springer.com/article/10.1140/epjb/e2006-00203-1[eur .
j. b * 51 * , 157 - 160 ( 2006 ) ] .
e. fradkin , _ critical behavior of disordered degenerate semiconductors .
ii . spectrum and transport properties in mean - field theory _ , http://dx.doi.org/10.1103/physrevb.33.3263[phys .
b * 33 * , 3263 - 3268 ( 1986 ) ] . k. kobayashi , t. ohtsuki , k .-
imura and i. f. herbut , _ density of states scaling at the semimetal to metal transition in three dimensional topological insulators _ , http://journals.aps.org/prl/abstract/10.1103/physrevlett.112.016402[phys .
lett . * 112 * , 016402 ( 2014 ) ] .
s. v. syzranov , l. radzihovsky and v. gurarie , _ critical transport in weakly disordered semiconductors and semimetals _ , http://journals.aps.org/prl/abstract/10.1103/physrevlett.114.166601[phys .
lett . * 114 * , 166601 ( 2015 ) ] .
b. sbierski , e. j. bergholtz and p. w. brouwer , _ quantum critical exponents for a disordered three - dimensional weyl node _ , http://link.aps.org/doi/10.1103/physrevb.92.115145[phys .
b * 92 * , 115145 ( 2015 ) ] .
j. h. pixley , p. goswami and s. das sarma _
anderson localization and the quantum phase diagram of three dimensional disordered dirac semimetals _ , http://journals.aps.org/prl/abstract/10.1103/physrevlett.115.076601[phys .
* 115 * , 076601 ( 2015 ) ] .
a. weie , g. wellein , a. alvermann and h. fehske , _ the kernel polynomial method _ , http://link.aps.org/doi/10.1103/revmodphys.78.275[rev .
phys . * 78 * , 275306 ( 2006 ) ] .
y. ominato and m. koshino , _ quantum transport in a three - dimensional weyl electron system _ , http://dx.doi.org/10.1103/physrevb.89.054202[phys .
b * 89 * , 054202 ( 2014 ) ] . | although lorentz invariance forbids the presence of a term that `` tilts '' the energy - momentum relation in the weyl hamiltonian , a tilted dispersion is not forbidden and , in fact , generic for condensed matter realizations of weyl semimetals .
we here investigate the combined effect of such a tilted weyl dispersion and the presence of potential disorder .
in particular , we address the influence of a tilt on the disorder - induced phase transition between a quasi - ballistic phase at weak disorder , in which the disorder is an irrelevant perturbation , and a diffusive phase at strong disorder .
our main result is that the presence of a tilt leads to a reduction of the critical disorder strength for this transition or , equivalently , that increasing the tilt at fixed disorder strength drives the system through the phase transition to the diffusive strong - disorder phase .
notably this obscures the tilt induced lifshitz transition to an over tilted type - ii weyl phase at any finite disorder strength .
our results are supported by analytical calculations using the self - consistent born approximation and numerical calculations of the density of states and of transport properties . |
several persistent x - ray binaries show large flux variations in their x - ray lightcurves on timescales significantly longer than their orbital periods ( wen et al .
highly periodic flux variaitons at time scales larger than their respective orbital periods are seen in her x-1 ( 35 d : tananbaum et al .
1972 , still & boyd 2004 ) , lmc x-4 ( 30.28 d : paul & kitamoto 2002 ) , 2s 0114 + 650 ( 30.75 d : farrell , sood & oneill 2006 ) , ss 433 ( 164 d : eikenberry et al . 2001 ; fabrika 2004 ) , xte j1716389 ( 99.1 d : cornelisse , charles & robertson 2006 ) , 4u 1820303 ( 172 d : smale & lochner 1992 ; zdziarski et al .
2007a ) and cyg x-1 ( 150 d : kitamoto et al .
2000 ; lachowicz et al .
the first three of these seven sources , known with stable superorbital period are accretion powered pulsars .
several other x - ray binaries show quasi periodic long term flux variations .
smc x-1 has a long term period in the range of 50 - 70 d ( grubers & rothschild 1984 ; clarkson et al .
2003 , wen et al .
grs 1747 - 312 shows quasi periodic flux variations of @xmath0 150 d @xcite .
cyg x-2 shows quasi periodic long term variability with period in the range of 60 - 90 d @xcite and lmc x-3 also shows long term flux variations with periodicity in the range of 100 - 500 d @xcite that is unstable over a longer period @xcite .
long term quasi periodic variations with a period of about 217 d are also seen from the rapid burster ( x1730333 ) which is in the form of recurrent outburst rather than a gradual change in x - ray flux @xcite .
cen x-3 is a high mass x - ray binary pulsar with a spin period of @xmath1 s and an orbital period of @xmath2 d , discovered with @xmath3 @xcite .
this x - ray binary is known to have strong long term flux variations with transient quasi - periodicity @xcite .
the long term lightcurve of this source obtained with the all sky monitor ( asm ) of the _ rossi x - ray timing explorer _ ( _ rxte _ ) showed very strong aperiodic flux variations along with two different accretion modes @xcite . in 12 years of monitoring data obtained with the _ rxte_-asm in three different energy bands , the hardness ratio between the 5 - 12 kev and 3 - 5 kev bands was found have a larger value during december 2000 to april 2004 , compared to the same outside this period .
long term observations with burst and transient source experiment ( batse ) of the _ compton gamma ray observatory ( cgro ) _ found that cen x-3 has alternate spin - up and spin - down intervals which last from about 10 to 100 days @xcite .
however , _ ginga _
observations revealed that there is no correlation between the observed x - ray flux and pulse period derivative of cen x-3 @xcite and it was suggested that the observed x - ray flux of cen x-3 does not represent its mass accretion rate .
the periodic long term variations seen in x - ray binaries are explained by variable obscuration of the central x - ray source by an accretion disk precessing under tidal force .
the tidal precession can be in the form of forced precession of accretion disks in the gravitational field of the companion star @xcite or slaved precession of accretion disc due to the rotation axis of the star being inclined @xcite .
tilted and twisted disks due to coronal winds @xcite and radiation pressure induced warped precessing disks @xcite can have variable or chaotic precession periods .
iping and petterson ( 1990 ) simulated the temporal evolution of radiatively warped accretion disk and found that it can give rise to aperiodic variability in the x - ray lightcurves .
they suggested that the long term variability of cen x-3 could be caused by such a radiatively warped accretion disk .
other more detailed disk warping models have also been developed where warping instabilities are incorporated into the models which give rise to unstable and chaotic disk precession with no stable long term period @xcite .
accretion torque induced precession of magnetic axis can also be a possible explanation for the observed flux variations @xcite in some x - ray binaries .
third body mechanism is also known to cause periodic x - ray flux variation by modulating the mass accretion rate @xcite . in the present work
we investigate the orbital modulation of the 1.5 - 12 kev x - ray flux of cen x-3 when the source is in high , intermediate and low states to understand if the aperiodic variations are occuring due to a variable mass accretion rate or due to the variable obscurtion of the central x - ray source by an accretion disk .
the eclipse structure is useful to know the size of the observed x - ray emission region in different flux states .
we have also measured the pulsed and total x - ray emission of the source in its different flux states using many observations by the proportional counter array ( pca ) of the @xmath4 .
evolution of the pulsed fraction with observed x - ray flux is useful to know the relative importance of scattered x - ray emission in different flux states .
the asm on board @xmath4 has three detectors which scan the sky in a series of 90 s dwells in 3 energy bands , namely 1.5 - 3 , 3 - 5 and 5 - 12 kev @xcite .
we have used the 1.5 - 12 kev lightcurves of four sources cen x-3 , her x-1 , smc x-1 and lmc x-4 , covering about 4100 days from january 1996 to study and compare the orbital modulation of these sources in three different flux states .
the lightcurves of the 4 sources binned with their respective orbital periods after excluding the eclipse data are shown in figure 1 , for 500 days .
the distinctly aperiodic flux variation of cen x-3 is in sharp contrast with the periodic modulation in her x-1 ( with one main - on and one short - on state ) , lmc x-4 ( with smaller signal to noise ratio ) and the quasi - periodic modulation in smc x-1 ( with some scatter within each high state ) .
for each of the three sources cen x-3 , her x-1 and smc x-1 , we have made 3 separate lightcurves , one each for the high , intermediate and low flux states .
to do this , we first calculated the average count rate per orbit after excluding the eclipse data ( see paul et .
2005 for more details ) .
depending on the average count rate in a binary orbit , the data points available during that orbit were collected in one of the three high , intermediate or low state lightcurves ( see table 1 for the interval of count rates defining different flux states the three sources ) .
these 3 lightcurves of each source were then folded with the respective orbital period of the source to get the orbital modulation lightcurves .
we have adopted a slightly different analysis for lmc x-4 since its signal to noise ratio is small .
we first folded the asm lightcurve with the superorbital period and determined the ephemeris for superorbital modulation as follows .
@xmath5 } = 50092.8 + ( 30.31\pm 0.01 ) n , \label{eq : so_ephemeris}\ ] ] where @xmath6 is an integer . the high , intermediate and low states
were determined based on the suerorbital phase being 0.35 to 0.55 , ( 0.20 to 0.35 , 0.55 to 0.80 ) , and @xmath70.20 to 0.20 , respectively .
data points from the full asm lightcurve belonging to the three different states were then used to make three different lightcurves and folded with the orbital period of lmc x-4 to get the orbital modulation lightcurves .
in figure 2 we have shown the orbital moduation of the four sources in different flux states .
for each soure , the orbital modulation lightcurves of the high and intermediate flux states are shown in the top panel , high state points are indicated by circles .
the low state modulations are shown in the bottom panels .
all the orbital modulation curves are normalised by dividing the original lightcurves by the average count rate calculated over an orbital phase of 0.2 near the peak flux of the respective orbital modulation curve .
the low state asm lightcurve for lmc x-4 does not show any orbtial modulation . in table 1 we have also given the orbital periods and eclipse durations of these 4 sources determined from the _
rxte_-asm lightcurves . in figure 3 , we have shown the flux dependent orbital modulation lightcurve of cen x-3 separately for the hard spectral state , that started in december 2000 and ended in april 2004 ( paul et al .
2005 ) .
we have also investigated the pulsation characteristics of cen x-3 at different flux levels in the 2 - 60 kev band .
cen x-3 was observed by _
rxte_-pca many times during the years 1996 , 1997 and 1998 .
all these observations have been obtained when cen x-3 was in the soft spectral state .
we have chosen 18 pca observations depending on the orbit averaged asm count rates at that time such that a wide range of x - ray flux is covered .
the 2 - 60 kev band lightcurves were obtained from the standard-1 mode data of the pca .
background lightcurves were generated using the background models provided for _
rxte_-pca by heasarc and were subtracted from standard-1 lightcurves to get the source lightcurves .
the source lightcurves were first barycenter corrected and then searched for the spin period of the neutron star .
we did not detect any pulsations when the orbit averaged asm count rate of cen x-3 was less than 0.8 count / sec ( equivalent to about 50 count / sec per proportional counter unit ) with a 90 per cent upper limit of 0.8 % on the pulsed fraction .
the pulse profiles were then generated by folding the barycenter corrected lightcurves by the respective spin period found in that lightcurve .
to avoid smearing of the pulse profiles due to the orbital motion of the pulsar , the pulse profiles were generated from short data segments of duration of a hundred pulses . in figure 4
we have shown six pulse profiles obtained at different source flux levels including a lightcurve folded in the low flux state with a period of 4.81 s when no pulses were detected .
figure 5 is a plot of the maximum - minimum count rate per proportional counter unit ( pcu ) against the maximum count rate per pcu for the 18 pulse profiles .
the points marked with the circles are for the pulse profiles shown in figure 4 .
epochs of the six pca observations , pulse profiles from which are shown in figure 4 , are marked in the top panel of figure 1 .
a two component function was fitted to the points in figure 5 .
@xmath8 the pulse fraction @xmath9 of the pulsating component was determined to be 90% , while the unpulsed component grows upto a count rate of @xmath10 = 175.5 per pcu . as seen in figure 4 , apart from the changing ratio of the unpulsed component of the flux to the total flux
, the pulse shape is also varying from one observation to another . to see whether the pulse shape changes are related to the flux , we selected three different observations with very similar eclipse subtracted orbit average asm count rates of @xmath11 , @xmath12 and @xmath13 . within these three observations ,
the pulse shape changed from a double peak profile to a broad single peak profile but the pulsed fraction remained the same .
we therefore conclude that while the pulsed flux of cen x-3 is related to the total flux , the pulse shape is independent of the x - ray flux .
we have found that in the four sources cen x-3 , her x-1 , smc x-1 and lmc x-4 , the binary orbital modulation of the x - ray flux shows remarkable dependence on the x - ray flux state ( figure 2 ) .
x - ray eclipses are found to be sharp in the high flux state which becomes more gradual in the intermediate state . in the low state , instead of sharp eclipse ingress and egress , there is a smooth flux variation with orbital phase . though the orbital modulation of lmc x-4 is not detectable in the low state of lmc x-4 with asm data , we note that a weak and smooth orbital modulation in the low state of lmc x-4 was clearly detected earlier with _ bepposax _ observations ( naik & paul , 2003 ) . the her x-1 orbital lightcurves show some pre - eclipse dips ( figure 2 ) . in the intermediate state of cen x-3 the eclipse egress starts earlier in phase by 0.03 compared to the high state as is shown with a vertical dashed line . using a subset of the _ rxte_-asm lightcurve for her
x-1 , scott & leahy ( 1999 ) also found that the orbital modulation is shallower in the short - on state compared to the same in the main - on state . in the _ bepposax _ observations of lmc x-4 , the absence of clear eclipse transitions in low state
was interpreted as due to obscuration of the central x - ray source by the precessing accretion disk .
the weak orbital modulation of the lightcurve in the low state is due to an extended x - ray scttering region , emission from which dominates the detectable x - rays in the low state .
the flux dependent orbital modulations of these four sources indicate that at lower flux , an increasing fraction of the observed x - rays are from a larger region , comparable to the size of the companion star .
the larger emission region may have different visibility at different orbital phases , leading to the smooth orbital modulation in the low state .
reprocessing of x - rays emitted from the compact star by scattering from stellar wind of the companion star is one likely scenario .
however , for her x-1 , which has a low mass companion star , the scattering medium is more likely to be part of the accretion disk and disk outflows than the stellar wind .
zdziarski et al ( 2007a ) have perfomed a similar analysis of _
rxte_-pca lightcurve of the x - ray binary 4u 1820303 and found a significant dependence of the profile of the orbital modulation on the average count rate .
however , in 4u 1820303 , the superorbital modulation is associated with an accretion rate modulation , probably due to third body interaction .
poutanen , zdziarski & ibragimov ( 2008 ) discovered a superorbital phase dependence of the soft x - ray orbital modulation in cyg x-1 in its hard spectral state , that is related to the size of a bulge in the outer accretion disk .
the ratio of the x - ray flux when the source is in eclipse and when it is out - of - eclipse is a measure of the relative scattering efficiency . only for cen
x-3 there is good enough statistics to compare the ratios .
we find that the out - of - eclipse count rates differ by large factor ( @xmath14 , @xmath15 and @xmath16 for high , intermediate and low states respectively ) while the in - eclipse count rates of the three states are comparable ( @xmath17 for high , @xmath18 for intermediate and @xmath19 for low state ) .
the eclipse count rate varies by a factor of only about @xmath02.5 while the out of eclipse count rate varies by a factor of @xmath018 .
the ratio of x - ray flux of cen x-3 during eclipse and out - of - eclipse is larger in the low state by a factor of @xmath20 compared to the same in the high state .
this behaviour is similar to smc x-1 , in which the eclipse count rate was found to be comparable in high and low states whereas the out - of - eclipse count rate varied by more than a factor of 20 @xcite . the measurement of pulsed x - ray flux as a function of the peak x - ray flux of cen x-3 as presented in the figures 4 and 5 , is also consistent with a scenario in which the measured x - ray flux has two components .
one component is highly variable with a pulsed fraction of about 90% and a second component that is unpulsed .
similar result was obtained for smc x-1 over a wide range of its measured x - ray flux @xcite . in the low state ,
the unpulsed component becomes dominant leading to non - detection of pulses .
we also note that in all the four sources mentioned here , the x - ray pulsations ( pulse period of @xmath04.8 s in cen x-3 , 1.24 s in her x-1 , 0.7 s in smc x-1 and 13.5 s in lmc x-4 ) have never been detected during the low state of the superorbital period , indicating that most of the radiation observed in low state is probably reprocessed emission from a large region @xcite .
the non - detection of pulsations is not due to faintness of the sources . even in the low state ,
cen x-3 , her x-1 and smc x-1 are bright enough for detection of a few percent pulse modulation with the _ rxte_-pca .
we point out the possibility that the intrinsic x - ray luminosity of the central x - ray source may remain unchanged for a long time .
it can be seen in figure 1 that except during the anomalous low state events of her x-1 , the peak luminosity of superorbital modulation of her x-1 , smc x-1 and lmc x-4 does not change very much and the peak of the 5 - 12 kev x - ray luminosity of cen x-3 also has a ceiling ( see figure 1 of paul et al .
2005 ) . though the results presented here indicate that the different flux states of cen x-3 are largely due to varying degree of obscuration , as is the case with her x-1 , smc x-1 and lmc x-4 , we can not completely rule out some contribution to the variability from a varying mass accretion rate , especially the variations associated with spectral change . but
as shown in figure 3 , even during the hard spectral state of cen x-3 from december 2000 to april 2004 , the orbital modulation lightcurve indicate an extended x - ray source at low flux level .
detailed observations over a wide spectral range with future missions like _ astrosat _ can throw more light on these aperiodic variations and spectral mode changes seen in cen x-3 lightcurves .
the binary orbital modulation of x - ray from cen x-3 is similar to that seen in the other three accreting x - ray pulsars . from sharp eclipse in high state
, it turns to a gradual modulation in the low state .
cen x-3 eclipse egress starts earlier in the intermediate state compared to the high state .
these observations indicate a larger emission region in the low state of cen x-3 . the ratio of x - ray flux of cen x-3 during eclipse and out - of - eclipse is larger in the low state by a factor of @xmath20 compared to the same in the high state .
a measurement of the pulsed x - ray flux in different flux states of cen x-3 is consistent with the x - ray flux having two components , one with a large pulsed fraction and a second unpulsed component that dominates in the low state .
we propose that the long term intensity variations in cen x-3 are mostly due to aperiodic obscuration of the compact source by the accretion disk .
the unpulsed x - ray emission from an extended region appears to be due to scattering of the x - rays from the central source by the stellar wind .
we thank the referee a. zdziarski for very useful suggestions that helped us very much to improve this paper .
this research has made use of data obtained from the high energy astrophysics science archive research center ( heasarc ) , provided by nasa s goddard space flight centre .
clarkson , w. i. , chales , p. a. , coe , m. j. , laylock , s. , tout , m. d. , wilson , c. a. , 2003 , mnras , 339 , 447 cornelisse , r. , charles , p. a. , robertson , c. , 2006 , mnras , 366 , 918 eikenberry , s. s. , cameron , p. b. , fierce , b. w. , kull , d. m. , dror , d. h. , houck , j. r. , margon , b. , 2001 , apj , 561 , 1027 fabrika , s. , 2004 , asprv , 12 , 1 farrell , s. a. , sood , r. k. , oneill , p. m. , 2006 , mnras , 367 , 1457 finger , m. h. , wilson , r. b. , fishman , g. j. , 1994 , second compton symposium , p 304 giacconi , r. , gursky , h. , kellogg , e. , schreier , e. , tananbaum , h. , 1971 , apj , 167 , 67 gruber , d. e. , rothschild , r. e. , 1984 , apj , 283 , 546 int zand , j. j. m. , hulleman , f. , markwardt , c. b. , mendez , m. , kuulkers , e. , cornelisse , r. , heise , j. , strohmayer , t. e. , verbunt , f. , 2003 , a&a , 406 , 233 iping , r. c. , petterson , j. a. , 1990 , a&a , 239 , 221 katz , j. i. , 1973 , nature physical sciences , 246 , 87 kaur , r. , paul , b. , raichur , h. , sagar , r. , 2007 , apj , 660 , 1409 kitamoto , s. , egoshi , w. , miyamoto , s. , tsunemi , h. , ling , j. c. , wheaton , w. a. , paul , b. , 2000 , 531 , 546 lachowicz , p. , zdziarski , a. a. , schwarzenberg - czerny , a. , pooley , g. g. , kitamoto , s. , 2006 , mnras , 368 , 1025 levine , a. m. , bradt , h. , cui , w. , jernigan , j. g. , morgan , e. h. , remillard , r. , shirey , r. e. , smith , d. a. , 1996 , apj , 469 , 33 maloney , p. r. , begelman , m. c. , pringle , j. e. , 1996 , apj , 472 , 582 masetti , n. 2002 , a&a , 381 , l45 naik , s. , paul , b. , 2003 , a&a , 401 , 265 ogilvie , g. i. , dubus , g. , 2001 , mnras , 320 , 485 paul , b. , kitamoto , s. , 2002 , jaa , 23 , 33 paul , b. , kitamoto , s. , makino , f. , 2000 , apj , 528 , 410 paul , b. , raichur , h. , mukherjee , u. , 2005 , a&a , 442 , l15 poutanen , j , zdziarski , a. a. , ibragimov , a. 2008 , arxiv0802.1391 , submitted to mnras priedhorsky , w. c. ; terrell , j. , 1983 , apj , 273 , 709 roberts , w. j. , 1974 , apj , 187 , 575 schandl , s. , 1996 , a&a , 307 , 95 schandl , s. , meyer , f. , 1994 , a&a , 289 , 149 scott , d. m. , leahy , d. a. , 1999 , apj , 510 , 974 smale , a. p. , lochner , j. c. , 1992 , apj , 395 , 582 still , m. , boyd , p. , 2004 ,
apj , 606 , l135 .
tananbaum , h. , gursky , h. , kellogg , e. m. , levinson , r. , schreier , e. , giacconi , r. , 1972 , apj , 174 , l143 truemper , j. , kahabka , p. , oegelman , h. , pietsch , w. , voges , w. , 1986 , apj , 300 , 63 tsunemi , h. , kitamoto , s. , tamura , k. , 1996 , apj , 456 , 316 wen , l. , levine , a. m. , corbet , r. h. d. , bradt , h. v. , 2006 , apjs , 163 , 372 wijers , r. a. m. j. , pringle , j. e. , 1999 , mnras , 308 , 207 wojdowski , p. , clark , g. w. , levine , a. m. , woo , j. w. , zhang , s. n. , 1998 , apj , 502 , 253 zdziarski , a. a. , gierliski , m. , wen , l. , kostrzewa , z. , 2007a , mnras , 377 , 1017 zdziarski , a. a. , wen , l. , gierliski , m. , 2007b , mnras , 377 , 1006 | we have investigated the long term flux variation in cen x-3 using orbital modulation and pulsed fraction in different flux states using observations made with the all sky monitor and the proportional counter array on board the _ rossi x - ray timing explorer_. in the high state , the eclipse ingress and egress are found to be sharp whereas in the intermediate state the transitions are more gradual . in the low state , instead of eclipse ingress and egress , the lightcurve shows a smooth flux variation with orbital phase .
the orbital modulation of the x - ray lightcurve in the low state shows that the x - ray emission observed in this state is from an extended object .
the flux dependent orbital modulations indicate that the different flux states of cen x-3 are primarily due to varying degree of obscuration .
measurement of the pulsed fraction in different flux states is consistent with the x - ray emission of cen x-3 having one highly varying component with a constant pulsed fraction and an unpulsed component and in the low state , the unpulsed component becomes dominant .
the observed x - ray emission in the low state is likely to be due to scattering of x - rays from the stellar wind of the companion star .
though we can not ascertain the origin and nature of the obscuring material that causes the aperiodic long term flux variation , we point out that a precessing accretion disk driven by radiative forces is a distinct possibility .
stars : neutron
( stars : ) pulsars : individual : cen x-3 x - rays : stars
( stars : ) binaries : general x - rays : individual : cen x-3 x - rays : binaries |
the study of lattice quantum many body systems by the exact diagonalisation technique has proven popular at zero temperature ( @xmath0 ) where only the ground state is required , but it is of less use at finite temperature .
the reason can be attributed to the different system sizes applicable , where for spin-1/2 the ground state can be found for up to _ n_@xmath130 lattice sites , but the entire spectrum can readily be achieved for systems only up to _ n_@xmath116 .
the intrinsic difficulties associated with applying the finite - size scaling method on such small systems severely limit finite temperature applications . at @xmath0
the continued fraction technique @xcite allows accurate calculations of dynamical correlations using only the machinery of the lanczos algorithm , but unfortunately this technique has not been extended to finite @xmath2 where mostly full diagonalisation has been employed . as well as direct applications of the canonical ensemble
, there is also a hybrid method which employs the canonical representation of dynamical correlation functions but uses a lanczos basis to provide a set of orthogonal states @xcite .
this method allows access to larger systems than are accessible to full diagonalisation techniques but to smaller systems than the current proposal , which does not need details of all the states even in the lanczos basis . in this article
we extend the @xmath0 formalism to finite temperature by applying a _
microcanonical ensemble _ approach combined with the lanczos method ( mclm ) that provides smooth predictions for dynamical correlation functions at least at high temperatures .
the physical advance is to appreciate that in the thermodynamic limit the microcanonical ensemble is equivalent to the canonical one @xcite , but for finite systems this is much easier to work with .
the statistical fluctuations engendered by the microcanonical choice are a drawback for small systems but become controllable for large systems . in practice ,
as the finite @xmath2 calculations are much smoother , it is more natural to contemplate applying finite - size scaling than for their @xmath0 counterparts .
the exponentially dense nature of a many particle spectrum in the bulk is the property that smoothes our calculations , a characteristic that is lost near the ground state where the spectrum is sparse .
besides the computational interest of this proposal it is worth pointing out that , to our knowledge , no studies of the fundamental equivalence between the microcanonical and canonical ensemble for quantum dynamic correlations exist in the literature .
thus this work is a step in numerically exploring this basic postulate of nonequilibrium statistical mechanics ; clearly , analytical studies are needed to clarify , for instance , the meaning of the microcanonical ensemble for a quantum system with dense spectrum as an average over a single quantum state ( or a narrow window of states ) and the finite size corrections inherent in this ensemble .
in this section we will discuss how an arbitrary probability distribution can be used , under reasonable assumptions , to represent the canonical ensemble in the thermodynamic limit .
the choice of a ( unnormalized ) distribution @xmath3 in the thermodynamic limit can be examined by considering its laplace transform , @xmath4 where @xmath5 controls the properties of a distribution designated by @xmath6 . as examples , the following choices of @xmath5 lead to , @xmath7 the canonical ensemble , @xmath8 the microcanonical ensemble at energy @xmath9 , @xmath10 the gaussian " ensemble at energy @xmath9 and width @xmath11 . if we examine the partition function @xmath12 , then , @xmath13 where @xmath14 is the canonical partition function . the physical idea behind
the thermodynamic limit is that the partition function becomes immensely sharp when considered as a function of state space and becomes dominated by the large number of states with the correct thermodynamics ; in practice fluctuating quantities can be replaced by their thermodynamic average with negligible error .
mathematically , an integral such as ( [ z ] ) may be approximated in the asymptotic thermodynamic limit using the idea of ` steepest descents ' with negligible error , @xmath15 here @xmath16 is chosen so that , @xmath17 and the average energy at the desired temperature is crucial . for the particular case of the microcanonical distribution @xmath18 and we need to employ states whose energy is the thermodynamic average as one might naively guess .
provided that @xmath5 has only a weak dependence on the system parameters , then the two partition functions are essentially equivalent .
it is also clear that provided @xmath5 has the required properties that the ` steepest descents ' is a good approximation , then any appropriate ensemble will provide the thermodynamic limit , for a particular temperature .
it is quite natural to employ @xmath19 where @xmath20 is extrinsic and @xmath21 is intrinsic , in order to limit towards the microcanonical ensemble .
the usually studied quantities of direct physical interest are the dynamic structure function , @xmath22 and dynamic susceptibility , @xmath23 \rangle , \label{chi}\end{aligned}\ ] ] where @xmath24 , the angle brackets denote a canonical ensemble thermal average and the commutator plays a central role in the linear response theory ( or kubo ) formulation of transport .
the two quantities are related by the fluctuation - dissipation relation , @xmath25 where @xmath26 is the inverse temperature .
note that @xmath27 satisfies the symmetry relation @xmath28 while the sum - rule , @xmath29 makes it natural to consider the normalised to a unit area correlation function , @xmath30 we have presented the dynamical correlation functions in the canonical ensemble and now we will establish their form in the microcanonical one . starting from equation ( [ s ] ) and employing solely the idea that our distribution has a restricted energy @xmath9 we can generate a correlation function @xmath31 in the microcanonical ensemble , @xmath32 here , we have used the relation , @xmath33 ( @xmath34 a function of @xmath35 ) and a decomposition using the eigenbasis @xmath36 . the expression ( [ sm ] ) integrates to provide , @xmath37 that can be re - represented as the basic correlation in the microcanonical ensemble , @xmath38 ^{-1 } x_{-{\bf q}}\rangle .
\label{basic}\ ] ] notice that this expression is exact in the zero temperature limit where the expectation value is to be taken over the ground state wavefunction .
now let us imagine that we could find a single eigenstate at will , with an energy arbitrarily close ( in the thermodynamic limit ) to a target energy , @xmath9 say .
it is in principle straightforward then to determine @xmath39 ^{-1 } x_{-{\bf q}}\mid * \rangle,\ ] ] exactly as before , where @xmath40 is the known eigenstate with @xmath41 .
if the microcanonical ensemble is equivalent to the canonical ensemble and if a single eigenstate is representative of the microcanonical one , then provided that @xmath18 for the desired temperature , we can expect that @xmath42 in the thermodynamic limit .
this amounts to the physical idea behind our calculations .
furthermore , from ( [ fd ] ) , it follows that @xmath43 and from the symmetry of @xmath44 in the canonical ensemble we can deduce that , @xmath45 this relation then provides an alternative , cross - checking technique for determining the temperature for a particular value of @xmath20 .
although we might like to believe that a single eigenstate corresponds to the microcanonical ensemble , based on a putative ergodicity assumption for the eigenstate , in practice it is not possible to find such an eigenstate .
so we relax the eigenstate hypothesis and go back to a distribution of eigenstates close to the desired value @xmath9 .
we simply use the formalism as though we had such an eigenstate .
in principle we must construct a particular eigenstate with energy @xmath9 that equals the canonical expectation value of the energy , @xmath46 , at the desired temperature . in practice
, we employ the well known lanczos algorithm that is an efficient way of diagonalising large hamiltonians using as variational subspace ( truncated basis ) the set of states , @xmath47 where @xmath48 is a ( usually random ) initial state and @xmath49 the number of lanczos steps . to obtain an eigenstate close to energy
@xmath20 one might expect to use the closest eigenstate to @xmath20 in the truncated basis , but this is totally incorrect . in practice ,
only the states at the edge of the spectrum converge and the other ` eigenstates ' in the truncated subspace have the suggested energies but are usually far from eigenstates . in order to apply the lanczos method idea
, one can simply push the energetic region of interest to the edge of the spectrum by choosing an appropriate new operator .
one natural choice is to use , @xmath50 which is positive definite and pushes the eigenstates with energy close to @xmath20 towards the minimal , zero , eigenvalue of @xmath51 .
another way to understand this technique is to consider expanding the ground state of @xmath51 , that we will call @xmath52 , as a probability distribution over eigenstates .
choosing @xmath9 establishes the appropriate mean for this distribution but minimising @xmath51 corresponds to minimising the variance of the distribution , and consequently localising the distribution near @xmath20 .
one can perform a lanczos calculation based upon the operator @xmath51 or , more efficiently , one can evaluate the operator @xmath51 ( now a pentadiagonal matrix ) in a previously constructed lanczos basis using @xmath35 ( l - projection " method ) . note that , @xmath53 ( the expectation value is over @xmath52 ) and so a small variance guarantees a narrow distribution of energies around @xmath9 . in any lanczos calculation
the mathematical orthogonality between states becomes lost at some stage as numerical errors build up . in practice
only the well - separated converged states suffer from this disease and for us these states , which are at the edge of the spectrum , do not gain any significant weight in the correlation functions and so do not manifest in our results . the states at low frequency are all well behaved and maintain their orthogonality .
it is straightforward to implement these ideas numerically , with a ` double - lanczos ' calculation ; the first run through a lanczos procedure of @xmath54 steps is employing the operator @xmath51 starting from a random state and it is used to find the state @xmath55 which plays the role of the microcanonical distribution .
the second run of @xmath56 steps through lanczos is made using @xmath57 as the initial state and then the resulting tridiagonal matrix can be diagonalised to form the dynamical correlations directly or by employing the continued fractions method which is numerically more efficient but introduces a loss of resolution .
all the analysis so far has been subject to several caveats ; firstly , that the microcanonical ensemble is equivalent to the canonical one in the thermodynamic limit and in the context of quantum dynamic correlations .
secondly , that a single eigenstate is equivalent to the microcanonical ensemble and thirdly that we can find such an eigenstate at will . the first two assumptions , as we have mentioned in the introduction , should be the focus of analytical studies as fundamental issues of nonequibrium statistical mechanics . regarding the third assumption , it is clearly problematic as it is well known that although the lanczos method converges quite easily at the sparse edges of the spectrum , in the denser inner regions of the spectrum , of interest at finite temperature , it takes the lanczos procedure an exponentially large number of iterations to converge .
a many - body spectrum has an exponential number of states , e.g. for spin-1/2 the # ( states)@xmath58 , and for a bounded hamiltonian the eigenstates are compressed into an energy region that grows only linearly with system size . although the low energy region maintains a sparse density of states , the eigenstates become exponentially close together in the area of interest and essentially become unattainable . at first sight
this appears an insurmountable difficulty , but in practice this issue allows the technique its success .
the first lanczos procedure provides a single quantum state @xmath55 , that is not an eigenstate , but which when decomposed in an eigenstate basis , it is represented by a narrow distribution @xmath59 around @xmath9 ; @xmath60 gives for the expectation value of an operator @xmath61 , @xmath62 this state , used in the evaluation of expectation values , acts as a statistical average over an energy window .
it is important to notice , that by employing a single quantum state ( not eigenstate ) for evaluating an expectation value ( as a substitute for a statistical average over a narrow energy window of eigenstates ) , we assume that the appearing off - diagonal terms ( second term in eq.([rpa ] ) ) cancel each other
. this assumption can be justified ( and numerically verified ) by invoking a random phase decomposition of the used quantum state . from this discussion
we can expect two types of fluctuations in the obtained spectra ; first , intrinsic fluctuations due to the finite size of the system , present even when a single eigenstate is used for the evaluation of the expectation value .
second , statistical fluctuations entering by the off - diagonal terms in eq .
( [ rpa ] ) due to the use of a single pure state that is not an eigenstate ; this type of fluctuations can be reduced by averaging over orthogonal states @xmath55 ( e.g. corresponding to different translational symmetry @xmath63 subspaces as we will show below ) .
in the following we present a test on the rate of convergence of the projection to a single quantum state with energy close to @xmath9 . due to the innate complexity of an implicit scheme like lanczos
, we develop the theory of a simpler technique briefly to exhibit the ideas .
a rather simple method of numerically solving for the ground state is by an iterative sequence of applications of the scaled hamiltonian . for us
this amounts to iterative applications of the operator , @xmath64 where @xmath65 is chosen to be large enough so that @xmath66 for the full spectrum . repeated applications of this operator exponentially suppresses all states except those for which @xmath67 which remain unaffected . we can start out with a set of random states and then for @xmath68 applications of our operator we can build a distribution , @xmath69 \langle \psi \mid p^m\ ] ] ( @xmath70 is the step function ) and if we were to perform an average over an orthogonal basis , @xmath71 , then this would converge to , @xmath72 p^{2m}.\ ] ] elementary analysis provides : @xmath73 ( 2m)!2^{2m}2\mu \left[\frac{1}{x}\frac{d}{dx}\right]^{2 m } \frac{\sinh x}{x}\vert _ { x=\beta \mu } .\ ] ] in the limit that @xmath74 we find that , @xmath75 and we converge to a narrow gaussian probability distribution , @xmath76.\ ] ] the width of this distribution is under our control , @xmath77 where @xmath78 is the natural energy scale for the model and we see that @xmath68 needs to scale with the square of the system size @xmath79 to maintain resolution .
the lanczos method is clearly much more sophisticated and provides a much narrower distribution .
we have examined the distribution obtained in a lanczos calculation and we find that it is well represented by a gaussian distribution with a variance controlled by the ` eigenvalue ' of @xmath51 attained by the calculation . in practice
this is about two orders of magnitude better in energy than the result obtained from the projection analysis ( eq.([p ] ) ) , which however it is analytically controllable ; indeed , we find that the lanczos method scales as @xmath80 so that the intrinsic resolution , @xmath81 , is inversely proportional to the number of iterations .
the convergence properties of the three schemes we discussed are depicted in figure [ fig1 ] for a representative calculation of the study that we present in the next section . ) , ( ii ) dotted line , using @xmath82 , ( iii ) continuous line , l - projection " ( see text).,width=8 ] the application of the technique should now be transparent ; employing a single random state , or averaging over a sequence of orthogonal random states , one performs a first lanczos calculation of @xmath54 steps to find the approximate ground state @xmath52 for the operator @xmath83 . the value of @xmath9 must be pre - selected so that @xmath18 for the chosen temperature ; several techniques are available for reliably determining this energy versus temperature relation as the bethe ansatz ( for integrable systems ) , the finite temperature lanczos ( ftlm ) @xcite , the transfer matrix renormalization group ( tmrg ) or quantum monte carlo method .
the degree of convergence can be measured using the eventual ` eigenvalue ' of @xmath51 ; it plays the role of the variance of the chosen distribution and its square - root is an intrinsic energy resolution @xmath11 .
this scale , @xmath84 , can never drop below the distance to the nearest eigenvalue . for a usual size system , e.g. @xmath85 , and temperature ,
this limit is unattainable but a resolution of @xmath86 ( @xmath87 ) is readily available with a thousand or so @xmath54 iterations .
once one has found this state @xmath88 , that plays the role of the state @xmath89 , a second lanczos projection sequence is generated employing the state @xmath90 as the initial state .
the resolution of the eventual result is controlled by the intrinsic dependence on the microcanonical ensemble and the degree of convergence measured by @xmath11 .
this can be seen from relation ( [ basic ] ) as the eigenstates over which the state @xmath91 is decomposed have a spread in energy @xmath11 with respect to the reference energy @xmath9 .
the resolution also depends on the convergence achieved in the 2nd lanczos procedure where the number of iterations @xmath56 denotes the finite number of poles which are used to try to represent the dynamical correlations . at the sparse edges of the spectrum
these poles denote the eigenvalues of the system but in the bulk of the spectrum , when grouped into bins of a given frequency width , they are fairly uniformly spread and offer a further natural energy resolution for the calculation .
more lanczos steps provide more poles and a finer spectral ` grid ' for the correlation functions , until the graininess of the real system is achieved .
we have elected to use a few thousand poles in our calculations with very little improvement obtained by increasing this number as we shall see .
the final resolution is self - imposed and is the @xmath92 of ( [ basic ] ) which we choose to be of order of the spectral grid in order to smooth our calculations .
we are now in a position to test our proposed technique and uncover its strengths and weaknesses .
we have chosen to investigate the finite temperature dynamics of the prototype spin-1/2 heisenberg model ( equivalent to the fermionic @xmath93 " model ) .
this choice was dictated by its central role in low dimensional quantum magnetism ; an exact solution of the thermodynamics and elementary excitations is known using the bethe ansatz method @xcite , the spin dynamics probed by nmr is of current experimental interest @xcite and several numerical and analytical studied have been devoted to the study of finite temperature dynamic correlations @xcite .
the hamiltonian is given by , @xmath94 where @xmath95 are spin-1/2 operators on site @xmath96 and we take @xmath97 as the unit of energy and frequency ( @xmath98 ) .
in particular , we will look at the high temperature spin conductivity in the antiferromagnetic regime , @xmath99 , for which several studies exist and some exact results are known @xcite . to discuss magnetic transport ,
we first define the relevant spin current , @xmath100 , by the continuity equation of the corresponding local spin density @xmath101 ( provided the total @xmath102 component is conserved ) , @xmath103 thus , we obtain for the spin current @xmath100 , ( that plays the role of the operator @xmath104 ) , @xmath105 the real part of the spin conductivity " @xmath106 ( corresponding to the charge conductivity of the fermionic model ) includes two parts , the drude weight @xmath107 and the regular part @xmath108 @xcite , @xmath109 the regular contribution is given by , @xmath110 where @xmath111 are the boltzmann weights and @xmath112 .
to compare the presented data on the conductivity we normalize them using the well known optical sum - rule that in the @xmath113 limit takes the form , @xmath114 the normalized conductivity , @xmath115 , in this high temperature limit is given by , @xmath116 that can be calculated using our microcanonical ensemble procedure by , @xmath117 in principle this expression includes also the contribution from the zero frequency drude weight @xmath118function , but in practice as the second lanczos procedure can not fully converge , the drude peak appears as a low frequency contribution . as we will discuss below ,
sorting out this low frequency part , in general allows us to reliably extract the drude weight value . , @xmath119 , @xmath120 ; ( a ) @xmath0 , ( b ) @xmath121.,width=8 ] in general
, we can employ the translational symmetry of the hamiltonian and study spectra in a given pseudomomentum @xmath63 subspace or average the results over different @xmath63 subspaces ; in the following we typically employ @xmath122 and @xmath123 lanczos iterations at @xmath121 unless otherwise stated . in figure [ fig2 ]
we compare a zero temperature @xcite with an infinite temperature ( @xmath113 ) calculation for a fairly large system in the @xmath124 subspace .
the zero temperature calculation finds a few poles with exact weights whereas the infinite temperature calculation provides a much smoother result
. there is clear structure in the infinite temperature result but also apparently some noise . to interpret this result
we must consider the issue of the veracity of the microcanonical ensemble for such small systems namely the extent to which the microcanonical ensemble is equivalent to the canonical one .
, @xmath125 , @xmath126 , ( b ) @xmath127 , @xmath128 , @xmath126.,title="fig:",width=8 ] , @xmath125 , @xmath126 , ( b ) @xmath127 , @xmath128 , @xmath126.,title="fig:",width=8 ] in figure [ fig3 ] we present a comparison , extremely encouraging , of some microcanonical calculations with the analogous canonical ones .
there is ` noise ' in all calculations , the origin and magnitude of which we will now discuss .
the canonical calculations are essentially a direct evaluation of expression ( [ sigma ] ) , where we applied a binning " procedure on the @xmath118function weights over an energy scale of about 0.01 .
the number of contributing matrix elements are of the order of the dimension @xmath129 of the hilbert space squared , @xmath130 , e.g. @xmath131 @xmath118 functions , with no continuity in the weights .
the results are not smooth and the resulting intrinsic fluctuations are heavily smoothed by our binning procedure . in the microcanonical calculations we employ our scheme , further averaging over translational symmetry @xmath63subspaces .
now , only of @xmath132 @xmath118functions are essentially contributing , multiplied by the number of states involved in the decomposition of the state @xmath88 ( a few thousand depending on the convergence ) and the number of @xmath63subspaces .
we could average over initial random states , but we find that this has only a small smoothing effect , because the underlying poles are at the same energies .
notice that the observed fluctuations are not associated with any of our different resolution processes which are much smaller than the observed scale of fluctuations ; they are due to the finite size of our system and thus to the effective smaller number of matrix elements contributing to the construction of the spectra . this seemingly new problem associated with our technique turns out to be dominant for small system sizes ; very soon however it becomes negligible as larger systems are achieved , specially considering that the dimension of hilbert space grows exponentially fast with the system size @xmath79 . in order to assess these fluctuations and simultaneously the role of our smoothing parameter @xmath92 , we performed some basic calculations involving only a single @xmath63subspace state @xmath133 . in figure [ fig4 ]
we offer a comparison of calculations involving just the poles evaluated using the 2nd lanczos procedure eigenstates against smoothed versions of the same data but employing the continued fraction technique .
= 2 ; ( a ) n=22 , ( b ) n=28.,title="fig:",width=8 ] = 2 ; ( a ) n=22 , ( b ) n=28.,title="fig:",width=8 ] the fluctuations clearly decay with system size with the final system being surprisingly smooth .
the limitations of the smoothing process are clear , the sharper features are slightly washed out although the ease of assessing the data makes such a smoothing advisable
. the weights for these microcanonical calculations are truly quite continuous in comparison to the intrinsic properties of the canonical calculation which is necessarily ragged .
obviously for our largest calculations we are nowhere near converged to the true spectrum which is a possible explanation for the observed continuity .
we can now fairly safely conclude that our technique is a viable way to calculate dynamical correlation functions at high temperature for the same systems accessible by the lanczos method at @xmath0 . by its very nature ,
the finite @xmath2 correlations are much smoother and more regular to interpret .
our technique introduces new statistical fluctuations which make small system sizes ragged but appear to leave large system sizes essentially unaffected .
although we can now investigate finite temperature dynamic correlations using the lanczos method , we are still restricted to @xmath134 for a spin-1/2 system .
the key to making useful physical deductions is the procedure of finite - size scaling , the attempt to deduce the properties of the infinite size system using assumed properties of the size , @xmath79 , dependence .
this method has been widely and succesfully applied in the evaluation of ground state energies or gap values using data provided by the exact diagonalization , lanczos or density matrix renormalization group technique .
but to extract information on finite temperature dynamic correlations one would need to know the form of the curves before fitting and scaling could take place mathematically . as it is clear from figure [ fig5 ] this might be a challenging task considering the statistical fluctuations inherent in the spectra @xcite ; however , from ongoing studies on other systems using this method , we find that the behavior of the spectra might greatly depend on the model hamiltonian and correlations under study ( e.g. it is far more structurless for energy current dynamic correlations ) .
note that the high frequency behavior is generally rather weakly size dependent while the low frequency one is the most subtle to determine .
the last however is the most physically interesting as it determines , for instance , the diffusive or ballistic behavior of the conductivity . ,
width=8 ] the basic properties of the @xmath113 current - current correlations are now available and so we provide in figure [ fig6 ] a few examples of the frequency dependence of the conductivity at @xmath121 as a function of @xmath135 . for @xmath136 ;
( a ) @xmath137 , ( b ) @xmath138,width=8 ] although we have devoted most of our effort to infinite temperature ( @xmath113 ) , our technique is valid at essentially any temperature ( provided that we remain at a dense region of the spectrum ) . analysing the pure heisenberg model
, we look at a couple of finite temperature @xmath63averaged calculations in figure [ fig7 ] . the temperature has been deduced from a least - squares fit of the quantity , @xmath139 to a linear ansatz , and although the statistical fluctuations are compounded , an almost vanishing intercept and a clear slope indicate the feasibility of the strategy .
the obtained @xmath140 values compare favorably with those corresponding to the canonical ensemble in the thermodynamic limit , evaluated using @xmath141 ; for @xmath142 , @xmath143 vs. @xmath144 , for @xmath145 , @xmath146 vs. @xmath147 .
, @xmath148=1 , @xmath149 ; ( a ) @xmath150 , @xmath142 , ( b ) temperature fit @xmath151 , ( c ) @xmath150 , @xmath145 , ( d ) temperature fit @xmath152.,title="fig:",width=8 ] , @xmath148=1 , @xmath149 ; ( a ) @xmath150 , @xmath142 , ( b ) temperature fit @xmath151 , ( c ) @xmath150 , @xmath145 , ( d ) temperature fit @xmath152.,title="fig:",width=8 ] although we have compared numerical evaluation of dynamic correlations obtained by a canonical and microcanonical method , we have yet to compare with an exact solution .
recently even non - zero temperature dynamical correlations have become partially accessible , with a calculation of the drude weight for the @xmath153 heisenberg model at finite temperature @xcite .
in particular , the drude weight in the @xmath121 limit is given analytically @xcite by , @xmath154 the drude weight , strictly speaking , is defined as the weight of a zero frequency @xmath118function , eq.([real ] ) ; it is a particularity of the heisenberg model that it appears as a narrow peak at low frequencies , of the order of the inverse lattice size @xcite , in contrast to the fermionic
t - v " version where it is accounted for only by the diagonal energy elements ( @xmath155 ) . in extracting the drude weight by the above described procedure we must take into account the problem caused by the intrinsic resolution of our calculations , which is of order @xmath156 . although our chosen resolution of @xmath157 is almost invisible for the smooth background , for the drude weight the resolution is essentially limited by that of our ` microcanonical ' distribution , viz @xmath84 .
an example of these ideas is provided in figure [ fig8 ] , from which it is clear that the drude peak is the only contribution for which the change in resolution is relevant .
these calculations involve a single state and are much improved by @xmath63averaging , also the energy window is so small that the individual poles in the 2nd lanczos procedure are visible and have been smoothed out with an @xmath158=0.005 which adds to the observed resolution . in the inset
, the scale of the conductivity clearly signals a low frequency peak ( notice the difference in scale between figure [ fig8 ] and its inset ) ; still in order to extract the drude weight from the smooth background , we must integrate the peak up to at least as far as it is resolved and that necessitates the inclusion of some of the background .
we have elected to err on the side of inclusion and tend to integrate past where the drude peak appears to become small . , for @xmath159 and @xmath160 ; inset , low frequency range.,width=8 ] in figure [ fig9 ] we offer a comparison of the analytical and numerically extracted drude weights in the @xmath113 limit .
the quantitative agreement is reasonably satisfactory , becoming rather poor near @xmath161 because of our technique for extracting the drude weight ; due to the finite resolution of our calculation we need to sample a finite width around @xmath162 . for the case
@xmath128 there is no drude weight but there does appear to be a power - law like divergence which we pick up in our finite window leading to the observed corrupted behaviour .
drude weight , @xmath163 ; numerical evaluation ( points ) vs analytical expression eq .
( [ cjj ] ) ( continuous line).,width=8 ]
our investigation appears to validate the use of the lanczos algorithm to analyse finite temperature dynamical properties of strongly correlated systems ; the crucial step is to employ the microcanonical ensemble , which essentially allows the thermodynamic average to be replaced by an elementary expectation value .
all the simplicity of the zero temperature formalism can then be taken over to the finite temperature calculation .
the comparison of canonical with microcanonical procedures indicates that the thermodynamic limit is reached with quite modest system sizes and consequently there appears to be little systematic error coming from our choice of ensemble .
there are intrinsic statistical fluctuations in our calculations but these are severely curtailed by increasing the system size and are an implicit difficulty with canonical calculations too .
we believe that we can calculate the high enough temperature dynamical correlations for a finite system with an excellent tolerance .
the statistical fluctuations in our results require to be controlled if an error analysis is to be contemplated .
although we have not got analytical control , we do have experience at various approaches to reducing the statistical fluctuations .
the crucial point is that , when taking a statistical average , one should use orthogonal " states ( @xmath88 s decomposed into different sets of eigenstates @xmath164 ) . averaging over random starting vectors in the same subspace is not very effective , even if they are originally orthogonal , because the resulting distribution involves the same states and consequently an overlap .
performing a @xmath63 average , over translational symmetry subspaces , is an excellent procedure , since the states are automatically orthogonal and intellectually one is reverting back towards the real physical statistical average .
another possibility is to use several of the eigenstates of the first lanczos procedure ; although the orthogonality is guaranteed , there is an induced loss in resolution due to the larger @xmath84 s of the higher lanczos states .
a final possibility is to employ the parameter @xmath9 , where the average over different @xmath9 s must be limited within a window that corresponds to the energy fluctuations at the studied temperature in the given size system .
providing that the @xmath9 s are further apart than the chosen @xmath11 , the orthogonality is essentially guaranteed .
although we believe we have access to the temperature behaviour of finite - size systems , this does not give immediate access to the dynamics in the thermodynamic limit because finite - size scaling must be performed ; figure 6 exhibits clear peaks of unknown form , plausible ` cusps ' and regions where the correlations vanish .
unless we can guess or deduce the form of these structures , finite - size scaling appears problematic .
we should note however from our experience , that not all models and dynamic correlations exhibit so involved spectra ; in forthcoming works we will present analysis of charge / spin / energy current correlations for other ( non- ) integrable systems of current interest ( higher spin , ladder models ) where the obtained spectra are far more structrurless .
finally , besides the finite frequency behavior , our method allows the reliable study of scalar quantities as the drude weight .
part of this work was done during visits of ( p.p . ) and ( m.l . ) at irrma as academic guests of epfl .
x.z acknowledge support by the swiss national foundation , the university of fribourg and the university of neuchtel .
r. haydock , v. heine and m.j .
kelly , j. phys .
c*5 * , 2845 ( 1972 ) .
cagliano and c.a .
balseiro , phys .
* 59 * , 2999 ( 1987 ) .
j. jakli , p. prelovek , phys .
b*49 * , 5065 ( 1994 ) ; adv . phys .
* 19 * , 1 ( 2000 ) .
landau and e.m .
lifshitz , course of theoretical physics : statistical mechanics " , * 5 * , 377 ( footnote ) , pergamon press , london - paris ( 1959 ) .
lebowitz , j.k .
percus and l. verlet , phys . rev . *
153 * , 250 ( 1967 ) .
quantum inverse scattering method and correlation functions " , v.e .
korepin , n.m .
bogoliubov and a.g .
izergin , cambridge univ . press ( 1993 ) .
m. takigawa et al . , phys .
rev . lett . * 76 * , 4612 ( 1996 ) .
thurber et al . , phys .
lett . * 87 * , 247202 ( 2001 ) .
x. zotos and p. prelovek , phys .
* , 983 ( 1996 ) .
x. zotos , f. naef and p. prelovek , phys .
rev . b*55 * 11029 ( 1997 ) .
k. fabricius and b.m .
mccoy , phys .
rev . b*57 * , 8340 ( 1998 ) and references therein .
narozhny , a.j .
millis and n. andrei , phys .
rev . b*58 * , 2921 ( 1998 ) .
x. zotos , phys .
* 82 * , 1764 ( 1999 ) . j. v. alvarez and c. gros , phys . rev . lett . * 88 * , 077203 ( 2002 ) . the low frequency behavior of the spin-1/2 heisenberg model , in particular the finite - size dependence as extracted from such a microcanonical study , will be adressed in a forthcoming publication .
a. klmper , private communication .
s. fujimoto and n. kawakami , j. phys .
a. * 31 * , 465 ( 1998 ) . | we show how to generalise the zero temperature lanczos method for calculating dynamical correlation functions to finite temperatures .
the key is the microcanonical ensemble , which allows us to replace the involved canonical ensemble with a single appropriately chosen state ; in the thermodynamic limit it provides the same physics as the canonical ensemble but with the evaluation of a single expectation value .
we can employ the same system sizes as for zero temperature , but whereas the statistical fluctuations present in small systems are prohibitive , the spectra of the largest system sizes are surprisingly smooth .
we investigate , as a test case , the spin conductivity of the spin-1/2 anisotropic heisenberg model and in particular we present a comparison of spectra obtained by the canonical and microcanonical ensemble methods . |
spontaneous chiral symmetry breaking ( s@xmath6sb ) is one of the important characteristics of qcd .
the standard scenario assumes that the mechanism of s@xmath6sb is triggered by the formation of a large @xmath4 in the qcd vacuum .
and there is an alternative scenario regarding that s@xmath6sb can be realized even with a small or vanishing @xmath4 in confining vector - like gauge theories @xcite @xcite .
@xmath0 scattering at threshold will provide a test to the two pictures of s@xmath6sb . for a summary of current experimental information can refer ref .
@xcite . in the near future
, there will be several experiments dedicated to the precise determination of the @xmath0 scattering lengths , including semileptonic @xmath7 decays from e865 @xcite and kloe @xcite , and the measurement of the pionium lifetime ( dirac ) @xcite , which directly determines the s - wave @xmath0 scattering length difference @xmath8 . on the theoretical side ,
the s - wave @xmath0 scattering lengths @xmath1 and @xmath2 have been investigated in the standard scenario by using the standard chiral perturbation theory up to two - loops @xcite @xcite and the nambu - jono - lasinio ( njl ) model to the leading order of @xmath9 expansion where @xmath10 is the number of color freedoms @xcite - @xcite .
also @xmath1 and @xmath2 have been calculated in the alternative scenario @xcite by using the generalized chiral perturbation theory formulated by orsay group @xcite .
the standard scenario of s@xmath6sb is usually motivated by the njl model @xcite , in which s@xmath6sb is associated with large quark condensate .
the effect of explicit chiral symmetry breaking induced by a small current quark mass @xmath11 which is important to discuss the process of @xmath0 scattering at threshold . as we have analyzed in @xcite , only beyond mean - field approximation ,
can the effect of massive pions be seen clearly .
our interest in this paper is to study the s - wave @xmath0 scattering lengths @xmath1 and @xmath2 in a self - consistent su(2 ) njl model @xcite with explicit chiral symmetry breaking @xcite .
the two - flavor njl model is defined through the lagrangian density , @xmath12,\end{aligned}\ ] ] where @xmath13 is the effective coupling constant with dimension @xmath14 , @xmath11 the current quark mass , and @xmath15 quark fields with flavor , colour and spinor indices suppressed , assuming isospin degeneracy of the @xmath16 and @xmath17 quarks .
we first briefly review the general scheme of the njl model beyond mean - field approximation @xcite . including current quark mass
explicitly , the quark self - energy @xmath18 expanded to @xmath19 order can be expressed as @xmath20 where @xmath21 and @xmath22 are the leading @xmath23 and subleading @xmath19 contributions shown in fig .
1 . the solid lines in fig . 1 indicate quark propagator @xmath24 with full @xmath18 .
the quark condensate @xmath4 is a one - loop quark integral @xmath25 the corresponding meson propagator @xmath26 ( m means @xmath27 or @xmath28 ) has the form @xmath29 where @xmath30 is the meson polarization function , which includes the leading order @xmath31 and subleading order @xmath32 , shown in fig .
1 . the above constituent quark mass and the meson propagator , namely the eqs .
( [ gap ] ) and ( [ prog ] ) , or the feynman diagrams in fig . 1 , form a self - consistent description of the su(2 ) njl model to the subleading order of @xmath9 expansion , which is different from the earlier calculations @xcite - @xcite where only the mean - field quark mass @xmath21 and the random - phase - approximation ( rpa ) meson polarization function @xmath31 are considered .
the meson mass @xmath33 satisfies the total meson propagator s pole condition @xmath34 and the meson - quark coupling constant @xmath35 is determined by the residue at the pole @xmath36 another important quantity in the meson sector is the pion decay constant @xmath37 which generally satisfies @xmath38 in the chiral limit , @xmath37 satisfies the goldberger - treiman relation @xmath39 .
( 500,100)(0,0 ) ( 120,50)(15,0,360 ) ( 120,35)2.5 ( 120,19)[]@xmath21 ( 200,50)(15,0,360 ) ( 200,70)(15,320,220)4 ( 200,35)2.5 ( 189,61)2.5 ( 210,61)2.5 ( 200,19)[]@xmath22 ( 200,75)[]@xmath27,@xmath28 ( 500,85)(0,0 ) ( 105,50)(15,0,360 ) ( 90,50)2.5 ( 120,50)2.5 ( 110,15)[]@xmath40 ( 180,50)(15,0,360 ) ( 180,65)(180,35)4 ( 165,50)2.5 ( 180,65)2.5 ( 180,35)2.5 ( 195,50)2.5 ( 180,15)[]@xmath41 ( 180,50)[]@xmath27,@xmath28 ( 260,50)(15,0,360 ) ( 260,70)(15,320,220)4 ( 249,60)2.5 ( 248,41)2.5 ( 271,60)2.5 ( 272,41)2.5 ( 260,75)[]@xmath27,@xmath28 ( 260,15)[]@xmath42 ( 340,50)(10,0,360 ) ( 400,50)(10,0,360 ) ( 330,50)2.5 ( 340,60)2.5 ( 340,40)2.5 ( 410,50)2.5 ( 400,60)2.5 ( 400,40)2.5 ( 370,65)[]@xmath27 ( 370,35)[]@xmath28 ( 370,15)[]@xmath43
now we turn to the calculation of the s - wave @xmath0 scattering length @xmath1 and @xmath2 .
the invariant scattering amplitude can be generally written as @xcite @xmath44 where @xmath45 and @xmath46 are the isospin labels of the initial and final states respectively , and @xmath47 and @xmath16 are the mandelstam variables , @xmath48 , @xmath49 , @xmath50 , where @xmath51 are initial and final isospin momentum . using perfect crossing symmetry , one can project out isospin amplitudes @xmath52 in the limit of scattering at threshold , @xmath53 , t = u=0 , the s - wave scattering lengths @xmath54 ( given in units of @xmath55 are : @xmath56 we will calculate the scattering length to the lowest order in @xmath9 expansion of the process @xmath57 in the njl model with quark mass @xmath18 and meson polarization function @xmath58 including the subleading order of @xmath9 expansion .
the feynman diagrams include box and @xmath59 exchange diagrams , and are the same as those in @xcite , but in@xcite quark mass @xmath21 in mean - field approximation and meson polarization function in rpa @xmath60 .
to keep the lowest order of the @xmath28 exchange diagrams , the internal @xmath28 propagator should be in leading @xmath19 order , which has the same form as that in rpa @xmath60 @xcite .
so what we need to do is using the expression formulae given in @xcite directly , and replacing the quantities in the mean - field approximation with those beyond mean - field approximation , i.e. , the external pion propagator becomes the total propagator eq .
( [ prog ] ) , the pole for quark propagator becomes the total quark self - energy @xmath18 defined in eq .
( [ gap ] ) , and the coupling constant @xmath61 now is the one expressed in eq .
( [ couple ] ) .
for the numerical calculations , we adopt the external momentum expansion method as discussed in detail in @xcite .
we introduce a quark momentum cut - off @xmath62 in pauli - villars regularization and a meson momentum cut - off @xmath63 in covariant regularization for the divergent momentum integrals . using only the two experimental observables
@xmath64 and @xmath65 and the reasonable quark condensate @xmath66 in the empirical range of @xmath67 as input , we can not give fixed values of the four parameters in the model , namely the current quark mass @xmath11 , coupling constant @xmath13 , and the two momentum cuts @xmath62 and @xmath63 .
so we should introduce one more free parameter @xmath68 , which characterizes the meson cloud contributions .
especially , in the limit of @xmath69 , the model goes back to the mean - field approximation automatically .
as we discussed in @xcite , for each @xmath70 , quark condensate @xmath71 has a minimum @xmath72 , around which there is a region quark condensate changes slowly with quark mass @xmath18 , and in this region of @xmath18 , @xmath11 and @xmath62 are in plateaus too . in table . 1
, we list the values of @xmath18 , @xmath11 and @xmath62 corresponding to @xmath72 ( in units of gev ) .
0.2 cm ( 320,100)(0,0 ) ( 0,0)(320,0 ) ( 0,80)(320,80 ) ( 0,100)(320,100 ) ( 100,90)[]@xmath69 ( 40,70)[]@xmath71 ( 40,50)[]@xmath18 ( 40,30)[]@xmath11 ( 40,10)[]@xmath62 ( 280,70)[]0.256 ( 220,70)[]0.237 ( 160,70)[]0.216 ( 160,90)[]@xmath73 ( 220,90)[]@xmath74 ( 280,90)[]@xmath75 ( 100,70)[]0.209 ( 280,50)[]0.34 ( 220,50)[]0.36 ( 160,50)[]0.47 ( 100,50)[]0.49 ( 280,30)[]0.0067 ( 220,30)[]0.0075 ( 160,30)[]0.0086 ( 100,30)[]0.0089 ( 280,10)[]0.82 ( 220,10)[]0.73 ( 160,10)[]0.63 ( 100,10)[]0.615 for each @xmath70 , with changing @xmath4 , we get a series of @xmath18 and @xmath61 , from which we can calculate the s - wave @xmath76 scattering length @xmath1 and @xmath2 . in the quark condensate plateau region around @xmath77 , there is also a plateau for @xmath1 and @xmath2 respectively .
we choose the values of @xmath1 and @xmath8 corresponding to @xmath72 to compare with those analyzed from experimental data and predicted by other theories , which are listed in table .
2 , where all the values are in units of @xmath78 . in this table ,
the experimental values of @xmath1 and @xmath2 without star were obtained in 1979 in a comprehensive phase shift analysis of peripheral @xmath3 reactions and @xmath7 decays @xcite ; those marked with star were from the chew - low - goebel analyses of the recent chaos data @xmath79 , and lampf e1179 data @xmath80 @xcite . and the experimental value of @xmath81 was obtained from the universal curve in @xcite . 0.2 cm ( 460,120)(0,0 ) ( 0,120)(460,120 ) ( 0,100)(460,100 ) ( 0,0)(460,0 ) ( 0,60)(460,60 ) ( 0,20)(460,20 ) ( 20,10)[]@xmath8 ( 20,40)[]@xmath2 ( 20,80)[]@xmath1 ( 300,110)[]@xmath82 ( 300,90)[][email protected] ( 300,50)[][email protected] ( 300,70 ) [ ] * [email protected] ( 300,30 ) [ ] * [email protected] ( 300,10)[][email protected] ( 370,110)[]@xmath84 ( 370,80)[]0.217 ( 370,40)[]-0.0413 ( 370,10)[]0.258 ( 430,110)[]@xmath85 ( 430,80)[]0.263 ( 430,40)[]-0.027 ( 430,10)[]0.29 ( 70,110)[]njl(@xmath69 , ( 70,80)[]0.16 ( 70,40)[]-0.044 ( 70,10)[]0.20 ( 120,110)[]@xmath73 , ( 120,80)[]0.17 ( 120,40)[]-0.046 ( 120,10)[]0.216 ( 170,110)[]@xmath74 , ( 170,80)[]0.196 ( 170,40)[]-0.052 ( 170,10)[]0.248 ( 220,110)[]@xmath75 ) ( 220,80)[]0.221 ( 220,40)[]-0.057 ( 220,10)[]0.278 from table .
2 we can see that , @xmath1 and @xmath2 calculated from njl model at a certain @xmath70 between @xmath86 are in agreement with the recent experimental values marked with star , and the scattering lengths at @xmath69 are the same as the weinberg values @xmath87 and @xmath88 @xcite
. comparing with those at @xmath69 , the scattering lengths at @xmath89 are all modified by @xmath90 . in the region of @xmath70 between @xmath86 in table .
1 , the quark mass @xmath18 is about 333 mev , and current quark mass @xmath11 is in the empirical region @xmath91 , and the quark condensate @xmath92 is between @xmath93 , which is in agreement with that @xmath94 from the lattice simulation @xcite and that @xmath95 from the sum rule calculation ( see @xcite and references therein ) .
it is can be seen from table .
2 , that the recent experimental results support the standard chiral mechanism , i.e. , s@xmath6sb is triggered by the formation of large quark condensate in qcd vacuum . while the scattering lengths predicted by two - loop generalized chiral perturbation theory , in which the quark condensate @xmath4 is around @xmath96 , are in agreement with the old experimental values without star .
anyway , which s@xmath6sb mechanism is right , should be answered after the analyses of the new experimental data in the near future .
if the elastic pion - pion scattering experimental data in the near future support the mechanism of large quark condensate , the free parameter @xmath70 in the njl model introduced to characterize the mesonic contributions can be fixed by the scattering lengths . then one can determine the four parameters @xmath11 , @xmath97 , @xmath62 and @xmath63 . from the above analyses
, it seems that @xmath70 is between @xmath86 for the njl model .
the authors would like to thank dr .
bing - song zou for his kind help during the work .
this work is supported by nnsf of china ( nos .
19677102 and 19845001 ) .
2 l. maiani and g. martinelli , phys . lett . * b 178 * , 265 ( 1986 ) .
+ d. daniel et al . , phys . rev . *
d 46 * , 3130 ( 1992 ) .
+ l. giusti , f. rapuano , m. talevi and a. vladikas , hep - lat/9807014 , nucl . phys . * b 538 * , 249 ( 1999 ) . | within the general framework of su(2 ) nambu - jona - lasinio model beyond mean - field approximation , the s - wave @xmath0 scattering lengths @xmath1 and @xmath2 are calculated .
the numerical results are in agreement with that analyzed from recent experiment @xmath3 data by using chew - low - goebel technique , and the corresponding chiral condensate @xmath4 is around @xmath5 , which is close to the ones obtained from lattice and sum rules . |
a current sheet structure with reversed magnetic fields can be seen everywhere in the universe . among
the magnetic field dissipation processes in the current sheet , magnetic reconnection may be the most popular one . breaking the topology of magnetic field lines , it releases the magnetic energy into particles kinetic energy in a relatively short time - scale ( several to several tens of an alfvn transit time @xmath1 , where @xmath2 is the half thickness of the current sheet and @xmath3 is the typical alfvn speed ) .
it plays an essential role in solar terrestrial sites such as the earth s
magnetotail @xcite and source regions of the solar flare @xcite .
moreover , it is expected in astrophysical sites such as active galactic nuclei @xcite and pulsar winds , especially their well - studied example of the crab nebula @xcite . in case of the crab nebula , energy conversion in a relativistic wind of @xmath0 plasmas
has been a long - standing problem ( the so - called the `` @xmath4-problem '' , where @xmath4 is the ratio of the poynting flux energy to the particle kinetic flux ) @xcite . the wind is originally poynting - dominated ( @xmath5 ) close to the neutron star @xcite ; but
it is kinetic - dominated near the termination shock ( @xmath4 is far less than the unity ; @xmath6 by @xcite and @xmath7 by @xcite ) .
recently , relativistic magnetic reconnection in the striped current sheets has been discussed as the most possible dissipation mechanism @xcite .
the properties of relativistic reconnection in @xmath0 plasmas have been investigated by an analytical study @xcite and by computer simulations @xcite .
one of the most important features of relativistic reconnection is the generation of non - thermal particles due to direct acceleration by dc electric field around the @xmath8-type region @xcite .
however , the above discussion is based on a two - dimensional picture of reconnection .
instabilities in a cross - field plane are also of critical importance to understanding a realistic current sheet problem that includes reconnection . does reconnection grow faster than cross - field instabilities ?
is reconnection the fastest dissipation process ? in the case of non - relativistic ion - electron plasmas ,
the following instabilities have been discussed : the lower hybrid drift instability ( lhdi ) @xcite , the kelvin - helmholtz instability ( khi ) @xcite , and the drift kink instability ( dki ) @xcite .
they lead to plasma heating and particle acceleration as well as the triggering of reconnection .
the purpose of this letter is to search for the basic properties of such cross - field instabilities and of the non - thermal particle generation in a relativistic @xmath0 current sheet . in this letter ,
we report our 2d simulations in a cross - field plane of a relativistic @xmath0 current sheet and we report a new acceleration process related to the relativistic drift kink instability ( rdki ) @xcite .
the simulation code used here is a three - dimensional particle - in - cell code .
the system size is composed of 1 ( @xmath8 ) @xmath9 256 ( @xmath10 ) @xmath9 512 ( @xmath11 ) grids .
we consider periodic boundaries in the @xmath8 , @xmath10 directions , and we periodically set two simulation boxes in the @xmath11 direction .
the typical scale of the current sheet @xmath2 is set to 10 grids , so that the y boundaries are located at @xmath12 , and the z boundaries are located at @xmath13 .
we take the relativistic harris model as the initial current sheet .
the magnetic field , the density of plasmas and their distribution functions are described by @xmath14 , @xmath15 and @xmath16 $ ] , respectively . in the above equations
, @xmath17 is the magnitude of magnetic field in the lobe ( background region ) , @xmath18 is the number density of plasmas in the current sheet , @xmath19 are the drift velocities for each species ; @xmath20 for positrons , @xmath21 for electrons , @xmath22 is the lorentz factor for @xmath23 [ @xmath24^{-1/2}$ ] ] , @xmath25 is the particle energy , @xmath26 is the relativistic four velocity of @xmath27^{-1/2 } \cdot \bm{v}$ ] and @xmath28 is the temperature . in the first simulation , we set @xmath29 and @xmath30 . we set no driving force to excite instabilities , so that instabilities arise from thermal noise .
we set no background plasmas .
the total energy is conserved within an error of @xmath31 throughout the simulation run .
snapshots at three characteristic stages of the simulation are presented in figure [ fig : kink ] .
the three left panels in figure [ fig : kink ] show color contours of the plasma density at ( @xmath32 ) @xmath33 , ( @xmath34 ) @xmath35 and ( @xmath36 ) @xmath37 , where time is normalized by the light transit time @xmath38 .
the three middle panels show color contours of the electric field @xmath39 at the corresponding time ; the white lines show contours of @xmath40 .
the three panels show the particles energy spectra for the whole simulation .
the horizontal axis shows the particles energy normalized by @xmath41 . as shown in the top three panels in fig .
[ fig : kink ] , we observe a kink - type modulation of the current sheet .
we confirmed that this is a signature of the rdki , a relativistic extension of the dki , which is a long - wavelength , current - driven instability in a thin current sheet . in the top middle panel , one can see the typical structure of @xmath39 components of the polarization of electric fields of rdki .
the sign of @xmath39 is positive in the yellow regions , while it is negative in the blue regions .
two types of regions appear alternatively along the current sheet and they are anti - symmetric with the neutral plane ( @xmath42 ) .
note that the @xmath43 direction using the lobe magnetic fields and the observed sign of @xmath39 is consistent with the @xmath11-displacement of the plasma bulk motion .
the symmetric @xmath44 components are also observed and their values are a little less than or comparable to those of @xmath39 .
the @xmath45 components are negligible .
since the system is uniform in the @xmath8 direction , there is no electrostatic @xmath45 fields .
after @xmath46 , the instability turns into its non linear stage .
the three middle panels in fig .
[ fig : kink ] are snapshots of the current sheet at @xmath47 .
the current sheet is strongly folded within a zone of @xmath48 . due to the @xmath11-displacement by rdki ,
red ( originally yellow ) regions , where @xmath49 , are placed in a row around the neutral plane ( @xmath50 ) in the midmost panel of fig .
[ fig : kink ] .
the total kinetic energy in the system has increased by 30% at this time .
importantly , we observe a clear sign of the particle acceleration in a high - energy tail in the energy spectrum of the middle - right panel in fig .
[ fig : kink ] .
we will discuss the acceleration process later .
the kinked - shape of the current sheet is not stable , because fragments of the anti - parallel currents ( @xmath51 ) pull each other .
as they start to collide , the system turns into the `` mixed '' stage .
the snapshots at this stage are presented in the bottom three panels in fig .
[ fig : kink ] at @xmath52 . due to the enhanced diffusion by sheet collisions ,
the total kinetic energy has increased by 170% at this time . in the energy spectrum of the bottom - right panel
, one can recognize both a remnant of the non - thermal tail and global plasma heating .
finally , the mixed current sheet slowly evolves into the broadened current sheet , which is 3 - 4 times thicker than the initial state .
the saturated level of the total kinetic energy is 330%-340% at @xmath53 or later .
the spectrum looks unchanged from the mixed stage ; one can still observe the non thermal tail .
now , we again take a look at the instability during its linear stage . the observed growth rate ( @xmath54 )
is plotted as a function of @xmath55 in fig .
[ fig : growth ] , where @xmath56 is the wavenumber and @xmath57 is the gyro frequency @xmath58 .
the growth rate and wave length of the most dominant mode of rdki are @xmath59 and @xmath60 , respectively .
we have also calculated eigen functions and their growth rates of the rdki , by linearizing relativistic two - fluid equations .
the theoretical growth rates for the present case are shown as a solid line in fig .
[ fig : growth ] .
the simulation results are consistent with them except for shorter wavelength of @xmath61 , where the instability is suppressed by kinetic effects .
we observe a slight signature of the relativistic drift sausage instability ( rdsi ) along with rdki .
however , rdsi is less influential on the current sheet deformation and particle acceleration .
note that no velocity - shear - driven mode such as khi is excited , ant that is because there is no velocity shear between drifting plasmas in the current sheet and empty background plasmas . also note that lhdi is not excited in an @xmath0 case , because there is no delay of a positron s response to an electron s .
next , we discuss an acceleration mechanism that is found at the non linear stage of rdki . by analyzing trajectories of highly accelerated particles , we found that the main site of particle acceleration is the central channel around @xmath62 , where the current sheet was originally located . for simple discussion
, we call this channel the `` acceleration channel ( ac ) '' .
we illustrate the acceleration mechanism in schematic views in fig .
[ fig : accel ] . in the upper view
, the structure of the polarization of @xmath39 fields at the linear stage are illustrated .
they are anti symmetric with respect to @xmath11 and both positive and negative regions are alternatively located along the sheet .
passing through their meandering orbits , some particles can resonate with the @xmath39 fields .
some of them gain their energy and others lose their energy . at the non - linear stage , due to the @xmath11-displacement of the current sheet ,
the positive @xmath39 regions are now located around the ac .
nside the positive @xmath39 regions , typical values of the fields are @xmath63 and @xmath64 .
contrary to the reconnection @xcite in which @xmath65 in an acceleration site , the value of @xmath66 is less than unity inside the positive @xmath39 regions , and so low - energy particles travel into the @xmath67 direction .
however , fewer populations of high - energy particles can travel across the positive @xmath39 regions , because their larmor radii @xmath68 are larger than the half - wavelength of rdki ( @xmath69 ) .
they gain their energy from the electric field @xmath39 , cross the current sheet and move into the neighboring positive @xmath39 regions . since @xmath68 becomes larger as particles gain more energy ( @xmath70 in non - relativistic case and @xmath71 in relativistic case ) , accelerated particles usually satisfies the crossing condition of @xmath72 in the following positive @xmath39 regions . in this way , they are successively accelerated in the ac , passing through multiple positive @xmath39 regions into the @xmath73 direction .
the above condition can be interpreted as the `` trapping condition '' inside the ac .
higher accelerated particles tend to stay in the ac , because larger @xmath68 prevents them from escaping .
this is why the non - thermal tail is enhanced in the energy spectrum .
we think this type of enhanced acceleration can be found in any kink - type instability in a thin current sheet , as long as its wavelength is comparable to a particle s larmor radius . for example , it may be applicable to acceleration of ions , when a thin current sheet of ion - electron plasmas is modulated by khi .
the folded structure evolves into the mixed stage when its @xmath11-displacement ( @xmath74 ) is nearly the same as the half - wavelength of rdki ( @xmath75 ) . assuming that @xmath76 is typically @xmath77 in the positive @xmath39 regions , the time scale of the non - linear stage ( @xmath78 ) is approximated by @xmath79 .
after the system turns into the mixed stage , a few number of high - energy particles and a lot of low - energy particles in the current sheet interact with the positive @xmath39 regions . then the magnetic fields are dissipated in a time - scale of tens of @xmath80 .
roughly speaking , the magnetic field energy is consumed by joule heating of the @xmath39 fields and the total current @xmath81 in the broadened current sheet .
the magnetic energy stored between the folded curves is approximated by @xmath82 .
we assume that the average value of the electric fields @xmath83 , where the factor 0.5 represents the fact that positive @xmath39 is not uniform in the ac .
the total current @xmath81 is represented by the zeroth order current @xmath84 .
thus the dissipation time scale ( @xmath85 ) is approximated by @xmath86 .
next , we briefly compare rdki and magnetic reconnection @xcite . in table
[ table ] , the parameters for several simulation runs of rdki and for relevant runs of reconnection are presented . throughout
the simulation runs , we calculate the ratio of the non - thermal kinetic energy to the total kinetic energy ( @xmath87 ) and their maximum values in time are presented in table [ table ] . in case of run
d3 , @xmath88 is not resolved by our method .
the maximum energies ( @xmath89 ) are also presented in table [ table ] . in the case of rdki ,
the highest energy particles run in the ac into the @xmath73 direction with the light speed during the non - thermal stage and the mixed stage .
their maximum energies are estimated by @xmath90 and the estimated values agree with the simulation results . in three cases of reconnection
, the evolution seems to be restricted by the periodic system size of @xmath91 , because two outflow jets travel into the @xmath92 direction .
due to this restriction , the upper limits of particle energy are estimated by @xmath93 in table [ table ] in reconnection cases .
however , the obtained maximum energies ( @xmath89 ) by reconnection are far larger than the maximum energies by rdki . from the viewpoint of field dissipation ( e.g. striped pulsar winds )
, rdki can dissipate more magnetic energy .
the linear growth rates of the energy conversion ( @xmath94 ) are presented in table 1 .
generally speaking , rdki grows _
faster _ in a relativistic current sheet ( @xmath95 ) and it converts @xmath96 times as much energy as the initial kinetic energy in the current sheet in a time scale of @xmath85 . moreover , rdki spontaneously occurs everywhere in the @xmath8 direction , while single reconnection requires a wide spatial range in the @xmath8 direction in order for the outflow jets to escape .
therefore , even in runs r3 and d3 , in which rdki grows slower , reconnection takes twice as long to dissipate the same amount of the magnetic energy as rdki . on the contrary , reconnection is the more favored candidate as the origin of non - thermal particles .
it produces more non - thermal energy of @xmath88 than rdki as a result of the high ratio of @xmath87 .
our initial results of a three - dimensional simulation show that rdki occurs almost uniformly in the current sheet . also , @xcite have presented their three - dimensional results of triggered reconnection .
they have referred to the importance of rdki in the three - dimensional case and a signature of a two - dimensional - like rdki is observed outside the reconnection region . in case of multiple current sheets structure , such as striped pulsar winds , there are several possibilities of enhanced field dissipation .
if the distance between two currents is less than the critical distance of @xmath97 , the two current sheets collide with each other and 90% or more of the magnetic energy will be dissipated into the particle energy . in typical cases , @xmath98 7 - 8 , but its dependency on @xmath28 or @xmath23 should be further investigated .
we also note that the secondary rdki takes place in the broadened current sheet in some occasions .
let us summarize this letter .
first , we investigate the rdki in a thin @xmath0 current sheet .
second , we find a new acceleration process that uses a folded structure of a current sheet .
the unique point of this process is that ac electric fields evolve into a dc acceleration channel .
third , due to the enhanced acceleration , a non - thermal particles are produced .
fourth , comparison with reconnection shows that rdki is more favorable to dissipate the magnetic energy in the current sheet structure .
we believe that rdki provides a crucial clue to the energy dissipation problem in the current sheets in the universe , including striped pulsar winds .
j. arons 1979 , space sci .
, 24 , 437 e. g. blackman and g. b. field 1994 , phys .
lett . , 72
, 494 f. v. coroniti 1990 , apj , 349 , 538 w. daughton 1998 , j. geophys .
, 103 , 29429 r. c. davidson et al .
1977 , phys .
fluids , 20 , 301 j. w. dungey 1961 , phys . rev .
, 6 , 47 c. h. jaroschek , r. a. treumann , h. lesch and m. scholer 2004 , phys .
plasmas,11 , 1151 c. f. kennel and f. v. coroniti 1984 , apj , 283 , 694 j. g. kirk and o. skjraasen 2003 , apj , 591 , 366 n. a. krall 1971 , phys .
a4 , 2094 h. lesch and g. t. birk 1998 , apj , 499 , 167 y. lyubarsky and j. g. kirk 2001 , apj , 547 , 437 k. mori , d , n , burrows , j. j. hester , g. g. pavlov , s. shibata and h. tsunemi 2004 , apj , 609 , 186 e. n. parker 1957 , j. geophys .
, 62 , 509 p. l. pritchett , f. v. coroniti and v. k. decyk 1996 , j. geophys .
res . , 101 , 27413 r. schopper , h. lesch and g. t. birk 1998 , a&a , 335 , 26 i. shinohara , h. suzuki , m. fujimoto and m. hoshino 2001 , phys .
, 8709 , 5001 p. yoon and j. f. drake 1996 , j. geophys .
res . , 101 , 27327 s. zenitani and m. hoshino 2001 , apj , 562 , l63 s. zenitani and m. hoshino 2003 , proc . of the 28th icrc z. zhu and r. m. winglee 1996 , j. geophys .
res . , 101 , 4885 | two - dimensional particle simulations of a relativistic harris current sheet of pair ( @xmath0 ) plasmas have demonstrated that the system is unstable to the relativistic drift kink instability and that a new kind of acceleration process takes place in the deformed current sheet .
this process contributes to the generation of non - thermal particles and contributes to the fast magnetic dissipation in the current sheet structure .
the acceleration mechanism and a brief comparison with relativistic magnetic reconnection are presented . |
a morse function on a smooth manifold is called perfect if the number of critical points equals the sum of betti numbers .
not every manifold possesses a perfect morse function .
homological spheres ( that are not spheres ) do not possess it ; manifolds with torsions in homologies do not possess it , etc . on the other hand , the celebrated millnor - smale cancelation technique for critical points with neighbor indices ( or , equivalently , cancelation of handles )
provides a series of existence - type theorems @xcite which are the key tool in smale s proof of generalized poincare conjecture @xcite . in the paper we focus on one particular example of a perfect morse function and
discuss some related problems .
namely , we bound ourselves by the space of configuration of the equilateral polygonal linkage with odd number @xmath0 of edges . as the ambient space it makes sense to take either @xmath1 or @xmath2 , which gives us the spaces @xmath3 and @xmath4 . in bigger dimension
the configuration space is not a manifold .
the number @xmath5 is chosen to be odd by the same reason : for even @xmath5 , the configuration space of the equilateral polygonal linkage has singular points .
we are interested in finding a `` natural '' perfect morse function , that is , a function that has a transparent physical or geometrical meaning .
the first candidate of a `` natural '' morse function on @xmath3 was the oriented area function @xmath6 .
indeed , it a morse function with an easy description of its critical points as cyclic polygons ( that is , polygons with a superscribed circle ) , and a simple formula for the morse index of a critical point @xcite .
however , for @xmath3 @xmath6 is not perfect . in particular , for the equilateral pentagonal linkage it has one extra local maximum ( except for the global maximum ) and one extra local minimum , see example [ expentagon ] . for the equilateral heptagonal linkage
the number of morse points greatly exceeds the sum of betti numbers of the configuration space . to build up a perfect morse function
, we take the space @xmath4 and _ decorate _ it .
the decorated space @xmath7 is well adjusted for an appropriate generalization @xmath8 of the area function @xmath6 .
its critical points ( loosely speaking ) are again cyclic polygons .
surprisingly , the function @xmath8 is a perfect morse function . as a direct corollary we interpret cyclic equilateral polygons as independent generators of the homology groups of the configuration space @xmath7 .
for an odd @xmath0 an equilateral _ polygonal @xmath5-linkage _ should be interpreted as a collection of rigid bars of lengths @xmath9 joined consecutively by revolving joints in a chain . _ a configuration _ of the polygonal @xmath5-linkage in the euclidean space @xmath10
, @xmath11 is a sequence of points @xmath12 with @xmath13 and @xmath14 modulo the action of orientation preserving isometries of the space @xmath15 .
we also call @xmath16 a _ polygon_. a configuration carries a natural orientation which we indicate in figures by an arrow .
the set @xmath17 of all configurations up to an orientation - preserving isometry of the ambient space is _ the moduli space , or the configuration space of the polygonal linkage _ @xmath18 .
for @xmath19 the space @xmath17 is a smooth manifold .
we explain below in this paragraph what is known about planar configurations and the signed area function as the morse function on the configuration space .
[ dfn_area ] the _ signed area _ of a polygon @xmath20 with the vertices @xmath21 is defined by @xmath22 a polygon @xmath16 is _ cyclic _ if all its vertices @xmath23 lie on a circle .
cyclic polygons arise here as critical points of the signed area : a polygon @xmath16 is a critical point of the signed area function @xmath6 iff @xmath16 is a cyclic configuration .
their morse indices are known @xcite .
[ expentagon ] @xcite the equilateral pentagonal linkage has 14 cyclic configurations indicated in fig .
the convex regular pentagon and its mirror image are the global maximum and minimum of the signed area @xmath6 .
their morse indices are @xmath24 and @xmath25 respectively .
the starlike configurations are a local maximum and a local minimum of @xmath6 .
there are ten configurations that have three consecutive edges aligned .
their morse indices equal @xmath9 .
[ pentagon ]
the _ decorated configuration space _ is defined as the set of pairs @xmath27 factorized by the diagonal action of the orientation preserving isometries of @xmath2 .
here @xmath28 is the unit sphere centered at the origin @xmath29 . 1 .
the space @xmath7 is an orientable fibration over @xmath4 whose fiber is @xmath30 .
the euler class of the fibration equals zero . proof .
( 1 ) the set of all polygons with fixed sidelengths ( before factorization by isometries ) is known to be orientable .
therefore the set of the pairs ( a polygon , a vector ) is also orientable as a trivial fibration .
since we take a factor by the action of orientation preserving isometries , the result is also orientable .
\(2 ) @xmath31 defines a non - zero section .
the gysin sequence implies : [ cor_gisin ] 1 .
we have the following short exact sequence : @xmath32 2 . the homology groups @xmath33 are free abelian . for the betti numbers we have : @xmath34 [ dfn_arear3 ] let @xmath35 , let @xmath36 be the vertices of @xmath16 .
the _ area _ of the pair @xmath37 is defined as the scalar product : @xmath38 an alternative equivalent definition is : @xmath39 where @xmath40 is the plane orthogonal to @xmath41 and cooriented by @xmath41 .
[ thm_crirical_3d ] for an equilateral polygon with odd number of edges , critical points @xmath42 of the function @xmath8 are pairs @xmath42 such that @xmath16 is a planar cyclic polygon , and @xmath41 is orthogonal to the affine hull of @xmath16 .
if @xmath42 is a critical point , then @xmath43 is critical as well .
the paper @xcite contains a characterization of all critical points for a generic polygonal linkage ( which is not necessarily equilateral ) . in our particular case
it implies that critical points @xmath42 of the function @xmath8 fall into two classes : 1 .
* planar cyclic configurations .
* these are pairs @xmath42 such that @xmath16 is a planar cyclic polygon , and @xmath41 is orthogonal to the affine hull of @xmath16 .
* non - planar configurations .
* they are characterized by the three following conditions : 1 .
the vectors @xmath41 and @xmath44 are parallel ( but they can have opposite directions ) .
2 . the orthogonal projection of @xmath16 onto the plane @xmath45 is a cyclic polygon .
3 . for every @xmath46 ,
the vectors @xmath47 , @xmath48 , and @xmath49 are coplanar . + here
@xmath49 is the @xmath46-th short diagonal , @xmath47 is the vector area of the triangle @xmath50 .
let us show that the second class ( non - planar configurations ) is empty .
indeed , given a non - planar critical configuration , introduce a cartezian system with the @xmath51-axes parallel to @xmath41 .
the three conditions ( a ) , ( b ) , and ( c ) implie that the absolute value of the slope of an edge with respect to the plane @xmath52 does not depend on the edge .
this implies a contradiction with the closing condition : @xmath53 where the indices are modulo @xmath5 .
[ cor_perfectmorse ] 1 .
the function @xmath8 is a perfect morse function on the decorated configuration space @xmath54 for an equilateral linkage with odd number of edges . 2 .
each critical point of the function @xmath8 is a pair @xmath37 , where @xmath16 is a planar cyclic configuration , @xmath41 is a unit vector orthogonal to @xmath16 .
each planar cyclic configuration @xmath16 gives two critical points of the function @xmath8 ( with two different choices of the normal vector @xmath41 ) .
3 . the morse index of a critical point @xmath37 is @xmath55 where @xmath56 is the winding number @xmath16 around the center of the circumscribed circle,@xmath57 is the number of edges that go counterclockwise .
sets an orientation on the plane of the polygon , so it makes sense to speak of `` edges going clockwise '' and `` edges going counterclockwise '' . ] proof .
( _ i _ ) the second statement is already proven .
let us show that the number of critical points equals the sum of betti numbers .
the latter are already known due to a. klyachko @xcite : @xmath58 by corollary [ cor_gisin ] we have @xmath59 each equilateral cyclic @xmath5-gon with an orthogonal vector @xmath41 is defined by its winding number @xmath56 and by the set of edges that go clockwise .
assume that the winding number is positive ( negative values are treated by symmetry ) .
if the number of edges that go clockwise is @xmath57 , then the winding number ranges from @xmath60 to @xmath61 . for @xmath62 denote by @xmath63
the number of such cyclic equilateral polygons for which @xmath64 then @xmath65 ( _ iii _ ) straightforward analysis of the hesian matrix is very complicated ( probably impossible ) , so we use derive a combinatorial approach .
we prove the formula for morse index by induction by @xmath5 .
the base is given by equilateral pentagon .
by symmetry , we assume in whatfollows that we have a critical point @xmath37 such that @xmath66 , or , equivalently , with winding number @xmath67 . proof is based on two observations : \(1 ) there is a natural embedding of a neighborhood of @xmath16 in space @xmath3 to @xmath7 .
it sends a configuration @xmath16 to @xmath37 with the same @xmath16 and @xmath41 orthogonal to the affine hull of @xmath16 .
the direction of @xmath41 we choose in the way such that @xmath8 is positive .
let us choose a basis of the tangent space @xmath68 which starts by the pushforward of a basis of @xmath69 and ends by coordinates of @xmath41 .
the hessian matrix related to this basis is a block matrix : @xmath70 where @xmath71 is the @xmath72 hessian matrix of the planar polygon @xmath16 related to the area function @xmath6 and the space @xmath3 ; @xmath73 is the the @xmath74 unit matrix .
\(2 ) for an @xmath75-gon and a number @xmath76 , consider the embedding @xmath77 which keeps @xmath41 and replaces the edge number @xmath46 by a fold of three edges ( see fig . 1 , ( 3 ) for an equilateral triangle with an edge by a three - fold ) .
each critical point @xmath78 induces a critical point @xmath79 .
since the embedding has codimension two , and all morse indices can only be even , we have either @xmath80 , @xmath81 , or @xmath82
. more precisely , replacement of an edge replaced by a three - fold adds two extra columns to @xmath71 and two extra columns to @xmath83 .
we know from @xcite that the morse index of @xmath16 related to @xmath3 and the oriented area @xmath6 ( that is , the number of negative eigenvalues of @xmath71 ) equals @xmath84 .
it increases by one after replacement of an edge by a three - fold .
this only leaves the case @xmath81 .
these arguments allows to make an induction step @xmath85 and thus proves ( _ iii _ ) for all cyclic configurations with triple edges .
it remains to prove the formula for configurations without triple edges , that is , with all the edges going counterclockwise .
we keep assuming that @xmath66 , so there are exactly @xmath86 such polygons : with @xmath87 .
their morse indices should be @xmath88 , @xmath89 , @xmath90 , etc .
the only question is which configuration has one or the other index .
we know that @xmath71 contributes @xmath91 , @xmath92 , @xmath93 , etc . to each of the morse indices .
the block @xmath83 contributes not more than @xmath91 , and @xmath94 contributes exactly two .
the statement follows . as an illustration
, we list below all cyclic equilateral pentagons and heptagons .
the first column depicts a combinatorial type , the third one tells the number of configurations of this type , and the last one tells the morse index .
[ figure_7gon ]
decorated configuration space and the ares @xmath8 can be defined for a polygonal linkage that is not necessarily equilateral .
however , generically , the function @xmath8 is not a perfect morse function . in the paper
we omit the discussion about non - degenericity of critical points , since it appears to be somewhat technical .
however the following arguments work : we may replace the equilateral linkage @xmath95 by its perturbation @xmath96 has non - degenerate critical configurations that are close to above described equilateral ones .
this work is supported by the russian science foundation under grant 16 - 11 - 10039 .
klyachko a. , _ spatial polygons and stable configurations of points in the projective line .
_ tikhomirov , alexander ( ed . ) et al . , algebraic geometry and its applications .
proceedings of the 8th algebraic geometry conference , yaroslavl , russia , august 10 - 14 , 1992 .
braunschweig : vieweg .
aspects math .
e 25 , 67 - 84 ( 1994 ) . | we show that an appropriate generalization of the _ oriented area function _ is a perfect morse function on the space of three - dimensional configurations of an equilateral polygonal linkage with odd number of edges
. therefore cyclic equilateral polygons ( which appear as morse points ) are interpreted as independent generators of the homology groups of the ( decorated ) configuration space . |
in wolf - rayet ( wr ) galaxies the presence of the broad heii @xmath7 4686 feature ( wr feature ) signals very recent massive star formation ( armus , heckman & miley 1988 ; conti 1991 ) .
wr galaxies therefore offer the unique possibility to study high mass star formation ( maeder & conti 1994 ) , since the progenitors of wr stars have initial masses @xmath8 20 m@xmath4 and the wr stars themselves are younger than 10 myr .
however , it is often unclear whether the wr line emission is originating from a single isolated hii region or is associated with the nucleus of the parent galaxy itself .
so far , 139 objects are listed in the new wolf - rayet ( wr ) galaxies catalog ( schaerer , contini & pindao 1999 ) , as wolf - rayet emission features are rarely observed in galaxies .
the study of wr galaxies is therefore essential for understanding the starburst phenomenon in galaxies , as it offers the unique opportunity to analyze the stellar population at a well - defined evolutionary stage .
+ ngc 6764 is a nearby ( 32 mpc for @xmath9=75 km s@xmath5 ; 1@xmath10=160 pc ) s - shaped barred spiral galaxy ( sbb ) , classified as a liner galaxy based on optical spectroscopy ( osterbrock & cohen 1982 ) .
ngc 6764 is unusual because it also displays a prominent 466 nm wolf - rayet emission feature at the nucleus ( osterbrock and cohen 1982 ) and was already in the first wr catalog of conti ( 1991 ) .
the galaxy contains a nuclear stellar optical continuum with a width of @xmath11 or about 260 pc ( rubin , thonnard , & ford 1975 ) .
+ eckart et al . ( 1991 ) presented single dish measurements of the j=1 - 0 and j=2 - 1 rotational transitions of @xmath12co and @xmath13co obtained with the iram 30 m telescope , as well as the first @xmath14 images of ngc 6764 .
eckart et al . ( 1996 ) presented x - ray observations , high resolution optical , millimeter - wave interferometer measurements of the 115 ghz @xmath12co(1 - 0 ) line emission , the first imaging observations of the 2.12 @xmath15 m h@xmath2 emission , and measurements of the 2.06 @xmath15 m he i and the 2.17 @xmath15 m br@xmath6 line emission toward the nucleus of ngc 6764 .
their nir and optical spectroscopic data which include the 466 nm wolf - rayet feature revealed a dense concentration of molecular gas in the center and a very recent ( few 10@xmath16 yr ) starburst at the nucleus of ngc 6764 .
this starburst is most likely responsible for the presence of a few thousand wolf - rayet stars in the nucleus of ngc 6764 ( eckart et al .
+ + in this paper we present the first @xmath0 band integral field spectroscopy and rosat x - ray observations using the high - resolution imager ( hri ) of the nucleus of ngc 6764 .
we describe our observations in 2 , and present an analysis of the x - ray data in 3 . in 4 we discuss the result from the integral field spectroscopy , and in 5 we analyze the nuclear star formation history of ngc 6764 .
we present a brief summary in 6 .
ngc 6764 was observed in the @xmath0 band ( 2.2 @xmath15 m ) on 1997 , july 26 and 28 with the mpe integral field spectrograph 3d ( weitzel et al . 1996 ) combined with the tip - tilt corrector rogue ( thatte et al .
1995 ) at the 3.5 m telescope in calar alto , spain .
the image scale was 0.5/pixel .
the total integration time on source was 2880 s. two fields off - set by 1.5 in east - west direction were combined to cover the nuclear region . since the seeing was different in both nights ( @xmath17 1 on july 26 and @xmath3 1.2 on july 28 ) , the data set from july 26 was convolved to the resolution of the second night for the combined data cube with a larger field of view . + 3d obtains simultaneous spectra for each pixel ( r@xmath18 750 for @xmath0 band ) of an 8 @xmath19 8 field .
this is done using an image slicer which rearranges the two - dimensional focal plane onto a long slit of a grism .
the dispersed spectra were then detected on a nicmos 3 array .
a detailed description of the instrument and the data reduction is given in weitzel et al .
the data reduction procedure converts each two dimensional image into a three dimensional data cube with two spatial and one spectral axes .
the data cubes were co - added and centered on the continuum peak .
all images were dark - current and sky - background subtracted , corrected for dead and hot pixels , and spatially and spectrally flat - fielded . to correct for the effects of the earth s atmosphere on the @xmath0 band spectrum ,
a standard star was observed .
this standard spectrum was first divided by a template spectrum of the same spectral type ( kleinmann & hall 1986 ) in order to remove stellar features .
the effect due to different zenith distance of the source and the standard star was minimized using the atran atmospheric model ( lord 1992 ) , mainly to correct for the different atmospheric absorption .
the source data were then divided by the atmospheric transmission spectrum .
+ we flux calibrated the data adopting the integrated @xmath0 band flux density value given by rieke ( 1978 ) in a 8.5 aperture centered on the nucleus . due to the different spatial resolution and s
/ n each data cube was flux calibrated individually before combination . a comparison to the line fluxes of eckart et al .
( 1996 ) in a 5 aperture showed good agreement within the errors for the hei @xmath72.06@xmath15 m and h@xmath2s(1 ) @xmath72.12@xmath15 m lines . for [email protected]@xmath15
m the obtained flux was by a factor of 2 higher .
this is probably related to the fact that the br@xmath6 line has a weak broad component ( fwzi @xmath3 2100 km / s ) .
ngc 6764 was observed using the hri detector ( david et al .
1996 ) on board the rosat satellite ( trmper 1983 ) between 1995 september 2 ( 10:23:45 ut ) and 1995 september 19 ( 10:23:38 ut ) .
the total exposure time was 44660 s and the source was centered on - axis in the field of view .
the centroid position of ngc 6764 in the hri image , computed from a gaussian fit to the spatial distribution is @xmath21 , @xmath22 .
the internal rosat hri pointing error is of the order of @xmath23 5 .
the source cell size was obtained from a radial profile of counts within a ring centered on the centroid position of ngc 6764 . for distances larger than 25 from the centroid position the radial profile is consistent with the background level .
the number of source plus background photons within a radius of 25 is 457 @xmath23 21 .
the background was determined in a ring around the source with an inner radius of 40 and an outer radius of 80 .
the number of background photons normalized to the source cell size is 109 @xmath23 10 .
the resulting number of source photons is therefore 348 @xmath23 25 corresponding to a mean count rate of @xmath24 .
our hri data show both the presence of an x - ray source that strongly varies on the time scale of several days and an additional extended x - ray component that looks similar to published vla radio continuum maps .
both findings confirm the composite nature of the center of ngc 6764 : the presence of a compact active nucleus and recent nuclear star formation ( see also gonalves , veron - cetty & veron 1999 ) . for the extragalactic sources studied by walter & fink ( 1993 ) with the rosat pspc the weighted mean photon index is 2.3 with an uncertainty of 0.03 . due to the limited photon statistics the pspc observations of the rosat all - sky survey of ngc 6764
do not constrain the spectral x - ray properties . therefore , we have used the spectral shape of a simple power - law model with an photon index @xmath25 of 2.3 and an absorbing column equal the galactic absorption by neutral hydrogen of @xmath26 towards ngc 6764 ( dickey & lockman 1990 ) .
we derive that the mean hri count rate of @xmath27 corresponds to an absorption - corrected 0.1 - 2.4 kev flux of @xmath28 ( see fig .
[ xxa01 ] ) . using equation ( 7 ) of schmidt & green ( 1986 ) ( with @xmath29 and @xmath30 )
we find that the mean hri count rate of ngc 6764 corresponds to an isotropic luminosity of @xmath31 or about 2.3@xmath1910@xmath32l@xmath4 fully consistent with the luminosity derived by eckart et al .
( 1996 ) . here
we assume that the radiation is isotropically radiated and that relativistic beaming effects are not relevant for the derived value of the luminosity .
kriss , canizares & ricker ( 1980 ) give an upper limit of the x - ray flux of ngc 6764 of l@xmath33@xmath1 8.6@xmath34ergs s@xmath5 over the ( @xmath35 kev ) energy band .
+ + an examination of the arrival times of the x - ray photons shows deviations from the mean count rate and suggests the presence of variability on very short time scales .
[ xxa01 ] displays the rosat hri light curve obtained for ngc 6764 between 1995 september 2 and 19 .
the exposure times for the individual data points range between 1575 seconds ( offset julian date 7.9 ) and 26520 seconds ( offset julian date 0.0 ) .
the integration time per bin is always larger than 1550 seconds to avoid apparent count rate variations due to the rosat wobble .
ngc 6764 shows count rate variations by a factor of about 2 on a time scale of 7 days .
a constant light curve model can be rejected with @xmath36 per cent confidence based on a @xmath37 statistic .
the continuous decrease of the rosat hri count rate starting at day 7.9 of the observations further accounts for true changes of the x - ray emission in the distant wolf - rayet liner galaxy ngc 6764 .
the presence of an extended x - ray component ( discussed in section [ soft ] ) that accounts for about half of the total luminosity suggests that the observed variabilitiy ( if entirely attributed to a nuclear source ) was significantly higher than a factor of 2 .
+ + the minimum count rate of @xmath38 was observed at day 1.5 within a 4760 second exposure time interval .
the maximum count rate is @xmath39 observed at day 7.9 within a 1575 s exposure interval .
the light curve of ngc 6764 shows a continuous decline of the count rate starting at day 7.9 up to the end of the observations .
such count rate variations suggest that a major fraction of the variable x - ray emission originates close to a super - massive black hole . assuming that the variability is not affected by beaming or relativistic motions the upper limit for the size of the variable x - ray emission is @xmath40 or only about 10@xmath41au .
the x - ray emission originating from ngc 6764 shows a complex structure : in addition to a source component ( see fig .
[ xxa02 ] ) at the nuclear position , we find an extended structure with indications for different centers of enhanced x - ray emission north - west and south - west of the central position .
the extended emission accounts for about half of the total x - ray luminosity in the 0.1 - 2.4 kev band derived from the mean count rate .
if parts of the extended emission are due to a soft x - ray spectrum , this suggests that the x - ray luminosity of this extended component is either due to bremsstrahlung emission or an emission line dominated spectrum . with the rosat hri and also the pspc observations we are not able to differentiate between bremsstrahlung models and an emission line dominated spectrum from hot diffuse gas based on the model calculations of raymond & smith ( 1977 ) .
the derived plasma temperatures of both models are constistent within the errors
. this would be in agreement with emission from a very diffuse component in addition to a contribution from compact sources indicated by our data and expected from supernovae , hot stars , and x - ray binaries .
eckart et al . (
1996 ) have shown that the x - ray luminosity could be entirely explained by the energy input of supernovae , and that the total mass of hot gas probably amounts to several 10@xmath42m@xmath4 .
+ starburst galaxies show typical values of 10@xmath43 to 10@xmath44 for their l@xmath33/l@xmath45 ratios ( boller 1999 ) .
the l@xmath33/l@xmath45 ratio for ngc 6764 lies within the range expected for starburst , if we use the x - ray luminosity for the extended component of l@xmath33 @xmath3 1.15 @xmath1910@xmath16 l@xmath4 and a total far - infrared luminosity of l@xmath45 @xmath3 2@xmath1910@xmath46 l@xmath4 . as discussed in section [ xxlbol ] about
40% of the far - infrared luminosity can be attributed to the nuclear starburst ( inner @xmath3 250 pc ) .
this leaves still enough luminosity for the extended component . therefore , we find that the l@xmath33/l@xmath45 ratio of the extended x - ray emission in ngc 6764 is in agreement with those found for starburst galaxies .
the spatial extent of the extented emission of @xmath3 10 fwhm ( @xmath3 1.6 kpc ) is similar to the extent of starburst rings of about 1 to 2 kpc .
+ we also note that the extended x - ray structure is similar to the structure found in the vla radio maps ( ulvestad , wilson , & sramek 1981 ; wilson & willis 1980 ) .
baum et al . ( 1993 ) discussed the origin of the radio emission suggesting strong star formation and due to the alignment of the radio emission with the minor kinematic axis pressure driven outflow .
a large - scale h@xmath47 map shows extended nuclear h@xmath47 emission over the inner 10
( d. frayer , private communication ) .
this suggests that the hot x - ray emitting gas and the synchrotron radiation emitting relativistic gas component are spatially coexisting and probably both are a result of strong recent circum - nuclear star formation .
the nuclear @xmath0 band spectrum is dominated by emission lines from warm molecular .
the slope of the continuum indicates no strong dilution from non - stellar sources like dust emission or agn power law contribution .
the @xmath0 band continuum map gives a source size of about fwhm 2.2 @xmath19 1.7
( pa @xmath3 95@xmath48 ) , larger than the seeing of 1.2 .
this implies that the @xmath0 band continuum source is extended with an intrinsic source size of about 1.7
@xmath19 1.2 ( pa @xmath3 95@xmath48 ) .
this is quite consistent with the july 26 data , where we obtained an uncorrected source fwhm of 1.8 @xmath19 1.3 ( pa @xmath3 95@xmath48 ) .
+ we made line maps for all the lines indicated in fig .
[ xxa03 ] by averaging the adjacent continuum channels and subtracting the resulting mean from all channels .
these line channels were subsequently co - added .
all fluxes and equivalent widths ( ews ) are measured in circular apertures , the corresponding errors are the standard deviations ( 1@xmath49 ) for the given apertures in the line maps ( tab .
[ xxt01 ] and [ xxt03 ] ) and do not include calibration uncertainties or systematic effects which we estimate to be of the order of 10% . all calculations for the stellar population synthesis are made for a 3 ( 480 pc ) aperture .
the br@xmath50 , br@xmath6 and hei line emission is spatially unresolved at our resolution .
the comparison of the fhwm in the line maps and the adjacent continuum channels shows clearly that the extent of the emission lines is more compact than the continuum emission ( fig .
[ xxa05 ] ) .
this indicates that the line emission is arising in a region considerably smaller than 150 pc ( about 1 ) .
the ratio of br@xmath50/br@xmath6 @xmath3 ( [email protected] ) is consistent with values of 0.65 - 0.67 expected for case b recombination ( osterbrock 1989 ) which suggests that extinction only plays a minor role at these wavelengths ( see also eckart et al .
+ assuming that the hei 2.06 @xmath15 m and br@xmath6 line emission is arising in hii regions around hot , young stars , the hei / br@xmath6 ratio can be used to estimate the mean effective temperature t@xmath51 of these ionizing stars .
we observe a ratio of 0.46 @xmath23 0.04 indicating t@xmath51 @xmath3 35 000 k , representative of an o8 star dominated ionizing stellar population .
note that we neglected effects like mixed - in dust and differences in the velocity structure of the strmgren spheres ( see lanon & rocca - volmerage 1996 , doherty et al .
since ngc 6764 is classified as a liner , we could expect some agn contribution to the br@xmath6 line emission , making the effective temperature a lower limit ( see also section [ xxsb ] ) .
this is consistent with the finding that about 66% ( 72% ) of the nuclear h@xmath47 ( h@xmath52 ) line emission which is coming from a narrow line component is of stellar origin ( i.e. consistent with hii region line ratios ; gonalves , veron - cetty & veron 1999 ) .
the 1 - 0 s(1 ) line is the strongest of all the six lines detected with a s / n @xmath53 3 in the @xmath0 band spectrum .
the 1 - 0 s(1 ) emission is extended in the east - west direction with a fwhm of @xmath3 3.3 ( 500 pc corrected for seeing ) in the line maps ( see also eckart et al .
the extent is even larger than that in the @xmath0 band continuum of 2.2 .
the maximum of the line emission is off - set by about 0.25
east from the continuum peak .
the 1 - 0 s(1 ) line has a fwzi of @xmath3 950 km / s .
line maps of the blue and red line wing ( centered @xmath23 270 km / s of the line center with widths of 400 km / s ) show a shift in the emission distribution from the south - west to the north - east indicating a rotating gas disk or ring .
the emission peaks of the blue and red line wings are about 1.2 ( 200 pc ) apart .
these observed kinematics and distribution of the warm molecular gas are similar to those of the cold molecular gas observed in its @xmath12co line emission ( schinnerer et al . in prep . ) .
the co absorption band heads at @xmath72.29 @xmath15 m are clearly visible in the spectrum ( fig .
[ xxa03 ] ) .
in addition the nai absorption at @xmath72.21 @xmath15 m is most likely detected , all those lines indicate the presence of cool , evolved stars .
the spatial extent of the stellar absorption lines coincides with the extent of the @xmath0 band continuum ( fig .
[ xxa05 ] ) , in contrast to the h and he recombination lines which are spatially unresolved at our resolution .
the situation therefore indicates that the hot , young stars and the cool , evolved stars are distributed differently . + + using the measured equivalent widths ( ews ) of the detected stellar absorption features and their ratios , it is possible to classify the mean stellar type ( e.g. , origlia , moorwood & oliva 1993 , oliva et al .
1995 , frster - schreiber 2000 ) .
the @xmath12co 2 - 0/nai ratio of ( [email protected] ) is consistent with effective temperatures t@xmath51 between 3000 k and 4800 k. the @xmath12co(2 - 0)/@xmath13co(2 - 0 ) ratio indicates a t@xmath51 @xmath54 4500 k , since these lines are vanishing for effective temperatures @xmath8 4500 k ( frster - schreiber 2000 ) . assuming that the total @xmath0 band continuum is due to cool stars and that the measured equivalent widths ( table [ xxt03 ] ) are not diluted , a comparison to ew of standard stars ( e.g. origlia , moorwood , oliva 1993 , schreiber 1998 ) suggests the presence of either late k supergiants or early m giants . under the assumption that there is a continuum flux density contribution from younger stars ( indicated by the hei and brackett emission ) and from the agn , the mean stellar type has to be even cooler since in this case the co band heads appear diluted and are therefore deeper than actually measured .
the cool , evolved stars can be either members of a few myr old starburst ( being mostly red supergiants , rsgs ) or part of an older stellar population ( being mostly red giants , rgs ) .
to investigate the nuclear star formation history further , we use the population synthesis code stars ( sternberg & kovo , private communication ) which was successfully applied to a number of galaxies ( ngc 1808 : krabbe , sternberg & genzel 1994 , ngc 7469 : genzel et al . 1995 , ngc 7552 : schinnerer et al .
1997 , ic 342 : bker et al .
1997 , circinus : maiolino et al .
1998 , m 82 : schreiber 1998 , ngc 3227 : schinnerer et al . 2000 ) in conjunction with the nir spectral synthesis code specsyn ( schinnerer et al .
1997 , 2000 ) .
+ the model is similar to other stellar population synthesis models ( e.g. lanon & rocca - volmerange 1996 , leitherer et al .
1999 ) and includes the most recent stellar evolutionary tracks ( schaerer et al .
1993 , meynet et al .
a summary of the systematic uncertainties in the population synthesis models is given in leitherer et al .
( 1999 ) ; for a discussion of the agb phase which is implemented in stars see oliva & origlia ( 2000 ) .
stars gives output observable parameters such as the bolometric luminosity l@xmath55 , the @xmath0 band luminosity l@xmath56,the lyman continuum luminosity l@xmath57 , and the supernova rate @xmath58 , as well as the diagnostic ratios between these quantities : l@xmath55/l@xmath57 , l@xmath56/l@xmath57 and 10@xmath59/l@xmath57 .
all three ratios are measures of the time evolution and the shape of the imf , with slightly different dependencies on the exponential slope @xmath47 and the upper mass cut - off m@xmath60 .
hertzsprung - russell diagrams ( hrds ) representing the distribution of these luminosities are calculated .
specsyn uses the distribution of the @xmath0 band luminosity l@xmath56 within a hrd ( from stars ) to weight standard star spectra of different spectral type and luminosity ( schinnerer et al .
1997 , 2000 ) .
in addition , effects like extinction or non - stellar contributions from dust emission or an agn can be included as well . to estimate _
l@xmath55 _ for the central 3
( @xmath3 500 pc ) of ngc 6764 we followed the approach of eckart et al .
( 1996 ) . for starburst regions the fir luminosity can be assumed to be equivalent to the bolometric luminosity ( telesco , dressel & wolstencroft 1993 ) . under the assumption that the relative contribution from the star forming regions is similar at all the iras wavelengths , we scaled the iras fluxes by 0.4 at 12 @xmath15 m ( 0.36 jy ) , 25 @xmath15 m ( 1.29 jy ) , 60 @xmath15 m ( 6.33 jy ) and 100 @xmath15 m ( 11.56 jy ) accordingly to the measured 10.6 @xmath15 m flux in a 5 aperture ( rieke 1978 ) . since the 10.6 @xmath15 m measurement of rieke ( 1978 ) and the iras 12 @xmath15 m measurement have similar filter width of @xmath3 5 @xmath15 m and the strong absorption and emission features in the mid - ir spectrum of ngc 6764 ( rigopoulou et al .
1999 ) are covered in both measurements , this approach seems justified to obtain the nuclear mid- and far - ir emission . to derive _
l@xmath55 _ the relation given by sanders & mirabel ( 1996 ) was used .
this approach was examined in the case of ngc 7552 , there similar values for l@xmath55 were obtained by using different methods ( schinnerer et al .
+ _ l@xmath61 _ and _ l@xmath57 _ were estimated via the equations given in genzel et al .
( 1995 ) ( using a @xmath62 = 0.6@xmath15 m as the @xmath0 band width ) and taking the values measured with 3d in a 3 aperture . for _
_ we used the empirical relation between the 5 ghz flux and _ @xmath58 _ given by condon ( 1992 ) . to derive the actual value
, we only used the nuclear 5 ghz flux of 3.4 mjy from baum et al .
( 1993 ) , since the h / he recombination lines as well as the stellar absorption lines are smaller in extent than the nuclear radio source size .
the derived quantities are listed in tab .
[ xxt04 ] .
( all equation are described in schinnerer , eckart , & tacconi 1998 . )
eckart et al . ( 1996 ) found that the wr feature at 4660 ( niii @xmath74640 and heii @xmath74686 ) is spatially extended .
however , the 3d hei and br@xmath6 line maps show no indication for extended emission which may be due to our sensitivity . the optical line emission is mainly due to n - rich wr stars ( subtype wn ) ( eckart et al .
1996 ) . in their recent optical data
kunth & conti ( 1999 ) were able to identify a wr feature at ciii @xmath75696 and civ @xmath75808 arising from c - rich wr stars ( subtype wc ) .
+ in the @xmath0 band , the and lines at 2.08 @xmath15 m and 2.11 @xmath15 m , respectively , are expected from wc stars ( figer , mclean , & najarro 1997 ) which are not detected in the 3d data at our sensitivity .
we see indication for the hei / heii line complex at 2.16 @xmath15 m ( fig .
[ xxa03 ] and [ xxa06 ] ) indicative of the presence of wn stars ( figer , mclean , & najarro 1997 ) .
since in late wn stars this line complex is almost as prominent as the 2.06 @xmath15 m hei line ( figer , mclean , & najarro 1997 ) , this implies that most of the line emission seen at 2.06 @xmath15 m is due to hii regions and not wr stars .
+ population synthesis results of wr populations show that considerable numbers of wn stars ( in addition to wc stars ) are present after about 3 myrs ( schaerer & vacca 1998 ) . at about 5 myrs
wn stars clearly dominate the wr star population and the strength of the @xmath75808 line ( indicating wc stars ) decreases whereas the wr features at @xmath74660 ( indicating wn stars ) start to increase significantly .
the presences of both wn and wc stars therefore suggests an age between 3 and 5 myrs for the nuclear stellar population in ngc 6764 associated with the wr stars . in our starburst model calculations , we used a salpeter imf with a lower mass cut - off of 1 m@xmath4 and an upper mass cut - off of 120 m@xmath4 .
the wr stars signal that massive star formation has occured recently or is still ongoing . on the other hand , an older stellar population that produces most of the @xmath0 band continuum must be present .
therefore only two simple scenorios can be considered producing the observed properties : ( 1 ) two starburst events describing discontinuous star formation due to episodic fueling with molecular gas and ( 2 ) continuous star formation assuming continuous fueling with molecular gas and/or a large nuclear reservoir of gas .
+ in a first step we tried to fit the diagnostic ratios for the two different scenarios . when the predicted l@xmath61 and l@xmath57 luminosities were matched to the observed values .
the spectral synthesis was used to address the non - stellar properties of the nuclear region such as extinction , dust emission or agn continuum contribution which also effect the overall values .
however , the derived values are rather small : for the spectral synthesis fitting ( which is independent of l@xmath61 , since it is tied to the calibrated spectrum ) , we tried apparent extinction values of @xmath63=1@xmath64 , 2@xmath64 , 3@xmath64 , 4@xmath64 and 5@xmath64 .
for the two star formation scenarios considered here , we find an extinction of ( 2 - 3)@xmath64 ( mainly affecting the slope of the @xmath0 band spectrum at shorter wavelength end ) combined with a @xmath3 5% non - stellar contribution to the @xmath0 band continuum which is either due to emission from warm dust ( @xmath65 @xmath3 500 k ) or agn power law emission ( affecting the slope of the @xmath0 band spectrum at the longer wavelength end ) .
we used the mean flux density at large radii ( r @xmath3 3.5 ) to estimate the contribution of the underlying old bulge population to about 10 % of the @xmath0 band continuum flux density in a 3 aperture .
however , the bulge population has a much larger spatial extent than the stellar cluster mapped in its stellar absorption lines .
+ both star formation possibilities are explored in the following .
the model parameters and results of the star forming events for the different scenarios are given in tab .
[ xxt04 ] .
as discussed in section [ xxwr ] the wr stars have an age of about 3 - 5 myr .
this low age immediately implies a second starburst or star formation event that is responsible for the cool evolved stars indicated by the stellar absorption features ( s. section [ xxco ] ) .
+ + _ starburst about 3 - 5 myr ago ( sb#1 ) : _ assuming a starburst event with a decay time of 3 myr seems reasonable given the fact that sn explosions might disturb the ism and prevent further star formation . for a starburst in that age range , stars gives a mean ionizing hot star of type o7 to o5 with a mean effective temperature of t@xmath51 @xmath66 40000 - 45000k .
neglecting dilution effects ( s. section [ xxheh ] ) this corresponds to an hei / br@xmath6 ratio of 0.7 ( lanon & rocca - volmerage 1996 ) .
a comparison of @xmath0 band spectra from galaxies with agn and pure starburst galaxies ( vanzi et al .
1998 ) suggests that a strong hei emission line is only present in the pure starburst galaxies without an agn .
this probably indicates that most of the hei line flux observed towards the nucleus of ngc 6764 is due to young , hot stars and not due to an agn component . under the assumption that all hei line flux is associated with the starburst , we can correct the observed hei / br@xmath6 ratio of [email protected] to obtain the br@xmath6 line flux associated with the starburst event . to fit the model ratio of 0.7 about 65% of the observed br@xmath6 line flux
has to be due to the 5 myr old starburst .
this finding is in excellent agreement with the value of 66% for the h@xmath47 line emission of gonalves , veron - cetty & veron ( 1999 ) . by fitting l@xmath57
, we obtained the contibutrions of this starburst to l@xmath61 , l@xmath55 and @xmath58 . the remaining differences in these quantities have to be attributed to the second starburst ( or star formation ) event and are used for the further model fits .
+ + _ starburst about 25 myr ago ( sb#2 ) : _ the first starburst contributes only a few percent to the @xmath0 band luminosity , and the old bulge population delivers about 10 % ( see section [ xxold ] ) .
therefore , this second starburst has to account for about 85% of the total l@xmath61 and the total amount of @xmath58 without further large contributions to l@xmath57 .
a @xmath3 15 myr old starburst can account for most of the remaining l@xmath57 .
however , if the age is about 25 myr almost all of the remaining 35% of the br@xmath6 line flux could be associated with the liner nucleus ( s. table [ xxt04 ] ) . at an age of 50 myr ,
the contribution of this starburst to the lyman continuum is already negligible and only small to the nuclear radio continuum .
we also tested ages of @xmath3 1 gyr where agb stars have a prominent contribution to the @xmath0 band continuum ( lanon 1999 , lanon et al .
1999 ) . this possibility can be ruled out , since at that age the radio emission could no longer be explained by star formation , i.e. synchrotron emission from sn remnants .
the age for this second starburst event can not be so strict as in the case of the wr starburst , since it is not possible to absolutely constrain the contribution of this starburst to the @xmath0 band luminosity , the lyman continuum luminosity and the sn rate .
however , its age has to be larger than 15 myr and below 50 myr which is well below 1 gyr .
also , an underlying low - level constant star formation is impossible , since such a population can not produce the l@xmath61 needed without large contributions to l@xmath57 ( see section [ xxcs ] ) .
+ + this scenario is quite opposite to the one found in ic 342 where a 70 pc diameter starburst ring with an age of about 5 myr surrounds a nuclear starburst of about 15 myr ( bker , frster - schreiber & genzel 1997 ) . in ngc 6764 the younger component ( _ sb#1 _ ) is concentrated in the nucleus and surrounded or embedded in the older star formation event ( _ sb#2 _ ) .
higher angular resolution would allow us to investigate if the _ sb#1 _ component is indeed concentrated on the nucleus or is distributed in a small ring like in ic 342 .
a comparison of the contribution of the two starburst events to the optical continuum shows that the wr burst contributes about 15% to the synthesized stellar @xmath67 band continuum .
this is in agreement with the 11% continuum contribution derived by osterbrock & cohen ( 1982 ) .
if we assume continuous star formation in the nuclear region the nuclear stellar population must have an age of @xmath53 1 gyr , since only then enough cool evolved stars have been produced to obtain the observed ratio of l@xmath56 to l@xmath57 .
it is necessary to reach at least the agb phase for a reasonable fit , since only then enough of the observed cool luminous stars are present .
this analysis is , however , hampered by the fact , that the evolutionary tracks of the high mass stars do not produce very cold red supergiants ( e.g. of type m ) and that specsyn does not use spectra of agb stars .
the spectra of agb stars are approximated by those of a red supergiant ( rsg ) of type m4iab .
this may affect the slope at the beginning of the @xmath0 band , however , specsyn spectra start only at 2.0267 @xmath15 m there this effect is already weak . since only a small fraction of the total @xmath0 band light is due to agb stars ,
this is neglegible and for the purpose of the nir spectral synthesis sufficient .
+ the derived star formation rate ( sfr ) of this scenario is @xmath3 0.3 m@xmath4/yr . for an age of the stellar population of 1 gyr this translates into about 3@xmath1910@xmath68 m@xmath4 of molecular gas that have been transformed into stars .
the average star formation effiency ( sfe ) in giant molecular cloud complexes ( gmcs ) is of the order of a few percent ( duerr et al .
1982 , tenorio - tagle & bodenheimer 1988 ) , up to values of 30 - 40 % only in the densest cores of these complexes ( lada 1982 ; see summary article by lada , strom & myers 1993 ) . if we assume an average value of up to 10% for molecular clouds in starburst environments , the mass inflow rate needed is about 3 - 6 m@xmath4/yr .
bars are relative efficient in transporting gas down to small radii , jogee , kenney & smith ( 1999 ) found a mass inflow rate of about 1 m@xmath4/yr for the central starburst in ngc 2782
. however , eckart et al . (
1991 ) estimated a total molecular gas mass of 2 - 6 @xmath1910@xmath68 m@xmath4 for ngc 6764 . together with the fact that bars are short - lived features in the evolution of galaxy ( @xmath3 10@xmath68 yrs ; combes 1998 )
, such a high inflow rate over 1 gyr seems highly unlikely . +
another aspect of this scenario is that the wr emission ( hei and br@xmath6 line ) and the stellar absorption bands should have similar distributions even at highest spatial resolution .
however , this is already in contradiction to the observed spatial distribution of the h and he recombination lines ( unresolved ) versus the extent of the stellar absorption lines ( similar to @xmath0 band continuum ) .
it might be a possible scenario that due to dispersion of the nuclear stellar cluster the older stars have moved to larger radii .
a comparison to the upper limit of @xmath3 3@xmath1910@xmath69 m@xmath4 for the total mass in the inner 5
( eckart et al . 1991 ) shows that the converted gas mass is only a small fraction ( @xmath3 10% ) and therefore we expect the gravitational potential to minimize such an effect .
in addition , a continuous star formation would account for the total h recombination line emission without any emission arising from the agn .
therefore continuous star formation is higly unlikely , although the spectral fit ( see fig . [ xxa07 ] ) is equally satisfying .
ngc6764 shows strong variations in its x - ray flux density by at least a factor of 2 on time - scales of 7 days .
this suggests the presence of a compact agn with an upper size estimate of @xmath40 or only about 10@xmath41au .
in addition there is evidence for an extended and possibly soft diffuse component of the x - ray emission .
the hot x - ray emitting gas and the synchrotron radiation emitting relativistic gas component are spatially coexisting and are probably both the result of strong recent nuclear star formation .
+ nir integral field spectroscopy of the nuclear region in the wr liner galaxy ngc 6764 reveals that the nuclear star formation is probably confined to two areas , one of less than 100 pc ( wr stars ) and the other with an extent of @xmath3 200 pc ( rsgs ) : the wr stars and the evolved cool stars are not co - spatial and the younger starburst resides inside the older starburst .
this picture is quite contrary to what is observed in many starburst galaxies where the younger component is arranged in a ring around the older nucleus ( e.g. ngc 7552 : schinnerer et al .
1997 , ic 342 : bker , frster - schreiber & genzel 1997 ) .
+ application of a population synthesis in conjunction with nir spectral synthesis infers an extinction towards the nuclear stellar cluster of about ( 2 - 3)@xmath64 in agreement with the observed br@xmath50/br@xmath6 ratio and earlier findings by eckart et al .
in addition we see evidence for a 5% non - stellar contribution to the @xmath0 band continuum either from warm dust or a power law contribution from the agn itself . in a 3 aperture
the @xmath0 band contribution of the bulge population is about 10% .
+ the nuclear star formation history allows for two simple possibilities : ( 1 ) two starbursts with ages of 3 to 5 myr and between 15 and @xmath1 50 myr and decay times of 3 myr which produce the wr stars and red supergiants that contribute to a large amount of the @xmath0 band continuum .
( 2 ) continuous star formation with a sfr of @xmath3 0.3 m@xmath4/yr for at least 1 gyr .
+ in the case of the two starburst scenario , an analysis of the h and he recombination lines shows that about 65% of the br@xmath6 line emission is associated with the young starburst , leaving the rest for the liner nucleus or / and the second starburst event .
comparison of our data and data from the literature with population synthesis results for wr dominated cluster suggests an age of 3 - 5 myr for the wr component in ngc 6764 .
the age for the second starburst event is in the range of 15 myr to well below 1 gyr depending on the contribution of this starburst to the lyman continuum and the radio continuum observed .
+ the continuous star formation scenario seems highly unlikely given the larger amount of molecular gas needed to be transported down to radii of @xmath3 160 pc over 1 gyr .
in addition , the spatial distribution of the line emission is expected to be cospatial with the stellar absorption lines - this is not observed .
even if we allowed for an unresolved 35% agn contribution to the br@xmath6 such a case seems unlikely , because there is no indication for extended emission on a lower level in the radial profile of the br@xmath6 line emission .
therefore , the most likely scenario is that two recent starbursts have occured in the nuclear region of ngc 6764 .
+ the nucleus of ngc 6764 exhibits an interesting star formation scenario , since the younger component seems more compact than the older one .
one explanation could be that the bar is transporting the molecular gas very close to the nucleus ( @xmath1 50 pc ) . in that case
the star formation might directly compete with the agn for the fuel
. it may also influence or even control the agn activity .
the presence of a compact agn ( from the x - ray data ) and violent recent nuclear star formation underline the composite nature of the nucleus of ngc 6764 .
we like to thank the mpe 3d group , especially n. thatte , j.f .
gallimore , m. tecza and n.m .
frster - schreiber , for their help taking the data .
we also thank the calar alto 3.5 m telescope staff for hospitality and support .
we are grateful to a. sternberg for providing us with the population synthesis code stars .
we also want to thank the anonymous referee for her / his valuable comments which helped to improve the paper .
boer , b. , schulz , h. , 1989 , in extranuclear activity in galaxies , eso workshop , held in garching , f.r.g . , may 16 - 18 , 1989 , ed .
meurs and r.a.e .
fosbury ; publisher , european southern observatory , garching bei mnchen , p. 299
the observed values are derived assuming a@xmath70=2@xmath64 and using the following measured values in a 3 aperture : s@xmath56=6.74mjy , s@xmath71=3.4mjy , f@xmath72=64.0@xmath1910@xmath73 ergs cm@xmath43 .
the equations can be found in schinnerer , eckart & tacconi ( 1998 ) .
m@xmath75 and m@xmath76 denote the present - day stellar mass and the consumed gas mass to date , respectively .
+ all starburst models are calculated using a salpeter imf with mass cut - offs of 1 m@xmath4 to 120 m@xmath4 .
+ sb # 1@xmath77 : age of 3 myr and decay time of 3 myr ( wr stars ) .
+ sb # 1@xmath78 : age of 5 myr and decay time of 3 myr ( wr stars ) .
+ sb # 2@xmath79 : age of 15 myr and decay time of 3 myr ( rsgs ) .
+ sb # 2@xmath80 : age of 25 myr and decay time of 3 myr ( rsgs ) .
+ sb # 2@xmath81 : age of 50 myr and decay time of 3 myr ( rsgs ) .
age @xmath53 1 gyr and continuous star formation . | near - infrared @xmath0 band imaging spectroscopy of the central 8 ( 1.3 kpc ) in the wolf - rayet liner galaxy ngc 6764 shows that the most recent star formation is most likely still unresolved at sub - arcsecond resolution ( @xmath1 100 pc ) .
the @xmath0 band continuum source has a size of about 1.5 ( 240 pc ) .
in addition to stellar co and na absorption lines as well as recombination lines of h and he , the @xmath0 band spectrum shows several strong emission lines from molecular hydrogen ( ) .
the h@xmath2 line emission is spatially and spectrally resolved showing a rotating ring / disk of @xmath3 1.2 ( 200 pc ) diameter .
+ an analysis of the nuclear ( 3 @xmath3 480 pc ) stellar light using population synthesis models in conjunction with nir spectral synthesis models suggests following star formation history : two starbursts with decay times of 3 myr occurred 3 to 5 myr and 15 to about 50 myr ago . continuous star formation with a sfr of @xmath3 0.3 m@xmath4yr@xmath5 over at least 1 gyr can also explain the observed parameter .
however , the mass relocation and consumption involved as well as the different spatial distribution of the lines associated with the star formation strongly favor the two starburst scenario . in that scenario ,
up to 35% of the total observed br@xmath6 flux could still be due to the agn , depending on the assumed age of the older starburst event .
in contrast to other starburst galaxies ( e.g. m82 , ngc 7552 ) , the younger starburst in ngc 6764 is very likely located closer to the nucleus and surrounded by the older starburst .
one possible explanation can be that the stellar bar still transports gas down to radii close to the nucleus .
this suggests that the massive star formation activity is directly competing with the agn for the fuel .
+ we also present the results of a 44 ksec .
hri rosat exposure .
the hri data show the presence of a x - ray source ( probably an agn ) which varies by more than a factor of 2 over a time scale of 7 days .
this implies the presence of a compact source with a discrete or at most 1000 au source size .
in addition , we find an extended x - ray component which looks similar to the radio continuum emission presented in published vla maps .
both data sets confirm the composite nature of the center of ngc 6764 : the presence of a compact agn as well as recent violent nuclear star formation . |
inflation @xcite , a rapid accelerated expanding phase of the universe in its very early stage of evolution , provides a mechanism of generation of tiny density fluctuations over the homogeneous and isotropic background , which later evolve into large scale structures like galaxies and clusters of galaxies in the universe .
it also explains the observed nearly scale invariant spectrum of these density fluctuations on superhorizon scales .
parameters , like scalar amplitude , scalar spectral index and tensor - to - scalar ratio , predicted by inflationary paradigm are being very accurately determined by the high - precision cmbr observations , like wmap and planck .
the recent planck-2015 data predicts the scalar amplitude and the scalar spectral index as @xmath2 and @xmath3 respectively at ( @xmath4 cl , plancktt+lowp ) @xcite .
recently bicep2/keck array cmb polarization experiments combined with planck analysis of cmb polarization and temperature data have put a bound on tensor - to - scalar ratio as @xmath5 cl ) @xcite .
planck also confirms that the primordial perturbations , to a good approximation , are indeed gaussian in nature by constraining the primordial non - gaussianity ( ng ) amplitude as @xmath6 @xcite .
there is a plethora of inflationary models which explain the key inflationary parameters like scalar amplitude , scalar spectral index and tensor - to - scalar ratio as observed in the cmb measurements and also account for the smallness of the non - gaussianity parameter simultaneously .
but it is well known that the simplest single field slow - roll inflation models with quartic @xmath7 and quadratic @xmath8 potentials are ruled out from the observations as they produce large tensor - to - scalar ratio @xmath9 and @xmath10 , respectively , for @xmath11 e - foldings @xcite .
one novel way of making the quartic potential of inflaton viable is through what is now known as the higgs inflation scenario @xcite , where the inflaton field @xmath12 is non - minimally coupled to the curvature scalar @xmath13 ( coupling term looks like @xmath14 ) .
this gives rise to very small tensor - to - scalar ratio @xmath15 for @xmath16 .
such a model however encounters the problem of unitarity violation in higgs - higgs scattering via graviton exchange at planck energy scales because of very large curvature coupling @xmath17 required to obtain the observed cmb amplitude @xcite .
starobinsky model of inflation which is @xmath18 correction to einstein gravity is mathematically equivalent to higgs inflation model with quartic potential and therefore it produces the similar small tensor - to - scalar ratio @xmath15 @xcite . on the other hand ,
generalized non - minimally coupled models with coupling term @xmath19 and quantum corrected quartic potential @xmath20 , which are equivalent to power law inflation model @xmath21 , are studied in @xcite which produce large @xmath22 , and thus are disfavored by present status of the cmb data @xcite . in the context of upcoming experiments like keck / bicep3 and cmbpol , it would be of interest to come up with models which predict @xmath23 close to the current bound of @xmath24 . from the theoretical perspective it would be desirable to examine scenarios where the @xmath7 or @xmath8 potentials , which generically occur in particle physics , are compatible with cmb observations . in this work we study a two - field inflationary model where the inflaton field @xmath12
is assisted by a dilaton field @xmath25 and has a non - canonical kinetic term due to the presence of a dilaton field .
supergravity theories which are low energy limit of string theory contains several scalar fields which can be of cosmological interest .
the action of the model can be generically written as s & = & 12d^4 x,[action - sigma - einstein2 ] where @xmath26 and @xmath27 are arbitrary independent parameters .
brans - dicke ( bd ) gravity in einstein frame ( ef ) is a special case where @xmath28 @xcite
. however bd gravity predicts @xmath23 larger than the observed limit , therefore generalization of bd theory is necessary for application to inflation . in this paper
we generalize the bd theory in ef to a two - parameter scalar - tensor theory where we treat @xmath26 and @xmath27 as two independent arbitrary parameters .
addition of one extra parameter allows us to obtain viable inflation with otherwise ruled out quadratic and quartic potentials as we can have tensor - to - scalar ratio @xmath23 in the range of interest for forthcoming experiments . to note
, it was shown by ellis et al .
@xcite that the inflaton field accompanied by a moduli field @xmath29 , which appear in string theories and have a no - scale supergravity form , give a potential for inflation equivalent to the @xmath30 starobinsky model , producing @xmath31 .
we show in this paper that the above mentioned two parameter scalar - tensor theory can be obtained from no - scale supergravity theories @xcite which now can produce @xmath23 much larger than @xmath32 and thus are observationally falsifiable by future experiments . also in contrast to the supergravity embedding of the starobinsky model , studied in @xcite , where the imaginary part of the superfield @xmath29 @xmath33 axion @xmath34 decreases rapidly and its real part @xmath33 dilaton @xmath35 drives the inflation , we will see that in our model dilaton - axion pair evolves sufficiently during inflation and the axion acts as the inflaton .
first we look at the background dynamics of our model . starting from the action
( [ action - sigma - einstein2 ] ) , the equations of motion of the fields @xmath12 and @xmath25 and the friedmann equations can be obtained as @xmath36 where an over dot represents derivatives w.r.t .
time and prime denotes derivative with respect to @xmath12 . in the slow - roll regime
when both the fields slow - roll , terms with double time derivatives can be neglected and therefore the background equations reduce to @xmath37 here the full potential @xmath38 can be regarded as the product of potentials of the individual fields , @xmath39 and @xmath40 , and thus we define the slow - roll parameters for both the fields in a usual way ( following @xcite ) : @xmath41 where @xmath42 . to ensure the smallness of the slow - roll parameters we demand that the hubble slow - roll parameter @xmath43 during inflation .
we notice that @xmath44 which implies that @xmath45 and @xmath46 during inflation . again , taking a time - derivative of the second equation of ( [ phidot ] ) to obtain @xmath47 , one obtains @xmath48 , which implies that @xmath49 and @xmath50 ( as @xmath26 is to be taken smaller than unity ) .
we would show later on that the dilaton field @xmath25 evolve slower than the inflaton field @xmath12 throughout the inflationary phase which would enable us to treat @xmath25 as a background field .
furthermore , we would require the initial field values to calculate the inflationary observables such as @xmath51 , @xmath23 and @xmath52 .
from the first equation of ( [ hdot ] ) , we notice that @xmath53 , \label{phi - evo } % \phi^2&=&\phi^2_0+\frac{16}{\beta^2}e^{\frac{\beta\sigma_0}{2}}\left[1-\left(\frac{a}{a_0}\right)^{\frac{\beta^2}{2}}\right ] \label{sigma}\end{aligned}\ ] ] where subscript @xmath54 indicates the values of the quantities 60 e - foldings prior to end of inflation . defining @xmath55 and requiring @xmath56 ( subscript @xmath57 denoting the quantities at the end of inflation ) for sufficient inflation
, one gets @xmath58
\gtrsim \delta n. \label{efolds}\end{aligned}\ ] ] the perturbation analysis of such a model has been extensively discussed in @xcite , where the comoving curvature perturbation is defined as @xmath59 where @xmath60 is the scalar metric perturbation in the longitudinal gauge . as we are dealing with a multi - field inflationay model
, @xmath61 is not a frozen quantity on superhorizon scales and its time evolution is given by = + , where @xmath62 represents isocurvature ( or entropy ) perturbations given by -0.3 cm = ( - ) .
[ iso - pert ] under slow - roll approximation , one can solve for the scalar perturbations of the model , @xmath63 , @xmath64 and @xmath60 , on superhorizon scales , which turn out to be @xcite @xmath65 , \label{super - horizon - eqns3}\end{aligned}\ ] ] where @xmath66 and @xmath67 are the time independent integration constants and can be fixed using initial conditions . in the above expression ( [ super - horizon - eqns3 ] ) , terms proportional to @xmath66 represent the adiabatic modes while those proportional to @xmath67 represent the isocurvature modes . using eq.s ( [ phidot])-([slowparameters ] ) , the comoving curvature perturbation ( [ comov - curv - pert ] ) can be simplified to the form : @xmath68 since from eq.([super - horizon - eqns3 ] ) , it is clear that all the terms in @xmath60 are proportional to @xmath69 slow - roll parameters , therefore we will ignore the potential @xmath60 compared to @xmath66 and @xmath67 in @xmath70 . from eq .
( [ super - horizon - eqns ] ) , we can calculate @xmath66 and @xmath67 . substituting @xmath66 and @xmath67 ,
the comoving curvature perturbation on super horizon scales becomes @xmath71 where @xmath72 and @xmath73 .
now we look at the observables predicted by this model .
the mode functions for superhorizon fluctuations of @xmath25 and @xmath12 , evaluated at horizon crossing @xmath74 , are @xmath75 and @xmath76 .
therefore , the power spectrum of comoving curvature perturbations becomes @xmath77 the tensor power spectrum retains its generic form , because the action involves only minimal einstein curvature term @xmath13 , given by @xmath78 , which yields the tensor - to - scalar ratio as @xmath79 the scalar spectral index @xmath51 in this model is -0.3 cm n_s-1a -b ^2.[ns ] + it can be noted that in the limit @xmath80 and @xmath81 ( i.e. @xmath82 and @xmath83 ) the forms of power spectrum , tensor - to - scalar ratio and spectral index reduce to their standard forms in the single field slow - roll inflation .
it is important to note that in our model the amplitude of the isocurvature perturbations , @xmath84 ^ 2}{e^{(2\beta-\gamma)\sigma}(\dot\sigma^{2 } + e^{-\gamma\sigma } \dot\phi^2)^3 } , \label{powers}\end{aligned}\ ] ] vanishes in the slow - roll approximation ( using ( [ phidot ] ) ) .
the multifield models with non - canonical kinetic term posses a strong single - field attractor solution @xcite as has also been observed in this case .
but generally these multifield models produce isocurvature perturbations which can also account for a large angular scale suppression of the power spectrum .
but our case differs from such multifield models as it does not produce any isocurvature perturbations upto slow - roll approximation .
in our model , we would treat the @xmath12 field as the inflaton field which is assisted by the dilaton field @xmath25 during the inflationary evolution .
this can only be ensured if we confirm that the @xmath25 field evolves slower than the @xmath12 field during the entire inflationary epoch . to show this
, we first study the background evolution of both the scalar fields @xmath12 and @xmath25 by numerically solving the field eq.s ( [ fullsigmadot]-[fullhdot ] ) .
we first treat the case when the inflaton field has quadratic potential @xmath85 . as some representative initial conditions , we choose @xmath86 and @xmath87 ( solid ) , @xmath88 ( dotted ) , @xmath89 ( dashed ) corresponding to @xmath90 , @xmath91 , @xmath92 respectively .
also we fix @xmath93 for each case , which is required for the sugra derivation of this model studied in the later part of this paper .
the initial conditions are chosen carefully such that we get correct @xmath51 and @xmath23 for @xmath94 e - folds . in fig .
[ fig1 ] , we show the time evolution of the fields @xmath12 and @xmath25 where time is given in the units of @xmath95 . the different colors in the figure correspond to different initial conditions as described above . in fig .
[ fig2 ] ( upper panel ) shows the evolution of the fields in @xmath96 plane .
this plot shows that during @xmath11-efolds inflation , dilaton @xmath25 evolves much slower compared to inflaton @xmath12 .
after the end of inflation , inflaton goes to its minimum value @xmath97 . such background evolution of the fields also ensure that the background spatial metric evolves ( quasi-)exponentially during inflation which has been depicted in the lower panel of fig .
also we checked that for the case of quartic potential @xmath98 , the fields evolve in a similar way ensuring that @xmath12 can be treated as an inflaton field .
we now analyze the observable parameters for inflaton potential @xmath99 . from eq .
( [ sigma ] ) , we find @xmath100 .
using @xmath101 , which is the condition for the end of inflation , we obtain @xmath102 . from eq.([efolds ] ) , the field value @xmath103 can be expressed in terms of @xmath104 and @xmath105 as @xmath106 now we substitute @xmath103 into eq.s ( [ powerr1 ] ) , ( [ tensor ] ) and ( [ ns ] ) to give @xmath51 , @xmath23 and @xmath107 in terms of @xmath105 , @xmath108 , @xmath109 , @xmath26 and @xmath27 . for @xmath110 e - folds and for the choice @xmath111 with various choices of the parameters @xmath26 and @xmath27 ,
the @xmath112 predictions for quadratic @xmath113 and quartic @xmath114 potentials are shown in the fig .
[ fig3 ] . for @xmath86 , @xmath110 and for the range of the parameter values of ( @xmath115 ) as shown in fig .
[ fig3 ] , we find inflaton mass in the range @xmath116 and self - coupling in the range @xmath117 .
e.g. for the choice @xmath118 , @xmath119 , which can produce @xmath120 , @xmath121 , gives @xmath122 . and for @xmath123 , @xmath124 , which produces @xmath125 , @xmath126 , gives @xmath127 .
therefore , in this model with quadratic and quartic potentials , similar to the case of single - field slow - roll inflation , we require light inflaton mass and fine - tuning of the inflaton self - coupling in order to fit the observed cmb amplitude .
however , unlike the higgs inflation which predicts very small @xmath128 and standard single - field inlation with quadratic and quartic potentials give large @xmath23 , the two field model can give @xmath23 close to the present bound @xmath24 .
for the above mentioned initial conditions , coupling constants and parameter values , the running of the spectral index @xmath129 comes out to be @xmath130 and @xmath131 for quadratic and quartic potentials , respectively , fully consistant with the planck observation @xmath132 @xcite .
besides yielding @xmath23 , @xmath51 and @xmath133 within the observational bounds , a viable inflationary model should not produce large non - gaussianity ( ng ) to remain in accordance with observations .
ng in multifield models where the fields have non - canonical kinetic terms has been calculated in ref.s @xcite .
following these ref.s , we calculated the non - linearity parameter @xmath134 which characterizes the amplitude of ng .
we find that for the range of parameters values as shown in fig .
[ fig1 ] , @xmath135 consistent with the observations .
also we find that @xmath136 does not depend on initial value of the dilaton @xmath105 and coupling constants for the considered chaotic form of potential .
in this section we would show that , such a two - field inflationary model can be realised in the realm of no - scale supergravity .
the two - field models of inflation with string motivated tree - level no - scale khler potential in no - scale supergravity framework are analyzed in @xcite .
the f - term scalar potential in ef is determined from khler function given in terms of khler potential @xmath137 and superpotential @xmath138 as @xmath139 , where @xmath140 are the chiral superfields . in the supergravity action ,
potential and kinetic terms in ef are given by -0.2 cm v = e^g , _ k = k_i^j * _ ^i ^^*_j , [ lvlk ] where @xmath141 is the inverse of the khler metric @xmath142 .
we consider the khler potential of the following form k=-3 + , [ kp ] here @xmath29 is the two component chiral superfield whose real part is the dilaton and imaginary part is an axion .
we identify axion as the inflaton of the model , and @xmath143 is an additional matter field with modular weight @xmath144 . in typical orbifold string compactifications with three moduli fields ,
the modular weight @xmath144 has value @xmath145 @xcite .
here we shall treat @xmath144 as a phenomenological parameter whose value can have small deviation from the canonical value @xmath145 which may be explained via string loop corrections to the effective supergravity action @xcite . in this model to obtain the correct cmb observables , the parameter @xmath146 has to be fine tuned to the order of @xmath0 . for the complete specification of supergravity
, we assume the superpotential as @xmath147 .
we can decompose @xmath29 field in its real and imaginary parts parametrized by two real fields @xmath25 and @xmath12 , respectively , as @xmath148 the evolution of the matter field @xmath143 is constrained by the exponential factor @xmath149 via @xmath150 in the scalar potential ( [ lvlk ] ) as @xmath151 . since @xmath152 for @xmath153 and @xmath154 during inflation .
therefore field @xmath143 , due to its exponentially steep potential , is rapidly driven to zero at the start of the inflation and stabilizes at @xmath155 @xcite . in fig .
[ fig4 ] , we show the stabilization of the field @xmath143 for different initial conditions as discussed before for @xmath156 .
we also checked the evolution of @xmath143 for @xmath157 and found that it stabilizes in the similar fashion .
therefore for vanishing @xmath143 , the scalar potential and kinetic term ( [ lvlk ] ) takes the simple form v= , _
k= , [ vlk ] which upon using the decomposition of @xmath29 becomes _ k&= & ^_+ e^- ^_,[lk1 ] + v&= & e^- ^m,[v1 ] where @xmath158 and @xmath159 .
since during inflation , dilaton @xmath25 evolves much slower compared to inflaton @xmath12 , see fig .
[ fig2 ] , therefore @xmath160 and hence the first term inside the bracket in ( [ v1 ] ) can be neglected compared to second term .
therefore , from ( [ lk1 ] ) and ( [ v1 ] ) , the lagrangian in ef becomes _
m= ^_+ e^- ^_+ e^- v ( ) , where @xmath161 and we set @xmath162 for quadratic potential ( @xmath163 ) and @xmath164 for quartic potential ( @xmath165 ) .
we see that the parameter @xmath166 is no new parameter and can be given in terms of @xmath144 .
for @xmath110 and @xmath111 , the @xmath167 predictions for a fixed value of @xmath93 and with varying @xmath26 are shown in fig .
to summarize , our two - field two - parameter inflationary model , where the inflaton field has a non - canonical kinetic term due to the presence of the dilaton field , renders quartic and quadratic potentials of the inflaton field viable with current observations . unlike higgs - inflationary scenario which predicts very small tensor - to - scalar ratio @xmath128
, this model can produce large @xmath23 in the range @xmath168 which would definitely be probed by future @xmath1mode experiments and thus such a model can be put to test with the future observations .
also this model produces no isocurvature perturbations upto slow - roll approximation and predicts negligible non - gaussianity consistent with the observations .
in addition , we showed that this model can be obtained from a no - scale sugra model which makes this model of inflation phenomenologically interesting from the particle physics perspective .
work of sd is supported by department of science and technology , government of india under the grant agreement number ifa13-ph-77 ( inspire faculty award ) .
we thank the anonymous referees for their critical comments which improved the discussion of the model .
a. r. ade _ et al .
[ planck collaboration ] , arxiv:1502.01589 [ astro-ph.co ] . p. a. r. ade _ et al .
_ [ planck collaboration ] , arxiv:1502.01592 [ astro-ph.co ] .
f. l. bezrukov and m. shaposhnikov , phys .
b * 659 * , 703 ( 2008 ) .
g. f. giudice and h. m. lee , phys .
b * 694 * , 294 ( 2011 ) .
g. chakravarty , s. mohanty and n. k. singh , int .
j. mod .
d * 23 * , no . 4 , 1450029 ( 2014 ) ; j. joergensen , f. sannino and o. svendsen , arxiv:1403.3289 [ hep - ph ] ; a. codello , j. joergensen , f. sannino and o. svendsen , arxiv:1404.3558 [ hep - ph ] ; g. k. chakravarty and s. mohanty , phys .
b * 746 * , 242 ( 2015 ) . c. brans and r. h. dicke , phys .
rev . * 124 * , 925 ( 1961 ) .
a. a. starobinsky and j. yokoyama , gr - qc/9502002 . j. garcia - bellido and d. wands , phys .
d * 52 * , 6739 ( 1995 ) .
f. di marco , f. finelli and r. brandenberger , phys .
d * 67 * , 063512 ( 2003 ) .
e. cremmer , s. ferrara , c. kounnas and d. v. nanopoulos , phys .
b * 133 * , 61 ( 1983 ) .
j. r. ellis , a. b. lahanas , d. v. nanopoulos and k. tamvakis , phys .
b * 134 * , 429 ( 1984 ) .
a. b. lahanas and d. v. nanopoulos , phys
. rept . * 145 * , 1 ( 1987 ) .
r. kallosh , a. linde , b. vercnocke and w. chemissany , jcap * 1407 * , 053 ( 2014 ) doi:10.1088/1475 - 7516/2014/07/053 [ arxiv:1403.7189 [ hep - th ] ] .
k. hamaguchi , t. moroi and t. terada , phys .
b * 733 * , 305 ( 2014 ) doi:10.1016/j.physletb.2014.05.006 [ arxiv:1403.7521 [ hep - ph ] ] .
d. i. kaiser and e. i. sfakianakis , phys .
lett . * 112 * , no . 1 , 011302 ( 2014 ) . k. schutz , e. i. sfakianakis and d. i. kaiser , phys . rev .
d * 89 * , no . 6 , 064044 ( 2014 ) .
d. seery and j. e. lidsey , jcap * 0509 * , 011 ( 2005 ) doi:10.1088/1475 - 7516/2005/09/011 [ astro - ph/0506056 ] .
k. y. choi , l. m. h. hall and c. van de bruck , jcap * 0702 * , 029 ( 2007 ) doi:10.1088/1475 - 7516/2007/02/029 [ astro - ph/0701247 ] .
j. a. casas , in * jerusalem 1997 , high energy physics * 914 - 917 [ hep - ph/9802210 ] .
j. ellis , m. a. g. garcia , d. v. nanopoulos and k. a. olive , jcap * 1408 * , 044 ( 2014 ) .
j. ellis , m. a. g. garca , d. v. nanopoulos and k. a. olive , jcap * 1501 * , no . 01 , 010 ( 2015 ) . s. ferrara , a. kehagias and a. riotto , fortsch .
* 62 * , 573 ( 2014 ) . | we present a two - field inflation model where inflaton field has a non - canonical kinetic term due to the presence of a dilaton field .
it is a two - parameter generalization of one - parameter brans - dicke gravity in the einstein frame .
we show that in such an inflation model the quartic and quadratic inflaton potentials , which are otherwise ruled out by the present planck-_keck_/bicep2 data , yield scalar spectral index and tensor - to - scalar ratio in accordance with the present data .
such a model yield tensor - to - scalar ratio of the order of @xmath0 which is within the reach of @xmath1mode experiments like _ keck_/bicep3 , cmbpol and thus can be put to test in the near future .
this model yields negligible non - gaussianity and no isocurvature perturbations upto slow - roll approximation .
finally , we show that such a model can be realised in the realm of no - scale supergravity . |
the shape of the @xmath0 spectra of secondaries is of great importance in all investigations concerning developments of cosmic ray cascades @xcite ( cf . also @xcite ) .
the crucial problem of practical importance is the _ existence _ or _ nonexistence _ of the feynman scalling , which says that @xmath1-spectra of secondaries are energy independent . because in cosmic ray applications one is sensitive essentially to large @xmath1 region ( of , say , @xmath2 )
, the commonly used formula @xcite @xmath3 stresses this fact by power - like form with exponent obtained from the @xmath4 limit of fragmentation data @xcite . here
@xmath5 is parameter responsible for the feynman scaling violation whereas @xmath6 and @xmath7 are additional parameters obtained from the fit to data - all are energy dependent ( cf .
@xcite for details ) .
we shall demonstrate that ( [ eq : ao ] ) emerges in natural way from the information theory approach to hadronization processes presented in @xcite and extended to allow for the possible nonextensivity of the hadronization process by applying tsallis entropy @xcite .
( see @xcite for details of notion of nonextensivity and @xcite for its application to high energy reactions . ) .
in order to describe the hadronization process in information theory one uses @xcite the _ least biased and most plausible _ single particle distribution @xmath8 ( where @xmath9 is rapidity ) resulting from hadronization process in which a mass @xmath10 hadronizes into @xmath11 secondaries of mean transverse mass @xmath12 each ( for simplicity we consider only one - dimensional hadronization with limited transverse momenta which can , however , depend on @xmath10 and @xmath11 )
. this is done maximalizing the shannon ( or boltzmann - gibbs ) information entropy , @xmath13 , under constraints of normalization ( @xmath14 ) and energy conservation ( @xmath15 ) , which leads to the following ( extensive ) distribution function cf .
@xcite for details ) : @xmath16 .
\label{eq : maxent}\ ] ] here @xmath17 comes from the normalization of @xmath18 whereas the lagrange multiplier @xmath19 is to be calculated from the energy conservation constraint .
notice that _ there is no free parameter here_. this should be contrasted with the popular use of eq .
( [ eq : maxent ] ) as a `` thermodynamical parametrizations '' with inverse `` temperature '' @xmath20 beeing a free , positively defined parameter . in @xcite it was shown that , in a wide range of energies and multiplicities , @xmath21 .
however , the comparison with data can not be done in such model independent way because of the fluctuations of inelasticity @xmath22 ( given by inelasticity distribution @xmath23 @xcite ) causing fluctuations of @xmath10 and because of fluctuations of number of particles @xmath24 produced from a mass @xmath10 given by multiparticle distribution @xmath25 @xcite ( its actual shape accounts here for the possible multisource structure of the production process ) .
one has therefore @xmath26 , \label{eq : full}\ ] ] where @xmath27 is still calculated from the energy conservation constraint but now for a given event , i.e. , it is a fluctuating quantity with fluctuations resembling gamma - like distribution ( and characterised by mean value @xmath28 and normalised variance @xmath29 ) .
it has been shown in @xcite that such fluctuations of the parameter of the exponential distribution convert it to a power - like distribution : @xmath30 with @xmath31 in our case where @xmath32 and @xmath33 is a parameter connected with the mean value of @xmath34 .
this is precisely the form corresponding to distribution obtained using non - extensive tsallis @xmath35-entropy @xcite , instead of shannon one ( and reproducing it in the limit @xmath36 ) , @xmath37^q)/(1 - q ) \rightarrow
s_{q=1 } = - \int dy f(y ) \ln f(y)$ ] , together with the modified constraint equation @xmath38^q = m / n$ ] . using it
one gets instead of ( [ eq : maxent ] ) its nonextensive version : @xmath39^{\frac{1}{1-q } } .
\label{eq : maxentq}\ ] ] nonextensivity means that the entropy of the composition @xmath40 of two independent systems @xmath41 and @xmath42 is equal to @xcite @xmath43 proceeding to its usual additive form only in the @xmath44 limit .
it arises whenever in the system one enconters long - range correlations , memory effects or fractal structure of the corresponding space - time or phase space .
such situation is expected to occur also in the hadronization processes @xcite .
( 10.,11 . ) fig .
1 . comparison of eq .
( [ eq : form ] ) with ua5 @xcite data for @xmath44 and @xmath45 .
we make therefore the following conjecture : the convolution ( [ eq : full ] ) can , in practice , be replaced by the following simple one parameter formula of the type of eq .
( [ eq : ao ] ) which should be used to describe the existing data of @xcite : @xmath46^{\frac{1}{1-q}}. \label{eq : form}\ ] ] the only free parameter ( characterising the strenght of fluctuations in the system according to the previous discussion ) is the nonextensivity parameter @xmath35 . here
@xmath47 and @xmath48 is the mean ( single non - diffractive ) charged multiplicity at given energy @xmath49 .
notice that because @xmath50 is calculated from the energy conservation constraint which involves all produced particles ( i.e. , total multiplicity ) it is calculated here for @xmath51 particles . for the same reason care
must be taken when one addresses data at @xmath52 gev as part of them is for charged and part for neutral particles only . in both cases
the @xmath50 must be the same ( calculated for total multiplicity at given energy ) whereas multiplicity in front of the formula has to be chosen accordingly to the actual situation . in fig .
1 we show our results both for @xmath44 and our best fit with @xmath53 . in all calculations
the experimentally observed variation of @xmath54 with energy has also been accounted for by using the following simple interpolation formula : @xmath55 gev .
the results are reasonable , especially for @xmath56 and @xmath57 gev . for higher energies
our distributions start to be broader than data and this can not be improved by changing @xmath35 as diminishing its value in order to make distributions narrower will spoil the agreement with data for small rapidities .
it turns out that ua7 and p238 data can not be fitted together with ua5 data , as they demand a slightly bigger value of @xmath58 , cf .
2 . notice that p238 data are for charged and ua7 data for neutral particles therefore they must be described with , respectively , @xmath59 and @xmath60 in eq .
( [ eq : form ] ) , this leads to differences clearly seen in fig .
2 . ( 6.5,5.5 ) ( 6.5,5.5 ) ( 6.5,3 ) ( 6.5,3 )
our approach consists in noticing the striking similarity of eq .
( [ eq : maxentq ] ) to the eq .
( [ eq : ao ] ) of @xcite ( @xmath61 ) . fig .
3 shows the characteristic features of @xmath62 for the case of hadronization of fireball of mass @xmath63 gev into @xmath64 particles of transverse mass @xmath65 gev .
notice that for @xmath66 one obtains an increase of particle densities in the central region connected with its decrease at the edges of rapidity range - clearly resembling the characteristic pattern of feynman scaling violation .
we have replaced therefore the `` exact '' formula , ( [ eq : full ] ) by a one parameter fit represented by ( [ eq : form ] ) where @xmath35 summarizes the action of the averaging over the fluctuations caused by the initial conditions of reaction represented by the inelasticity and multiplicity distributions .
we find it very amazing that such simple approach coincides practically with empirical formula ( [ eq : ao ] ) and with only one parameter @xmath35 describes fairly well all data .
notice that @xmath53 is not very far from @xmath67 , which gives power @xmath68 in ( [ eq : ao ] ) .
notice also that our general formula allows for description of small @xmath1 region as well
. it would certainly be interesting to connect @xmath35 directly to the parameters describing inelasticity and multiplicity distributions or to descriptions of leading particles , for example of the type of that presented in @xcite .
our results seem to indicate that most probably parameter @xmath35 should be @xmath1-dependent reflecting different character of fluctuations of the quantity @xmath69 @xcite in the central ( mostly pionization ) and fragmentation regions .
this problem will be addressed elsewhere .
authors zw and gw are gratefull to the organizers of the _ chacaltaya meeting on cosmic ray physics _ for their support and hospitality . the partial support of polish committee for scientific research ( grants 2p03b 011 18 and 621/e-78/spub / cern / p-03/dz4/99 ) is acknowledged . | we demonstrate that the apparently _ ad hoc _ parametrization of the particle production spectra discussed in the literature and used in description of cosmic ray data can be _ derived _ from the information theory approach to multiparticle production processes .
in particular , the violation of the feynman scaling law can be interpreted as a manifestation of nonextensivity of the production processes .
[ 1999/12/01 v1.4c il nuovo cimento ] |
constituent quark models have been widely used to predict the properties of baryon resonances . however , a large number of resonance states which appear in the constituent quark models have not been observed in pion - nucleon scattering .
it has been argued @xcite that such `` missing resonances '' either decouple from the pion - nucleon channel or are masked by the neighboring resonances which have large @xmath1 decay widths .
this has led to suggestions that other reaction channels such as vector meson productions@xcite might be more sensitive to these missing states .
however , all possibilities have not been exhausted in the @xmath2 n channel .
one such possibility is to consider the polarization observables .
while efforts@xcite are being made to measure the single polarizations , it will be interesting to explore double polarizations which could be more effective in revealing some of the missing resonances through the interference between the resonant and background(non - resonant ) amplitudes .
investigations of these double polarization observables is timely and topical as they coincide with the availability of new opportunities to measure them at jefferson lab and other laboratories . as a step in this direction ,
we will explore in this paper how the double polarization observables of pion photoproduction can be used to investigate nucleon resonances in the higher energy region ( 1.6 @xmath3 2.1 gev ) in which a lot of missing resonances have been suggested . in this energy region , a complete dynamical approach to pion photoproduction , such as that developed@xcite in the investigation of the @xmath4 resonance , is still being developed .
the main difficulty is due to the large number of resonance states and the complexities of the couplings between several meson - baryon channels . for our present very limited purpose
, we will introduce some rather drastic simplifications in using a formulation which is a direct extension of the dynamical formulation of ref.@xcite .
this will then allow us to address the question concerning the missing resonances by relating the resonant amplitude directly to the predictions from a constituent quark model .
as an example , we will consider the predictions by capstick and roberts@xcite . in section
ii , we will give explicitly the formulation used in our calculations .
our findings will be discussed in section iii .
the differential cross section of @xmath5 in the center of mass ( c.m . ) frame can be written as , @xmath6 and @xmath7 are respectively the momenta of the incident photon and outgoing pion , and @xmath8 , @xmath9 , and @xmath10 are the spin projections of the initial nucleon , final nucleon , and incoming photon , respectively . for a total invariant mass @xmath11 ,
the momenta @xmath12 and @xmath13 are @xmath14 , \nonumber \\ q&=&\frac{1}{2\sqrt{s}}[(s - m_n^2-m_\pi^2)^2 - 4m_n^2m_\pi^2]^{1/2},\nonumber\end{aligned}\ ] ] where @xmath15 and @xmath16 are the masses for the nucleon and pion respectively .
the photon laboratory energy @xmath17 is @xmath18\end{aligned}\ ] ] following the formulation of ref .
@xcite , we write the pion photoproduction amplitude as @xmath19.\end{aligned}\ ] ] in this work , we assume that the total amplitude @xmath20 of eq.(3 ) can be calculated from the multipole amplitude generated by the said program@xcite .
these empirical amplitudes are obtained by the following procedure . for each partial wave ,
the multipole amplitude is written as a sum of the conventional born term , vector - meson - exchange , and a many - parameter phenomenological term .
the amplitudes are unitarized by using the empirical @xmath1 phase shifts which are also available in the said program .
the empirical pion photoproduction multipole amplitude are then obtained by adjusting the parameters of the phenomenological term to perform @xmath21-fits to all of the existing data of @xmath22 , @xmath23 and @xmath24 reactions from threshold up to invariant mass w=2.1 gev . in this work
we use the sm01 solutions@xcite from their energy - dependent fits .
more details about the said program can be found in refs.@xcite and @xcite . following the formulation of ref .
@xcite , we write the resonant amplitude as @xmath25 @xmath26 where @xmath27 is the mass of a @xmath28 with spin quantum numbers @xmath29 . here
we neglect the effect due to the non - resonant mechanisms on the @xmath28 decay amplitude and the shift of the resonance position .
then the resonance mass @xmath30 and the @xmath28 decay amplitude @xmath31 can be identified with the predictions from a quark model , as discussed in refs .
@xcite .
we however do not have information about the total decay width @xmath32 for most of the @xmath28 s considered here . for simplicity , we assume that its energy - dependence is similar to the width of the @xmath33 decay within the oscillator constituent quark model .
following ref . @xcite and neglecting the real part of the mass shift , we then have @xmath34 , \label{n*:decay}\ ] ] where @xmath35 is the orbital angular momentum of the considered @xmath1 state and @xmath36 in the above equations , @xmath37 is the pion momentum at energy @xmath38 while the on - shell momentum @xmath39 is evaluated at @xmath40 .
the energies are defined by @xmath41 and @xmath42 .
our choice of the total average width @xmath43 and cutoff parameter @xmath44 for eq .
( [ n*:decay ] ) will be specified in sec .
iii . by setting the photon momentum in the @xmath45-direction , the @xmath46 amplitude ( @xmath47 ) in eq .
( [ t : n * ] ) can be calculated from the helicity amplitude @xmath48 listed in ref .
@xcite @xmath49 the @xmath50 amplitude ( @xmath51 ) takes the following form @xcite @xmath52 where @xmath53 is the @xmath2 meson momentum at @xmath54 .
here we also include an extrapolation factor @xmath55 $ ] similar to that for defining the width eq.(5 ) . by using the information in refs.@xcite
, we can get helicity amplitude @xmath56 of eq.(7 ) and partial - wave transition strength @xmath57 of eq.(8 ) predicted by capstick and roberts .
the resulting values for the considered 30 @xmath28 and 26 @xmath58 are listed in tables i - iv .
our first task is to calculate the resonant amplitude defined by eqs .
( [ resampeq])-([omegan * ] ) . with the helicity amplitude @xmath56 and partial - wave transition strength @xmath57 given in tables i - iv
, this amplitude depends only on the choice of the total width @xmath43 and the cutoff @xmath44 in eqs.(5 ) and ( 8) . for simplicity
, we set @xmath44 = 650 mev for all resonances considered .
this value is close to what was determined in the investigation of @xmath59@xcite .
the widths of the known @xmath28 resonances listed in @xcite range from 200 - 400 mev .
we therefore take their average value and set @xmath60 mev for all @xmath28 resonances listed in tables i - ii . for the similar reason
, we set @xmath61 mev for the considered @xmath58 resonances listed in tables iii - iv . with these specifications , the resonant amplitude eq.(4 ) can be computed with no adjustable parameters . since the full amplitude @xmath20 is taken from
said program , the non - resonant(background ) amplitude @xmath62 for our investigation can then be fixed by @xmath63.\end{aligned}\ ] ] with the background amplitude defined above , we then use eq.(2 ) to investigate the role of each resonance in determining the resonance amplitude . in fig .
[ amp ] , we first observe that the differential cross sections(solid curves ) calculated from the employed said amplitude agree well with the available data .
the dashed curves are obtained when the the effects due to all of the considered 56 resonance states are removed .
we see that the nucleon resonance effects on the unpolarized differential cross sections are not very pronounced .
furthermore none of the considered resonances listed in tables i - iv dominate the difference between the full calculation(solid curves ) and the background only calculation(dashed curves ) .
obviously , it is not easy to investigate the nucleon resonances by only considering the unpolarized differential cross sections .
this is similar to the finding of ref.@xcite in a investigation of @xmath64 photoproduction .
we now turn to reporting our investigations of the polarization observables .
before we discussing the double polarization observables , we show in fig . 2 some typical results from our calculations of single polarization observables .
as expected , the effects due to nucleon resonances are much more visible here .
hopefully , these predictions can be tested in the near future when the data from the on - going experiments become available . where @xmath67 is the cross - section @xmath68 and the superscripts @xmath69 denote the polarizations of the beam , target ; recoil and meson respectively .
@xmath70 corresponds to a circularly polarized photon beam with helicity + 1 , @xmath71 represents the direction of the nucleon polarization , and u represents the unpolarized particle .
the coordinate system is chosen such that the incoming photon is in the z - direction and x - z plane is the reaction plane .
thus y - direction is perpendicular to the reaction plane .
we have performed extensive calculations of the two double polarization observables , @xmath72 and @xmath73 , in the energy region 1.6 gev @xmath74 w @xmath75 2.1 gev . as expected ,
the resonance effects are much more pronounced than what can be observed in fig.1 for the unpolarized differential cross sections or fig.2 . for the single polarization observables .
this is illustrated in fig.3 for the beam - target polarization and fig.4 for the beam - recoil polarization for two typical angles at 90 and 120 degrees .
our next task is to explore the favorable kinematic regions where only one or just a limited number of nucleon resonances among the 56 considered resonance states will dominant . we have found that at angles close to 120 degrees the resonance effects on the beam - recoil polarization ( eq.(11 ) ) are mainly due to the four - star resonances @xmath76 of table i and @xmath77 and @xmath78 of table ii , and the two - star resonance @xmath79 of table ii . furthermore the @xmath79 state has the largest effect among about 50 two - star and missing resonances in this higher energy region .
this is shown in fig.5 .
we see that by adding only the two - star @xmath79 to the calculation including only background and three four - star resonances , the predicted beam - recoil polarizations are shifted from the dashed curves to the dot - dashed curves which are close to the results from full calculations including all resonances listed in tables i - iv .
the investigation of the two - star @xmath79 is clearly most favorable in the @xmath22 channel .
our results suggest that data on the double polarization observables of pion photoproduction in the region 1.6 @xmath80 w @xmath802.1 gev could be very useful in investigating nucleon resonances .
the experiments on these observables would be complementary to other similarly motivated ongoing experiments on vector meson photoproduction and kaon photoproduction .
moreover , these double polarization observables should be accessible at facilities such as jefferson lab and spring-8 . to end , we like to mention that some experimental data on double polarization observables have recently become available in the @xmath81 channel for w @xmath82 1.6 gev @xcite , but there is no data on the other two charged channels .
z. li , ph.d thesis * an analysis of pion photoproduction * , virginia polytechnic institute and state university , 1992 .
t. yoshimoto , t. sato , m. arima , and t .- s .
h. lee , phys .
c * 61 * , 065203 ( 2000 ) . .
parameters for positive parity nucleon resonances from refs .
the helicity amplitude @xmath56 is given in unit of @xmath83 gev@xmath84 .
@xmath57 is in unit of mev@xmath85 .
the resonance mass @xmath27 is in unit of mev . [
cols="^,^,^,^,^,^,^,^,^",options="header " , ] | the role of the nucleon resonances in the double polarization observables of pion photoproduction is investigated by using the resonance parameters predicted by capstick and roberts . as an example
, we show that the not - well - determined two - star resonance n@xmath0(1960 ) can be examined by performing experiments on beam - recoil polarization at large angles . |
throughout this note , @xmath0 will denote a second countable , locally compact groupoid with haar system @xmath4 .
also , we shall fix a bundle - bundles . however , we adopt the increasingly popular convention that bundles are to be denoted by calligraphic letters .
the fibres in a given bundle are then denoted by the corresponding roman letter .
thus if @xmath1 is a bundle of @xmath2-algebras , say , over a space @xmath5 , then the fibre over @xmath6 will be denoted @xmath7 . ]
@xmath1 of ( separable ) @xmath2-algebras over the unit space @xmath8 of @xmath0 .
we shall write @xmath9 for the projection of @xmath1 onto @xmath8 .
we shall assume that there is a continuous action , denoted @xmath10 , of @xmath0 on @xmath1 .
this means the following : first of all , @xmath10 must be a homomorphism from @xmath0 into the isomorphism groupoid of @xmath1 , @xmath11 , so that , in particular , @xmath12 is a @xmath2-isomorphism for each @xmath13 .
secondly , let @xmath14 and @xmath15 be the bundles on @xmath0 obtained by pulling back @xmath1 via @xmath16 and @xmath17 , so that @xmath18 and similarly for @xmath19
. then @xmath10 determines a bundle map @xmath20 by the formula @xmath21 .
the continuity assumption that we make is that for each continuous section @xmath22 of @xmath23 , @xmath24 is a continuous section of @xmath15 .
let @xmath25 denote the space of continuous sections of @xmath15 with compact support and for @xmath26 , set@xmath27 and set@xmath28 then , with respect to these operations and pointwise addition and scalar multiplication , @xmath25 becomes a topological @xmath29-algebra in the inductive limit topology to which renault s disintegration theorem @xcite applies .
the enveloping @xmath2-algebra of @xmath25 is called the _ crossed product of _
acting on _ @xmath1 and is denoted @xmath3 @xcite .
the basic problem we study in this note is [ fundquest]under what circumstances on @xmath0 and @xmath1 is @xmath3 a continuous trace @xmath2-algebra ? the first systematic investigation into questions of this type that we know of is green s pioneering study @xcite .
there , green deals with the case where @xmath0 is the transformation group groupoid obtained by letting a locally compact _ group _
@xmath30 act on a locally compact hausdorff space @xmath5 and where @xmath1 is the trivial line bundle over @xmath31 .
his principal result asserts that if @xmath30 acts freely on @xmath5 , then @xmath3 is continuous trace if and only if the action of @xmath30 is proper . in this event , in fact , @xmath3 is strongly morita equivalent to @xmath32 .
another precedent to this investigation is the work of the second two authors in @xcite . here , the hypothesis is that @xmath0 is a _ principal _ groupoid and the bundle @xmath1 is again the trivial line bundle .
it was shown that @xmath33 has continuous trace if and only if the action of @xmath0 on @xmath8 is proper . in this event , again , @xmath34 is strongly morita equivalent to @xmath35 .
the first example where the bundle @xmath1 is nontrivial was considered by raeburn and rosenberg in @xcite .
they considered a locally compact _ group _ acting on a continuous trace @xmath2-algebra @xmath36 and showed that if the natural action of @xmath0 on the spectrum of @xmath36 , @xmath37 , is free and proper , then the cross product @xmath2-algebra @xmath38 has continuous trace . in @xcite ,
olesen and raeburn proved a conditioned converse : if the group @xmath0 is abelian and acts freely on @xmath37 , then if @xmath38 is continuous trace , the action of @xmath0 on @xmath37 must be proper . quite recently , deicke @xcite used non - abelian duality theory to remove the hypothesis that @xmath0 is abelian . thus , the best result in this direction is : if @xmath0 acts on @xmath36 yielding a free action on @xmath37 , then @xmath38 is continuous trace if and only if the action of @xmath0 on @xmath37 is proper .
our objective in this note , theorem [ main ] , is to prove a result that contains all of these examples as special cases and much more , as well .
it is based on two hypotheses and some ancillary considerations that we will elaborate .
the first hypothesis is [ hyp1 ] the @xmath2-algebra @xmath39 is continuous trace . here
, @xmath39 denotes the @xmath2-algebra of continuous sections of the bundle @xmath1 that vanish at infinity on @xmath8 .
an hypothesis on the bundle @xmath1 of this nature is natural and reasonable , in view of the fact that in the trivial case @xmath40 , we have @xmath41 .
furthermore , we note that a compact group can act on an antiliminal @xmath2-algebra in such a way that the crossed product is continuous trace @xcite .
to state our second hypothesis , we need a couple of remarks about the spectrum of @xmath39 .
we shall denote it by @xmath5 throughout this note .
observe that @xmath5 may be expressed as the disjoint union of spectra @xmath42 and the natural projection @xmath43 from @xmath5 to @xmath8 is continuous and open @xcite .
the groupoid @xmath0 acts on @xmath5 ( using the map @xmath43 ) as follows . if @xmath6 , we shall write @xmath44 $ ] in order to specify a particular irreducible representation in the equivalence class represented by @xmath45 . then if @xmath46 denotes the space @xmath47 and if @xmath48 , then @xmath49 is defined to be @xmath50 $ ] .
the fact that the action of @xmath0 on @xmath5 is well defined and is continuous is easily checked ( as in ( * ? ? ?
* lemma 7.1 ) for example ) .
our second hypothesis is [ hyp2]the action of @xmath0 on @xmath5 is free . for @xmath6 ,
we shall write @xmath51 for the quotient @xmath52 . then @xmath51 is well defined ( i.e. , it is independent of the choice of @xmath53 ) and is an elementary @xmath54-algebra .
hypothesis [ hyp1 ] guarantees that the collection @xmath55 may be given the structure of an elementary @xmath2-algebra bundle over @xmath5 satisfying fell s condition ( * ? ? ?
* proposition 10.5.8 ) . as a result
, we find that @xmath39 is naturally isomorphic to @xmath56 .
the action of @xmath0 on @xmath5 induces one on @xmath57 that we shall use .
we find it preferable to express this in terms of the _ action groupoid _
@xmath46 defined as follows . as a set , @xmath58 and the groupoid operations
are defined by the formulae @xmath59 @xmath60 , @xmath61 .
note , in particular , that the unit space of @xmath46 may be identified with @xmath5 via : @xmath62 .
note , too , that the range and source maps on @xmath46 , denoted @xmath63 and @xmath64 , are given by the equations @xmath65 and @xmath66 .
the groupoid @xmath46 in the product topology is clearly locally compact , hausdorff , and separable .
it has a haar system @xmath67 given by the formula @xmath68 .
observe that the action of @xmath0 on @xmath5 is free ( resp .
proper ) iff @xmath46 is principal ( resp .
proper ) .
the groupoid @xmath46 acts on @xmath57 via the formula@xmath69 where @xmath70 lies in @xmath71 .
note that this action is well defined since @xmath72 .
we promote this action of @xmath46 on @xmath57 to one on @xmath73 : @xmath74 . if the action of @xmath0 on @xmath5 is free and proper , then the action of @xmath46 on @xmath75 is also free and proper . in this case
, we write @xmath76 for the quotient space @xmath77 .
then @xmath76 is naturally a bundle of elementary @xmath2-algebras over @xmath78 .
in fact , using the methods of theorem 1.1 of @xcite , it is easy to see that @xmath76 satisfies fell s condition .
thus , in particular , @xmath79 has continuous trace , if @xmath0 acts on @xmath5 freely and properly . with these preliminaries at our disposal , we are able to state the main result of this paper as [ main]under hypotheses [ hyp1 ] and [ hyp2 ]
, @xmath80 has continuous trace if and only if the action of @xmath0 on @xmath5 is proper . in this event
, @xmath3 is strongly morita equivalent to @xmath79 , where @xmath76 is the elementary @xmath2-bundle over @xmath81 satisfying fell s condition , that was just defined .
first we reduce the proof of theorem [ main ] to the case when @xmath0 is a principal groupoid .
this reduction is accomplished with the aid of [ reduction]in the notation established above , @xmath3 is isomorphic to @xmath82 . for @xmath83 ,
identify @xmath84 with @xmath85 .
also , set @xmath86 has compact support in @xmath87 .
recall that @xmath88 and , likewise , @xmath89 .
so , in the definition of @xmath90 , @xmath91 , which nominally is in @xmath92 , is to be viewed in @xmath85 . in particular , @xmath93 lies in @xmath94 , when @xmath95 .
now it is straightforward to check that @xmath90 is a subalgebra of @xmath25 that is dense in @xmath96 in the inductive limit topology .
further , if we define @xmath97 by the formula@xmath98(x)=f(x,\gamma)\text{,}%\ ] ] @xmath99 , then @xmath100 is an algebra @xmath101-homomorphism that is continuous in the inductive limit topologies on @xmath102 and @xmath103 .
hence by renault s disintegration theorem ( * ? ? ?
* theorem 4.1 ) , @xmath100 extends to a @xmath2-homomorphism from @xmath82 into @xmath3 .
the inverse @xmath104 to @xmath100 is given formally by the formula@xmath105 @xmath106 ; i.e. , at the level of @xmath107 , @xmath108 and @xmath109 are the identity maps on the appropriate algebras .
the problem is that _ a priori _ @xmath110 is not continuous in the inductive limit topologies .
however , @xmath110 _ is _ manifestly continuous with respect to the so - called @xmath111-norms on @xmath112 and @xmath102 , where for @xmath113 the @xmath111-norm of @xmath22 is @xmath114 and similarly for @xmath102 . since every representation of one of these algebras is continuous in the @xmath111-norm by renault s disintegration theorem (
* theorem 4.1 ) , we conclude that @xmath110 extends to a @xmath2-homomorphism from @xmath3 to @xmath82 that is the inverse of @xmath100 .
_ on the basis of theorem [ reduction ] we may and shall assume from now on that _
@xmath115 _ and that _ @xmath84 _ is an elementary _
@xmath116__-algebra for every _ _
@xmath117 [ proof of the sufficiency ]
if @xmath0 acts freely and properly on @xmath8 , then @xmath8 is an equivalence between @xmath0 and the quotient space @xmath118 ( viewed as a cotrivial groupoid ) in the sense of @xcite .
further , if @xmath119 , then @xmath120 serves as a bundle of morita equivalences between @xmath1 and @xmath121 in the sense of @xcite .
( see @xcite , too . ) indeed , recall that @xmath122 is the quotient @xmath123 , where @xmath124 .
then the @xmath1-valued inner product on @xmath120 is given by the formula @xmath125 while the @xmath121-valued inner product on @xmath126 is given by the formula@xmath127\text{,}\ ] ] where @xmath128 $ ] denotes the image of @xmath129 in @xmath130 .
by corollaire 5.4 of @xcite,@xmath131 this second algebra is a continuous - trace @xmath2-algebra since @xmath121 is a bundle of elementary @xmath2-algebras satisfying fell s condition , as we noted earlier .
the proof of the necessity in theorem [ main ] is modeled on the proofs in @xcite and @xcite , which in turn are inspired by ideas in @xcite .
there are , however , a number of new difficulties that must be overcome . for each @xmath83 , we fix an irreducible representation @xmath132 of @xmath84 on a hilbert space @xmath133 . since each @xmath84 is elementary the @xmath132 s are unique up to unitary equivalence .
we define @xmath134 to be the representation of @xmath3 on the hilbert space @xmath135 according to the formula@xmath136 where @xmath137 and @xmath138 .
( recall that @xmath139 is the image of @xmath140 under inversion . )
thus , @xmath134 is the representation of @xmath141 _ induced _ by the irreducible representation @xmath132 viewed as a representation of @xmath39 .
this implies , in particular , that replacing @xmath132 by a unitarily equivalent representation does not affect the unitary equivalence class of @xmath134 .
the following lemma and corollary capture the salient features of the @xmath134 that we shall use .
the proof follows the lines of the arguments in ( * ? ? ?
* lemma 2.4 and proposition 2.5 ) .
only minor changes need to be made to accommodate the presence of @xmath1 .
the key point is that @xmath134 is unitarily equivalent to the representation @xmath146 of @xmath3 defined by the formula@xmath147 @xmath137 , @xmath148,\mu_{\lbrack u]})\otimes h_{u}$ ] , where @xmath149 $ ] denotes the orbit of @xmath143 and @xmath150}$ ] is the image of @xmath139 under the map @xmath151 .
the fact that the action of @xmath0 on @xmath8 is free ( i.e. , @xmath0 is a principal groupoid ) implies that @xmath151 is a bijection between @xmath152 and @xmath149 $ ] .
it is a borel isomorphism , of course , because of our separability hypotheses and the fact that @xmath151 is continuous .
the value of @xmath146 for us lies in the fact that it is evident how to express @xmath146 as the integrated form of a representation of @xmath153 in the sense of ( * ? ? ?
* definition 3.4 ) .
the measure class on @xmath8 is , of course , that determined by @xmath150}$ ] and the hilbert bundle @xmath154 is the constant bundle determined by @xmath133 over the orbit of @xmath143 , i.e. , @xmath155{cc}% \{v\}\times h_{u } & v\in\lbrack u]\\ 0 & \text{otherwise}% \end{array } \right .
\text{.}%\ ] ] thus , @xmath156}$ ] is identified with @xmath157,\mu_{\lbrack u]})\otimes h_{u}$ ] in the standard fashion .
the groupoid @xmath0 is represented on @xmath154 according to the formula@xmath158 @xmath159 , @xmath160 $ ] , i.e. , @xmath161 is just the translation representation , and @xmath1 is represented on @xmath154 according to the formula@xmath162 @xmath163 , @xmath164 , where @xmath165 is the unique element in @xmath0 with source @xmath144 and range @xmath143 .
observe that the @xmath39 acts as multipliers on @xmath141 according to the formula @xmath166 for @xmath167 and @xmath137 .
the extension @xmath168 of @xmath146 to the multiplier algebra @xmath3 represents @xmath169 on @xmath170,\mu_{\lbrack u]})\otimes h_{u}$ ] via the equation@xmath171 again , where @xmath165 is the unique element in @xmath0 with source @xmath144 and range @xmath143 .
it is clear from this that the weak closure of the algebra @xmath172 is the full algebra of decomposable operators on @xmath170,\mu_{\lbrack u]})\otimes h_{u}$ ] .
it follows that any projection that commutes with @xmath173 must be diagonal . on the other hand
, it follows from the definition of the representation of @xmath0 , @xmath161 , that a diagonal operator commuting with @xmath173 must commute with @xmath161 , and therefore must be constant a.e .
@xmath150}$ ] .
this proves that @xmath146 , and hence @xmath134 , is irreducible .
if @xmath143 and @xmath144 lie in the same orbit , it is clear that translation by the ( unique ) @xmath165 with source @xmath144 and range @xmath143 implements an equivalence between @xmath134 and @xmath142 . on the other hand ,
if @xmath143 and @xmath144 lie in different orbits , then @xmath134 and @xmath142 are disjoint .
indeed , the representations @xmath174 and @xmath175 of @xmath176 obtained by restricting @xmath168 and @xmath177 to @xmath176 , viewed as a subalgebra of @xmath178 , are supported on the disjoint sets @xmath149 $ ] and @xmath179 $ ] .
arguing just as we did in the proof of ( * ? ? ?
* proposition 2.5 ) , using ( * ? ? ?
* lemma 4.15 ) , we conclude @xmath134 and @xmath142 are disjoint . finally , to see that the map @xmath180 is continuous , observe that hypothesis [ hyp1 ] guarantees that for each point @xmath181 we can find a neighborhood @xmath182 of @xmath143 on which the @xmath183 s can be chosen to be the fibres of a ( topological ) hilbert bundle @xmath184 and on which we can choose the @xmath185 s so that for any section @xmath186 that is supported on @xmath182 and any two @xmath187-sections of @xmath188 over @xmath182 , @xmath189 and @xmath190 , the function @xmath191 is continuous .
it follows from the continuity of the haar system that given such sections @xmath189 and @xmath190 of @xmath188 and any two functions @xmath192 and @xmath193 in @xmath194 , the function @xmath195 ( where the inner products are taken in @xmath196 ) is continuous for all @xmath197 .
this shows that the map @xmath198 is continuous .
[ l^ubis]assume that @xmath0 is principal and that @xmath1 is an elementary @xmath2-bundle over @xmath8 , satisfying fell s condition , on which @xmath0 acts .
if @xmath3 has continuous trace , then the map that sends @xmath83 to the unitary equivalence class of @xmath134 defines a continuous open surjection of @xmath8 onto @xmath199 that is constant on @xmath0-orbits .
in particular , orbits are closed and @xmath118 is homeomorphic to @xmath200 .
the proof is also essentially the same as the proof in ( * ? ? ?
* proposition 2.5 ) . here is an outline .
write @xmath100 for the map @xmath201 $ ] . then by lemma [ l^u ] , @xmath100 is continuous and constant on @xmath0-orbits
. thus @xmath100 passes to continuous map on @xmath118 with the quotient topology ( no matter how bad that might be ) .
since , however , @xmath200 is hausdorff by hypothesis , we conclude that @xmath118 is hausdorff .
suppose that @xmath202 is an irreducible representation of @xmath3 and let @xmath203 be the representation of @xmath176 obtained by extending @xmath202 to the multiplier algebra of @xmath3 and then restricting to @xmath176 .
the kernel @xmath204 of @xmath203 is the set of functions in @xmath176 that vanish on a closed set @xmath205 in @xmath8 .
then @xmath205 is easily seen to be invariant .
indeed , one may do this directly or use the fact that it supports the quasi - invariant measure associated to the disintegrated form of @xmath202 guaranteed by @xcite . further , since @xmath202 is irreducible , @xmath205 can not be expressed as the union of two disjoint , closed , @xmath0-invariant sets .
since the quotient map from @xmath8 to @xmath118 is continuous and open , we may apply the lemma on page 222 of @xcite to conclude that @xmath205 is an orbit closure . since we now know that orbits are closed , @xmath205 is , in fact , an orbit .
thus @xmath202 factors through @xmath206};\mathcal{a})$ ] .
however , @xmath207}$ ] is a _ transitive _ principal groupoid . a little reflection ,
using theorem 3.1 of @xcite , reveals that every irreducible representation of @xmath206};\mathcal{a})$ ] is unitarily equivalent to @xmath134 .
we now assume the action of @xmath0 on @xmath8 is not proper and use lemma 2.6 of @xcite to choose a sequence @xmath208 such that @xmath209 in the sense that @xmath210 eventually escapes each compact subset of @xmath0 , and such that @xmath211 , for some @xmath212 .
we shall fix this sequence for the remainder of the proof .
we also choose a relatively compact neighborhood @xmath213 of @xmath212 and a section @xmath192 in the pedersen ideal of @xmath39 such that @xmath192 is non - negative , compactly supported , and satisfies @xmath214 on @xmath213 .
the fact that @xmath1 satisfies fell s condition guarantees that such choices are possible . with these ingredients fixed ,
we want to build a special neighborhood @xmath215 of @xmath216 in @xmath0 , following the analysis on pages 236238 of @xcite .
first observe , as we have above , that since @xmath0 is principal , @xmath17 maps @xmath217 bijectively onto @xmath218 $ ] while @xmath16 maps @xmath219 bijectively onto @xmath218 $ ] , where , recall , @xmath218 $ ] denotes the orbit of @xmath216 . since @xmath218 $ ]
is closed by corollary [ l^ubis ] while @xmath17 is continuous and open on @xmath0 , we see that @xmath17 maps @xmath217 homeomorphically onto @xmath218 $ ] .
likewise , @xmath16 maps @xmath219 homeomorphically onto @xmath218 $ ] .
also , since @xmath0 is principal , multiplication induces a homeomorphism between @xmath220 and @xmath221}$ ] .
let @xmath222 be the closed support of @xmath192 , a compact subset of @xmath8 , and set @xmath223\cap n)$ ] and @xmath224\cap n)$ ] , obtaining compact subsets of @xmath217 and @xmath219 , respectively .
then we see that if @xmath225}$ ] and if @xmath226 and @xmath227 , then @xmath228 .
according to lemma 2.7 of @xcite , we may select symmetric , conditionally compact open neighborhoods @xmath229 and @xmath230 of @xmath8 such that @xmath231 .
( recall that a neighborhood @xmath232 of @xmath8 is conditionally compact in case @xmath233 and @xmath234 are relatively compact subsets of @xmath0 for each relatively compact subset @xmath235 in @xmath0 . )
we select such a pair , as we may , with the additional property that @xmath236 .
then from the preceding paragraph , we see that if @xmath237 , then either @xmath238 or @xmath239 . by construction ,
@xmath240 so we may find relatively compact open neighborhoods @xmath241 and @xmath242 of @xmath216 in @xmath0 so that @xmath243 , @xmath244 , and so that @xmath245 with these @xmath241 and @xmath242 so chosen , the special open neighborhood @xmath215 of @xmath216 in @xmath0 that we want is defined to be @xmath246 .
observe that we have @xmath247 set@xmath248{cc}% g(r(\gamma)),\ ; & \gamma\in\overline{w_{1}}^{7}\overline{v_{1}}\,\overline { w_{1}}^{7}\\ 0,\ ; & \gamma\notin e \end{array } \text{.}\right.\ ] ] then , since @xmath249 , @xmath250 is a continuous section of @xmath15 on @xmath0 that vanishes outside @xmath251 observe the following containment relations among relatively compact sets : @xmath252 .
hence , we may find a compactly supported function @xmath253 on @xmath0 such that @xmath254 for all @xmath165 , @xmath255 on @xmath256 and @xmath257 off @xmath258 .
replacing @xmath253 by @xmath259 , if necessary , we may assume that @xmath253 is a selfadjoint element of the convolution algebra of scalar - valued functions @xmath194 .
define @xmath260 . by our choices of @xmath192 and @xmath253
, @xmath205 belongs to @xmath25 , @xmath205 is selfadjoint and @xmath261 for all @xmath143 by the definition of @xmath134 ( cf . ) .
let @xmath262 be the projection onto @xmath263 @xmath264 and let @xmath265 be the projection onto the orthocomplement , @xmath266 @xmath267 . then , if @xmath268,@xmath269we see that@xmath269@xmath270 for all @xmath271 because @xmath253 is identically @xmath272 on @xmath256 .
however , by definitions of @xmath215 and @xmath250 , the equation persists when @xmath273 , yielding @xmath274 .
thus , @xmath262 commutes with @xmath275 .
moreover , when @xmath276 , these formulas show that @xmath277 .
we now want to show that @xmath278 and we want to analyze the trace , @xmath279 .
however , when @xmath189 is in the range of @xmath280 , the formula for @xmath281 shows that @xmath282 as for the trace , observe that if @xmath283 is defined by the formula@xmath284 on @xmath285 , then our calculations show that @xmath283 is continuous , positive semidefinite and supported on @xmath286 , and that @xmath287 .
consequently , we may use duflo s generalization of mercer s theorem , ( * ? ? ?
* proposition 3.1.1 ) , and the fact that @xmath288 to conclude that@xmath289@xmath290 by our choice of @xmath192 , this expression is continuous in @xmath143 and when @xmath276 yields the value @xmath291 .
we will show that there is a positive number @xmath292 such that @xmath293 eventually , where @xmath294 denotes the positive part of the selfadjoint operator @xmath295 therefore , eventually @xmath296 assume that we have shown this , and set @xmath297{cc}% 0 , & t\leq a\\ 2(t - a ) , & a\leq t\leq2a\\ t , & 2a \end{array } \right.\ ] ] then @xmath298 is a positive element in the pedersen ideal of @xmath299 , and so the function @xmath300@xmath301 is finite and continuous in @xmath143 , with value @xmath302 at @xmath276 .
( recall that @xmath303 and so @xmath304 . ) on the other hand , we showed that @xmath305 is positive .
since @xmath262 commutes with @xmath275 , @xmath306 .
but also we showed that @xmath307 is continuous at @xmath216 .
consequently , so is @xmath308 .
since @xmath309 , the function @xmath310 is continuous by lemma 4.4.2(i ) in @xcite , with value @xmath311 at @xmath276 , also .
therefore @xmath312 since the largest eigenvalue of @xmath313 , the largest eigenvalue of @xmath314 also .
consequently , @xmath315 this contradiction will complete the proof .
so we will finish by verifying the asserted inequality ( [ norm ] ) . to this end , choose an open neighborhood of @xmath216 in @xmath0 , @xmath316 , that is contained in @xmath241 and choose a conditionally compact neighborhood @xmath317 of @xmath8 such that if @xmath318 , then @xmath17 maps @xmath319 into @xmath213 . without loss of generality
, we may assume that @xmath320 .
observe that if @xmath321 then for @xmath322 , @xmath323 . indeed ,
if @xmath324 , then @xmath325 contrary to assumption .
so , since @xmath326 and @xmath327 are tending to @xmath216 , while @xmath328 eventually escapes @xmath329 , we can conclude that for @xmath330 sufficiently large , whenever @xmath165 lies in @xmath331 , then @xmath323 while @xmath332 and @xmath333 lie in @xmath213 .
from now on , we will assume that @xmath330 is sufficiently large so that these conditions are satisfied .
next observe that since for each @xmath330 , the map @xmath334 defines a continuous family of rank @xmath272 projections on the hilbert space @xmath335 , we can find a borel family of unit vectors @xmath336 such that @xmath337 is the rank @xmath272 projection determined by @xmath338 .
let @xmath339 where @xmath340 denotes the characteristic function of @xmath331 .
observe that if @xmath341 , then @xmath342 .
consequently , @xmath343 and we may calculate to find that if @xmath322 , then@xmath344 hence @xmath345 .
however , by our assumption on @xmath328 , @xmath346 lies in @xmath347 , so @xmath348 .
this shows that @xmath349 provided @xmath330 is sufficiently large .
but the continuity of the haar system implies that @xmath350 , as @xmath351 .
this verifies equation ( [ norm ] ) and completes the proof of theorem [ main ] .
# 1 [ 2 ] http://www.ams.org/mathscinet-getitem?mr=#1[#2 ] [ 2]#2 p. bernat , n. conze , m. duflo , m. lvy - nahas , m. ras , p. renouard , and m. vergne , _ reprsentations des groupes de lie rsolubles _ , dunod , paris , 1972 , monographies de la socit mathmatique de france , no . | let @xmath0 be a second countable , locally compact groupoid with haar system , and let @xmath1 be a bundle of @xmath2-algebras defined over the unit space of @xmath0 on which @xmath0 acts continuously .
we determine conditions under which the associated crossed product @xmath3 is a continuous trace @xmath2-algebra . |
hadrons are composed of the fermions called quarks and the gauge bosons called gluons , and the dynamics of these particles are described by the quantum chromodynamics ( qcd ) .
the lagrangian of qcd is as follows : @xmath6 here @xmath7 is the field strength , @xmath8 the quark field , @xmath9 the coupling constant , @xmath10 the covariant derivative , @xmath11 the structure constant of @xmath12 , @xmath13 the current quark mass , respectively .
the subscript @xmath14 in the quark field stands for the flavor . in the real world , @xmath15 , @xmath16 ( up , down , charm , strange , top , and bottom ) , and the values for @xmath13 are as follows scheme . for @xmath17 , @xmath18 , and @xmath19 quarks ,
the scale is @xmath20gev while for @xmath21 , @xmath22 , and @xmath23 quarks , the scale is set to that of the mass of each particle . ]
@xcite : @xmath24 2 mev , @xmath25 6 mev , @xmath26 1.2 gev , @xmath27 95 mev , @xmath28 4.1 gev , @xmath29 175 gev .
let us briefly review the present theoretical understandings on the phase structure of qcd matter . since the quark number is conserved in each flavor ,
there exist the chemical potential ( @xmath30 ) which are related to the conserved quark number .
it is possible to analyze the case that the chemical potentials of each flavor are different , but in the following we will review only the case that they have the same value : @xmath31 .
we can make rather reliable predictions in the case of extremely high temperature or chemical potential , because of the asymptotic freedom . at high chemical potential ,
smallness of @xmath9 justifies us to expect that the system should be a fermi liquid with a fermi sphere of the quarks .
however , there is an attractive interaction between the two quarks in the color anti - triplet channel via gluon exchange .
this attractive interaction causes an instability of the fermi surface as is well known , and it results in the permanent formation of the cooper pair composed of quarks leading to the superconducting phase called color superconducting ( csc ) phase .
also in the case that temperature is extremely high , the quarks and the gluons are expected to be deconfined @xcite .
that phase is called quark - gluon plasma ( qgp ) @xcite . since @xmath9 is not so small when temperature and chemical potential are not extremely large determined by experiments in some energy scales are as follows @xcite : @xmath9(1 gev)@xmath32 , @xmath9(10 gev)@xmath33 , and @xmath9(100 gev)@xmath34 . ] , it is difficult to what phase is realized in that region . nevertheless , the first principle calculation using the monte carlo simulation , which is called lattice qcd , is possible when the chemical potential is zero .
the lattice results tell us that the transition from the confined phase to the deconfined one is crossover , not the phase transition in the usual sense , and that the transition temperature ( @xmath35 ) is approximately 200 mev .
analyses in the small-@xmath36 region have been tried by using some methods such as the taylor expansion method @xcite and the imaginary chemical potential method . in the case that the chemical potential is finite , the lattice calculation is unreliable at present , so we have to use some effective models such as the nambu - jona - lasino ( njl ) model and its improved version , the chiral random matrix model @xcite , and the strong coupling expansion .
many of the analyses @xcite suggest the following : there is a second - order critical point at finite @xmath0 and @xmath36 .
the first order phase transition line exist , whose one end is connected with the critical point and the other is on the @xmath36 axis . summarizing the suggestions above and the expectations based on the lattice qcd and the asymptotic freedom , a schematic figure of a possible phase diagram of qcd is drawn in fig .
[ fig : phasediagram ] .
in this thesis , we focus on the analysis of the qgp phase .
qgp is considered to be realized by heavy ion collision experiment @xcite , which is performed in the relativistic heavy ion collider in brookhaven national laboratory and large hadron collider in cern .
qgp was also realized in the early universe since it is considered that temperature was higher than @xmath35 when the age of universe is @xmath37 seconds or younger . for this reason ,
the properties of qgp are relevant to the analysis of the early universe .
qgp is the only experimentally realizable fermion - boson system in which the masses are negligible compared with temperature . because of this property , fermionic collective excitations appear in qgp , which is the main topic of this thesis .
we note that the yukawa model and qed at high temperature have that property , so we also treat these models , which are simpler than qcd , in this thesis . throughout this thesis ,
the yukawa model means the model in which a fermion ( denoted by @xmath38 ) is coupled with a scalar field @xmath39 .
the lagrangian is @xmath40= \frac{1}{2}(\partial^\mu\phi)^2+\overline{\psi}i({\ooalign{\hfil/\hfil\crcr$\partial$}}+i{g}\phi)\psi.\end{aligned}\ ] ] we do not include the possible self - coupling of the scalar fields for simplicity . in the following ,
we consider the case that the temperature ( @xmath0 ) is so high that the current quark mass and the nonperturbative effect are negligible : @xmath41 , @xmath42 . here
@xmath42 is a scale parameter , which is of order 200 mev .
the inequality @xmath43 implies that @xmath44 so that the perturbative expansion in terms of the coupling constant is applicable .
we consider the case that the chemical potential is zero ( @xmath45 ) .
some of @xmath46 are very large , so it seems that the temperature at which the massless approximation is valid is too high to be realized .
however , the heavy quarks are almost decoupled from the system if their masses are much larger than @xmath0 , so that approximation is expected to be valid for light quarks in that case .
one may also have doubt on the validity of the perturbative expansion : the coupling constant is not small enough for justification of the perturbative expansion in realistic temperature . nevertheless , there are some discussions which suggest that the perturbative expansion is at least partially applicable if the temperature is higher than @xmath47 . for example
, the entropy calculated from approximately self - consistent scheme which is motivated by the existence and the properties of the collective excitations predicted from the perturbation theory , agrees very well with that calculated from the lattice qcd @xcite . in the temperature region
described above , the perturbative method suggests that the system has multi - energy scale structure @xcite , which will be reviewed in the following subsections and summarized in fig .
[ fig : intro ] .
we note that the structure appears not only in qcd but also in the yukawa model and qed . among the properties of qgp
, we focus on the spectrums of the quarks and the gluons .
these quantity are quite important to elucidate the picture how the quarks and the gluons , which are the basic building blocks of qcd , behave in the system .
naively , that problem seems to be trivial in the weak coupling regime : a naive expectation might be that the spectrum of the particle seems to be almost the same as that of free particle since @xmath9 is small .
we will see that it is not the case when the energy of the particle is @xmath48 or @xmath49 in the following subsections .
[ [ properties - of - hard - particles ] ] properties of hard particles + + + + + + + + + + + + + + + + + + + + + + + + + + + + the energy scale @xmath0 is called hard scale .
one example of appearance of this scale in the theory is the average inter - particle distance , which can be shown as follows @xcite : for simplicity , we show in the case of the boson in the yukawa model .
the number density of the boson in the free limit is @xmath50 where @xmath51 is the bose - einstein distribution function , and @xmath52 is the riemann zeta function .
@xmath53 is approximately equal to 1.202 .
we note that the volume occupied by one particle is @xmath54 .
since @xmath55 is of order @xmath0 , the average inter - particle distance is of order @xmath56 .
the de broglie wavelength of the hard particle , whose momentum is of order @xmath0 , is also of order @xmath56 .
this fact suggests that the properties of the hard particle is not so different from those of the free particle because the overlap among the particles is not large enough to compensate the smallness of the coupling constant .
we show that this suggestion is valid in the following .
one modification of the dispersion relation of the hard particle is the momentum - independent mass called asymptotic thermal mass @xcite , which is of order @xmath57 .
the expressions of the asymptotic thermal masses for the fermions and the bosons , in the yukawa model and qed / qcd , are as follows : @xmath58 here @xmath59 ( @xmath60 ) , @xmath61 ( @xmath62 ) , and @xmath63 ( @xmath64 ) are the squares of the masses for the fermion ( boson ) in the yukawa model @xcite , electron ( photon ) in qed @xcite , quark ( gluon ) in qcd @xcite , respectively , and we have introduced @xmath65 .
we use @xmath9 as a coupling constant not only in qcd but also in the yukawa model and qed instead of the standard notation ( @xmath66 in the case of qed ) , to make it clear that the same order counting appears as that in qcd in this thesis . to demonstrate that the momentum - independent masses appear as a result of the perturbative analysis ,
here we perform an analysis at the one - loop order in the yukawa model . in the present and the next chapter ,
we employ the real - time formalism in keldysh basis .
the retarded self - energy of a massless fermion ( the boson ) @xmath67 ( @xmath68 ) at @xmath69 is defined as follows : @xmath70 where @xmath71 ( @xmath72 ) is the retarded fermion ( boson ) propagator .
the pole positions of the fermion and the boson are determined by the conditions that the inverse of the retarded propagators are zero : @xmath73 note that since @xmath69 and the fermion is massless , @xmath67 can be expressed in terms of the scalar functions @xmath74 and @xmath75 as @xmath76 with @xmath77 .
thus we see that the self - energy gives the modification of the pole position due to the interaction .
first , we derive the expression for the asymptotic thermal mass of the boson .
the boson retarded self - energy at the one - loop order is given by @xmath78 , \end{split}\end{aligned}\ ] ] whose diagrammatic expression is shown in fig .
[ fig : boson - selfenergy ] .
@xmath79 are the bare propagators of the fermion which are defined as @xmath80 here we have introduced the fermi - dirac distribution function @xmath81 and the free spectral function @xmath82 which is given by @xmath83 since the dominant contribution to @xmath84 comes from the region @xmath85 , as will be confirmed later , @xmath86 , which is smaller than @xmath87 .
for this reason , @xmath88 which satisfies the pole condition ( eq .
( [ eq : intro - boson - polecondition ] ) ) is expected to be approximately equal to @xmath89 , so @xmath90 is negligible in the calculation of the asymptotic thermal mass . by neglecting @xmath90 , we get @xmath91 where we have used eq .
( [ eq : integral - formula - f1 ] ) . from this expression and eq .
( [ eq : intro - boson - polecondition ] ) , we confirm eq .
( [ eq : mass - boson ] ) .
next we evaluate the asymptotic thermal mass of the fermion . the diagram for the retarded fermion self - energy at the one - loop order
is drawn in fig .
[ fig : fermion - selfenergy ] .
its expression is @xmath92 , \end{split}\end{aligned}\ ] ] where @xmath93 are the bare propagators of the scalar boson defined as @xmath94 inserting eqs .
( [ eq : bare - propagators ] ) , ( [ eq : bare - propagatorsf ] ) , ( [ eq : bare - propagatorer ] ) , and ( [ eq : bare - propagatore ] ) into ( [ eq : one - loop - bare ] ) , we obtain @xmath95 , \end{split}\end{aligned}\ ] ] where we have used the on - shell conditions for the bare particles , @xmath96 , and @xmath97 in @xmath98 and @xmath99 . to obtain the expression of the asymptotic thermal mass
, we only have to calculate @xmath100 , which is easier than calculation of @xmath101 when @xmath102 is hard .
let us see this . by using the fact @xmath103 , eq .
( [ eq : intro - fermion - polecondition ] ) becomes @xmath104 where the terms which are of order @xmath105 have been neglected . thus we see that the square of the asymptotic thermal mass is equal to @xmath100 . by neglecting @xmath90 as in the boson case , @xmath100 reads @xmath106 where @xmath107 note that @xmath108 is independent of @xmath102 . by using the formulae eqs .
( [ eq : integral - formula - f1 ] ) and ( [ eq : integral - formula - b1 ] ) , @xmath100 becomes @xmath109 from this expression and eq .
( [ eq : intro - fermion - polecondition - calculate ] ) , we confirm eq .
( [ eq : mass - fermion ] ) . .
, scaledwidth=40.0% ] the other modification of the dispersion relation of the hard particles is the damping rate , which comes from the imaginary part of the self - energies . in the yukawa model ,
the imaginary parts of the self - energies at the leading order have the following forms : @xmath110 where @xmath111 and @xmath112 , which will be found to be the damping rate of the fermion and the boson at the leading order later , are the real numbers .
the leading contribution to @xmath112 and @xmath111 is found to come from the two - loop diagrams containing the collision effect among the hard particles , not from the one - loop diagrams @xcite .
the damping rates of the hard particles are of order @xmath113 ( the fermion and the boson in the yukawa model @xcite , photon ) or @xmath114 ( electron , quark , gluon ) .
the damping rates of the electron , the quark , and the gluon at the leading order are known to be independent of momentum @xcite .
the expressions of the damping rates are obtained for the electron ( @xmath115 ) , the quark ( @xmath116 ) , and the gluon ( @xmath117 ) , and only at the leading - log order as a large quantity .
therefore @xmath118 which appears in the order estimate of the damping rate is sometimes omitted . ] : @xmath119 , \\ { \zeta_q}&={g}^2t{c_f}\left[\frac{1}{4\pi}\ln\left(\frac{1}{{g}}\right ) + o(1)\right ] , \\ { \zeta_g}&={g}^2tn\left[\frac{1}{4\pi}\ln\left(\frac{1}{{g}}\right )
+ o(1)\right ] .\end{aligned}\ ] ] the asymptotic thermal mass and the damping rate indeed modify the pole position of the hard particle . in the case of the yukawa model , eqs .
( [ eq : intro - fermion - polecondition ] ) , ( [ eq : intro - boson - polecondition ] ) , ( [ eq : intro - thermalmass - boson - final ] ) , ( [ eq : intro - fermion - thermalmass - final ] ) , ( [ eq : intro - imaginary - selfenergy - boson ] ) , and ( [ eq : intro - imaginary - selfenergy - fermion ] ) lead to @xmath120 where @xmath121 or @xmath22 , and @xmath122 is the damping rate of each particle .
the terms of the asymptotic thermal mass and the damping rate , which give the corrections to the dispersion relation and come from the self - energy , are of order @xmath123 and much smaller than the momentum term , which is of order @xmath124 .
this fact is consistent with the expectation that the medium correction to the properties of the hard particle is small . in this sense
, we can regard the fermion and the boson with the hard energy as independent particle excitations .
[ [ relevance - of - hard - particles - to - other - quantities ] ] relevance of hard particles to other quantities + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we make a comment on relevance of the hard particle to the physical quantities which is not sensitive to the infrared energy region . whereas the number density of the hard particle is of order @xmath125 due to the large phase space as was shown before
, that of the fermion ( boson ) whose energy is of order @xmath57 is much smaller and of order @xmath126 ( @xmath127 ) @xcite , as will be shown in the case of the boson in the yukawa model in the following : first we introduce the cutoff parameter @xmath128 which satisfies @xmath129 . from eq .
( [ eq : intro - boson - number ] ) , the number density of the free particle whose energy is less than @xmath128 is @xmath130 here we have used @xmath131 for @xmath132 .
this order estimate implies that the dominant contribution comes from @xmath133 because @xmath134 is of order @xmath125 and much larger than @xmath135 . by setting @xmath136
, we see that the contribution from @xmath137 is of order @xmath138 .
this fact implies that most of the particles have the hard energy .
for this reason , it seems that the hard particles determine physical quantities which are not sensitive to the infrared region at the leading order .
actually , the leading contribution to some thermodynamic quantities ( pressure and energy @xcite ) and dynamical quantities ( transport coefficients , self - energies of the fermion and the boson ) , comes from a part in which the energy of the particle in the loop integral is hard .
we demonstrate this in the case of the hard and on - shell boson self - energy . by using eq .
( [ eq : intro - thermalmass - boson - final ] ) , the contribution to @xmath84 from the energy region @xmath139 is of order @xmath140 here we have used @xmath141 for @xmath132 .
this order estimate means that the contribution from the hard particle , @xmath142 , is much larger than that from the particle with smaller energy , @xmath143 .
nevertheless , we note that there are some quantities which are sensitive to the energy region @xmath144 .
thus in analyzing such quantities , it is necessary to take into account the contribution from the soft particles even at the leading order .
examples of such quantities will be introduced in the next subsection .
[ [ properties - of - soft - particles - and - hard - thermal - loop - approximation ] ] properties of soft particles and hard thermal loop approximation + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the energy scale @xmath57 is called soft scale . in this scale , the self - energies and the inverses of the bare propagators
have the same order of the magnitude in eqs .
( [ eq : intro - fermion - polecondition ] ) and ( [ eq : intro - boson - polecondition ] ) , as will be shown later .
this fact suggests that the medium effects can not be neglected even at the weak coupling when the energy scale is soft ( @xmath145 ) .
thus , bosonic and fermionic collective excitations can appear in this energy region .
there are transverse and longitudinal excitations @xcite in the bosonic sector , and the latter is called plasmon and does not exist in the vacuum . to confirm that fact and to see the expression of the dispersion relations and the residues of the collective excitations , we here calculate the photon self - energy at the one - loop order in qed , adopting the coulomb gauge .
the retarded photon self - energy at the one - loop level is given by @xmath146\\ & \simeq { g}^2\int\frac{d^4k}{(2\pi)^4 } { n_f}(k^0)\rho^0(k ) { \mathrm{tr}}\left[\gamma^\mu\left(\frac{2k^\nu{\ooalign{\hfil/\hfil\crcr$k$}}+{\ooalign{\hfil/\hfil\crcr$k$}}\gamma^\nu{\ooalign{\hfil/\hfil\crcr$p$}}}{2k\cdot p+p^2 } -\frac{2k^\nu{\ooalign{\hfil/\hfil\crcr$k$}}-{\ooalign{\hfil/\hfil\crcr$p$}}\gamma^\nu{\ooalign{\hfil/\hfil\crcr$k$}}}{2k\cdot p - p^2}\right)\right ] , \end{split } \end{aligned}\ ] ] where we have neglected the contribution from @xmath0-independent part since it can be eliminated by the renormalization of photon wave function in the vacuum .
we note that not only the quark but also the gluon contributes to the gluon self - energy in qcd in contrast to the case of qed . since we are focusing on the soft region , we can utilize the useful condition @xmath147 .
it is justified since @xmath148 while @xmath85 , which was shown in the previous subsection . by neglecting the terms which are of order or much smaller than @xmath149 ,
we arrive at @xmath150 \\ & = 4{g}^2\int\frac{d^4k}{(2\pi)^4 } { n_f}(k^0)\rho^0(k ) \frac{1}{k\cdot
p}\left(p^\mu k^\nu+k^\mu p^\nu-\frac{p^2}{k\cdot p}k^\mu k^\nu \right ) -\frac{3{\omega_{0\gamma}}^2}{2}g^{\mu\nu } , \end{split}\end{aligned}\ ] ] where @xmath151 .
this approximation is called the hard thermal loop ( htl ) approximation . here
we decompose the photon propagator into the transverse and the longitudinal component in the coulomb gauge : @xmath152 where @xmath153 and @xmath154 are the transverse and the longitudinal components of the retarded self - energy .
here we omitted the subscript `` bare '' for simplicity . from eq .
( [ eq : htl - boson - selfenergy - result ] ) , the retarded self - energy in each component is calculated as @xmath155 , \\
\label{eq : htl - selfenergy - longitudinal } \varpi^{{r}l}(p)&\simeq 3{\omega_{0\gamma}}^2\left(1-\frac{p^0}{2|{{\mathbf{p}}}|}\ln\frac{p^0+|{{\mathbf{p}}}|}{p^0-|{{\mathbf{p}}}|}\right),\end{aligned}\ ] ] where we have used eq .
( [ eq : integral - formula - cos2 ] ) in the computation of the transverse component .
thus , the dispersion relations in the transverse and longitudinal sectors read @xmath156&= 0,\\ \label{eq : htlresult - dispersion - longitudinal } respectively . since eqs .
( [ eq : htlresult - dispersion - transverse ] ) and ( [ eq : htlresult - dispersion - longitudinal ] ) are invariant for the transformation @xmath157 , we see that the positive energy solution and the negative energy solution are degenerated .
we analyze only the positive energy solutions from now on .
the dispersion relations of the excitations in the transverse ( @xmath158 ) and the longitudinal sector ( @xmath159 ) are plotted in the left panel of fig .
[ fig : htl - dispersion ] .
the longitudinal excitation is called plasmon . by using @xmath145
, we see that the terms coming from the inverse of the free propagator and the terms coming from the self - energy in eqs .
( [ eq : htlresult - dispersion - transverse ] ) and ( [ eq : htlresult - dispersion - longitudinal ] ) have the same order of magnitude in contrast to the case of the hard particle .
in this sense , the medium effect can not be regarded as a weak perturbation in the soft region .
the damping rates of the two excitations are zero in the htl approximation , and their leading contribution comes from the two - loop diagrams .
they are of order @xmath160 @xcite , so the damping rate of the bosonic excitations are much smaller than the excitation energies , which is of order @xmath57 .
the expressions for the residues of the two excitations are given by @xmath161 the asymptotic forms of the dispersion relations for momenta which satisfies @xmath162 are as follows : @xmath163 both expressions coincide at @xmath164 because it is impossible to distinguish between transverse and longitudinal excitations in that case . from this
expressions , we see that the energies of both branches are @xmath165 at @xmath164 . thus , we understand the physical meaning of @xmath165 : the plasma frequency . on the other hand , for @xmath166 which satisfies @xmath167 ,
the dispersion relations are approximated as @xmath168}\right).\end{aligned}\ ] ] we see that both dispersion relation approaches the light cone as the momentum becomes large .
however , we note that the residue of the plasmon is exponentially small for large momenta : @xmath169 these observations support the expectation that the medium correction is suppressed by powers of @xmath9 when the momentum is much larger than @xmath57 .
equation ( [ eq : htl - result - dispersion - largep - transverse ] ) tells us that the mass of the transverse photon with large momenta is @xmath170 , which coincides the asymptotic mass given in eq .
( [ eq : mass - photon ] ) .
this coincidence is unexpected since the htl approximation is not valid when @xmath171 . , while the horizontal axis is the momentum @xmath166 .
the dotted lines denote the light cone . left panel : the dispersion relation in the bosonic sector . the solid line with larger energy represents the transverse mode , while the other the longitudinal mode ( plasmon ) .
the residue of the longitudinal mode becomes exponentially small for @xmath172 , so the plot of that mode does not represent physical excitation in that momentum region .
right panel : the dispersion relation in the fermionic sector .
the solid line in the positive energy region corresponds to the normal fermion , while that in the negative energy region corresponds to the antiplasmino .
note that since we focus on the fermion sector , the antiplasmino appears instead of the plasmino .
the residue of the antiplasmino becomes exponentially small for @xmath172 , so the plot of the antiplasmino does not represent physical excitation in that momentum region .
, title="fig:",scaledwidth=45.0% ] , while the horizontal axis is the momentum @xmath166 .
the dotted lines denote the light cone .
left panel : the dispersion relation in the bosonic sector .
the solid line with larger energy represents the transverse mode , while the other the longitudinal mode ( plasmon ) .
the residue of the longitudinal mode becomes exponentially small for @xmath172 , so the plot of that mode does not represent physical excitation in that momentum region .
right panel : the dispersion relation in the fermionic sector .
the solid line in the positive energy region corresponds to the normal fermion , while that in the negative energy region corresponds to the antiplasmino .
note that since we focus on the fermion sector , the antiplasmino appears instead of the plasmino .
the residue of the antiplasmino becomes exponentially small for @xmath172 , so the plot of the antiplasmino does not represent physical excitation in that momentum region .
, title="fig:",scaledwidth=45.0% ] on the other hand , in the fermionic sector , there are also two independent excitations : one is called normal fermion and the other plasmino . the plasmino does not exist in the vacuum as the plasmon does not .
let us perform the calculation of the self - energy using the htl approximation in qed , to show the existence and the properties of the fermionic excitations introduced above .
the retarded self - energy is expressed as @xmath173\gamma^\nu,\end{aligned}\ ] ] where @xmath174 is the bare propagators of the photon defined as @xmath175 where @xmath176 and @xmath177 is the projection operator on the transverse direction , @xmath178 with @xmath179 . the diagram corresponding to eq .
( [ eq : intro - oneloop - fermion - qed ] ) is shown in fig .
[ fig : fermion - selfenergy ] . before proceeding the calculation ,
let us show that the term which contains @xmath180 in @xmath181 does not contribute to the result in the htl approximation .
the contribution from that term becomes @xmath182 here we have used @xmath147 and neglected the @xmath183 contribution .
after @xmath184 integration , this expression vanishes .
thus by using the condition @xmath185 as in the bosonic case , eq .
( [ eq : intro - oneloop - fermion - qed ] ) becomes @xmath186 from this expression , we see that @xmath187 on account of @xmath148 .
this order - estimate implies that @xmath188 , so the computation of @xmath189 does not give us the dispersion relation of the electron in contrast to the case of @xmath171 .
therefore we proceed the calculation of @xmath190 instead of evaluation of @xmath189 .
equation ( [ eq : intro - htl - selfenergy - electron ] ) becomes @xmath191\right ] , \end{split}\end{aligned}\ ] ] in the last line we have used eqs .
( [ eq : integral - formula - f1 ] ) and ( [ eq : integral - formula - b1 ] ) and introduced @xmath192 .
this expression can be rewritten as @xmath193.\end{aligned}\ ] ] by substituting this expression into eq .
( [ eq : intro - propagator - selfenrgy - fermion ] ) , we obtain @xmath194 from the spinor structure in the numerators of the two terms in the right - hand side , we see that these terms are eigenstates of ( chirality)/(helicity )
. the eigenvalue of the first term is @xmath195 while that of the second term is @xmath196 . in the vacuum , the fermion number of the former is @xmath195 while that of the latter is @xmath196 . from eq .
( [ eq : intro - htl - propagator - fermion ] ) , the dispersion relations of the collective excitations which has the fermion number @xmath195 is @xmath197 = 0.\end{aligned}\ ] ] the two dispersion relations are obtained as solutions of this equation , which are denoted by @xmath198 and @xmath199 .
these dispersion relations are plotted in the right panel of fig .
[ fig : htl - dispersion ] .
we note that the excitations whose dispersion relations are @xmath200 and @xmath201 also appear in the fermion number @xmath196 sector .
the excitation whose dispersion relation is @xmath198 is called normal fermion , @xmath200 the plasmino , and @xmath199 the antiplasmino , respectively .
the residue of each branch is @xmath202 eq .
( [ eq : htl - result - dispersion - fermion ] ) can be solved explicitly for @xmath166 which satisfies @xmath162 or @xmath203 . in the case of @xmath162 ,
the dispersion relations become @xmath204 from this expression , we understand the physical meaning of @xmath205 : the counterpart of plasma frequency for fermion .
the residue becomes @xmath206 this expression tells us that both branches have the same strength when they are at rest . on the other hand , for @xmath167 ,
the dispersion relations become @xmath207 and the residue is @xmath208,\\ z_-(|{{\mathbf{p}}}|)&\simeq \frac{2|{{\mathbf{p}}}|^2}{{\omega_{0e}}^2 } e^{-2|{{\mathbf{p}}}|^2/{\omega_{0e}}^2 - 1}.\end{aligned}\ ] ] from these equations , we see that the residue of the plasmino becomes exponentially small as @xmath166 becomes large .
we also see that the dispersion relation of the normal fermion approaches the light cone and the residue approaches unity , as @xmath166 becomes large .
these two facts can be understood from the expectation that the medium effect is suppressed for @xmath209 , as explained in the previous subsection .
equation ( [ eq : htlresult - dispersion - plus - largep ] ) can be expressed as @xmath210 if we neglect higher order term , which implies that the mass of the normal fermion with @xmath211 is equal to the asymptotic thermal mass @xmath212 at the leading order .
it is nontrivial because the htl approximation is not valid when @xmath171 .
we emphasize that the plasmino is a novel excitation which reflects the fact that qgp is a fermion - boson system at ultrarelativistic temperature : if each of the mass of the quark and the gluon is not negligible , the fermion self - energy would be suppressed and the plasmino would not appear . since @xmath213 as was shown before , we note that the fermion retarded self - energy and the momentum of the soft particle in eq .
( [ eq : intro - propagator - selfenrgy - fermion ] ) have the same order of magnitude ( @xmath214 ) , as in the case of the boson .
the damping rates of these excitations are zero in the htl approximation . by evaluating the two - loop diagrams containing the effect of the collision , they are found to be of order @xmath215 @xcite .
thus the damping rates of the fermionic excitations are much smaller than the excitation energy @xmath216 . the normal fermion and the plasmino exists not only in @xmath69 case but also in the case that @xmath36 is finite .
it is not straightforward to give an interpretation of the two fermionic excitations at finite @xmath0 since it does not have their classical counterparts , but at zero temperature and finite chemical potential case , it is possible to clarify the state which forms the plasmino @xcite . from that analysis
, it was suggested that the state of the plasmino is superposition of an antifermion state and the state composed of a boson and a hole .
the extension of such analysis to the finite temperature case is an interesting task .
we also note that there is another attempt of interpretation in terms of the idea of resonant scattering @xcite .
in addition to collective excitations , the htl approximation also leads to the debye screening @xcite .
the screening mass is @xmath217 which is of order @xmath57 .
[ [ relevance - of - soft - particles - to - other - quantities ] ] relevance of soft particles to other quantities + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the knowledge of the spectra of the soft excitations is necessary not only to establish the picture of the particles , but also to calculate the quantities which are sensitive to the soft energy region .
examples include the gluon damping rate at rest .
that quantity was first calculated by using the bare perturbation expansion and found to be dependent on the gauge - fixing , though that quantity should be gauge - independent .
the solution was given from the following observation : since properties such as the dispersion relation of the soft particles are different largely from that in the free limit because of the medium effect , we have to use the resummed propagators ( eqs .
( [ eq : htl - photon - propagator ] ) and ( [ eq : intro - htl - propagator - fermion ] ) ) which include the information of the result of the htl approximation instead of the bare propagators , when soft particles appear in the loop integral .
vertex function also needs to be resummed in such a way .
this resummation is called htl resummation , and it is essentially important to get a sensible result . by using the htl resummation ,
the sensible and gauge - independent result was obtained for the gluon damping rate at rest .
that method was also applied to the analysis of the quark damping rate at zero spatial momentum . [ [ vlasov - equation ] ] vlasov equation + + + + + + + + + + + + + + +
the htl approximation is a diagrammatic technique for quantities at equilibrium .
nevertheless , due to the linear response theory , the htl results should be reproduced from the analysis of the nonequilibrium time evolution at the leading order in the case where the system is close to the equilibrium state .
actually , the htl approximation corresponds to the linearized vlasov equation , which is a collisionless kinetic equation , as is shown in the case of the photon self - energy in the following : the vlasov equation which describes the time evolution of the electron distribution function under the background electromagnetic field reads @xmath218 where @xmath219 is the four - velocity of a massless particle , @xmath220 the electron distribution function at the nonequilibrium state , @xmath221 the background electric field , @xmath222 the background magnetic field with @xmath223 being the background gauge field , respectively .
the left - hand side describes the time evolution in the free limit , and is called the drift term .
the right - hand side contains the effect of the interaction between the electron and the background electromagnetic field , and is called the force term . by linearizing eq .
( [ eq : vlasov ] ) , we get @xmath224 where @xmath225 is the deviation of the electron distribution function from the value at equilibrium .
we note that the background magnetic field does not contribute in this order .
the induced current @xmath226 can be expressed in terms of @xmath227 as follows : @xmath228 where the degeneracy of the spin of electron and the contribution from the positron have been taken into account . by using this expression , we get the following expression of the polarization tensor by performing the fourier transformation ( @xmath229 ) : @xmath230 here we have used the relation in the linear response theory , @xmath231 , and introduced @xmath232 .
the infinitesimal number @xmath233 comes from the condition that the background field is introduced adiabatically : @xmath234 and @xmath235 vanish when @xmath236 .
this expression yields eqs .
( [ eq : htl - selfenergy - transverse ] ) and ( [ eq : htl - selfenergy - longitudinal ] ) after the @xmath184 integration , which means that the linearized vlasov equation reproduces the photon self - energy in the htl approximation . we can also generalize the vlasov equation to the case where the background field is fermionic .
in this situation , we are to analyze the time evolution of the amplitude of the process in which a hard fermion becomes a hard boson and its inverse process , instead of the distribution function whose time evolution was analyzed above ( see chapter [ chap : kinetic ] ) . by performing such analyses , the fermion retarded self - energy eq .
( [ eq : htl - selfenergy - fermion - result ] ) in the htl approximation , is reproduced .
[ [ resummed - perturbation - and - boltzmann - equation ] ] resummed perturbation and boltzmann equation + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the energy scale @xmath49 is called ultrasoft scale .
this scale appears in the damping rate , or equivalently , the inverse of the relaxation time of the hard particle , as was shown in sec .
[ ssc : intro - hard ] .
for this reason we expect that the collision effect becomes important in the ultrasoft region ( @xmath237 ) . to demonstrate that the collision effect is not negligible in the ultrasoft region , we consider the linearized boltzmann equation , which is the kinetic equation containing the collision effect , in the relaxation time approximation .
the situation is the same as in eq .
( [ eq : vlasov ] ) : the particle is electron , and the background field is the electromagnetic one .
that equation reads @xmath238 the second term in the left - hand side is the collision term in the relaxation time approximation . here
@xmath239 is a typical relaxation time for the hard electron , whose inverse has the same order of magnitude as the damping rate of the hard electron ( @xmath115 ) , which is of order @xmath240 , so it seems that we do not need to take into account the interaction among the hard particles in the ultrasoft region .
however , the difference of the thermal masses of the fermion and the boson , which is of order @xmath123 , plays the similar effect to that of the mean free path , as will be shown in chapter [ chap : ultrasoft ] , so we can not neglect the interaction effect when we analyze the fermion self - energy with ultrasoft momentum . ] . from this equation , we see that the collision term is not negligible in the case of ultrasoft region ( @xmath241 ) since that term and the drift term have the same order of magnitude .
in fact , the boltzmann equation should be used instead of the vlasov equation in the calculation of the gluon @xmath4-point function in this energy region .
we will show in chapter [ chap : kinetic ] that the other interaction effects such as the asymptotic thermal mass should be taken into account as well as the collision , in the analysis of the ultrasoft fermion propagator .
we note that the asymptotic thermal mass and a part of the collision effect correspond to the real and the imaginary part of eq .
( [ eq : intro - dispersion - hard ] ) , respectively .
the boltzmann equation in the analysis of the gluon @xmath4-point function can be translated into a diagrammatic language , and the resultant diagrammatic method is not a simple one - loop approximation ( htl approximation ) , but the resummed perturbation which resums the damping rate of the hard particles and sums up the ladder diagrams .
in addition to the gluon @xmath4-point function , there are other quantities whose calculation need considering the interaction effect among the hard particles . for example , the transport coefficient is such quantity . that quantity can be obtained from the kubo formula , by taking the zero energy limit of the green function as follows @xcite : the shear viscosity @xmath242 , the bulk viscosity @xmath243 , the electrical conductivity @xmath244 in qed , and the flavor diffusion constant @xmath245 in qcd are given by @xmath246 \rangle , \\
\zeta&= \frac{1}{2}\lim_{\omega\rightarrow 0 } \frac{1}{\omega } \int d^4x e^{i\omega x^0}\langle[p(x ) , p(0 ) ] \rangle , \\
\sigma&= \frac{1}{6}\lim_{\omega\rightarrow 0 } \frac{1}{\omega } \int d^4x e^{i\omega x^0}\langle[j_{i}(x ) , j_{i}(0 ) ] \rangle , \\
d_{\alpha\beta}&= \frac{1}{6}\lim_{\omega\rightarrow 0 } \frac{1}{\omega } \int d^4x e^{i\omega x^0}\langle[j^\alpha_{i}(x ) , j^\gamma_{i}(0 ) ] \rangle \varxi^{-1}_{\gamma\beta},\end{aligned}\ ] ] respectively , where @xmath247 is the local pressure , @xmath248 the traceless part of the energy - momentum tensor with @xmath249 being the energy - momentum tensor , @xmath250 the electromagnetic current , @xmath251 the flavor current , respectively , and we have introduced @xmath252 . because the zero energy limit is taken
, the energy scale of that quantity is much less than @xmath49 .
for this reason , the analysis of that quantity also needs including the interaction effect among the hard particles .
accordingly , the transport coefficient can be calculated either by the boltzmann equation or by the resummed perturbation - particle irreducible formalism .
all of these methods produce the same result in the leading order of the coupling constant . ] . using the correspondence between that perturbation theory and the boltzmann equation , the resummation scheme in the resummed perturbation theory
is interpreted with the language of the kinetic theory .
[ [ suggestion - on - existence - of - ultrasoft - fermion - mode ] ] suggestion on existence of ultrasoft fermion mode + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + when the energy scale is much below @xmath49 , the hydrodynamics works well , and the bosonic hydrodynamic modes such as the phonon and the diffusion mode appear . on the other hand ,
the fermionic sector whose energy is of order @xmath49 has not been well investigated , due to the difficulty of taking into account of the interaction among the hard particles .
for this reason , it has not been studied well whether there are any modes in this region .
nevertheless , there have been some suggestive works for supporting the existence of such an ultrasoft fermion mode at finite temperature .
some of them are based on supersymmetry , and others are based on analyses using effective model , as will be shown later .
first we introduce the works related to supersymmetry .
historically , the ultrasoft fermion mode at finite @xmath0 was found in supersymmetric models as nambu - goldstone fermions called goldstino associated with spontaneous breaking of supersymmetry at @xmath253 @xcite , by using ward - takahashi identity and a diagrammatic technique .
the complete calculation at the leading order was first performed by lebedev and smilga @xcite .
because the supercurrent and the energy - momentum tensor are in the same supermultiplet , the goldstino was also regarded as the supersymmetric analogue of the phonon , phonino . here
we note that the analysis in @xcite was performed in the temperature region @xmath254 , where @xmath255 is the bare mass .
it implies that their analysis is only valid for @xmath256 , but not for @xmath257 for which our analysis in the present paper is concerned .
the analysis was extended to qcd at so high temperature that the coupling constant is weak @xcite , in which a supersymmetry is still assigned at the vanishing coupling , and hence , the supersymmetry is , needless to say , explicitly broken by the interaction .
thus , there exists no exact fermionic zero mode but only a pseudo - zero mode does . although these analyses are suggestive , it is still obscure whether a genuine ultrasoft fermion mode exists when supersymmetry is absent , in particular , at extremely high temperature .
here we note that there have been suggestions of the existence of ultrasoft fermion mode at finite @xmath0 even without supersymmetry .
it was shown in one - loop calculations @xcite that when a fermion is coupled with a massive boson with mass @xmath255 , the spectral function of the fermion gets to have a novel peak in the far - low - energy region in addition to the normal fermion and the antiplasmino , when @xmath258 , irrespective of the type of boson ; it means that the spectral function of the fermion has a three - peak structure in this temperature region .
it was suggested that such a three - peak structure may persist even at the high temperature limit in the sense @xmath259 , for the massive vector boson on the basis of a gauge - invariant formalism , again , at the one - loop order @xcite .
thus , one may expect that the novel excitation may exist in the far - infrared region also for a fermion coupled with a massless boson , although the one - loop analysis admittedly may not be applicable at the ultrasoft momentum region .
there are also the works suggesting the existence of the ultrasoft fermion mode using the schwinger - dyson equation ; we note , however , that it is difficult to keep gauge symmetry in the schwinger - dyson approach at finite @xmath0 .
the analysis of the quark spectrum around @xmath35 using the njl model was also performed @xcite , and as a result , the existence of the ultrasoft fermion mode was suggested due to the coupling between the quark and the mesons .
finally let us give a generic argument supporting the existence of an ultrasoft fermion mode at finite @xmath0 on the basis of the symmetry of the self - energy for a massless fermion .
it was shown that the chiral , parity , and the charge symmetry make the fermion retarded propagator have the following structure @xcite : @xmath260 here we have introduced @xmath261 , and these functions satisfy @xmath262 by setting @xmath164 , we get @xmath263 eqs .
( [ eq : symmetry - s1 ] ) and ( [ eq : symmetry - s2 ] ) imply @xmath264 by using @xmath265 , we get @xmath266 where @xmath88 is real .
this property implies that @xmath267 , which is the real part of the inverse of the retarded fermion propagator , is zero at @xmath268 if there is no singularity .
for this reason , it is suggested that the fermion retarded propagator always have the pole around the origin , provided that the imaginary part of the fermion retarded self - energy is small enough .
this argument suggests that the existence of the ultrasoft pole may be a universal phenomenon at high temperature in the theory composed of massless fermion coupled with a boson , as long as chiral , parity , and charge symmetry exist .
we note that the attenuation of the pole at the origin in the case that the fermion has finite bare mass @xcite , is consistent with that argument because the finite fermion mass breaks the chiral symmetry .
[ [ what - we - do - in - the - thesis ] ] what we do in the thesis + + + + + + + + + + + + + + + + + + + + + + + + in this thesis , we analyze the properties in the ultrasoft region , focusing on the fermion spectrum : we calculate the ultrasoft fermion spectrum by using the resummed perturbation which enables the calculation of the fermion propagator with ultrasoft momentum , and show that a novel fermionic mode exists in that energy region .
we also obtain the expressions for the dispersion relation , the damping rate , and the strength of that mode . as explained before , there is an equivalence between the diagrammatic method and the kinetic equation such as the equivalence between htl approximation and vlasov equation , or the resummed perturbation treating the gluon @xmath4-point function and boltzmann equation .
because of this equivalence , we expect that the resummed perturbation used in the analysis of the ultrasoft fermion is expected to be equivalent to some kinetic equation which contains the interaction effect among the hard particles .
we derive the generalized boltzmann equation that is equivalent to the basic equation of the resummed perturbation theory treating the ultrasoft fermion , from the kadanoff - baym equation , which describes the time evolution of the nonequilibrium system based on the field theory .
this thesis is organized as follows : chapter [ chap : ultrasoft ] is devoted to the analysis on the spectrum of the ultrasoft fermion using the resummed perturbation theory in the yukawa model and qed / qcd .
the analysis in the yukawa model and qed is based on ref .
@xcite . as a result of the analysis
, we show that a novel fermionic mode which we call ultrasoft fermion mode exists in that energy region , and obtain the expressions of the pole position and the strength of that mode .
we also show that the resultant fermion propagator and the vertex function satisfy the ward - takahashi identity in qed / qcd . in chapter [ chap :
kinetic ] , we derive the linearized and generalized boltzmann equation for ultrasoft fermion excitations , from the kadanoff - baym equation in a yukawa model and qed . we show that this equation is equivalent to the self - consistent equation in the resummed perturbation theory used in the analysis of the ultrasoft fermion spectrum at the leading order .
furthermore , we derive the equation that determines the @xmath4-point function with external lines for a pair of fermions and @xmath5 bosons with ultrasoft momenta in qed
. we also showed that the ward - takahashi identity is satisfied , and that identity can be derived the conservation law of the current .
the analysis in that chapter is based on ref .
finally we summarize this thesis and give the concluding remarks in chapter [ chap : summary ] . in appendix
[ app : temporal ] , we show that the results obtained in the coulomb gauge in this thesis , also can be obtained in the temporal gauge .
we write some formulae used in the text in appendix [ app : formula ] .
though there are some suggestion on existence of the ultrasoft mode as is written in the previous chapter , it is not a simple task to establish that fermionic mode exist in the ultrasoft region on the complete leading order calculation because of the infrared divergence called pinch singularity that breaks a naive perturbation theory , as will be briefly reviewed in the next section .
we remark that the same difficulty arises in the calculation of transport coefficients and the gluon self - energy in the ultrasoft energy region .
therefore , in this chapter , we analyze the fermion propagator in the ultrasoft energy region in yukawa model and qed / qcd using a similar diagrammatic technique in refs .
@xcite to regularize the pinch singularity .
we shall show that the retarded fermion propagator has a pole at @xmath269 ( @xmath270 for yukawa model and @xmath271 for qed / qcd ) with the residue @xmath272 for ultrasoft momentum @xmath273 taking into account the ladder summation .
this chapter is organized as follows : in sec .
[ sec : ultrasoft - yukawa ] , we discuss the ultrasoft fermion mode in yukawa model as a simple example without supersymmetry . in sec .
[ sec : ultrasoft - qed ] , we examine the ultrasoft fermion mode in qed .
we analytically sum up the ladder diagrams giving the vertex correction in the leading order , and find the existence of the ultrasoft fermion mode as in the yukawa model .
we shall also show that the resultant ultrasoft fermion propagator and the vertex satisfy the ward - takahashi identity .
we perform the analysis in qcd using the method which is similar to that in qed in sec . [
sec : ultrasoft - qcd ] .
section [ sec : ultrasoft - summary ] is devoted to a brief summary of this chapter .
the analysis in secs .
[ sec : ultrasoft - yukawa ] and [ sec : ultrasoft - qed ] is based on ref .
@xcite .
let us start with the yukawa model , which is the simplest model to study the ultrasoft fermion mode . generalization to gauge theory will be discussed in sec .
[ sec : ultrasoft - qed ] and [ sec : ultrasoft - qcd ] .
we calculate the fermion retarded self - energy and obtain the fermion retarded green function with an ultrasoft momentum @xmath237 .
we first see that the naive perturbation theory breaks down in this case .
then , we shall show that a use of a dressed propagator gives a sensible result in the perturbation theory and that the resulting fermion propagator has a new pole in the ultrasoft region .
the retarded self - energy in the one - loop level is given by eq .
( [ eq : one - loop - bare ] ) , and its diagrammatic representation is shown in fig .
[ fig : fermion - selfenergy ] . for small @xmath102 , the self - energy , which is expressed as eq .
( [ eq : intro - selfenergy - fermion - oneloop ] ) after some manipulations , is reduced to @xmath274 where @xmath108 is defined in eq .
( [ eq : kernel - definition ] ) .
this approximation is equivalent to the htl approximation .
the htl approximation is , however , only valid for @xmath148 , and not applicable in the ultrasoft momentum region . in fact , the retarded self - energy in the one - loop level obtained with use of the bare propagators is found to diverge when @xmath275 , since the integrand contains @xmath276 .
this singularity is called `` pinch singularity '' .
the origin of this singularity is traced back to the use of the bare propagators because the singularity is caused by the fact that the dispersion relations of the fermion and the boson are the same and the damping rates are zero in these propagators .
for this reason , one may suspect that this singularity can be removed by adopting the dressed propagators taking into account the asymptotic masses and damping rates of the quasiparticles , as will be shown to be the case shortly .
since the leading contribution comes from the hard ( @xmath277 ) internal and almost on - shell ( @xmath278 ) momentum ) or smaller is not relevant : in fact , the htl - resummed propagators should be used for soft momenta . however , the dispersion relations of the fermion and the boson obtained from these propagators are different from each other , so the pinch singularity will not appear in this case .
] , we are led to employ the following dressed propagators for the fermion and boson : @xmath279 where the expressions of @xmath280 and @xmath281 are given in eqs .
( [ eq : mass - fermion ] ) and ( [ eq : mass - boson ] ) . the damping rates of the hard particles , @xmath282 and @xmath283 , are of order @xmath284 .
the logarithmic enhancement for the damping rate is caused by the soft - fermion exchange , which is analogous to that of the hard photon .
note that these resummed propagators are the same as those used in @xcite , except for the smallness of the damping rates : we remark that such a smallness is not the case in qed / qcd , where the damping rate is anomalously large and of order @xmath285 ( `` anomalous damping '' ) . using these dressed propagators ,
we obtain @xmath286 for small @xmath102 , where @xmath287 , @xmath288 , @xmath289 , and @xmath290 .
we have used the modified on - shell condition of the quasi - particles , @xmath291 and @xmath292 , to obtain the denominator of the integrand in eq .
( [ eq : one - loop - pc ] ) .
we have also neglected @xmath293 , @xmath294 , @xmath295 , and @xmath296 in @xmath108 , since the leading contribution comes from hard momenta @xmath85 .
it is worth emphasizing that thanks to @xmath297 and @xmath243 , @xmath298 given in eq .
( [ eq : one - loop - pc ] ) does not diverge in the infrared limit , @xmath275 . before evaluating eq .
( [ eq : one - loop - pc ] ) , we introduce the the following dimensionless value : @xmath299 it can be computed as follows : @xmath300 in the last line , we have used eqs .
( [ eq : integral - formula - f1 ] ) and ( [ eq : integral - formula - b1 ] ) . from eq .
( [ eq : ultrasoft - lambda - computation ] ) , we see that @xmath301 is of order unity .
this value will characterize the strength of residue of the pole for both yukawa model and qed / qcd .
we expand the self - energy in terms of @xmath302 instead of @xmath303 itself .
this is the key point of our expansion , which enables us to analytically find the pole of the ultrasoft fermion mode .
then , the leading contribution is @xmath304 by using eqs .
( [ eq : integral - formula - f3 ] ) and ( [ eq : integral - formula - b3 ] ) , we get @xmath305 with @xmath306 and @xmath307 . note that the zeroth - order term is absent , which implies that there is no mass term . actually , it is guaranteed by the symmetries , as discussed in sec .
[ ssc : intro - ultrasoft ] .
thus , we obtain the fermion propagator in the ultrasoft region as @xmath308 here we have used @xmath309 and decomposed the fermion propagator into the fermion number @xmath195 and @xmath196 sectors in the second line .
these two sectors are symmetric under the transformation @xmath310 and @xmath311 , so we analyze only the fermion number @xmath195 sector in the following . from eq .
( [ eq : result - propagator ] ) , we find a pole at @xmath312 the dispersion relation of the real part , @xmath313 , is shown in the left panel of fig .
[ fig : dispersion ] together with the htl results for comparison , where the coupling constant is chosen as @xmath314 .
the imaginary part of the pole reads @xmath315 which is much smaller than those of the normal fermion and the antiplasmino @xcite .
since the real part and the imaginary part of the pole are finite for @xmath316 , this mode is a damped oscillation mode .
the residue of the pole is evaluated to be @xmath317 which means that the mode has only a weak strength in comparison with those of the normal fermion and the antiplasmino , whose residues are order of unity .
it is worth mentioning that such smallness of the residue is actually compatible with the results in the htl approximation : the sum of the residues of the normal fermion and the antiplasmino modes obtained in the htl approximation is unity and thus the sum rule of the spectral function of the fermion is satisfied in the leading order .
therefore , one could have anticipated that the residue of the ultrasoft mode can not be the order of unity but should be of higher order . the pole given by eq .
( [ eq : result - dispersion ] ) gives rise to a new peak in the spectral function of the fermion as @xmath318 which is depicted in the right panel of fig .
[ fig : dispersion ] , where @xmath166 is set to zero . since the expression of @xmath243 for the yukawa model is not available in the literature , we simply adopt @xmath319 in the figure . .
the vertical axis is the energy @xmath320 , while the horizontal axis is the momentum @xmath166 .
the solid ( blue ) lines correspond to the normal fermion and the antiplasmino , while the bold solid ( red ) one to the ultrasoft mode .
note that since we focus on the fermion sector , the antiplasmino appears instead of the plasmino .
the dotted lines denote the light cone . since our analysis on the ultrasoft mode is valid only for @xmath321 , the plot for @xmath322 may not have a physical meaning .
the residue of the antiplasmino becomes exponentially small for @xmath172 , so the plot of the antiplasmino does not represent physical excitation for @xmath172 , either .
right panel : the spectral function in the fermion sector , eq .
( [ eq : spectral ] ) , as a function of energy @xmath320 at zero momentum . ,
title="fig:",scaledwidth=40.0% ] .
the vertical axis is the energy @xmath320 , while the horizontal axis is the momentum @xmath166 .
the solid ( blue ) lines correspond to the normal fermion and the antiplasmino , while the bold solid ( red ) one to the ultrasoft mode .
note that since we focus on the fermion sector , the antiplasmino appears instead of the plasmino .
the dotted lines denote the light cone . since our analysis on the ultrasoft mode is valid only for @xmath321 , the plot for @xmath322 may not have a physical meaning .
the residue of the antiplasmino becomes exponentially small for @xmath172 , so the plot of the antiplasmino does not represent physical excitation for @xmath172 , either .
right panel : the spectral function in the fermion sector , eq .
( [ eq : spectral ] ) , as a function of energy @xmath320 at zero momentum . ,
title="fig:",scaledwidth=45.0% ] so far , we have considered the one - loop diagram .
we need to check that the higher - order loops are suppressed by the coupling constant
. this task would not be straightforward because , @xmath323 appears in the denominator , as seen in eq .
( [ eq : one - loop - pc ] ) , which could make invalid the naive loop expansion .
the possible diagrams contributing in the leading order are ladder diagrams shown in fig .
[ fig : ladder ] because the pair of the fermion and the boson propagators gives a contribution of order @xmath324 , and the vertex gives @xmath325 .
however , there is a special suppression mechanism in the present case with the scalar coupling . for example
, let us evaluate the first diagram in fig .
[ fig : ladder ] , at small @xmath102 .
the self - energy is evaluated to be @xmath326 since there are four vertices and two pairs of the propagators whose momenta are almost the same , the formula would apparently yield the factor , @xmath327 .
one can easily verify that this order estimate would remain the same in any higher - loop diagram , so any ladder diagram seems to contribute in the leading order as explained .
however , this is not the case for yukawa model with the scalar coupling .
an explicit evaluation of the numerators of the fermion propagators gives @xmath328 , which turns out to be of order @xmath329 .
this is because the internal line is almost on - shell , i.e. , @xmath330 , @xmath331 , which comes from the asymptotic masses squared .
an analysis shows that the same suppression occurs in the higher - order diagrams such as the second diagram in fig .
[ fig : ladder ] .
thus , the ladder diagrams giving a vertex correction do not contribute in the leading order in the scalar coupling , and hence , the one - loop diagram in fig .
[ fig : fermion - selfenergy ] with the dressed propagators solely suffices to give the self - energy in the leading - order .
we remark that a similar suppression occurs in the effective three - point - vertex at @xmath148 @xcite .
we also note that this suppression mechanism is quite similar to that found in a supersymmetric model for an intermediate temperature region in the sense that @xmath256 @xcite , whereas we are dealing with extremely high-@xmath0 case .
it should be emphasized that this suppression of the vertex correction is not the case in qed / qcd , where all the ladder diagrams contribute in the leading order and must be summed over @xcite , as will be shown in the following sections .
next we explore whether the ultrasoft fermion mode also exists in qed at high @xmath0 .
one might expect that the analysis would be done in much the same way as in the yukawa model .
it turns out , however , that the analysis is more complicated and involved .
it is necessary to sum up the contributions from all the ladder diagrams even apart from the complicated helicity structure of the photon . in this section
, we successfully perform the summation of the ladder diagrams in an analytic way , and obtain the fermion propagator that is valid in the ultrasoft region .
then we evaluate the pole in the ultrasoft region explicitly and examine the properties of the ultrasoft fermion mode in qed .
we also discuss whether the resummed vertex satisfies the ward - takahashi identity .
[ [ one - loop - calculation ] ] one - loop calculation + + + + + + + + + + + + + + + + + + + + first , we evaluate the contribution from the one - loop diagram .
the dressed propagators with hard momenta read @xmath332 the expressions of @xmath333 and @xmath334 are given in eqs .
( [ eq : mass - electron ] ) and ( [ eq : mass - photon ] ) . the damping rates of electron and photon are estimated as @xmath335 and @xmath336 .
note that @xmath115 is much larger than that in the yukawa model , which is called `` anomalous damping '' .
this large electron damping makes the damping rate of the ultrasoft mode much larger than in the yukawa model .
here we have adopted the coulomb gauge , in which the analysis becomes simple thanks to the transversality of the photon propagator .
we note that the @xmath337 term has been omitted in eqs .
( [ eq : qed - photon - retarded - propagator ] ) and ( [ eq : qed - photon - s - propagator ] ) because that term vanishes after the @xmath338 integral , as in eq .
( [ eq : htl - longitudinal - contribution ] ) . by using these resummed propagators , the one - loop contribution in the ultrasoft region
is evaluated as @xmath339\gamma^\nu\\ & \simeq2\int\frac{d^{4}k}{(2\pi)^{4}}\tilde{{x}}(k)\frac{{\ooalign{\hfil/\hfil\crcr$k$}}}{1 + 2\tilde{p}\cdot k/{\delta m}^2 } , \end{split}\ ] ] where @xmath340 , @xmath341 .
here we have used the same notation for @xmath297 and @xmath243 as those in the yukawa model , although their parametrical expressions are different from each other . the factor two in the last line of eq .
( [ eq : pairpropagator ] ) comes from two degrees of freedom of photon polarization . at the one - loop order
, we obtain @xmath342 we note that this expression has the same structure as that for the yukawa model ; see eq .
( [ eq : sigma - p1 ] ) .
+ [ [ ladder - summation ] ] ladder summation + + + + + + + + + + + + + + + + as already mentioned , the ladder diagrams contribute to the leading - order in contrast to the case of the yukawa model . in this subsection ,
we sum up all the ladder diagrams , and obtain the analytical expressions of the pole position and the residue of the ultrasoft mode . for this purpose
, we introduce the vertex @xmath343 defined through the following self - energy : @xmath344 here the vertex contains the contributions from all the ladder diagrams by imposing the following self - consistent equation for the vertex function : @xmath345 equations ( [ eq : selfenergy ] ) and ( [ eq : vertex ] ) are represented diagrammatically in fig . [ fig : selfconsistentequation ] .
we have used the same approximation as that used in the derivation of eq .
( [ eq : bareselfenergy ] ) for the propagator of fermion and photon .
we should remark here that this summation scheme using the self - consistent equation was first constructed in ref .
however , we also note that the equation has never been solved either analytically nor numerically . in the following ,
we solve this self - consistent equation analytically for small @xmath346 , and show that the dispersion relation does not change from that in the one - loop order even after incorporating all the ladder diagrams . at small @xmath346 , eq . ( [ eq : vertex ] )
reduces to @xmath347 where we have dropped the term which is proportional to @xmath348 because it only gives higher order contribution after being multiplied by @xmath348 in the numerator of eq .
( [ eq : selfenergy ] ) , because @xmath349 .
let us solve the self - consistent equation ( [ eq : vertex2 ] ) .
we expand the vertex function as @xmath350 where @xmath351 is of order unity and @xmath352 is of order @xmath353 .
we first evaluate @xmath351 , which can be decomposed as follows : @xmath354 where @xmath355 , @xmath356 , and @xmath357 are @xmath358 matrices .
then the self - consistent equation for @xmath359 becomes @xmath360 where @xmath361 and @xmath362 in the right hand side vanish due to transversality of the photon propagator : @xmath363 . by assuming that @xmath364 is a constant ,
the integral becomes @xmath365 by imposing @xmath96 as before , this expression becomes @xmath366 where we have used eqs .
( [ eq : integral - formula - f1 ] ) and ( [ eq : integral - formula - b1 ] ) , and dropped @xmath348 term in the last line .
then , from eq .
( [ eq : ultrasoft - qed - bs-0 ] ) , we find @xmath367 where @xmath368 is the unit @xmath358 matrix .
next we evaluate @xmath369 . from eq .
( [ eq : vertex2 ] ) , expanding @xmath370 in terms of @xmath371 , we find the self - consistent equation for @xmath369 as @xmath372 .
\end{split } \label{eq : deltavargamma}\ ] ] it is not easy to analytically solve eq .
( [ eq : deltavargamma ] ) directly .
so we instead calculate the following function : @xmath373 then eq . ( [ eq : deltavargamma ] ) leads to the following closed equation , @xmath374 \tilde{x}(k ' ) p^t_{\rho\lambda}(k')\gamma^\rho{\ooalign{\hfil/\hfil\crcr$k$}}'\delta\vargamma^\lambda(p , k ' ) \\ = & -\lambda a\varsigma_\text{one - loop}(p ) -\lambda\delta\varpi(p ) .
\end{split}\ ] ] here we have used eqs .
( [ eq : integral ] ) and ( [ eq : deltapi ] ) in the second line .
the solution to this equation is readily found to be @xmath375 , where eq .
( [ eq : vargammaabc ] ) is used for @xmath355 .
then , the self - energy is evaluated as @xmath376 where the residue is @xmath377 the pole of the ultrasoft fermion mode has the velocity @xmath307 , damping rate @xmath378 , and the residue @xmath379 .
the dispersion of the mode is the same as that in the one - loop level whereas the residue is changed .
this is our main result for qed . in this subsection
, we examine whether the summation scheme , eqs .
( [ eq : selfenergy ] ) and ( [ eq : vertex ] ) , is consistent with the @xmath380 gauge symmetry .
concretely , we check that our resummed vertex function and self - energy satisfy the ward - takahashi ( wt ) identity in the leading order of the coupling constant .
the wt identity reads @xmath381 since @xmath382 contains two separated scales @xmath85 and @xmath237 , we need to treat them carefully . for the hard part ,
@xmath383 is of order @xmath49 , which is negligible compared to @xmath348 . in addition , the momentum dependent part @xmath384 is also negligible compared to @xmath298 .
therefore the wt identity reduces to @xmath385 on the other hand , multiplying eq .
( [ eq : vertex ] ) by @xmath386 , we have @xmath387 where we have dropped the terms of order @xmath388 , and used eq .
( [ eq : selfenergy ] ) in the last line .
this expression coincides with eq .
( [ eq : mod - wt ] ) , so our self - consistent equation satisfies the wt identity .
we note that this proof was made without using the expansion in terms of @xmath389 . for this reason , eq .
( [ eq : mod - wt ] ) is generally valid for @xmath237 , as well as for @xmath390 .
next , we check whether the explicit solution eq .
( [ eq : vargammaabc ] ) of the self - consistent equation ( [ eq : vertex ] ) at zeroth order in @xmath353 satisfies the wt identity to see the consistency with the gauge symmetry of the following two conditions adopted to obtain eq .
( [ eq : vargammaabc ] ) : one is that terms proportional to @xmath348 in @xmath391 is dropped , since they are negligible in the self - energy due to @xmath392 .
the other is that we imposed the on - shell condition @xmath96 , because the internal photon in the self - energy is almost on - shell .
using the same conditions , we expect that the vertex function satisfies the wt identity in the leading order of the coupling constant .
in fact , we have for the zeroth order in @xmath346 @xmath393 here we have dropped @xmath348 and @xmath330 in the last equality . finally , we check that the equation determining the vertex function eq .
( [ eq : deltavargamma ] ) , which is first order in @xmath346 , satisfies the wt identity . by multiplying this equation by @xmath386
, we obtain @xmath394\\ & = \varsigma^r(p ) , \end{split}\ ] ] where we have dropped @xmath348 as in the previous equation and used eq .
( [ eq : selfenergy-1st ] ) .
therefore , our analytic solution of the self - consistent equation satisfies the wt identity in the leading order of the coupling constant .
we note that without the summation of the ladder diagrams , the wt identity is not satisfied when the external momentum of fermion is ultrasoft .
by contrast , the ladder summation was unnecessary in the yukawa model , in which the gauge symmetry is absent . in this section ,
we show the existence of the ultrasoft fermion mode , and obtain the expressions of the dispersion relation , the damping rate , and the strength in qcd , in a similar way to that in qed .
the differences between qed and qcd are as follows : first , the damping rate of the hard gluon is much larger than that of hard photon .
this is because the gluon can collides via the soft gluons @xcite owing to the self - coupling while the photon can not .
second , new kind of the ladder diagrams shown in fig .
[ fig : ladder - qcd ] contribute in qcd in addition to the ladder diagrams drawn in fig .
[ fig : ladder ] , again due to the self - coupling of the gluons . despite these differences
, we will see that the properties of the ultrasoft fermion mode are the same qualitatively in both gauge theories . to sum up the ladder diagrams
, we use the following self - consistent equation for the quark - gluon vertex function @xmath395 , which can be derived as in the qed case : @xmath396 here we have used the coulomb gauge as in the analysis in qed , and introduced @xmath397 $ ] and @xmath398 . the expression for @xmath399 and @xmath400
are given by eqs .
( [ eq : mass - quark ] ) and ( [ eq : mass - gluon ] ) . @xmath116 and @xmath117 are of order @xmath240 , which is much larger than @xmath401 due to the self - interaction of the gluon .
we note that the presence of the third term in the right - hand side , which is absent in the case of qed , is caused by the self - coupling of the gluons .
the quark self - energy is written in terms of @xmath395 as @xmath402 the diagrams for these two equations are drawn in fig .
[ fig : bs - qcd ] . ) and ( [ eq : selfenergy - qcd ] ) .
the notation is the same as in fig .
[ fig : ladder - qcd ] .
, scaledwidth=90.0% ] we note that this self - consistent equation is the same as that in qed except for @xmath415 in the right - hand side : there are no quantitatively novel effects which result from the fact that the gauge group is non - abelian .
we also note that the same equation was obtained in the temporal gauge @xcite , which is nontrivial if we consider the following facts : in qed , the photon propagator in the resummed perturbation scheme is always on - shell , so only the transverse component is considered . in that case , apparently there is no difference between the coulomb gauge and the temporal gauge in this scheme .
however , in qcd , the off - shell gluon appears due to the self - interaction of the gluon .
then , the components other than the transverse one appear in the calculation , and the propagator of those components are different in both gauges .
we can solve the self - consistent equation in the same way as in qed .
the solution is as follows : we expand the vertex function in terms of @xmath416 as in eq .
( [ eq : ultrasoft - qed - vertex - expand ] ) .
the zeroth order solution is @xmath417 here @xmath355 , @xmath361 , and @xmath362 are defined by eq .
( [ eq : vargamma ] ) , and @xmath301 is defined by eq .
( [ eq : ultrasoft - lambda ] ) . the self - consistent equation which determines the vertex function at the first order @xmath352 is written as @xmath418 then @xmath419 defined by eq .
( [ eq : deltapi ] ) becomes @xmath420 owing to eq .
( [ eq : selfenergy - qcd ] ) , the retarded quark self - energy is found to be @xmath421 where the expression of the residue @xmath379 will be given shortly .
we note that this expression is the same as that in qed except for the numerical factor .
thus , the expression for the pole position of the ultrasoft mode in qcd is the same as in qed , while the residue of that mode is not : the residue is @xmath422 we can also show that the analytic solution of the self - consistent equation satisfies the wt identity in the leading order as in qed , by checking that the counterparts of eqs .
( [ eq : ultrasoft - wt - zeroth ] ) and ( [ eq : ultrasoft - wt - first ] ) are satisfied . in this chapter
, we developed the resummed perturbation theory which was originally constructed in ref .
@xcite and enables us to successfully regularize the infrared singularity . by using this method , we analyzed the quark spectrum whose energy is ultrasoft in qgp .
since the yukawa model and qed are simpler than qcd but have some similarity to qcd , we also worked in these models . in qed
/ qcd , the summation of the ladder diagrams had to be done whereas that procedure was unnecessary in the yukawa model .
that summation was necessary also from the point of view of the gauge symmetry .
as a result , we established the existence of a novel fermionic mode in the ultrasoft energy region , and obtained the expressions of the pole position and the strength of that mode .
the expressions for the dispersion relation , the damping rate , and the residue of the ultrasoft mode in the yukawa model , qed , and qcd are summarized in table [ tab : ultrasoft - expression ] .
we also showed that the ward - takahashi identity is satisfied in the resummed perturbation theory in qed / qcd .
.the expressions of the dispersion relation , the damping rate , and the residue of the ultrasoft mode in the yukawa model , qed , and qcd . @xmath423 and @xmath424 are set to three . [ cols="<,^,^,^",options="header " , ] [ tab : correspondence ] here , let us show the equivalence between eq .
( [ eq : qed - coulomb - result - k ] ) and the self - consistent equation in the resummed perturbation theory @xcite in chapter [ chap : ultrasoft ] . to this end
, we rewrite eq .
( [ eq : qed - vertex - correction ] ) in terms of the off - diagonal self - energy using eq .
( [ eq : qed - coulomb - k - vertex ] ) : @xmath425}{2ik'\cdot \partial_x + 2i{\zeta_e}k'^{0}+\delta m^2 } { \tilde{\vargamma}}^\alpha(k',x ) . \end{split}\end{aligned}\ ] ] here we set @xmath426 . by performing the fourier transformation , eq .
( [ eq : fourier - p ] ) , we get @xmath427\vargamma^\alpha(k',p)}{2k'\cdot p + 2i{\zeta_e}k'^{0}+\delta m^2 } , \end{split}\end{aligned}\ ] ] where @xmath428 .
we note that @xmath429 is the vertex function @xcite introduced in chapter [ chap : ultrasoft ] whose momenta are hard and ultrasoft .
equation ( [ eq : qed - bs ] ) is none other than eq .
( [ eq : vertex ] ) , which is the self - consistent equation in the resummed perturbation @xcite .
the retarded fermion self - energy is also written in terms of the vertex function : from eq .
( [ eq : qed - coulomb - k - vertex ] ) , we arrive at @xmath430}{2k\cdot p + 2i{\zeta_e}k^0 + \delta m^2 } \vargamma^\nu(k , p ) .
\end{split}\end{aligned}\ ] ] this expression equals to eq .
( [ eq : selfenergy ] ) , which is the expression of the fermion retarded self - energy in the resummed perturbation theory ) is analytically solved for the energy region @xmath431 in the previous chapter @xcite . ]
thus we see that eq .
( [ eq : qed - coulomb - result - k ] ) , derived in the non - equilibrium state in a linear response regime from the kadanoff - baym equation , is equivalent to the self - consistent equation in the resummed perturbation theory , which is constructed in the thermal equilibrium state . here
we discuss the correspondence between each prescription in the resummed perturbation theory @xcite and each term in the kinetic equation . as in the yukawa model
, the resummation of the thermal mass difference in the resummed perturbation theory corresponds to the mass difference term in the kinetic equation .
the damping rate corresponds to the collision term since both of them contain the damping rate of the hard electron , @xmath115 .
the ladder summation in the resummed perturbation theory @xcite is caused by the correction to the force term in the kinetic equation .
thus , the ladder summation corresponds to the correction to the force term .
these interpretations are summarized in the table .
[ tab : correspondence ] .
the fermionic induced source @xmath433 generates the @xmath4-point functions with @xmath434 , not only the fermion self - energy . in this subsection
, we derive the self - consistent equation determining the @xmath4-point function whose external lines consist of two fermions ( @xmath435 ) and @xmath5 bosons ( @xmath436 ) with ultrasoft external momenta , and make an order estimate of the quantity .
for example in the case of @xmath437 , we obtain the correction to the bare three - point function , @xmath438 , from @xmath433 : @xmath439 here we have expanded @xmath440 around @xmath426 : @xmath441 where @xmath442 contains one @xmath435 while @xmath443 contains one @xmath435 and one @xmath436 .
@xmath443 can be obtained by the following way .
collecting terms that contain one @xmath436 in eq .
( [ eq : qed - coulomb - result - k ] ) , we obtain @xmath444 \int\frac{d^4 k'}{(2\pi)^4}\frac{k^\alpha \gamma_\nu+\gamma^\alpha k'_\nu}{k\cdot k ' } \delta k_\alpha(k',x ) .
\label{eq : deltaki } \end{split}\end{aligned}\ ] ] since @xmath442 is determined by setting @xmath426 in eq .
( [ eq : qed - coulomb - result - k ] ) , this equation is closed and @xmath443 can be determined .
let us estimate the order of @xmath443 . from eq .
( [ eq : qed - coulomb - result - k ] ) and @xmath445 , we find @xmath446 therefore , the vertex correction is estimated as @xmath447 which is much larger than the bare vertex , @xmath448 .
similar order estimate for the @xmath4-point function with @xmath449 can be done with the same procedure ; as a result , we find that the order of the @xmath4-point - vertex function is @xmath450 .
we show that the off - diagonal self - energy given in eq .
( [ eq : qed - vertex - correction ] ) satisfies the ward - takahashi ( wt ) identity . from eq .
( [ eq : qed - vertex - correction ] ) , we get @xmath451 here we have used @xmath452 , which can be confirmed by multiplying eq .
( [ eq : qed - coulomb - result - k ] ) by @xmath348 from the left .
this equation generates the wt identity ( eq .
( [ eq : mod - wt ] ) ) derived in chapter [ chap : ultrasoft ] by setting @xmath426 and differentiating with respect to @xmath435 .
the wt identity implies that the vertex correction is not negligible because the identity relates the vertex correction to the fermion self - energy , which is much larger than the inverse of the free fermion propagator with an ultrasoft momentum . in the yukawa model ,
the wt identity associated with gauge symmetries is absent from the outset , so the smallness of the vertex correction is not in contradiction with any identity .
equation ( [ eq : qed - wt ] ) can be derived from the conservation law of the induced current , @xmath453 by differentiating eq .
( [ eq : conservationlaw ] ) with respect to @xmath454 , we obtain @xmath455 here @xmath456 is the delta function defined along the contour @xmath357 . by multiplying this equation by @xmath457 , we get @xmath458 here we have set @xmath459 , @xmath460 and neglected the sub - leading terms .
the first term in the left - hand side has the same order of magnitude as the hard fermion self - energy @xmath461 times @xmath462 , so that term is negligible .
thus the left - hand side becomes @xmath463 .
the right - hand side becomes @xmath464 we see that the first term in the last line is negligible because of the on - shell condition .
thus we obtain @xmath465 if we remember that @xmath466 .
this equation is nothing but eq .
( [ eq : qed - wt ] ) . in this chapter , we derived the linearized off - diagonal boltzmann equation for the ultrasoft fermion average field in the kadanoff - baym formalism , and showed that the resultant equation is equivalent to the self - consistent equation in the resummed perturbation theory at the leading order of the perturbative expansion , while it has been known that the vlasov equation is equivalent to the basic equation of the htl approximation .
this equivalence shows that the resummation scheme of that perturbation theory contains the interaction effect among the particles beyond the mean field approximation .
we note that the off - diagonal boltzmann equation contains not only the collision term but also the thermal mass terms and the off - diagonal self - energy term , in contrast to the usual boltzmann equation .
we also derived the equation which determines the @xmath4-point functions with external lines for a pair of fermions and @xmath5 bosons with ultrasoft momenta , by considering the non - linear response regime using the gauge symmetry .
we also derived the ward - takahashi identity from the conservation law of the electromagnetic current in the kadanoff - baym formalism .
in this thesis , we developed the resummed perturbation theory which takes into account the separation of the scale and enables us to successfully regularize the infrared singularity . by using that method
, we analyzed the spectral properties of the ultrasoft fermion , including its existence itself , in the yukawa model and qed / qcd .
the procedure of that method consists of the resummation of the asymptotic thermal masses . in qed / qcd , the summation of the ladder diagrams is also necessary . as a result of the analysis , we established the existence of the novel fermionic mode in that energy region , and obtain the expressions of the pole position and the residue of that mode , which are summarized in table .
[ tab : ultrasoft - expression ] .
we also showed that the resultant fermion propagator and the vertex function satisfy the ward - takahashi identity in qed / qcd .
we also derived the linearized and generalized boltzmann equation for ultrasoft fermion excitations , which describe the time - evolution of the amplitude of the process of changing the fermion to the boson , from the kadanoff - baym equation in a yukawa model and qed .
we showed that this equation is equivalent to the self - consistent equation in the resummed perturbation theory used in the analysis of the ultrasoft fermion spectrum at the leading order .
it helps us to establish the foundation of the resummed perturbation theory .
we derived the ward - takahashi identity from the conservation law of the electromagnetic current in qed .
furthermore , we derived the equation that determines the @xmath4-point function with external lines for a pair of fermions and @xmath5 bosons with ultrasoft momenta in qed , by using @xmath380 gauge symmetry .
we analyzed the finite temperature and zero chemical potential ( @xmath36 ) case in this thesis , so it is quite natural to ask how our result on the ultrasoft fermion mode is modified in other region .
such analysis is already done by the author and his collaborator @xcite , and we hope to report its results elsewhere .
we just describe the analysis we performed and its results in the following : we constructed the resummed perturbation theory in high-@xmath36 and @xmath183 case , and analyzed the fermion spectrum with the ultrasoft momentum ( @xmath467 ) in the yukawa model . as a result , we found that the ultrasoft fermion mode does not exist in this case .
this result is consistent with the absence of the charge symmetry in high-@xmath36 case , because the charge symmetry is essentially important for generating the ultrasoft fermion mode , as discussed in sec .
[ ssc : intro - ultrasoft ] .
we also investigate how large @xmath36 kills the ultrasoft fermion mode , and found that the maximum chemical potential in which the mode persists is of order @xmath0 . the resummed perturbation scheme used in this thesis
can be applied to other quantities in the low energy region , in general multi - component system .
we briefly discuss some examples in the following : * as we wrote in chapter [ chap : intro ] , the computation of the transport coefficient also needs taking into account the interaction effect among the hard particles .
usually , only the collision effect need to be resummed .
however , our analysis in this thesis suggests that the difference of the dispersion relations of the particles also needs to be taken into account , when two kinds of particle appear in the loop integral .
it is the case in the computation of the flavor diffusion constant @xcite : the quarks with different flavor appear in the integral .
since the current quark masses are different in each flavor , it is quite interesting to investigate the effect of the difference of the current quark mass by performing the analysis which takes into account the current quark mass . *
the perturbative analysis of the neutrino spectrum around the critical temperature of the electroweak transition , which is expected to be realized at the early universe , needs the resummation used in this thesis since the bare masses of the weak bosons becomes negligible compared with the temperature .
it is an interesting task to perform such analysis , and investigate the effect of the fermion spectrum in the ultrasoft region , on the properties of the early universe . *
another example is the supersymmetric system in both of the relativistic and nonrelativistic system .
in such system , the susy is spontaneously broken by the finite temperature / density effect @xcite , and as a result , the fermionic zero mode , goldstino , appears . as is described in sec .
[ ssc : intro - ultrasoft ] , the analysis of the goldstino , which uses the resummation scheme , was performed in relativistic system .
+ though the susy has not been observed as a fundamental symmetry in the real world , it was suggested to simulate the supersymmetric system using the cold atom system , which is nonrelativistic .
such system also needs the resummation scheme @xcite , and it may be surprising since the power counting rule in the nonrelativistic system is different from that in the relativistic system .
we note that , however , there is an important difference between the relativistic system and the nonrelativistic one : since the energy - momentum tensor and the supercurrent are not in the same supermultiplet in the nonrelativistic system , the goldstino can not be interpreted as the phonino .
the dynamical analysis on the goldstino in the nonrelativistic system was performed @xcite , but without the bose - einstein condensation ( bec ) .
the analysis in the system with bec is quite attractive task .
the author is very grateful to his advisor , teiji kunihiro , for the collaboration to this work and the supervision .
he supported and encouraged the author s research .
the author thanks yoshimasa hidaka for the collaboration to this work , and his kind encouragement to the author .
the author acknowledges jean - paul blaizot , who gave the author the hospitality during the author s staying in his group , and contributed to the author s research as a collaborator .
the author also thanks the members of nuclear theory groups in kyoto university and in yukawa institute for theoretical physics .
the author is supported by japan society for the promotion of science and a grant - in - aid for scientific research [ ( c ) no .
24@xmath46856384 ] .
[ app : temporal ] in this thesis , the calculation was performed in the coulomb gauge . however , since the resummed perturbation theory was first proposed in the temporal gauge @xcite , we need to check that the results in this thesis also can be obtained in the temporal gauge . to this end , we show that the equation determining @xmath469 in the temporal gauge is the same as that in the coulomb gauge , eq .
( [ eq : qed - coulomb - result - k ] ) , in qed in this appendix . showing that equation can be obtained also in the temporal gauge suffices to show that the expression of the properties of the ultrasoft fermion mode obtained in the coulomb gauge do not change in the temporal gauge , because that equation is equivalent to the self - consistent equation in the resummed perturbation theory , which was used to analyze the ultrasoft fermion mode .
the gauge - fixing condition in the temporal gauge is @xmath470=a^0 $ ] and @xmath471 .
this condition is equivalent to the constraint @xmath472 .
because of this constraint , we have @xmath473 the equations governing @xmath474 are @xmath475 from these equations , we obtain @xmath476 here let us evaluate @xmath477 , which is the longitudinal component of @xmath478 . by multiplying eq .
( [ eq : qed - temporal - k ] ) by @xmath479 , we get @xmath480 here we have neglected the terms that are of order @xmath481 .
we see that @xmath482 and thus the longitudinal component of @xmath483 , @xmath484 , is negligible compared with the transverse component of @xmath483 .
we note that @xmath485 , which is the result of the gauge - fixing condition , and @xmath486 are valid also in the coulomb gauge in the leading order .
furthermore , also the free photon propagator at equilibrium is the same as that in the coulomb gauge .
thus , we can obtain eq .
( [ eq : qed - coulomb - result - k ] ) in the same way as in the sec .
[ ssc : kinetic - qed - derivation ] .
[ app : formula ] j. beringer _ et al
. _ [ particle data group collaboration ] , phys . rev .
d * 86 * , 010001 ( 2012 ) .
j. b. kogut and d. k. sinclair , phys .
d * 70 * , 094501 ( 2004 ) [ hep - lat/0407027 ] .
d. k. sinclair and j. b. kogut , pos lat * 2006 * , 147 ( 2006 ) [ hep - lat/0609041 ] .
de forcrand , m. a. stephanov and u. wenger , pos lat * 2007 * , 237 ( 2007 ) [ arxiv:0711.0023 [ hep - lat ] ] . w. detmold , k. orginos and z. shi , phys .
d * 86 * , 054507 ( 2012 ) [ arxiv:1205.4224 [ hep - lat ] ] . b. c. barrois , nucl .
b * 129 * , 390 ( 1977 ) .
d. bailin and a. love , phys .
rept . *
107 * , 325 ( 1984 ) .
m. iwasaki and t. iwado , phys .
b * 350 * , 163 ( 1995 ) .
m. g. alford , k. rajagopal and f. wilczek , phys .
b * 422 * , 247 ( 1998 ) [ hep - ph/9711395 ] .
r. rapp , t. schafer , e. v. shuryak and m. velkovsky , phys . rev .
lett . * 81 * , 53 ( 1998 ) [ hep - ph/9711396 ] .
m. g. alford , a. schmitt , k. rajagopal and t. schafer , rev .
phys . * 80 * , 1455 ( 2008 ) [ arxiv:0709.4635 [ hep - ph ] ] .
j. c. collins and m. j. perry , phys .
* 34 * ( 1975 ) 1353 .
k. yagi , t. hatsuda and y. miake , camb .
monogr . part .
cosmol .
* 23 * , 1 ( 2005 ) .
a. bazavov , t. bhattacharya , m. cheng , n. h. christ , c. detar , s. ejiri , s. gottlieb and r. gupta _ et al .
_ , phys .
d * 80 * , 014504 ( 2009 ) [ arxiv:0903.4379 [ hep - lat ] ] . c. w. bernard _ et al .
_ [ milc collaboration ] , phys . rev .
d * 55 * , 6861 ( 1997 ) [ hep - lat/9612025 ]
. m. cheng , n. h. christ , s. datta , j. van der heide , c. jung , f. karsch , o. kaczmarek and e. laermann _ et al .
_ , phys .
d * 77 * , 014511 ( 2008 ) [ arxiv:0710.0354 [ hep - lat ] ] .
j. b. kogut , m. stone , h. w. wyld , w. r. gibbs , j. shigemitsu , s. h. shenker and d. k. sinclair , phys .
lett . * 50 * , 393 ( 1983 ) .
z. fodor and s. d. katz , jhep * 0404 * , 050 ( 2004 ) [ hep - lat/0402006 ] . c. r. allton , s. ejiri , s. j. hands , o. kaczmarek , f. karsch , e. laermann , c. schmidt and l. scorzato , phys .
d * 66 * , 074507 ( 2002 ) [ hep - lat/0204010 ] .
a. roberge and n. weiss , nucl .
b * 275 * , 734 ( 1986 ) .
y. sakai , k. kashiwa , h. kouno and m. yahiro , phys .
d * 78 * , 036001 ( 2008 ) [ arxiv:0803.1902 [ hep - ph ] ] .
y. nambu and g. jona - lasinio , phys .
* 122 * , 345 ( 1961 ) .
y. nambu and g. jona - lasinio , phys .
rev . * 124 * , 246 ( 1961 ) .
t. hatsuda and t. kunihiro , phys .
* 247 * , 221 ( 1994 ) [ hep - ph/9401310 ] .
k. fukushima , phys .
b * 591 * , 277 ( 2004 ) [ hep - ph/0310121 ] .
k. fukushima , phys .
d * 77 * , 114028 ( 2008 ) [ erratum - ibid .
d * 78 * , 039902 ( 2008 ) ] [ arxiv:0803.3318 [ hep - ph ] ] . c. ratti , m. a. thaler and w. weise , phys .
d * 73 * , 014019 ( 2006 ) [ hep - ph/0506234 ] .
e. v. shuryak and j. j. m. verbaarschot , nucl .
a * 560 * , 306 ( 1993 ) [ hep - th/9212088 ] .
g. munster , nucl .
b * 180 * , 23 ( 1981 ) .
p. de forcrand and m. fromm , phys .
* 104 * , 112005 ( 2010 ) [ arxiv:0907.1915 [ hep - lat ] ] .
t. z. nakano , k. miura and a. ohnishi , prog .
phys . *
123 * , 825 ( 2010 ) [ arxiv:0911.3453 [ hep - lat ] ] .
m. asakawa and k. yazaki , nucl .
phys . a * 504 * , 668 ( 1989 ) .
a. barducci , r. casalbuoni , s. de curtis , r. gatto and g. pettini , phys .
b * 231 * , 463 ( 1989 ) .
f. wilczek , int .
j. mod .
phys . a * 7 * , 3911 ( 1992 ) [ erratum - ibid .
a * 7 * , 6951 ( 1992 ) ] .
j. berges and k. rajagopal , nucl .
b * 538 * , 215 ( 1999 ) [ hep - ph/9804233 ] .
m. gyulassy and l. mclerran , nucl .
a * 750 * , 30 ( 2005 ) [ nucl - th/0405013 ] .
b. muller and j. l. nagle , ann .
nucl . part .
sci . * 56 * , 93 ( 2006 ) [ nucl - th/0602029 ] .
m. strickland , j. o. andersen , l. e. leganger and n. su , prog .
theor . phys .
suppl . * 187 * , 106 ( 2011 ) [ arxiv:1011.0416 [ hep - ph ] ] .
j. o. andersen , e. braaten and m. strickland , phys .
d * 61 * , 014017 ( 2000 ) [ hep - ph/9905337 ]
. j. o. andersen , e. braaten and m. strickland , phys .
lett . * 83 * , 2139 ( 1999 ) [ hep - ph/9902327 ] .
j. p. blaizot , e. iancu and a. rebhan , phys .
lett . *
83 * , 2906 ( 1999 ) [ hep - ph/9906340 ] .
j. p. blaizot , e. iancu and a. rebhan , phys .
b * 470 * , 181 ( 1999 ) [ hep - ph/9910309 ] .
j. p. blaizot , e. iancu and a. rebhan , phys .
d * 63 * , 065003 ( 2001 ) [ hep - ph/0005003 ] .
j. p. blaizot , e. iancu and a. rebhan , phys .
d * 68 * , 025011 ( 2003 ) [ hep - ph/0303045 ] .
j. -p .
blaizot and e. iancu , phys .
* 359 * , 355 ( 2002 ) [ hep - ph/0101103 ] .
m. le bellac , `` thermal field theory , '' cambridge , uk : univ .
( 1996 ) .
m. h. thoma , z. phys .
c * 66 * , 491 ( 1995 ) [ hep - ph/9406242 ] .
f. flechsig and a. k. rebhan , nucl .
b * 464 * , 279 ( 1996 ) [ hep - ph/9509313 ] .
j. s. schwinger , j. math .
* 2 * , 407 ( 1961 ) .
l. v. keldysh , zh . eksp .
fiz . * 47 * , 1515 ( 1964 ) [ sov .
jetp * 20 * , 1018 ( 1965 ) ] .
aurenche , f. gelis and h. zaraket , phys .
d * 62 * , 096012 ( 2000 ) [ hep - ph/0003326 ] .
p. aurenche , f. gelis and h. zaraket , phys .
d * 61 * , 116001 ( 2000 ) [ hep - ph/9911367 ] .
r. baier , h. nakkagawa , a. niegawa and k. redlich , z. phys .
c * 53 * , 433 ( 1992 ) .
j. i. kapusta , p. lichard and d. seibert , phys . rev .
d * 44 * , 2774 ( 1991 ) [ erratum - ibid .
d * 47 * , 4171 ( 1993 ) ] .
p. b. arnold , g. d. moore and l. g. yaffe , jhep * 0111 * , 057 ( 2001 ) [ hep - ph/0109064 ] .
b. arnold , g. d. moore and l. g. yaffe , jhep * 0112 * , 009 ( 2001 ) [ hep - ph/0111107 ] .
b. arnold , g. d. moore and l. g. yaffe , jhep * 0206 * , 030 ( 2002 ) [ hep - ph/0204343 ] .
j. -p .
blaizot and e. iancu , phys .
d * 55 * , 973 ( 1997 ) [ hep - ph/9607303 ] .
j. -p .
blaizot and e. iancu , phys .
* 76 * , 3080 ( 1996 ) [ hep - ph/9601205 ] .
v. v. lebedev and a. v. smilga , physica a * 181 * , 187 ( 1992 ) . v. v. lebedev and a. v. smilga , phys . lett .
b * 253 * , 231 ( 1991 ) .
r. d. pisarski , phys .
d * 47 * , 5589 ( 1993 ) .
s. jeon , phys .
d * 52 * , 3591 ( 1995 ) [ hep - ph/9409250 ] .
e. wang and u. w. heinz , phys . lett .
b * 471 * , 208 ( 1999 ) [ hep - ph/9910367 ] .
m. e. carrington , d. -f .
hou and r. kobes , phys .
d * 62 * , 025010 ( 2000 ) [ hep - ph/9910344 ] . m. a. valle basagoiti , phys
. rev .
d * 66 * , 045005 ( 2002 ) [ hep - ph/0204334 ] .
e. wang and u. w. heinz , phys . rev .
d * 67 * , 025022 ( 2003 ) [ hep - th/0201116 ] .
j. -s . gagnon and s. jeon ,
d * 75 * , 025014 ( 2007 ) [ erratum - ibid .
d * 76 * , 089902 ( 2007 ) ] [ hep - ph/0610235 ] .
j. -s . gagnon and s. jeon ,
phys . rev .
d * 76 * , 105019 ( 2007 ) [ arxiv:0708.1631 [ hep - ph ] ] .
y. hidaka and t. kunihiro , phys . rev .
d * 83 * , 076004 ( 2011 ) [ arxiv:1009.5154 [ hep - ph ] ] .
e. braaten and r. d. pisarski , nucl .
b * 339 * , 310 ( 1990 ) . v. v. klimov , sov .
j. nucl .
phys . * 33 * , 934 ( 1981 ) [ yad .
* 33 * , 1734 ( 1981 ) ]
. h. a. weldon , phys .
d * 26 * , 2789 ( 1982 ) .
h. a. weldon , phys .
d * 40 * , 2410 ( 1989 ) .
h. a. weldon , phys .
d * 26 * , 1394 ( 1982 ) .
j. c. taylor and s. m. h. wong , nucl .
b * 346 * , 115 ( 1990 ) .
e. braaten and r. d. pisarski , phys .
d * 45 * , 1827 ( 1992 ) .
j. frenkel and j. c. taylor , nucl .
b * 374 * , 156 ( 1992 ) .
t. altherr and u. kraemmer , astropart .
* 1 * , 133 ( 1992 ) . h. vija and m. h. thoma , phys .
b * 342 * , 212 ( 1995 ) [ hep - ph/9409246 ] . c. manuel , phys .
d * 53 * , 5866 ( 1996 ) [ hep - ph/9512365 ] .
j. -p .
blaizot and j. -y .
ollitrault , phys .
d * 48 * , 1390 ( 1993 ) [ hep - th/9303070 ] .
m. kitazawa , t. kunihiro and y. nemoto , prog .
phys . * 117 * , 103 ( 2007 ) [ hep - ph/0609164 ] .
o. k. kalashnikov and v. v. klimov , sov .
j. nucl .
phys . * 31 * , 699 ( 1980 ) [ yad .
fiz . * 31 * , 1357 ( 1980 ) ] . t. h. hansson and i. zahed , phys . rev .
lett . * 58 * , 2397 ( 1987 ) .
t. h. hansson and i. zahed , nucl .
b * 292 * , 725 ( 1987 ) .
r. kobes and g. kunstatter , phys .
lett . * 61 * , 392 ( 1988 ) .
h. t. elze , u. w. heinz , k. kajantie and t. toimela , z. phys .
c * 37 * , 305 ( 1988 ) .
r. kobes , g. kunstatter and a. rebhan , phys .
lett . * 64 * , 2992 ( 1990 ) .
r. kobes , g. kunstatter and a. rebhan , nucl .
b * 355 * , 1 ( 1991 ) .
r. d. pisarski , phys .
lett . * 63 * , 1129 ( 1989 ) .
e. braaten and r. d. pisarski , phys .
lett . * 64 * , 1338 ( 1990 ) .
e. braaten and r. d. pisarski , nucl .
b * 337 * , 569 ( 1990 ) .
e. braaten and r. d. pisarski , phys .
d * 42 * , 2156 ( 1990 ) .
e. braaten and r. d. pisarski , phys .
d * 46 * , 1829 ( 1992 ) .
r. kobes , g. kunstatter and k. mak , phys .
d * 45 * , 4632 ( 1992 ) .
j. p. blaizot and e. iancu , nucl .
b * 417 * , 608 ( 1994 ) [ hep - ph/9306294 ] .
j. p. blaizot and e. iancu , phys .
lett . * 70 * , 3376 ( 1993 ) [ hep - ph/9301236 ] .
j. -p .
blaizot and e. iancu , nucl .
b * 390 * , 589 ( 1993 ) .
d. bodeker , nucl .
b * 559 * , 502 ( 1999 ) [ hep - ph/9905239 ] .
d. bodeker , phys .
b * 516 * , 175 ( 2001 ) [ hep - ph/0012304 ] .
d. bodeker , phys .
b * 426 * , 351 ( 1998 ) [ hep - ph/9801430 ] .
d. buchholz and i. ojima , nucl .
b * 498 * , 228 ( 1997 ) [ hep - th/9701005 ] .
k. kratzert , annals phys .
* 308 * , 285 ( 2003 ) [ hep - th/0303260 ] .
d. boyanovsky , phys .
d * 29 * , 743 ( 1984 ) .
h. aoyama and d. boyanovsky , phys .
d * 30 * , 1356 ( 1984 ) .
r. gudmundsdottir and p. salomonson , nucl .
b * 285 * , 1 ( 1987 ) .
r. gudmundsdottir and p. salomonson , phys .
b * 182 * , 174 ( 1986 ) .
v. v. lebedev and a. v. smilga , nucl .
b * 318 * , 669 ( 1989 ) .
m. harada and y. nemoto , phys .
d * 78 * , 014004 ( 2008 ) [ arxiv:0803.3257 [ hep - ph ] ] .
m. harada , y. nemoto and s. yoshimoto , prog .
phys . * 119 * , 117 ( 2008 ) arxiv:0708.3351 [ hep - ph ] .
s. -x .
qin , l. chang , y. -x .
liu and c. d. roberts , phys .
d * 84 * , 014017 ( 2011 ) [ arxiv:1010.4231 [ nucl - th ] ] .
h. nakkagawa , h. yokota and k. yoshida , arxiv:1208.6386 [ hep - ph ] .
h. nakkagawa , h. yokota and k. yoshida , phys .
d * 85 * , 031902 ( 2012 ) [ arxiv:1111.0117 [ hep - ph ] ] .
m. kitazawa , t. kunihiro and y. nemoto , phys .
b * 633 * , 269 ( 2006 ) [ hep - ph/0510167 ] .
h. a. weldon , phys .
d * 61 * , 036003 ( 2000 ) [ hep - ph/9908204 ]
. m. kitazawa , t. kunihiro , k. mitsutani and y. nemoto , phys .
d * 77 * , 045034 ( 2008 ) [ arxiv:0710.5809 [ hep - ph ] ] .
y. hidaka , d. satow and t. kunihiro , nucl .
a * 876 * , 93 ( 2012 ) [ arxiv:1111.5015 [ hep - ph ] ] .
d. satow and y. hidaka , phys .
d * 85 * , 116009 ( 2012 ) [ arxiv:1204.6532 [ hep - ph ] ] .
a. j. niemi and g. w. semenoff , nucl .
b * 230 * , 181 ( 1984 ) .
h. matsumoto , i. ojima and h. umezawa , annals phys .
* 152 * , 348 ( 1984 ) .
l. kadanoff , g. baym , `` quantum statistical mechanics , '' benjamin , new york ( 1962 ) .
b. s. dewitt , phys .
rev . * 162 * , 1195 ( 1967 ) .
b. s. dewitt , phys .
* 19 * , 295 ( 1975 ) .
l. f. abbott , nucl .
b * 185 * , 189 ( 1981 ) .
k. a. meissner , acta phys .
b * 17 * , 409 ( 1986 ) .
l. f. abbott , acta phys .
b * 13 * , 33 ( 1982 ) .
y. a. markov and m. a. markova , phys .
105009 ( 2001 ) [ hep - ph/0101167 ] . j. -p .
blaizot and d. satow , in preparation .
d. boyanovsky , phys .
d * 72 * , 033004 ( 2005 ) [ hep - ph/0505186 ] . g. f. giudice , a. notari , m. raidal , a. riotto and a. strumia , nucl .
b * 685 * , 89 ( 2004 ) [ hep - ph/0310123 ] .
d. notzold and g. raffelt , nucl .
b * 307 * , 924 ( 1988 ) .
y. yu and k. yang , phys .
lett . * 100 * , 090404 ( 2008 ) [ arxiv:0707.4119 [ cond-mat.other ] ] .
m. snoek , s. vandoren and h. t. c. stoof , phys .
a * 74 * , 033607 ( 2006 ) [ cond - mat/0604671 ]
. y. yu and k. yang , phys .
lett . * 105 * , 150605 ( 2010 ) [ arxiv:1005.1399 [ cond-mat.quant-gas ] ] .
m. snoek , m. haque , s. vandoren , and h. t. c. stoof , phys .
lett . * 95 * , 250401 ( 2005 ) . | at so high temperature ( @xmath0 ) that the coupling constant ( @xmath1 ) is small and the masses of the particles are negligible , different scheme has to be applied in each energy scale in the analysis of the quark - gluon plasma ( qgp ) . in the soft energy region ( @xmath2 ) , the simple perturbative expansion called the hard thermal loop ( htl ) approximation can be applied , and that approximation expects the existence of the bosonic and fermionic collective excitations called plasmon and plasmino . on the other hand , in the ultrasoft energy region ( @xmath3 ) , the htl approximation is inapplicable due to infrared singularity , so the question whether there are any excitation modes in that energy region has not been studied well . in this thesis
, we analyze the quark spectrum whose energy is ultrasoft in qgp , using the resummed perturbation theory which enables us to successfully regularize the infrared singularity . since the yukawa model and qed are simpler than qcd but have some similarity to qcd , we also work in these models . as a result
, we establish the existence of a novel fermionic mode in the ultrasoft energy region , and obtain the expressions of the pole position and the strength of that mode .
we also show that the ward - takahashi identity is satisfied in the resummed perturbation theory in qed / qcd .
furthermore , we derive the linearized and generalized boltzmann equation for the ultrasoft fermion excitations in the kadanoff - baym formalism , and show that the resultant equation is equivalent to the self - consistent equation in the resummed perturbation theory .
we also derive the equation which determines the @xmath4-point functions with external lines for a pair of fermions and @xmath5 bosons with ultrasoft momenta , by considering the non - linear response regime using the gauge symmetry .
we also derive the ward - takahashi identity from the conservation law of the electromagnetic current in the kadanoff - baym formalism . |
the emergence of the internet has changed the way of communication radically and , especially , the development of web 2.0 applications has led to some extremely popular online social sites , such as facebook @xcite , flickr @xcite , youtube @xcite , twitter @xcite , livejournal @xcite , orkut @xcite and xiaonei @xcite .
these sites provide a powerful means of sharing information , finding content and organizing contacts @xcite for ordinary people .
users can consolidate their existing relationships in the real world through publishing blogs , photos , messages and even states .
they also have a chance to communicate with strangers that they have never met on the other end of the world .
based on the development and prevalence of the internet , online social sites have reformed the structure of the traditional social network to a new complex system , called the online social network , which attracts a lot of research interests recently as a new social media .
recent works about online social networks mainly focus on probing and collecting network topologies @xcite , structural analysis @xcite , user interactions @xcite and content generating patterns @xcite . at the same time
, some concepts and methods of traditional social networks have also been introduced into current researches : the strength of ties is one of them .
the strength of ties was first proposed by granovetter in his landmark paper @xcite in 1973 , in which he thought the strength of ties could be measured by the relative overlap of the neighborhood of two nodes in the network .
it was interesting that different from the common sense , he found that loose acquaintances , known as weak ties , were helpful in finding a new job @xcite .
this novel finding has become a hot topic of research for decades . in @xcite ,
a predictive model was proposed to map social media data to the tie strength . in @xcite ,
onnela et al . gave a simple but quantified definition to the overlap of neighbors of nodes @xmath0 and @xmath1 as follows : @xmath2 where @xmath3 is the number of common acquaintances , @xmath4 and @xmath5 are the degrees of @xmath0 and @xmath1 , respectively .
* in this paper , we define @xmath6 as the strength of the tie between @xmath7 and @xmath8*. the lower @xmath9 is , the weaker the strength of tie between @xmath0 and @xmath1 is . as a social media ,
the core feature of online social networks is the information diffusion .
however , the mechanism of the diffusion is different from traditional models , such as susceptible - infected - susceptible ( sis ) , susceptible - infected - recovered ( sir ) @xcite and random walk @xcite . at the same time
, few works have been done to reveal the coupled dynamics of both the structure and the diffusion of online social networks @xcite . to meet this critical challenge , in this paper
, we aim to investigate the role of weak ties in the information diffusion in online social networks . by monitoring the dynamics of @xmath10 where @xmath11 is the number of connected clusters with @xmath12 nodes , and
@xmath13 is the size of the network , a phase transition was found in the mobile communication network during the removal of weak ties first @xcite .
we find that this phase transition is pervasive in online social networks , which implies that weak ties play a special role in the structure of the network .
this interesting finding inspires us to investigate the role of weak ties in the information diffusion . to this end
, we propose a model @xmath14 to characterize the mechanism of the information diffusion in online social networks and associate the strength of ties with the process of spread . through the simulations on large - scale real - world data sets , we find that selecting weak ties preferentially to republish can not make the information diffuse quickly , while the random selection can .
nevertheless , further analysis and experiments show that the coverage of the information will drop substantially during the removal of weak ties even for the random diffusion case .
so we conclude that weak ties play a subtle role in the information diffusion in online social networks .
we also discuss their potential use for the information diffusion control practices .
the rest of this paper is organized as follows .
section [ sec : datasets ] introduces the data sets used in this paper . in section
[ sec : sroleofweakties ] , we study the structural role of weak ties .
the model @xmath14 is proposed in section [ sec : droleofweakties ] , and the role of weak ties in the information diffusion is then investigated .
section [ sec : dcontrol ] discusses the possible uses of weak ties in the control of the virus spread and the private information diffusion .
finally , we give a brief summary in section [ sec : summary ] .
we use two data sets in this paper , i.e. , ` youtube ` and ` facebook ` in new orleans . `
youtube ` is a famous video sharing site , and ` facebook ` is the most popular online social site which allows users to create friendships with other users , publish blogs , upload photos , send messages , and update their current states on their profile pages .
all these sites have some privacy control schemes which control the access to the shared contents .
the data set of ` youtube ` includes user - to - user links crawled from ` youtube ` in 2007 @xcite .
the data set of ` facebook ` contains a list of all the user - to - user links crawled from the new orleans regional network in ` facebook ` during december 29th , 2008 and january 3rd , 2009 @xcite . in both two data sets
, we treat the links as undirected . in these data sets ,
each node represents a user , while a tie between two nodes means there is a friendship between two users . in general , creating a friendship between two users always needs mutual permission .
so we can formalize each data set as an undirected graph @xmath15 , where @xmath16 is the set of nodes and @xmath17 is the set of ties .
we use @xmath18 to denote the size of the network , and @xmath19 to denote the size of ties .
some characteristics of the data sets are shown in table [ tab : datasets ] .
the @xmath20 of the strength of ties is shown in fig .
[ fig : cdfofstrength ] .
.[tab : datasets]data sets [ cols="^,^,^",options="header " , ] ( color online ) @xmath21 of the strength of ties . ] as we know , online social networks are divided into two types : knowledge - sharing oriented and networking oriented @xcite . for the data sets we use , ` youtube ` belongs to the former , while ` facebook ` belongs to the latter , both of which are scale - free networks .
+ in this section , we study the structural role of weak ties . as shown in fig .
[ fig : facebook_average_s ] and fig .
[ fig : youtube_average_s ] , we find a phase transition ( characterized by @xmath22 ) similar to the one in @xcite in online social networks during the removal of weak ties first .
this phase transition , however , disappears if we remove the strong ties first .
furthermore , it is also found in fig .
[ fig : facebook_gcc ] and fig .
[ fig : youtube_gcc ] that the relative size of giant connected cluster ( gcc ) , denoted by @xmath23 , shows different dynamics between the removals of weak ties first and strong ties first .
we denote the critical fractions of the removed ties at the phase transition point by @xmath24 .
it is interesting to note that @xmath25 for ` youtube ` and @xmath26 for ` facebook ` when @xmath22 reaches the submit , which are very close to the case when @xmath27 . in the percolation theory
, the existence of the above phase transition means that the network is collapsed , while the network is just shrinking if there is no phase transition when removing the ties @xcite .
so the above experiments tell us that weak ties play a special role in the structure of online social networks , which is different from the one strong ties play .
in fact , they act as the important bridges that connect isolated communities . in
what follows , we build a model that associates the weak ties with the information diffusion , to discuss the coupled dynamics of the structure and the information diffusion .
the information diffusing in online social networks includes blogs , photos , messages , comments , multimedia files , states , etc . because of the privacy control and other features of online social sites , the mechanism of the information diffusion in online social networks is different from traditional models , such as sis , sir and random walk .
we start by discussing the procedure of information diffusion in online social networks .
the procedure of the diffusion in online social networks can be briefly described as follows : * the user @xmath0 publishes the information @xmath28 , which may be a photo , a blog , etc . *
friends of @xmath0 will know @xmath28 when they access the profile page of @xmath0 or get some direct notifications from the online social site .
we call this scheme as _ push_. * some friends of @xmath0 , may be one , many or none , will comment , cite or reprint @xmath28 , because they think that it is interesting , funny or important
. we call this behavior as _ republish_. * the above steps will be repeated with @xmath0 replaced by each of those who have republished @xmath28 . it is easy to find that the key feature of the information diffusion in online social networks is that the information is pushed actively by the site and only part of friends will republish it . take ` facebook ` as an example , in which _ news feed _ and _ live feed _ are two significant and popular features . news feed constantly updates a user s profile page to list all his or her friends news in ` facebook ` .
the news includes conversations taking place between the walls of the user s friends , changes of profile pages , events , and so on @xcite .
live feed facilitates the users to access the details of the contents updated by news feed .
it is updated in a real - time manner after the user s login to the web @xcite . in fact , news feed aggregates the most interesting contents that a user s friends are posting , while live feed shows to the user all the actions his or her friends are taking in ` facebook ` @xcite .
the feature of pushing and republishing we have discussed above is indeed more obvious in ` twitter ` , in which all the words you post will be pushed immediately to your followers terminals , including a pc or even a mobile phone , and then they can republish it if they like . however , in real - world situations , the trace of the information is hard to collect @xcite , especially for large - scale networks .
so it is quite reasonable to build a model to characterize the mechanism and simulate the diffusion .
based on the procedure described above , we propose a simple model @xmath14 , where @xmath29 is the navigating factor and @xmath30 represents the strength of the information . in this model , @xmath29 determines how to select neighbors to republish the information , while @xmath31 $ ] is a physical character of the information , which describes how interesting , novel , important , funny or resounding it is .
the model is defined as follows : * step 1 : suppose there comes information @xmath28 .
set the state of all the nodes in @xmath16 to @xmath32 .
the state @xmath32 of a node means @xmath28 is not known to it , otherwise the state is @xmath33 .
* step 2 : randomly select a seed node @xmath0 from the network .
the degree of @xmath0 is @xmath4 .
set @xmath0 to @xmath33 .
it publishes the information @xmath28 with strength equal to @xmath30 at time @xmath34 . *
step 3 : increase the time by one unit , i.e. , @xmath35 .
set each node in the neighborhood of @xmath0 to @xmath33 .
add @xmath0 to the set of nodes that have published @xmath28 , denoted by @xmath36 .
so @xmath37 . *
step 4 : calculate the number of nodes that will republish @xmath28 in the next round : @xmath38 * step 5 : select one node @xmath1 from the neighborhood of @xmath0 with the probability @xcite @xmath39 if @xmath1 is not in @xmath36 , then add it to the set of nodes that will republish @xmath28 in the next round , denoted by @xmath40 .
so @xmath41 .
repeat this step for @xmath42 times .
* step 6 : for each node in @xmath40 , execute from step 3 to step 5 recursively until @xmath40 is null or all the nodes in @xmath16 have known @xmath28 .
it is easy to find from eq .
( [ eq : rounds ] ) that during the diffusion , the number of republishing nodes selected from the neighborhood of @xmath0 is decided by @xmath4 and @xmath30 .
it is consistent with the real situation that the user with more friends tends to attract more other users to visit and republish the information .
the more interesting or important the information is , the higher the chance that it will be republished .
we use parameter @xmath29 in eq .
( [ eq : probability ] ) to associate the diffusion with the strength of the ties , which means different values of @xmath29 will lead to different selections of ties as paths for republishing information in the next round .
in fact , when @xmath43 , weak ties are to be selected preferentially as paths for republishing .
the selection is random when @xmath44 , and the strong ties will be selected with higher priority when @xmath45 .
we define the fraction of nodes with the state @xmath33 as the coverage of @xmath28 , denoted by @xmath46 . since it is found that only 1 - 2% friends will republish the information in flickr @xcite , we let @xmath47 in the simulations .
[ fig : id_visited_t ] shows the numeric experimental results on ` facebook ` and ` youtube ` networks .
as can be seen , @xmath46 reaches the maximum when @xmath44 . in other words ,
compared with weak or strong ties , selecting the republishing nodes randomly from the neighborhood will make the information spread faster and wider .
this is indeed out of our expectation , since previous studies show that weak ties can facilitate the information diffusion in social networks . to understand this
, we further explore the process of the information diffusion in details . by eq .
( [ eq : tiestrength ] ) , we can easily have @xmath48 assume that as @xmath5 increases , @xmath3 increases proportionately , i.e. , @xmath49 .
then given a node @xmath0 and its neighbor node @xmath1 , we have @xmath50 , and vice versa .
this implies that a neighbor node of @xmath0 tends to have a higher degree if it has a stronger strength of ties with @xmath0 .
therefore , when selecting the republishing nodes for the next round from the neighborhood , different @xmath29 will select nodes with different degrees preferentially . for example , when @xmath43 , the weak ties will be selected with higher priority , which means that the nodes with lower degrees will be selected preferentially . however , it is easy to learn from eq .
( [ eq : rounds ] ) that , for the node with lower degree , the republishing nodes selected from its neighborhood will be less , which will eventually reduce the total number of republishing nodes and impede the information from further spreading in the network .
as to the case of selecting strong ties preferentially , although it will tend to select the nodes with higher degrees to republish , the local trapping @xcite will limit the scope of selected nodes into some local areas and make it harder to propagate the information further in the network . to validate the analysis above
, we also observe the fraction of the nodes that have published @xmath28 during the diffusion , denoted by @xmath51 . as shown in fig .
[ fig : fpub ] , @xmath51 increases more slowly when @xmath43 , and the time - varying properties of @xmath51 are similar to those of @xmath46 in fig .
[ fig : id_visited_t ] for different @xmath29 values , respectively .
we also monitor the fraction of the nodes that have published @xmath28 in each hop away from the source node , denoted by @xmath52 . as shown in fig .
[ fig : flocal ] , when @xmath43 , @xmath52 decreases faster than other cases , in particular the @xmath44 case .
it means when @xmath43 , the number of republishing nodes selected from the neighborhood decreases sharply as the information spreading far away from the source , which agrees with our former analysis . as for the case of @xmath45 ,
@xmath51 increases more and more slowly during the diffusion , because the nodes selected to republish are trapped in some local clusters .
in other words , it is hard to find some new nodes to republish the information to the outer space .
based on the above results , we can conclude that selecting weak ties preferentially as the path to republish information can not make it diffuse faster .
however , this does not mean that weak ties play a trivial role in the information diffusion in online social networks , especially when we recall its special role in the network structure in section [ sec : sroleofweakties ] .
let @xmath44 in @xmath14 , we compare the variation of @xmath46 under the situation of removing weak ties first with that of removing strong ties first . as shown in fig .
[ fig : remove_ties ] , for the case of removing weak ties first , the coverage of the information decreases rapidly , e.g. , from 0.8 to 0.4 in ` facebook ` when the fraction of removed weak ties reaches about 0.4 .
this implies that weak ties are indeed crucial for the coverage of information diffusion in online social networks .
to further study the effect of @xmath30 , we conduct experiments with different @xmath30 values , as shown in fig . [
fig : coverage_beta ] . as can be seen ,
no matter what the @xmath30 value is , random selection ( @xmath44 ) is still the fastest mode for the information diffusion , although the gap tends to shrink with higher @xmath30 values .
it is also shown that when @xmath30 grows , @xmath46 will also rise for all @xmath29 values .
that is , the greater the strength of the information is , the more nodes will be attracted to republish it , and the wider it will spread in the network .
until now we can conclude that weak ties play a subtle role in the information diffusion in online social networks .
on one hand , they are bridges that connect isolated communities and break through the trapping of information in local areas @xcite . on the other hand , selecting weak ties preferentially as the path of republishing can not make the information diffuse faster and wider .
the growing popularity of the online social networks does not mean that it is safe and reliable .
on the contrary , the virus spread and the private information diffusion have made it become a massive headache for it administrators and users @xcite .
for example , `` kooface '' is a trojan worm on facebook , which spreads by leaving a comment on profile pages of the victim s friends to trap a click on the malicious link @xcite .
about 63% of system administrators worry that their employees will share too much private information online @xcite .
so as time goes by , it becomes more and more important and urgent to control the virus spread and the private information diffusion in online social networks . in the light of this
, we can make use of the weak ties for the information diffusion control .
that is , in the real - world practices , we can assume that the behavior of republishing information is random , i.e. , @xmath44 . then according to the results in fig .
[ fig : remove_ties ] , we can make the virus or the private information trapped in local communities by removing weak ties and stop them from diffusing further in the network .
online social sites have become one of the most popular web 2.0 applications in the internet . as a new social media ,
the core feature of online social networks is the information diffusion .
we investigate the coupled dynamics of the structure and the information diffusion in the view of weak ties .
different from the recent work @xcite , we do not focus on the trace collection and analysis of the real data flowing in the network . instead , inspired by @xcite
, we propose a model for online social networks and take a closer look at the role of weak ties in the diffusion .
we find that the phase transition found in the mobile communication network exists pervasively in online social networks , which means that the weak ties play a special role in the network structure .
then we propose a new model @xmath14 , which associates the strength of ties with the diffusion , to simulate how the information spreads in online social networks .
contrary to our expectation , selecting weak ties preferentially to republish can not facilitate the information diffusion in the network , while the random selection can . through extra analysis and experiments
, we find that when @xmath43 , the nodes with lower degrees are preferentially selected for republishing , which will limit the scope of the distribution of republishing nodes in the following rounds .
however , even for the random selection case , removal of the weak tie can make the coverage of the information decreases sharply , which is consistent with its special role in the structure .
so we conclude that weak ties play a subtle role in the information diffusion in online social networks .
on one hand , they play a role of bridges , which connect isolated communities and break through the trapping of information in local areas . on the other hand , selecting weak ties preferentially to republish can not make the information diffuse faster in the network . for potential applications ,
we think that the weak ties might be of use in the control of the virus spread and the private information diffusion .
this work was supported by national 973 program of china ( grant no.2005cb321901 ) and the fund of the state key laboratory of software development environment ( sklsde-2008zx-03 ) .
the second author was supported partially by national natural science foundation of china ( grant no .
70901002 and 90924020 ) and beihang innovation platform funding ( grant .
no ymf-10 - 04 - 024 ) . | as a social media , online social networks play a vital role in the social information diffusion .
however , due to its unique complexity , the mechanism of the diffusion in online social networks is different from the ones in other types of networks and remains unclear to us .
meanwhile , few works have been done to reveal the coupled dynamics of both the structure and the diffusion of online social networks .
to this end , in this paper , we propose a model to investigate how the structure is coupled with the diffusion in online social networks from the view of weak ties . through numerical experiments on large - scale online social networks ,
we find that in contrast to some previous research results , selecting weak ties preferentially to republish can not make the information diffuse quickly , while random selection can achieve this goal .
however , when we remove the weak ties gradually , the coverage of the information will drop sharply even in the case of random selection .
we also give a reasonable explanation for this by extra analysis and experiments .
finally , we conclude that weak ties play a subtle role in the information diffusion in online social networks .
on one hand , they act as bridges to connect isolated local communities together and break through the local trapping of the information . on the other hand , selecting them as preferential paths to republish can not help the information spread further in the network . as a result , weak ties might be of use in the control of the virus spread and the private information diffusion in real - world applications . |
with the integrated luminosity of about 50 pb@xmath2 collected at hera during the years 19941997 a new kinematic domain of large @xmath0 and @xmath1 becomes accessible for the study of deep inelastic scattering in @xmath3 collisions .
measurements by zeus of the @xmath4 single differential neutral current ( nc ) and charged current ( cc ) born cross sections for @xmath5 gev@xmath6 have recently become available @xcite .
standard model ( sm ) predictions calculated with , for instance , the parton distribution set cteq4 @xcite are in good agreement with the nc cross sections but fall below the cc measurements at large @xmath0 and @xmath1 .
this could be an indication of new physics beyond the sm but might also be due to an imperfect knowledge of the parton densities in this kinematic region .
for instance in @xcite it is shown that a modification of the cteq4 down quark density yields sm predictions in agreement with the cc @xmath4 data . to investigate these issues
we have performed a global nlo qcd analysis @xcite of structure function data to obtain the parton densities in the proton .
a full error analysis provides estimates of the uncertainties in these parton densities , the structure functions and the sm predictions of the nc and cc cross sections .
the data used in the fit are @xmath7 from zeus @xcite and h1 @xcite together with @xmath7 , @xmath8 @xcite and @xmath9 @xcite from fixed target experiments .
also included were neutrino data on @xmath10 @xcite for @xmath11 and data on the difference @xmath12 @xcite .
after cuts the data cover a kinematic range of @xmath13 , @xmath14 gev@xmath6 and @xmath15 gev@xmath6 .
both @xmath8 and @xmath9 as well as @xmath10 were corrected for nuclear effects using the parameterization of @xcite for deuterium ( with an assumed uncertainty of 100% ) and that of @xcite for iron ( assumed uncertainty 50% ) .
the qcd predictions for the structure functions were obtained by solving the qcd evolution equations in nlo in the scheme @xcite . at the scale @xmath16 gev@xmath6 the gluon density ( @xmath17 ) , the sea and valence quark densities ( @xmath18 ) and the difference @xmath12 were parameterized in the standard way @xcite ( 16 free parameters ) .
the strange quark density was taken to be a fraction @xmath19 of the sea @xcite .
the normalizations of the parton densities were constrained such that the momentum sum rule and the valence quark counting rules are satisfied .
the charm and bottom quarks were assumed to be massless and were generated dynamically above the thresholds @xmath20 and @xmath21 gev@xmath6 respectively .
the value of the strong coupling constant was set to @xmath22 .
higher twist contributions to @xmath7 and @xmath8 were taken into account phenomenologically by describing these structure functions as @xmath23\ ] ] where @xmath24 obeys the nlo qcd evolution equations and where @xmath25 was parameterized as a fourth degree polynomial in @xmath0 ( 5 free parameters ) .
the normalizations of the zeus , h1 and nmc data were kept fixed to unity whereas those of the remaining data sets were allowed to float within the quoted normalization errors ( 7 parameters ) .
the uncertainties in the parton densities , structure functions and related cross sections were estimated taking into account all correlations . included in the error calculation
are the experimental statistical errors , 57 independent sources of systematic uncertainty ( propagated using the technique described in @xcite ) and the errors on the input parameters of the fit .
the fit yielded a good description of the data with a @xmath26 of 1540 for 1578 data points and 28 free parameters .
the quality of the fit is illustrated in fig .
[ fig : f2pvsq ] where we show the fixed target @xmath7 data for @xmath11 .
the higher twist coefficient @xmath25 in eq .
( [ eq : htwist ] ) is found to be negative for @xmath27 and becomes large and positive at high @xmath0 .
our result on the higher twist contribution is very close to that obtained by mrst @xcite .
nuclear effects in neutrino scattering were investigated by calculating the ratios @xmath28 and @xmath29 of the ccfr data to the qcd predictions of scattering on a free nucleon .
both these ratios do not significantly depend on @xmath1 in accordance with measurements of nuclear effects in muon scattering @xcite .
the ratios , averaged over @xmath1 , are plotted in fig .
[ fig : nuceff ] and show the typical @xmath0 dependence of nuclear effects , including the rise at large @xmath0 due to fermi motion .
however , within the present accuracy it is not possible to establish whether these nuclear effects are significantly different from those measured in charged lepton scattering ( full curves in fig .
[ fig : nuceff ] ) . are shown the parton densities obtained from the qcd fit ( full curves ) .
we have verified that the strange quark density is compatible with the measurements from ccfr @xcite and also that the qcd prediction of @xmath30 agrees well with the zeus data @xcite above @xmath31 gev@xmath6 .
this supports the assumption made in this analysis that , at least for @xmath32 , quark mass effects do not spoil the qcd extrapolations to large @xmath1 .
also shown in fig .
[ fig : allpdf ] are the parton densities from cteq4 and cteq5 @xcite .
there is , within errors , good agreement between these results and the qcd fit , although both the present analysis and cteq5 yield a slightly harder @xmath33 density than cteq4 .
bodek and yang @xcite also obtain a harder @xmath34 density by modifying cteq4 ( where the ratio @xmath35 as @xmath36 ) such that @xmath37 they find @xmath38 which implies that @xmath39 as @xmath36 .
the @xmath40 ratio is shown in fig .
[ fig : dou ] .
it is seen that the result from the qcd fit with @xmath41 ( full curve ) is for @xmath42 close to the modified cteq4 distribution with @xmath43 ( dotted curve ) ; if we leave @xmath44 a free parameter in the fit , we obtain @xmath45 ( statistical error ) , close to zero . in any case
, the large error band clearly indicates that the exact behavior of @xmath40 at large @xmath0 is not well constrained .
it might go to zero ( cteq ) , to a constant ( bodek and yang ) or may even diverge ( this analysis ) as @xmath46 .
the cc @xmath47 cross section is predominantly sensitive to the @xmath48 quark density @xmath49 where @xmath50 with @xmath51 the @xmath3 center of mass energy ( @xmath5210@xmath53 gev@xmath6 at hera ) . in fig .
[ fig : ratios ] we show the zeus measurements @xcite of the cc @xmath4 single differential cross sections @xmath54 and @xmath55 for @xmath56 gev@xmath6 , normalized to the nlo predictions calculated with the cteq4 parton distribution set .
the full ( dotted ) curves correspond to the predictions from the qcd fit ( cteq5 ) which both achieve a better description of the data at large @xmath1 and @xmath0 due to an improved determination of the @xmath48 quark density .
sm predictions of the @xmath4 nc cross sections , calculated with the parton densities obtained in this analysis , also agree very well with the recent zeus data @xcite ( not shown ) .
99 zeus collab . , desy 99 - 056 .
zeus collab .
, desy 99 - 059 .
cteq collab .
lai et al .
d55 ( 1997 ) 1280 .
yang and a. bodek , phys .
82 ( 1999 ) 2467
. m. botje , nikhef 99 - 011 , desy 99 - 038 ( in preparation ) .
zeus collab .
, m. derrick et al .
, z. phys .
c72 ( 1996 ) 399 .
, s. aid et al .
b470 ( 1996 ) 3 .
e665 collab .
adams et al . , phys .
d54 ( 1996 ) 3006 .
nmc , m. arneodo et al .
b483 ( 1997 ) 3 .
bcdms collab .
benvenuti et al . , phys .
lett . b223 ( 1989 ) 485 and phys .
b237 ( 1990 ) 592 .
slac , l.w .
whitlow et al . , phys .
b282 ( 1992 ) 475 .
nmc , m. arneodo et al . ,
b487 ( 1997 ) 3 .
ccfr collab .
, w. seligman et al .
79 ( 1997 ) 1213 .
e866 collab .
hawker et al .
80 ( 1998 ) 3715 .
j. gomez et al . , phys .
d49 ( 1994 ) 4348 .
eskola , v.j .
kolhinen and c.a .
salgado , jyfl-8/98 , us - ft/14 - 98 , hep - ph/9807297 ; k.j .
eskola , v.j .
kolhinen and p.v .
ruuskanen , nucl .
535 ( 1998 ) 351 .
g. gurci , w. furmanski and r. petronzio , nucl .
b175 ( 1980 ) 27 ; w. furmanski and r. petronzio , phys .
lett . 97b ( 1980 ) 437 and z. phys .
c11 ( 1982 ) 293 .
a.d martin , w.j .
stirling and r.g .
roberts , phys .
d50 ( 1994 ) 6734 .
ccfr collab .
bazarko et al . , z. phys .
c65 ( 1995 ) 189 .
c. pascaud and f. zomer , lal 95 - 05 .
martin et al .
lett . b443
( 1998 ) 301 .
nmc , a. amaudruz et al .
b441 ( 1995 ) 3 ; m. arneodo et al .
b441 ( 1995 ) 12 .
i. redondo , in these proceedings .
cteq collab .
lai et al .
, hep - ph/9903282 . | the parton momentum densities in the proton are determined from a nlo qcd analysis of structure functions measured by hera and fixed target experiments . also included
are data on the difference of the up and down anti - quark densities .
the uncertainties in the parton densities , structure functions and related cross sections are estimated from the experimental errors , taking into account all correlations .
standard model predictions of the charged current born cross sections at large @xmath0 and @xmath1 are calculated and compared with recent data from zeus . |
let @xmath9 be a field of characteristic @xmath10 .
the total milnor @xmath11-ring @xmath12 was introduced in @xcite .
the total galois cohomology ring @xmath13 is canonically isomorphic , under our hypothesis on the characteristic of @xmath9 , to the total galois cohomology ring @xmath14 with coefficients in the trivial galois module @xmath15 .
the witt ring @xmath16 of nondegenerate quadratic forms modulo hyperbolic forms has an decreasing filtration @xmath17 generated by powers of the _ fundamental ideal _
@xmath18 of even rank forms .
the milnor conjecture relates these three objects : milnor @xmath11-theory , galois cohomology , and quadratic forms .
the quotient map @xmath19 induces a graded ring homomorphism @xmath20 called the _ norm residue symbol _ by bass
tate @xcite .
the _ pfister form _
map @xmath21 given by @xmath22 induces a group homomorphism @xmath23 ( see scharlau ( * ? ? ?
* lemma 12.10 ) ) , which extends to a surjective graded ring homomorphism @xmath24 , see milnor ( * ? ? ?
let @xmath9 be a field of characteristic @xmath10 .
there exists a graded ring homomorphism @xmath25 called the _ higher invariants _ of quadratic forms , which fits into the following diagram @xmath26^{h^{{\bullet } } } \ar[d]_{s^{{\bullet } } } & h^{{\bullet}}(f,{\bm{\mu}}_2^{{\otimes}{\bullet } } ) \\
i^{{\bullet}}(f)/i^{{\bullet}+1}(f ) \ar[ur]_{e^{{\bullet } } } & } \ ] ] of isomorphisms of graded rings .
many excellent introductions to the milnor conjecture and its proof exist in the literature .
for example , see the surveys of kahn @xcite , friedlander rapoport
suslin @xcite , friedlander @xcite , pfister @xcite , and morel @xcite .
the conjecture breaks up naturally into three parts : the conjecture for the norm residue symbol @xmath27 , the conjecture for the pfister form map @xmath28 , and the conjecture for the higher invariants @xmath29 .
milnor ( * ? ? ?
* question 4.3 ,
6 ) originally made the conjecture for @xmath27 and @xmath28 , which was already known for finite , local , global , and real closed fields , see ( * ? ? ?
* lemma 6.2 ) . for general fields ,
the conjecture for @xmath30 follows from hilbert s theorem 90 , and for @xmath31 and @xmath32 by elementary arguments .
the conjecture for @xmath33 is easy , see pfister @xcite .
merkurjev @xcite proved the conjecture for @xmath34 ( hence for @xmath35 as well ) , with alternate proofs given by arason @xcite , merkurjev @xcite , and wadsworth @xcite .
the conjecture for @xmath36 was settled by merkurjev
suslin @xcite ( and independently by rost @xcite ) . the conjecture for @xmath29
can be divided into two parts : to show the existence of maps @xmath37 ( which are _ a priori _ only defined on generators , the pfister forms ) , and then to show they are surjective .
the existence of @xmath38 was proved by arason @xcite , @xcite .
the existence of @xmath39 was proved by jacob
rost @xcite and independently szyjewski @xcite .
voevodsky @xcite proved the conjecture for @xmath27 .
voevodsky @xcite proved the conjecture for @xmath28 , with different proofs given by morel @xcite and kahn
sujatha @xcite .
the theory of quadratic forms over a general field has its genesis in witt s famous paper @xcite .
because of the assumption of characteristic @xmath10 , we do not distinguish between quadratic and symmetric bilinear forms .
the _ orthogonal sum _ @xmath40 and _ tensor product _ @xmath41 give a semiring structure on the set of isometry classes of symmetric bilinear forms over @xmath9 .
the _ hyperbolic plane _ is the symmetric bilinear form @xmath42 , where @xmath43 and @xmath44 .
the _ witt ring _ of symmetric bilinear forms is the quotient of the grothendieck ring of nondegenerate symmetric bilinear forms over @xmath9 with respect to @xmath45 and @xmath46 , modulo the ideal generated by the hyperbolic plane , see scharlau ( * ? ? ?
* ch . 2 ) .
the _ rank _ of a bilinear form @xmath47 is the @xmath9-vector space dimension of @xmath48 .
since the hyperbolic plane has rank 2 , the rank modulo 2 is a well defined invariant of an element of the witt ring , and gives rise to a surjective ring homomorphism @xmath49 whose kernel is the _ fundamental ideal _ @xmath18 . the _ signed discriminant _ of a non - degenerate bilinear form @xmath47 is defined as follows .
choosing an @xmath9-vector space basis @xmath50 of @xmath48 , we consider the gram matrix @xmath51 of @xmath52 , i.e. the matrix given by @xmath53 .
then @xmath52 is given by the formula @xmath54 , where @xmath55 .
the gram matrix of @xmath52 , with respect to a different basis for @xmath48 with change of basis matrix @xmath56 , is @xmath57 .
thus @xmath58 , which depends on the choice of basis , is only well - defined up to squares . for @xmath59 ,
denote by @xmath60 its class in the abelian group @xmath61 .
the signed discriminant of @xmath47 is defined as @xmath62 . introducing the sign into the signed discriminant ensures its vanishing on the ideal of hyperbolic forms ,
hence it descents to the witt group .
while the signed discriminant is not additive on @xmath16 , its restriction to @xmath18 gives rise to a group homomorphism @xmath63 which is easily seen to be surjective .
it s then not difficult to check that its kernel coincides with the square @xmath64 of the fundamental ideal . see scharlau @xcite for more details .
the _ clifford invariant _ of a non - degenerate symmetric bilinear form @xmath47 is defined in terms of its clifford algebra .
clifford algebra _
@xmath65 of @xmath47 is the quotient of the tensor algebra @xmath66 by the two - sided ideal generated by @xmath67 .
if @xmath47 has rank @xmath68 , then @xmath69 inherits the structure of a @xmath15-graded semisimple @xmath9-algebra of @xmath9-dimension @xmath70 , see scharlau @xcite . by the structure theory of the clifford algebra ,
@xmath65 or @xmath71 is a central simple @xmath9-algebra depending on whether @xmath47 has even or odd rank , respectively .
clifford invariant _
@xmath72 is then the brauer class of @xmath65 or @xmath71 , respectively .
since the clifford algebra and its even subalgebra carry canonical involutions of the first kind , their respective classes in the brauer group are of order 2 , see knus @xcite .
while the clifford invariant is not additive on @xmath16 , its restriction to @xmath64 gives rise to a group homomorphism @xmath73 see knus ( * ? ? ?
* iv prop .
8.1.1 ) .
any symmetric bilinear form @xmath47 over a field of characteristic @xmath10 can be diagonalized , i.e. a basis can be chosen for @xmath48 so that the gram matrix @xmath51 is diagonal . for @xmath74 , we write @xmath75 for the standard symmetric bilinear form with associated diagonal gram matrix . for @xmath76 , denote by @xmath77 the ( quaternion ) @xmath9-algebra generated by symbols @xmath78 and @xmath79 subject to the relations @xmath80 , @xmath81 , and @xmath82 .
for example , @xmath83 is hamilton s ring of quaternions .
then the discriminant and clifford invariant can be conveniently calculated in terms of a diagonalization . for @xmath84
, we have @xmath85 and @xmath86 where @xmath87 see lam @xcite , scharlau ( * ? ? ? *
ii.12.7 ) or esnault kahn levine viehweg @xcite .
generalizations ( what we will call _ globalizations _ ) of the milnor conjecture to the context of rings and schemes have emerged from many sources , see parimala @xcite , colliot - thlne
parimala @xcite , parimala
sridharan @xcite , monnier @xcite , pardon @xcite , elbaz - vincent mller - stach @xcite , gille @xcite , and kerz @xcite . to begin with
, one must ask for appropriate globalizations of the objects in the conjecture : milnor @xmath11-theory , galois cohomology theory , and the witt group with its fundamental filtration .
while there are many possible choices of such globalizations , we will focus on two types : global and unramified .
let @xmath9 be a field of characteristic @xmath10 .
let @xmath88 ( resp .
@xmath89 ) be the category of fields ( resp.commutative unital rings ) with an @xmath9-algebra structure together with @xmath9-algebra homomorphisms .
let @xmath90 be the category of @xmath9-schemes , and @xmath91 the category of smooth @xmath9-schemes .
we will denote , by the same names , the associated ( large ) zariski sites .
let @xmath92 ( resp .
@xmath93 ) be the category of abelian groups ( resp.graded abelian groups ) , we will always consider @xmath92 as embedded in @xmath93 in degree 0 .
let @xmath94 be a functor .
a _ globalization _ of @xmath4 to rings ( resp .
schemes ) is a functor @xmath95 ( resp .
contravariant functor @xmath96 ) extending @xmath4 .
if @xmath97 is a globalization of @xmath4 to rings , then we can define a globalization to schemes by taking the sheaf @xmath98 associated to the presheaf @xmath99 on @xmath90 ( always considered with the zariski topology ) . for a commutative unital ring @xmath100 , mimicking milnor s tensorial construction ( with the additional relation that @xmath101 , which is automatic for fields ) yields a graded ring @xmath102 , which should be referred to as `` nave '' milnor @xmath11-theory .
this already appears in guin @xcite and later studied by elbaz - vincent mller - stach @xcite .
nave milnor @xmath11-theory has some bad properties when @xmath100 has small finite residue fields , see kerz @xcite who also provides a improved milnor @xmath11-theory repairing these defects .
thomason @xcite has shown that there exists no globalization of milnor @xmath11-theory to smooth schemes which satisfies @xmath103-homotopy invariance and has a functorial homomorphism to algebraic @xmath11-theory .
tale cohomology provides a natural globalization of galois cohomology to schemes .
we will thus consider the functor @xmath104 on @xmath90 .
for a scheme @xmath105 , the _ global _ witt group @xmath106 of regular symmetric bilinear forms introduced by knebusch @xcite provides a natural globalization of the witt group to schemes .
other possible globalizations are obtained from the witt groups of triangulated category with duality introduced by balmer @xcite , @xcite , @xcite , @xcite .
these include : the _ derived _ witt group of the bounded derived category of coherent locally free @xmath107-modules ; the _ coherent _ witt group of the bounded derived category of quasicoherent @xmath107-modules with coherent cohomology ( assuming @xmath105 has a dualizing complex , see gille @xcite , @xcite ) ; the _ perfect _ witt group of the derived category of perfect complexes of @xmath107-modules .
the global and derived witt groups are canonically isomorphic by balmer ( * ? ? ?
all of the above witt groups are isomorphic ( though not necessarily canonically ) if @xmath105 is assumed to be regular .
globalizations of the classical invariants of quadratic forms are defined as follows .
let @xmath108 be a regular symmetric bilinear form of rank @xmath109 on @xmath105 .
the rank ( modulo 2 ) of @xmath110 gives rise to a functorial homomorphism @xmath111 whose kernel @xmath112 is called the _ fundamental ideal _ of @xmath106 .
the _ signed discriminant form _
@xmath113 gives rise to a functorial homomorphism @xmath114 see knus ( * ? ? ?
* iii 4.2 ) and milne ( * ? ? ?
* iii 4 ) .
alternatively , the center of the ( even ) clifford @xmath107-algebra of @xmath108 defines a class in @xmath115 called the _
arf invariant _ , which coincides with the signed discriminant under the canonical morphism @xmath116 ( see knus ( * ? ? ?
* iv prop.4.6.3 ) or parimala srinivas @xcite ) . denote the kernel of @xmath32 by @xmath117 , which is an ideal of @xmath106 .
note that @xmath117 may not be the square of the ideal @xmath112 .
the clifford @xmath107-algebra @xmath118 gives rise to a functorial homomorphism @xmath119 called the _
clifford invariant _ , see knus
ojanguren @xcite and parimala
srinivas @xcite .
denote the kernel of @xmath35 by @xmath120 , which is an ideal of @xmath106 .
as parimala
srinivas @xcite point out , there is no functorial map @xmath121 lifting the clifford invariant .
instead , we can work with grothendieck
witt groups .
the rank ( modulo 2 ) gives rise to a functorial homomorphism @xmath122 with kernel denoted by @xmath123 .
the signed discriminant gives rise to a functorial homomorphism @xmath124 with kernel denoted by @xmath125 .
the class of the clifford @xmath107-algebra , together with it s canonical involution ( via the `` involutive '' brauer group construction of parimala
srinivas @xcite ) , gives rise to a functorial homomorphism @xmath126 also see knus
parimala sridharan @xcite .
denote the kernel of @xmath127 by @xmath128 , which is an ideal of @xmath129 .
[ lem : gw ] let @xmath105 be a scheme satisfying @xmath130 . then under the quotient map
@xmath131 , the image of the ideal @xmath132 is precisely the ideal @xmath133 , for @xmath134 . for @xmath135
this is a consequence of the following diagram with exact rows and columns @xmath136 & 0 \ar[d ] & & \\ & k_0(x ) \ar@{=}[r ] \ar[d]^h & k_0(x ) \ar[d]^h & & \\ 0 \ar[r ] & gi^{n}(x ) \ar[d ] \ar[r ] & gi^{n-1}(x ) \ar[r]^(.41){ge^{n-1 } } \ar[d ] & { h_{{\mathrm{\acute{e}t}}}}^{n-1}(x,{\mathbb z}/2{\mathbb z } ) \ar@{=}[d ] \ar[r ] & 0 \\ 0 \ar[r ] & i^{n}(x ) \ar[d ] \ar[r ] & i^{n-1}(x ) \ar[d ] \ar[r]^(.41){e^{n-1 } } & { h_{{\mathrm{\acute{e}t}}}}^{n-1}(x,{\mathbb z}/2{\mathbb z } ) \ar[d ] \ar[r ] & 0 \\ & 0 & 0 & 0 & \\ } \ ] ] which is commutative since hyperbolic spaces have even rank and trivial signed discriminant . here
, @xmath137 is the grothendieck group of locally free @xmath107-modules of finite rank and @xmath138 is the hyperbolic map @xmath139 . for @xmath140
, we have the formula @xmath141 due to esnault kahn
viehweg ( * ? ? ?
5.5 ) ( combined with ) . here
@xmath142 is the 1st chern class modulo 2 , defined as the first coboundary map in the long - exact sequence in tale cohomology @xmath143 arising from the kummer exact sequence @xmath144 see grothendieck @xcite .
the claim then follows by a diagram chase through @xmath145 & 0 \ar[d ] & 0 \ar[d ] \\
0 \ar[r ] & k_0'(x ) \ar[d]^h \ar[r ] & k_0(x ) \ar[d]^h \ar[r]^{\det } & \operatorname{\mathrm{pic}}(x)/2 \ar[d]^{c_1 } \ar[r ] & 0\\ 0 \ar[r ] & gi^{3}(x ) \ar[d ] \ar[r ] & gi^{2}(x ) \ar[d ] \ar[r]^(.45){ge^2 } & { h_{{\mathrm{\acute{e}t}}}}^2(x,{\bm{\mu}}_2 ) \ar[d ] & \\ 0 \ar[r ] & i^{3}(x ) \ar[d ] \ar[r ] & i^{2}(x ) \ar[d]\ar[r]^(.45){e^2 } & { { } _ 2\!\operatorname{\mathrm{br}}}(x)\ar[d ] & \\ & 0 & 0 & 0 & \\ } \ ] ] where the right vertical column arises from the kummer sequence , and @xmath146 is the subgroup of @xmath137 generated by locally free @xmath107-modules whose determinant is a square . the hypothesis that @xmath147 is satisfied if @xmath105 is a quasi - compact quasi - separated scheme admitting an ample invertible sheaf by de jong s extension @xcite ( see also ( * ? ? ?
2.2.2.1 ) ) of a result of gabber @xcite . the existence of global globalizations of the higher invariants ( e.g.a globalization of the arason invariant ) remains a mystery .
viehweg @xcite have shown that for a regular symmetric bilinear form @xmath108 that represents a class in @xmath128 , the obstruction to having an arason invariant in @xmath148 is precisely the 2nd chern class @xmath149 in the chow group modulo 2 ( note that the invariant @xmath150 of @xcite is trivial if @xmath108 represents a class in @xmath128 ) .
they also provide examples where this obstruction does not vanish .
on the other hand , higher cohomological invariants always exist in unramified cohomology .
a functorial framework for the notion of `` unramified element '' is established in colliot - thlne @xcite .
see also the survey by zainoulline @xcite .
rost ( * ? ? ?
* rem.5.2 ) gives a different perspective in terms of cycle modules , also see morel @xcite .
assume that @xmath105 has finite krull dimension and is equidimensional over a field @xmath9 .
for simplicity of exposition , assume that @xmath105 is integral .
denote by @xmath151 its set of codimension @xmath152 points .
denote by @xmath153 the category of local @xmath9-algebras together with local @xmath9-algebra morphisms . given a functor @xmath154 ,
call @xmath155 the group of _ unramified elements _ of @xmath4 over @xmath105 .
then @xmath156 is a globalization of @xmath4 to schemes .
given a functor @xmath157 , there is a natural map @xmath158 .
if this map is injective , surjective , or bijective we say that the _ injectivity _ , _ weak purity _ , or _
purity _ property hold , respectively . whether these properties hold for various functors @xmath4 and schemes
@xmath105 is the subject of many conjectures and open problems , see colliot - thlne @xcite for examples .
define the _ unramified milnor @xmath11-theory ( resp .
modulo 2 ) _ of @xmath105 to be the graded ring of unramified elements @xmath159 ( resp .
@xmath160 ) of the `` nave '' milnor @xmath11-theory ( resp . modulo 2 ) functor @xmath161 ( resp .
@xmath162 ) restricted to @xmath153 , see [ subsec : global_theories ] .
let @xmath163 be the zariski sheaf on @xmath90 associated to `` nave '' milnor @xmath11-theory and @xmath164 the associate sheaf quotient , which is the zariski sheaf associated to the presheaf @xmath165 , see morel ( * ? ? ?
* lemma 2.7 ) . then @xmath166 and @xmath167 when @xmath105 is smooth over an infinite field ( compare with the remark in [ subsec : global_theories ] ) by the bloch ogus gabber theorem for milnor @xmath11-theory ,
see colliot - thlne hoobler kahn ( * ? ? ?
* cor.5.1.11 ,
7.3(5 ) ) . also , see kerz @xcite .
note that the long exact sequence in zariski cohomology yields a short exact sequence @xmath168 still assuming @xmath105 is smooth over an infinite field .
define the _ unramified tale cohomology ( modulo 2 ) _ of @xmath105 to be the graded ring of unramified elements @xmath169 of the functor @xmath170 .
letting @xmath171 be the zariski sheaf on @xmath90 associated to the functor @xmath170 , then @xmath172 when @xmath105 is smooth over a field of characteristic @xmath10 by the exactness of the gersten complex for tale cohomology , see bloch ogus ( * ? ? ?
* , thm .
4.2 ) .
define the _ unramified witt group _ of @xmath105 to be the abelian group of unramified elements @xmath173 of the global witt group functor @xmath174 .
letting @xmath175 be the zariski sheaf associated to the global witt group functor , then @xmath176 when @xmath105 is regular over a field of characteristic @xmath10 by ojanguren
panin @xcite ( also see morel ( * ? ? ?
* thm . 2.2 ) ) .
writing @xmath177 , then the functors @xmath178 are zariski sheaves ( still assuming @xmath105 is regular ) , denoted by @xmath179 , which form a filtration of @xmath175 , see morel ( * ? ? ?
note that the long exact sequence in zariski cohomology yields short exact sequences @xmath180 where @xmath181 and we are still assuming @xmath105 is regular over a field of characteristic @xmath10 .
if the obstruction group @xmath182 is nontrivial , then not every element of @xmath183 is represented by a quadratic form on @xmath105 .
if @xmath105 is the spectrum of a regular local ring , then @xmath184 , see morel ( * ? ? ?
* thm.2.12 ) .
[ rem : warning ] as before , the notation @xmath185 does not necessarily mean the @xmath109th power of @xmath186 .
this is true , however , when @xmath105 is the spectrum of a regular local ring containing an infinite field of characteristic @xmath10 , see kerz
mller - stach ( * ? ? ?
* cor.0.5 ) .
gersten complexes ( cousin complexes ) exists in a very general framework , but for our purposes , we will only need the gersten complex for milnor @xmath11-theory , tale cohomology , and witt groups .
let @xmath105 be a regular excellent integral @xmath9-scheme .
let @xmath187 denote the gersten complex for milnor @xmath11-theory @xmath188 & { k_{\text{m}}}^n(f(x ) ) \ar[r]^(.4){{\partial}^{{k_{\text{m}}}^n } } & \smash{\underset{{x \in x^{(1)}}}{\textstyle\bigoplus } } { k_{\text{m}}}^{n-1}(f(x ) ) \ar[r]^(.54){{\partial}^{{k_{\text{m}}}^{n-1 } } } & \smash{\underset{y \in x^{(2)}}{\textstyle\bigoplus } } { k_{\text{m}}}^{n-2}(f(y ) ) \ar[r ] & \dotsm } \ ] ] where @xmath189 is the `` tame symbol '' homomorphism defined in milnor ( * ? ? ?
* lemma 2.1 ) .
we have that @xmath190 .
see rost @xcite or fasel ( * ? ? ?
2 ) for more details . we will also consider the gersten complex @xmath191 for milnor @xmath11-theory modulo 2 , for which we have that @xmath192 .
let @xmath105 be a smooth integral @xmath9-scheme , with @xmath9 of characteristic @xmath10 .
let @xmath193 denote the gersten complex for tale cohomology @xmath188 & h^{n}(f(x ) ) \ar[r]^(.4){{\partial}^{h^n } } & \smash{\underset{{x \in x^{(1)}}}{\textstyle\bigoplus } } h^{n-1}(f(x ) ) \ar[r]^(.55){{\partial}^{h^{n-1 } } } & \smash{\underset{{y \in x^{(2)}}}{\textstyle\bigoplus } } h^{n-2}(f(y ) ) \ar[r ] & \dotsm } \ ] ] where @xmath194 and @xmath195 is the homomorphism induced from the spectral sequence associated to the coniveau filtration , see bloch
ogus @xcite .
then we have that @xmath193 is a resolution of @xmath196 .
let @xmath105 be a regular integral @xmath9-scheme of finite krull dimension .
let @xmath197 denote the gersten
witt complex @xmath188 & w(f(x ) ) \ar[r]^(.44){{\partial}^w } & \smash{\underset{{x \in x^{(1)}}}{\textstyle\bigoplus } } w(f(x ) ) \ar[r]^{{\partial}^w } & \smash{\underset{{y \in x^{(2)}}}{\textstyle\bigoplus } } w(f(y ) ) \ar[r ] & \dotsm } \ ] ] where @xmath198 is the homomorphism induced from the second residue map for a set of choices of local parameters , see balmer
walter @xcite .
because of these choices , @xmath197 is only defined up to isomorphism , though there is a canonical complex defined in terms of witt groups of finite length modules over the local rings of points .
we have that @xmath199 .
the filtration of the gersten complex for witt groups induced by the fundamental filtration was first studied methodically by arason
jacob @xcite , see also parimala sridharan @xcite , gille @xcite , and fasel @xcite .
the differentials of the gersten complex for witt groups respect the fundamental filtration as follows : @xmath200 see fasel ( * ? ? ?
9.2.4 ) and gille @xcite .
thus for all @xmath201 we have complexes @xmath202 @xmath188 & i^{n}(f(x ) ) \ar[r]^(.4){{\partial}^{i^n } } & \smash{\underset{{x \in x^{(1)}}}{\textstyle\bigoplus } } i^{n-1}(f(x ) ) \ar[r]^(.55){{\partial}^{i^{n-1 } } } & \smash{\underset{{y \in x^{(2)}}}{\textstyle\bigoplus } } i^{n-2}(f(y ) ) \ar[r ] & \dotsm } \ ] ] which provide a filtration of @xmath197 in the category of complexes of abelian groups . here
we write @xmath203 for @xmath204 .
we have that @xmath205 .
the canonical inclusion @xmath206 respects the differentials , and so defines a cokernel complex @xmath207 @xmath208 & i^n / i^{n+1}(f(x ) ) \ar[r]&\smash{\underset{{x \in x^{(1)}}}{\textstyle\bigoplus } } i^{n-1}/i^n(f(x ) ) \ar[r]&\smash{\underset{{y \in x^{(2)}}}{\textstyle\bigoplus } } i^{n-2}/i^{n-1}(f(y ) ) \ar[r ] & \dotsm } \ ] ] see fasel ( * ? ? ?
9.2.10 ) , where @xmath209 for a field @xmath210 .
we have that @xmath211 the norm residue symbol for fields provides a morphism of complexes @xmath212 , where the map on terms of degree @xmath213 is @xmath214 . by the milnor conjecture for fields ,
this is an isomorphism of complexes . upon restriction
, we have an isomorphism @xmath215 . upon further restriction
, we have an injection @xmath216 .
the pfister form map for fields provides a morphism of complexes @xmath217 , where the map on terms of degree @xmath213 is @xmath218 . upon restriction
, we have a homomorphism @xmath219 .
see fasel ( * ? ? ?
* thm.10.2.6 ) . by the milnor conjecture for fields
, there exists a higher cohomological invariant morphism of complexes @xmath220 , where the map on terms of degree @xmath213 is @xmath221 . upon restriction ,
we have homomorphisms @xmath222 factoring through to @xmath223 . furthermore , on the level of complexes , the higher cohomological invariant morphisms factors through to a morphism of complexes @xmath224 , which by the milnor conjecture over fields , is an isomorphism .
upon restriction , we have isomorphisms @xmath225 . also see morel @xcite .
there is another important globalization of milnor @xmath11-theory and galois cohomology , but we only briefly mention it here .
conjectured to exist by belinson @xcite and lichtenbaum @xcite , and then constructed by voevodsky @xcite , motivic complexes modulo 2 give rise to zariski and tale motivic cohomology groups modulo 2 @xmath226 and @xmath227 . for a field @xmath9 ,
suslin @xcite and totaro @xcite establish a canonical isomorphism @xmath228 while the work of bloch , gabber , and suslin ( see the survey by geisser @xcite ) establishes an isomorphism @xmath229 ( for @xmath9 of characteristic @xmath10 ) . the natural pullback map @xmath230 induced from the change of site @xmath231 is then identified with the norm residue homomorphism .
thus @xmath232 and @xmath233 provide _ motivic _ globalizations of the mod 2 milnor @xmath11-theory and galois cohomology functors , respectively . on the other hand
, there does not seem to exist a direct motivic globalization of the witt group or its fundamental filtration .
let @xmath9 be a field of characteristic @xmath10 . summarizing the results cited in 2.22.3 , we have a commutative diagram of isomorphisms @xmath234^{h^{{\bullet } } } \ar[d]_{s^{{\bullet } } } & { \mathcal{h}}_{{\mathrm{\acute{e}t}}}^{{\bullet } } \\ { \mathcal{i}}^{{\bullet}}/{\mathcal{i}}^{{\bullet}+1 } \ar[ur]_{e^{{\bullet } } } & } \ ] ] of sheaves of graded abelian groups on @xmath91 .
in particular , we have such a commutative diagram of isomorphisms on the level of global sections . what we will consider as a globalization of the milnor conjecture
the _ unramified milnor question _ is a refinement of this .
* let @xmath105 be a smooth scheme over a field of characteristic @xmath10 .
then restricting @xmath235 gives rise to a homomorphism @xmath236 .
[ question ] let @xmath105 be a smooth scheme over a field of characteristic @xmath10 .
consider the following diagram : @xmath237 _ { ? } \ar@{^{(}->}[r]^(.54){i_k^{{\bullet } } } & { { \mathcal{k}}_{\text{m}}}^{{\bullet}}/2(x ) \ar[r]^{h_{{\mathrm{ur}}}^{{\bullet } } } \ar[d]_{s_{{\mathrm{ur}}}^{{\bullet } } } & { h_{{\mathrm{ur}}}}^{{\bullet}}(x,{\bm{\mu}}_2^{{\otimes}{\bullet } } ) \\
i_{{\mathrm{ur}}}^{{\bullet}}(x)/i_{{\mathrm{ur}}}^{{\bullet}+1}(x ) \ar@{^{(}->}[r]^(.54){i_i^{{\bullet } } } & { \mathcal{i}}^{n}/{\mathcal{i}}^{n+1}(x ) \ar[ur]_{e_{{\mathrm{ur}}}^{{\bullet } } } & } \ ] ] 1 . is the inclusion @xmath238 surjective ? 2 .
is the inclusion @xmath239 surjective ?
does the restriction of @xmath235 to @xmath240 have image contained in @xmath241 ?
if so , is it an isomorphism ?
note that in degree @xmath109 , questions [ question ] ( 1 ) , ( 2 ) , and ( 3 ) can be rephrased in terms of the obstruction groups , respectively : does @xmath242 vanish ; does @xmath243 vanish ; and does the restriction of @xmath244 yield a map @xmath245 and is it an isomorphism ? from now on we shall focus mainly on the unramified milnor question for quadratic forms ( i.e. question [ question](1 ) ) , which was already explicitly asked by parimala
sridharan ( * ? ? ?
* question q ) .
we mention a global globalization of the milnor conjecture for quadratic forms . because of the conditional definition of the global cohomological invariants , we restrict ourselves to the classical invariants on grothendieck witt groups defined in [ subsec : global_theories ] . [
question : gw ] let @xmath105 be a regular scheme with 2 invertible . for @xmath246 , consider the homomorphisms , @xmath247 induced from the ( classical ) cohomological invariants on grothendieck
witt groups .
are they surjective ?
this can be viewed as a globalization of merkurjev s theorem . indeed , first note that the cases @xmath248 of question [ question : gw ]
next , a consequence of lemma [ lem : gw ] is that @xmath249 is surjective if and only @xmath250 is surjective .
this , in turn , is a consequence of a positive answer to question [ question](1 ) for any @xmath105 satisfying weak purity for the witt group ( see [ subsec : purity ] for examples ) .
the globalization of the milnor conjecture for milnor @xmath11-theory using zariski and tale motivic cohomology modulo 2 ( see [ motivic ] ) is the @xmath251 modulo 2 case of the belinson
lichtenbaum conjecture : for a smooth variety @xmath105 over a field , the canonical map @xmath252 is an isomorphism for @xmath253 .
the combined work of suslin voevodsky @xcite and geisser
levine @xcite show the belinson
lichtenbaum conjecture to be a consequence of the bloch
kato conjecture , a proof of which has been announced by rost and voevodsky . in this section
we review some of the purity results ( see [ subsec : unramified_replacements ] ) relating the global and unramified witt groups and cohomology . for a survey on purity results for witt groups , see zainoulline @xcite .
purity for the global witt group means that the natural map @xmath254 is an isomorphism .
[ purity_witt ] let @xmath105 be a regular integral noetherian scheme with 2 invertible .
then purity holds for the global witt group functor under the following hypotheses : 1 .
@xmath105 is dimension @xmath255 , 2 .
@xmath105 is the spectrum of a regular local ring of dimension @xmath256 , 3 .
@xmath105 is the spectrum of a regular local ring containing a field . for part ( 1 ) , the case of dimension @xmath257 is due to colliot - thlne sansuc ( * ? ? ?
* cor.2.5 ) , the case of dimension @xmath258 and @xmath105 affine is due to ojanguren
suresh @xcite , and for the general case ( as well as ( 2 ) ) see balmer walter @xcite .
for ( 3 ) , see ojanguren panin @xcite . as a consequence , for any scheme over which purity for the witt group holds , the unramified milnor question for @xmath259 ( with @xmath260 ) is equivalent to the analogous global milnor question .
this is especially useful for the case of curves . for @xmath105
geometrically locally factorial and integral , the purity property holds for unramified cohomology in degree @xmath261 , i.e.@xmath262 see colliot - thlne sansuc ( * ? ? ?
3.2 , prop.4.1 ) . for @xmath105
smooth over a field of characteristic @xmath10 , we have a canonical identification @xmath263 by bloch
ogus @xcite such that the canonical map @xmath264 is the map arising from the kummer exact sequence already considered in the proof of lemma [ lem : gw ] .
we now survey some of the known positive cases of the unramified milnor question in the literature .
let @xmath100 be a local ring with infinite residue field of characteristic @xmath10 .
then the unramified milnor question ( all parts of question [ question ] ) has a positive answer over @xmath265 .
hoobler @xcite had already proved this in degree @xmath266 .
the following result was communicated to us by stefan gille ( who was inspired by totaro @xcite ) .
let @xmath105 be a proper smooth integral variety over a field of characteristic @xmath10 .
if @xmath105 is @xmath9-rational then the unramified milnor question ( all parts of question [ question ] ) has a positive answer over @xmath105 .
the groups @xmath267 , @xmath196 , and @xmath185 are birational invariants of smooth proper @xmath9-varieties .
to see this , one can use colliot - thlne ( * ? ? ? * prop .
2.1.8e ) and the fact that the these functors satisfy weak purity for regular local rings ( see theorem [ purity_witt ] ) .
another proof uses the fact that the complexes @xmath187 , @xmath193 , and @xmath202 are cycle modules in the sense of rost , see ( * ? ? ?
* cor.12.10 ) . in any case , by colliot - thlne ( * ? ? ?
2.1.9 ) the pullback induces isomorphisms @xmath268 ( first proved by milnor ( * ? ? ?
2.3 ) for @xmath269 ) , @xmath270 , and @xmath271 for all @xmath201 and @xmath272 .
in particular , @xmath273 and @xmath274 , and the theorem follows from the milnor conjecture over fields .
the following positive results are known for low dimensional schemes .
recall the notion of _ cohomological dimension _
@xmath275 of a field ( see serre ( * ? ? ?
* i 3.1 ) ) , _ virtual cohomological 2-dimension _ @xmath276 and their 2-primary versions . denoting by @xmath277 any of these notions of dimension , note that if @xmath278 and @xmath279 then @xmath280 .
[ thm : curves ] let @xmath105 be a smooth integral curve over a field @xmath9 of characteristic @xmath10 .
then the unramified milnor question for quadratic forms ( question [ question](1 ) ) has a positive answer over @xmath105 in the following cases : 1 .
@xmath281 , 2 .
@xmath282 , 3 .
@xmath283 and @xmath105 is affine , 4 .
@xmath284 and @xmath105 is affine . for ( 1 )
, this follows from parimala sridharan ( * ? ? ?
* lemma 4.1 ) and the fact that @xmath32 is always surjective . for ( 2 ) ,
the case @xmath285 ( i.e. @xmath9 is real closed ) is contained in monnier ( * ? ? ? * cor .
3.2 ) and the case @xmath286 follows from a straightforward generalization to real closed fields of the results in @xcite for the real numbers . for ( 3 ) , see ( * ? ? ?
* lemma 4.2 ) . for ( 4 ) , the statement follows from a generalization of ( * ? ? ?
* thm.6.1 ) .
we wonder whether @xmath287 can be replaced by @xmath288 in theorem [ thm : curves ] .
parimala sridharan ( * ? ? ?
* rem.4 ) ask whether there exist affine curves ( over a well - chosen field ) over which the unramified milnor question has a negative answer . for surfaces ,
there are positive results are in the case of @xmath289 .
if @xmath9 is algebraically closed , then the unramified milnor question for quadratic forms ( question [ question](1 ) ) has a positive answer by a direct computation , see fernndez - carmena @xcite .
if @xmath9 is real closed , one has the following result .
let @xmath105 be smooth integral surface over a real closed field @xmath9 . if the number of connected components of @xmath290 is @xmath261 ( i.e. in particular if @xmath291 ) , then the unramified milnor question for quadratic forms ( question [ question](1 ) ) has a positive answer over @xmath105 .
examples of surfaces with many connected components over a real closed field , and over which the unramified milnor question still has a positive answer , are also given in monnier @xcite .
finally , as a consequence of ( * ? ? ?
3.4 ) , the unramified milnor question for quadratic forms ( question [ question](1 ) ) has a positive answer over any scheme @xmath105 satisfying : @xmath292 is generated by quaternion azumaya algebras ( i.e. index@xmath293period for 2-torsion classes ) ; or @xmath292 is generated by azumaya algebras of degree dividing 4 ( i.e.index@xmath293period@xmath294 for 2-torsion classes ) and @xmath295 is 2-divisible . in particular
, this recovers the known cases of curves over finite fields ( by class field theory ) and surfaces over algebraically closed fields ( by de jong @xcite ) .
alex hahn asked if there exists a ring @xmath100 over which the global merkurjev question ( question [ question : gw ] ) has a negative answer , i.e. @xmath296 is not surjective .
the results of parimala , scharlau , and sridharan @xcite , @xcite , @xcite , show that there exist smooth complete curves @xmath105 ( over @xmath1-adic fields @xmath9 ) over which the unramified milnor question ( question [ question](1 ) ) in degree 2 ( and hence , by purity , the global merkurjev question ) has a negative answer .
[ rem : wrong ] the assertion ( in gille @xcite and pardon @xcite ) that the unramified milnor question ( question [ question](1 ) ) has a positive answer over any smooth scheme ( over a field of characteristic @xmath10 ) is incorrect . in these texts , the distinction between the groups @xmath297 and @xmath298 is not made clear .
a scheme @xmath105 over a field @xmath9 has the _ extension property _ for quadratic forms if there exists @xmath299 such that every regular quadratic form on @xmath300 extends to @xmath105 .
let @xmath9 be a field of characteristic @xmath10 and with @xmath301 and @xmath105 a smooth integral @xmath9-curve .
then the unramified milnor question for quadratic forms ( question [ question](1 ) ) has a positive answer for @xmath105 if and only if @xmath105 has the extension property .
the extension property is guaranteed when a residue theorem holds for the witt group .
the reside theorem for @xmath302 is due to milnor @xcite . for nonrational curves ,
the choice of local uniformizers inherent in defining the residue maps is eliminated by considering quadratic forms with values in the canonical bundle @xmath303 .
let @xmath105 be a scheme and @xmath304 an invertible @xmath107-module .
an _ ( @xmath304-valued ) symmetric bilinear form _ on @xmath105 is a triple @xmath305 , where @xmath110 is a locally free @xmath107-module of finite rank and @xmath306 is an @xmath107-module morphism .
let @xmath105 be a smooth proper integral curve over a field @xmath9 of characteristic @xmath10 . then there is a canonical complex ( which is exact at the first two terms ) @xmath307 & w(x,\omega_{x / f } ) \ar[r ] & w(f(x),\omega_{f(x)/f } ) \ar[r]^(.45){{\partial}^{\omega_x } } & \smash{\underset{{x \in x^{(1)}}}{\textstyle\bigoplus } } w(f(x),\omega_{f(x)/f } ) \ar[r]^(.72){\operatorname{\mathrm{tr}}_{x / f } } & w(f ) } \ ] ] and thus in particular @xmath308 has a residue theorem .
now any choice of isomorphism @xmath309 , induces a group isomorphism @xmath310 via @xmath311 .
thus in particular , if @xmath303 is a square in @xmath295 , then @xmath105 has the extension property .
conversely : [ thm : nonsurjectivity ] let @xmath9 be a local field of characteristic @xmath10 and @xmath105 a smooth integral hyperelliptic @xmath9-curve of genus @xmath312 with @xmath313 .
then the unramified milnor question for quadratic forms ( question [ question](1 ) ) holds over @xmath105 if and only if @xmath314 is a square .
let @xmath105 be the smooth proper hyperelliptic curve over @xmath315 with affine model @xmath316 .
one can show using ( * ? ? ?
2.4 ) that @xmath303 is not a square .
the point @xmath317 defines a @xmath315-rational point of @xmath105 .
hence by theorem [ thm : nonsurjectivity ] , the unramified milnor question has a negative answer over @xmath105 .
note that possible counter examples which are surfaces could be extracted from the following result .
let @xmath105 be a smooth integral surface over a real closed field @xmath9 .
then the unramified milnor question for quadratic forms ( question [ question](1 ) ) has a positive answer over @xmath105 if and only if the cokernel of the mod 2 signature homomorphism is 4-torsion .
let @xmath105 be a smooth @xmath9-scheme .
let @xmath318 be the witt group of @xmath304-valued symmetric bilinear forms on @xmath105 and @xmath319 the _ total _ witt group , where the sum is taken over a set of representatives of @xmath320 . while this group is only defined up to non - canonical isomorphism depending on our choice of representatives , none of our cohomological invariants depend on such isomorphisms , see @xcite .
furthermore , we will not consider any ring structure on this group .
thus we will not need to descend into most of the important considerations of balmer
calms @xcite .
let @xmath321 be the usual rank modulo 2 map @xmath322 and @xmath323 its kernel .
then the signed discriminant ( see @xcite ) defines a surjective map @xmath324 finally , denote by @xmath325 the subgroup generated by forms of trivial arf invariant and @xmath326 .
then there exists a _
total clifford invariant _ for line bundle - valued quadratic forms @xmath327 defined in ( * ? ? ?
the surjectivity of the total clifford invariant can be viewed as a version of the global merkurjev question ( question [ question : gw ] ) for line bundle - valued quadratic forms .
let @xmath105 be a smooth proper integral curve over a local field @xmath9 of characteristic @xmath10 or a smooth proper integral surface over a finite field @xmath9 of characteristic @xmath10 .
then the total clifford invariant @xmath328 is surjective .
the surjectivity of the total clifford invariant can also be reinterpreted as the statement that while not every class in @xmath329 is represented by a quadratic form on @xmath105 , every class is represented by a line bundle - valued quadratic form on @xmath105 .
j. arason , _
quadratic forms and galois - cohomology _ , sminaire de thorie des nombres , 19721973 ( univ .
bordeaux i , talence ) , exp .
thorie des nombres , centre nat .
recherche sci . ,
talence , 1973 , p. 5 .
arason , richard elman , and bill jacob , _ the graded witt ring and galois cohomology .
i _ , quadratic and hermitian forms ( hamilton , ont . , 1983 ) ,
cms conf . proc . ,
vol . 4 , amer .
soc . , providence , ri , 1984 , pp .
to3em , _ on the witt ring of elliptic curves _ , @xmath11-theory and algebraic geometry : connections with quadratic forms and division algebras ( santa barbara , ca , 1992 ) , proc . sympos .
pure math .
58 , amer .
soc . , providence , ri , 1995 , pp .
125 .
h. bass and j. tate , _ the milnor ring of a global field _ , algebraic @xmath11-theory , ii : `` classical '' algebraic @xmath11-theory and connections with arithmetic ( proc .
seattle , wash . , battelle memorial inst . , 1972 ) ,
springer , berlin , 1973 , pp .
lecture notes in math .
, vol . 342 .
colliot - thlne , _ birational invariants , purity and the gersten conjecture _ , @xmath11-theory and algebraic geometry : connections with quadratic forms and division algebras ( santa barbara , ca , 1992 ) , proc . sympos . pure math .
58 , amer .
soc . , providence , ri , 1995 , pp .
164 .
jean - louis colliot - thlne , raymond t. hoobler , and bruno kahn , _ the bloch - ogus - gabber theorem _ , algebraic @xmath11-theory ( toronto , on , 1996 ) , fields inst .
16 , amer .
soc . , providence , ri , 1997 , pp .
3194 .
alexander grothendieck , _ dix exposs sur la cohomologie des schmas _ , ch .
viii classes de chern et reprsentations linaires des groupes discrets , pp . 215305 , north - holland publishing company , amsterdam ; masson & cie , diteur , paris , 1968 .
m. knebusch , _ symmetric bilinear forms over algebraic varieties _ , conference on quadratic forms1976 ( kingston , ont . ) ( g. orzech , ed . ) , queen s papers in pure and appl .
, no . 46 ,
queen s univ . , 1977 ,
103283 .
alexander merkurjev , _ on the norm residue homomorphism of degree two _ , proceedings of the st .
petersburg mathematical society .
xii ( providence , ri ) , amer .
2 , vol . 219 , amer .
soc . , 2006 , pp . 103124 . m. ojanguren , r. parimala , r. sridharan , and v. suresh , _
witt groups of the punctured spectrum of a 3-dimensional regular local ring and a purity theorem _ , j. london math .
( 2 ) * 59 * ( 1999 ) , no . 2 , 521540 . r. parimala and w. scharlau , _ on the canonical class of a curve and the extension property for quadratic forms
, recent advances in real algebraic geometry and quadratic forms ( berkeley , ca , 1990/1991 ; san francisco , ca , 1991 ) , contemp . math .
155 , amer .
soc . , providence , ri , 1994 , pp .
339350 .
to3em , _ some remarks on the historical development of the algebraic theory of quadratic forms _ , quadratic and hermitian forms ( hamilton , ont . , 1983 ) ,
cms conf . proc .
, vol . 4 , amer .
soc . , providence , ri , 1984 , pp .
h. g. quebbemann , r. scharlau , w. scharlau , and m. schulte , _ quadratische formen in additiven kategorien _ , bull .
france suppl .
( 1976 ) , no . 48 , 93101 , colloque sur les formes quadratiques ( montpellier , 1975 ) .
andrei suslin and vladimir voevodsky , _ bloch - kato conjecture and motivic cohomology with finite coefficients _ , the arithmetic and geometry of algebraic cycles ( banff , ab , 1998 ) , nato sci .
ser . c math .
, vol . 548 , kluwer acad .
, dordrecht , 2000 , pp .
117189 .
adrian r. wadsworth , _ merkurjev s elementary proof of merkurjev s theorem _ , applications of algebraic @xmath11-theory to algebraic geometry and number theory , part i , ii ( boulder , colo . , 1983 ) , contemp .
55 , amer .
soc . , providence , ri , 1986 , pp . | the milnor conjecture has been a driving force in the theory of quadratic forms over fields , guiding the development of the theory of cohomological invariants , ushering in the theory of motivic cohomology , and touching on questions ranging from sums of squares to the structure of absolute galois groups . here
, we survey some recent work on generalizations of the milnor conjecture to the context of schemes ( mostly smooth varieties over fields of characteristic @xmath0 ) .
surprisingly , a version of the milnor conjecture fails to hold for certain smooth complete @xmath1-adic curves with no rational theta characteristic ( this is the work of parimala , scharlau , and sridharan ) .
we explain how these examples fit into the larger context of the unramified milnor question , offer a new approach to the question , and discuss new results in the case of curves over local fields and surfaces over finite fields .
the first cases of the ( as of yet unnamed ) milnor conjecture were studied in pfister s habilitationsschrift @xcite in the mid 1960s . as pfister @xcite himself points out , ``
[ the milnor conjecture ] stimulated research for quite some time . '' indeed , it can be seen as one of the driving forces in the theory of quadratic forms since milnor s original formulation @xcite in the early 1970s .
the classical cohomological invariants of quadratic forms ( rank , discriminant , and clifford hasse
witt invariant ) have a deep connection with the history and development of the subject .
in particular , they are used in the classification ( hasse minkowski local - global theorem ) of quadratic forms over local and global fields .
the first `` higher invariant '' was described in arason s thesis @xcite , @xcite .
the celebrated results of merkurjev @xcite and merkurjev
suslin @xcite settled special cases of the milnor conjecture in the early 1980s , and served as a starting point for voevodsky s development of the theory of motivic cohomology .
other special cases were settled by arason elman
jacob @xcite and jacob
rost @xcite .
voevodsky s motivic cohomology techniques @xcite ultimately led to a complete solution of the milnor conjecture , for which he was awarded the 2002 fields medal .
the consideration of quadratic forms over rings ( more general than fields ) has its roots in the number theoretic study of lattices ( i.e.quadratic forms over @xmath2 ) by gauss as well as the algebraic study of division algebras and hermitian forms ( i.e. quadratic forms over algebras with involution ) by albert .
a general framework for the study of quadratic forms over rings was established by bass @xcite , with the case of ( semi)local rings treated in depth by baeza @xcite .
bilinear forms over dedekind domains ( i.e. unimodular lattices ) were studied in a number theoretic context by frhlich @xcite , while the consideration of quadratic forms over algebraic curves ( and their function fields ) was initiated by geyer , harder , knebusch , scharlau @xcite , @xcite , @xcite , @xcite .
the theory of quadratic ( and bilinear ) forms over schemes was developed by knebusch @xcite , @xcite , and utilized by arason @xcite , dietel @xcite , parimala @xcite , @xcite , fernndez - carmena @xcite , sujatha @xcite , @xcite , arason
elman
jacob @xcite , @xcite , and others .
a theory of symmetric bilinear forms in additive and abelian categories was developed by quebbemann
scharlau
schulte @xcite , @xcite .
further enrichment came eventually from the triangulated category techniques of balmer @xcite , @xcite , @xcite , and walter @xcite .
this article will focus on progress in generalizing the milnor conjecture to these contexts .
these remarks grew out of a lecture at the rims - camp - style seminar `` galois - theoretic arithmetic geometry '' held october 19 - 23 , 2010 , in kyoto , japan .
the author would like to thank the organizers for their wonderful hospitality during that time .
he would also like to thank stefan gille , moritz kerz , r. parimala , and v. suresh for many useful conversations .
the author acknowledges the generous support of the max plank institute for mathematics in bonn , germany where this article was written under excellent working conditions .
this author is also partially supported by national science foundation grant msprf dms-0903039 .
a graded abelian group or ring @xmath3 will be denoted by @xmath4 .
if @xmath5 is a decreasing filtration of a ring @xmath6 by ideals , denote by @xmath7 the associated graded ring .
denote by @xmath8 the elements of order 2 in an abelian group @xmath6 .
all abelian groups will be written additively . |
spacecraft sent to mars in recent years have revealed new evidence suggesting that the environmental conditions on early mars were very different than those today , with liquid water flowing on the surface at least episodically .
geomorphological evidence of past water flow in ancient terrains includes valley networks ( remnant of widespread fluvial activity ) ( carr , 1996 ) , extensive sedimentary layered deposits ( including delta - like landforms)(malin and edgett , 2000 , 2003 ) and erosion rate much higher than today affecting the most ancient landforms ( craddock and maxwell , 1993 ) .
satellite remote sensing and in - situ analysis of the surface mineralogy have also revealed the local presence of minerals that require liquid water for their formation : clay / phyllosilicates ( poulet et al . 2005 , bibring et al .
2006 , ehlmann et al .
2011 ) , sulfates ( gendrin et al .
2005 , squyres et al .
2004 ) , opaline silica ( squyres et al . 2008 ) , carbonate ( ehlmann et al .
2008 , boynton et al .
2009 , morris et al . 2010 ) and chloride ( osterloo et al . , 2008 ) .
all these observations have provided a wealth of information about early mars . in particular , they suggest that the conditions strongly varied throughout the noachian and hesperian era
. nevertheless , we still do not know if the conditions suitable for liquid water were stable on long timescales , or if they were the consequence of episodic , possibly catastrophic events .
what is clear is that the conditions which allowed such extensive alteration by liquid water occurred early in the history of mars and not later . in particular
, the fact that some alteration minerals can still be observed today , although they formed several billion years ago and are easily transformed into other materials by , e.g. , diagenesis , suggests that water has been very limited on the martian surface from soon after their formation until today ( tosca and knoll , 2009 ) ) .
similarly the majority of the valley networks are found almost exclusively in the highly cratered ancient southern highlands , which date back from the end of the heavy bombardment period some 3.5 - 3.8 gyr ago ( fasset and head , 2008 ) . although the role of hydrothermalism ( resulting from volcanism or impacts ) with respect to atmospheric induced processes is not clear , climate conditions were certainly different on early mars .
it is possible that the atmosphere could have been thicker than today , providing an environment more suitable for liquid water thanks to a surface pressure well above the triple point of water , and possibly a greenhouse effect capable of warming the surface closer to the melting points at 0@xmath3c . in analogy with earth and venus , even after taking into account the differences in size , it can be estimated that the initial inventory of volatiles on mars included at least several bars of atmospheric gases , mostly co@xmath0 ( see haberle et al . 1998 ) . on this basis , since the 1980s many studies have been performed to characterize the possible climate on early mars assuming a co@xmath0 atmosphere thicker than today , and taking into account the fact that according to stellar evolution models the sun s luminosity at 3.8 ga was 75% of its present value ( e.g. , gough , 1981 ) .
almost all of these studies were performed using one - dimensional ( 1-d ) radiative convective models ( see section [ sc : previous ] ) .
characterization has been challenging however , because of the complex co@xmath0 gas spectroscopy , the likely formation of co@xmath0 ice clouds in the atmosphere , and the four - dimensional aspects of planetary climate .
the primary goal of the present paper is to update these calculations using a full 3-d climate model including a parameterization of co@xmath0 ice clouds and state of the art spectroscopic data . in a companion paper ( wordsworth et al . , 2012 ) we extend these calculations by including a model of the possible water cycle that takes into account the radiative effects of water vapor and clouds , and can predict precipitation and the formation of lakes and glaciers on early mars . to start with , in section [ sc : previous ] and [ sc : which_atm ] we review previous work on modelling early mars climate , and briefly discuss the possible early mars atmosphere composition and thickness . section [ sc : model ] describes our global climate model ( gcm ) .
the simulated climates for various pressures , cloud microphysics parameters , obliquity values , orbital parameters , and possible atmospheric dust loading scenarios are analysed in section [ sc : results ] .
finally , in section [ sc : discussion ] , we discuss and summarise our results .
the climate modelling studies of early mars performed before 1998 are reviewed in detail in haberle ( 1998 ) . by the end of the 1980s ,
the paradigm for early martian climate was based on the results of pollack et al .
( 1987 ) . using a 1-d radiative convective model ,
they had shown that a 5-bar gaseous co@xmath0 atmosphere would raise the global mean surface temperature to 0@xmath3c , allowing `` warm and wet '' conditions ( though large , 5 bars of co@xmath0 was still consistent with contempory estimates of the available inventory ) .
this scenario was later challenged when kasting ( 1991 ) published his reanalysis of the pollack et al .
( 1987 ) greenhouse calculations .
kasting used the same model , but took into account the fact that at higher pressures co@xmath0 can condense in the middle atmosphere .
this seriously decreases the greenhouse effect because of the latent heat warming of the middle atmosphere .
furthermore kasting suggested that the resulting co@xmath0 ice clouds would probably raise the planetary albedo and further cool the planet .
however , still working with the same model , forget and pierrehumbert ( 1997 ) later showed that the inclusion of the co@xmath0 ice cloud radiative effect at both solar and thermal infrared wavelengths led to a net warming of the surface rather than a cooling . indeed , assuming that the co@xmath0 ice cloud particles are larger than a few micrometers
, they can readily scatter infrared radiation and reflect outgoing thermal radiation back to the surface . the resulting warming effect more than compensates for the albedo increase due to the clouds .
these calculations were later confirmed by mischna et al .
( 2000 ) , who used a more sophisticated and accurate 1-d radiative - convective transfer model based on correlated k - distribution methods .
they noted that co@xmath0 ice clouds could also cool the surface if they are low and optically thick , and concluded that `` estimating the actual effect of co@xmath0 clouds on early martian climate will require three - dimensional models in which cloud location , height , and optical depth , as well as surface temperature and pressure , are determined self - consistently '' .
nevertheless , the physics of carbon dioxide clouds in a dense co@xmath0 atmosphere was further explored by colaprete and toon ( 2003 ) still using a 1-d atmospheric model , but including a detailed microphysical cloud model .
they took into account laboratory studies by glandorf et al .
( 2002 ) showing that high critical supersaturations are required for co@xmath0 cloud particle nucleation and that surface kinetic growth is not limited .
under such conditions , colaprete and toon ( 2003 ) predicted large carbon dioxide ice particles with radii greater than 500 @xmath4 m , and thus low cloud opacities . because of this , and because of the warming of the atmosphere associated with cloud formation , they estimated that the greenhouse effect of co@xmath0 clouds would probably be limited to 5 - 10 k surface warming .
all these studies took into account the radiative effect of co@xmath0 and water vapor .
postawko and kuhn ( 1986 ) also explored the possible greenhouse effect of so@xmath0 .
yung et al .
( 1997 ) resumed this investigation , and studied the radiative effect of a very small amount ( 0.1 ppmv ) of so@xmath0 in a 2 bar co@xmath0 atmosphere in a 1d model .
they showed that it would raise the temperature of the middle atmosphere by approximately 10 k , so that the upper atmosphere would globally remain above the condensation temperature of co@xmath0 .
exploration of the impact of so@xmath0 was further motivated by the discovery of sulfate sediments in mars ancient terrains ( gendrin et al .
2005 , squyres et al . 2005 , halevy et al .
johnson et al . ( 2008 ) investigated the impact of larger amount of sulfur volatiles ( h@xmath0s and so@xmath0 mixing ratios of 1 to 1000 ppmv ) in a martian atmosphere of 50 and 500 mbar of co@xmath0 ( with h@xmath0o ) .
for this purpose they used the 3d mars weather research and forecasting ( marswrf ) gcm ( richardson et al . , 2007 ) .
their control simulations , achieved with a pure co@xmath0 atmosphere or with co@xmath0 + h@xmath0o ( but neglecting the radiative effects of co@xmath0 clouds or h@xmath0o clouds ) are of interest in the present paper for comparison .
calculations performed including the sulfur volatile influxes suggested that these gases could have been responsible for greenhouse warming up to 25 k above that caused by co@xmath0 .
however , tian et al . ( 2010 ) showed that such large amounts of so@xmath0 would inevitably lead to the formation of sulfate and large reduced sulfur aerosols ( s@xmath5 ) .
they concluded that these aerosols would have raised the planetary albedo and that the resulting cooling would more than outweigh the gaseous greenhouse effect . in this paper
, we do not include the effect of sulfur volatiles or aerosols .
our goal is to investigate the details of the climate induced by a co@xmath0 atmosphere as simulated by a 3d gcm in which co@xmath0 cloud location , height , and optical depth , as well as surface temperature and pressure , are determined self - consistently for the first time . in comparison to the previous studies mentioned above
, we also benefit from an improved parameterization of the collision - induced absorption of co@xmath0 ( important for pressure larger than a fraction of a bar ) that we have developed for this project , as described in wordsworth et al .
( 2010 ) ( see details in section [ sc : cia ] ) . to calculate the collision - induced opacity in a co@xmath0 atmosphere , all the studies mentioned above relied without modification on a parameterization originally derived for the venus atmosphere by james pollack ( pollack et al . , 1980
; see detailed description in kasting et al .
it was based on the measurements of ho et al .
( 1971 ) from 7 to 250 @xmath6 , and on a simple parameterization of collision induced opacity in the other spectral domains described in an unpublished phd thesis by john f. moore ( moore , 1971 ) , still available as a nasa report ) .
this parameterization included significant opacities between 526 and 295 @xmath6 resulting from the pressure induced wings of the strong 15 @xmath4 m bands .
unfortunately , this feature was kept in subsequent models , despite the fact that these opacities were probably overestimated and , in most cases , already accounted for in the codes chosen to calculate the radiative transfer in the 15 @xmath4 m band . as a result
, most previous studies probably overestimated the greenhouse warming of co@xmath0 by several kelvins .
by scaling the earth or venus volatile inventory , one can estimate that the amount of co@xmath0 brought to mars during accretion was probably larger than 10 bars
. however , tian et al . (
2009 ) showed that the extreme ultraviolet flux from the young sun was so high that it would have strongly warmed the thermosphere and induced thermal escape of a primordial co@xmath0-dominated martian atmosphere . in their calculation
, a co@xmath0 atmosphere could not have been maintained until about 4.1 billion years ago . under such conditions ,
how much atmosphere could have been degassed late enough after 4.1 ga during the mid to late noachian ?
phillips et al .
( 2001 ) estimated that the total release of gases from the magmas that formed tharsis during the noachian era could have produced the integrated equivalent of a 1.5-bar co@xmath0 atmosphere .
however , they assumed a magmatic co@xmath0 content of 0.65 weight percent , consistent with hawaiian basaltic lavas .
since then , several authors have suggested that this was probably overestimated . in particular , hirschmann and withers ( 2008 ) calculated that post-4.5 ga magmatism could have provided only 0.1 to at most 1 bar of co@xmath0 . to refine these conclusions , grott et al .
( 2011 ) combined the hirschmann and withers ( 2008 ) model for the solubility of co@xmath0 with a thermo - chemical evolution model to self consistently calculate the dissolved amount of co@xmath0 in martian magmas .
they estimated that during noachian , about 250 mbar of co@xmath0 were outgassed between 4.1 and 3.7 ga . in spite of these studies ,
if one assumes that more than one bar of atmospheric co@xmath0 was present during the noachian era , where did it go ? estimating the amount of atmosphere that could have escaped to space in the last 4 gyr is difficult because many different processes may have been involved , including photochemical escape , ions dragged by the solar wind field , sputtering , and impact erosion ( chassefiere and leblanc , 2004 ) . up to now , only the present - day mars ion loss by solar wind interaction has been measured , and found to be very small ( barabash et al . , 2007 ) .
the possible loss of volatiles resulting from large impacts ( melosh and vickery 1989 ) is also difficult to model and constrain , but recent studies suggest that it must have been small ( pham et al . , 2011 ) .
altogether , the most recent estimations of the amount of co@xmath0 lost to space in the past 3.5 gyr are below a few hundreds of millibars ( leblanc and johnson , 2002 , lammer et al .
2008 , 2010 ) .
alternatively , one classical hypothesis is to assume that , as on earth , large amounts of co@xmath0 could be stored in form of carbonates in the martian crust after chemical precipitation .
however , almost no carbonates were initially detected by the omega imaging spectrometer in spite of its high sensitivity to the spectral signature of carbonates ( bibring et al . , 2005 ) .
recently , several observations from orbiters ( ehlmann et al . , 2008 ) and landers ( boynton et al .
2009 , morris et al .
2010 ) have revived the carbonate hypothesis and reasserted the importance of carbon dioxide in martian climate history ( harvey , 2010 ) . in this paper ,
we have chosen to explore the possible climate on early mars for a wide range of surface pressures , up to 7 bars , as assumed in previous works on the same topic . however
, when interpreting our model results , it is important to keep in mind that atmospheres thicker than one bar may be unlikely .
another issue for an early mars co@xmath0 atmosphere is its photochemical stability . using a 1-d photochemical model of the martian atmosphere , zahnle et al .
( 2008 ) showed that in a thick , cold and dry co@xmath0 atmosphere in which a surface sink is assumed for reactive oxidized gases ( like h@xmath0o@xmath0 and o@xmath7 ) , co@xmath0 would tend to be reduced into co. they noted that the process is very slow , and that `` co@xmath0 atmospheres can be unstable but persistent simply because there is nt time enough to destroy them '' . finally , in this work we did not take into account the presence of other possible greenhouse gases .
such gases are not expected to be photochemically stable since they should have been photodissociated or oxydized , but they may be present if geochemical sources were active enough .
this is further discussed in section [ sc : othergases ] .
we have used a new generic " version of the lmd global climate model recently developed to simulate any kind of atmosphere with the goal of studying early climates in the solar system ( this paper ) as well as possible climates on extrasolar planets ( e.g. wordsworth et al .
in practice , the model is derived from the lmd present - day mars gcm ( forget et al .
1999 ) , with several new parameterizations ( see below and wordsworth et al . ( 2012 ) ) .
this mars gcm has been used successfully to simulate mars meteorology ( e.g. forget et al .
1998 , montmessin et al .
2004 , madeleine et al . 2011 ) and photochemistry ( lefevre et al . 2004 , 2008 ) from the surface to the thermosphere , and to simulate recent climate changes induced by the oscillations of mars rotational and orbital parameters ( levrard et al .
2004 , forget et al .
2006 , montmessin et al .
2007a , madeleine et al . 2009 ) .
the lmd gcm solves the primitive equations of meteorology using a finite difference dynamical core on an arakawa c grid .
this dynamical core has been used and tested successfully in many kind of atmospheres thicker than present - day mars such as the earth ( e.g. hourdin et al .
2004 ) , venus ( lebonnois et al . 2010 ) , and titan ( hourdin et al . 1995 , lebonnois et al . 2012 ) . in this paper ,
simulations were performed with two horizontal resolutions : 32@xmath824 ( corresponding to resolutions of 7.5@xmath3 latitude by 11.25@xmath3 longitude ) and 64@xmath848 ( 3.75@xmath3 latitude by 5.625@xmath3 longitude ) .
in the vertical , the model uses hybrid coordinates , that is , a terrain - following @xmath9 coordinate system near the surface and lower atmosphere ( @xmath9 is pressure divided by surface pressure ) , and pressure levels in the upper atmosphere . in this work , with the exception of one sensitivity study described in section [ sc : cloud ]
, we used 15 layers with the lowest mid - layer levels at about 18 m , 60 m , 150 m , 330 m , 640 m etc . , and the top level at 0.3% of the surface pressure , that is about 5.3 scale heights ( @xmath1050 km ) above the zero datum .
nonlinear interactions between explicitly resolved scales and subgrid - scale processes are parameterized by applying a scale - selective horizontal dissipation operator based on an @xmath11 time iterated laplacian @xmath12 .
this can be written @xmath13^{n}/ { \tau_{\mbox{\scriptsize diss } } } ) ( \delta x)^{2n } \delta^{n } q$ ] where @xmath14 is the smallest horizontal distance represented in the model , @xmath15 is the dissipation timescale for a structure of scale @xmath14 , and @xmath16 a variable like temperature , meridional wind , or zonal wind . as in most gcms , this dissipation operator is necessary to ensure the numerical stability of the dynamical core .
however , it must be used with moderation ( i.e. the dissipation timescales must be kept as high as possible ) . in particular , during this study we found that , when used with large topography variations and with atmospheric pressure larger than about one bar , our dissipation scheme tended to artificially produce some heat in the lower atmosphere .
this production is usually completely negligible , but with high pressure and a strong greenhouse effect like in some of our simulations we found that the impact on surface temperatures could be non - negligible if the dissipation timescale were chosen smaller than necessary .
this meant that our initial results overestimated the warming possible due to co@xmath0 clouds in the early martian atmosphere , and hence the likelihood of warm , wet conditions under a pure co@xmath0 atmosphere ( wordsworth et al .
fortunately , we have been able to identify this issue , reduce the dissipation , and demonstrate that this problem does not affect the energy balance of the results presented in this paper .
subgrid - scale dynamical processes including turbulent mixing and convection are parameterized as in forget et al .
( 1999 ) . in practice ,
the boundary layer dynamics are accounted for by mellor and yamada s ( 1982 ) unstationary 2.5-level closure scheme plus a `` convective adjustment '' which rapidly mixes the atmosphere in the case of unstable temperature profiles .
turbulence and convection mixes energy ( potential temperature ) , momentum ( wind ) , and tracers ( gases and aerosols ) .
the subgrid - scale orography and gravity wave drag schemes of the present - day mars gcm ( forget et al .
1999 ) were not applied .
surface temperature evolution is governed by the balance between radiative and sensible heat fluxes ( direct solar insolation , thermal radiation from the atmosphere and the surface , and turbulent fluxes ) and thermal conduction in the soil .
the parameterization of this last process was based on an 18-layer soil model solving the heat diffusion equation using finite differences .
the depth of the layers were chosen to capture diurnal thermal waves as well as the deeper annual thermal wave .
vertically homogeneous soil was assumed . for most simulations ,
the thermal inertia was set to 250 j s@xmath17 m@xmath18 k@xmath19 everywhere ( a value slightly higher than the mean value on present - day mars , to account for the higher gaseous pressure in the pore space , which increases the soil conductivity ) . for the ground albedo , as well as for the topography
, we chose to keep the same distributions as observed on mars today .
both fields may have been quite different three billion years ago , but we assume that the present - day values can be representative of the range and variability of the noachian - hesperian eras . in section [ sc : tsurf ] , the sensitivity of our results to surface thermal inertia and albedo is discussed .
our radiative scheme is based on the correlated - k model , with the absorption data calculated directly from high resolution spectra computed by a line - by - line model from the hitran 2008 database ( rothman et al . , 2009 ) .
these were then converted to correlated - k coeficients for use in the radiative transfer calculations . in practice , at a given pressure and temperature , correlated - k coefficients in the gcm are interpolated from a matrix of coefficients stored in a @xmath20 temperature and log - pressure grid : @xmath21 k , @xmath22 , 10@xmath23 , 10@xmath24 , .... , 10@xmath25 pa . we used 32 spectral bands in the thermal infrared and 36 at solar wavelengths .
sixteen points were used for the g - space integration , where g is the cumulated distribution function of the absorption data for each band .
rayleigh scattering by co@xmath0 molecules was included using the method described in hansen and travis ( 1974 ) , and using the toon et al .
( 1989 ) scheme to compute the radiative transfer . as mentioned in section [ sc : previous ] , a key improvement of our radiative transfer model compared to previous models is the use of an improved parameterization for the co@xmath0 collision - induced absorption ( cia ) .
it was specially developed for the present study using the results of baranov et al .
( 2004 ) and gruszka and borysow ( 1998 ) .
the method is described and justified in wordsworth et al .
the sublorentzian profiles of perrin and hartmann ( 1989 ) were used for the co@xmath0 far line absorption .
the model includes a parameterization to account for the possible condensation of co@xmath0 when temperatures fall below the condensation temperature @xmath26 . to compute @xmath26 as a function of pressure @xmath27 ( pa )
, we used the expression provided by fanale et al .
( 1982 ) : @xmath28\ ] ] for @xmath29 pa , @xmath30 for @xmath31 pa ( 518000 pa is the pressure of the triple point of co@xmath0 ) .
as on mars today , co@xmath0 can directly form on the ground .
when the surface temperature falls below the condensation temperature , co@xmath0 condenses , releasing the latent heat required to keep the solid - gas interface at the condensation temperature .
conversely , when co@xmath0 ice is heated , it partially sublimates to keep its temperature at the frost point . when co@xmath0 ice is present ,
the surface albedo and emissivity are set to 0.5 and 0.85 respectively .
on present - day mars , this albedo varies a lot in space and time .
0.5 is consistent with the various estimations of the average co@xmath0 ice cap albedo used to fit the seasonal pressure variations induced by the condensation and sublimation in the polar caps ( forget et al .
1998 , haberle et al . 2008 ) in our simulations , co@xmath0 ice clouds also form in the atmosphere .
such clouds have been observed on present - day mars during the polar night in the lower atmosphere ( see e.g. pettengill and ford , 2000 , tobie et al . 2003 ) and in the equatorial mesosphere around 60 - 80 km ( montmessin et al .
2007b , mttnen et al . 2010 ) .
these mesospheric co@xmath0 ice clouds may share some similarities with those predicted in our early mars simulations .
however they form more rarely ( probably because the atmosphere is not often below the frost point ( gonzalez - galindo 2010 , forget et al .
2009 ) ) and at significantly lower pressures ( 0.1 to 0.01 pa ) . in spite of this , the observed optical depths and particle radii are large ( typically around 0.2 and 3 @xmath4 m , respectively ) , suggesting that the growth of co@xmath0 ice particles in a co@xmath0 gas atmosphere is a very efficient process . in theory , to properly model the formation of such clouds , one must take into account various microphysical processes such as supersaturation , nucleation and crystal growth ( wood 1999 , colaprete and toon 2003 ) . in most of our simulations
, however , we assumed that condensation would occur as soon as the temperature @xmath32 predicted from the dynamical and radiative cooling rates dropped below @xmath26 . in section [ sc : microphysic ] , we also present results from a test case in which a 30@xmath33 supersaturation is required before condensing . at each timestep , the mass mixing ratio @xmath34 ( kg / kg ) of condensed ice in a model box ( or its evolution @xmath35
if ice is already present ) is simply deduced from the amount of latent heat needed to keep @xmath36 : @xmath37 with @xmath35 the mass mixing ratio of ice that has condensed or sublimated ( @xmath100 when condensing ) , @xmath38 the specific heat at constant pressure ( we took 735.9 j kg@xmath19 k@xmath19 ) and @xmath39 the latent heat of co@xmath0 ( @xmath40 j kg@xmath19 ) .
co@xmath0 ice is transported by the large - scale circulation and by turbulent and convective mixing .
the transport scheme used in our gcm is a `` van - leer i '' finite volume scheme ( hourdin and armengaud , 1999 ) . to estimate the size of the cloud particles , we assumed that the number mixing ratio of cloud condensation nuclei [ ccn ] ( kg@xmath19 ) is constant throughout the atmosphere . assuming that the cloud particle size distribution is monodisperse in each box , the cloud particle radius @xmath41
is then given by : @xmath42 } } ) ^{1/3}\ ] ] with @xmath43 the co@xmath0 ice density , set to 1620 kg m@xmath44 in our model . once @xmath41 is known , the cloud particle sedimentation velocity @xmath45 is calculated using stokes law ( rossow , 1978 ) .
sedimentation is then computed separately from the transport , but using a similar van - leer i transport scheme in the vertical .
sedimentation and condensation are strongly coupled .
within one physical timestep ( 1/48 of a sol , or 1850 s in our model ) , we found that a significant part of the ice that forms within a cloud can sediment to the layers below , where it can condense at a different rate or even sublimate .
for these reasons , in the model , condensation and sedimentation schemes are coupled together and integrated with a sub - timestep equal to @xmath46 of the physical timestep .
the cloud particle sizes are also used to calculate the cloud radiative effect .
refractive indices for co@xmath0 ice are taken from hansen ( 2005 ) and mie theory is assumed to retrieve the single scattering properties .
scattering and absorption are computed at both solar and thermal infrared wavelengths using the toon et al .
( 1989 ) scheme .
[ ccn ] is clearly a key parameter which directly controls the properties and the impact of the modeled clouds .
exploring the sensitivity of the results to this poorly known parameter allows us to account for most of the uncertainties related to the co@xmath0 ice clouds microphysics and particles size distribution .
what is the possible range of [ ccn ] ? on the earth , the number mixing ratio of cloud condensation nuclei in the troposphere ranges between 10@xmath47 kg@xmath19 ( for low saturation in clean polar air ) and 10@xmath48 kg@xmath19 ( polluted air mass)(hudson and yun , 2002,andreae , 2009 ) .
it is significantly lower for icy cirrus clouds ( @xmath4910@xmath50 kg@xmath19 ) ( e.g. demott et al .
2003 ) . even in the absence of surface or chemical sources ,
a minimum number of nuclei would be provided by meteoritic dust and smoke particles , which have been suggested to be possible condensation nuclei for terrestrial polar mesospheric clouds ( gumbel and megner , 2009 ) . such particles must have been abundant four billion years ago . in this paper , we explore [ ccn ] values ranging from 10@xmath51 kg@xmath19 to 10@xmath52 kg@xmath19 , with a baseline value of 10@xmath53 kg@xmath19 .
early mars winds were probably able to lift and transport mineral dust from the surface .
if the planet was dry enough , and in the absence of oceanic sinks as on the earth , we must take into account the possibility that the atmosphere was laden with mineral aerosols , as on mars today . to explore the impact of atmospheric dust on the early mars climate , we added a second type of aerosol in addition to the co@xmath0 ice clouds particles .
we assume that the dust is similar to that observed on present - day mars and use wolff et al .
( 2009 ) optical properties , with a constant effective radius of 1.5 @xmath4 m . instead of simulating the lifting and transport of dust by the winds ,
we prescribe the dust mixing ratio in the atmosphere and assume that it does not vary with time .
the column averaged dust optical depth at the mean pressure level is horizontally uniform .
in the vertical , the dust mixing ratio @xmath16 is taken to be constant in the lower atmosphere up to a level above which it rapidly declines , as in forget et al .
( 1999 ) : @xmath54 \right\ } & \ ; \ ; \ ; p \leq p_0 \\ q = q_0 & \ ; \ ; \ ; p >
p_0 \label{eq : newconrath}\end{aligned}\ ] ] where @xmath55 is a constant determined by the prescribed optical depth at the global average surface pressure @xmath56 , and @xmath57 the altitude ( km ) of the top of the dust layer
. in this paper simulations were performed with @xmath58 and @xmath59 km ( see section [ sc : dustresults ] ) .
to explore the range of climates that may have occurred on a planet like early mars with a thick co@xmath0 atmosphere , we have performed multiple simulations with different values of key model parameters : 1 ) mean surface pressure , 2 ) cloud microphysics parameters ( i.e. cloud condensation nuclei density ) , 3 ) obliquity and eccentricity , 4 ) surface properties , 5 ) atmospheric dust loading .
the impact of water vapor is also discussed .
solar luminosity is set to 75% of the present value ( see section [ sc : faint ] for a discussion of this assumption ) .
the solar spectrum was assumed to be the same as today .
all results correspond to the last year of a 10-year simulation . in all cases ,
the modelled planet has reached equilibrium by this time ( results are repeatable from year to year ) , except in the case of permanent co@xmath0 condensation , which induces a slow decrease of the atmospheric mass and pressure ( atmospheric collapse ; see below ) . in this section ,
we describe simulations performed assuming a circular orbit , 25@xmath3 obliquity as on present - day mars , no dust , and a constant cloud condensation nuclei number density [ ccn ] set to @xmath60 kg@xmath19 . figure [ fg : ts_ps ] presents global / annual mean surface temperature ( k ) as a function of mean surface pressure @xmath61 .
results obtained without taking into account the radiative effects of co@xmath0 ice clouds ( the clouds are assumed to be `` transparent '' at all wavelengths ) are in agreement with previous 1d model calculations performed with the same radiative transfer parameterization ( wordsworth et al .
surface temperature increases up to @xmath62 bar . above 2 - 3
bar rayleigh scattering by co@xmath0 gas more than compensates for the increased thermal infrared opacity of the atmosphere .
increasing the atmospheric thickness does not result in an increase of the mean surface temperature .
taking into account the radiative effect of co@xmath0 ice clouds results in a global warming of the surface by more than 10 k resulting from the co@xmath0 ice cloud scattering greenhouse effect ( forget and pierrehumbert , 1997 ) . in the gcm , this effect is significant , but not as much as it could have potentially been according to the estimation by forget and pierrehumbert ( 1997 ) .
this is mostly because the cloud opacities remain relatively low compared to what was assumed in that study ( see below ) .
figure [ fg : olr ] shows the radiative budget corresponding to figure [ fg : ts_ps ] simulations .
one can see that the clouds strongly raise the planetary albedo and thus decrease the absorbed solar radiation . on this plot
, one can verify that the absorbed solar energy is equal to the emitted infrared energy ( i.e. the simulated planet is in radiative equilibrium as expected ) , except at high pressure when the atmosphere collapses on the surface and releases latent heat , as detailed below .
figure [ fg : ts_ls ] presents the seasonal and latitudinal variation of zonal mean surface temperature and surface co@xmath0 ice for the different surface pressure experiments .
the maximum accumulation of co@xmath0 ice is not found exactly at the poles because co@xmath0 ice clouds are predicted to be especially thick there , and their effect on the incident thermal infrared flux tend to limit the surface cooling and thus the surface condensation . with @xmath63 bar , seasonal co@xmath0 ice caps
are predicted to form at high latitudes during fall , winter and spring as in the mars northern hemisphere today . for @xmath64
bar , however , permanent surface co@xmath0 ice glaciers are predicted to form at high latitudes in both the radiatively active and inactive cloud cases . in these simulations , the atmosphere is collapsing and the results after 10 years may not represent a realistic long - term solution .
many more years would be required to reach a steady state in which the permanent co@xmath0 ice caps would be in solid - gas equilibrium with the atmosphere , with a significant part of the atmosphere trapped on the surface .
figure [ fg : co2min1 ] show maps of the extension of the permanent co@xmath0 caps after 10 and 25 years . between 1 and 25 years , the atmospheric mass collapses by 1% per year in the ( initially ) 0.1-bar simulations , whereas the pressure drops by only 0.032% per year in the 0.3-bar case . to properly simulate an equilibrated atmosphere / permanent co@xmath0 ice cap system
, it would also be necessary to take into account the ability of co@xmath0 ice glaciers to flow and spread ( co@xmath0 ice is known to be much softer than h@xmath0o ice at the same temperature ( kreslavsky and head , 2011 , durham et al .
1999 ) ) , and probably to include the effect of slopes and roughness ( kreslavsky and head , 2005 ) .
the formation of permanent polar caps at low pressure is also very dependent on obliquity .
this is discussed in detail in section [ sc : obliquity ] . for @xmath65 bar , permanent surface co@xmath0 ice glaciers
are also predicted to form on the surface in both the radiatively active and inactive cloud cases .
co@xmath0 ice tends to condense permanently in the colder areas at high altitudes and latitude . in the @xmath66 bar case ,
the permanent co@xmath0 ice glaciers are restricted to a couple of locations on the tharsis bulge near 30@xmath3s-100@xmath3w , and in the southern high latitudes around 75@xmath3s-100@xmath3e ( figure [ fg : co2min2]a ) . for @xmath67
bar and more , the permanent co@xmath0 ice sheets are much more extensive , covering high latitudes and most of the tharsis bulge ( figure [ fg : co2min2]b ) .
the fact that permanent co@xmath0 ice cap form at either low or high pressure can be interpreted as follows : at low pressures , the greenhouse effect and heat transport are weak and thus temperatures easily reach the frost point . at high pressures
the increase in the frost point temperature overcomes the increased greenhouse effect and co@xmath0 also condense easily .
in addition , heat transport is then so efficient that horizontal temperature gradients are small
. consequently , co@xmath0 ice can form in non - polar regions .
the formation of extensive permanent and seasonal polar caps at high pressures buffers the surface temperature and explains the increase of global mean surface temperature for @xmath684 bar in figure [ fg : ts_ps ] . in these simulations ,
thick co@xmath0 ice clouds form just above the surface .
such clouds tend to cool the surface ( mischna et al . , 2000 ) because they reflect solar radiation throughout their entire depth ( the planetary albedo is increased ) whereas only the upper layers can contribute to reflect the thermal infrared flux emitted by the atmosphere below them . as a result ,
the radiatively active clouds simulations show mean surface temperatures even lower than when the clouds are assumed to be transparent .
figures [ fg : mapts]a and [ fg : mapts]b present annual mean surface temperature maps in the more realistic @xmath690.5 and 2 bar cases .
one can notice that surface temperatures strongly depend on the local topography , especially in the 2 bar simulations in which the lowest plains are the warmest places on mars .
this is not due to a stronger greenhouse warming where the atmosphere is thicker .
as illustrated on figure [ fg : ts_ps ] , the increase of surface temperature with co@xmath0 pressure due to the greenhouse effect is weaker at high pressure and negligible above 2 bars .
the opposite is found in our maps .
in fact , earthlings are familiar with this situation , which results from adiabatic cooling and warming of the atmosphere when it moves vertically and the fact that the atmosphere can influence the surface temperature when it is dense enough . on present - day mars ,
the atmosphere is too thin to affect the surface temperature .
the local topography has no significant effect on surface temperature . at which atmospheric pressure
does the transition between a `` present - day mars regime '' and an `` earth - like regime '' occur ? to address this question , we have computed the mean surface temperature lapse rate as a function of mean surface pressure ( figure [ fg : lapserate ] ) .
the lapse rate is simply estimated by performing a linear regression between annual mean surface temperatures and ground altitude at every grid point for latitudes lower than 30@xmath3 , to avoid the influence of seasonal co@xmath0 ice condensation .
we can see that the surface temperature lapse rate quickly decreases with increasing pressure .
the effect is already significant ( -1.5 k / km ) for @xmath70 bar , and reaches an asymptotic value ( -2.5 k / km ) above about 2 bars .
this is half the theoretical dry atmospheric lapse rate equal to -5.05 k / km in our model .
this can be compared to the earth case , where surface temperature lapse rates ranging between around -3 and -6.5 k / km have been reported ( see minder et al .
, 2010 , and references therein ) , while the reference dry atmospheric lapse rate is -9.8 k / km
. however , on the earth water condensation can decrease the lapse rate by a factor of two . in our simulations , annual mean surface temperatures are always significantly below 0@xmath3c . in theory , this means that potential oceans and near - surface groundwater should be permanently frozen in ice sheets and permafrost . could surface lakes and rivers form ? to investigate this possibility , a better criterion is the average temperature in summer . on the earth , for instance , intense fluvial activity takes place in northern siberia in the summertime , despite annual mean surface temperatures below -15@xmath3c ( 258 k ) .
maps of peak summertime daily mean temperatures are shown in figures [ fg : mapts]c and [ fg : mapts]d .
it appears that , unlike those in siberia , these temperatures never reach the melting point of pure water in our baseline simulations . in fact , the climate predicted by our model should be , at best , analogous to antarctica s upper dry valleys ( marchant and head , 2007 ) . water may melt , but only for a few hours during a few afternoons in summer ( as on mars today ) .
the locations where above - freezing temperatures are predicted are shown in figures [ fg : mapts]e and [ fg : mapts]f , which presents the maximum surface temperatures in the @xmath690.5 and 2 bar cases .
melting would be possible in many areas below 60@xmath3 latitude .
although the mean temperatures are lower in the 0.5 bar than in the 2 bar simulation , the diurnal amplitude is larger and the maximum temperatures are not very dependent on pressure .
the amount of water which can melt during such short episodes should be limited ( see wordsworth et al . , 2012 ) .
one can speculate that the meltwater could create supraglacial channels below the glaciers , maintain perennially ice - covered lakes ( notably by bringing energy in the form of the latent heat of fusion ; mckay et al .
1985 ) , or that brines may play a major role .
however , one must keep in mind that in our baseline simulations we modeled a dry , desertic planet mars .
first , we do not take into account the latent heat losses which tend to cool the surface if ice were present .
second , surface ice could possibly have higher albedo and thermal inertia than bare ground . to assess the sensitivity of our results to surface properties and the presence of ice
, we performed additional simulations with a global albedo of 0.4 ( instead of the present day mars ground albedo which is near 0.22 on average ) and a global thermal inertia of 1000 j s@xmath17 m@xmath18 k@xmath19 ( instead of 250 j s@xmath17 m@xmath18 k@xmath19 assumed for dry ground ) .
as shown in figure [ fg : maptsice ] , annual mean temperatures are lowered by about 8 and 12 k at 0.5 and 2 bar respectively .
this reflects the effect of lowering the albedo , since thermal inertia has little effect on annual mean temperatures .
however , increasing the thermal inertia dampens the maximum temperatures , which are lowered by about 30 k. this suggests that extensive ice deposits would not reach 0@xmath3c , and that , as in antarctica upper dry valleys today , only ice lying on dark rocks may sometimes melt ( marchant and head , 2007 ) . figure [ fg : sectiont ] presents cross - sections of zonal mean temperatures and zonal winds in northern summer and winter in the 2 bar baseline simulation . at this pressure ,
the atmospheric thermal structure is relatively homogeneous , with little latitudinal and seasonal variations compared to present - day mars or even the earth .
temperatures decrease monotonically with altitude , even near the surface , as discussed above .
the small latitudinal gradients nevertheless induce a meridional circulation .
plotting the mass streamfunction ( not shown ) reveals that it is characterized by earth - like overturning hadley cells between 30@xmath3s and 30@xmath3n , and which extend up to about 15 km . in a given season , the cross - equatorial cell with rising motion in the spring - summer hemisphere and descending motion in the fall - winter hemisphere dominates .
the corresponding zonal wind structure is characterized by an earth - like subtropical prograde jet in the winter hemisphere .
the summer hemisphere retrograde jet is much more mars - like . in the northern hemisphere ,
the large longitudinal variations in topography induce a strong stationary wave which modulates the zonal wind . in this 2-bar simulation ,
temperatures below the co@xmath0 condensation point are predicted above about 11 km , and co@xmath0 ice clouds form at all seasons and latitudes ( figure [ fg : section_clouds ] ) .
however , at any given time , co@xmath0 ice clouds typically cover about half of the planet ( when counting visible opacity above 0.2 ) , as illustrated in figure [ fg : map_cloud_inst ] .
the locations of the clouds evolve constantly , with a combination of transient cloud condensed and transported by travelling waves , and stationary clouds forming above topography features ( such as tharsis , the edge of hellas or arabia terra ) .
this most likely results from resolved gravity waves .
is topography a key driver of clouds and climate ? to address this question
, we performed a similar simulation , but with topography removed .
we found that the cloud coverage is relatively similar , with a global mean visible opacity of 6 compared to 4.5 with topography .
the mean surface temperature is almost identical ( 231.5 k compared to 231.0 k ) .
figures [ fg : section_clouds]a and c show the annual mean section of the cloud density and cloud particles radii in our 2-bar baseline simulation with topography .
the average particle radius is well above 10 @xmath4 m , the minimum size to readily scatter thermal infrared radiation and induce the scattering greenhouse effect ( forget and pierrehumbert , 1997 ) .
however , with a mean optical depth near 4.5 and fractional cloud cover most of the time , the clouds only induce a 10 k greenhouse warming , as discussed above .
because modeled clouds properties can be expected to be sensitive to model resolution , we performed an additional 2-bar simulation with doubled resolution in latitude and longitude .
results are compared in figures [ fg : section_clouds ] , [ fg : map_cloud_inst ] , and [ fg : co2cloud]e and f. in the high - resolution simulation , the impact of topography and gravity waves seems to be stronger .
the cloud distribution ( figure [ fg : co2cloud]f ) exhibits more structure in relation to local topography .
nevertheless , the mean cloud cover , optical depth , and particle sizes are about the same on average ( average visible opacity of 4.6 vs 4.5 in the baseline simulation ) .
the mean surface temperature is only 1 k warmer , which is not significant .
similarly , we explored the sensitivity to the model vertical resolution by doubling the number of layers and thus the resolution above the boundary layer ( reaching a vertical resolution of about 2.5 km between 5 and 25 km ) , and found very similar results .
figure [ fg : co2cloud ] presents the annual mean coverage in optical depth at other pressures in the baseline cases . the average cloud visible optical depth @xmath71 growths with mean pressure @xmath61 , but at the same time the altitude of the bottom of the cloud layer @xmath72 decreases .
for instance , @xmath71=0.7 and @xmath7314 km for @xmath61=0.5 bar ; @xmath71=1.8 and @xmath7311 km for @xmath61=1 bar ; @xmath71=4.5 and @xmath738 km for @xmath61=2 bar ; @xmath71=9 and @xmath735 km for @xmath61=3 bar ; and @xmath71=16 and @xmath730 km for @xmath61=4 bar . as mentioned above , the scattering greenhouse effect for low lying clouds is reduced . above 4 bars ,
the clouds cool the planet rather than warming it .
as explained in section [ sc : modelcloud ] , we consider that many of the uncertainties related to the co@xmath0 ice clouds microphysics and particle sizes can be accounted for by varying the prescribed number mixing ratio of cloud condensation nuclei [ ccn ] .
we explored the sensitivity of the model to this parameter in the @xmath690.5 and 2 bar cases by performing 7 simulations with [ ccn ] set to 10@xmath51 , 10@xmath74 , ... , 10@xmath25 , 10@xmath52 kg@xmath19 ( see section [ sc : modelcloud ] ) .
the primary effect of decreasing [ ccn ] is to increase the size of the cloud particles when they form . on the one hand , this enhances the sedimentation rate and thus decreases the mass of the cloud ( and thus its opacity ) . on the other hand , for a given cloud mass , this affects the cloud opacity by changing the number of particles and their single scattering properties .
we found that the first effect dominates : almost no clouds are present with [ ccn]=10@xmath51 kg@xmath19 . with higher [ ccn ] , the average cloud mass increased by a factor of about 3 for each order of magnitude increase in [ ccn ] .
meanwhile , the mean cloud particle sizes ( computed by taking into account sedimentation and sublimation below the condensation level ) remains near 10 to 20 @xmath4 m . as a result , with @xmath692 bar ,
the mean cloud optical depth reaches 16 , 67 and 156 for [ ccn]=10@xmath47 , 10@xmath25 and 10@xmath52 kg@xmath19 , respectively . in the last two cases , the cloud optical depth is too thick to allow the scattering greenhouse effect to take place ( forget and pierrehumbert , 1997 ) .
the clouds are so thick in the visible that albedo increases outweigh the scattering of upwelling ir .
the clouds reflect most of the incoming solar radiation , and their net effect is to cool the surface as seen in figure [ fg : nmix ] . with @xmath690.5 bar , the mass of co@xmath0 ice clouds is lower and the cloud warming is maximized with more nuclei , up to 10@xmath25 kg@xmath19 ( figure [ fg : nmix ] ) as mentioned in section [ sc : previous ] , based on the measured constraints on the critical saturation level and the microphysical properties of the formation of carbon dioxide cloud particles ( glandorf et al . , 2002 ) , colaprete and toon ( 2003 ) pointed out that , since nucleation efficiency decreases when the size of the nucleating particles decreases , only the biggest particles are efficiently nucleated and are able to condensate co@xmath0 .
they concluded that , as a result , co@xmath0 clouds on early mars should contain few , large particles with average radii greater than 500 @xmath4 m .
in addition , they showed that a supersaturation of nearly 35% is needed to allow for such an efficient nucleation .
while the first effect could be well mimicked by reducing [ ccn ] as done above , we have run another set of simulations to quantify the impact of the critical supersaturation needed to form clouds in a cloud free medium . in these simulations ,
the number density of ccn is prescribed as above . however , in a grid cell where no significant amount of carbon dioxide ice is present at the beginning of a timestep , co@xmath0 is allowed to condensate only if the saturation @xmath75 reaches a critical value @xmath76 , i.e. if the temperature in the cell is below the nucleation temperature @xmath77 , where @xmath78 is the equilibrium condensation temperature for a co@xmath0 partial pressure @xmath27 . the average temperatures reached in these simulations are also shown in figure [ fg : nmix ] as a function of [ ccn ] .
as can be seen , the need for supersaturation decreases the warming effect of co@xmath0 clouds by a few degrees . however , this variation is weaker than the one obtained by varying the density of ccn .
the obliquity and the eccentricity of mars have strongly varied throughout its existence .
as the evolution of these parameters is strongly chaotic , it is not possible to know their values before a few million years ago ( laskar et al . 2004 ) .
nevertheless , laskar et al . (
2004 ) showed that the obliquity could have varied between less than 5@xmath3 and up to more than 60@xmath3 or even @xmath79 .
in fact , the average value of the obliquity over 5 gyr was estimated to be 37.625@xmath3 with a standard deviation of 13.82@xmath3 .
the average eccentricity was 0.0690 , with standard deviation 0.0299 .
more recently , brasser and walsh ( 2011 ) reanalysed the stability of the martian obliquity for the noachian era , taking into account the fact that before the late heavy bombardment the giant planets may have been on drastically different orbits than today , according to gomes et al .
( 2005 ) . for such conditions
, they found that the martian obliquity would have remained chaotic for its most probable mean values , between about 30@xmath2 and 60@xmath2 , but more stable for the less probable mean obliquities below 30@xmath2 ( with oscillation amplitude still as high as 20@xmath2 ) and above 60@xmath2 ( with amplitude of 9@xmath2 ) varying the obliquity or the season of perihelion in models of the present - day mars climate system has a profound impact on the surface temperatures and the water cycle ( e.g. haberle et al .
2003 , forget et al .
2006 , montmessin et al .
the impact is especially strong at high latitudes , with the average insolation at the pole proportional to the sine of the obliquity .
the permanent polar caps predicted to form for @xmath80 and @xmath81 bar in the baseline case ( 25@xmath3 obliquity ) and discussed in section [ sc : tsurf ] do not form at 35@xmath3 obliquity .
conversely , we found that a southern permanent polar cap forms when running with @xmath690.5 bar and 10@xmath3 obliquity ( whereas there is no such cap in the baseline case with 25@xmath3 obliquity shown in figure [ fg : ts_ls]c ) . with @xmath62
bar , the thick atmosphere prevents the formation of permanent polar caps even in the @xmath82 obliquity case .
in fact , in that simulation co@xmath0 does not condense on the surface at all ! figure [ fg : cap_ob ] summarizes the range of mean surface pressures and obliquities for which permanent co@xmath0 ice caps are predicted to form , possibly leadingto atmospheric collapse .
in general , a higher obliquity corresponds to a stronger seasonal cycle , with a warmer summer and a colder winter at mid and high latitudes .
could this significantly increase maximum temperatures and allow seasonal melting ?
figure [ fg : ts_ob ] shows the global mean , maximum daily mean and planetary maximum temperatures in the @xmath70 and 2 bar cases for various obliquities between 10@xmath3 and 60@xmath3 . in both cases , the global mean temperature slightly decreases with obliquity .
this is due to the raising of the albedo resulting from the seasonal polar caps which are more and more extensive when obliquity increases .
as expected , the maximum daily mean temperature strongly increases with obliquity , especially in the @xmath70 bar case for which temperature above 0@xmath3c are predicted at very high obliquity .
looking at temperature maps ( not shown ) , we can see that such high temperatures are reached in both hemisphere ( more in the north ) , above about 70@xmath3 latitude .
insolation can be further increased near perihelion in the case of an eccentric orbit . to explore this effect , and in order to create a realistic `` optimum case '' we performed two additional simulations with @xmath70 and 2 bar , assuming an eccentricity of 0.1 , a perihelion during northern summer solstice , and an obliquity set to 41.8@xmath3 ( the most likely obliquity for mars according to laskar et al .
( 2004 ) ) .
the corresponding maximum daily mean temperatures are shown in figure [ fg : maptslsp ] . in the 0.5 bar case ,
seasonal temperatures above 0@xmath3c are predicted everywhere above 70@xmath3 latitude .
if valley networks and layered deposits were mostly observed at such high latitudes , it would be interesting to relate their formation to this type of orbital configurations .
of course , this is not the case . to explore the radiative effects of airborne mineral aerosols ,
we have used the dust parameterization described in section [ sc : dustmodel ] and performed a series of simulations with various amount of dust in the @xmath69 0.5 and 2 bar cases .
the impact of dust opacity on the global averaged surface temperature is shown in figure [ fg : ts_dust ] .
two vertical distributions are employed . when the``dust top '' is set to 100 km , the dust mixing ratio decreases slowly in the lower 50 km ( 80% of the surface value at 50 km ) and more steadily above , up to 100 km .
similarly , with a dust top at 30 km , most of the dust is confined below 20 km . in most cases ,
it is found that dust can slightly warm the surface . with a dust
visible opacity @xmath71 set to @xmath835 , surface temperatures are increased by a couple of kelvins in the @xmath69 0.5 bar case , and by more than 10 k with @xmath69 2 bar . with dust opacity @xmath8310 ,
the impact is reduced . to interpret these results , we can note that airborne dust modifies the energy balance in four ways : 1 ) dust absorbs solar radiation and thus warms the atmosphere , 2 ) as a result , it decreases the co@xmath0 ice clouds opacity 3 ) dust decreases the planetary albedo , 4 ) dust increases the atmospheric infrared opacity . the first two effects tend to reduce the atmospheric greenhouse surface warming and thus the mean surface temperatures .
however , the other two effects work in the opposite direction and actually dominate .
figure [ fg : tprofile_dust ] illustrates the impact of dust on the temperature profiles in the @xmath69 2 bar case .
temperatures above 10 km are warmed by 15 k with an opacity @xmath835 and by more than 20 k with @xmath8310 .
this should decrease the greenhouse effect by increasing the outgoing infrared radiation , but this is more than compensated by the fact that the dust infrared opacity allows the mean infrared emission to take place at higher and colder altitude .
the dominant efect is thus an increase in surface temperature to keep the energy budget in balance .
this is less true with @xmath8310 , and one can see that the surface temperature is then lower than with @xmath835 . in the @xmath835 case ,
the co@xmath0 ice cloud mean opacity is reduced from about 5 with no dust down to about 1.5 and 0.17 with the dust top at 30 and 100 km , respectively . the colder upper atmosphere temperatures and the thicker clouds explain why surface temperatures are a few kelvins warmer when the dust is confined in the lower atmosphere . in this paper we have assumed that mars was mostly dry , and neglected the additional greenhouse effect of water vapor .
however , even in the case of a water - covered mars , this effect is limited because our predicted atmospheric temperatures are in most cases so low that even at saturation the concentration of water vapor remains small . in the companion paper , wordsworth et al .
( 2012 ) present simulations similar to our baseline cases with various co@xmath0 surface pressure , but in which the greenhouse effect of water vapor is maximized by assuming 100% relative humidity everywhere . as seen on their figures 2 and 3 , it is found that even in this overestimated case , water vapor only increases surface temperatures by a few kelvins , and thus does not qualitatively change our conclusions . in reality
, water also affects the climate by forming water ice clouds which can either cool or warm the planet .
this is further discussed in wordsworth et al .
in this work , our purpose has been to develop a climate model complete enough to properly answer the question `` what is the climate on a mars - like planet with a thick co@xmath0 atmosphere and a faint young sun ? ''
previous work had found contradictory answers to this apparently simple question ( pollack et al . 1987 , kasting et al .
1991 , forget and pierrehumbert , 1997 , mischna et al .
2000 , colaprete and toon , 2003 ) . by using a complete 3d gcm , revised spectrocopic properties for collision - induced co@xmath0 , and by systematically exploring the model sensitivity to the possible surface pressures , cloud microphysics properties , obliquity and orbital properties , atmospheric dust loading and model resolution
, we hope to have provided an improved answer to the question . in our simulations , a wide range of dynamical , meteorological and climatic phenomenons are observed .
unlike on present day mars , and more like on the earth , for pressure higher that a fraction of bar , surface temperatures varies with altitude because of the adiabatic cooling and warming of the atmosphere when it moves vertically and the increased coupling between the atmosphere and the surface under higher atmospheric pressures than today . in most simulations ,
co@xmath0 ice clouds cover a major part of the planet but not all .
their behavior is controlled by a combination of large - scale ascent and descent of air , stationary and travelling waves , and resolved gravity waves related to the topography .
the formation of co@xmath0 ice seasonal or perennial deposits is also a key process in the climates that we have explored . in particular , atmospheric collapse in permanent co@xmath0 ice caps
could have buffered the thickness of the early mars atmosphere to pressures lower than about 3 bar ( in the unlikely case that enough co@xmath0 was available ; see section [ sc : which_atm ] ) .
conversely at pressures lower than one bar , we find that if the obliquity is lower than a threshold value between about @xmath82 and @xmath84 ( depending on the initial pressure ) , a part of the atmosphere may be trapped in the high latitudes , as on mars today .
however , as mentioned above , to properly simulate an equilibrated atmosphere / permanent co@xmath0 ice cap system , it would also be necessary to take into account the ability of co@xmath0 ice glaciers to flow and spread , and probably include the effect of slopes and roughness ( kreslavsky and head 2005 , 2011 ) .
our most striking result is that no combination of parameters can yield surface temperatures consistent with the melting and flow of liquid water as suggested by the available geological evidence on early mars .
previous studies had suggested that co@xmath0 ice clouds could have strongly warmed the planet thanks to their scattering greenhouse effect .
looking at figures [ fg : ts_ps ] and [ fg : nmix ] , one can see that the optimum case exhibiting maximum surface temperatures in our simulations is with @xmath61=2 bar and [ ccn ] near 10@xmath53 - 10@xmath47 kg@xmath19 .
this actually corresponds to one of the reference casess that we have studied in detail . even in that case , the mean cloud warming remains lower than 15 k because of the partial cloud coverage and the limited cloud optical depth .
we conclude that a co@xmath0 atmosphere could not have raised the annual mean temperatures above 0@xmath2c anywhere on the planet .
summertime diurnal mean surface temperatures above 0@xmath2c ( a condition which could have allowed rivers and lakes to form ) are predicted for obliquity larger than 40@xmath2 at high latitudes but not in locations where most valley networks or layered sedimentary units are observed . at most latitudes , above melting temperatures
occur , but only for a few hours during summer afternoons , and assuming albedo and thermal inertia values which may be unrealistic for snow or ice deposits . in the companion paper , wordsworth et al .
( 2012 ) present further results on this subject using a full water cycle model that predicts the stability and evolution of surface ice deposits .
like in most previous model studies , we assumed that the early solar flux was 75% than today , on the basis of the `` standard solar model '' ( e.g. gough , 1981 ) .
can this model be questioned ?
the fundamental reason is very robust : `` the gradual increase in luminosity during the core hydrogen burning phase of evolution of a star is an inevitable consequence of newtonian physics and the functional dependence of the thermonuclear reaction rates on density , temperature and composition '' ( gough , 1981 ) . in practice
, it agrees very well with solar neutrinos and helioseismology observations . the only way to explain a brighter early sun is to assume that it was more massive initially and that it subsequently lose the excess mass in an intense solar wind during its first billion years ( whitmire et al .
a few percent of the mass would be sufficient because a star luminosity strongly depends on its mass .
interestingly , a younger , more massive sun has also been suggested recently to explain inconsistencies between the standard solar model predictions and recently revised measurements of the solar elemental abundances ( guzik and mussack 2010 , turck - chieze et al . , 2011 ) .
however , observations of mass loss in young stars do not suggest that they lose significant mass after the first 0.1 gyr ( wood et al .
2005 , minton and malhotra , 2007 ) . nonetheless , the faint young sun is still a topic of active investigation .
could supplemental trace greenhouse gases boost the greenhouse power of a pure co@xmath0 ( and h@xmath0o ) atmosphere ?
past studies have notably explored the effects of sulfur dioxide ( so@xmath0 ) , hydrogen sulfide ( h@xmath0s ) , methane ( ch@xmath85 ) , and ammonia ( nh@xmath7 ) .
the key to the success of these gases in solving the faint young sun paradox for early mars rests on two main conditions : their ability to absorb infrared radiation in parts of the spectrum not covered by co@xmath0 and h@xmath0o , and their ability to sustain the needed concentrations given plausible sources and sinks . of the gases listed ,
nh@xmath7 has the potential to plug up the so - called `` window '' region ( 800 - 1200 @xmath6 ) as it exhibits strong absorption between 700 and 1300 @xmath6 .
this makes it a very powerful greenhouse gas .
kasting et al .
( 1992 ) suggested that 500 ppm of nh@xmath7 in a 4 - 5 bar co@xmath0 atmosphere could raise surface temperatures to 273 k. however , nh@xmath7 is photochemically unstable ( kuhn and atreya , 1979 ) and would require shielding to survive ( e.g. , sagan and chyba 1997 ; wolf and toon , 2010 ) .
methane , which is also unstable and whose photolysis products could provide the shield , has absorption features at somewhat higher wavenumbers ( 1200 - 1500 @xmath6 ) , which could further help reduce the outgoing infrared .
however , recent calculations indicate that even at concentrations of 500 ppm ch@xmath85 does not significantly boost the greenhouse effect of a pure co@xmath0/h@xmath0o atmosphere ( jim kasting , private communication ) .
this is partly the result of an overlap with water bands , partly because these bands are on the wings of the planck function , and partly due to the fact that ch@xmath85 also absorbs in the near infrared , which has an anti - greenhouse effect . and like nh@xmath7 , ch@xmath85 would require strong sources to sustain the above concentrations . as noted in the introduction , the detection of noachian - hesperian aged sulfate deposits has renewed interest in so@xmath0 and h@xmath0s as potential greenhouse gases .
an obvious source for these gases is volcanic activity , which was much higher in the past than it is today .
h@xmath0s has absorption features in the window region but they are weak , and its stronger bands at lower wavenumbers occur at the tail end of the planck function . on the other hand ,
so@xmath0 @xmath86 fundamental near 520 @xmath6 is near the peak of a 273 k planck function , and it also has significant opacity at the high wavenumber end of the window region .
thus , so@xmath0 has received most of the attention ( e.g. , halevy et al .
, 2007 ; johnson et al . , 2008 ) . to be effective , so@xmath0 needs to build up to concentrations around the 10 ppm level or higher in moderately thick atmospheres ( surface pressures greater than 0.5 bar ) . in the model of johnson et al
( 2008 ) global mean surface temperatures well above freezing are achieved with so@xmath0 concentrations in this range .
this is the only modern model to predict such warm conditions with a pure gaseous greenhouse effect .
however , in practice the so@xmath0 greenhouse faces significant challenges as so@xmath0 readily converts to aerosols regardless of the redox state of the atmosphere . and
as on earth , these aerosols should have a net cooling effect on surface temperatures ( tian et al . , 2010 ) .
furthermore , so@xmath0 is highly soluble and will washout quickly when conditions become warm enough for rainfall .
thus , while the supplemental greenhouse gases considered thus far in the literature can not be ruled out , they require special circumstances to play a significant role in warming the surface of mars to the melting point of liquid water . to date , no model has been published that convincingly demonstrates a gaseous greenhouse can solve the faint young sun problem for early mars . within that context , using our global climate model , we plan to explore the radiative effect and the chemical cycle of these gases , starting with the sulfur cycle .
the idea that early mars climate may have been cold is not new .
several studies have suggested that the geomorphological evidence related to liquid water could be explained by assuming hydrothermal convection and the formation of aquifers driven by geothermal heat and heat associated with impacts ( squyres and kasting , 1994 , gaidos and marion , 2003 ) .
erosion and surface alteration may have occurred by saline and acidic liquid solutions ( brines ) which can be stable at sub - zero temperatures ( fairen , 2010 ) .
impacts may have induced melting and rainfall on a global scale ( toon et al . 2010 ) or locally ( e.g. mangold , 2012 , and reference therein ) which may explain the formation of fluvial valleys even under cold background climate conditions . similarly , while clay minerals have been thought to be an indication of surface weathering by liquid water and thus an indication of warm and wet past conditions , ehlmann et al .
( 2011 ) showed that , in many locations , available data instead indicates clay formation by hydrothermal groundwater circulation possibly operating in cold , arid surface conditions .
furthermore , meunier et al . (
2012 ) proposed that the iron - and - magnesium - rich clay formed direcly by the precipitation from water - rich magama - derived fluids .
this new paradigm is not without problems . in particular , it is more and more clear that `` early '' mars encompasses several epochs . while a cold early planet scenario may be consistent with relatively recent late noachian or hesperian geological features such as delta fans or the most recent fluvial features , several authors have argued that it could not explain the morphology of large valley networks , the very high global erosion rate inferred for the early noachian period , or the formation of extensive deposits of phyllosilicates like in the mawrth vallis region ( e.g. craddock and howard , 2002 , bibring et al .
2006 , mckeown et al .
2009 ) .
in any case , there is no doubt that conditions on mars during the first billion years of its history were dramatically different from those at present .
a thicker atmosphere was probably necessary to allow water to flow for long distances across the surface .
our simulations of the resulting climate provide clues on the environment , and notably show that with surface pressure above a fraction of a bar , the southern highlands become cold traps where ice may migrate and replenish water sources .
this hydrological cycle is modeled and discussed in detail in the companion paper ( wordsworth et al . , 2012 ) .
we are grateful for the comments on the text and scientific suggestions provided by james w. head .
we thank n. mangold , b. charnay , f. leblanc , a. spiga , j - p .
bibring for helpful discussions .
we also appreciate the reviews provided by e. kite and s. clifford .
, y. i. , lafferty , w. j. , fraser , g. t. , 2004 . infrared spectrum of the continuum and dimer absorption in the vicinity of the o@xmath87 vibrational fundamental in o@xmath87/co@xmath87 mixtures .
journal of molecular spectroscopy 228 , 432440 . ,
p . , langevin , y. , gendrin , a. , gondet , b. , poulet , f. , berth , m. , soufflot , a. , arvidson , r. , mangold , n. , mustard , j. , drossart , p. , 2005 . mars surface diversity as revealed by the omega / mars express observations .
science 307 , 15761581 . , j .-
p . , langevin , y. , mustard , j. f. , poulet , f. , arvidson , r. , gendrin , a. , gondet , b. , mangold , n. , pinet , p. , forget , f. , 2006 .
global mineralogical and aqueous mars history derived from omega / mars express data .
science 312 , 400404 . , w. v. , ming , d. w. , kounaves , s. p. , m. , y. s. m. , arvidson , r. e. , hecht , m. e. , hoffman , j. , niles , p. b. , k. , h. d. , quinn , r. c. , smith , p. h. , sutter , b. , catling , d. c. , moriis , r. v. , 2009 .
evidence for calcium carbonate at the phoenix landing site .
science 325 , 61 . , p. j. , cziczo , d. j. , prenni , a. j. , murphy , d. m. , kreidenweis , s. m. , thomson , d. s. , borys , r. , rogers , d. c. , 2003 .
measurements of the concentration and composition of nuclei for cirrus formation .
proceedings of the national academy of science 100 , 1465514660 . , b. l. , mustard , j. f. , murchie , s. l. , bibring , j .-
p . , meunier , a. , fraeman , a. a. , langevin , y. , 2011 .
subsurface water and clay mineral formation during the early history of mars .
nature 479 , 5360 . , b. l. , mustard , j. f. , murchie , s. l. , poulet , f. , bishop , j. l. , brown , a. j. , calvin , w. m. , clark , r. n. , des marais , d. j. , milliken , r. e. , roach , l. h. , roush , t. l. , swayze , g. a. , wray , j. j. , 2008 .
orbital identification of carbonate - bearing rocks on mars .
science 322 , 1828. forget , f. , hourdin , f. , fournier , r. , hourdin , c. , talagrand , o. , collins , m. , lewis , s. r. , read , p. l. , huot .
improved general circulation models of the martian atmosphere from the surface to above 80 km .
j. geophys
104 , 24,15524,176 .
, f. , montmessin , f. , bertaux , j .- l .
, gonzlez - galindo , f. , lebonnois , s. , qumerais , e. , reberac , a. , dimarellis , e. , lpez - valverde , m. a. , 2009 .
density and temperatures of the upper martian atmosphere measured by stellar occultations with mars express spicam .
journal of geophysical research ( planets ) 114 ( e13 ) , 1004+ . , a. , mangold , n. , bibring , j .-
langevin , y. , gondet , b. , poulet , f. , bonello , g. , quantin , c. , mustard , j. , arvidson , r. , lemoulic , s. , 2005 .
sulfates in martian layered terrains : the omega / mars express view . science 307 , 15871591 . , f. , bougher , s. w. , lpez - valverde , m. a. , forget , f. , murphy , j. , 2010 .
thermal and wind structure of the martian thermosphere as given by two general circulation models .
planetary and space sciences 58 , 18321849 . , r. m. , forget , f. , colaprete , a. , schaeffer , j. , boynton , w. v. , kelly , n. j. , chamberlain , m. a. , 2008 .
the effect of ground ice on the martian seasonal co@xmath0 cycle .
space sci .
56 , 251255 .
hourdin , f. , forget , f. , talagrand , o. , 1995 . the sensitivity of the martian surface pressure to various parameters : a comparison between numerical simulations and viking observations .
j. geophys .
100 , 55015523 . ,
s. , hourdin , f. , eymet , v. , crespin , a. , fournier , r. , forget , f. , jun .
superrotation of venus atmosphere analyzed with a full general circulation model . journal of geophysical research ( planets ) 115 , 6006 . ,
a. , montmessin , f. , gondet , b. , scholten , f. , hoffmann , h. , gonzlez - galindo , f. , spiga , a. , forget , f. , hauber , e. , neukum , g. , bibring , j. , bertaux , j. , 2010 . mapping the mesospheric co@xmath87 clouds on mars : mex / omega and mex / hrsc observations and challenges for atmospheric models .
icarus 209 , 452469 .
, j .- b . ,
forget , f. , millour , e. , montabone , l. , wolff , m. j. , nov . 2011 . revisiting the radiative impact of dust on mars using the lmd global climate model .
journal of geophysical research ( planets ) 116 , 11010 . , n. k. , bishop , j. l. , noe dobrea , e. z. , ehlmann , b. l. , parente , m. , mustard , j. f. , murchie , s. l. , swayze , g. a. , bibring , j .- p . , silver , e. a. , 2009 . characterization of phyllosilicates observed in the central mawrth vallis region , mars , their potential formational processes , and implications for past climate . journal of geophysical research ( planets ) 114 , 0 . ,
a. , petit , s. , ehlmann , b. l. , dudoignon , p. , westall , f. , mas , a. , el albani , a. , ferrage , e. , 2012 .
magmatic precipitation as a possible origin of noachian clays on mars .
nature geoscience 5 , 739743 .
, f. , forget , f. , rannou , p. , cabane , m. , haberle , r. m. , 2004 . origin and role of water ice clouds in the martian water cycle as inferred from a general circulation model .
journal of geophysical research ( planets ) 109 ( e18 ) , 10004 . , f. , gondet , b. , bibring , j .-
p . , langevin , y. , drossart , p. , forget , f. , fouchet , t. , 2007 .
hyperspectral imaging of convective co@xmath87 ice clouds in the equatorial mesosphere of mars .
journal of geophysical research ( planets ) 112 , 11+ . , f. , haberle , r. m. , forget , f. , langevin , y. , clancy , r. t. , bibring , j .-
p . , 2007 .
on the origin of perennial water ice at the south pole of mars : a precession - controlled mechanism ?
journal of geophysical research ( planets ) 112 ( e11 ) , 8+ . , r. v. , ruff , s. w. , gellert , r. , ming , d. w. , arvidson , r. e. , clark , b. c. , golden , d. c. , siebach , k. , klingelhfer , g. , schrder , c. , fleischer , i. , yen , a. s. , squyres , s. w. , 2010 .
identification of carbonate - rich outcrops on mars by the spirit rover .
science 329 , 421. , m. m. , hamilton , v. e. , bandfield , j. l. , glotch , t. d. , baldridge , a. m. , christensen , p. r. , tornabene , l. l. , anderson , f. s. , 2008 .
chloride - bearing materials in the southern highlands of mars .
science 319 , 1651. , m. y. , hartmann , j. m. , 1989 .
temperature - dependent measurements and modeling of absorption by co2-n2 mixtures in the far line - wings of the 4.3-micron co2 band .
j. quant .
spectrosc .
transfer 42 , 311317 .
, r. j. , zuber , m. t. , solomon , s. c. , golombek , m. p. , jakosky , b. m. , banerdt , w. b. , smith , d. e. , williams , r. m. e. , hynek , b. m. , aharonson , o. , hauck , s. a. , 2001 .
ancient geodynamics and global - scale hydrology on mars .
science 291 , 25872591 . ,
f. , bibring , j .- p . ,
mustard , j. f. , gendrin , a. , mangold , n. , langevin , y. , arvidson , r. e. , gondet , b. , gomez , c. , 2005 .
phyllosilicates on mars and implications for early martian climate .
nature 438 , 623627 . , l. s. , gordon , i. e. , barbe , a. , benner , d. c. , bernath , p. f. , birk , m. , boudon , v. , brown , l. r. , campargue , a. , champion , j .-
, chance , k. , coudert , l. h. , dana , v. , devi , v. m. , fally , s. , flaud , j .- m .
, gamache , r. r. , goldman , a. , jacquemart , d. , kleiner , i. , lacome , n. , lafferty , w. j. , mandin , j .- y . , massie , s. t. , mikhailenko , s. n. , miller , c. e. , moazzen - ahmadi , n. , naumenko , o. v. , nikitin , a. v. , orphal , j. , perevalov , v. i. , perrin , a. , predoi - cross , a. , rinsland , c. p. , rotger , m. , v simekov , m. , smith , m. a. h. , sung , k. , tashkun , s. a. , tennyson , j. , toth , r. a. , vandaele , a. c. , vander auwera , j. , 2009 .
the hitran 2008 molecular spectroscopic database .
j. quant .
spectrosc .
transfer 110 , 533572 .
, s. w. , arvidson , r. e. , ruff , s. , gellert , r. , morris , r. v. , ming , d. w. , crumpler , l. , farmer , j. d. , des marais , d. j. , yen , a. , mclennan , s. m. , calvin , w. , bell , j. f. , clark , b. c. , wang , a. , mccoy , t. j. , schmidt , m. e. , de souza , p. a. , 2008 .
detection of silica - rich deposits on mars .
science 320 , 1063. , s. w. , grotzinger , j. p. , arvidson , r. e. , bell , j. f. , calvin , w. , christensen , p. r. , clark , b. c. , crisp , j. a. , farrand , w. h. , herkenhoff , k. e. , johnson , j. r. , klingelhfer , g. , knoll , a. h. , mclennan , s. m. , mcsween , h. y. , morris , r. v. , rice , j. w. , rieder , r. , soderblom , l. a. , 2004 . in situ evidence for an ancient aqueous environment at meridiani planum , mars .
science 306 , 17091714 . , f. , claire , m. w. , haqq - misra , j. d. , smith , m. , crisp , d. c. , catling , d. , zahnle , k. , kasting , j. f. , 2010 .
photochemical and climate consequences of sulfur outgassing on early mars . earth and planetary science letters 295 , 412418 .
toon , o. b. , mckay , c. p. , ackerman , t. p. , santhanam , k. , 1989 .
rapid calculation of radiative heating rates and photodissociation rates in inhomogeneous multiple scattering atmospheres .
j. geophys
94 , 16,28716,301 .
, r. , forget , f. , millour , e. , madeleine , j .- b . , haberle , r. m. , eymet , v. , feb .
modelling past mars climates and water cycle with a thicker co2 atmosphere . in : f.
forget & e. millour ( ed . ) , mars atmosphere : modelling and observation .
. 447448 . , r. d. , forget , f. , selsis , f. , millour , e. , charnay , b. , madeleine , j .- b . , 2011 .
gliese 581d is the first discovered terrestrial - mass exoplanet in the habitable zone .
the astrophysical journal letters 733 , l48 . | on the basis of geological evidence , it is often stated that the early martian climate was warm enough for liquid water to flow on the surface thanks to the greenhouse effect of a thick atmosphere .
we present 3d global climate simulations of the early martian climate performed assuming a faint young sun and a co@xmath0 atmosphere with surface pressure between 0.1 and 7 bars .
the model includes a detailed representation of radiative transfer using revised co@xmath0 gas collision induced absorption properties , and a parameterisation of co@xmath1 ice cloud microphysical and radiative properties .
a wide range of possible climates is explored using various values of obliquities , orbital parameters , cloud microphysic parameters , atmospheric dust loading , and surface properties . unlike on present - day mars , for pressures higher than a fraction of a bar surface temperatures
vary with altitude because of adiabatic cooling / warming of the atmosphere . in most simulations ,
co@xmath0 ice clouds cover a major part of the planet .
previous studies suggested that they could have warmed the planet thanks to their scattering greenhouse effect .
however , even assuming parameters that maximize this effect , it does not exceed + 15 k. combined with the revised co@xmath0 spectroscopy and the impact of surface co@xmath0 ice on the planetary albedo , we find that a co@xmath0 atmosphere could not have raised the annual mean temperature above 0@xmath2c anywhere on the planet .
the collapse of the atmosphere into permanent co@xmath0 ice caps is predicted for pressures higher than 3 bar , or conversely at pressure lower than one bar if the obliquity is low enough .
summertime diurnal mean surface temperatures above 0@xmath2c ( a condition which could have allowed rivers and lakes to form ) are predicted for obliquity larger than 40@xmath2 at high latitudes but not in locations where most valley networks or layered sedimentary units are observed . in the absence of other warming mechanisms , our climate model results are thus consistent with a cold
early mars scenario in which non climatic mechanisms must occur to explain the evidence for liquid water . in a companion paper by wordsworth et al .
, we simulate the hydrological cycle on such a planet and discuss how this could have happened in more detail .
mars , atmospheres , evolution , mars , climate , mars , polar caps , mars , surface |
let @xmath1 be a smooth compact and connected symplectic manifold of dimension @xmath5 and let @xmath0 be a torus which acts effectively on @xmath1 by means of symplectomorphisms .
we furthermore assume that some principal @xmath0-orbit is a coisotropic submanifold of @xmath1 , which implies that @xmath6 if @xmath7 denotes the dimension of @xmath0 .
see lemma [ coisotropicorbitlem ] for alternative characterizations of our assumptions . in this paper
we will classify the compact connected symplectic manifolds with such torus actions , by constructing a list of explicit examples to which each of our manifolds is equivariantly symplectomorphic .
see theorem [ invariantthm ] , theorem [ existencethm ] and corollary [ crowncor ] for our main result . in many integrable systems in classical mechanics , we have an effective hamiltonian action of an @xmath8-dimensional torus on the @xmath5-dimensional symplectic manifold , but also non - hamiltonian actions occur in physics , see for instance novikov @xcite . if the effective action of @xmath0 on @xmath1 is hamiltonian , then @xmath9 and the principal orbits are lagrange submanifolds . moreover , the image of the momentum mapping is a convex polytope @xmath10 in the dual space @xmath11 of @xmath12 , where @xmath12 denotes the lie algebra of @xmath0 .
@xmath10 has the special property that at each vertex of @xmath10 there are precisely @xmath8 codimension one faces with normals which form a @xmath13-basis of the integral lattice @xmath14 in @xmath15 , where @xmath14 is defined as the kernel of the exponential mapping from @xmath12 to @xmath0 .
the classification of delzant @xcite says that for each such polytope @xmath10 there is a compact connected symplectic manifold with hamiltonian torus action having @xmath10 as image of the momentum mapping , and the symplectic manifold with torus action is unique up to equivariant symplectomorphisms .
such polytopes @xmath10 and corresponding symplectic @xmath0-manifolds @xmath16 are called _ delzant polytopes _ and _ delzant manifolds _ in the exposition of this subject by guillemin @xcite , after delzant @xcite .
each delzant manifold has a @xmath0-invariant khler structure such that the khler form is equal to @xmath17 . because critical points of the hamiltonian function correspond to zeros of the hamiltonian vector field , a hamiltonian action on a compact manifold always has fixed points .
therefore the other extreme case of a symplectic torus action with coisotropic principal orbits occurs if the action is _
free_. in this case , @xmath18 is a principal torus bundle over a torus , hence a nilmanifold for a two - step nilpotent lie group as described in palais and stewart @xcite .
if the nilpotent lie group is not commutative , then @xmath18 does not admit a khler structure , cf .
benson and gordon @xcite . for four - dimensional manifolds @xmath18 ,
these were the first examples of compact symplectic manifolds without khler structure , introduced by thurston @xcite .
see the end of remark [ freerem ] .
the general case is a combination of the hamiltonian case and the free case , in the sense that @xmath18 is an associated @xmath19-bundle @xmath20 over @xmath21 with a @xmath22-dimensional delzant submanifold @xmath23 of @xmath16 as fiber . here
@xmath24 is the unique maximal subtorus of @xmath0 which acts in hamiltonian fashion on @xmath1 .
it has dimension @xmath25 and its lie algebra is denoted by @xmath26 .
@xmath19 is a two - step nilpotent lie group , and @xmath27 is a commutative closed lie subgroup of @xmath19 , which acts on @xmath28 via @xmath29 .
the base space @xmath21 is a torus bundle over a torus , see remark [ horem ] .
this leads to an explicit model of @xmath16 in terms of the ingredients 1 )
6 ) in definition [ ingredientdef ] . see proposition [ aprop ] and proposition [ omegalem ] .
the model allows explicit computations of many aspects of @xmath16 . as an example we determine the fundamental group of @xmath18 in proposition [ pi1prop ] , and the chern classes of the normal bundle in @xmath30 of the fixed point set of the action of @xmath24 on @xmath30 in proposition [ chernprop ] . here
@xmath31 is a complementary subtorus to @xmath24 in @xmath0 , which acts freely on @xmath18 .
the main result of this paper is that the compact connected symplectic manifolds with symplectic torus action with coisotropic principal orbits are completely classified by the ingredients 1 )
6 ) in definition [ ingredientdef ] , see theorem [ invariantthm ] and theorem [ existencethm ] .
the proof starts with the observation that the symplectic form on the orbits is given by a two - form @xmath32 on @xmath12 , see lemma [ constlem ] .
write @xmath33 .
the inner product of the symplectic form @xmath17 with the infinitesimal action of @xmath0 defines a closed basic @xmath34-valued one - form @xmath35 on @xmath18 , which turns the orbit space @xmath3 into a locally convex polyhedral @xmath34-parallel space , as defined in definition [ cornerdef ] .
the locally convex polyhedral @xmath34-parallel space @xmath3 is isomorphic to @xmath36 , in which @xmath10 is a delzant polytope in @xmath37 and @xmath38 is a cocompact discrete additive subgroup of the space @xmath39 of all linear forms on @xmath40 which vanish on @xmath26 .
[ m / tprop ] .
the main step in the proof of the classification is the construction of lifts to @xmath18 of the constant vector fields on the @xmath34-parallel manifold @xmath3 with the simplest possible lie brackets and symplectic products of the lifts .
see proposition [ liftprop ] .
this construction uses calculations involving the de rham cohomology of @xmath3 .
all the proofs become much simpler in the case that the action of @xmath0 on @xmath18 is free .
we actually first analyzed the free case with lagrangian principal orbits , meaning that the principal orbits are lagrange submanifolds of @xmath18 .
next we treated the case with lagrangian principal orbits where @xmath18 is fibered by delzant manifolds , and only after we became aware of the article of benoist @xcite , we generalized our results to the case with coisotropic principal orbits . in @xcite ,
6.6 states that every compact connected symplectic manifold with a symplectic torus action with coisotropic principal orbits is isomorphic to the cartesian product of a delzant manifold and a compact connected symplectic manifold with a free symplectic torus action .
however , even in the special case that the principal orbits are langrange submanifolds of @xmath18 , this conclusion appears to be too strong , if the word `` isomorphic '' implies `` equivariantly diffeomorphic '' , see remark [ productrem ] and benoist @xcite .
the paper is organized as follows . in section [ stasec ]
we discuss the condition that some ( all ) principal orbits are coisotropic submanifolds of @xmath1 . in section [ m / tsec ] we analyze the space of @xmath0-orbits in all detail , where we use the definitions and theorems in the appendix section [ vsec ] concerning what we call `` @xmath4-parallel spaces '' .
section [ twolemmasec ] contains a lemma about basic differential forms and one about equivariant diffeomorphisms which preserve the orbits . in section [ liftsec ]
we construct our special lifts of constant vector fields on the orbit space .
these are used in section [ delzantsec ] in order to construct the delzant submanifolds of @xmath1 and in section [ normalformsec ] for the normal form of the symplectic @xmath0-manifold .
the classification is completed by means of the theorems in section [ invariantsec ] . in the first appendix , section [ vsec ] , we prove that every complete connected locally convex @xmath4-parallel space is isomorphic to the cartesian product of a closed convex subset of a finite - dimensional vector space and a torus
. see theorem [ cornerclassthm ] for the precise statement .
this result is a generalization of the theorem of tietze @xcite and nakajima @xcite , which states that every closed and connected locally convex subset of a finite - dimensional vector space is convex . in the second appendix , section [ sympltubes ] , we describe the local model of benoist ( * ? ? ?
1.9 ) and ortega and ratiu @xcite for a proper symplectic action of an arbitrary lie group on an arbitrary symplectic manifold .
there are many other texts on the classification of symplectic torus actions on compact manifolds which in some way are related to ours .
the book of audin @xcite is on hamiltonian torus actions , with emphasis on the topological aspects .
orlik and raymond @xcite and pao @xcite classified actions of two - dimensional tori on four - dimensional compact connected smooth manifolds . because they do not assume an invariant symplectic structure , our classification in the four - dimensional case forms only a tiny part of theirs . on the other hand the completely integrable systems with local torus actions of kogan
@xcite form a relatively close generalization of torus actions with lagrangian principal orbits .
the classification of hamiltonian circle actions on compact connected four - dimensional manifolds in karshon @xcite , and of centered complexity one hamiltonian torus actions in arbitrary dimensions in karshon and tolman @xcite , are also much richer than our classification in the case that @xmath41 .
mcduff @xcite and mcduff and salamon @xcite studied non - hamiltonian circle actions , and ginzburg @xcite non - hamiltonian symplectic actions of compact groups under the assumption of a `` lefschetz condition '' . in another direction symington @xcite and leung and symington @xcite classified four - dimensional compact connected symplectic manifolds which are fibered by lagrangian tori where however the fibration is allowed to have elliptic or focus - focus singularities .
we are very grateful to yael karshon for her suggestion of the problem .
a. pelayo thanks her for moral and intellectual support during this project .
let @xmath1 be a smooth compact and connected symplectic manifold and let @xmath0 be a torus which acts effectively on @xmath1 by means of symplectomorphisms . in this section we show that some principal @xmath0-orbit is a coisotropic submanifold of @xmath1 if and only if the poisson brackets of any pair of smooth @xmath0-invariant functions on @xmath18 vanish if and only if every principal @xmath0-orbit is a coisotropic submanifold of @xmath1 .
see lemma [ coisotropicorbitlem ] , remark [ multfreerem ] and remark [ orbtyperem ] below .
this follows from the local model of benoist ( * ? ? ?
1.9 ) , see theorem [ gthm ] , which in the case of symplectic torus actions with coisotropic principal orbits assumes a particularly simple form , see lemma [ modellem ] .
if @xmath42 is an element of the lie algebra @xmath12 of @xmath0 , then we denote by @xmath43 the infinitesimal action of @xmath42 on @xmath18 .
it is a smooth vector field on @xmath18 , and the invariance of @xmath17 under the action of @xmath0 implies that @xmath44 here @xmath45 denotes the lie derivative with respect to the vector field @xmath46 , and @xmath47 the inner product of a differential form @xmath48 with @xmath46 , obtained by inserting @xmath46 in the first slot of @xmath48 .
the first identity in ( [ lxsigma ] ) follows from the homotopy identity @xmath49 combined with @xmath50
. if @xmath51 is a smooth real - valued function on @xmath18 , then the unique vector field @xmath46 on @xmath18 such that @xmath52 is called the _ hamiltonian vector field _ of @xmath51 , and will be denoted by @xmath53 .
given @xmath46 , the function @xmath51 is uniquely determined up to and additive constant , which implies that @xmath51 is @xmath0-invariant if and only if @xmath46 is @xmath0-invariant . if @xmath54 , then @xmath43 is hamiltonian if and only if the closed two - form @xmath55 is exact .
the following lemma says that the pull - back to the @xmath0-orbits of the symplectic form @xmath17 on @xmath18 is given by a constant antisymmetric bilinear form on the lie algebra @xmath12 of @xmath0 .
there is a unique antisymmetric bilinear form @xmath56 on @xmath12 , such that such that @xmath57 for every @xmath58 and every @xmath59 .
[ constlem ] it follows from benoist ( * ? ? ?
* lemme 2.1 ) that if @xmath60 and @xmath46 are smooth vector fields on @xmath18 such that @xmath61 and @xmath62 , then @xmath63={\!\!\mathop{\rm ~ham}\nolimits}_{\sigma ( u,\ , v)}$ ] . we repeat the proof .
@xmath64}\sigma = { \!\!\mathop{\rm ~l}\nolimits}_u({\!\!\mathop{\rm ~i}\nolimits}_v\sigma ) = { \!\!\mathop{\rm ~i}\nolimits}_u({\!\!\mathop{\rm ~d}\nolimits}({\!\!\mathop{\rm ~i}\nolimits}_v\,\sigma ) ) + { \!\!\mathop{\rm ~d}\nolimits}({\!\!\mathop{\rm ~i}\nolimits}_u({\!\!\mathop{\rm ~i}\nolimits}_v\sigma ) ) = \ , -{\!\!\mathop{\rm ~d}\nolimits}(\sigma ( u,\ , v)).\ ] ] here we used @xmath61 in the first equality , the homotopy formula for the lie derivative in the second identity , and finally @xmath50 , the homotopy identity and @xmath62 in the third equality . applying this to @xmath65 , @xmath66 for @xmath58 , and using that @xmath67=0 $ ] ,
hence @xmath68=\ , -[x,\ , y]_m=0 $ ] , it follows that @xmath69 .
thus @xmath70 , and the function @xmath71 is constant on @xmath18 , because @xmath18 is connected . in the further discussion
we will need some basic facts about proper actions of lie groups , see for instance ( * ? ? ? * sec .
2.62.8 ) .
for each @xmath59 we write @xmath72 for the _ stabilizer subgroup _ of the @xmath0-action at the point @xmath73 .
@xmath74 is a closed lie subgroup of @xmath0 , it has finitely many components and its identity component is a torus subgroup of @xmath0 . the lie algebra @xmath75 of @xmath74 is equal to the space of all @xmath54 such that @xmath76 . in other words , @xmath75 is the kernel of the linear mapping @xmath77 from @xmath12 to @xmath78 .
the image of @xmath79 is equal to the tangent space at @xmath73 of the @xmath0-orbit through @xmath73 , and will be denoted by @xmath80 .
the linear mapping @xmath81 induces a linear isomorphism from @xmath82 onto @xmath80 . for each closed subgroup @xmath27 of @xmath0 which can occur as a stabilizer subgroup , the _ orbit type _
@xmath83 is defined as the set of all @xmath59 such that @xmath74 is conjugate to @xmath27 , but because @xmath0 is commutative this condition is equivalent to the equation @xmath84 . each connected component @xmath85 of @xmath83 is a smooth @xmath0-invariant submanifold of @xmath18 .
the connected components of the orbit types in @xmath18 form a finite partition of @xmath18 , which actually is a whitney stratification .
this is called the _ orbit type stratification _ of @xmath18 .
there is a unique open orbit type , called the _ principal orbit type _
, which is the orbit type of a subgroup @xmath27 which is contained in every stabilizer subgroup @xmath74 , @xmath59 .
because the effectiveness of the action means that the intersection of all the @xmath74 , @xmath59 is equal to the identity element , this means that the principal orbit type consists of the points @xmath73 where @xmath86 , that is where the action is free .
if the action is free at @xmath73 , then the linear mapping @xmath87 from @xmath12 to @xmath78 is injective .
the points @xmath59 at which the @xmath0-action is free are also called the _
regular points _ of @xmath18 , and the principal orbit type , the set of all regular points in @xmath18 is denoted by @xmath88 .
the principal orbit type @xmath88 is a dense open subset of @xmath18 , and connected because @xmath0 is connected , see ( * ? ?
the _ principal orbits _ are the orbits in @xmath88 , the principal orbit type . in our situation ,
the principal orbits are the orbits on which the action of @xmath0 is free .
let @xmath40 be the kernel in @xmath12 of the two - form @xmath32 on @xmath12 defined in lemma [ constlem ] , the set of all @xmath54 such that @xmath89 for every @xmath90 .
then @xmath91 for every @xmath59 .
[ txinslem ] if @xmath92 , then @xmath76 , hence @xmath93 for every @xmath90
. the linear subspace @xmath40 of @xmath12 will play an important part in the classification of the symplectic torus actions with coisotropic principal orbits .
a submanifold @xmath85 of @xmath18 is called _ coisotropic _ , if for every @xmath94 , @xmath95 , the condition that @xmath96 for every @xmath97 implies that @xmath98 . in other words , if the @xmath99-orthogonal complement @xmath100 of @xmath101 in @xmath78 is contained in @xmath101 .
every symplectic manifold has an even dimension , say @xmath5 , and if @xmath85 is a coisotropic submanifold of dimension @xmath102 , then @xmath103 shows that @xmath104 .
@xmath85 has the minimal dimension @xmath8 if and only if @xmath105 , if and only if @xmath85 is lagrange submanifold of @xmath18 , an isotropic submanifold of @xmath18 of maximal dimension @xmath8
. the next lemma is basically the implication ( iv ) @xmath106 ( ii ) in benoist ( * ? ? ?
let @xmath1 be a connected symplectic manifold , and @xmath0 a torus which acts effectively and symplectically on @xmath1 .
then every coisotropic @xmath0-orbit is a principal orbit .
furthermore , if some @xmath0-orbit is coisotropic , then every principal orbit is coisotropic , and @xmath107 .
[ coisotropicorbitlem ] we use theorem [ gthm ] with @xmath108 , where we note that the commutativity of @xmath0 implies that the adjoint action of @xmath109 on @xmath12 is trivial , which implies that the coadjoint action of @xmath27 on the component @xmath110 is trivial as well .
let us assume that the orbit @xmath111 is coisotropic , which means that @xmath112 , or equivalently the subspace @xmath113 defined in ( [ wdef ] ) is equal to zero .
this implies that the action of @xmath27 on @xmath114 is trivial , and the vector bundle @xmath115 is @xmath0-equivariantly isomorphic to @xmath116 , where @xmath0 acts by left multiplications on the first factor .
it follows that in the model all stabilizer subgroups are equal to @xmath27 , and therefore @xmath117 for all @xmath118 in the @xmath0-invariant open neighborhood @xmath119 of @xmath73 in @xmath18 . because the principal orbit type is dense in @xmath18 , there are @xmath120 such that @xmath121 , and it follows that @xmath122 , that is , @xmath111 is a principal orbit .
we note in passing that this implies that @xmath107 .
when @xmath123 , we read off from ( [ gsigma ] ) with @xmath124 given by @xmath32 in lemma [ constlem ] , and ( [ etadef ] ) , that the symplectic form @xmath125 is given by @xmath126 for all @xmath127 , and @xmath128 . in this model , the tangent space of the @xmath0-orbit is the set of all @xmath129 such that @xmath130 , of which the symplectic orthogonal complement is equal to the set of all @xmath131 such that @xmath132 and @xmath133 , which implies that in this model every @xmath0-orbit is coisotropic and therefore the orbit @xmath134 is coisotropic for every @xmath120 .
this shows that the set of all @xmath59 such that @xmath111 is coisotropic is an open subset of @xmath18 . because for all @xmath135 the tangent spaces of the orbits @xmath111 have the same dimension , equal to @xmath136 , the set of all @xmath135 such that @xmath111 is coisotropic is closed in @xmath88 .
because @xmath88 is connected , it follows that @xmath111 is coisotropic for all @xmath135 as soon as @xmath111 is coisotropic for some @xmath135 . in the proof of lemma [ coisotropicorbitlem ] , linear forms on @xmath110
were identified with linear forms on @xmath40 . for any linear subspace @xmath137 of a finite - dimensional vector space @xmath138
we have the canonical projection @xmath139 , and its dual mapping @xmath140 . because @xmath141 is surjective
, @xmath142 is injective , and its image @xmath143 is equal to the space @xmath144 of all @xmath145 such that @xmath146 .
this leads to a canonical linear isomorphism @xmath142 from @xmath147 onto @xmath144 , which will be used throughout this paper to identify @xmath147 with the linear subspace @xmath144 of @xmath148 .
[ e / frem ] let @xmath135 . because the principal orbit type @xmath88 is fibered by the @xmath0-orbits ,
the tangent space @xmath80 at @xmath73 of @xmath111 is equal to the common kernel of the @xmath149 , where @xmath51 ranges over the @xmath0-invariant smooth functions on @xmath18 . because @xmath150 , it follows that @xmath151 is equal to the set of all @xmath152 , @xmath153 . here
@xmath154 denotes the space of all @xmath0-invariant smooth functions on @xmath18 .
suppose that the principal orbits are coisotropic and let @xmath153 .
then we have for every @xmath135 that @xmath155 , or @xmath156 for a uniquely determined @xmath157 .
it follows that the @xmath53-flow leaves every principal orbit invariant , and because @xmath88 is dense in @xmath18 , the @xmath53-flow leaves every @xmath0-orbit invariant .
because a point @xmath59 is called a _ relative equilibrium _ of a @xmath0-invariant vector field @xmath46 if the @xmath46-flow leaves @xmath111 invariant , the conclusion is that all points of @xmath18 are relative equilibria of @xmath53 , and the induced flow in @xmath3 is at standstill .
moreover the @xmath0-invariance of @xmath53 implies that @xmath158 is constant on each principal @xmath0-orbit , which implies that the @xmath53-flow in @xmath88 is _ quasiperiodic _ , in the direction of the infinitesimal action of @xmath40 on @xmath88 .
if @xmath159 and @xmath135 , then @xmath152 and @xmath160 both belong to @xmath161 , and it follows that the _ poisson brackets _ @xmath162 of @xmath51 and @xmath163 vanish at @xmath73 . because @xmath88 is dense in @xmath18 , it follows that @xmath164 for all @xmath159 if the principal orbits are coisotropic . if conversely @xmath165 for all @xmath159 , then we have for every @xmath135 that @xmath166 , which means that @xmath111 is coisotropic . therefore the principal orbits are coisotropic if and only if the poisson brackets of all @xmath0-invariant smooth functions vanish . in guillemin and sternberg @xcite , a symplectic manifold with a hamiltonian action of an arbitrary compact lie group is called a _ multiplicity - free space _ if the poisson brackets of any pair of invariant smooth functions vanish .
because in @xcite the emphasis is on representations of noncommutative compact lie groups , which do not play a role in our paper , and because on the other hand we allow non - hamiltonian actions , we did not put the adjective `` multiplicity - free '' in the title .
[ multfreerem ] the next lemma is statement ( 1 ) ( a ) in benoist ( * ? ?
* lemma 6.7 ) . for general symplectic torus actions
the stabilizer subgroups need not be connected .
for instance , there exist symplectic torus actions with symplectic orbits and nontrivial finite stabilizer subgroups .
let @xmath1 be a connected symplectic manifold , and @xmath0 a torus which acts effectively and symplectically on @xmath1 , with coisotropic principal orbits .
then , for every @xmath59 , the stabilizer group @xmath74 is connected , that is , a subtorus of @xmath0 .
[ txconnectedlem ] as in the proof of lemma [ coisotropicorbitlem ] , we use theorem [ gthm ] with @xmath108 , where @xmath27 acts trivially on the factor @xmath167 in @xmath168 . recall that @xmath169 acts on @xmath170 by sending @xmath171 to @xmath172 .
when @xmath173 , then @xmath174 because @xmath0 is commutative , and we see that the action of @xmath27 on @xmath170 is represented by the linear symplectic action of @xmath27 on @xmath113 , where @xmath113 is defined by ( [ wdef ] ) . because @xmath175 and because the assumption that the principal orbits are coisotropic implies that @xmath176 , see lemma [ coisotropicorbitlem ] , it follows that @xmath177
write @xmath178 .
the action of the compact and commutative group @xmath27 by means of symplectic linear transformations on the @xmath179-dimensional symplectic vector space @xmath180 leads to a direct sum decomposition of @xmath113 into @xmath181 mutually @xmath182-orthogonal two - dimensional @xmath27-invariant linear subspaces @xmath183 , @xmath184 . for @xmath185 and every @xmath184 ,
let @xmath186 denote the restriction to @xmath187 of the action of @xmath188 on @xmath138 .
note that @xmath189 , because @xmath186 preserves the restriction to @xmath190 of @xmath182 , which is an area form on @xmath183 .
averaging any inner product in each @xmath183 over @xmath27 , we obtain an @xmath27-invariant inner product @xmath191 on @xmath183 , and @xmath192 is a homomorphism of lie groups from @xmath27 to @xmath193 , the group of linear transformations of @xmath183 which preserve both @xmath191 and the orientation . on the other hand , if @xmath185 and @xmath194 , then @xmath195 therefore @xmath196 for all @xmath184 implies that @xmath197 , hence @xmath198 .
this implies that the homomorphism of lie groups @xmath199 , defined by @xmath200 is injective .
because both the source group @xmath27 and the target group are @xmath181-dimensional lie groups , and the target group is connected , it follows that @xmath199 is an isomorphism of lie groups .
this implies in turn that @xmath27 is connected .
the @xmath27-invariant inner product @xmath191 on @xmath183 , introduced in the proof of lemma [ txconnectedlem ] , is unique , if we also require that the symplectic inner product of any orthonormal basis with respect to @xmath182 is equal to @xmath201 . in turn
this leads to the existence of a unique complex structure on @xmath183 such that , for any unit vector @xmath202 in @xmath203 , we have that @xmath202 , @xmath204 is an orthonormal basis in @xmath205 and @xmath206 . here
this leads to an identification of @xmath183 with @xmath208 , which is unique up to multiplication by an element of @xmath209 . in turn
this leads to an identification of @xmath113 with @xmath210 , with the symplectic form @xmath182 defined by @xmath211 the element @xmath212 acts on @xmath213 by sending @xmath214 to the element @xmath215 such that @xmath216 for every @xmath184 .
there is a unique isomorphism of lie groups @xmath217 such that @xmath185 acts on @xmath218 by sending @xmath214 to @xmath219 .
the identification of @xmath113 with @xmath213 is unique up to a permutation of the coordinates and the action of an element of @xmath220 .
[ complexrem ] in the local model of lemma [ modellem ] below , we will use that any subtorus of a torus has a complementary subtorus , in the following sense .
let @xmath119 be a @xmath221-dimensional subtorus of a @xmath7-dimensional torus @xmath0 .
let @xmath222 and @xmath14 denote the integral lattice , the kernel of the exponential mapping , in the lie algebra @xmath223 and @xmath12 of @xmath119 and @xmath0 , respectively .
let @xmath224 , @xmath225 , be a @xmath13-basis of @xmath222 .
then there are @xmath226 , @xmath227 , such that the @xmath224 and @xmath226 together form a @xmath13-basis of @xmath14 .
if we denote by @xmath228 the span of the @xmath226 , then @xmath229 is a subtorus of @xmath0 with lie algebra equal to @xmath228 .
@xmath4 is a complementary subtorus of @xmath119 in @xmath0 in the sense that the mapping @xmath230 is an isomorphism from @xmath231 onto @xmath0 .
the @xmath226 form a @xmath13-basis of the integral lattice @xmath232 in the lie algebra @xmath228 of @xmath4 .
[ tbaselem ] we repeat the well - known argument . if @xmath233 , @xmath234 , @xmath235 , and @xmath236 , then @xmath237 and @xmath238 in @xmath0 , hence @xmath238 in @xmath119 , and it follows that @xmath239 .
this means that the finitely generated commutative group @xmath240 is torsion - free , and therefore has a @xmath13-basis @xmath241 , @xmath242 , cf .
hungerford ( * ? ? ?
* th . 6.6 on p. 221 ) .
we have that @xmath243 for some @xmath244 .
if @xmath233 , then there are unique @xmath245 such that @xmath246 , which means that @xmath247 . but
this implies that there are unique @xmath248 such that @xmath249 , which shows that the @xmath224 and @xmath226 together form a @xmath13basis of @xmath14 , which in turn implies that @xmath250 .
the last statement follows from the fact that the mapping @xmath251 from @xmath252 to @xmath0 induces an isomorphism from @xmath253 onto @xmath0 which maps @xmath254 onto @xmath119 and @xmath255 onto @xmath4 .
the complementary subtorus @xmath4 in lemma [ tbaselem ] is by no means unique .
the @xmath226 can be replaced by any @xmath256 in which the @xmath257 are integers .
this leads to a bijective correspondence between the set of all complementary subtori of a given subtorus @xmath119 and the set of all @xmath258-matrices with integral coefficients .
[ tbaserem ] let @xmath109 be the subtorus of @xmath0 in lemma [ txconnectedlem ] .
let @xmath259 be a complementary subtorus of @xmath27 in @xmath0 and , for any @xmath169 , let @xmath260 and @xmath261 be the unique elements in @xmath27 and @xmath259 , respectively , such that @xmath262 .
let @xmath263 be a linear projection from @xmath12 onto @xmath40 .
we also use the identification of @xmath113 with @xmath213 as in remark [ complexrem ] . with these notations
, we have the following local model for our symplectic @xmath0-space with coisotropic principal orbits . under the assumptions of lemma [ txconnectedlem ]
, there is an isomorphism of lie groups @xmath199 from @xmath27 onto @xmath220 , an open @xmath220-invariant neighborhood @xmath264 of the origin in @xmath265 , and a @xmath0-equivariant diffeomorphism @xmath266 from @xmath267 onto an open @xmath0-invariant neighborhood @xmath119 of @xmath73 in @xmath18 , such that @xmath268 . here
@xmath169 acts on @xmath269 by sending @xmath270 to @xmath271 .
in addition , the symplectic form @xmath125 on @xmath267 is given by @xmath272 for all @xmath273 , and @xmath274 .
here we identify each tangent space of the torus @xmath259 with @xmath275 and each tangent space of a vector space with the vector space itself .
finally , @xmath276 is the symplectic form on @xmath213 defined in ( [ sigmaw ] ) .
[ modellem ] as in the proof of lemma [ txconnectedlem ] , we use theorem [ gthm ] with @xmath108 , where @xmath27 acts trivially on the factor @xmath167 and @xmath185 acts on @xmath218 by sending @xmath214 to @xmath219 . here
@xmath217 is the isomorphism from the torus @xmath27 onto the standard torus @xmath220 introduced in remark [ complexrem ] , and the symplectic form @xmath276 on @xmath213 is given by ( [ sigmaw ] ) . because @xmath259 is a complementary subtorus of @xmath27 in @xmath0 ,
the manifold @xmath277 is a global section of the vector bundle @xmath278 . indeed ,
if @xmath279 , then @xmath280 is the unique element in @xmath281 .
furthermore , if @xmath169 and @xmath282 , then @xmath283 is the unique element in @xmath284 .
this exhibits @xmath170 as a trivial vector bundle over @xmath259 , which is a homogeneous @xmath0-bundle , where @xmath169 acts on @xmath277 by sending @xmath285 to @xmath283 .
finally , if in ( [ etadef ] ) we restrict ourselves to @xmath286 , then the right hand side simplifies to @xmath287 , which leads to ( [ modelsigma ] ) . in the local model of lemma
[ modellem ] , we have that @xmath288 if and only if @xmath289 is a fixed point of @xmath290 if and only if @xmath291 . because @xmath292 is a symplectic submanifold of @xmath293 , it follows that every orbit type is a smooth symplectic submanifold of @xmath1 .
moreover , @xmath294 is a coisotropic submanifold of @xmath293 , and we conclude that every @xmath0-orbit is a coisotropic submanifold of its orbit type .
the discussion of the relative equilibria in remark [ multfreerem ] , with @xmath88 replaced by any orbit type @xmath83 , leads to the conclusion that for every @xmath153 the flow of the hamiltonian vector field @xmath53 in @xmath83 is quasiperiodic , in the direction of the infinitesimal action of @xmath110 in @xmath83 .
[ orbtyperem ] we conclude this section with a discussion of the special case that the two - form @xmath56 in lemma [ constlem ] is equal to zero
. we have @xmath295 if and only if @xmath296 if and only if some @xmath0-orbit is isotropic if and only if every @xmath0-orbit is isotropic . also , every principal orbit is a lagrange submanifold of @xmath1 if and only if some principal orbit is a lagrange submanifold of @xmath1 if and only if @xmath297 and @xmath298 .
[ isotropicorbitlem ] the equivalence of @xmath295 and @xmath299 is obvious , whereas the equivalence between @xmath295 and the isotropy of some ( every ) @xmath0-orbit follows from lemma [ constlem ] . if @xmath135 and @xmath111 is a lagrange submanifold of @xmath1 , then @xmath300 , and @xmath298 follows in view of the first statement in the lemma .
conversely , if @xmath297 and @xmath298 , then every orbit is isotropic and for every @xmath135 we have @xmath301 , which implies that @xmath111 is a lagrange submanifold of @xmath1 .
in this section we investigate the orbit space of our action of the torus @xmath0 on the compact connected symplectic manifold @xmath1 with coisotropic principal orbits .
the main results are that the closed basic one - form @xmath35 of lemma [ alphalem ] exhibits the orbit space as a locally convex polyhedral @xmath34-parallel space , see definition [ cornerdef ] and lemma [ parallellem ] , and that as such @xmath3 is isomorphic to the cartesian product of a delzant polytope and a torus , see proposition [ m / tprop ] .
the assumption that the principal orbits are coisotropic will be assumed throughout this section , unless explicitly stated otherwise . in this subsection
we exhibit the space of @xmath0-orbits as an @xmath34-parallel space in the sense of definition [ cornerdef ] .
we denote the space of all orbits in @xmath18 of the @xmath0-action by @xmath3 , and by @xmath302 the canonical projection which assigns to each @xmath59 the orbit @xmath111 through the point @xmath73 .
the orbit space is provided with the maximal topology for which the canonical projection is continuous ; this topology is hausdorff . for each connected component @xmath85 of an orbit type @xmath83 in @xmath18 of the subgroup @xmath27 of @xmath0 , as introduced in the paragraphs preceding lemma [ txinslem ]
, the action of @xmath0 on @xmath85 induces a proper and free action of the torus @xmath303 on @xmath85 , and @xmath304 has a unique structure of a smooth manifold such that @xmath305 is a principal @xmath303-bundle .
@xmath306 is called the _ orbit type of @xmath27 in @xmath3 _ and @xmath304 is a connected component of @xmath307 . the connected components of the orbit types in the orbit space form a finite stratification of the orbit space , cf .
2.7 ) . although @xmath3 is equal to the union of the finitely many strata of the orbit type stratification in @xmath3 , where each of these strata is a smooth manifold , the orbit space @xmath3 is not a smooth manifold , unless the action of @xmath0 on @xmath18 is free .
in general the principal orbit type @xmath308 is a smooth manifold of dimension @xmath309 , which is an open and dense subset of @xmath3,and @xmath3 will have singularities at the lower dimensional strata , the strata in the complement of @xmath310 in @xmath3 .
however , in this section we will obtain a much more explicit description of the orbit space @xmath3 .
a smooth differential form @xmath48 on @xmath18 is called _ basic _ with respect to the @xmath0-action if it is @xmath0-invariant , that is @xmath311 for every @xmath54 , and if @xmath312 for every @xmath54 .
the basic differential forms constitute a module over the algebra @xmath154 of @xmath0-invariant smooth functions on @xmath18 , the basic forms of degree zero on @xmath18 .
a smooth differential form @xmath48 on @xmath18 is basic if and only if the restriction of @xmath48 to the principal orbit type is equal to @xmath313 for a smooth differential form @xmath314 on the principal orbit type in @xmath3 .
a theorem of koszul @xcite says that the ech (= sheaf ) cohomology group @xmath315 of @xmath3 is canonically isomorphic to the de rham cohomology of the basic forms on @xmath18 , that is , the space of closed basic @xmath102-forms on @xmath18 modulo its subspace consisting of the @xmath316 in which @xmath314 ranges over the basic @xmath317-forms on @xmath18 .
this theorem holds for any proper action of a lie group on any smooth manifold , and in particularly it does not need the compactness of @xmath18 .
recall that @xmath40 is the kernel of the antisymmetric bilinear form @xmath32 which had been introduced in lemma [ constlem ] . for each @xmath132
, @xmath318 is a closed basic one - form on @xmath18 .
[ alphalem ] that @xmath319 is closed follows from ( [ lxsigma ] ) . because @xmath132 , we have for each @xmath320 that @xmath321 . also we have for every @xmath90 that @xmath322}\sigma + { \!\!\mathop{\rm ~i}\nolimits}_{x_m}({\!\!\mathop{\rm ~l}\nolimits}_{y_m}\sigma ) = 0.\ ] ] here we have used the leibniz identity for the lie derivative , the commutativity of @xmath15 , and the @xmath0-invariance of @xmath17 which implies that @xmath323 . for each @xmath59 , @xmath324 is a linear form on @xmath78 which depends linearly on @xmath132 , and therefore @xmath325 is an @xmath34-valued linear form on @xmath78 , which we denote by @xmath326 . in this way
@xmath327 is an @xmath34-valued one - form on @xmath18 , which we denote by @xmath35 . with these conventions
, we have @xmath328 note that the @xmath34-valued one - form @xmath35 on @xmath18 is basic and closed .
let @xmath54 and suppose that @xmath329 for some @xmath330 .
then we have for every @xmath90 that @xmath331 and it follows that @xmath153 if and only if @xmath332 . the @xmath0-action on @xmath1
is called a _ hamiltonian @xmath0-action _ if for every @xmath54 there exists an @xmath153 such that @xmath329 .
note that if @xmath333 , that is , if @xmath295 , then the @xmath0-action is hamiltonian if and only if if for every @xmath54 there exists an @xmath334 such that @xmath329 .
we recall the delzant manifolds , mentioned in section [ intrsec ] .
koszul s theorem now implies the following .
we do not assume that the principal orbits are coisotropic . let @xmath54 .
then @xmath329 for some @xmath153 , if and only if @xmath335 and the cohomology class @xmath336\in { \!\!\mathop{\rm ~h}\nolimits}^1(m / t,\,{{\mathbb r}})$ ] is equal to zero .
if the @xmath0-action is hamiltonian , then @xmath295 . finally , if @xmath295 and @xmath337 , then the @xmath0-action is hamiltonian and @xmath16 is a delzant manifold .
[ hamcor ] in the local model of lemma [ modellem ] , the @xmath0-orbit space of @xmath267 is equal to @xmath338 , which is contractible by using the radial contractions in @xmath264 .
it follows that for every @xmath339 there is a @xmath0-invariant open neighborhood @xmath119 of @xmath340 in @xmath18 such that the open subset @xmath341 of the orbit space @xmath3 is contractible . because of koszul s theorem , and because the ech cohomology of @xmath341 is trivial , it follows that the infinitesimal action of @xmath40 on @xmath119 is hamiltonian .
therefore , if @xmath295 , then the @xmath0-action is _ locally hamiltonian _ in the sense that every element in @xmath18 has a @xmath0-invariant open neighborhood in @xmath18 on which the @xmath0-action is hamiltonian .
[ lochamrem ] in the local model of lemma [ modellem ] , we write @xmath342 with @xmath343 for each @xmath184 .
then the symplectic form @xmath182 with @xmath218 in ( [ sigmaw ] ) is equal to @xmath344 the mapping @xmath345 induces a homeomorphism from the @xmath0-orbit space @xmath346 onto @xmath347 , in which @xmath348 note that @xmath349 acts on @xmath213 by sending @xmath350 to @xmath351 and leaving @xmath352 fixed .
if we identify the lie algebra of @xmath220 with @xmath353 , then the infinitesimal action of @xmath354 in @xmath355-coordinates is equal to the constant vector field @xmath356 .
the tangent mapping at @xmath357 of the isomorphism @xmath217 is a linear isomorphism from @xmath358 onto @xmath353 , which we we also denote by @xmath199 .
for every @xmath90 , the infinitesimal action @xmath359 of @xmath360 on @xmath361 is equal to the vector field @xmath362 , see the description of the action of @xmath0 on the model in lemma [ modellem ] .
write @xmath363 with @xmath364 and @xmath365 . here
@xmath275 denotes the lie algebra of the complementary torus @xmath259 to @xmath27 in @xmath0 , which implies that @xmath366 . because @xmath367 , we have @xmath368 and @xmath369 . because @xmath370 is a linear form on @xmath40 which is equal to zero on @xmath358 , it follows that @xmath371 therefore , if in ( [ modelsigma ] ) we substitute @xmath372 with @xmath373 , then we obtain @xmath374 here we have used that @xmath375 implies that @xmath376 .
furthermore ( [ gotkgotl ] ) with @xmath373 implies that @xmath377 .
finally the formula for the @xmath276-term follows from ( [ rhodef ] ) , as @xmath378 , @xmath379 , and @xmath380 because the infinitesimal action of @xmath381 is equal to @xmath382 .
consider the linear mapping @xmath383 \label{aiota}\ ] ] from @xmath384 onto @xmath34 .
@xmath385 is a linear isomorphism , because the source space and the target space have the same dimension , and @xmath386 : testing with arbitrary @xmath387 yields that @xmath388 , and then testing with arbitrary @xmath373 yields that @xmath133 .
let @xmath389 denote the element of @xmath367 such that @xmath390 , in which @xmath202 denotes the @xmath391-th standard basis vector in @xmath392 .
note that the @xmath393 , @xmath184 , form a @xmath13-basis of the integral lattice of the lie algebra of @xmath220 , and because @xmath217 is an isomorphism of tori , it follows that the @xmath389 , @xmath184 , form a @xmath13-basis of the integral lattice of the lie algebra @xmath358 of @xmath27 .
also note that @xmath394 here it is essential that we use the coordinates @xmath395 instead of their infinitesimal displacements @xmath396 , because in lemma [ chartlem ] below we are interested in the the consequences of the inequalities @xmath397 .
this leads to the following conclusion .
let @xmath266 be the @xmath0-equivariant symplectomorphism from @xmath398 onto the open @xmath0-invariant neighborhood @xmath119 of @xmath73 in @xmath18 as introduced in lemma [ modellem ] .
then the smooth mapping @xmath399 , which consists of @xmath400 , followed by the @xmath401-map and then @xmath385 , induces a homeomorphism @xmath402 from @xmath403 onto an open neighborhood of @xmath404 in the corner @xmath405 in @xmath34 , such that @xmath406 . here
the @xmath389 , @xmath184 , form a @xmath13-basis of the integral lattice of the lie algebra @xmath407 of @xmath27 .
[ chartlem ] for every @xmath373 , the right hand side of ( [ iysigma ] ) is equal to @xmath408 .
combined with the definitions of @xmath409 and @xmath385 , this yields that @xmath406 . because @xmath410 is a homeomorphism from @xmath411 onto @xmath412
, @xmath402 is a homeomorphism from @xmath403 onto an open neighborhood of 0 in the corner in @xmath34 which is determined by the inequalities @xmath413 , @xmath184 .
if @xmath414 is mapping as in lemma [ chartlem ] , with corresponding chart @xmath415 , then @xmath416 shows that @xmath417 is locally constant on @xmath418 , which implies that @xmath419 is locally constant on @xmath420 . in terms of definition [ cornerdef ] , we have proved with the @xmath402 of lemma [ chartlem ] as local charts on @xmath3 , the orbit space @xmath3 is a locally convex polyhedral @xmath34-parallel space .
the linear forms @xmath421 , @xmath184 , in definition [ cornerdef ] are the @xmath422 , where the @xmath389 , @xmath184 , form a @xmath13-basis of the integral lattice of the lie algebra @xmath367 of a stabilizer group @xmath109 of an element @xmath59 .
[ parallellem ] in the next lemma we will introduce the subtorus @xmath24 of @xmath0 which later will turn out to be the unique maximal subtorus of @xmath0 which acts on @xmath18 in a hamiltonian fashion . for this reason @xmath24 will be called the _
hamiltonian torus_. there are only finitely many different stabilizer subgroups of @xmath0 , each of which is a subtorus of @xmath0 .
the product @xmath24 of all the different stabilizer subgroups is a subtorus of @xmath0 , and the lie algebra @xmath26 of @xmath24 is equal to the sum of the lie algebras of all the different stabilizer subgroups of @xmath0 .
it follows from lemma [ txinslem ] that @xmath423 .
[ stabilizerlem ] in the local model of lemma [ modellem ] , the stabilizer subgroup of @xmath270 is equal to the set of all @xmath185 such that @xmath424 for every @xmath391 such that @xmath425 .
it follows that we have @xmath426 different stabilizer subgroups @xmath427 , @xmath120 , namely one for each subset of @xmath428 . because @xmath18 is compact ,
is follows that there are only finitely many different stabilizer subgroups of @xmath0 . for the last statement
we observe that the product of finitely many subtori is a compact and connected subgroup of @xmath0 and therefore a subtorus of @xmath0 .
also the image under the exponential mapping of the sum of the finitely many different lie algebras of the stabilizer subgroups of @xmath0 is equal to @xmath24 , which proves that the lie algebra of @xmath24 is equal to the sum of the finitely many different @xmath75 , @xmath59 .
see for instance ( * ? ? ?
1.12 ) for the general facts about lie subgroups of tori , which we have used here .
the orbit @xmath429 of any @xmath135 is called a _ regular point of @xmath3_. recall from the paragraph preceding lemma [ txinslem ] that @xmath135 if and only if @xmath430 if and only if @xmath431 .
therefore the set @xmath310 of all regular points in @xmath3 is just the principal orbit type , which is a smooth manifold of dimension @xmath309 . in the local model of lemma [ modellem ] with @xmath135 , where @xmath432 and @xmath433 , at each point the @xmath34-valued one - form @xmath35 corresponds to the projection @xmath434 , and
@xmath435 is equal to the tangent space of the @xmath0-orbit .
it follows that for every @xmath436 the induced linear mapping @xmath437 is a linear isomorphism .
more generally , the orbit type stratification , introduced in the paragraph preceding lemma [ txinslem ] , leads to a corresponding decomposition of @xmath3 .
the strata for the @xmath0-action in @xmath438 are of the form @xmath439 in which @xmath440 is a subset of @xmath441 and @xmath439 is the set of all @xmath270 such that @xmath442 if and only if @xmath443 . in terms of the @xmath444-coordinates
, this corresponds to @xmath445 for all @xmath443 and @xmath446 for @xmath447 .
the lie algebra of the corresponding stabilizer subgroup of @xmath0 corresponds to the span of the vector fields @xmath448 with @xmath443 .
therefore , if @xmath449 is a connected component of the orbit type in @xmath3 defined by the subtorus @xmath27 of @xmath0 with lie algebra @xmath358 , then for each @xmath450 we have @xmath451 for all @xmath452 , and @xmath453 may be viewed as an element of @xmath454 , the set of all linear forms on @xmath40 which vanish on @xmath358 , see remark [ e / frem ] .
the linear mapping @xmath455 is a linear isomorphism .
[ regrem ] in the following proposition [ m / tprop ] , the orbit space @xmath3 is viewed as a locally convex polyhedral @xmath34-parallel space , as in definition [ cornerdef ] with @xmath456 and @xmath457 .
see lemma [ parallellem ] .
let the subset @xmath458 of @xmath459 and the mapping @xmath460 from @xmath458 to @xmath3 be defined as in definition [ ddef ] .
we have the linear subspace @xmath39 of @xmath457 , which acts on @xmath456 by means of translations , and the period group @xmath38 of the @xmath39-action on @xmath461 , as defined in lemma [ nlem ] and lemma [ perlem ] , respectively . with the choice of a base point @xmath462 , we write @xmath463 .
let @xmath464 be a linear complement of @xmath26 in @xmath12 and let @xmath462 .
with these definitions , and the identification of @xmath465 with the space of linear forms on @xmath40 which vanish on @xmath26 , see remark [ e / frem ] , we have the following conclusions .
let @xmath85 be a linear complement of @xmath465 in @xmath34 .
* @xmath466 , @xmath38 is a cocompact discrete subgroup of the additive group @xmath39 , and @xmath467 is a @xmath468-dimensional torus .
* there is a delzant polytope @xmath10 in @xmath469 , such that @xmath470 . *
the mapping @xmath471 is an isomorphism of locally convex polyhedral @xmath34-parallel spaces from @xmath36 onto @xmath3 .
[ m / tprop ] the linear forms @xmath472 which appear in the characterization of @xmath39 in theorem [ cornerclassthm ] are equal to the collection of all the @xmath473 which appear in @xmath13-bases of integral lattices of lie algebras @xmath358 of stabilizer subgroups @xmath27 of @xmath0 . because @xmath39 is equal to the common kernel of all the @xmath472
, @xmath39 is equal to the set @xmath465 of all elements of @xmath34 which vanish on the sum @xmath474 of the finitely many different lie algebras @xmath358 of stabilizer subgroups of @xmath0 . because @xmath85 is a linear complement of @xmath465 in @xmath34 , the mapping @xmath475 induces an isomorphism from @xmath85 onto @xmath37 .
@xmath10 is a delzant polytope in @xmath37 in the sense of guillemin @xcite , because each @xmath13-basis of the integral lattice of @xmath75 can be extended to a @xmath13-basis of the integral lattice of @xmath26 , see lemma [ tbaselem ] . because @xmath85 is a linear complement of @xmath476 in @xmath34 , proposition [ m / tprop ] now follows from lemma [ parallellem ] and theorem [ cornerclassthm ] .
let @xmath1 be a compact connected @xmath5-dimensional symplectic manifold and suppose that we have an effective symplectic action of an @xmath8-dimensional torus @xmath0 on @xmath1 , where we do not assume that the principal orbits are coisotropic .
then the following conditions are equivalent . *
the action of @xmath0 has a fixed point in @xmath18 . *
the sum of the lie algebras of all the different stabilizer subgroups of @xmath0 is equal to the lie algebra of @xmath0 . *
@xmath295 and @xmath3 is homeomorphic to a convex polytope . *
@xmath295 and @xmath337 . *
the action of @xmath0 is hamiltonian .
[ delzantprop ] if @xmath73 is a fixed point , then @xmath477 , hence @xmath478 , which implies ii ) .
write @xmath479 for the sum of the lie algebras of all the different stabilizer subgroups of @xmath0 .
if @xmath92 , then @xmath76 and it follows from lemma [ constlem ] that @xmath480 for every @xmath90 .
this shows that @xmath480 for every @xmath481 and every @xmath90 .
now ii ) means that @xmath482 , hence @xmath298 , and lemma [ isotropicorbitlem ] implies that every principal orbit is a lagrange submanifold of @xmath1 , and therefore coisotropic .
it follows that we may apply proposition [ m / tprop ] with @xmath483 , and conclude that @xmath484 is a homeomorphism from the delzant polytope @xmath10 onto @xmath3 .
\iii ) @xmath106 iv ) because any convex polytope is contractible .
\iv ) @xmath106 v ) follows from corollary [ hamcor ] .
finally v ) @xmath106 i ) follows from the fact that the image of the momentum mapping is equal to the convex hull of the images under the momentum mapping of the fixed points , cf .
atiyah ( * ? ? ?
1 ) or guillemin and sternberg ( * ? ? ?
the implication i ) @xmath106 v ) has also been obtained by giacobbe ( * ? ? ?
* th . 3.13 ) .
note that if the conditions i ) v ) in corollary [ delzantprop ] hold , then @xmath1 together with the @xmath0-action on @xmath18 is a delzant manifold , and @xmath3 is the corresponding delzant polytope .
if a compact lie group @xmath259 acts linearly and continuously on a vector space @xmath4 , then the _ average _ of @xmath485 over @xmath259 is defined as @xmath486 in which @xmath181 denotes any haar measure on @xmath259 . with the notation of proposition _ [ m / tprop ]
_ , let @xmath487 be the mapping @xmath488 followed by the projection from @xmath36 onto the second factor .
let @xmath489 be defined by @xmath490 .
then we have the following conclusions . for each nonnegative integer @xmath102 ,
the mapping @xmath491 is an isomorphism , with inverse equal to @xmath492 .
the mapping which assigns to any @xmath493 the cohomology class of the constant @xmath102-form @xmath494 on @xmath467 is an isomorphism from @xmath495 onto @xmath496 , and every closed @xmath102-form on @xmath467 is cohomologous to its average over the torus @xmath467 .
[ cohomcor ] the first statement follows because @xmath10 is a convex subset of @xmath11 and hence it is contractible .
the second statement is a well - known characterization of the cohomology of tori .
the fact that a closed differential form on a compact connected lie group is cohomologous to its average goes back to lie cartan @xcite .
any finite - dimensional vector space @xmath113 carries a positive translation - invariant measure @xmath181 , which is unique up to a positive factor . for any non - negligible compact subset @xmath385 of @xmath113 , the _ center of mass _ of @xmath385
is defined as @xmath497 which is independent of the choice of the positive translation - invariant measure @xmath181 on @xmath113 .
. then @xmath43 is hamiltonian if and only if @xmath498 .
furthermore , the image of any momentum mapping of the hamiltonian action of @xmath24 on @xmath18 is equal to a translate of the delzant polytope @xmath10 in proposition _ [ m / tprop ] _ , where we note that any two momentum mappings for the same torus action differ by a constant element of @xmath499 .
the translational ambiguity of @xmath10 can be removed by putting the center of mass of @xmath10 at the origin . here
a momentum mapping for the hamiltonian action of @xmath24 is a smooth @xmath499-valued function @xmath500 on @xmath18 such that for every @xmath498 the @xmath42-component of @xmath501 is equal to @xmath502 .
[ tstabcor ] it follows from corollary [ hamcor ] and corollary [ cohomcor ] that the vector field @xmath43 is hamiltonian if and only if @xmath336=0 $ ] if and only if @xmath503=\iota_p^*[\widehat{\sigma}(x)]=0 $ ] . now constant one - forms on @xmath467 are canonically identified with linear forms on @xmath466 , which are identified with elements of @xmath504 . with this identification , @xmath505 corresponds to @xmath506 , which is equal to zero if and only if @xmath498 .
the second statement in the corollary follows from the fact that if @xmath500 is a momentum mapping for the hamiltonian @xmath24-action , then @xmath507 for every @xmath498 . in other words
, @xmath500 differs from the @xmath26-component of any canonical local chart on @xmath3 by a constant vector in @xmath499 .
therefore the image of @xmath500 corresponds to @xmath508 , the orbit space of the translational @xmath39-action on @xmath3 .
here we use that restriction to @xmath26 of linear forms on @xmath40 leads to a canonical identification of @xmath509 with @xmath499 .
mcduff @xcite proved that a symplectic circle action on a four - dimensional compact connected symplectic manifold is hamiltonian , if and only if it has a fixed point , but that in higher dimensions there exist non - hamiltonian symplectic circle actions with fixed points .
corollary [ tstabcor ] follows from @xcite if @xmath510 , but not if @xmath511 .
our proof of corollary [ tstabcor ] uses in an essential way that @xmath43 is an infinitesimal action of a symplectic action of an @xmath8-dimensional torus with a lagrange orbit . because a hamiltonian torus action has fixed points , it follows from corollary [ tstabcor ] that the action of @xmath24 on @xmath18 has fixed points , that is , there exist @xmath59 such that @xmath512 , hence @xmath513 because the definition of @xmath24 in lemma [ stabilizerlem ] implies that @xmath514 for every @xmath59 .
in other words , @xmath24 can also be characterized as the unique maximal stabilizer subgroup of @xmath0 .
actually the fixed points in @xmath18 for the action of @xmath24 are the @xmath59 such that @xmath515 is a vertex of the delzant polytope @xmath10 , where @xmath516 denotes the momentum map of the hamiltonian @xmath24-action .
[ t1rem ] let @xmath517 .
it follows from lemma [ stabilizerlem ] that for every @xmath518 the vector field @xmath43 has no zeros in @xmath18 , and we conclude that the euler characteristic @xmath519 of @xmath18 is equal to zero .
furthermore the localization formula of berline - vergne and atiyah - bott in equivariant cohomology , in the form of ( * ? ? ?
* ( 4.13 ) ) , yields for every @xmath0-equivariantly closed @xmath0-equivariant differential form @xmath48 on @xmath18 that the integral of @xmath48 over @xmath18 is equal to zero , when evaluated at @xmath518 .
because @xmath520 is dense in @xmath12 , it follows that the integral over @xmath18 of each @xmath0-equivariantly closed @xmath0-equivariant differential form is identically equal to zero . if @xmath498 , then lemma [ tstabcor ] implies that @xmath43 is hamiltonian , and the zeros of @xmath43 are the critical points of its hamiltonian function , which form a non - empty subset of @xmath18 . in this case the localization formula ( * ? ? ? * ( 4.13 ) ) yields that the sum over the connected components @xmath137 of the zeroset of @xmath43 of the integrals over @xmath137 of @xmath521 is equal to zero .
the generalization of ginzburg ( * ? ? ?
6.1 ) of the duistermaat - heckman formula is related to these observations . on the other hand
the integral over @xmath18 of a @xmath24-equivariantly closed @xmath24-equivariant differential form , such as @xmath522 , is usually nonzero .
if @xmath523 , then it follows from corollary [ delzantprop ] and corollary [ hamcor ] that @xmath16 is a delzant manifold , and @xmath519 is equal to the number of vertices of the delzant polytope @xmath10 .
this can be proved by observing that for a generic @xmath54 the momentum map is bijective from the zeroset of @xmath43 to the set of vertices of @xmath10 , and each zero of @xmath43 has poincar index equal to one .
see also guillemin ( * ? ? ?
* exerc . 4.15 ) .
[ eulerrem ]
the following lemmas will be used later in the paper .
lemma [ basicformlem ] is used in the proof of proposition [ liftprop ] , whereas lemma [ hslem ] is used in the proof of lemma [ lphilem ] and lemma [ tplem ] .
the proofs of lemma [ basicformlem ] and lemma [ tplem ] are based on the local models of lemma [ modellem ] . throughout this section
, @xmath1 is a symplectic manifold with an effective symplectic action of a torus @xmath0 with coisotropic principal orbits .
let @xmath389 , @xmath524 , be a basis of @xmath40 .
the basic @xmath102-forms on @xmath18 are the @xmath102-forms @xmath525 in which @xmath526 and @xmath35 is defined as in lemma [ alphalem ] .
[ basicformlem ] because the one - forms @xmath527 are basic , any @xmath48 as in ( [ basicform ] ) with @xmath526 is a basic form .
using partitions of unity with elements of @xmath154 , it is sufficient to prove the converse statement in a local model as in lemma [ modellem ] .
let @xmath48 be a basic @xmath102-form .
in the principal stratum where @xmath528 for every @xmath184 , we have that @xmath529 in which the @xmath530 are uniquely determined smooth @xmath531-forms on @xmath167 , depending smoothly on @xmath352 .
we are done if we can prove that the @xmath532 extend smoothly over the boundary where some of the @xmath395 are equal to zero .
recall that @xmath533 , if @xmath534 with @xmath535 .
then @xmath536 shows that @xmath537 has the component @xmath538 and therefore the smoothness of @xmath48 implies that for each @xmath539 and each sequence @xmath540 with @xmath541 the form @xmath542 depends smoothly on @xmath543 . applying the differential operator @xmath544 and putting @xmath545
, we obtain that @xmath546 depends smoothly on @xmath141 , and moreover is invariant under each of the reflections @xmath547 .
whitney @xcite proved that this implies that the function @xmath532 extends smoothly over the boundary where some of the @xmath395 are equal to zero . for any smooth mapping @xmath51 from a smooth manifold @xmath18 to a smooth manifold @xmath39 , the tangent mapping @xmath548 is the linear mapping from @xmath549 to @xmath550 which in local coordinates corresponds to the jacobi matrix of @xmath51 at the point @xmath73 .
let @xmath551 be a @xmath0-equivariant diffeomorphism which preserves the @xmath0-orbits .
then there is a unique smooth @xmath0-invariant mapping @xmath552 such that @xmath553 for every @xmath59 .
if @xmath266 preserves the symplectic form @xmath17 , then @xmath554 for each @xmath59 and @xmath95 . here @xmath40 is the kernel of the antisymmetric bilinear form @xmath32 introduced in lemma [ constlem ] , and we identify each tangent space of @xmath0 with @xmath12 . [ hslem ] the first statement has been proved for arbitrary torus actions on orbifolds by haefliger and salem ( * ? ? ?
3.1 ) , but in our case the proof is elementary .
the statement is obvious if we replace @xmath18 by the set @xmath88 on which the action is free and defines a principal @xmath0-fibration , and it remains to be proved that @xmath555 has a smooth extension to @xmath18 . in the local model of lemma
[ modellem ] , we have @xmath556 the smoothness of @xmath266 implies that @xmath557 has a smooth extension .
write @xmath558 .
it remains to prove that the fact that @xmath559 has a smooth extension , implies that @xmath560 has a smooth extension , because the fact that @xmath217 is an isomorphism of lie groups then implies that @xmath561 has a smooth extension .
now the function @xmath562 has a smooth extension , of which the restriction to the `` real domain '' @xmath545 is an even function in each of the variables @xmath563 .
it therefore follows from whitney @xcite that there is a smooth function @xmath564 such that @xmath565 .
however @xmath566 when @xmath567 , and it follows that @xmath568 in which @xmath569 is smooth . because @xmath570 it follows that @xmath571 when @xmath528 , which extends smoothly over the boundary @xmath567 .
write , for each @xmath59 , @xmath572 , viewed as a linear mapping from @xmath78 to @xmath12 , and @xmath573 , which is a symplectic linear mapping from @xmath78 to @xmath574 .
then it follows from the sum rule for differentiation of an expression in which a variable occurs at several places , that @xmath575 if @xmath54 , then the @xmath0-equivariance of @xmath266 implies that @xmath576 . on the other hand
, the commutativity of @xmath0 implies that @xmath577 for every @xmath169 , and differentiating this with respect to @xmath578 at @xmath579 in the direction of @xmath42 , we obtain @xmath580 .
the condition @xmath581 implies that we have , for every @xmath59 , @xmath95 , and @xmath54 , @xmath582 which implies that @xmath583 because @xmath584 is symplectic . because @xmath585 for every @xmath54 , it follows that @xmath586 . one can prove that @xmath266 is a @xmath0-equivariant symplectomorphism of @xmath1 which preserves the @xmath0-orbits , if and only if for every @xmath59 there exists a @xmath0-invariant open neighborhood @xmath119 of @xmath73 in @xmath18 , a @xmath0-invariant smooth function @xmath51 on @xmath119 , and an element @xmath169 , such that @xmath587 on @xmath119 .
the `` if '' part follows from remark [ multfreerem ] .
if we identify each of the tangent spaces of @xmath310 with @xmath34 as in remark [ regrem ] , then any @xmath588 can be viewed as a constant vector field on @xmath310 .
a vector field @xmath589 in @xmath88 is called a _ lift _ of @xmath590 , if @xmath591 for all @xmath135 . here
the tangent mapping @xmath592 of @xmath593 is identified with the linear mapping @xmath594 , defined by the @xmath34-valued one - form @xmath35 on @xmath18 . in view of the definition of @xmath35 in lemma [ alphalem ] and ( [ widehatsigma ] ) ,
the condition that @xmath589 is a lift of @xmath590 therefore is equivalent to @xmath595 if @xmath589 , @xmath588 , is a family of smooth @xmath0-invariant vector fields on @xmath88 , which depends linearly on @xmath590 and are lifts in the sense of ( [ lift ] ) , then for each @xmath135 the vectors @xmath596 , @xmath588 , span a linear subspace @xmath597 of @xmath78 which is complementary to the tangent space @xmath80 at @xmath73 of the orbit @xmath111 .
the @xmath597 , @xmath135 , are the horizontal spaces for a unique @xmath0-invariant _ infinitesimal connection _
@xmath598 for the principal @xmath0-bundle @xmath599 .
this connection is @xmath0-invariant , if and only if each of the lifts @xmath589 , @xmath588 , is @xmath0-invariant .
conversely , if we have given a @xmath0-invariant infinitesimal connection @xmath598 for the principal @xmath0-fibration in @xmath88 , with horizontal spaces @xmath600 , @xmath135 , then we have for each @xmath588 a unique lift @xmath589 of @xmath590 such that @xmath601 for every @xmath135 .
@xmath589 is called the _ horizontal lift _ of @xmath590 defined by the connection @xmath598 , and denoted by @xmath602 in the literature on connections . because the mapping @xmath603 is linear , `` lifts @xmath589 which depend linearly on @xmath590 '' and `` connections '' are equivalent objects .
we will use the somewhat simpler notation @xmath589 instead of @xmath602 , because it is the lifts which we will be using to construct our global model . in this section
we construct lifts @xmath589 , @xmath588 , depending linearly on @xmath590 , which are admissible in the sense of definition [ liftdef ] , and have lie brackets and symplectic products which are as simple as we can get them
. see proposition [ liftprop ] below .
this construction is based on a computation in the cohomology of the closed basic differential forms on @xmath18 , which according to the theorem of koszul @xcite is canonically isomorphic to the sheaf (= ech ) cohomology of the orbit space @xmath3 with values in @xmath604 .
the lifts in proposition [ liftprop ] form the core of the construction of the model for the symplectic @xmath0-manifold @xmath16 , given in proposition [ aprop ] and proposition [ omegalem ] .
let , in the local model of lemma [ modellem ] with the diffeomorphism @xmath266 , the lift @xmath589 be equal to the image under @xmath605 of the vector field @xmath606 .
then , in terms of the @xmath355-coordinates in @xmath213 , we obtain in view of ( [ modelsigma ] ) and ( [ aiota ] ) that the equation ( [ lift ] ) is equivalent to @xmath607 .
let @xmath608 , and let @xmath609 be the image under @xmath605 of the `` constant '' vector field @xmath610 , where we use the @xmath355-coordinates in @xmath213 .
then @xmath609 is a smooth @xmath0-invariant vector field on @xmath611 , and a lift of @xmath590 .
we call @xmath609 the _ local model lift defined by the local model with the diffeomorphism @xmath266_. the local model lift @xmath609 extends to a smooth @xmath0-invariant vector field on @xmath119 when @xmath612 , that is , when @xmath613 on @xmath358 . on the other hand , if we write @xmath614 , then @xmath615 .
this shows that @xmath616 has a pole singularity at any point @xmath270 for which there exists a @xmath184 such that @xmath617 and @xmath618 .
[ localmodelliftdef ] let @xmath619 be another local model as in lemma [ modellem ] , where we use the same projection @xmath620 . then there is a smooth @xmath0-invariant mapping @xmath621 , such that @xmath622 for every @xmath623 .
here @xmath616 and @xmath624 are the local model lifts definition [ localmodelliftdef ] .
[ lphilem ] let @xmath625 and write @xmath626 , where we use the @xmath444-coordinates in @xmath213 . by permuting the coordinates in @xmath213
, we can arrange that @xmath627 for @xmath628 and @xmath629 for @xmath630 .
then @xmath631 is equal to the subgroup @xmath632 of @xmath27 , where @xmath199 denotes the isomorphism from @xmath27 onto @xmath220 , introduced in remark [ complexrem ] . here
@xmath109 as in lemma [ modellem ] .
let @xmath633 .
then @xmath634 is a complementary subtorus to @xmath635 in @xmath27 , and @xmath636 is a complementary subtorus to @xmath635 in @xmath0 which contains @xmath259 .
let @xmath637 and @xmath638 be the first @xmath639 and the last @xmath640 of the @xmath641-coordinates , respectively .
then the rotation of @xmath642 over @xmath643 , for each @xmath630 , defines an element @xmath644 of @xmath645 , and @xmath646 . on the other hand @xmath647
is a linear form on @xmath358 which is equal to zero on the lie algebra @xmath648 of @xmath635 .
this linear from has a unique extension to a linear form @xmath649 on @xmath40 which is equal to zero on @xmath650 .
in this way we obtain an element @xmath651 .
a straightforward computation shows that the mapping @xmath652 when restricted to the the domain where @xmath528 for all @xmath630 , defines a smooth @xmath0-equivariant symplectomorphism from @xmath653 to @xmath654 .
moreover , @xmath655 belongs to the @xmath0-orbit of @xmath656 in @xmath654 . because the tangent mapping of @xmath409 maps @xmath657 to @xmath658 , we have @xmath659 .
similarly we have a smooth @xmath0-equivariant symplectomorphism @xmath660 from a @xmath0-invariant open neighborhood of @xmath661 in @xmath662 onto a @xmath0-invariant open neighborhood of @xmath656 in @xmath663 , such that @xmath664 and @xmath665 . here
@xmath666 is another complementary subtorus to @xmath635 in @xmath0 .
the mapping @xmath667 is a @xmath0-equivariant symplectomorphism which maps @xmath656 to @xmath656 .
here we have written , for each @xmath668 , @xmath669 with @xmath670 and @xmath671 . because @xmath672 is the unique element in @xmath635 such that @xmath673 , the fact that @xmath674 is a @xmath0-equivariant symplectomorphism follows from the proof of lemma [ modellem ] , with @xmath675 replaced by @xmath676 and by @xmath677 , respectively . because the tangent mapping of @xmath674 maps @xmath678 to @xmath678 , we have that @xmath679 the mapping @xmath680 is a smooth @xmath0-equivariant symplectomorphism from an open @xmath0-invariant neighborhood of @xmath656 in @xmath654 , onto an open @xmath0-invariant neighborhood of @xmath656 in @xmath654 , which preserves the @xmath0-orbit of @xmath656 .
recall the @xmath34-valued one form @xmath35 defined in lemma [ alphalem ] , which we used to identify all tangent spaces of the orbit space with @xmath34 . because every @xmath0-equivariant symplectomorphism preserves @xmath35 , its induced transformation of the orbit space has derivative equal to the identity at every point .
therefore @xmath681 is a translation on each connected open subset of the @xmath0-orbit space by means of a constant element @xmath46 of @xmath34 .
because @xmath681 preserves the @xmath0-orbit of @xmath656 , we have @xmath682 on the connected component of @xmath656 of the domain of definition @xmath683 of @xmath681 .
that is , @xmath681 preserves all the @xmath0-orbits in a @xmath0-invariant open neighborhood of @xmath656 .
it now follows from lemma [ hslem ] that there there is a smooth @xmath0-invariant @xmath40-valued function @xmath555 on @xmath683 , such that @xmath684 and @xmath685 for every @xmath686 and @xmath687 .
it follows from ( [ tphitau ] ) , with @xmath266 and @xmath688 replaced by @xmath681 and @xmath689 , respectively , where @xmath690 and @xmath691 , that @xmath692 maps the vector field @xmath678 to the sum of @xmath678 and @xmath693 because @xmath694 , @xmath695 , and @xmath659 , the conclusion of the lemma follows with @xmath696 if @xmath697 .
we use the atlas of local models as in lemma [ modellem ] , with a fixed linear projection @xmath263 from @xmath12 onto @xmath40 . for every @xmath588 ,
an _ admissible lift of @xmath590 _ is a smooth @xmath0-invariant vector field @xmath589 on @xmath88 such that for each local model as in lemma [ modellem ] there is a smooth @xmath0-invariant @xmath40-valued function @xmath698 on @xmath119 , such that @xmath699 for every @xmath700 . here
@xmath616 is the local model lift introduced in definition [ localmodelliftdef ] .
if we are at an orbit type @xmath449 with stabilizer group @xmath27 , and @xmath590 is equal to zero on the lie algebra @xmath367 of @xmath27 , then @xmath589 has a unique smooth @xmath0-invariant extension to an open neighborhood of @xmath449 in @xmath18 , which will also be denoted by @xmath589 .
in particular , if @xmath701 , the space of linear forms on @xmath40 which vanish on @xmath26 , then @xmath702 is a smooth @xmath0-invariant vector field on the whole manifold @xmath18 .
an _ admissible connection _ for the principal @xmath0-bundle @xmath599 is a linear mapping @xmath703 from @xmath34 to the space of smooth vector fields on @xmath88 , such that , for each @xmath588 , @xmath589 is an admissible lift of @xmath590 . because we work with a fixed action of the torus @xmath0 , we will just write `` admissible connection '' in the sequel . in the literature ,
the term `` admissible connection '' has been used in various different frameworks and with correspondingly different meanings .
our usage of the term `` admissible connection '' continues this .
[ liftdef ] there exist admissible connections @xmath703 .
for each admissible connection @xmath703 , we have @xmath704 [ liftexistencelem ] if we piece the local model lifts @xmath616 , introduced in definition [ localmodelliftdef ] , together by means of a partition of unity consisting of smooth @xmath0-invariant functions with supports in the local model neighborhoods @xmath119 , then it follows from lemma [ lphilem ] that the resulting connection is admissible . in the local model of lemma
[ modellem ] , where we use the @xmath444-coordinates in @xmath213 as in ( [ rhodef ] ) , ( [ aiota ] ) , we have @xmath705 with @xmath608 .
furthermore @xmath706 .
it follows that @xmath707 here we have used ( [ gotkgotl ] ) with @xmath360 replaced by @xmath42 . on the other hand @xmath708 for every @xmath54
if @xmath709 , and ( [ stronglift ] ) now follows from definition [ liftdef ] .
the equation ( [ stronglift ] ) improves upon ( [ lift ] ) if @xmath40 is a proper linear subspace of @xmath12 , that is , if @xmath710 . if the principal orbits are lagrangian submanifolds of @xmath18 , then @xmath333 and ( [ stronglift ] ) is the same as ( [ lift ] ) .
recall the hamiltonian torus @xmath24 , the unique maximal stabilizer subgroup @xmath24 of @xmath0 as in remark [ t1rem ] and lemma [ stabilizerlem ] , with lie algebra @xmath711 . in our quest for nice admissible lifts , we will use a decomposition of @xmath0 into the subtorus @xmath24 and a complementary subtorus @xmath31 , as in lemma [ tbaselem ] with @xmath712 . note that the torus @xmath31 acts freely on @xmath18 , because if @xmath59 , then @xmath514 , hence @xmath713 .
this explains our choice of the subscript f in @xmath31 .
note also that the choice of a complementary subtorus @xmath31 to @xmath24 is far from unique if @xmath714 .
see remark [ tbaserem ] .
we will refer to @xmath31 as _ a freely acting complementary torus to the hamiltonian torus _ @xmath24 .
if @xmath715 denotes the lie algebra of @xmath31 , then we have a corresponding direct sum decomposition @xmath716 of lie algebras .
each linear form on @xmath499 has a unique extension to a linear form on @xmath40 which is equal to zero on @xmath715 .
this leads to an isomorphism of @xmath499 with the linear subspace @xmath717 of @xmath34 .
this isomorphism depends on the choice of the complementary freely acting torus @xmath31 to the hamiltonian torus in @xmath0 .
note that the direct sum decomposition @xmath718 implies the direct sum decomposition @xmath719 let @xmath720 denote the projection @xmath302 , followed by the projection from @xmath721 onto the first factor .
here we use the isomorphism @xmath722 of proposition [ m / tprop ] , with @xmath466 and @xmath723 . note that @xmath724 is a momentum mapping for the hamiltonian @xmath24-action on @xmath18 as in corollary [ tstabcor ] . with these notations ,
proposition [ liftprop ] below yields the existence of an admissible connection for which both the lie brackets and the symplectic products of the @xmath589 take an extremely simple form .
( we are tempted to call such a connection a `` minimal admissible connection '' , see also subsection [ minimalcouplingss ] , but we do not have a proposal for a functional which is minimized exactly by the connections in proposition [ liftprop ] . ) in remark [ curvrem ] and remark [ integralcrem ] we discuss the topological meaning of the antisymmetric bilinear form @xmath725 . from these remarks
it follows that @xmath726 is unique .
the freedom in the choice of the admissible connection in proposition [ liftprop ] will be described in lemma [ liftfreedomlem ] .
there exists an admissible connection @xmath727 as in definition [ liftdef ] , and an antisymmetric bilinear mapping @xmath725 , with the following properties . * @xmath728 = 0 $ ] for all @xmath729 , * @xmath730=0 $ ] for all @xmath731 and @xmath732 , * @xmath733 = c(\zeta , \,\zeta ' ) _ m$ ] for all @xmath734 , * @xmath735 for all @xmath729 , * @xmath736 for all @xmath731 and @xmath732 , and finally * @xmath737 for all @xmath734 and @xmath59 . here
@xmath738 denotes the @xmath26-component of @xmath739 in the direct sum decomposition @xmath740 .
the antisymmetric bilinear mapping @xmath725 in part iii ) satisfies the relation @xmath741 for every @xmath742 .
note that @xmath743 is a linear form on @xmath40 which vanishes on @xmath26 , and therefore @xmath744 is a real number which only depends on the projection of @xmath739 to @xmath504 .
[ liftprop ] we start with an arbitrary admissible connection @xmath727 , which exists according to lemma [ liftexistencelem ] , and first simplify the lie brackets .
we use the isomorphism @xmath722 of proposition [ m / tprop ] , in order to identify @xmath3 with @xmath745 . in view of lemma [ basicformlem ] ,
the smooth basic @xmath102-forms on @xmath18 satisfy @xmath746 on @xmath88 for uniquely determined smooth @xmath102-forms @xmath314 on @xmath747 , such that @xmath314 extends to a smooth @xmath102-form on @xmath748 .
this leads to an identification of the space of all smooth basic @xmath102-forms on @xmath18 with the space of all restrictions to @xmath749 of smooth @xmath102-forms on @xmath748 . if we view @xmath588 as a constant vector field on @xmath748 , then the fact that @xmath750 maps @xmath589 to @xmath590 implies that @xmath751 because @xmath593 intertwines the flow of @xmath589 with the flow of the constant vector field @xmath590 , we also have the identity @xmath752 for the lie derivatives .
in particular the differentiation of @xmath0-invariant smooth functions on @xmath18 in the direction of the vector field @xmath589 corresponds to the differentiation @xmath753 of smooth functions on @xmath3 in the direction of the constant vector field @xmath590 .
let @xmath754 , @xmath755 , be a basis of @xmath40 .
we will write @xmath756 , in which the real numbers @xmath757 are the coordinates of @xmath758 with respect to this basis , and we use einstein s summation convention when summing over indices which run from @xmath357 up to @xmath759 .
the local model lifts @xmath616 and @xmath760 , introduced in definition [ localmodelliftdef ] , commute because they are constant vector fields on @xmath267 in the local model of lemma [ modellem ] , where we use the @xmath444-coordinates for @xmath289 .
if @xmath761 as in definition [ liftdef ] , then the fact that the vector fields @xmath616 and @xmath760 commute as well as the vector fields @xmath762 and @xmath763 , implies that @xmath764=\beta ^i_{\xi , \,\xi ' } \ , ( x_i)_m$ ] , in which the uniquely determined @xmath0-invariant functions @xmath765 on @xmath766 are given by @xmath767 that is , @xmath768 on @xmath119 , if we define the basic @xmath40-valued one - form @xmath769 and two - form @xmath770 by @xmath771 and @xmath772 , respectively . because @xmath770 is locally exact , it follows that the globally defined smooth basic two - form @xmath770 is closed .
any other connection @xmath773 is admissible , if and only if @xmath774 in which @xmath775 defines a smooth basic @xmath34-valued one - form @xmath776 on @xmath18 . with the same reasoning as above we obtain that @xmath777 = \widetilde{\beta}^i_{\xi , \,\xi ' } \ , ( x_i)_m$ ] , in which @xmath778 defines a smooth basic @xmath34-valued two - form @xmath779 , such that @xmath780 . according to corollary [ cohomcor ] , the de rham cohomology class of @xmath770 contains a unique @xmath781 , such that @xmath782 when @xmath783 or @xmath784 , and for @xmath785 the @xmath40-valued function @xmath786 is a constant , equal to the average of @xmath787 over any @xmath467-orbit in @xmath3 .
this leads to the desired properties of the lie brackets , where the uniqueness of @xmath726 follows from the injectivity of the mapping @xmath788 in corollary [ cohomcor ] .
we now turn to the symplectic inner products .
let @xmath789 .
it follows from definition [ localmodelliftdef ] that @xmath790 . in view of definition [ liftdef ] and formula ( [ stronglift ] )
we conclude that @xmath791 is a smooth function on @xmath18 , which moreover is @xmath0-invariant .
therefore @xmath792 defines a smooth basic two - form @xmath793 on @xmath18 . if in ( [ tildeu ]
) we take @xmath794 for @xmath795 , that is , @xmath796 , then the lie brackets do not change , but @xmath797 where we have used that @xmath798 , because @xmath799 .
this means that @xmath800 . in order
investigate the exterior derivative of @xmath793 , we recall the identity @xmath801)+\omega(v,\ , [ w,\ , u])+\omega(w,\ , [ u,\ , v ] ) , \label{domega}\end{aligned}\ ] ] which holds for any smooth two - form @xmath48 and smooth vector fields @xmath802 .
it follows that @xmath803 ) -\sigma(l_{\xi ' } , \ , [ l_{\xi ' ' } , \ , l_{\xi } ] ) -\sigma(l_{\xi ' ' } , \ , [ l_{\xi},\ , l_{\xi ' } ] ) \\ & = & -\xi ( c(\xi ' , \,\xi ' ' ) ) -\xi ' ( c(\xi ' ' , \,\xi ) ) -\xi ' ' ( c(\xi , \,\xi ' ) ) , \end{aligned}\ ] ] where in the third identity we have used that @xmath804 , and in the last identity we have inserted @xmath764=c(\xi , \,\xi ' ) ^i\ , ( x_i)_m$ ] and ( [ lift ] ) .
this shows that @xmath805 is constant . in the notation of corollary [ cohomcor ]
, we have that @xmath806 is constant , and cohomologically equal to zero , which in view of the first part of the last statement in corollary [ cohomcor ] implies that @xmath807 .
that is , @xmath808 when @xmath809 , which in turn is equivalent to ( [ ft ] ) . on the other hand it follows from the already proved statements about the lie brackets that @xmath782
if @xmath783 or @xmath784 , and hence @xmath808 unless one of the vectors @xmath810 belongs to @xmath85 and the other two belong to @xmath39 . moreover , if @xmath783 and @xmath811 , then we obtain that @xmath812 . in other words , the smooth basic two - form @xmath813 is closed . here
@xmath500 is viewed as a @xmath499-valued @xmath0-invariant function on @xmath18 , and the pairing with the @xmath26-valued antisymmetric bilinear form @xmath814 yields a smooth basic two - form @xmath815 on @xmath18 . according to corollary [ cohomcor ] , the smooth basic one - form @xmath816 can be now chosen such that if @xmath817 , then @xmath818 when @xmath783 or @xmath784 , and for @xmath785 the function @xmath819 is constant .
we finally observe that if @xmath820 is a linear mapping from @xmath34 to @xmath40 , which is viewed as a constant , hence closed one - form on the @xmath34-parallel space @xmath3 , then the lie brackets of the @xmath589 s do not change if we replace @xmath589 by @xmath821 .
however , @xmath791 then gets replaced by @xmath822 .
because any antisymmetric bilinear form on @xmath34 is of the form @xmath823 , for a suitable linear mapping @xmath824 , we can arrange that @xmath825 , which leads to vi ) in proposition [ liftprop ] . because the left hand side of ( [ ft ] ) is antisymmetric in @xmath826 , it is automatically equal to zero when @xmath827 .
however , when @xmath828 , then the equations ( [ ft ] ) impose nontrivial conditions on the @xmath40-valued two - form @xmath726 on @xmath39 .
[ ftrem ] for every @xmath135 , let @xmath597 denote the linear span in @xmath78 of the vectors @xmath596 , @xmath588 .
then the @xmath597 , @xmath135 , define a @xmath0-invariant infinitesimal connection of the principal @xmath0-bundle @xmath88 over @xmath829 . here @xmath830 denotes the interior of the delzant polytope @xmath10 .
any connection of this principal @xmath0-bundle has a _ curvature form _ which is a smooth @xmath12-valued two - form on @xmath831 .
the cohomology class of the curvature form is an element of @xmath832 , which is independent of the choice of the connection .
the action of @xmath39 on @xmath3 leaves @xmath833 invariant , with orbits isomorphic to the torus @xmath467 , and the pull - back to the @xmath39-orbits defines an isomorphism from @xmath832 onto @xmath834 , which in turn is identified with @xmath835 as in corollary [ cohomcor ] .
the proof of proposition [ liftprop ] shows that the element @xmath836 is equal to the negative of the pull - back to an @xmath39-orbit of the cohomology class of the curvature form .
this proves in particular that the antisymmetric bilinear mapping @xmath725 in proposition [ liftprop ] is independent of the choice of the freely acting complementary torus @xmath31 to the hamiltonian torus in @xmath0 .
more precisely , if @xmath837 is another symplectic manifold with an effective symplectic @xmath0-action with coisotropic principal orbits , then proposition [ liftprop ] with @xmath16 replaced by @xmath838 yields an antisymmetric bilinear mapping @xmath839 instead of @xmath726 .
if there exists a @xmath0-equivariant symplectomorphism @xmath266 from @xmath16 onto @xmath838 then @xmath840 .
[ curvrem ] the retrivializations of the principal @xmath0-bundle @xmath841 define a one - cocycle of smooth @xmath12-valued functions on @xmath831 , of which the sheaf (= ech ) cohomology class @xmath555 in @xmath842 classifies the principal @xmath0-bundle @xmath841 . because the sheaf @xmath843 is fine
, the short exact sequence @xmath844 induces an isomorphism @xmath845 . here
exp denotes the exponential mapping @xmath846 .
the cohomology class @xmath847 is called the _
chern class _ of the principal @xmath0-bundle @xmath841 .
it is a general fact , see for instance the arguments in ( * ? ? ?
15.3 ) , that the image of @xmath848 in @xmath832 under the coefficient homomorphism @xmath849 is equal to the negative of the cohomology class of the curvature form of any connection in the principal @xmath0-bundle . in view of remark [ curvrem ] , we therefore conclude that @xmath726 represents the chern class of the principal @xmath0-bundle @xmath841 . in view of the canonical isomorphism between sheaf cohomology and singular cohomology , this implies that the integral of @xmath726 over every two - cycle in @xmath310 belongs to @xmath14 . if @xmath850 , then for every @xmath436 the mapping @xmath851 defines a two - cycle in @xmath310 , and @xmath852 it follows that @xmath853 for every @xmath850 . in lemma [ tplem ]
, this conclusion will be proved by means of a group theoretical consideration .
other topological interpretations of @xmath726 will be given in proposition [ pi1prop ] and proposition [ chernprop ] .
[ integralcrem ] let @xmath854 denote the space of all linear mappings from a vector space @xmath138 to a vector space @xmath137 .
in the following lemma we use that a smooth @xmath0-invariant mapping @xmath855 corresponds to a unique smooth basic @xmath40-valued one - form on @xmath18 , which we also denote by @xmath769 , such that @xmath856 for every @xmath95 such that @xmath857 . in view of lemma [ basicformlem ]
, @xmath769 can also be viewed as the restriction to @xmath3 of a smooth @xmath40-valued one - form on @xmath748 , if we identify the @xmath34-parallel space @xmath3 with a subset of @xmath748 as in proposition [ m / tprop ] .
let @xmath727 be a connection as in proposition [ liftprop ] .
then @xmath858 is an admissible connection , if and only if there exists a smooth @xmath0-invariant mapping @xmath859 from @xmath18 to @xmath860 , such that @xmath861 for every @xmath59 and @xmath588 .
proposition [ liftprop ] holds with @xmath862 replaced by @xmath863 , if and only if @xmath769 is closed when considered as a smooth basic @xmath40-valued one - form on @xmath18 , and moreover @xmath769 is symmetric in the sense that @xmath864 for all @xmath789 and all @xmath59 .
[ liftfreedomlem ] the first statement follows from definition [ liftdef ] , the definition of admissible connections .
it follows from the proof of proposition [ liftprop ] that @xmath865 \equiv [ l_{\xi},\ , l_{\xi'}]$ ] if and only if @xmath769 is closed . in view of the uniqueness of @xmath726 ,
see remark [ curvrem ] , we have iv ) , v ) , vi ) in proposition [ liftprop ] with @xmath862 replaced by @xmath863 , if and only if @xmath866 , which is equivalent to ( [ alphasym ] ) .
recall the especially nice admissible connection introduced in proposition [ liftprop ] , the construction of which is based on the identification in proposition [ m / tprop ] of the orbit space @xmath3 with the @xmath34-parallel space @xmath36 .
proposition [ embedprop ] below implies that the vector fields @xmath867 , @xmath868 , and @xmath869 , @xmath731 , are tangent to the fibers of a fibration of @xmath18 by delzant submanifolds . from this section on , the word _
fibration _ is short for a locally trivial smooth fiber bundle .
the remainder of this section is devoted to the proof and further precision of proposition [ embedprop ] . for any subset @xmath360 of a set @xmath42 , the _ inclusion mapping _
@xmath870 is the identity on @xmath360 , viewed as a mapping from @xmath360 to @xmath42 .
let @xmath727 be an admissible connection as in proposition [ liftprop ] .
then there is a unique smooth @xmath0-invariant distribution @xmath458 on @xmath18 such that , for every @xmath135 , @xmath871 is equal to the linear span in @xmath78 of the vectors @xmath872 with @xmath868 and @xmath873 , @xmath874 .
the distribution @xmath458 is integrable .
each integral manifold manifold @xmath875 of @xmath458 is invariant under the action of the hamiltonian torus @xmath24 , and @xmath876 is a delzant manifold with the delzant polytope @xmath10 introduced in proposition [ m / tprop ] . here
@xmath877 is the inclusion mapping from @xmath875 into @xmath18 .
the integral manifolds of @xmath458 form a smooth fibration of @xmath18 into delzant submanifolds with delzant polytope @xmath10 .
[ embedprop ] this follows from lemma [ embedlem ] below , which in turn uses lemma [ pi2lem ] .
let @xmath878 be the mapping which is equal to @xmath302 , followed by the inverse @xmath879 of the isomorphism @xmath484 in proposition [ m / tprop ] , iii ) , followed by the projection @xmath880 onto the second factor .
then @xmath878 defines a smooth fibration of @xmath18 over the torus @xmath467 .
each fiber @xmath137 of @xmath878 is a connected compact @xmath0-invariant smooth submanifold of @xmath18 . for each fiber @xmath137 of @xmath878 , @xmath881 is dense in @xmath137 .
[ pi2lem ] because @xmath599 and the projection from @xmath882 onto @xmath467 are smooth fibrations with connected fibers , it follows that the restriction to @xmath88 of @xmath883 is a smooth fibration with connected fibers . in the local model of lemma [ modellem ] ,
the mapping @xmath883 corresponds to the mapping @xmath884 where we have used the direct sum decomposition ( [ ldec ] ) .
this shows that @xmath883 is a smooth submersion .
moreover , for each fiber @xmath137 of @xmath878 , @xmath881 is dense in @xmath137 , because the point @xmath270 is regular if and only if @xmath425 for every @xmath391 .
because the fiber @xmath881 of the restriction to @xmath88 of @xmath883 is connected , it follows that @xmath137 is connected .
because @xmath18 is compact , the submersion @xmath883 is proper , and because every proper submersion is a fibration , it follows that @xmath883 is a fibration .
as observed in the beginning of subsection [ nicess ] , the action on @xmath18 of the complementary torus @xmath31 to @xmath24 is free .
this exhibits each fiber @xmath137 of @xmath883 as a principal @xmath31-bundle @xmath885 , in which the @xmath31-orbit space @xmath886 is a compact , connected smooth manifold , on which we still have the action of the hamiltonian torus @xmath24 .
the following lemma says that there is a symplectic form @xmath887 on @xmath886 such that @xmath888 is a delzant manifold defined by the delzant polytope @xmath10 , and that the fibration @xmath889 is trivial , exhibiting @xmath137 as the cartesian product of the delzant manifold @xmath886 with @xmath31 .
there is a unique smooth distribution @xmath458 on @xmath18 such that , for every @xmath135 , @xmath871 is equal to the linear span in @xmath78 of the vectors @xmath872 , @xmath868 , and @xmath873 , @xmath731 .
the distribution @xmath458 is integrable and @xmath0-invariant . for every fiber @xmath137 of the fibration @xmath883 in lemma [ pi2lem ] , we have @xmath890 , which implies that @xmath891 or @xmath892 for every integral manifold @xmath875 of @xmath458 .
let @xmath893 and let @xmath894 be the integral manifold of @xmath458 such that @xmath895 .
for each @xmath896 there is a unique @xmath897 such that @xmath898 .
the mapping @xmath899 is the inverse of a trivialization @xmath555 of the principal @xmath31-fibration @xmath885 .
the trivialization @xmath555 is @xmath0-equivariant , where @xmath169 acts on @xmath900 by sending @xmath901 to @xmath902 , if @xmath903 , with @xmath904 and @xmath905 .
finally , there is a unique symplectic form @xmath887 on @xmath886 such that , for any integral manifold @xmath875 of @xmath458 in @xmath137 , @xmath906 if @xmath907 denotes the inclusion mapping from @xmath875 into @xmath137 . with this symplectic form , ( [ f / tdelzant ] )
is a delzant manifold with delzant polytope @xmath10 .
for each integral manifold @xmath875 of @xmath458 in @xmath137 , @xmath908 is a delzant manifold with delzant polytope @xmath10 , and @xmath909 is a @xmath24-equivariant symplectomorphism from @xmath908 onto the delzant manifold ( [ f / tdelzant ] ) .
[ embedlem ] in order to investigate the @xmath871 with @xmath135 near a singular point @xmath340 , we use a local model as in lemma [ modellem ] , with the @xmath444-coordinates in @xmath213 as in ( [ rhodef ] ) . here
@xmath910 is a subtorus of the hamiltonian torus @xmath24 .
let @xmath911 be a complementary subtorus to @xmath27 in @xmath24 .
we will take @xmath912 as the complementary subtorus to @xmath27 in @xmath0 .
for the lie algebras we have the corresponding direct sum decompositions @xmath913 and @xmath914 .
the span of the infinitesimal actions of the elements @xmath387 is equal to the span of the vector fields @xmath915 , @xmath184 , and the vector fields @xmath916 with @xmath917 .
the @xmath918 , @xmath731 , are the linear combinations of the vector fields @xmath919 , @xmath184 , and the vector fields @xmath920 , with constant @xmath921 . according to definition [ liftdef ] , the definition of admissible lifts , @xmath922 in which the vector field @xmath923 is smooth on @xmath18 , of the form @xmath924 for a smooth @xmath0-invariant @xmath40-valued function @xmath769 on @xmath18 .
we write @xmath925 instead of @xmath923 if @xmath926 is such that @xmath927 .
the problem is that the vector fields @xmath448 and @xmath928 have a zero and a pole at @xmath929 .
now @xmath930 imply that @xmath931 these two vector fields are smooth and converge to @xmath932 and @xmath933 , respectively , as @xmath934 .
this proves the first statement in the lemma .
we also obtain for every @xmath59 that @xmath935 , if @xmath936 denotes the linear span of the @xmath937 , @xmath938 , and the @xmath873 , @xmath732 . in view of ( [ lift ] ) , conclusion i ) in proposition [ liftprop ] , and the commutativity of the infinitesimal action of @xmath26 on @xmath18 , the vector fields @xmath867 and @xmath869 all commute with each other .
this implies that on @xmath88 the distribution @xmath458 satisfies the frobenius integrability condition .
because @xmath88 is dense in @xmath18 , it follows by continuity that @xmath458 is integrable on @xmath18 . because the vector fields @xmath867 , and @xmath869 are @xmath0-invariant , the restriction to @xmath88 of @xmath458 is @xmath0-invariant , and it follows by continuity that @xmath458 is @xmath0-invariant . for each @xmath135 , the vectors @xmath939 , @xmath54 , and @xmath873 , @xmath940 , together span @xmath941 , hence @xmath942 . because @xmath943 is dense in @xmath137 , see lemma [ pi2lem ] , it follows by continuity that @xmath890 .
this implies in turn that if @xmath875 is an integral manifold of @xmath458 in @xmath18 and @xmath944 , then @xmath891 and @xmath875 is an integral manifold of @xmath945 . because for every @xmath946 the linear subspaces @xmath871 and @xmath947 of @xmath948 have zero intersection and their dimensions add up to the dimension of @xmath137 , we have that @xmath871 is a complementary linear subspace to @xmath947 in @xmath948 , and it follows that @xmath945 defines a @xmath31-invariant infinitesimal connection for the principal @xmath31-bundle @xmath885 .
it follows from ( [ stronglift ] ) and the conclusion v ) in proposition [ liftprop ] , that for every @xmath135 the complementary linear subspaces @xmath871 and @xmath936 of @xmath78 are @xmath949-orthogonal , and by continuity the same conclusion follows for every @xmath59 .
this implies that , for every @xmath59 , @xmath871 is a symplectic vector subspace of @xmath78 , and therefore every integral manifold @xmath875 of @xmath458 is a symplectic submanifold of @xmath1 .
if @xmath875 is an integral manifold of @xmath458 , then the restriction to @xmath875 of @xmath950 is a covering from @xmath875 onto @xmath886 .
because @xmath17 is invariant under the action of @xmath31 , there is a unique two - form @xmath887 on @xmath886 such that ( [ piiotasigma ] ) holds , and because @xmath909 is a covering , it follows that @xmath887 is a smooth symplectic form on @xmath886 .
the mapping from @xmath886 to @xmath10 induced by ( [ pi^1 ] ) , which we also denote by @xmath500 , is a momentum mapping for the @xmath24-action on the symplectic manifold @xmath951 . because for any @xmath952 the pre - image of @xmath953 under the projection from @xmath721 onto the second factor is equal to @xmath954 , and @xmath500
forgets the second factor , we have that @xmath955 , and therefore @xmath956 . because @xmath137 is compact and connected , see lemma [ pi2lem ]
, the image @xmath886 of @xmath137 under the continuous projection @xmath957 is also compact and connected .
the conclusion is that ( [ f / tdelzant ] ) is a delzant manifold defined by the delzant polytope @xmath10 . because @xmath886 is simply connected in view of lemma [ pi1delzantlem ]
, @xmath958 is a diffeomorphism .
the other statements in the lemma now readily follow .
every delzant manifold is simply connected .
[ pi1delzantlem ] every delzant manifold can be provided with the structure of a toric variety defined by a complete fan , cf .
delzant @xcite and guillemin ( * ? ? ?
1 ) , and danilov ( * ? ? ?
9.1 ) observed that such a toric variety is simply connected .
the argument is that the toric variety has an open cell which is isomorphic to @xmath959 , of which the complement is a complex subvariety of complex codimension one .
therefore any loop can be deformed into the cell and contracted within the cell to a point .
the pull - back to each @xmath31-orbit of the symplectic form @xmath17 on @xmath18 is given by @xmath960 because @xmath961 and @xmath962 , we have that this pull - back is equal to zero if and only @xmath963 , that is , the principal @xmath0-orbits are lagrangian . in this case
the tangent spaces of the @xmath31-orbits in @xmath137 are the kernels of the pull - back to @xmath137 of @xmath17 , and the symplectic form @xmath887 on @xmath886 is the reduced form of the pull - back to @xmath137 of @xmath17 . in other words , @xmath964 is a reduced phase space for the `` momentum mapping '' @xmath878 for the @xmath31-action , where the word momentum mapping is put between parentheses because the free @xmath31-action is not hamiltonian .
the @xmath0-invariant projection @xmath878 induces a @xmath24-invariant projection @xmath965 , of which the fibers are canonically identified with the @xmath886 , where the @xmath137 are the fibers of @xmath878 .
if @xmath295 , then the symplectic leaves in @xmath30 of the poisson structure on @xmath966 are equal to the fibers @xmath886 of @xmath965 , provided with the symplectic forms @xmath887 .
it is quite remarkable that the symplectic leaves form a fibration , because in general the symplectic leaves of a poisson structure are only immersed submanifolds , not necessarily closed .
[ fredrem ]
let @xmath1 be our compact connected symplectic manifold , together with an effective action of the torus @xmath0 by means of symplectomorphisms of @xmath1 , such that some ( all ) principal orbits of the @xmath0-action are coisotropic submanifolds of @xmath1 . in subsection [ gss ] we will show that the @xmath0-action together with the infinitesimal action of the vector fields @xmath702 , @xmath732 , introduced in proposition [ liftprop ] , lead to an action on @xmath18 of a two - step nilpotent lie group @xmath19 , where @xmath19 is explicitly defined in terms of the antisymmetric bilinear mapping @xmath725 introduced in proposition [ liftprop ] . subsection [ gss ]
is a sequence of definitions , together with some of their immediate consequences .
recall the fibration of @xmath18 into delzant submanifolds introduced in proposition [ embedprop ] . the action of @xmath19 on @xmath18 will be used to exhibit this fibration as a @xmath19-homogeneous bundle over the homogeneous space @xmath21 with fiber equal to a delzant manifold defined by the delzant polytope @xmath10 . here
@xmath27 is a closed lie subgroup of @xmath19 which is explicitly defined in terms of @xmath726 and the period group @xmath38 in @xmath39 , defined in lemma [ perlem ] with @xmath456 , @xmath457 , and @xmath466 .
see proposition [ aprop ] .
the symplectic form on this bundle of delzant manifolds is given explicitly by means of the formula ( [ a*sigma ] ) , in terms of the antisymmetric bilinear form @xmath32 on @xmath12 introduced in lemma [ constlem ] , the antisymmetric bilinear mapping @xmath725 introduced in proposition [ liftprop ] , and the symplectic form @xmath967 on the delzant manifold @xmath28 . in this way
we obtain an explicit global model for our symplectic manifold @xmath1 with symplectic @xmath0-action . in the sequel
, @xmath968 denotes the lie algebra of all smooth vector fields on @xmath18 , provided with the lie brackets @xmath63 $ ] of @xmath969 such that @xmath63\ , f = u\ , ( v\ , f)-v\ , ( u\ , f)$ ] for every @xmath330 .
we denote the flow after time @xmath970 of @xmath971 by @xmath972 .
this defines an exponential mapping @xmath973 from @xmath968 to the group @xmath974 of all smooth diffeomorphisms of @xmath18 , which is analogous to the exponential mapping @xmath975 from the lie algebra of any lie group to the lie group .
let @xmath727 be an admissible lift as in proposition [ liftprop ] .
for each @xmath732 , @xmath702 is a smooth vector field on @xmath18 , see definition [ liftdef ] , and because @xmath18 is compact , its flow @xmath976 is defined for all @xmath970 .
a lie algebra @xmath977 is called _ two - step nilpotent _ if @xmath978,\ , z]=0 $ ] for all @xmath979 . because the vector fields @xmath702
, @xmath732 commute with the @xmath43 , @xmath54 , and the @xmath43 , @xmath54 , commute with each other , it follows from iii ) in proposition [ liftprop ] that the linear span of the @xmath43 , @xmath54 , and the @xmath702 , @xmath732 , is a two - step nilpotent lie subalgebra @xmath980 of @xmath968 .
moreover , if we provide @xmath981 with the structure of a two - step nilpotent lie algebra defined by @xmath982=\ , -(c(\zeta , \,\zeta ' ) , \ , 0 ) , \quad ( x,\,\zeta ) , \ , ( x',\,\zeta ' ) \in { \mathfrak{g}}={\mathfrak{t}}\times n , \label{gotg}\ ] ] then the mapping @xmath983 is an injective anti - homomorphism of lie algebras from @xmath977 to @xmath968 , with image equal to @xmath980 .
the vector space @xmath984 , provided with the product @xmath985 is a two - step nilpotent lie group with lie algebra equal to @xmath986 and the identity as the exponential mapping .
it follows that the mapping @xmath987 is a ( left ) action of the group @xmath984 on @xmath18 , that is a homomorphism from the group @xmath984 to the group @xmath974 , with infinitesimal action given by @xmath983 .
it follows that @xmath988 the kernel of the homomorphism ( [ gaction ] ) is equal to the discrete normal subgroup @xmath989 of @xmath990 , in which @xmath991 is the integral lattice in the lie algebra @xmath12 of @xmath0 .
it follows that the connected lie group @xmath992 acts smoothly on @xmath18 , where in @xmath993 we have the product @xmath994 and the action is given by @xmath995 note that the lie algebra of @xmath19 is equal to the previously introduced two - step nilpotent lie algebra @xmath996 .
also note that the @xmath0-orbit map @xmath302 intertwines the action of @xmath19 on @xmath18 with the translational action of @xmath39 on @xmath3 , in the sense that @xmath997 for every @xmath998 .
let @xmath727 be an admissible connection as in proposition [ liftprop ] .
for each @xmath999 and @xmath462 , the curve @xmath1000 , @xmath1001 , is a loop in @xmath3 .
if @xmath59 and @xmath1002 , then the curve @xmath1003 , @xmath1001 , is called the _ horizontal lift in @xmath18 of the loop @xmath1004 which starts at @xmath73 _ , because @xmath1005 , @xmath1006 is a horizontal tangent vector which is mapped by @xmath1007 to the constant vector @xmath743 , which implies that @xmath1008 , @xmath1001 .
the element of @xmath0 which maps the initial point @xmath1005 to the end point @xmath1009 is called the _
holonomy @xmath1010 of the loop @xmath1004 and the intial point @xmath73 _ with respect to the given connection . because @xmath1011 , we have @xmath1012 . in lemma [ tplem ] below we investigate the dependence of the holonomy element @xmath1013 on the point @xmath59 and the period @xmath999 .
let @xmath732 .
then the following conditions are equivalent .
* there exists an @xmath59 and a @xmath169 such that @xmath1014 . *
@xmath999 , where @xmath38 is the period group in @xmath39 for the translational action of @xmath39 on @xmath3 , as defined in lemma [ perlem ] with @xmath456 , @xmath457 , and @xmath466 . *
the diffeomorphism @xmath1015 leaves all @xmath0-orbits in @xmath18 invariant . for each @xmath999
there is a unique @xmath0-invariant smooth mapping @xmath1016 such that @xmath1017 for every @xmath59 .
we have @xmath1018 for every @xmath1019 .
we have @xmath853 whenever @xmath850 , and @xmath1020 is a commutative subgroup of @xmath19 . finally , the mapping @xmath1016 is constant on every fiber of the fibration of @xmath18 into delzant submanifolds introduced in proposition [ embedprop ] , and satisfies @xmath1021 [ tplem ]
because the action of @xmath1015 on the @xmath0-orbits is equal to the transformation @xmath1022 in @xmath3 , the equivalence between i ) , ii ) , iii ) follows from lemma [ perlem ] with @xmath456 , @xmath457 , and @xmath466 .
if @xmath999 , then @xmath1023 leaves each @xmath0-orbit invariant . because , for every @xmath732 , @xmath1023 commutes with the @xmath0-action , this implies the existence of the smooth mapping @xmath1024 in view of lemma [ hslem ] . in order to show that ( [ tauzeta ] ) holds , we observe that @xmath1025 } ( \tau _ { \zeta}(x)\cdot x ) = { \!\!\mathop{\rm ~e}\nolimits}^{l_{\zeta ' } } ( { \!\!\mathop{\rm ~e}\nolimits}^{c(\zeta , \,\zeta ' ) } \cdot ( \tau _ { \zeta}(x)\cdot x)),\end{aligned}\ ] ] which implies that @xmath1026 in combination with the @xmath0-invariance of @xmath1024 this yields ( [ tauzeta ] ) . if @xmath1027 , then we have for every @xmath59 that @xmath1028 , hence @xmath1029 , which in view of ( [ tauzeta ] ) implies that @xmath1030 , hence @xmath853 . the fact that @xmath853 for all @xmath850 implies in view of ( [ hproduct ] ) that @xmath1020 is a commutative subgroup of @xmath993 .
because @xmath702 commutes with all @xmath869 , @xmath1031 , see ii ) in proposition [ liftprop ] , we have @xmath1032 which for regular @xmath73 implies that @xmath1033 . by continuity
this identity extends to all @xmath59 . because also @xmath1034 for all @xmath1035
, it follows from the definition in proposition [ embedprop ] of the fibration of @xmath18 into delzant submanifolds , that @xmath1024 is constant on its fibers . if @xmath850 , then we obtain , using ( [ expproduct ] ) , that @xmath1036 which implies ( [ tauplus ] ) .
let @xmath1037 , @xmath1038 , be a @xmath13-basis of @xmath38 .
for any @xmath999 we have @xmath1039 for unique integral coordinates @xmath1040 . with the notation @xmath1041 ,
the formula ( [ tauplus ] ) leads to the formula @xmath1042 for @xmath1010 in terms of the elements @xmath1043 .
in other words , all holonomies at a given point @xmath59 can be expressed in terms of the holonomies of the basic loops @xmath1044 , @xmath1045 , by means of the formula ( [ tauzetaform ] ) . in this subsection , let @xmath1046 be one of the delzant submanifolds of @xmath16 in proposition [ embedprop ] .
that is , @xmath28 is an integral manifold @xmath875 of the distribution @xmath27 , and @xmath1047 , if @xmath1048 denotes the inclusion mapping from @xmath875 to @xmath18 .
recall that all delzant manifolds with the same delzant polytope are @xmath24-equivariantly symplectomorphic , which means that one may identify @xmath1049 with any favourite explicit model of a delzant manifold with delzant polytope @xmath10 .
we will construct a model for our symplectic @xmath0-manifold @xmath16 by means of the mapping @xmath1050 which is defined by @xmath1051 write @xmath1024 for the common value of the @xmath1010 for all @xmath1052 , see lemma [ tplem ] .
define @xmath1053 then @xmath27 is a closed lie subgroup of @xmath19 , commutative because @xmath1020 is commutative , see lemma [ tplem ] .
furthermore , @xmath1054 defines a smooth action of @xmath27 on the delzant manifold @xmath28 . the mapping _
( [ a ] ) _ induces a diffeomorphism @xmath769 from @xmath20 onto @xmath18 , where @xmath185 acts on @xmath1055 by sending @xmath1056 to @xmath1057 .
the diffeomorphism @xmath769 intertwines the action of @xmath19 on @xmath20 , which is induced by the action @xmath1058 of @xmath19 on @xmath1055 , with the action of @xmath19 on @xmath18 , and therefore also the action of the normal subgroup @xmath1059 of @xmath19 with the @xmath0-action on @xmath18 .
the projection @xmath1060 induces a @xmath19-equivariant smooth fibration @xmath1061 , and @xmath1062 is a @xmath19-equivariant smooth fibration of which the delzant submanifolds of @xmath18 introduced in proposition [ embedprop ] are the fibers .
[ aprop ] let @xmath1063 and @xmath1064 .
for each @xmath732 , the projection @xmath878 defined in lemma [ pi2lem ] intertwines the diffeomorphism @xmath1015 in @xmath18 with the translation in @xmath467 over the vector @xmath743 .
because these translations act transitively on @xmath467 , there exists a @xmath732 such that @xmath1065 , which implies that @xmath1066 belongs to the same fiber @xmath137 of @xmath1067 as @xmath340 .
with such a choice of @xmath743 , it follows from lemma [ embedlem ] that there exists @xmath905 and @xmath1052 such that @xmath1068 , or equivalently @xmath1069 .
this shows that already the restriction of @xmath385 to @xmath1070 is surjective .
let @xmath1071 , @xmath1072 and @xmath1073 . then @xmath1074 in which @xmath1075 .
write @xmath1076 with @xmath169 and @xmath732 .
then @xmath1077 implies that @xmath999 , and it follows from lemma [ tplem ] that @xmath1078 because @xmath1079 and @xmath1080 belong to the same integral manifold @xmath875 of @xmath27 , it follows from lemma [ embedlem ] that the element @xmath1081 of @xmath31 is equal to the identity element , hence @xmath1082 and @xmath1083 . in other words , @xmath185 , @xmath1084 and @xmath1083 .
this proves that the mapping @xmath385 induces a bijective mapping @xmath769 from @xmath20 onto @xmath18 .
the closedness of @xmath27 in @xmath19 implies that the right action of @xmath27 on @xmath19 is proper and free , hence the action of @xmath27 on @xmath1055 is proper and free , and the orbit space @xmath20 has a unique smooth structure for which the projection @xmath1085 is a principal @xmath27-bundle . with respect to this smooth structure on @xmath20 ,
the mapping @xmath1086 is smooth .
the transversality to @xmath1087 of the span of @xmath1088 , @xmath938 and the @xmath702 , @xmath732 , implies that at every point the tangent mapping of @xmath385 is surjective .
hence @xmath769 is a submersion , and because @xmath769 is bijective , it follows from the inverse mapping theorem that @xmath769 is a diffeomorphism .
the other statements in the proposition are general facts about induced fiber bundles @xmath20 over @xmath21 with fiber @xmath28 , see for instance ( * ? ? ?
on @xmath21 we still have the free action of the torus @xmath1089 , which exhibits @xmath21 as a principal @xmath1089-bundle over the torus @xmath1090 .
palais and stewart @xcite showed that every principal torus bundle over a torus is diffeomorphic to a nilmanifold for a two - step nilpotent lie group . in this remark
we will give an explicit nilmanifold description of @xmath21 .
the hamiltonian torus @xmath24 , or rather the identity component @xmath1091 of @xmath27 , is a closed normal lie subgroup of both @xmath1092 and @xmath27 , and the mapping @xmath1093 is a @xmath19-equivariant diffeomorphism from @xmath1094 onto @xmath21 .
the group structure in @xmath1095 is defined by @xmath1096 and @xmath1097 is equal to @xmath725 , followed by the projection @xmath1098 .
this exhibits @xmath1099 as a two - step nilpotent lie group with universal covering equal to @xmath1100 and covering group @xmath1101 .
also note that @xmath1102 is an isomorphism from the period group @xmath38 onto @xmath1103 . in view of ( [ hdef ] )
, we conclude that the compact homogeneous @xmath19-space @xmath21 is isomorphic to the quotient of the simply connected two - step nilpotent lie group @xmath1100 by the discrete subgroup of @xmath1100 which consists of all @xmath1104 such that @xmath1105 .
[ horem ] in proposition [ aprop ] we have described the global model @xmath1106 for the @xmath0-manifold @xmath18 , where the multiplication in the lie group @xmath1092 is defined by ( [ hproduct ] ) .
we now describe the symplectic form on @xmath1107 .
let @xmath48 be the pull - back of @xmath17 to @xmath1108 by means of the mapping @xmath385 in ( [ a ] ) .
let @xmath1109 and @xmath1110 be tangent vectors to @xmath1055 at @xmath1111 , where we identify each tangent space of the torus @xmath0 with @xmath12 .
write @xmath1112 and @xmath1113
. then @xmath1114 here @xmath1115 denotes the @xmath26-component of @xmath54 with respect to the direct sum decomposition @xmath1116 .
if @xmath1117 denotes the canonical projection from @xmath1055 onto @xmath1106 , then the @xmath0-invariant symplectic form @xmath1118 on @xmath1107 is the unique two - form @xmath770 on @xmath1107 such that @xmath1119 .
[ omegalem ] it follows from ( [ expproduct ] ) that @xmath1120 therefore , if we substitute @xmath1121 and differentiate with respect to @xmath1122 at @xmath1123 , we get the vector @xmath1124 at the image point under the mapping @xmath1015 .
because @xmath1015 commutes with the @xmath0-action , it follows , with the notations @xmath1125 and @xmath1126 , that @xmath1127 in which @xmath1112 .
if @xmath73 is a regular point in @xmath28 , then we can write @xmath1128 for uniquely determined @xmath868 and @xmath1129 .
the vector fields @xmath867 , @xmath868 , and @xmath869 , @xmath731 , commute with the vector fields @xmath43 , @xmath54 , and @xmath702 , @xmath732 ,
because of ii ) in proposition [ liftprop ] and the fact that all the vector fields are @xmath0-invariant . therefore @xmath1130 , and we obtain that @xmath1131 is equal to the value at @xmath1125 of the vector field @xmath1132 . in view of ( [ stronglift ] ) and
iv ) , v ) in proposition [ liftprop ] , the symplectic product of this vector with the one in which @xmath1133 , @xmath1134 , @xmath360 , @xmath926 are replaced by @xmath1135 , @xmath1136 , @xmath1137 , @xmath1138 , respectively , is equal to @xmath1139 in which @xmath1140 and @xmath1141 .
collecting terms and using the equations @xmath1142 , @xmath1143 , @xmath1144 , and vi ) in proposition [ liftprop ] , we arrive at ( [ a*sigma ] ) . because @xmath1145 , we have @xmath1146 . the uniqueness in the last statement
follows because @xmath1117 is a submersion .
let @xmath31 be a complementary torus to the hamiltonian torus @xmath24 in @xmath0 .
then the following conditions are equivalent .
* @xmath16 is @xmath0-equivariantly symplectomorphic to the cartesian product of a symplectic @xmath31- space @xmath1147 on which the @xmath31-action is free and a delzant manifold @xmath1148 . here
@xmath169 acts on @xmath1149 by sending @xmath1150 to @xmath1151 , if @xmath1152 with @xmath905 and @xmath904 .
* @xmath1153 .
* @xmath1154 . * the @xmath26-component @xmath814 of @xmath726 in the direct sum decomposition @xmath718 is equal to zero .
[ e0rem ] the equivalence between b ) and c ) follows from the fact that @xmath38 has a @xmath13-basis which is an @xmath604-basis of @xmath39 .
the equivalence between c ) and d ) is obvious .
if @xmath16 is equal to the cartesian product of a delzant manifold @xmath1046 and a symplectic @xmath31-space @xmath1147 for which the @xmath31-action on @xmath1155 is free , then we can choose the @xmath702 in the direction of the second component @xmath1155 .
in this case we have for every @xmath734 that @xmath1156\in { \mathfrak{t}}_{{\!\!\mathop{\mbox{\rm \scriptsize ~f}}\nolimits}}$ ] , which means that the antisymmetric bilinear mapping @xmath725 in proposition [ liftprop ] has the property that @xmath1157 , or equivalently @xmath1158 . for
the converse , assume that @xmath1157 , which implies that @xmath1158 .
then the lie group @xmath19 is equal to the cartesian product @xmath1159 , in which @xmath1160 , where the product in @xmath1161 is defined as in ( [ hproduct ] ) with @xmath0 replaced by @xmath31 . according to subsection [ holinvariantss ]
we can multiply the elements @xmath1162 , for @xmath1052 , by any element of @xmath1163 . because @xmath474 , it follows that we can arrange that @xmath1164 for every @xmath1045 , and then it follows from ( [ tauzetaform ] ) that @xmath1165 for every @xmath999 .
the mapping @xmath1166 is a homomorphism from @xmath38 onto a discrete cocompact subgroup of @xmath1161 .
write @xmath1167 .
it follows that the mapping @xmath1168 induces a diffeomorphism @xmath1169 from @xmath1170 onto @xmath18 .
moreover , it follows from ( [ a*sigma ] ) that the symplectic form @xmath1171 on @xmath1170 is equal to @xmath1172 , if @xmath1173 and @xmath1174 is the projection form @xmath1170 onto the first and the second factor , respectively , and the symplectic form @xmath1175 on @xmath1155 is given by @xmath1176 here @xmath1177 , the tangent vectors @xmath1178 and @xmath1179 are elements of @xmath1180 , the vectors @xmath1181 and @xmath1182 are elements of @xmath715 , and finally @xmath1183 because in vi ) in proposition [ liftprop ] we have @xmath1158 .
it follows that @xmath16 is @xmath0-equivariantly symplectomorphic to @xmath1184 , in which @xmath1147 is a compact connected symplectic manifold with a free symplectic action @xmath31-action . in the proof of lemma [ e0rem ]
we have also given a global model for @xmath16 in the case that the action of @xmath0 is free .
note that the mapping @xmath1185 defines an isomorphism from @xmath1161 onto the group @xmath1099 in remark [ horem ] , and an isomorphism from @xmath1186 onto @xmath1103 , which leads to an identification of @xmath1187 with the manifold @xmath21 . in remark [ horem ]
, @xmath21 has been described as a principal @xmath31-bundle over the torus @xmath467 , and as a nilmanifold for a two - step nilpotent lie group .
the first example of thurston @xcite is equal to @xmath1188 with @xmath1189 , @xmath1190 , @xmath1191 , @xmath295 , @xmath1192 , and @xmath1193 if @xmath1194 , @xmath1195 denotes the standard basis in @xmath1196 . for more examples ,
see mcduff and salamon ( * ? ? ?
3.8 on p.88 ) and the references therein .
[ freerem ] if @xmath1197 , then the choice of a complementary torus @xmath31 to @xmath24 in @xmath0 is far from unique , see remark [ tbaserem ] .
it can happen that for some choice of @xmath31 we have @xmath1157 , whereas for another choice we have not . however , if @xmath1198 and @xmath235 , then there is no choice of a complementary torus @xmath31 to @xmath24 such that @xmath1157 , and therefore @xmath16 is in no way @xmath0-equivariantly symplectomorphic to a cartesian product of a symplectic manifold with a free torus action and a delzant manifold .
[ makee0rem ] if @xmath1199 then @xmath1200 because every antisymmetric bilinear form on a one - dimensional space is equal to zero , and we conclude that @xmath16 is a cartesian product of a delzant manifold with a two - dimensional homogeneous symplectic torus .
if @xmath18 is four - dimensional , that is , @xmath1201 , then we have only few possibilities .
* the homogeneous symplectic torus , where @xmath0 acts freely and transitively on @xmath18 . cases where @xmath0 is replaced by a subtorus with coisotropic orbits are treated as subcases . * @xmath16 is @xmath0-equivariantly symplectomorphic with the cartesian product of a two - dimensional homogeneous symplectic torus and a sphere , provided with a rotationally invariant area form .
* @xmath1202 is a four - dimensional delzant manifold . *
the action of @xmath0 is free with lagrangian orbits , but not in case a ) .
see the proof of lemma [ e0rem ] for a more detailed description of this case .
the example of thurston mentioned in remark [ freerem ] was the first one in the literature .
[ fourdimrem ] in view of the surjectivity of the mapping @xmath385 in proposition [ aprop ] , lemma [ tplem ] contains the description of the dependence of the holonomy @xmath1203 on all points @xmath59 .
the only change which occurs is that if @xmath73 is replaced by @xmath1204 , @xmath1205 , then @xmath1010 is replaced by @xmath1206 , see ( [ tauzeta ] ) .
we now investigate the dependence of the holonomy on the choice of the admissible connection as in proposition [ liftprop ] .
it follows from lemma [ liftfreedomlem ] that @xmath858 is another connection as in proposition [ liftprop ] , if and only if there exists a smooth @xmath0-invariant mapping @xmath859 from @xmath18 to @xmath860 , which is _ closed _ when viewed as an @xmath40-valued one - form on @xmath3 , and _ symmetric _ in the sense of ( [ alphasym ] ) , such that @xmath861 for every @xmath59 and @xmath588 .
the change from @xmath862 to @xmath863 leads to a change from @xmath1207 to @xmath1208 here @xmath1209 , @xmath1001 , is a loop in @xmath3 because @xmath999 .
because @xmath769 is closed , the integral @xmath1210 only depends on the de rham cohomology class of @xmath769 , which means that for the effect on the @xmath1207 s we can restrict ourselves to constant @xmath40-valued one - forms on @xmath34 , that is , linear mappings from @xmath34 to @xmath40 , or equivalently , bilinear forms on @xmath34 . therefore , at a given point @xmath59 , the allowed changes in @xmath1207 , @xmath999 , consist of the multiplications with @xmath1211 , where @xmath769 ranges over the space of linear mappings @xmath1212 from @xmath34 to @xmath40 , which are symmetric in the sense of ( [ alphasym ] ) .
let @xmath1213 denote the space of mappings @xmath1214 such that @xmath1215 because @xmath1216 , the factor @xmath1217 in ( [ tauplushom ] ) is an element of order two in @xmath0 .
therefore the elements of @xmath1213 are quite close to being homomorphisms from @xmath38 to @xmath0 .
they are homomorphisms from @xmath38 to @xmath0 if @xmath1218 . [ homcdef ] if @xmath1219 is a homomorphism from @xmath38 to @xmath0 , then @xmath1220 for every @xmath1221 , and @xmath1222 defines a free , proper , and transitive action of @xmath1223 on @xmath1213 . if @xmath1037 , @xmath1038 is a @xmath13-basis of @xmath38 , then the mapping @xmath1224 is an isomorphism from @xmath1223 onto the torus @xmath1225 .
therefore @xmath1213 is diffeomorphic to a torus of dimension @xmath1226 . for each @xmath1205
, @xmath1227 is a homomorphism from @xmath38 to @xmath12 , actually @xmath40-valued .
write @xmath1228 for the set of all @xmath1229 such that @xmath1205 .
@xmath1228 is a linear subspace of the lie algebra @xmath1230 of @xmath1223 .
let @xmath1231 denote the space of all linear mappings @xmath1232 which are symmetric in the sense of ( [ alphasym ] )
. for each @xmath1233 , the restriction @xmath1234 of @xmath769 to @xmath38 is a homomorphism from @xmath38 to @xmath1235 . in this way
the set @xmath1236 of all @xmath1234 such that @xmath1233 is another linear subspace of @xmath1230 .
write @xmath1237 for the orbit space of the action of the lie subgroup @xmath1238 of @xmath1223 on @xmath1213 . because @xmath1238 need not be a closed subgroup of @xmath1223 , the quotient topology of @xmath1239 need not be hausdorff .
it follows from ( [ tauplus ] ) and ( [ tauplushom ] ) , that for every choice of a connection as in proposition [ liftprop ] and every @xmath59 , the mapping @xmath1240 is an element of @xmath1213 .
it is the point of ( [ calt ] ) , that the right hand side in @xmath1241 defines an invariant @xmath1242 of our symplectic @xmath0-space @xmath16 , in the sense that it neither depends on the choice of the point @xmath59 , nor on the choice of the connection as in proposition [ liftprop ] .
[ overlinetaudef ] in order to obtain some more insight in the vector space @xmath1243 in ( [ calt ] ) , we use the direct sum decomposition @xmath1244 , where @xmath474 is the lie algebra of the hamiltonian torus @xmath24 and @xmath715 is the lie algebra of a complementary torus @xmath31 to @xmath24 in @xmath0 .
this leads to an identification of @xmath466 with @xmath1245 and of its linear complement @xmath85 in @xmath34 with @xmath499 .
let @xmath1246 denote the space of all linear mappings @xmath1247 , which satisfy the symmetry condition ( [ alphasym ] ) with @xmath40 replaced by @xmath1248 .
the space @xmath1236 of all restrictions to @xmath1249 of linear mappings @xmath1232 which satisfy the the symmetry condition ( [ alphasym ] ) is equal to the direct sum of the space @xmath1250 of _ all _ homomorphisms from @xmath38 to @xmath26 , and the space @xmath1246 .
this means that in the space @xmath1239 in ( [ calt ] ) we dispose of the @xmath24- components , and in the computation of @xmath1243 we can replace @xmath726 by its @xmath1248-component @xmath1251 .
now suppose that @xmath1205 and @xmath1252 .
this is equivalent to the condition that @xmath1253 for all @xmath1254 .
here we have used ( [ ft ] ) in the first equality . in the second equality
we have used the antisymmetry of @xmath726 and the fact that the elements @xmath1255 are equal to zero on @xmath26 . in other words , @xmath1252 if and only if @xmath1256 on @xmath1257 , or equivalently @xmath1256 on the linear subspace of @xmath40 which is spanned by @xmath1258 .
let @xmath1259 denote the space of all @xmath1260 which are equal to zero on the linear span of @xmath1258 .
in view of ( [ ft ] ) we have @xmath1261 , and it follows that the dimension of @xmath1262 is equal to @xmath1263 , whereas the dimension of @xmath1264 is equal to @xmath1265 .
it follows that the dimension of @xmath1266 is equal to @xmath1267 , in which @xmath1268 .
therefore the codimension of @xmath1266 in the @xmath1269-dimensional space @xmath1270 is equal to @xmath1271 .
because all elements of @xmath1243 map to @xmath40 , it follows that @xmath1272 [ dimholinvariantrem ]
in this section , which is not needed for the classification in section [ invariantsec ] , we give some applications of proposition [ aprop ] to minimal coupling , the reduced phase spaces , the topology of the torus action , and to the universal covering of our symplectic @xmath0-space @xmath18 .
the fibration of @xmath18 by delzant manifolds is a fibration by symplectic submanifolds with a structure group @xmath27 which acts on the fiber by means of symplectomorphisms .
see proposition [ aprop ] . moreover , the distribution spanned by the @xmath1088 , @xmath938 , and the @xmath702 , @xmath732 , which we used in the construction of the model , is the symplectic orthogonal complement of the fibers .
this follows from the fact that at the regular points the tangent space to the delzant submanifold is spanned by the @xmath867 , @xmath868 , and the @xmath869 , @xmath731 , combined with lemma [ constlem ] and @xmath1273 , the equation ( [ stronglift ] ) , and v ) in proposition [ liftprop ] . because the @xmath1088 , @xmath938 , commute with each other and with the @xmath702 , @xmath732 , the only nonzero lie brackets of horizontal vector fields are the @xmath1274= c(\zeta , \,\zeta ' ) _ m$ ] , @xmath734 , see iii ) in proposition [ liftprop ] .
the vertical part of @xmath1274 $ ] is equal to @xmath1275 . because , for every @xmath868 , @xmath867 is the hamiltonian vector field defined by the function @xmath1276 , the vertical part of @xmath1274 $ ] is the hamiltonian vector field defined by the function @xmath1277 .
it follows from vi ) in proposition [ liftprop ] that the derivative of this function is equal to the negative of the derivative of @xmath1278 , and therefore the vertical part of @xmath1156 $ ] is equal to @xmath1279 .
this equation , which holds in great generality for the curvature of the symplectically orthogonal connection in a fibration by symplectic manifolds , is known as _ minimal coupling _ , see guillemin , lerman and sternberg ( * ? ? ?
in this way equation vi ) in proposition [ liftprop ] represents the minimal coupling term in the symplectic form on @xmath18 .
this observation was suggested to us by yael karshon .
recall that the fibration of @xmath18 by delzant manifolds was not a priori given .
it has been constructed using the special admissible connection introduced in proposition [ liftprop ] , and it is not unique if @xmath1197 . on the symplectic manifold @xmath1 we have the hamiltonian action of the torus @xmath24 , with momentum mapping @xmath1280 , where @xmath1281 .
let @xmath1282 . then , restricting the discussion to the orbit type stratum which contains @xmath1283 , we obtain that @xmath1283 is a compact and connected smooth submanifold of @xmath18 , on which @xmath1284 acts freely , where @xmath27 denotes the common stabilizer subgroup of the elements in @xmath1283 .
it follows that the orbit space @xmath1285 has a unique structure of a compact connected smooth manifold , such that the projection @xmath1286 is a principal @xmath1284-fibration . at each point of @xmath1283 ,
the kernel of the pull - back to @xmath1283 of @xmath17 is equal to the tangent space of the @xmath24-orbit through that point , and it follows that there is a unique symplectic form @xmath1287 on @xmath1288 such that @xmath1289 , if @xmath1290 denotes the inclusion mapping from @xmath1283 to @xmath18 .
the symplectic manifold @xmath1291 is called the _ reduced phase space at the @xmath500-value @xmath1292 _ for the hamiltonian action of @xmath24 on @xmath1 .
on @xmath1288 we still have the action of the torus @xmath1089 , which is free , leaves the symplectic form @xmath1287 invariant , and has coisotropic orbits . the vector fields @xmath702 , @xmath732 , are tangent to @xmath1283 , and are intertwined by @xmath1293 with unique smooth vector fields @xmath1294 on @xmath1288 . in combination with @xmath1289 ,
this implies that @xmath1295 , as an identity between constant functions on @xmath1283 .
it therefore follows from vi ) in proposition [ liftprop ] that @xmath1296 we now show that each of the reduced phase spaces @xmath1297 can be identified with the @xmath19-homogeneous space @xmath1298 discussed in remark [ horem ] .
moreover , if @xmath1198 , then ( [ sigmaq ] ) corresponds to the description of the variation of the cohomology class of the symplectic form of the reduced phase spaces in duistermaat and heckman @xcite . let @xmath1299 , and write @xmath1300 because @xmath74 is a closed lie subgroup of @xmath24 , @xmath597 is a closed lie subgroup of @xmath27 , see ( [ hdef ] ) , and @xmath1301 is a compact @xmath19-homogeneous space .
the mapping @xmath1302 induces an embedding @xmath79 from @xmath1301 into @xmath18 , with image equal to @xmath1303 .
this exhibits @xmath1303 as a compact and connected smooth submanifold of @xmath18 , and actually as a @xmath19-homogeneous space .
the pull - back to @xmath1301 of the symplectic form @xmath17 is given by the formula ( [ a*sigma ] ) , in which @xmath1304 . because @xmath1305 , the mapping @xmath79 induces a @xmath1089-equivariant diffeomorphism @xmath1306 from @xmath1307 onto the reduced phase space @xmath1308 . because the dimension of @xmath1303 jumps down
if @xmath1309 moves into a lower - dimensional orbit type stratum , it is quite remarkable that nevertheless the reduced phase spaces @xmath1288 for all @xmath1309 are isomorphic to the same space @xmath21 in a natural way . in this model
the principal @xmath1310-fibration @xmath1311 corresponds to the principal @xmath1312-fibration @xmath1313 , in which @xmath1305 is a torus .
if @xmath1314 , then the chern class of the principal @xmath1312-fibration @xmath1315 , which is an element of @xmath1316 is equal to @xmath1317 , in which @xmath1318 is the cohomology class corresponding to the antisymmetric bilinear form @xmath726 introduced in proposition [ liftprop ] , and @xmath1319 is the projection from @xmath21 onto @xmath1320 . therefore , in the case that @xmath1314 , formula ( [ sigmaq ] ) shows that the variation of the cohomology class of the symplectic form of the reduced phase spaces is equal to the cohomology class @xmath1321 of the curvature form .
the action of the hamiltonian torus @xmath31 on @xmath18 has fixed points , which are the @xmath59 such that @xmath515 is equal to a vertex @xmath46 of the delzant polytope @xmath10 , if @xmath1322 denotes the momentum mapping of the hamiltonian @xmath24-action as in ( [ pi^1 ] ) .
let @xmath46 be a vertex of @xmath10 . because @xmath1323 for every @xmath1324 , the reduced phase space @xmath1325 at the level @xmath46 , introduced in subsection [ redg / hrem ] , is equal to @xmath1326 . because the reduced phase spaces are connected , the @xmath1326 , where @xmath46 ranges over the vertices of @xmath10 , are the connected components @xmath1327 of the fixed point set @xmath1328 of the @xmath24-action in @xmath18 . because in this subsection we want to find invariants of the @xmath0-action , disregarding the symplectic structure , we use the notation @xmath1327 for the connected components of @xmath1328 , instead of @xmath1326 .
note that each @xmath1327 is a global section of the fibration @xmath1329 of @xmath18 by delzant submanifolds .
using morse theory with the hamiltonian functions of infinitesimal @xmath24-actions as bott - morse functions , this may lead to useful information about the topology of @xmath18 in terms of the connected components @xmath1327 of @xmath1328 .
let @xmath31 be a complementary torus to @xmath24 in @xmath0 .
because the action of @xmath31 is free , we have the principal @xmath31-fibration @xmath1330 , and because the actions of @xmath31 and @xmath24 commute , we have an induced action of @xmath24 on @xmath30 . the manifolds @xmath1331 are the connected components of the fixed point set @xmath1332 of the @xmath31-action in @xmath30 .
the fibration @xmath1329 induces a fibration @xmath1333 by delzant manifolds , of which each connected component @xmath1334 of the @xmath24-fixed point set is a global section , diffeomorphic to @xmath467 .
let @xmath1335 , and write @xmath1336 .
the tangent action of @xmath24 on the normal space @xmath1337 to @xmath1338 can be identified with the tangent action of @xmath24 on the tangent space of the delzant manifold through @xmath73 .
it follows from the local model in lemma [ modellem ] , with @xmath1339 , @xmath1340 , and @xmath1341 , that @xmath1342 has a direct sum decomposition into @xmath24-invariant two - dimensional linear subspaces @xmath1343 , @xmath1344 , a complex structure on each @xmath1343 , and a corresponding @xmath13-basis @xmath1345 , @xmath1344 of the integral lattice @xmath1346 in @xmath26 , such that the tangent action of @xmath1347 , @xmath868 on @xmath1342 corresponds to the multiplication with @xmath1348 in @xmath1343 , if @xmath1349 .
although for their existence we referred to the local model in lemma [ modellem ] for our symplectic @xmath0-space , all these ingredients are uniquely determined in terms of the linearized action of @xmath24 on the normal bundle @xmath39 of @xmath1350 in @xmath1351 , up to a permutation of the indices @xmath391 . that is , disregarding the symplectic structure . for each @xmath391 , the @xmath1343 , @xmath1352 , form a complex line bundle @xmath1353 over @xmath1354 , and
the normal bundle @xmath39 of @xmath1338 in @xmath361 is the direct sum of the complex line bundles @xmath1353 , @xmath1344 .
any smooth complex line bundle @xmath862 over a smooth manifold @xmath1355 has a chern class , which is defined as follows .
let @xmath1356 denote the multiplicative group of the nonzero complex numbers .
the transition functions of local trivializations define a 1-cocycle of germs of smooth @xmath1356-valued functions , and the bundle @xmath862 is classified by the sheaf (= ech ) cohomology class @xmath1357 of the 1-cocycle of the transition functions .
because the sheaf @xmath1358 is fine , the short exact sequence @xmath1359 induces an isomorphism @xmath1360 , and the _ chern class of the complex line bundle @xmath862 over @xmath1355 _ is defined as the cohomology class @xmath1361 . with these definitions , we have the following conclusions .
let the @xmath1362 , @xmath1344 , be defined by @xmath1363 let @xmath1344 .
viewing @xmath1364 as an element of @xmath1365 as in corollary [ cohomcor ] , we have that @xmath1364 is equal to the image in @xmath1366 under the coefficient homomorphism @xmath1367 , of the chern class @xmath1368 of the complex line bundle @xmath1353 over @xmath1369 . if @xmath18 is @xmath0-equivariantly diffeomorphic to @xmath1170 , in which @xmath24 acts only on @xmath28 with isolated fixed points , and @xmath31 acts only on @xmath1155 and freely , then @xmath1158 , and we have the conclusions c ) and a ) in lemma [ e0rem ] .
[ chernprop ] because the vector fields @xmath702 , @xmath732 , are invariant under the action of @xmath0 , hence under the action of @xmath24 and @xmath31 , they are intertwined by the projection @xmath1370 to uniquely determined @xmath24-invariant smooth vector fields on @xmath361 , which we also denote by @xmath702 .
the identity iii ) in proposition [ liftprop ] leads to the identity @xmath1156=
c_{{\!\!\mathop{\mbox{\rm \scriptsize ~h}}\nolimits}}(\zeta , \,\zeta ' ) _ { \overline{m}}$ ] for vector fields on @xmath361 . because the @xmath702 are @xmath24-invariant
, their flows leave each connected component @xmath1338 of the @xmath24-fixed point set @xmath1371 invariant , and their linearizations define automorphisms of the normal bundle @xmath39 of @xmath1338 in @xmath361 which commute with the linearized action of @xmath24 on @xmath39 . therefore these automorphisms leave each of the complex line bundles @xmath1353 invariant , and the corresponding infinitesimal automorphisms define vector fields on @xmath183 which we again denote by @xmath702 . because the @xmath702 are lifts of the constant vector fields @xmath743 on @xmath467 ,
we conclude that @xmath1372 is a @xmath1373-invariant connection in @xmath1353 , where @xmath1373 is the unit circle in @xmath208 . because the cohomology class in @xmath1374 of the negative of the curvature form is equal to the image of @xmath1368 in @xmath1366 under the coefficient homomorphism @xmath1375 , the first statement in the proposition follows from the combination of the above discussions with the identifications in remark [ curvrem ] and remark [ integralcrem ] , and the general facts about chern classes of complex line bundles as for instance in bott and tu @xcite .
for the second statement assume that @xmath1376 , in which @xmath24 acts only on @xmath28 and has isolated fixed points , and @xmath31 acts freely on @xmath1155 .
then @xmath1377 in which @xmath24 only acts on the second component .
it follows that the connected components of @xmath1371 are of the form @xmath1378 in which @xmath73 ranges over the isolated fixed points of the @xmath24-action on @xmath28 , and the normal bundle @xmath39 of @xmath1338 in @xmath361 is @xmath24-equivariantly isomorphic to @xmath1379 .
this shows that each of the complex line bundles @xmath1353 is trivial , which implies that @xmath1380 and therefore @xmath1381 in view of the first statement in the proposition .
because this holds for every @xmath1344 , it follows that @xmath1158 . in proposition [ pi1prop ]
below , we describe an explicit universal covering of the manifold @xmath18 by a cartesian product @xmath1382 of a vector space and the delzant manifold @xmath28 , which leads to an explicit description of the fundamental group of @xmath18 . in remark [ benoist6.16rem ]
we recover corollaire 6.16 of benoist @xcite , which states that the universal cover of a compact connected symplectic @xmath0-manifold with coisotropic principal orbits is @xmath1383-equivariantly symplectomorphic to the cartesian product of a symplectic vector space and a delzant manifold .
let @xmath1037 , @xmath1384 , be a @xmath13-basis of the period group @xmath38 in @xmath39 .
if @xmath1385 have coordinates @xmath1386 with respect to this basis , then we write @xmath1387 this defines a bilinear mapping @xmath1388 such that @xmath1389 .
we have @xmath850 if and only if @xmath1390 for all @xmath1391 .
therefore @xmath1216 , see lemma [ tplem ] , implies that @xmath1392 .
let @xmath1052 . for each @xmath1045
we choose @xmath1393 such that @xmath1394 .
then ( [ tauzetaform ] ) implies that , for each @xmath999 , @xmath1395 where in the second term in the exponent we use einstein s summation convention .
let @xmath31 be the complementary torus to the hamiltonian torus @xmath24 in @xmath0 which has been used in proposition [ liftprop ] .
finally , let @xmath226 , @xmath1396 , be a @xmath13-basis of the integral lattice @xmath1397 in the lie algebra @xmath715 of @xmath31 .
for any @xmath54 we denote by @xmath1115 and @xmath1398 the @xmath26-component and the @xmath715-component of @xmath42 , respectively . with these notations
, we have the following conclusions .
the lattice @xmath1399 is a group with respect to the multiplication defined by @xmath1400 let @xmath1401 .
let @xmath1402 act on @xmath1382 by sending @xmath1403 to @xmath1404 , where @xmath1405 this defines a proper and free action of @xmath1406 on @xmath1382 , and the mapping @xmath1407 is a universal covering of @xmath18 with the action of @xmath1406 on @xmath1382 as the covering group .
let @xmath1052 and let @xmath1408 be the fundamental group of @xmath18 with base point @xmath73 . for any homotopy class @xmath1409
$ ] of a closed loop @xmath1410 based at @xmath73 , let @xmath1411)$ ] be the element of @xmath1406 of which the action on @xmath1382 is equal to the covering transformation defined by @xmath1410 .
let @xmath1412 be the closed loop @xmath1413 , @xmath1001 , and let @xmath1414 be the closed loop based at @xmath73 which consists of @xmath1415 , @xmath1001 , followed by @xmath1416 , @xmath1001 .
then the isomorphism @xmath1417 is uniquely determined by the condition that @xmath1418)=(z_j,\ , 0)$ ] , @xmath1419 , and @xmath1420)=(0,\,\varepsilon ^l)$ ] , @xmath1045 .
[ pi1prop ] let @xmath1064 .
the surjectivity of the mapping @xmath385 in ( [ a ] ) , see proposition [ aprop ] , implies that there exist @xmath169 , @xmath732 , @xmath1052 , such that @xmath1421 .
we have @xmath1422 with @xmath904 and @xmath905 , and subsequently there exists @xmath938 such that @xmath1423 . because @xmath1424 commutes with @xmath1425 and leaves @xmath28 invariant , it follows that @xmath1426 , which proves that the mapping @xmath1427 is surjective .
let @xmath1428 .
the injectivity of the mapping @xmath769 in proposition [ aprop ] implies that there exist @xmath1429 such that @xmath1430 and @xmath1431 . in view of ( [ hproduct ] ) and ( [ hdef ] )
, this implies that @xmath1432 , @xmath1433 , @xmath1434 , @xmath1435 , and @xmath1436 . in view of @xmath1437 and ( [ xzeta ] ) with @xmath1438
, the @xmath31-part and the @xmath24-part of the equation for @xmath1439 mean that @xmath1440 and @xmath1441 , respectively .
it follows that the fibers of @xmath1427 are the @xmath1406-orbits , if we let @xmath1442 act on @xmath1382 as in ( [ gammaact ] ) .
note that @xmath1443 implies that @xmath1444 , and then @xmath1445 implies that @xmath1446 .
therefore the action of @xmath1406 on @xmath1382 is free , which implies that it is effective , in the sense that the mapping from @xmath1406 to the set of diffeomorphisms of @xmath1382 is injective .
there is a group structure on @xmath1406 for which the action of @xmath1406 is a group action , a homomorphism from @xmath1406 to the group of diffeomorphisms of @xmath1382 , if and only if the composition of the actions of two elements of @xmath1406 is an action of an element of @xmath1406 .
the effectiveness of the action implies that if this is the case , then the group structure on @xmath1406 for which this holds is unique . if we let @xmath1447 act on ( [ gammaact ] ) , then we arrive at @xmath1448 , in which @xmath1449 @xmath1450 , and @xmath1451 here we have used that @xmath1452 in the equation for @xmath1453 .
furthermore , in the equation for @xmath1454 we have used ( [ tauplus ] ) and the fact that if @xmath1455 , then @xmath1456 , hence @xmath1457 , and therefore @xmath1458 .
this proves that @xmath1406 is a group with respect to the multiplication defined by ( [ gammadef ] ) , and that ( [ gammaact ] ) defines a group action of @xmath1406 on @xmath1382 .
because the action of @xmath1406 on @xmath1382 is obviously proper and free , we conclude that @xmath1427 is a covering with covering group equal to the action of @xmath1406 . because @xmath1382 is simply connected as the cartesian product of a vector space and the simply connected delzant manifold @xmath28 , see lemma [ pi1delzantlem ]
, @xmath1382 is a universal covering of @xmath18 .
it follows from general facts about universal coverings , see for instance greenberg ( * ? ? ?
5 ) , that @xmath1459 is an isomorphism from @xmath1408 onto @xmath1406 . finally @xmath1427 maps the curve @xmath1460 , @xmath1001 , which runs from @xmath1461 to @xmath1462 , to @xmath1412 .
furthermore @xmath1427 maps the curve @xmath1463 , @xmath1001 , followed by the curve @xmath1464 , @xmath1001 , which runs from @xmath1465 to @xmath1466 to @xmath1414 .
this shows that @xmath1418)=(z_j,\ , 0)$ ] and @xmath1420)=(0,\,\varepsilon ^l)$ ] . because the elements @xmath1467 , @xmath1419 , and @xmath1468 , @xmath1045 , together generate @xmath1406 , this proves the last statement in the proposition . viewing @xmath715 as an additive group , the connected commutative lie group @xmath1469 acts on @xmath1382 , where @xmath1470 sends @xmath1403 to @xmath1471 . the covering map
@xmath1472 intertwines the @xmath119-action on @xmath1382 with the @xmath0-action on @xmath18 via the covering homomorphism @xmath1473 , in the sense that @xmath1474 for every @xmath1475 and @xmath1476 .
the fundamental group of @xmath18 is commutative if and only if @xmath1477 . the first homology group @xmath1478 of @xmath18 with coefficients in @xmath13 is isomorphic to @xmath1479 , in which @xmath681 denotes the additive subgroup of @xmath1397 which is generated by the elements @xmath1480 , such that @xmath1455 .
the first betti number @xmath1481 is equal to @xmath1482 .
[ pi1cor ] a straightforward computation shows that @xmath1483 , that @xmath1484 and that the commutator @xmath1485 is equal to @xmath1486
. therefore the subgroup of @xmath1487 generated by the commutators is equal to @xmath1488 , and @xmath1406 is commutative if and only if @xmath1489 if and only if @xmath1490 .
the canonical homomorphism @xmath1491 is surjective with kernel equal to the subgroup of @xmath1408 generated by the commutators , see greenberg ( * ? ? ?
this induces an isomorphism from @xmath1492 onto @xmath1478 . finally the universal coefficient theorem , cf . greenberg ( * ? ? ?
* th . 29.12 ) implies that for any principal ideal domain @xmath1493 , in particular for @xmath1494 , @xmath1495 is isomorphic to @xmath1496 .
therefore @xmath1497 which is equal to @xmath1498 in view of lemma [ coisotropicorbitlem ] .
it follows that the rank of @xmath681 is a purely topological feature of @xmath18 , disregarding both the @xmath0-action and the symplectic structure on @xmath18 .
note that the generators of @xmath1406 mentioned in proposition [ pi1prop ] were defined in terms of the action of @xmath31 and the @xmath1499 , where the latter were defined in terms of both the @xmath0-action and the symplectic form on @xmath18 .
the symplectic form @xmath1500 on the universal covering @xmath1501 is given by ( [ a*sigma ] ) , in which @xmath1502 , @xmath1503 , @xmath1504 are replaced by @xmath1403 , @xmath1505 , @xmath1506 , respectively , with @xmath1507 .
we view the linear form @xmath1508 on @xmath1509 as an element of @xmath1510 .
let @xmath1511 be defined by @xmath1512 then @xmath409 is a diffeomorphism from @xmath1382 onto @xmath1382 , and the symplectic form @xmath1513 on @xmath1514 is given by @xmath1515 that is , @xmath1516 is equal to the cartesian product of a symplectic vector space @xmath1517 and the delzant manifold @xmath1518 .
here @xmath1519 because @xmath409 is @xmath1383-equivariant , we have recovered benoist ( * ? ? ?
6.16 ) , in which the `` cocycle @xmath726 '' is equal to our @xmath32 .
[ benoist6.16rem ]
the model in proposition [ aprop ] , of a compact connected symplectic manifold @xmath1 with an effective symplectic action of a torus @xmath0 of which the principal orbits are coisotropic submanifolds of @xmath1 , has been described in terms of the following ingredients .
let @xmath0 be a given torus . a _ list of ingredients for @xmath0 _ consists of : * an antisymmetric bilinear form @xmath32 on the lie algebra @xmath12 of @xmath0 . * a subtorus @xmath24 of @xmath0 ,
of which the lie algebra @xmath26 is contained in @xmath33 . * a delzant polytope @xmath10 in @xmath1520 with center of mass at the origin . * a discrete cocompact subgroup @xmath38 of the additive subgroup @xmath1509 of @xmath34 . * an antisymmetric bilinear mapping @xmath725 with the following properties . * * for every @xmath850 , the element @xmath1521 belongs to the integral lattice @xmath14 in @xmath12 , the kernel of the exponential mapping @xmath1522 . * * for every @xmath1523 we have that @xmath1524 * an element @xmath1242 of the space @xmath1239 which has been defined in ( [ calt ] ) .
[ ingredientdef ] regarding the delzant polytope @xmath10 in 3 ) in definition [ ingredientdef ] , we have a corresponding delzant manifold @xmath1518 , which is a @xmath1525-dimensional compact connected symplectic manifold , equipped with an effective hamiltonian @xmath24-action on @xmath1518 , for which @xmath10 is equal to the image of the momentum map . in 5b ) , @xmath732 is viewed as a linear form on @xmath40 which vanishes on @xmath26 , so @xmath744 is a real number .
the ingredient 6 ) , the holonomy invariant , has been introduced in subsection [ holinvariantss ] .
as explained there , the space @xmath1239 to which it belongs can have a non - hausdorff quotient topology .
let @xmath18 be a compact and connected smooth manifold , @xmath17 a symplectic form on @xmath18 , and @xmath0 a torus acting effectively on @xmath1 by means of symplectomorphisms and with coisotropic principal orbits .
the _ list of ingredients of @xmath16 _ , as in definition [ ingredientdef ] , consists of : * @xmath1526 is the antisymmetric bilinear form @xmath32 on @xmath12 as defined in lemma [ constlem ] . * @xmath1527 is the hamiltonian torus @xmath24 , the unique maximal stabilizer subgroup @xmath24 for the @xmath0-action on @xmath18 , see remark [ t1rem ] and lemma [ stabilizerlem ] .
* @xmath1528 is the image @xmath1529 of the momentum map @xmath1530 of the @xmath24-action on @xmath1 , which is hamiltonian , cf .
corollary [ tstabcor ] , where we eliminated the translational ambiguity by putting the center of mass of @xmath10 at the origin .
@xmath10 is a translate of the delzant polytope @xmath1531 in proposition [ m / tprop ] .
* @xmath1532 is the period group @xmath38 defined in lemma [ perlem ] with @xmath456 , @xmath457 , and @xmath466 , which according to proposition [ m / tprop ] is a discrete cocompact additive subgroup of @xmath39 .
* @xmath1533 is the antisymmetric bilinear mapping @xmath725 defined in proposition [ liftprop ] .
* @xmath1534 is the holonomy invariant of @xmath16 , the element @xmath1242 of @xmath1239 defined in ( [ overlinetau ] ) .
note that all the ingredients in definition [ ingredientdef ] are defined only in terms of the torus @xmath0 .
[ invariantdef ] let @xmath0 be a torus .
the list of ingredients of @xmath16 is a complete set of invariants for the compact connected symplectic manifold @xmath1 with effective symplectic @xmath0-action with coisotropic principal orbits , in the following sense . if @xmath1535 is another compact connected symplectic manifold with effective symplectic @xmath0-action with coisotropic principal orbits , then there exists a @xmath0-equivariant symplectomorphism @xmath266 from @xmath16 onto @xmath1536 if and only if the list of ingredients of @xmath16 is equal to the list of ingredients of @xmath1536 .
[ invariantthm ] the property 5a ) in definition [ ingredientdef ] follows from remark [ integralcrem ] , and also from lemma [ tplem ] .
equation 5b ) in definition [ ingredientdef ] is the equation ( [ ft ] ) .
suppose that @xmath266 is a @xmath0-equivariant symplectomorphism from @xmath16 onto @xmath1536 .
we will check that the ingredients of @xmath16 and @xmath1537 are the same .
in other words , the ingredients are invariants of the symplectic @xmath0-spaces .
if @xmath58 , then the @xmath0-equivariance of @xmath266 implies that @xmath1538 and @xmath1539 . in combination with @xmath1540 ,
this implies in view of lemma [ constlem ] that @xmath1541 where we have used in the last equation that @xmath1542 is a constant on @xmath1543 .
this proves that @xmath1544 .
the @xmath0-equivariance of @xmath266 implies that @xmath1545 for every @xmath59 , and therefore @xmath1546 . in combination with @xmath1547
, the @xmath0-equivariance of @xmath266 implies that the @xmath34-valued closed basic one - form @xmath35 defined in lemma [ alphalem ] is equal to @xmath1548 .
it follows that @xmath266 induces an isomorphism of locally convex polyhedral @xmath34-parallel spaces from @xmath3 onto @xmath1549 . in view of proposition [ m / tprop
] this implies that @xmath1550 and that @xmath1551 is a translate of @xmath1528 in @xmath499 . because both @xmath1551 and @xmath1528 have their center of mass at the origin , it follows that @xmath1552 .
the @xmath0-equivariant symplectomorphism @xmath266 maps an admissible connection as in proposition [ liftprop ] to an admissible connection as in proposition [ liftprop ] with @xmath16 replaced by @xmath1536 .
it follows that @xmath1553 in view of remark [ curvrem ] and @xmath1554 in view of subsection [ holinvariantss ] .
this proves the `` only if '' part of the theorem . for the `` if '' part , the completeness of the invariants
, we observe that the manifold @xmath1106 and the @xmath0-invariant symplectic form @xmath1555 on @xmath1107 , see proposition [ aprop ] and proposition [ omegalem ] , are defined in terms of the ingredients 1 )
6 ) in definition [ ingredientdef ] , and the elements @xmath1556 , @xmath999 . let @xmath59 and choose an admissible connection for @xmath1557 as in proposition [ liftprop ] .
then @xmath1554 implies in view of subsection [ holinvariantss ] that there exist @xmath1558 and a choice of an admissible connection for @xmath1536 as in proposition [ liftprop ] , such that the holonomy @xmath1559 , @xmath999 , defined by this connection and with the initial point @xmath1079 , is equal to @xmath1560 , @xmath999 .
therefore the model for @xmath1536 in proposition [ aprop ] , with @xmath16 replaced by @xmath1536 , can be chosen to be equal to the model for @xmath16 in proposition [ aprop ] .
this implies the existence of a @xmath0-equivariant symplectomorphism @xmath1561 from @xmath1562 onto @xmath1536 , and it follows that @xmath1563 is a @xmath0-equivariant symplectomorphism from @xmath16 onto @xmath1564 . because @xmath1552 , the point @xmath1558 in the last paragraph of the proof of theorem [ invariantthm ] can be chosen such that @xmath1565 , where @xmath500 and @xmath1566 denote the momentum maps of the hamiltonian @xmath24-actions on @xmath1 and @xmath1567 , respectively .
this implies that there is a @xmath24-equivariant symplectomorphism @xmath1568 from the delzant submanifold of @xmath1 through @xmath73 onto the delzant submanifold of @xmath1535 through @xmath1079 , which maps @xmath73 to @xmath1079 . using @xmath1568 in order to identify both delzant manifolds with @xmath1046 , under which identifications @xmath73 and @xmath1079 are mapped to the same point of @xmath28 ,
we conclude that @xmath1569 , if @xmath266 is the @xmath0-equivariant symplectomorphism from @xmath16 to @xmath1536 , described in the last paragraph of the proof of theorem [ invariantthm ] .
let @xmath1570 denote the automorphism group of @xmath16 , the set of all @xmath0-equivariant symplectomorphisms from @xmath16 to @xmath16 .
each @xmath1571 induces a transformation @xmath1572 of @xmath3 , which is an isomorphism of @xmath34-parallel spaces , and therefore of the form @xmath1573 for a unique element @xmath1574 .
the mapping @xmath1575 is a homomorphism from the group @xmath1570 to the torus @xmath467 . using the previous paragraph with @xmath1576 and @xmath1577 and using subsection [ holinvariantss ]
, it can be proved that @xmath1578 is equal to the set of @xmath1579 , for which there exists an @xmath1233 such that @xmath1580 for all @xmath999 , where it is sufficient to satisfy these equations for all @xmath743 in a @xmath13-basis of @xmath38 . using
this one can prove that @xmath1578 is a lie subgroup of @xmath467 with lie algebra equal to @xmath1259 , the space of all elements of @xmath39 which are equal to zero on the span of @xmath1258 .
actually , @xmath702 is an inifinitesimal symplectomorphism if and only if @xmath1581 . in general @xmath1578 need
not be a closed subgroup of the torus @xmath467 , and it neither needs to be connected , but it has countably many connected components .
the kernel of the homomorphism @xmath314 from @xmath1570 to @xmath467 consists of the group + @xmath1582 of all @xmath0-equivariant symplectomorphisms @xmath1583 which preserve all the @xmath0-orbits .
this group can be analyzed starting from remark [ hsrem ] .
[ autrem ] the following existence theorem completes the classification .
every list of ingredients as in definition [ ingredientdef ] is equal to the list of invariants of a compact connected symplectic manifold @xmath1 with effective symplectic @xmath0-action with coisotropic principal orbits as in theorem [ invariantthm ] .
[ existencethm ] a straightforward verification shows that , for any antisymmetric bilinear mapping @xmath725 , ( [ gotg ] ) turns @xmath981 into a two - step nilpotent lie algebra , and that ( [ hproduct ] ) defines a product in @xmath1584 for which @xmath19 is a lie group with @xmath977 as its lie algebra .
choose an element @xmath1221 such that @xmath1585 , see ( [ calt ] ) . because the @xmath1024 , @xmath999 , satisfy ( [ tauplushom ] )
, it follows that ( [ hdef ] ) defines a closed lie subgroup @xmath27 of @xmath19 , and that ( [ hact ] ) defines a smooth action of @xmath27 on the delzant manifold @xmath28 .
here we have used a choice of a complementary torus @xmath31 to @xmath24 , which will be kept fixed in the remainder of the proof . because @xmath27 is a closed lie subgroup of @xmath19 , its right action on @xmath19 is proper and free , and
therefore the action of @xmath27 on @xmath1055 , for which @xmath185 sends @xmath1056 to @xmath1057 , is proper and free .
the orbit space @xmath1586 has a unique structure of a smooth manifold such that the canonical projection @xmath1587 is principal @xmath27-bundle .
because @xmath19 and @xmath28 are connected , @xmath18 is connected as the image of the connected set @xmath1055 under the continuous mapping @xmath1588 . the projection @xmath1589 induces a @xmath19-equivariant smooth fibration @xmath1590 with fiber @xmath28 , the fiber bundle induced from the principal fiber bundle @xmath1591 by means of the action of @xmath27 on @xmath28
. see ( * ? ? ?
because @xmath38 is cocompact in @xmath39 , the base space @xmath21 is compact , and because the fiber @xmath28 is a compact delzant manifold , it follows that @xmath18 is compact .
on @xmath1055 we define the smooth two - form @xmath48 by ( [ a*sigma ] ) .
( note that we can not use the equation @xmath1592 here , because we do not have the symplectic form @xmath17 on the manifold @xmath18 yet , we are in the process of defining it . ) in ( [ a*sigma ] ) we have used a choice of a linear projection @xmath263 from @xmath12 onto @xmath40 , which will be kept fixed in the remainder of the proof .
we first verify that @xmath48 is closed .
the part @xmath1593 of ( [ a*sigma ] ) is closed , because it is defined by a constant two - form on @xmath993 . for the part @xmath1594 of ( [ a*sigma ] ) ,
it follows from ( [ domega ] ) that @xmath1595 is equal to the cyclic sum over @xmath1596 of @xmath1597 , which is equal to zero because of 5b ) in definition [ ingredientdef ] .
if @xmath1598 denotes the action of @xmath24 on @xmath28 , then the part @xmath1599 of ( [ a*sigma ] ) is equal to the pull - back of @xmath1600 by means of the mapping @xmath1601 this part of ( [ a*sigma ] ) is closed , because @xmath1602 .
the remaining part of ( [ a*sigma ] ) is @xmath1603 because the action of @xmath24 on @xmath28 is hamiltonian with momentum mapping @xmath500 , we have for every @xmath868 that @xmath1604 is equal to the derivative of @xmath1276 in the direction of @xmath1605 .
if we apply this to @xmath1606 , then we obtain that the remaining part of ( [ a*sigma ] ) is equal to @xmath1607 , in which the one - form @xmath1410 is defined by @xmath1608 because @xmath1609 , this completes the proof that @xmath1610 .
the element @xmath1611 sends @xmath1612 to @xmath1613 with @xmath1614 , @xmath1615 , and @xmath1616 .
therefore the tangent map of the action of @xmath1617 sends @xmath1618 to @xmath1619 with @xmath1620 , @xmath1621 , and @xmath1622 . because @xmath1623 , and because @xmath1624 is a symplectomorphism on @xmath28 which leaves @xmath500 and infinitesimal @xmath24-actions invariant , it follows that the two - form @xmath48 defined by ( [ a*sigma ] ) is @xmath27-invariant .
the condition that @xmath1625 for every @xmath1626 is equivalent to @xmath1627 ( take @xmath1628 , @xmath1629 , and let @xmath1135 range over @xmath1248 ) , @xmath1630 ( take @xmath1628 , @xmath1629 , and let @xmath1135 range over @xmath715 , where we use that we already have @xmath1627 ) , @xmath1631 ( take in the remaining equation ( [ a*sigma ] ) @xmath1628 , @xmath1632 and let @xmath1633 range over @xmath1634 ) , and finally @xmath1635 , because the fact that the @xmath24-orbits in @xmath28 are isotropic now implies that @xmath1636 for all @xmath1637 .
it follows that the kernel of @xmath1638 is equal to @xmath1639 .
the conclusion is that @xmath48 is a basic two - form for the action of @xmath27 on @xmath1055 , which implies that there is a unique smooth two - form @xmath17 on @xmath1640 such that @xmath1641 .
because @xmath1642 and at every point the tangent mapping of @xmath1588 is surjective , we have that @xmath50 .
furthermore @xmath17 is nondegenerate at every point , because the kernel of @xmath48 is equal to the kernel of the tangent mapping of @xmath1588 at every point .
therefore @xmath17 is a symplectic form on @xmath18 .
on @xmath1055 we have the action of @xmath1643 which sends @xmath1612 to @xmath1644 .
this action clearly leaves @xmath48 invariant , and it follows that the induced action of @xmath0 on @xmath1586 leaves @xmath17 invariant .
the tangent vectors to the orbits in @xmath1055 are the @xmath1645 , @xmath1646 , and if we substitute these as @xmath1504 in ( [ a*sigma ] ) then we obtain @xmath1647 requiring that this is equal to zero for all @xmath1646 is equivalent to @xmath1627 ( let @xmath1648 range over @xmath1248 ) , @xmath1630 ( let @xmath1648 range over @xmath715 and use that @xmath1627 ) and @xmath1605 is symplectically orthogonal to @xmath1649 .
if @xmath1650 , then the last condition implies that @xmath1651 for a unique @xmath868 .
this shows that the principal orbits of the @xmath0-action are coisotropic submanifolds of @xmath1 .
we now verify that the invariants of the compact connected symplectic manifold @xmath1 with symplectic @xmath0-action with coisotropic principal orbits are the ingredients in definition [ ingredientdef ] we started out with . if we substitute @xmath1652 and @xmath1653 in ( [ a*sigma ] ) , then we get @xmath1654 , which shows that the pull - back of @xmath17 to the @xmath0-orbits is given by @xmath32 . if @xmath1643 and @xmath1612 are such that @xmath1655 for some @xmath1611 , then @xmath1444 , @xmath1656 , and @xmath1657 .
because @xmath1658 implies that @xmath1659 , it follows that @xmath1660 and @xmath1661 .
this shows that @xmath1527 , the maximal stabilizer subgroup of the @xmath0-action on @xmath18 , see remark [ t1rem ] , is equal to @xmath24 .
the action of the subtorus @xmath24 of @xmath0 on @xmath1640 is induced by the action of @xmath24 on the second factor @xmath28 of @xmath1055 .
it follows that the action of @xmath24 on @xmath18 is hamiltonian with image of the momentum mapping equal to a translate of @xmath10 .
this proves that @xmath1528 is equal to a translate of @xmath10 , and therefore equal to @xmath10 if we add a suitable constant to the momentum mapping . because @xmath1662 , we have that @xmath1663 . for each @xmath1664 ,
the infinitesimal action of @xmath1665 on @xmath18 defines a smooth vector field @xmath702 on @xmath18 .
if the vector fields @xmath1666 on @xmath28 are lifts of @xmath1667 as in proposition [ liftprop ] with @xmath16 replaced by @xmath1046 , then the vector field @xmath1668 is intertwined by @xmath1588 with a unique vector field @xmath869 on @xmath18 , and the @xmath869 , @xmath702 together form a collection of lifts of @xmath926 , @xmath743 as in proposition [ liftprop ] , with @xmath726 replaced by @xmath1533 in proposition [ liftprop ] , iii ) .
it now follows from ( [ gotg ] ) that @xmath1669 . finally , if @xmath999 , then @xmath1670 , and therefore @xmath1671 which implies that @xmath1672 . according to theorem [ invariantthm ] , the symplectic manifold @xmath1 is unique up to @xmath0-equivariant symplectomorphisms .
in particular the dimension of @xmath18 is determined in terms of the ingredients in definition [ ingredientdef ] .
lemma [ coisotropicorbitlem ] implies that @xmath176 . in the language of categories , see maclane @xcite , theorem [ invariantthm ] and theorem [ existencethm ] can be summarized as follows .
let @xmath0 be a torus .
let @xmath1673 denote the category of which the objects are the compact connected symplectic manifolds @xmath1 together with an effective symplectic @xmath0-action on @xmath1 with coisotropic principal orbits , and the morphisms are the @xmath0-equivariant symplectomorphisms .
let @xmath1674 denote the set of all lists of invariants as in definition [ ingredientdef ] , viewed as a category with only the identities as morphisms .
then the assignment @xmath199 in definition [ invariantdef ] is an equivalence of categories from @xmath1673 onto @xmath1674 . in particular
it follows that the proper class @xmath1675 of isomorphism classes in @xmath1673 is a set , and the functor @xmath1676 induces a bijective mapping @xmath1677 from @xmath1675 onto @xmath1674 .
[ crowncor ] that @xmath1678 is a functor and @xmath1677 is injective follows from theorem [ invariantthm ] .
the surjectivity of @xmath199 , hence of @xmath1679 , follows from theorem [ existencethm ] .
let @xmath31 be a complementary torus to @xmath24 in @xmath0 .
if @xmath18 is @xmath0-equivariantly diffeomorphic to @xmath1170 , in which @xmath24 acts only on @xmath28 with isolated fixed points , and @xmath1155 acts freely on @xmath1155 , then @xmath1157 .
see proposition [ chernprop ] .
conversely , if this condition is satisfied , then lemma [ e0rem ] implies the stronger statement that @xmath16 is @xmath0-equivariantly symplectomorphic to the cartesian product of a symplectic manifold @xmath1147 with a free symplectic @xmath31-action and a delzant manifold @xmath1046 .
let @xmath726 in 5 ) in definition [ ingredientdef ] be such that @xmath1680 for some @xmath734 . if @xmath16 is as in theorem [ existencethm ] , then @xmath18 is not @xmath0-equivariantly diffeomorphic to @xmath1170 , in which @xmath31 acts freely on @xmath1155 and @xmath24 acts only on @xmath28 and with isolated fixed points .
therefore such @xmath16 are counterexamples to benoist ( * ? ? ?
6.6 ) , if in ( * ? ? ?
6.6 ) the word `` isomorphic '' implies `` equivariantly diffeomorphic ''
. there exists @xmath726 in 5 ) in definition [ ingredientdef ] , such that for every choice of a complementary torus @xmath31 to @xmath24 in @xmath0 we have @xmath1680 for some @xmath734 .
for instance , if @xmath1681 and @xmath1682 , then there exists a nonzero antisymmetric bilinear mapping @xmath726 from @xmath1683 to @xmath26 , which maps @xmath1684 into the integral lattice @xmath1346 in the lie algebra @xmath26 of @xmath24 .
such a @xmath726 satisfies 5a ) because every @xmath732 is a linear form on @xmath40 which vanishes on @xmath26 , and it satisfies 5b ) by assumption .
on the other hand @xmath1685 as soon as @xmath1686 .
[ productrem ] let @xmath1687 and @xmath295 , that is , the action of @xmath0 is free , @xmath297 , and the orbits are lagrange submanifolds of @xmath1 . in this case
the admissible connections are just the smooth @xmath0-invariant infinitesimal connections for the principal @xmath0-bundle @xmath1688 over the torus @xmath1689 , as in remark [ curvrem ] .
the first step in the proof of proposition [ liftprop ] consists of the construction of an infinitesimal connection for the principal @xmath0-bundle @xmath18 over the torus @xmath1689 , of which the curvature form is a constant two - form on the torus @xmath1689 . in this construction
the symplectic form did not enter , and the principal @xmath0-bundle @xmath18 over @xmath1689 can be constructed from the ingredients 4 ) , 5 ) in which condition 5a ) is kept , but condition 5b ) is dropped
. however , if one has a @xmath0-invariant symplectic form @xmath17 on @xmath18 for which the @xmath0-orbits are lagrangian , then ( [ ft ] ) , that is 5b ) , holds . in combination with theorem [ existencethm ] ,
we conclude that this principal @xmath0-bundle @xmath18 over @xmath1689 admits a @xmath0-invariant symplectic form for which the @xmath0-orbits are lagrangian , if and only if 5b ) holds .
this interpretation of condition 5b ) was suggested to us by yael karshon . if @xmath828 , then there exist antisymmetric bilinear mappings @xmath1690 for which 5b ) does not hold , and it follows that the principal @xmath0-bundle over @xmath1689 defined by @xmath163 does not admit a @xmath0-invariant symplectic form for which the @xmath0-orbits are lagrangian .
[ cocyclerem ] a slightly different approach to the classification would be to allow morphisms @xmath1691 , in which @xmath266 is a symplectomorphism from @xmath1692 onto @xmath1535 , @xmath199 is an isomorphism of lie groups from @xmath0 onto @xmath1693 , and @xmath266 intertwines the @xmath0-action on @xmath18 with the @xmath1693-action on @xmath1543 in the sense that @xmath1694 for every @xmath59 and @xmath169 .
the isomorphisms between tori are classified by the choices of @xmath13-bases in the integral lattices .
for instance if we fix a @xmath13-basis @xmath1695 , @xmath1696 , of @xmath1697 , then the mapping which assigns to a @xmath13-basis @xmath1698 , @xmath1699 , of @xmath14 the isomorphism @xmath1700 such that the tangent mapping of @xmath199 at the identity element maps @xmath1698 to @xmath1695 , is a bijective mapping from the set of @xmath13-bases of @xmath14 onto the set of isomorphisms from @xmath0 onto @xmath1693 . in turn
the set of @xmath13-bases in @xmath14 is in bijective correspondence with the group @xmath1701 of all @xmath1702-matrices with integral coefficients and determinant equal to @xmath201 .
this can be applied in particular to @xmath1703 , with @xmath1704 , @xmath1699 , equal to the standard basis of @xmath1705 .
if we also choose a @xmath13-basis @xmath1037 , @xmath1706 , in @xmath38 , then the ingredients 1 )
6 ) in definition [ ingredientdef ] are determined by their coefficients with respect to these bases .
furthermore the groups @xmath19 and @xmath27 are identified with @xmath1707 and @xmath1708 , respectively .
this would lead to a presentation of the model in coordinates , except for the delzant manifold @xmath1709 .
such a model looks even more explicit than the one in proposition [ aprop ] .
the disadvantage of this approach is that the invariants are given by coefficient matrices , which are determined uniquely only up to the action on these matrices of the changes of @xmath13-bases .
also the notations become quite a bit heavier if we write out our objects in coordinates .
[ baserem ]
in this section we define the notion of a @xmath4-parallel space , and prove that every complete , connected and locally convex @xmath4-parallel space is isomorphic to the cartesian product of a closed convex subset of a vector space and a torus .
let @xmath4 be an @xmath8-dimensional vector space .
@xmath4-parallel space _ is a hausdorff topological space @xmath461 , together with an open covering @xmath1710 , @xmath1711 , of @xmath461 and homeomorphisms @xmath1712 from @xmath1710 onto subsets @xmath1713 of @xmath4 such that , for every @xmath1714 for which @xmath1715 , the mapping @xmath1716 is locally constant .
a subset @xmath119 of @xmath461 is a @xmath4-parallel space with the @xmath1717 replaced by their restrictions to @xmath1718 .
if @xmath113 is another vector space and @xmath1493 is another @xmath113-parallel space , then @xmath1719 is a @xmath1720-parallel space in an obvious way .
the @xmath4-parallel space @xmath461 is called _ locally convex _ if the @xmath1713 are convex subsets of @xmath4 .
the locally convex @xmath4-parallel space @xmath461 is called _ locally convex polyhedral _
if for every @xmath1711 there is a convex open subset @xmath1721 of @xmath4 and there are finitely many linear forms @xmath1722 @xmath1723 , on @xmath4 , such that @xmath1724 [ cornerdef ] if we have ( [ cornereq ] ) with linearly independent linear forms @xmath1725 , and ( [ pareq ] ) is replaced to the weaker condition that for every @xmath1726 the mapping @xmath1727 is a diffeomorphism , then @xmath461 is a `` manifold with corners '' as defined for instance in mather @xcite .
if the @xmath1713 are open subsets of @xmath4 and ( [ pareq ] ) is relaxed to the condition that for every @xmath1714 the mapping @xmath1727 is locally an affine linear transformation , the composition of a linear mapping and a translation , then @xmath461 is a `` manifold with affine covering '' as in auslander and markus @xcite or a `` locally affine manifold '' as in auslander @xcite . if on the other hand we have ( [ pareq ] ) and ( [ cornereq ] ) with @xmath1728 for all @xmath1711 , then @xmath461 is called an `` affine space modelled over @xmath4 '' in geometry . in order to avoid confusion about the interpretation of the term `` affine ''
, we have replaced the term `` affine '' by `` parallel '' , where the latter word reminds of a parallelism , which is defined as a global frame in the tangent bundle , or equivalently a trivialization of the tangent bundle .
[ cornerdefrem ] let @xmath4 and @xmath113 be finite - dimensional vector spaces , @xmath461 and @xmath1493 a locally convex @xmath4-parallel and @xmath113-parallel spaces with charts @xmath1717 , @xmath1711 and @xmath1729 , @xmath1730 , respectively
. if @xmath862 is a linear mapping from @xmath113 to @xmath4 , then an _
@xmath862-map from @xmath1493 to @xmath461 _ is a continuous map @xmath1731 such that for each @xmath1730 and @xmath1711 and each @xmath119 of @xmath1732 such that @xmath1733 is convex , we have that @xmath1734 with such maps as morphisms , the locally convex parallel spaces form a category .
in particular two locally convex parallel spaces are called _ isomorphic _ if there exists an @xmath862-map from the first one to the second one , which is a homeomorphism and for which the linear mapping @xmath862 is bijective . [ lmapdef ] let @xmath461 be a @xmath4-parallel space . for each neighborhood @xmath1735 of @xmath404 in @xmath4 ,
let @xmath1736 be the set of all @xmath1737 such that there exists an @xmath1711 for which @xmath1738 and @xmath1739 .
the @xmath1736 define a _
uniform structure _ on @xmath461 in the sense of bourbaki ( * ? ? ?
ii , 1 , no .
1 ) . with respect to this uniform structure
, we have the notion of _ cauchy filters _ in @xmath461 as in bourbaki ( * ? ? ?
ii , 3 , no .
the @xmath4-parallel space is called _ complete _ if every cauchy filter in @xmath461 converges to an element of @xmath461 , cf .
bourbaki ( * ? ? ?
ii , 3 , no .
3 ) . [ uniformdef ]
our next goal is to prove that every complete , connected and locally convex @xmath4-parallel space is isomorphic to the product of a closed convex subset of a vector space and a torus .
see theorem [ cornerclassthm ] below for the precise statement .
let @xmath461 be a locally convex @xmath4-parallel space , as in definition [ cornerdef ] , and let @xmath485 .
a _ motion in @xmath461 with constant velocity @xmath46 _ is a continuous mapping @xmath1410 from an interval @xmath875 in @xmath604 to @xmath461 such that for every @xmath1711 and interval @xmath440 in @xmath1740 , we have @xmath1741 in other words , @xmath1742 is an @xmath862-map , where @xmath862 is the linear mapping from @xmath604 to @xmath4 which sends @xmath1743 to @xmath1744 . if @xmath1742 and @xmath1745 are motions in @xmath461 with constant velocity @xmath46 , and @xmath1746 for some @xmath1747 , then @xmath1748 for all @xmath1749 .
[ uniquemotionlem ] let @xmath1750 .
then @xmath259 is a closed subset of @xmath1751 , because @xmath1410 and @xmath1752 are continuous .
suppose @xmath1753 . because the @xmath1710 , @xmath1711 , form a covering of @xmath461 , there exists an @xmath1711 such that @xmath1754 . because @xmath1710 is open in @xmath461 and @xmath1410 and @xmath1752 are continuous , @xmath1755 is an open neighborhood of @xmath793 in @xmath1751 . for every @xmath1756
we have @xmath1757 hence @xmath1748 because @xmath1717 is injective .
therefore @xmath1758 , and we have proved that @xmath259 is also an open subset of @xmath1751 . because @xmath1751 is connected , it follows that @xmath1759 or @xmath1760 .
let @xmath1761 and @xmath1762 .
let @xmath1763 denote the set of all motions @xmath1410 in @xmath461 with constant velocity @xmath46 , which are defined on an interval @xmath1764 in @xmath604 such that @xmath1765 and @xmath1766 .
then it follows from lemma [ uniquemotionlem ] that the @xmath1767 have a common extension @xmath1768 to the union @xmath1769 of all the intervals @xmath1764 , @xmath1767 .
@xmath1769 is an interval in @xmath604 and @xmath1770 is a motion in @xmath461 with constant velocity @xmath46 , the unique maximal motion @xmath1742 with constant velocity @xmath46 such that @xmath1771 and @xmath1766 .
@xmath458 is the set of all @xmath1772 such that @xmath1773 .
we write @xmath1774 when @xmath1775 .
note that @xmath1776 when @xmath1762 and @xmath485 , whereas @xmath1777 when @xmath1778 .
for every @xmath1762 we write @xmath1779 .
[ ddef ] assume that the locally convex @xmath4-parallel space @xmath461 is complete , see definition [ cornerdef ] and definition [ uniformdef ] .
then , insofar as defined , the mapping @xmath1780 is an action of the additive group @xmath1781 on @xmath461 , in the following sense .
* for every @xmath1762 we have @xmath1782 and @xmath1783 .
* if @xmath1784 then @xmath1785 and @xmath1786 .
* if @xmath1787 and @xmath1788 , then @xmath1789 and @xmath1790 .
[ actionlem ] the statements i ) and ii ) follow immediately from the definitions .
our proof of iii ) is surprisingly long . in order to prove iii ) , assume that @xmath1787 and @xmath1791 .
let @xmath862 be the linear mapping from @xmath1196 to @xmath4 which sends @xmath1792 to @xmath1793 and @xmath1794 to @xmath1795 .
for any @xmath1796 , let @xmath1797 denote the convex hull of @xmath1798 , @xmath1799 , and @xmath1800 in @xmath1196 , which is a solid triangle .
let @xmath875 denote the set of all @xmath1801 $ ] for which there exists @xmath1802 and an @xmath862-map @xmath1803 such that @xmath1804 .
note that if @xmath1805 , then @xmath1806\subset i$ ] .
note also that for every @xmath1807 , @xmath1808\ni u\mapsto f(t_1+u\ , ( t_1'-t_1),\ , t_2+u\ , ( t_2'-t_2))\ ] ] is a motion in @xmath461 with constant velocity @xmath1809 , hence @xmath1810 and @xmath1811 if we apply this with @xmath1812 , then we see that @xmath51 is uniquely determined by the formula @xmath1813 where @xmath1814 for every for every @xmath1815 , which in turn implies that @xmath1816 where @xmath1817 , @xmath1818 and @xmath1819 all belong to @xmath458 for every @xmath1820 .
let @xmath1821 denote the infimum of @xmath875 , which implies that @xmath1822 i_1,\ , 1\right]\subset i$ ] .
we will show that this implies that @xmath1823 , which means that @xmath1824 $ ] . because trivially @xmath1825 , we may assume that @xmath1826 . because @xmath1827 , there exists an @xmath1711 such that @xmath1828 . because the mapping @xmath1829 is continuous from @xmath1830 to @xmath461
, there exists an @xmath1831 i_1,\ , 1\right]\subset i$ ] such that @xmath1832 .
because @xmath1805 , there exists @xmath1802 and an @xmath862-map @xmath1803 such that @xmath1804 .
note that for each @xmath1833 $ ] we have @xmath1834 . because @xmath51 is continuous
, there exists @xmath1835 such that @xmath1836 , if we write @xmath1837 .
because @xmath1838 , @xmath1839 , and @xmath1840 all belong to @xmath1710 , the points @xmath1841 , @xmath1842 , and @xmath1843 all belong to the convex subset @xmath1713 of @xmath4 , which implies that their convex hull @xmath1844 in @xmath4 is contained in @xmath1713 .
let @xmath1355 be the convex hull of @xmath1845 , @xmath1798 , and @xmath726 in @xmath1196 .
the @xmath862-map from @xmath1355 onto @xmath1844 which sends @xmath1798 to @xmath1846 , followed by @xmath1847 , defines an @xmath862-map @xmath1848 from @xmath1355 to @xmath461 such that @xmath1849 .
because of the uniqueness of @xmath862-maps which map @xmath1798 to @xmath1839 , see ( [ f+ ] ) , we have that @xmath1850 on @xmath1851 , and therefore @xmath1848 and @xmath51 have a common extension @xmath1852 . in order to prove that @xmath163 is an @xmath862-map , we observe that the property of being an @xmath862-map is local , and because @xmath1848 and @xmath51 are @xmath862-maps on the open subsets @xmath1853 and @xmath1854 of @xmath1355 and @xmath1797 , respectively , we have that @xmath163 is an @xmath862-map on @xmath1855 . on the other hand , if @xmath1856 , then there are neighborhoods @xmath1857 and @xmath1858 of @xmath1859 in @xmath1355 and @xmath1797 , respectively , such that @xmath1860 when @xmath1861 and @xmath1862 when @xmath1863 .
it follows that @xmath1864 which implies ( [ lmapeq ] ) with @xmath51 replaced by @xmath163 .
let @xmath1865 be the intersection point of the straight line through @xmath1799 and @xmath1800 , and the straight line through @xmath1845 and @xmath726 .
a straightforward calculation shows that @xmath1866 , with @xmath1867 because @xmath726 is lying on the straight line between @xmath1845 and @xmath1865 , it follows that the convex hull @xmath1868 of @xmath1845 , @xmath1799 and @xmath1865 is equal to to the union of the convex hull @xmath1355 of @xmath1845 , @xmath1798 , @xmath726 , and the convex hull @xmath137 of the points @xmath1845 , @xmath726 , @xmath1799 , and @xmath1865 .
because the latter four points all lie in @xmath1869 , it follows that @xmath1870 .
because the restriction of @xmath163 to @xmath1871 is an @xmath862-map such that @xmath1872 , hence @xmath1873 in view of ( [ f+ ] ) with @xmath51 replaced by @xmath163 , it follows that @xmath1823 .
now suppose that @xmath1874 . with @xmath1711 such that @xmath1828 , there exists @xmath1875 such that @xmath1832 .
the same reasoning as above with @xmath1821 and @xmath1876 interchanged , where we use that @xmath1823 , leads to the conclusion that @xmath1805 , in contradiction with the definition @xmath1877 of @xmath1821 .
we conclude that @xmath1878 , or @xmath1879 , which means that there exists @xmath1802 and an @xmath862-map @xmath51 from @xmath1880 to @xmath461 such that @xmath1881 .
define @xmath440 as the set of all @xmath1882 $ ] for which there exists an @xmath862-map @xmath51 from @xmath1880 to @xmath461 such that @xmath1881 , and write @xmath1883 .
the uniqueness of the @xmath862-maps @xmath51 from @xmath1880 to @xmath461 such that @xmath1881 , see ( [ f+ ] ) , where @xmath1884 ranges over @xmath440 , implies that these @xmath862-maps have a common extension to the union @xmath85 over all @xmath1885 of the triangles @xmath1880 , which we also denote by @xmath51 .
note that the closure of @xmath85 in @xmath1196 is equal to @xmath1886 .
@xmath1887 is an @xmath862-map , hence uniformly continuous .
because a uniformly continuous mapping from a dense subset @xmath385 of a complete space @xmath42 to a complete space @xmath1888 has a unique extension to a continuous mapping from @xmath42 to @xmath1888 , see bourbaki ( * ? ? ?
* ch . 2 ,
3 , no . 6 , th .
2 ) , and because we had assumed that @xmath461 is complete , it follows that @xmath51 has a unique extension to a continuous mapping @xmath1889 . because the restriction of the continuous mapping @xmath1889 to the dense subset @xmath85 of @xmath1890 is an @xmath862-map
, it follows that @xmath1891 is an @xmath862-map , and we conclude that @xmath1892 , or equivalently @xmath1893 $ ] .
if @xmath1894 then the previous argument leading to @xmath1878 , with @xmath1793 and @xmath1795 replaced by @xmath1895 and @xmath1896 , respectively , shows that there exists @xmath1897 and an @xmath862-map @xmath1898 from the convex hull @xmath1899 of @xmath1900 , @xmath1901 , and @xmath1902 to @xmath461 , such that @xmath1903 .
the uniqueness of @xmath862-maps which send @xmath1900 to @xmath141 , see ( [ f+ ] ) , yields that @xmath1904 on @xmath1905 , which implies that @xmath51 and @xmath1898 have a common extension @xmath1906 to @xmath1907 . as in the argument leading to @xmath1878 , we have that @xmath1906 is an @xmath862-map , and because @xmath1908 it follows that @xmath1909 , in contradiction with the definition @xmath1910 of @xmath793
. we arrive at @xmath1911 , or equivalently @xmath1912 , which means that there is an @xmath862-map @xmath1913 such that @xmath1881 . now ( [ + ] ) with @xmath1914 , @xmath1915 and @xmath1916 implies that @xmath1917 and @xmath1918 .
this completes the proof of the statement iii ) in the lemma .
let @xmath461 be a complete locally convex @xmath4-parallel space .
then @xmath458 is a closed subset of @xmath1919 and @xmath1920 is a uniformly continuous @xmath862-map from @xmath458 to @xmath461 . here
@xmath1921 is the mapping which sends @xmath1922 to @xmath1777 . for each @xmath1762 ,
the mapping @xmath1923 is a local homeomorphism from @xmath1924 onto an open subset of @xmath461 .
[ contactionlem ] let @xmath1762 , and let @xmath1924 be defined as in definition [ ddef ] . if @xmath1711 and @xmath1710 is a neighborhood of @xmath141 in @xmath461 , then the translate @xmath1925 of @xmath1713 over the vector @xmath1926 is contained in @xmath1924 .
indeed , if @xmath485 , @xmath1927 , then the convexity of @xmath1713 implies that @xmath1928 for every @xmath1929 $ ] .
the continuity of @xmath1930 implies that @xmath1931 is a motion in @xmath461 which has constant velocity @xmath46 and is equal to @xmath141 at @xmath1932 , which shows that @xmath1933 .
note that in passing we have proved that @xmath1934 for all @xmath1001 , and in particular @xmath1935 we next claim that @xmath1925 is a neighborhood of @xmath404 in @xmath1924 . if this would not be the case
, then there is a sequence @xmath925 in @xmath1936 which converges to @xmath404 in @xmath4 as @xmath1937 .
because @xmath461 is a hausdorff space , there exists an open neighborhood @xmath1938 of @xmath141 in @xmath461 such that the closure @xmath1939 of @xmath1938 in @xmath461 is contained in @xmath1710 .
let @xmath1940 be the set of all @xmath1941 such that @xmath1942 and @xmath1943 , and write @xmath1944 .
note that @xmath1945 . because @xmath1946 is continuous from @xmath1947 to @xmath461 , and @xmath1948 is closed in @xmath461
, we have that @xmath1940 is a closed subset of @xmath1949 , hence @xmath1950 , which implies that @xmath1951 . on the other hand we have for every @xmath1952 that @xmath1953 , hence @xmath1954 . because @xmath1955 , this can not happen for @xmath579 , which proves that @xmath1956 . because @xmath1957 is continuous and @xmath1958 for every @xmath1959
, we have that @xmath1960 , hence @xmath1961 in @xmath4 , hence in @xmath1713 . because @xmath1962 is continuous ,
we conclude that @xmath1963 in @xmath1710 , hence in @xmath461 , in contradiction with the fact that @xmath1951 for every @xmath391 . because @xmath1935 for every @xmath1964 , the restriction to the neighborhood @xmath1925 of @xmath404 in @xmath1924 of the mapping @xmath1923 is equal to the continuous mapping @xmath1965 , from @xmath1925 onto the neighborhood @xmath1710 of @xmath141 in @xmath461 , with inverse equal to the continuous mapping @xmath1966 from @xmath1710 onto @xmath1925 . in particular we have that for every @xmath1967 the mapping @xmath1968 is a homeomorphism from a neighborhood of @xmath404 in @xmath1969 onto a neighborhood of @xmath1292 in @xmath461 .
now let @xmath1933 . if @xmath1970 such that @xmath1971 , then @xmath1972 with @xmath1973 hence @xmath1974 and @xmath1975 in view of iii ) in lemma [ actionlem ] .
furthermore ii ) in lemma [ actionlem ] implies that @xmath1976 and @xmath1977 .
we now have @xmath1976 and @xmath1978 , which in view of iii ) in lemma [ actionlem ] implies that @xmath1979 and @xmath1980 . combining the equations we arrive at @xmath1981
because @xmath1982 is continuous from a neighborhood of @xmath1983 in @xmath458 to @xmath4 , with value @xmath404 at @xmath1984 , and @xmath1968 is continuous from a neighborhood of @xmath404 in @xmath1985 to @xmath461 , with @xmath1986 , we conclude that @xmath1987 is continuous from a neighborhood of @xmath1983 in @xmath458 to @xmath461 .
because this holds for every @xmath1933 , we have proved that the mapping @xmath1920 is continuous from @xmath458 to @xmath461 , and therefore it is an @xmath862-map from @xmath458 to @xmath461 .
note also that ( [ qw ] ) with @xmath1988 implies that @xmath1923 is a local homeomorphism from @xmath1924 onto an open subset of @xmath461 . as an @xmath862-mapping ,
the mapping @xmath1989 is uniformly continuous from @xmath1990\times d$ ] to @xmath461 .
if @xmath1991 denotes the closure of @xmath458 in the complete space @xmath1919 , then @xmath1990\times d$ ] is dense in the complete space @xmath1990\times\overline{d}$ ] .
because also @xmath461 is complete , the mapping @xmath51 has a unique extension to a continuous mapping @xmath163 from @xmath1990\times \overline{d}$ ] to @xmath461 .
see bourbaki ( * ? ? ?
* ch . 2 ,
3 , no . 6 , th .
2 ) . by continuity
it follows that for each @xmath1992 the mapping @xmath1990\ni
t\mapsto g(t,\ , v,\ , p)$ ] is a motion in @xmath461 with constant velocity @xmath46 , equal to @xmath141 at @xmath1932 , and the conclusion is that @xmath1933 .
this shows that @xmath1993 , that is , @xmath458 is closed in @xmath1919 .
let @xmath461 be a complete locally convex @xmath4-parallel space which in addition is connected .
then the action of @xmath4 on @xmath461 is transitive , in the sense that for any @xmath1994 there exists @xmath485 such that @xmath1784 and @xmath1995 .
this implies that for every @xmath1762 the mapping @xmath1996 is a local homeomorphism from @xmath1924 , which is a closed subset of @xmath4 , onto @xmath461 .
[ surjlem ] we write @xmath1997 if there exists a @xmath485 such that @xmath1784 and @xmath1998 .
it follows from lemma [ actionlem ] that @xmath1999 is an equivalence relation in @xmath461 .
it follows from the last statement in lemma [ contactionlem ] that the equivalence classes are open subsets of @xmath461 .
because @xmath461 is connected , the equivalence relation @xmath1999 has only one equivalence class @xmath461 , which proves the transitivity of the action .
@xmath1924 is a closed subset of @xmath4 , as the preimage of the closed subset @xmath458 of @xmath1919 under the continuous mapping @xmath2000 .
let @xmath461 be a complete and connected locally convex @xmath4-parallel space . for any @xmath485 ,
the following conditions are equivalent .
* there exists a @xmath1762 such that @xmath2001 . * for every @xmath1762 we have @xmath2001 .
let @xmath39 denote the set of all @xmath485 such that a ) or b ) holds .
then @xmath39 is a linear subspace of @xmath4 , @xmath2002 , and the restriction to @xmath2003 of the mapping @xmath1920 defines an action of the additive group @xmath39 on @xmath461 .
[ nlem ] assume that a ) holds , which implies that @xmath2004 for every @xmath970 .
let @xmath2005 .
because of the transitivity in lemma [ surjlem ] , there exists a @xmath2006 such that @xmath1970 and @xmath2007 .
because of iii ) in lemma [ actionlem ] we have for every @xmath970 that @xmath2008 , which in view of definition [ ddef ] , the definition of @xmath458 , implies that @xmath2009 for every @xmath2010 .
if for any @xmath970 and @xmath2011 we replace @xmath578 by @xmath2012 and take the limit for @xmath2013 , then we obtain in view of the closedness of @xmath458 that @xmath2014 , which implies that @xmath2015 .
because this holds for every @xmath970 , it follows that @xmath2016 . if @xmath2017 then we have for every @xmath2018 that @xmath2019 .
moreover , we have for every @xmath1762 that @xmath2020 , @xmath2021 , hence @xmath2022 in view of iii ) in lemma [ actionlem ] .
if for any @xmath970 we replace @xmath1859 and @xmath793 by @xmath2023 and @xmath2024 , respectively , we obtain that @xmath2025 , hence @xmath2026 , and we conclude that @xmath39 is a linear subspace of @xmath4 .
furthermore b ) implies that @xmath1933 for every @xmath2027 and @xmath1762 , and it follows from lemma [ actionlem ] that the mapping @xmath1920 defines an action of @xmath39 on @xmath461 .
let @xmath461 be a complete and connected locally convex @xmath4-parallel space .
for any @xmath485 , the following conditions are equivalent .
* there exists @xmath1762 such that @xmath1933 and @xmath2028 .
* for all @xmath1762 we have @xmath1933 and @xmath2029 .
let @xmath38 denote the set of all @xmath485 such that i ) or ii ) holds .
then @xmath38 is a discrete additive subgroup of the linear subspace @xmath39 of @xmath4 which is defined in lemma [ nlem ] .
[ perlem ] assume that i ) holds .
lemma [ surjlem ] implies that for any @xmath2005 there exists @xmath2030 such that @xmath1972 , and therefore @xmath2031 which proves ii ) .
if @xmath2028 , then it follows by induction on @xmath102 that @xmath2032 and @xmath2033 for every positive integer @xmath102 , and using i ) , ii ) in lemma [ actionlem ] we obtain the same conclusions for all @xmath2034 .
this implies that @xmath2035 , which in turn implies that @xmath2036 , because @xmath2037 is an interval in @xmath604 . in view of lemma [ nlem ]
, we conclude that @xmath2027 .
it follows that @xmath38 is equal to the set of @xmath2027 such that the action of @xmath46 on @xmath461 is trivial , and therefore @xmath38 is an additive subgroup of @xmath39 .
the last statement in lemma [ surjlem ] implies that @xmath38 is a discrete subset of @xmath1924 , hence of the closed subset @xmath39 of @xmath1924 .
let @xmath461 be a complete and connected locally convex @xmath4-parallel space .
let @xmath38 be the period group in @xmath4 defined in lemma [ perlem ] .
denote by @xmath2038 the @xmath604-linear span of @xmath38 in @xmath4 , where we note that @xmath2039 is a torus .
let @xmath85 be a linear complement in @xmath4 of the linear span @xmath2038 of @xmath38 in @xmath4 .
let @xmath1762 .
then there is a convex closed subset @xmath1531 of @xmath85 such that @xmath2040 .
furthermore the mapping @xmath2041 defines an isomorphism of @xmath4-parallel spaces from @xmath2042 onto @xmath461 .
the projection from @xmath461 onto the @xmath2038-orbit space @xmath2043 induces an isomorphism from @xmath1531 , viewed as a @xmath85-parallel space , onto the @xmath2044-parallel space @xmath2043 .
there is a collection of linear forms @xmath2045 on @xmath4 and real numbers @xmath2046 , where @xmath2047 runs over some index set @xmath875 , such that @xmath1924 is equal to the set of all @xmath485 such that @xmath2048 for all @xmath2049 . for every such collection @xmath2050 , @xmath2046 ,
the linear subspace @xmath39 of @xmath4 defined in lemma [ nlem ] is equal to the common kernel of the linear forms @xmath2045 , @xmath2049 , on @xmath4 .
@xmath461 is compact if and only if @xmath1531 is compact , which implies that @xmath2051 , and @xmath38 is a cocompact discrete subgroup of the additive group @xmath39 . if @xmath461 is a compact connected locally convex polyhedral @xmath4-parallel space , then @xmath2052 is a convex polytope in @xmath2053 .
[ cornerclassthm ] let @xmath1933 , @xmath2054 , and @xmath2055 .
it follows from lemma [ actionlem ] that @xmath2056 , @xmath2057 and @xmath2058 , which in view of lemma [ perlem ] implies that @xmath2059 .
define @xmath2060 .
if @xmath2061 , then it follows from the transitivity in lemma [ surjlem ] that there exists a @xmath2062 such that @xmath2063 and @xmath2064 , which implies that @xmath2065 .
replacing @xmath289 by @xmath2066 we therefore can arrange that @xmath2067 .
now @xmath2068 implies that for every @xmath1929 $ ] we have @xmath2069 , which in combination with iii ) in lemma [ actionlem ] implies that @xmath2070 , and therefore @xmath2071 .
this proves that @xmath1531 is convex .
it is a closed subset of @xmath85 because @xmath1924 is a closed subset of @xmath4 , see lemma [ surjlem ] .
the fact that the additive group @xmath39 acts on @xmath461 , cf .
lemma [ perlem ] , and hence its subgroup @xmath2038 acts on @xmath461 , implies that @xmath2072 . because @xmath2073 and @xmath2055 imply that @xmath2074 , and @xmath85 is complementary to @xmath2038 , the mapping @xmath2075 is injective . on the other hand lemma [ surjlem ] implies that @xmath2076 is a surjective local homeomorphism .
because @xmath484 is an @xmath862-map with @xmath2077 , it follows that @xmath484 is an isomorphism from @xmath2042 onto @xmath461 .
the statement about @xmath39 , the @xmath2045 and the @xmath2046 follows because @xmath2078 is a closed convex subset of @xmath4 , and @xmath39 is equal to the lineality of @xmath1924 , the set of direction vectors of lines which are contained in @xmath1924 , cf .
rockafellar @xcite .
finally , a compact hausdorff topological space has a unique uniform structure which is compatible with its topology , and the compact space is complete with respect to this uniform structure , cf .
bourbaki ( * ? ? ?
* ch . 2 ,
clearly @xmath461 is compact if and only if @xmath2042 is compact if and only if @xmath1531 is compact . in view of @xmath2079
, the latter implies that @xmath2080 , hence @xmath2051 , and @xmath38 is a cocompact discrete subgroup of the additive topological group @xmath39 . a convex compact subset of a vector space @xmath85 which is a @xmath85-parallel manifold with corners is a convex polytope in @xmath85 .
theorem [ cornerclassthm ] is a generalization of the theorem of tietze @xcite and nakajima @xcite that any closed , connected , and locally convex subset of a finite - dimensional vector space is convex .
our proof of lemma [ actionlem ] is close to the proof of klee ( * ? ? ? * ( 5.2 ) ) of the generalization of the tietze - nakajima theorem to subsets of arbitrary topological vector spaces .
in this section we describe the local model of benoist ( * ? ? ? * prop .
1.9 ) and ortega and ratiu @xcite for a general proper symplectic lie group action .
see also ortega and ratiu ( * ? ? ?
7.27.4 ) for a detailed proof . for hamiltonian actions ,
such local models had been obtained before by marle @xcite and guillemin and sternberg ( * ? ? ?
let @xmath1 be a smooth symplectic manifold and @xmath19 a lie group which acts smoothly on @xmath1 by means of symplectomorphisms .
furthermore assume that the action is _ proper _ , which means that for any compact subset @xmath259 of @xmath18 the set of all @xmath2081 such that @xmath2082 is compact in @xmath2083 .
the action of @xmath19 is certainly proper if @xmath19 is compact . for every @xmath2084 we will write @xmath2085 for the action of @xmath163 on @xmath18 . for every element @xmath42 in the lie algebra @xmath977 of @xmath19 , the infinitesimal action on @xmath18
will be denoted by @xmath43 .
it is a smooth vector field @xmath43 on @xmath18 , the flow of which leaves @xmath17 invariant .
it follows from the properness of the action , that for every @xmath59 the stabilizer subgroup @xmath2086 of @xmath73 in @xmath19 is a compact , lie subgroup of @xmath19 , and the mapping @xmath2087 induces a @xmath19-equivariant smooth embedding @xmath2088 from @xmath21 into @xmath18 , with closed image , equal to the orbit @xmath2089 of @xmath19 through the point @xmath73 . here
@xmath2084 acts on @xmath21 by sending @xmath2090 to @xmath2091 .
the lie algebra @xmath2092 of @xmath2093 is equal to the set of @xmath2094 such that @xmath76 .
the linear mapping @xmath2095 induces induces a linear isomorphism from @xmath2096 onto @xmath2097 . for the description of the symplectic form in the local model , we begin with the closed two - form @xmath2098 on @xmath21 , which represents the `` restriction '' of @xmath17 to the orbit @xmath2099 through the point @xmath73 . here
@xmath2100 is defined in ( [ axdef ] ) .
the @xmath19-invariance of @xmath17 and the @xmath19-equivariance of @xmath79 imply that @xmath124 is @xmath19-invariant .
if we identify @xmath2101 with @xmath2102 , and @xmath2103 denotes the projection @xmath2104 , then @xmath2105 , and therefore the @xmath19-invariant two - form @xmath124 on @xmath21 , is determined by the antisymmetric bilinear form @xmath2106 on @xmath977 , which is invariant under the adjoint action of @xmath27 on @xmath977 .
it follows that the kernel @xmath2107 is an @xmath2108-invariant linear subspace of @xmath977 .
we have @xmath367 , because of ( [ sigmagdef ] ) and the fact that @xmath2109 vanishes on the kernel @xmath358 of @xmath141 .
if @xmath862 is a linear subspace of a symplectic vector space @xmath2110 , then the symplectic orthogonal complement @xmath2111 of @xmath862 in @xmath4 is defined as the set of all @xmath485 such that @xmath2112 for every @xmath2113 .
the restriction to @xmath2114 of @xmath949 defines a symplectic form @xmath182 on the vector space @xmath2115 and the mapping @xmath2116 defines a linear isomorphism from @xmath110 onto @xmath2117 .
the linearized action @xmath2118 of @xmath27 on @xmath78 is symplectic and leaves @xmath2119 invariant , acting on it via the adjoint representation . that is ,
@xmath2120 , if @xmath2094 and @xmath185 .
it therefore also leaves @xmath2114 invariant and induces an action of @xmath2121 on the symplectic vector space @xmath180 by means of symplectic linear transformations . for any action by linear transformations of a compact lie group @xmath259 on a vector space @xmath4
, any @xmath259-invariant linear subspace @xmath862 of @xmath4 has an @xmath259-invariant linear complement @xmath2125 in @xmath4 .
for instance , if @xmath770 is an inner product on @xmath4 , then the average @xmath2126 of @xmath770 over @xmath259 is a @xmath259-invariant inner product on @xmath4 , and the @xmath2126-orthogonal complement @xmath2125 of @xmath862 in @xmath4 has the desired properties .
choose @xmath2108-invariant linear complements @xmath275 and @xmath2127 of @xmath358 and @xmath40 in @xmath977 , respectively .
let @xmath2128 and @xmath2129 denote the linear projection from @xmath977 onto @xmath40 and @xmath358 with kernel equal to @xmath2127 and @xmath275 , respectively
. then these linear projections are @xmath2108-equivariant .
if @xmath2084 , then we denote by @xmath2130 the multiplication from the left by means of @xmath163 .
define the smooth one - form @xmath2131 on @xmath2132 by @xmath2133 for all @xmath2084 , @xmath2134 , @xmath2135 , and all @xmath2094 , @xmath2136 , @xmath2137 . here
we identify the tangent spaces of a vector space with the vector space itself and @xmath167 with the space of linear forms on @xmath40 which vanish on @xmath358 .
let @xmath138 be defined as in ( [ edef ] ) , with @xmath40 and @xmath113 as in ( [ ldef ] ) and ( [ wdef ] ) , respectively .
let @xmath2138 denote the orbit space of @xmath2132 for the proper and free action of @xmath27 on @xmath2132 , where @xmath185 acts on @xmath2132 by sending @xmath2139 to @xmath2140 .
the action of @xmath19 on @xmath2138 is induced by the action @xmath2141 of @xmath19 on @xmath2132 .
let @xmath2142 denote the mapping which is induced by the projection @xmath2143 onto the first component .
because @xmath27 acts on @xmath138 by means of linear transformations , this projection exhibits @xmath2138 as a @xmath19-homogeneous vector bundle over the homogeneous space @xmath21 , which fiber @xmath138 and structure group @xmath27 . with these notations , we have the following local normal form for the symplectic @xmath19-space @xmath2144 .
furthermore , there exists an open @xmath27-invariant neighborhood @xmath264 of the origin in @xmath138 and a @xmath19-equivariant diffeomorphism @xmath266 from @xmath2148 onto an open @xmath19-invariant neighborhood @xmath119 of @xmath73 in @xmath18 , such that @xmath2149 and @xmath2150 here the mapping @xmath2151 is induced by the projection onto the first component , and @xmath124 is defined in _ ( [ sigmagdef])_. [ gthm ] j.j .
duistermaat and g.j .
heckman : on the variation in the cohomology of the symplectic form of the reduced phase space .
_ * 69 * ( 1982 ) 259268 .
addendum in : _ invent .
_ * 72 * ( 1983 ) 153158 .
marle : classification des actions hamiltoniennes au voisinage dune orbite . _ c. r. acad .
paris sr .
_ * 299 * ( 1984 ) 249252 .
modle daction hamiltonienne dun groupe de lie sur une varit symplectique .
sem . mat .
* 43 * ( 1985 ) 227251 .
m. symington : four dimensions from two in symplectic topology .
153208 in _ topology and geometry of manifolds ( athens , ga , 2001)_. proc .
pure math .
, 71 , amer .
soc . , providence , ri , 2003 .
h. tietze : ber konvexheit i m kleinen und i m groen und ber gewisse den punkten einer menge zugeordete dimensionszahlen . _ math
. z. _ * 28 * ( 1928 ) 697707 .
h. whitney : differentiable even functions .
_ duke math .
j. _ * 10 * ( 1943 ) . | in this paper we completely classify symplectic actions of a torus @xmath0 on a compact connected symplectic manifold @xmath1 when some , hence every , principal orbit is a coisotropic submanifold of @xmath2 . that is
, we construct an explicit model , defined in terms of certain invariants , of the manifold , the torus action and the symplectic form .
the invariants are invariants of the topology of the manifold , of the torus action , or of the symplectic form . in order to deal with symplectic actions which are not hamiltonian
, we develop new techniques , extending the theory of atiyah , guillemin - sternberg , delzant , and benoist .
more specifically , we prove that there is a well - defined notion of constant vector fields on the orbit space @xmath3 . using a generalization of the tietze - nakajima theorem to what we call @xmath4-parallel spaces ,
we obtain that @xmath3 is isomorphic to the cartesian product of a delzant polytope with a torus .
we then construct special lifts of the constant vector fields on @xmath3 , in terms of which the model of the symplectic manifold with the torus action is defined . |
elastic objects pinned by quenched disorder are central to the physics of disordered systems . in the last decades
a considerable amount of research has been devoted to them . from the theory side
they are among the simplest , but still quite non - trivial , models of glasses with complex energy landscape and many metastable states .
they are related to a remarkably broad set of problems , from subsequences of random permutations in mathematics @xcite , random matrices @xcite to growth models @xcite and burgers turbulence in physics @xcite , as well as directed polymers @xcite and optimization problems such as sequence alignment in biology @xcite .
foremost , they are very useful models for numerous experimental systems , each with its specific features in a variety of situations . interfaces in magnets @xcite experience either short - range disorder ( random bond rb ) , or long range ( random field rf ) . charge density waves ( cdw ) @xcite or the bragg glass in superconductors @xcite are periodic objects pinned by disorder . the contact line of liquid helium meniscus on a rough substrate is governed by long range elasticity @xcite .
all these systems can be parameterized by a @xmath6-component height or displacement field @xmath7 , where @xmath8 denotes the @xmath9-dimensional internal coordinate of the elastic object ( we will use @xmath10 to denote fourier components ) .
an interface in the 3d random field ising model has @xmath11 , @xmath12 , a vortex lattice @xmath13 , @xmath14 , a contact - line @xmath15 and @xmath12 .
the so - called directed polymer ( @xmath15 ) has been much studied @xcite as it maps onto the kardar - parisi - zhang growth model @xcite for any @xmath6 .
the equilibrium problem is defined by the partition function @xmath16\ , \exp(-{\cal h}[u]/t)$ ] associated to the hamiltonian @xmath17= \int { { \mathrm{d}}}^d x\ , \frac{1}{2 } ( \nabla u)^2 + v(u_x , x ) \ , \ ] ] which is the sum of an elastic energy which tends to suppress fluctuations away from the perfectly ordered state @xmath18 , and a random potential which enhances them .
the resulting roughness exponent @xmath19 @xmath20 is measured in experiments for systems at equilibrium ( @xmath21 ) or driven by a force @xmath22 . here and below @xmath23
denote thermal averages and @xmath24 disorder ones . in some cases
, long - range elasticity appears e.g. for the contact line by integrating out the bulk - degrees of freedom @xcite , corresponding to @xmath25 in the elastic energy .
as will become clear later , the random potential can without loss of generality be chosen gaussian with second cumulant @xmath26 with various forms : periodic systems are described by a periodic function @xmath27 , random bond disorder by a short - range function and random field disorder of variance @xmath28 by @xmath29 at large @xmath30 . although this paper is devoted to equilibrium statics , some comparison with dynamics will be made and it is thus useful to indicate the equation of motion @xmath31 with friction @xmath32 .
the pinning force is @xmath33 of correlator @xmath34 in the bare model . despite some significant progress ,
the model ( [ ham ] ) has mostly resisted analytical treatment , and one often has to rely on numerics . apart from the case of the directed polymer in @xmath35 dimensions ( @xmath15 , @xmath12 ) , where a set of exact and rigorous results was obtained @xcite , analytical methods are scarce .
two main analytical methods exist at present , both interesting but also with severe limitations .
the first one is the replica gaussian variational method ( gvm ) @xcite .
it is a mean field method , which can be justified for @xmath36 and relies on spontaneous replica symmetry breaking ( rsb ) @xcite .
although useful as an approximation , its validity at finite @xmath6 remains unclear .
indeed , it seems now generally accepted that rsb does not occur for low @xmath9 and @xmath6 .
the remaining so - called weak rsb in excitations @xcite may not be different from a more conventional droplet picture .
another exactly solvable mean field limit is the directed polymer on the cayley tree , which also mimics @xmath37 and there too it is not fully clear how to meaningfully expand around that limit @xcite .
the second main analytical method is the functional renormalization group ( frg ) which attempts a dimensional expansion around @xmath38 @xcite .
the hope there is to include fluctuations , neglected in the mean field approaches .
however , until now this method has only been developed to one loop , for good reasons , as we discuss below . its consistency has never been checked or tested in any calculation beyond one loop ( i.e. lowest order in @xmath2 ) .
thus contrarily to pure interacting elastic systems ( such as e.g. polymers ) _ there is at present no quantitative method , such as a renormalizable field theory , which would allow to compute accurately all universal observables in these systems_. the central reason for these difficulties is the existence of many metastable states ( i.e. local extrema ) in these systems . although qualitative arguments show that they arise beyond the larkin length @xcite , these are hard to capture by conventional field theory methods .
the best illustration of that is the so called dimensional reduction ( dr ) phenomenon , which renders naive perturbation theory useless @xcite in pinned elastic systems as well as in a wider class of disordered models ( e.g. random field spin models ) .
indeed it is shown that to _ any _ order in the disorder at zero temperature @xmath39 , any physical observable is found to be _ identical _ to its ( trivial ) average in a gaussian random force ( larkin ) model , e.g.@xmath40 for rb disorder .
thus perturbation theory appears ( naively ) unable to help in situations where there are many metastable states .
the two above mentioned methods ( gvm and frg ) are presently the only known ways to _ escape dimensional reduction _ and to obtain non - trivial values for @xmath19 ( in two different limits but consistent when they can be compared @xcite ) .
the mean field method accounts for metastable states by rsb .
this however may go further than needed since it implies a large number of pure states ( i.e. low ( free ) energy states differing by @xmath41 in ( free ) energy ) .
the other method , the frg , captures metastability through a non - analytic action with a cusp singularity . both the rsb and the cusp arise dynamically , i.e. spontaneously , in the limits studied .
the 1-loop frg has had some success in describing pinned systems .
it was noted by fisher @xcite within a wilson scheme analysis of the interface problem in @xmath42 that the coarse grained disorder correlator becomes _ non - analytic _ beyond the larkin scale @xmath43 , yielding large scale results distinct from naive perturbation theory . within this approach
an infinite set of operators becomes relevant in @xmath44 , parameterized by the second cumulant @xmath27 of the random potential .
explicit solution of the 1-loop frg for @xmath27 gives several non - trivial attractive fixed points ( fp ) to @xmath45 proposed in @xcite to describe rb , rf disorder and in @xcite , periodic systems such as cdw or vortex lattices .
all these fixed points exhibit a `` cusp '' singularity as @xmath46 at small @xmath47 .
the cusp was interpreted in terms of shocks in the renormalized force @xcite , familiar from the study of burgers turbulence ( for @xmath15 , @xmath12 ) .
the dynamical frg was also developed to one loop @xcite to describe the depinning transition .
the mere existence of a non - zero critical threshold force @xmath48 ) .
extension to non - zero temperature @xmath49 suggested that the cusp is rounded within a thermal boundary layer @xmath50 .
this was interpreted to describe thermal activation and leads to a reasonable derivation of the celebrated creep law for activated motion @xcite . in standard critical phenomena a successful 1-loop calculation
usually quickly opens the way for higher loop computations , allowing for accurate calculation of universal observables and comparison with simulations and experiments , and eventually a proof of renormalizability . in the present context however , no such work has appeared in the last fifteen years since the initial proposal of @xcite , a striking sign of the high difficulties which remain .
only recently a 2-loop calculation was performed @xcite but since this study is confined to an analytic @xmath27 it only applies below the larkin length and does not consistently address the true large scale critical behavior .
in fact doubts were even raised @xcite about the validity of the @xmath5-expansion beyond order @xmath5 .
it is thus crucial to construct a renormalizable field theory , which describes statics and depinning of disordered elastic systems , and which allows for a systematic expansion in @xmath2 .
as long as this is not achieved , the physical meaning and validity of the 1-loop approximation does not stand on solid ground and thus , legitimately , may itself be called into question . indeed , despite its successes , the 1-loop approach has obvious weaknesses .
one example is that the frg flow equation for the equilibrium statics and for depinning are identical , while it is clear that these are two vastly different physical phenomena , depinning being irreversible .
also , the detailed mechanism by which the system escapes dimensional reduction in both cases is not really elucidated .
finally , there exists no convincing scheme to compute correlations , and in fact no calculation of higher than 2-point correlations has been performed .
another motivation to investigate the frg is that it should apply to other disordered systems , such as random field spin models , where dimensional reduction also occurs and progress has been slow @xcite .
insight into model ( [ ham ] ) will thus certainly lead to progresses in a broader class of disordered systems . in this paper
we construct a renormalizable field theory for the statics of disordered elastic systems beyond one loop .
the main difficulty is the non - analytic nature of the theory ( i.e. of the fixed point effective action ) at @xmath39 .
this makes it a priori quite different from conventional field theories for pure systems .
we find that the 2-loop diagrams are naively `` ambiguous '' , i.e. it is not obvious how to assign a value to them .
we want to emphasize that this difficulty already exists at one loop , e.g. _ even the simplest one loop correction to the two point function is naively `` ambiguous''_. thus it is not a mere curiosity but a fundamental problem with the theory , `` swept under the rug '' in all previous studies , but which becomes unavoidable to confront at 2-loop order .
it originates from the metastability inherent in the problem . for the related theory of the depinning transition ,
we have shown in companion papers @xcite how to surmount this problem and we constructed a 2-loop renormalizable field theory _ from first principles_. there , all ambiguities are naturally lifted using the known exact property that the manifold only moves forward in the slowly moving steady state .
unfortunately in the statics there is no such helpful property and the ambiguity problem is even more arduous . here
we examine the possible ways of curing these difficulties .
we find that the natural physical requirements , i.e.that the theory should be ( i ) _ renormalizable _
( i.e. that a universal continuum limit exists independent of short - scale details ) , ( ii ) that the renormalized force should remain _ potential _ , and ( iii ) that no stronger singularity than the cusp in @xmath51 should appear to two loop ( i.e. _ no `` supercusp '' _ ) , are rather restrictive and constrain possible choices .
we then propose a theory which satisfies all these physical requirements and is consistent to two loops .
the resulting @xmath0-function differs from the one derived in previous studies @xcite by novel static `` anomalous terms '' .
these are different from the dynamical `` anomalous terms '' obtained in @xcite showing that indeed depinning and statics differ at two loop , fulfilling another physical requirement .
we then study the fixed points describing several universality classes , i.e. the interface with rb and rf disorder , the random periodic problem , and the case of lr elasticity .
we obtain the @xmath4 corrections to several universal quantities .
the prediction for the roughness exponent @xmath19 for random bond disorder has the correct sign and order of magnitude to notably improve the precision as compared to numerics in @xmath52 and to match the exact result @xmath53 in @xmath15 . for random field disorder
we find @xmath3 which , for equilibrium is likely to hold to all orders .
by contrast , non - trivial corrections of order @xmath4 were found for depinning @xcite .
the amplitude , which in that case is a universal function of the random field strength is computed and it is found that the 2-loop result also improves the agreement as compared to the exact result known @xcite for @xmath54 . for the periodic cdw case we compare with the numerical simulations in @xmath13 and obtain reasonable agreement .
some of the results of this paper were briefly described in a short version @xcite and agree with a companion study using exact rg @xcite ) .
since the physical results also seem to favor this theory we then look for better methods to justify the various assumptions .
we found several methods which allow to lift ambiguities and all yield consistent answers .
a detailed discussion of these methods is given .
in particular we find that correlation functions can be unambiguously defined in the limit of a small background field which splits apart quasi - degenerate states when they occur .
this is very similar to what was found in a related study where we obtained the exact solution of the frg in the large @xmath6 limit @xcite .
finally , the methods introduced here will be used and developed further to obtain a renormalizable theory to three loops , and compute its @xmath0-function in @xcite .
let us mention that a first principles method which avoids ambiguities is to study the system at @xmath49
. however , this turns out to be highly involved .
it is attempted via exact rg in @xcite and studied more recently in @xcite where a field theory of thermal droplet excitation was constructed .
a short account of our work has appeared in @xcite , and a short pedagogical introduction is given in @xcite .
the outline of this paper is as follows . in section [ sec : model ]
we explain in a detailed and pedagogical way the perturbation theory and the power counting . in section [ sec : renprog ] we compute the 1-loop ( section [ sec : oneloop ] ) and 2-loop ( section [ sec : twoloop ] ) corrections to the disorder .
the calculation of the repeated 1-loop counter - term is given in section [ sec : rgdis ] . in section [ sec : finalbeta ] we identify the values for ambiguous graphs .
this yields a renormalizable theory with a finite @xmath0-function , which is potential and free of a supercusp .
the more systematic discussion of these ambiguities is postponed to section [ sec : ambiguities ] .
we derive the @xmath0-function and in section [ sec : fixedpoints ] present physical results , exponents and universal amplitudes to @xmath4 .
some of these quantities are new , and have not yet been tested numerically . in section [ sec : ambiguities ] we enumerate all the methods which aim at lifting ambiguities and explain in details several of them which gave consistent results . in section [ sec : correlations ] we detail the proper definition and calculation of correlation functions . in appendix
[ sec : symm ] and [ sec : nona ] we present two methods which seem promising but _ do not _ work , in order to illustrate the difficulties of the problem . in appendix [ sec : finitet ] we present a summary of all one and 2-loop corrections including finite temperature . in appendix
[ sec : sloopbc ] we give details of calculations for what we call the sloop elimination method .
the reader interested in the results can skip section [ sec : model ] and section [ sec : renprog ] and go directly to section [ sec : fixedpoints ] .
the reader interested in the detailed discussion of the problems arising in this field theory should read section [ sec : ambiguities ] .
we study the static equilibrium problem using replicas , i.e. consider the partition sum in presence of sources : @xmath55 = \int \prod_a { \cal d}[u_a]\ , \exp\left(- { \cal s}[u ] + \int_x \sum_a j_x^a u_x^a\right ) \ , \ ] ] from which all static observables can be obtained .
the action @xmath56 and replicated hamiltonian corresponding to ( [ ham ] ) are @xmath57 = \frac{{\cal h}[u]}{t } & = & \frac{1}{2 t } \int_x \sum_a [ ( \nabla u^a_x)^2 + m^2 u^a_x ] \nonumber \\ & & - \frac{1}{2 t^2 } \int_x \sum_{a b } r(u^a_x - u^b_x ) \ . \label{action}\end{aligned}\ ] ] @xmath58 runs from 1 to @xmath59 and the limit of zero number of replicas @xmath60 is implicit everywhere .
we have added a small mass which confines the interface inside a quadratic well , and provides an infrared cutoff .
we are interested in the large scale limit @xmath61 .
we will denote @xmath62 for periodic systems the integration is over the first brillouin zone .
a short - scale uv cutoff is implied at @xmath63 , but for actual calculations we find it more convenient to use dimensional regularization .
we also consider the effective action functional @xmath64 $ ] associated to @xmath56 .
it is , as we recall @xcite , the legendre transform of the generating function of connected correlations @xmath65
= \ln { \cal z}[j]$ ] , thus defined by eliminating @xmath66 in @xmath64 = j u - { \cal w}[j]$ ] , @xmath67 = u$ ] .
if we had chosen non - gaussian disorder additional terms with free sums over @xmath68 replicas ( called @xmath68-replica terms ) corresponding to higher cumulants of disorder would be present in ( [ action ] ) , together with a factor of @xmath69 .
these terms are generated in the perturbation expansion , i.e. they are present in @xmath64 $ ] .
we do not include them in ( [ action ] ) because , as we will see below , these higher disorder cumulants are not relevant within ( conventional ) power counting , so for now we ignore them .
the temperature @xmath70 appears explicitly in the replicated action ( [ action ] ) , although we will focus on the @xmath39 limit . because the disorder distribution is translation invariant , the disorder term in the above action is invariant under the so called statistical tilt symmetry @xcite ( sts ) , i.e. the shift @xmath71 .
one implication of sts is that the 1-replica replica part of the action ( i.e. the first line of [ action ] ) is uncorrected by disorder , i.e. it is the same in @xmath64 $ ] and @xmath72 $ ] @xcite .
since the elastic coefficient is not renormalized , we have set it to unity .
we first study perturbation theory , its graphical representation and power counting . everywhere in the paper
we denote the exact 2-point correlation by @xmath73 , i.e. in fourier : @xmath74 while the free correlation function ( from the elastic term ) used for perturbation theory in the disorder is denoted by @xmath75 and reads in fourier : @xmath76 which is represented graphically by a line : @xmath77{./figures / prop}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / prop}{\hspace*{0.8ex}}}}_{b } ~= \frac{t\delta_{ab}}{q^{2}+m^{2}}\ .\ ] ] each propagator thus carries one factor of @xmath78 .
each disorder interaction vertex comes with a factor of @xmath79 and gives one momentum conservation rule .
since each disorder vertex is a function , an arbitrary number of lines can come out of it .
@xmath80 lines coming out of a vertex result in @xmath80 derivatives @xmath81 after wick contractions @xmath82{./figures / rp}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / rp}{\hspace*{0.8ex}}}}= r^{(k)}\ .\ ] ] since each disorder vertex contains two replicas it is sometimes convenient to use `` splitted vertices '' rather than `` unsplitted ones '' .
thus we call `` vertex '' an unsplitted vertex and we call a `` point '' the half of a vertex . @xmath83{./figures / rab}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / rab}{\hspace*{0.8ex } } } } = \sum_{ab } \frac{r ( u_{a}-u_{b})}{2 t^{2}}\ .\ ] ] each unsplitted diagram thus gives rise to several splitted diagrams , as illustrated in fig .
[ f:1loopexpstat ] ) is a three replica term , title="fig:",width=264 ] one can define the number of connected components in a graph with splitted vertices .
since each propagator identifies two replicas , a @xmath68-replica term contains @xmath68 connected components .
when the 2-points of a vertex are connected , this vertex is said to be `` saturated '' .
it gives a derivative evaluated at zero @xmath84 .
standard momentum loops are loops with respect to unsplitted vertices , while we call `` sloops '' the loops with respect to points ( in splitted diagrams ) .
this is illustrated in fig .
( [ f : sloops ] ) .
the left vertex on diagram ( c ) is `` saturated '' : replica indices are constrained to be equal and thus the diagram does not depend on the left space point.,title="fig:",width=302 ] the momentum 1-loop and 2-loop diagrams which correct the disorder at @xmath39 are shown in fig .
[ alldiag2loop ] ( unsplitted vertices ) .
there are three types of 2-loop graphs a , b and c. since they have two vertices ( a factor @xmath85 each ) and three propagators ( a factor of @xmath70 each ) the graphs e and f lead to corrections to @xmath86 proportional to temperature and will not be studied here ( see however appendix [ sec : finitet ] ) .
it is important to distinguish between _ fully saturated diagrams _ and _ functional diagrams_. the fs diagrams are those needed for a full average , e.g. a correlation function .
there all fields are contracted and one is only left with the space dependence .
these are the standard diagrams in more conventional polynomial field theories such as @xmath87 . then all vertices are evaluated at @xmath18 , yielding products of derivatives @xmath84
. these are also the graphs which come in the standard expansion of @xmath64 $ ] in powers of @xmath30 which generate the `` proper '' or `` renormalized '' vertices , i.e. the sum over all 1-particle irreducible graphs with some external legs , from which all correlations can be obtained .
note that in the fully saturated diagrams there can be no free point , all points in a vertex have to be connected to some propagator ( and to some external replica ) otherwise there is a free replica sum yielding a factor of @xmath59 and a vanishing contribution in the limit of @xmath60 .
however , since we have to deal with a function @xmath27 we will more often consider functional diagrams .
a functional diagram still depends on the field @xmath30 .
it can depend on @xmath30 at several points in space ( multi - local term ) , as for example : @xmath88{0mm}{2mm}_{x}{\settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / func1}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / func1}{\hspace*{0.8ex}}}}_{y } \sim \sum_{abc } \frac{r'(u^a_x - u^b_x)}{2 t^2 } \frac { r'(u^a_y - u^c_y)}{2 t^2 } t g(x - y)\ ] ] such a graph with @xmath68 connected components corresponds to a @xmath68 replica functional term . or it can represent the projection of such a term onto a local part , as arises in the standard operator product expansion ( ope ) : @xmath89{./figures / func1}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / func1}{\hspace*{0.8ex } } } } \sim \sum_{abc } \frac{r'(u^a_x - u^b_x)}{2 t^2 } \frac{r'(u^a_x - u^c_x)}{2 t^2 } t \int_y g(x - y)\ .\ ] ] typically using functional diagrams we want to compute the effective action functional @xmath64 $ ] , or its local part , i.e. its value for a spatially uniform mode @xmath90 , which includes the corrections to disorder . specifying the two replicas on each connected component , one example of
a 1-particle irreducible diagram producing corrections to disorder is @xmath91{./figures / func2}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / func2}{\hspace*{0.8ex } } } } \sim \frac{t^2}{t^4 } r''(u^a - u^b ) r''(u^a - u^b ) \int_q g(q)^2\ .\ ] ] the complete analysis of these corrections will be made in section ( [ sec : renprog ] ) .
finally , note that functional diagrams may contain saturated vertices , whose space and field dependence disappears ( such as ( c ) in fig . [
f : sloops ] ) and that the limit @xmath92 does not produce constraints .
an example is the calculation of @xmath64 $ ] since one can always attach additional external legs to any point by taking a derivative with respect to the field @xmath30 .
if we consider fully saturated diagrams and analytic @xmath27 we find trivial results .
this is because at @xmath39 the model exhibits the property of dimensional reduction @xcite ( dr ) both in the statics and dynamics .
its `` naive '' perturbation theory , obtained by taking for the disorder correlator @xmath27 an _ analytic function _ of @xmath30 has a triviality property . as is easy to show using the above diagrammatic rules
( see a typical cancellation due to the `` mounting '' construction in fig .
[ f : dimred ] , see also appendix d in ref .
@xcite ) the perturbative expansion of any correlation function @xmath93 ( of any _ analytic _ observable ) in the derivatives @xmath84 yields to all orders the same result as that obtained from the gaussian theory setting @xmath94 ( the so called larkin random force model ) .
the 2-point function thus reads to all orders : @xmath95 ( at @xmath39 correlations are independent of the replica indices @xmath96 ) .
this dimensional reduction results in a roughness exponent @xmath40 which is well known to be incorrect .
one physical reason is that this @xmath39 perturbation theory amounts to solving in perturbation the zero force equation @xmath97 this , whenever more than one solution exists ( which certainly happens for small @xmath98 ) is clearly not identical to finding the lowest energy configuration . )
arises if one averages over multiple solutions @xmath99 with some random weights @xmath100 ( then using the representation of the delta function @xmath101 and averaging over disorder using ( [ corrstat ] ) . summing over multiple solutions @xmath99 requires instead to include the crucial weight @xmath102 in order to select the true ground state . ]
curing this problem within the field theory , is highly non - trivial .
coarse graining within the frg up to a scale at which the renormalized disorder correlator @xmath27 becomes _ non - analytic _ ( which includes some of the physics of multiple extrema ) is one possible route , although understanding exactly how this cures the problem within the field theory is a difficult open problem .
it is important to note that dimensional reduction is not the end of perturbation theory , since saturated diagrams remain non - trivial at finite temperature , so one way out is to study @xmath49 .
this is not the route chosen here , instead we will attempt to work at @xmath39 with a non - analytic action and focus on functional diagrams which remain non - trivial . .
due to dr only the first diagram ( a ) survives .
diagrams ( b ) and ( c ) cancel because by shifting the line one gets a minus sign .
diagram ( c ) is proportional to @xmath103 and vanishes in an analytic theory .
similar cancellations occur to all orders.,title="fig : " ] let us now consider power counting .
let us recall the conventional analysis within e.g. the wilson scheme @xcite .
the elastic term is invariant under @xmath104 , @xmath105 and @xmath106 , with @xmath107 .
@xmath19 is for now undetermined . under this transformation
the disorder function @xmath86 is multiplied by @xmath108 .
it becomes relevant for @xmath109 , provided @xmath110 which is physically expected ( for instance in the random periodic case , @xmath111 is the only possible choice , and for other cases @xmath112 ) .
the rescaled dimensionless temperature term scales as @xmath113 ( see below ) and is formally irrelevant near four dimension . in the end @xmath19 will be fixed by the disorder distribution at the fixed point . to be more precise , we want to determine in the field theoretic framework the necessary counter - terms to render the theory uv finite as @xmath114 .
the study of superficial divergences usually involves examining the irreducible vertex functions ( ivf ) : @xmath115 { \hskip0.1ex\raisebox{-1ex}[0ex][0.8ex]{\rule{0.1ex}{2.75ex}\hskip0.2ex}}_{u=0}\ ] ] with @xmath116 external fields @xmath30 ( at momenta @xmath117 , @xmath118 ) .
the perturbation expansion of a given ivf to any given order in the disorder is represented by a set of 1-particle irreducible ( 1pi ) graphs ( in unsplitted diagrammatics ) .
being the derivative of the effective action they are the important physical objects since all averages of products of fields @xmath30 can be expressed as tree diagrams of the ivf .
finiteness of the ivf thus implies finiteness of all such averages .
however since @xmath64 $ ] is non - analytic in some directions ( e.g. for a uniform mode @xmath119 ) , derivatives such as ( [ gammaexp ] ) may not exist at @xmath120 , and we have to be more general and consider _ functional diagrams_. the ( disorder part of the ) effective action is the sum of @xmath80-replica terms , noted @xmath121 $ ] @xmath122 = \sum_{k \geq 2 } t^{-k } \gamma_k[u]\ .\ ] ] each @xmath121 $ ] is the sum over 1pi graphs with @xmath80 connected components ( using splitted vertices ) , and itself depends on @xmath70 as @xmath123 = \sum_{l \geq 0 } t^{l } \gamma_{k , l}[u]\ , \ ] ] where @xmath124 is the number of sloops .
thus at @xmath39 there are no sloops and @xmath121 = \gamma_{k , l=0}[u]$ ] is the sum over 1pi _ tree _ graphs with @xmath80 connected components ( trees in replica - space , not position - space ) .
let us compute the superficial degree of uv divergence @xmath125 of a functional graph entering the expansion of the local part of the effective action .
we denote @xmath126 the number of unsplitted disorder vertices , @xmath127 the number of internal lines ( propagators ) , @xmath128 the number of loops and @xmath124 the number of sloops .
one has the relations @xmath129 the total factors of @xmath70 are @xmath130 . at @xmath39 ( @xmath131 )
the superficial degree of uv divergence is thus @xmath132 thus in @xmath38 the only graphs with positive superficial degree of divergence are for @xmath133 ( quadratic @xmath134 ) , and @xmath135 ( log divergence ) .
@xmath133 corresponds to a constant in the free energy .
because of sts all single replica terms are uncorrected and there is no wave - function renormalization in this model .
thus to renormalize the @xmath39 theory we need a priori to look only at graphs with @xmath136 connected components , which by definition are those correcting the second cumulant @xmath27 , compute their divergent parts , and construct the proper counter - term to the function @xmath27 .
as mentioned above , higher cumulants are irrelevant by power counting , and are superficially uv - finite .
the graphs which contribute to the 2-replica part @xmath137 $ ] have @xmath128 loops with @xmath138 . at zero temperature ,
@xmath131 , thus @xmath139 . the loop expansion thus corresponds to the expansion in power of @xmath27 and , as we will see below , to an @xmath5-expansion .
more generally using the above relation one has , schematically @xmath140 = \sum_{l \geq \max(1,2+l - k ) } \partial^{(4 l - 4 + 2 k - 2 l)}_u r^{(l-1+k - l ) } \ , \ ] ] where the number of internal lines gives the total number of derivatives acting on an argument @xmath30 of the functions @xmath86 .
for instance , the 2-replica part at @xmath39 is a sum over @xmath128-loop graphs of the type @xmath141 = \sum_{l \geq 1 } \partial^{4 l}_u r^{l+1}\ .\ ] ] if one now considers @xmath49 one finds that @xmath142 .
each additional power of @xmath70 yields an additional quadratic divergence , more generally a factor of @xmath143 .
thus to obtain a theory where observables are finite as @xmath144 one must start from a model where the initial temperature scales with the uv cutoff as @xmath145 this is similar to @xmath87-theory where it is known that a @xmath146 term can be present and yields a finite uv limit ( i.e.does not spoil renormalizability ) only if it has the form @xmath147 .
such a term , with precisely this cutoff dependence , is in fact usually present in the starting bare model , e.g. in lattice spin models .
it then produces only a finite shift to @xmath148 without changing universal properties . here
each factor of @xmath149 comes with a factor of @xmath150 which compensates the uv divergence from the graph .
thus the finite-@xmath70 theory may also be renormalizable .
computing the resulting shift in @xmath27 to order @xmath151 by resumming the diagrams @xmath152 and @xmath153 of fig .
[ alldiag2loop ] and all similar diagrams to any number of loops has not been attempted here ( see however appendix [ sec : finitet ] ) .
the `` finite shift '' here is , however , much less innocuous than in @xmath87-theory since it smoothes the cusp .
the effects of a non - zero temperature are explored in @xcite .
one can use the freedom to rescale @xmath30 by @xmath154 .
the dimensionless temperature @xmath155 is then defined .
the disorder term in @xmath64 $ ] is then is as in ( [ action ] ) with @xmath27 replaced by @xmath156 in terms of a dimensionless rescaled function @xmath157 of a dimensionless rescaled argument .
this will be further discussed below .
in this section we compute the effective action to 2-loop order at @xmath39 .
we are only interested in the part which contains uv divergences as @xmath158 . we know from the analysis of the last section that we only need to consider the local @xmath135 2-replica part , i.e. the corrections to @xmath27 .
these @xmath159 and @xmath160 loop corrections contain @xmath161 vertices .
higher @xmath126 yields higher number of replicas . to one loop at @xmath39
there is only one unsplitted diagram @xmath162 , corresponding to two splitted diagrams ( a ) and ( b ) as indicated in figure [ oneloop ] .
both come with a combinatorial factor of @xmath163 from taylor - expanding the exponential function and @xmath164 from the action .
( a ) has a combinatoric factor of @xmath165 and ( b ) of @xmath166 .
together , they add up to the 1-loop correction to disorder @xmath167
i_{1 } \label{1loop}\\ i_1^{m}=i_{1 } & : = & \int_q \frac{1}{(q^2+m^2)^2 } \nonumber\\ & = & \gamma \left ( 2 - \frac{d}{2}\right ) m^{-\epsilon } \int_q { { \mathrm{e}}}^{-q^2 } \nonumber \\ & = & \frac{1}{(4 \pi)^{d/2 } } \gamma \left ( 2 - \frac{d}{2}\right ) m^{-\epsilon}= o\left(\frac{1}\epsilon\right)\ .\label{i1}\qquad\end{aligned}\ ] ] note that ( b ) has a saturated vertex , hence the factor @xmath168 .
this does not lead to ambiguities in the 1-loop @xmath0-function , since the frg to one loop yields a discontinuity only in the third derivative and @xmath51 remains continuous .
, title="fig:",width=226 ] there are only three graphs correcting disorder at @xmath39 with @xmath160 loops and @xmath169 vertices .
they are denoted @xmath170 , @xmath171 and @xmath172 and we will examine each of them .
we begin our analysis with class a. ) contributing to 2-replica terms.,title="fig:",width=264 ] the possible diagrams with splitted vertices of type a are diagrams ( a ) to ( f ) given in fig .
[ static-2-loop - do ] .
the resulting correction to @xmath27 is written as : @xmath173 where the combinatorial factors are : @xmath174 from the taylor - expansion of the exponential function , @xmath175 from the explicit factors of @xmath164 in the interaction , a factor of 3 to chose the vertex at the top of the hat , and a factor of 2 for the possible two choices in each of the vertices . furthermore below some additional combinatorial factors
are given : a factor of @xmath165 for generic graphs and @xmath176 if it has the mirror symmetry with respect to the vertical axis .
each diagram symbol denotes the diagram including the symmetry factor .
the first two graphs are : @xmath177 to obtain the sign one can choose an `` orientation '' in each vertex ( @xmath178 ) , the final result does not depend on the choice .
the minus sign in @xmath58 comes because the two legs enter on opposite points in the top vertex .
define the 2-loop momentum integral ( see appendix a in ref .
@xcite ) @xmath179 graphs @xmath58 and @xmath180 are non - ambiguous .
they are the only contributions in an analytic theory .
the other graphs are @xmath181 and vanish if @xmath27 is analytic ( since then @xmath182 ) but a priori should be considered when @xmath27 is non - analytic .
we have indicated their `` natural '' sign and amplitude ( e.g. symmetry factor setting @xmath183 ) but have introduced factors @xmath184 to recall that they are _ ambiguous _ : since @xmath185 one is confronted to a choice each time one saturates a vertex and there is no obvious way to choose the sign at this stage .
we recall that we have defined _ saturated _ vertices as vertices evaluated at @xmath18 while _ unsaturated _ vertices still contain @xmath30 and do not lead to ambiguities . at this stage
we will not discuss in detail how to give a definite values to these contributions to disorder .
this will be done in section [ sec : ambiguities ] .
we will just use the most reasonable assumptions , which will be reevaluated , and justified later .
a natural step is to set @xmath186 since these graphs can not correct @xmath27 as they are _ odd _ functions of @xmath30 , which yields no contribution when inserted into the action @xmath187 .
we now turn to graphs of type b ( bubble - diagrams ) , @xmath188 to @xmath124 represented in fig
. [ banana ] .
we use the same convention as in ( [ corr ] ) , and start with the combinatorics .
there are 3 ways to choose the vertex in the middle . upon splitting the vertices , for @xmath189 and @xmath66
there are only two choices at the middle vertex whereas for @xmath188 there are four choices .
there are also four choices for @xmath190 , @xmath80 and @xmath124 .
there , one must also choose the rightmost vertex , leading to an extra factor of @xmath165 .
the final result is @xmath191 only @xmath80 and @xmath124 are ambiguous but it is also natural to set : @xmath192 which we do for now , and discuss later .
diagrams @xmath98 , @xmath59 , @xmath68 , @xmath193 of class c are represented in fig .
[ f : bonetdane ] @xmath194 with @xmath195 there it is natural to assume @xmath196 which we do for now and discuss it later .
this leaves no correction to disorder from graphs c , as is the case for depinning @xcite .
this is fortunate , since the integral @xmath197 has a quadratic uv - divergence in @xmath38 , while @xmath198 is uv - finite .
physically , it is unlikely that these could enter physical observables as the tadpole divergence can usually be eliminated by proper field reordering ( normal - ordering ) or vacuum subtraction . to summarize , for the equilibrium statics at @xmath39 in perturbation of @xmath199 , the contributions to the disorder to one and two loops , i.e. the corresponding terms in the effective action @xmath200 $ ] are @xmath201 i_{1 } \qquad \quad \label{d21 } \\ \delta^{2 } r(u ) & = & \left [ r'''(u)^2 ( r''(u ) - r''(0 ) ) \right ] i_a { \nonumber}\\ & & + \frac{1}{2 } \left [ ( r''(u ) - r''(0))^2 r''''(u ) \right ] i_1 ^ 2{\nonumber}\\ & & - \lambda r'''(0^+)^2 r '' ( u ) i_a \label{d22 } \ .\end{aligned}\ ] ] we have allowed for a yet undetermined constant @xmath202 .
we now show that requiring renormalizability allows to fix @xmath203 .
let us now recall the method , also used in our study of depinning @xcite , to renormalize a theory where the interaction is not a single coupling - constant , but a whole function , the disorder - correlator @xmath27 .
we denote by @xmath204 the bare disorder
this is the object in which perturbation theory is carried out
i.e. one considers the bare action ( [ action ] ) with @xmath205 .
we denote here by @xmath86 the renormalized dimensionless disorder i.e. the corresponding term in the effective action @xmath64 $ ] is @xmath206 ( i.e. the local 2-replica part of @xmath64 $ ] ) .
symbolically , we can write @xmath207 & \leftrightarrow & r_0 \\ \gamma[u ] & \leftrightarrow & m^\epsilon r\ .\label{lf106}\end{aligned}\ ] ] we define the dimensionless symmetric bilinear 1-loop and trilinear 2-loop functions ( see ( [ d21 ] ) and ( [ d22 ] ) ) such that @xmath208 they can be extended to non - equal argument using @xmath209 $ ] and a similar expression for the trilinear function . whenever possible we will use the shorthand notation @xmath210 and @xmath211 .
the expression of @xmath86 obtained perturbatively in powers of @xmath204 at 2-loop order reads : @xmath212 it contains terms of order @xmath213 and @xmath214 .
this is sufficient to calculate the rg - functions at this order . in principle
, one has to keep the finite part of the 1-loop terms , but we will work in a scheme , where these terms are exactly 0 , by normalizing all diagrams by the 1-loop diagram . inverting ( [ lf39 ] ) yields @xmath215 \ , \ ] ] where @xmath216 is the 1-loop repeated counter - term : @xmath217 the @xmath0-function is by definition the derivative of @xmath86 at fixed @xmath204 .
it reads @xmath218[0.8ex]{\rule{0.1ex}{2.75ex}\hskip0.2ex}}_{r_0 } & = & \epsilon \big [ m^{-\epsilon}r_0 + 2\delta ^{(1 ) } ( m^{-\epsilon}r_0 ) { \nonumber}\\ & & \ \ \ \ + 3 \delta ^{(2 ) } ( m^{-\epsilon}r_0 ) + \dots \big]\ .\qquad \label{betar0}\end{aligned}\ ] ] using the inversion formula ( [ do ] ) , the @xmath0-function can be written in terms of the renormalized disorder @xmath86 : @xmath219[0.8ex]{\rule{0.1ex}{2.75ex}\hskip0.2ex}}_{r_{0 } } & = & \epsilon \big [ r+\delta ^{(1 ) } ( r ) \nonumber \\ & & \quad+2\delta ^{(2 ) } ( r)-\delta ^{(1,1 ) } ( r ) + \dots \big]\ .
\qquad\qquad \label{beta - renor - do}\end{aligned}\ ] ] in order to proceed , let us calculate the repeated 1-loop counter - term @xmath220 .
we start from the 1-loop counter - term ( [ d21 ] ) , which has the bilinear form @xmath221 \tilde i_1 \qquad \qquad \label{lf42}\end{aligned}\ ] ] with the dimensionless integral @xmath222[0.8ex]{\rule{0.1ex}{2.75ex}\hskip0.2ex}}_{m=1}$ ] ; we will use the same convention for @xmath223[0.8ex]{\rule{0.1ex}{2.75ex}\hskip0.2ex}}_{m=1}$ ] .
thus @xmath224 reads @xmath225 \tilde i_1 ^ 2 { \nonumber}\ .\\ \label{1-loop - rep - cc}\end{aligned}\ ] ] in the course of the calculation the only possible ambiguity could come from @xmath226''\big|_{u \to 0 } \nonumber \\ & = & \big [ r'''(u)^2 - r''''(u ) ( r''(u ) - r''(0 ) ) \big]\big|_{u \to 0 } \nonumber
\\ & = & r'''(0^+)^2 \label{lf43}\end{aligned}\ ] ] but there is _ no ambiguity _ since the function @xmath227 is continuous at @xmath18 with value @xmath228 .
this is exactly the same calculation as is done to one loop when computing the non - trivial fixed point for the pinning force correlator @xmath229 yielding @xmath230 .
thus there is no doubt that the graph @xmath231 with the 1-loop counter - term inserted in a 1-loop diagram is _ non - ambiguous_. the 2-loop @xmath0-function ( [ beta - renor - do ] ) then becomes with the help of ( [ 1-loop - rep - cc ] ) @xmath232 ( \epsilon \tilde i_{1 } ) \nonumber \\ & & + \left [ ( { r}''(u ) - { r}''(0 ) ) { r}'''(u)^2 \right ] \,\epsilon\ ! \left ( 2\tilde i_{a}-\tilde i_{1}^{2 } \right){\nonumber}\\ & & - ( { r}'''(0^+))^2 { r}''(u ) \ , \epsilon\ ! \left ( 2 \lambda \tilde i_{a}-\tilde i_{1}^{2 } \right ) \label{rgdisorderunrescaled}\ .\end{aligned}\ ] ] the first result is that , apart from the last `` anomalous '' term , the @xmath233-terms cancel in the corrections to disorder . in the terms coming from graphs
a this works because , as we recall , @xmath234 so that the combination @xmath235 is finite .
graphs b cancel completely since we have chosen as counter - term the full 1-loop graph .
so for an analytic theory the above @xmath0-function would be finite .
this however is incomplete , since the flow of such a @xmath0-function leads to a non - analytic @xmath236 above the larkin scale .
thus we must consider the last , `` anomalous '' term in ( [ rgdisorderunrescaled ] ) .
it clearly appears that the only value of @xmath203 compatible with the cancellation of the @xmath233 poles is @xmath237 leading to a finite @xmath0-function .
thus the requirement that the theory be renormalizable ( i.e. yield universal large scale results independent of the short - scale details ) fixes the value @xmath238 .
note that the cancellation of the graphs b also works thanks to ( [ lf2 ] ) .
it is interesting to compare with what happens at depinning .
there the cancellation of the @xmath233-terms in the anomalous part is more complicated but automatic .
it requires a consistent evaluation of all anomalous non - analytic diagrams . in the depinning theory
the cancellation was unusual : a non - trivial bubble diagram ( called @xmath239 in @xcite ) was crucial in achieving the cancellation . in the statics the 2-loop bubble diagrams of type b appear to be simply the square of the 1-loop ones which is the usual situation .
this however is clearly a consequence of ( [ lf2 ] ) so the previous experience with depinning indicates that care is required and we will discuss some justification for ( [ lf2 ] ) below . in the search for a fixed point
it is convenient to write the @xmath0-function for the rescaled function @xmath240 defined through @xmath241 which amounts to rescale the fields @xmath30 by @xmath242 .
note that this is a simple field rescaling and different from standard wave - function renormalization , since as mentioned above there is none in this theory due to sts .
we have also included the 1-loop integral factor to simplify notations and further calculations ( equivalently it can be absorbed in the normalization of momentum or space integrals ) . with this
, the @xmath243-function takes the simple form : @xmath244 \nonumber \\ & & + \frac{1}{2 } x \left [ ( \tilde{r}''(u ) - \tilde{r}''(0 ) ) \tilde{r}'''(u)^2 \right ] { \nonumber}\\ & & - \frac{\lambda}{2 } x ( \tilde{r}'''(0^+))^2 \tilde{r}''(u ) \label{rgdisorder}\ . \\
\lambda = 1 \ , & & \qquad x=1\ .
\label{lf45}\end{aligned}\ ] ] we have left a @xmath203 for future use , but its value in the theory we study here is set to @xmath176 .
also for convenience we have introduced @xmath245 which is @xmath246 in the @xmath5-expansion studied here , but has a different value for lr elasticity , see below .
in fact it is shown in appendix [ app : universal ] that @xmath247 is independent of the particular infrared cutoff procedure ( here a massive scheme ) .
although the global rescaling factor of @xmath157 , @xmath248 , has @xmath249 corrections which depend on the infrared cutoff chosen , the frg equation above does not depend on it .
note that the above equation remains true in fixed dimension , with the appropriate value for @xmath250 , up to terms of order @xmath251 .
we will see that the value @xmath252 in ( [ rgdisorder ] ) has other highly desirable properties .
first this value is _ the only one _ which guarantees that the non - analyticity in @xmath240 does not become _ more severe _ at two loops than it is at one loop .
let us take one derivative of ( [ rgdisorder ] ) and take @xmath253 .
one finds : @xmath254 thus if @xmath255 the cusp in @xmath256 and the resulting finite value of @xmath257 immediately creates a cusp in @xmath258 . the singularity has become worse !
we call this a supercusp
. it must be avoided in the statics ( see also discussion in section [ sec : ambiguities ] ) .
interestingly it _ does occur in the driven dynamics _
, where it is a physical signature of irreversibility .
indeed this property is intimately related to another highly desirable property of the statics : _
potentiality_. this property is more conveniently described by considering the flow equation for the ( rescaled ) _ correlator of the pinning force _
@xmath259 , the second derivative of ( [ rgdisorder ] ) : @xmath260 '' \nonumber \\ & & + \frac{1}{2 } \left [ ( \tilde{\delta}(u ) - \tilde{\delta}(0 ) ) \tilde{\delta}'(u)^2 \right ] '' { \nonumber}\\ & & - \frac{\lambda}{2 } ( \tilde{\delta}'(0^+))^2 \tilde{\delta}''(u ) \label{rgdisorderdelta}\ .\end{aligned}\ ] ] formally , this equation could have been obtained directly from a study of the dynamical field theory . such an equation was indeed obtained at depinning but with a different value of @xmath203 : @xmath261 which shows that statics and dynamics differ not at one , but at two loops . integrating the equation for @xmath262
once yields a non - zero fixed point value for @xmath263 unless @xmath238 .
potentiality on the other hand requires that the force remains the derivative of a potential and that , for short - range disorder ( e.g. rb for interface ) one must have @xmath264 .
while violating potentiality is desirable at depinning where irreversibility is expected , this would be physically incorrect in the statics , and thus again points to the value @xmath238 as the physically correct one .
thus we will for now assume that this is the correct theory of the statics and explore its consequences in the next section . in section [ sec : ambiguities ] we will provide better justifications , and explain our understanding of the tantalizing problem of ambiguous diagrammatics in the non - analytic theory of pinned disordered systems .
especially we will present methods , which satisfy all the above constraints of renormalizability , absence of a supercusp and potentiality up to 3-loop order @xcite .
the frg - equation derived above describes several different physical situations , and admits a small number of fixed - point functions @xmath265 describing a few universality classes .
the fixed point associated to a periodic disorder correlator describes single component periodic systems ( such as charge density waves ) .
the fixed point associated to a short - range ( exponentially decaying ) correlator @xmath266 describes a class of systems with so called random bond disorder .
there is also a family of fixed points associated to long range , i.e. algebraic , correlations .
this includes , as one particular example , the random field disorder , which will be discussed separately .
we now give the results for these fixed points , first for short - range elasticity , then for lr elasticity , and compare with available numerical and exact results .
the most important quantity to compute is the roughness exponent @xmath19 . since we have shown that @xmath250 in ( [ rgdisorder ] ) is universal to dominant order this proves universality of @xmath19 to the order in @xmath5 studied here ( i.e.@xmath4 ) . for lr disorder and for periodic fixed points
we can also compute the universal amplitudes for the correlation function of displacements , and discuss their dependence on large scale boundary conditions . anticipating a bit ,
let us summarize the general result that we use in that case , which is derived in section [ sec : correlations ] .
the @xmath39 disorder - averaged 2-point function for @xmath267 , @xmath268 fixed , reads for any dimension @xmath9 , in fourier @xmath269 the amplitude @xmath270 is given by the relation ( exact to all orders in the present scheme ) : @xmath271 it is found to be universal only for long range and periodic disorder .
the scaling function , computed in section [ sec : correlations ] for sr and lr elasticity , is always universal ( independent of short scale details ) and satisfies @xmath272 and @xmath273 where @xmath180 is computed in section [ sec : correlations ] .
this gives us all we need for a calculation to @xmath4 of the universal amplitude , e.g for the propagator in the massless limit @xmath274 : @xmath275 the result for @xmath276 in presence of a mass is also interesting since it gives the fluctuations of the center of mass coordinate for an interface physically confined in a quadratic well .
although that situation would be interesting to study numerically , most numerical results are for finite size systems of volume @xmath277 ( and @xmath278 ) .
we thus also define in that case : @xmath279 with @xmath280 . for _ periodic _ boundary conditions @xmath281 , @xmath282 and @xmath283 .
the prime indicates that the value of this amplitude depends on the large scale boundary conditions , i.e. it depends on whether e.g. a mass is used or periodic boundary conditions as an infrared cutoff .
the ratio , computed in section [ sec : correlations ] for short range elasticity , @xmath284 is unity only for periodic disorder , in which case the amplitude is independent of both large and small scale details . before studying the different fixed points , let us mention an important property , valid under all conditions : if @xmath240 is solution of ( [ rgdisorder ] ) , then @xmath285 is also a solution ( for @xmath286 a constant independent of @xmath98 ) .
we can use this property to fix @xmath287 or @xmath288 in the case of non - periodic disorder .
( for periodic disorder the solution is unique , since the period is fixed . ) let us now look for a solution of our 2-loop frg equation which decays exponentially fast at infinity as expected for sr random - bond disorder . to this aim
, we have to solve order by order in @xmath5 the fixed - point equation ( [ rgdisorder ] ) numerically . making the ansatz @xmath289 the partial differential equation to be solved at
leading order is at 1-loop order .
we have plotted a numerical solution ( red ) as well as the taylor - expansion ( [ taylor1 ] ) about 0 up to order 25 ( blue).,title="fig : " ] @xmath290 where we have used our freedom to normalize @xmath291 .
( [ rrb1loop ] ) has a solution for any @xmath292 , but only for one specific value of @xmath292 does this solution decay exponentially fast to 0 , without crossing the axis , see figure [ f : rranbond1loop ] .
the strategy is thus the following : one guesses @xmath292 , and then integrates ( [ rrb1loop ] ) from 0 to infinity . in practice , however , there are numerical problems for small @xmath30 . one strategy , which we have adopted here , and which works very well , is to use the value of @xmath292 , to generate a taylor - expansion about 0 .
this taylor - expansion is then evaluated at 0.5 , where the numerical integration of ( [ rrb1loop ] ) is started , both forwards to infinity ( which in practice is chosen to be 25 ) and backwards to 0 .
this enables to control the accuracy of both the taylor - expansion and the numerical integration .
the result for the best value @xmath293 is given on figure [ f : rranbond1loop ] .
( note that in @xcite only the first four digits were given . ) on this scale , taylor - expansion and numerical integration are indistinguishable .
the error - estimate on the last digit comes from moving the starting - point of the numerical integration ( which was 0.5 above ) up to 1 , which allows for a crude estimate of the error .
we also reproduce the taylor - expansion up to order 25 below : @xmath294 at second order in @xmath295 , we have to solve @xmath296 where the last equation reflects our choice of @xmath297 .
note that to solve the 2-loop order equation , one has to feed in the solution at 1-loop order , both the taylor - expansion about 0 and the numerically obtained solution for larger @xmath30 . at 2-loop order .
we have plotted a numerical solution ( red ) as well as the taylor - expansion ( [ taylor2 ] ) about 0 up to order 25 ( blue).,title="fig : " ] again @xmath298 is determined from the condition that the solution decays at infinity .
following the same procedure as at 1-loop order , we find @xmath299 the function @xmath300 is plotted on figure [ f : rranbond2loop ] .
the taylor - expansion up to order 25 about 0 reads @xmath301 one observes that @xmath302 is necessarily bounded from above by @xmath303 as no sr solution can cross this value ( to any order ) without exploding .
this reflect the exact bound for sr disorder @xmath304 , which simply means that optimization of energy must lower energy fluctuations compared to a simple sum of random numbers .
equality is obtained for the trivial constant eigenmode @xmath305 corresponding to @xmath306 , associated to the fluctuation of the zero mode of the random potential .
we can now discuss our results for the roughness exponent .
these are summarized in table [ table ] and compared to numerical simulations in @xmath52 and the exact result for the directed polymer in @xmath15 .
a first observation is that the corrections compared to the 1-loop result have the correct sign and , further , that they improve the precision of the 1-loop result .
given the difficulties associated with this theory , this is a significant achievement .
second , the error bars given in table [ table ] are estimated as half the 2-loop contribution , which should not be taken too literally , as it is difficult to obtain a good precision from only two terms of the series and no currently available information about the large order behavior of this novel @xmath5-expansion .
third , one may try to improve the precision using the exact result @xmath307 in @xmath15 . estimating the third order correction in the three possible pade s in order to match @xmath53 for @xmath308
, we obtain consistently the values quoted in the fourth column of table [ table ] .
we hope that these predictions can be tested in higher precision numerics soon . [ cols="^,^,^,^,^,^",options="header " , ] let us first recall that at the level of the _ bare _ model the static random field disorder correlator obeys @xmath309 at large @xmath47 @xcite , where @xmath310 is proportional to the amplitude of the random field .
if one studies the large @xmath30 behavior in the frg equation ( [ rgdisorder ] ) one clearly sees that the non - linear terms do not contribute , thus one has : @xmath311 thus for a rf fixed point to exist , the @xmath4 correction to @xmath19 has to vanish .
@xmath312 this will presumably hold to all orders .
indeed it is clear that if there is a similar @xmath0-function to any order , since each @xmath86 carries at least two derivatives and at least one must be evaluated at @xmath313 , the sum of all non - linear terms to a given _
finite _ order decreases at least as @xmath314 .
( this does not strictly excludes that summing up all orders may yield a slower decay , although it appears far fetched and does not occur in the non - perturbative large-@xmath6 limit . )
the above value of @xmath19 ensures that @xmath315 in the effective action , i.e. non - renormalization of @xmath28 .
note that this argument based on long - range large @xmath30 behavior is a priori valid for any @xmath203 .
since it is made on the @xmath86 equation ( no such argument can be made on the equation for @xmath316 ) it uses the property of potentiality .
however , from ( [ lf3 ] ) with @xmath3 one sees that @xmath255 is incompatible with the existence of a fixed point , even of a fixed point with a supercusp .
thus , the only way to satisfy potentiality for the static random field problem seems to have @xmath28 unrenormalized , @xmath3 and @xmath238 ( the previous discussion of potentiality in section iii.d assumed short - range disorder ) .
this must be contrasted with the theory of depinning , where we found that : @xmath317 following from @xmath318 in ( [ rgdisorderdelta ] ) .
since in that case the rg - flow is non - potential , it is clear that no similar argument as above exists to protect the value @xmath319 .
( the force correlator is short range ) .
the conjecture of @xcite thus appears rather unphysical in that respect .
we first study the fixed - point equation for @xmath320 and later use the rescaling freedom to tune the solution to the correct value of @xmath28 at large scale @xmath30 .
the 2-loop frg equation ( [ rgdisorderdelta ] ) becomes ( @xmath238 ) : @xmath321 \nonumber \ .\end{aligned}\ ] ] one can then integrate once with respect to @xmath30 : @xmath322 \nonumber\ .\end{aligned}\ ] ] there is no integration constant here because the second line precisely vanishes at @xmath323 ( absence of supercusp ) .
the 1-loop solution involves the first line only .
dividing by @xmath324 and integrating over @xmath30 yields : @xmath325 i.e. an implicit equation for @xmath324 , which defines @xmath326 .
it satisfies @xmath327 we can put the 2-loop solution under a similar form . making the ansatz
@xmath328 one obtains @xmath329 at this order , one can replace @xmath324 by @xmath330 , i.e. use @xmath331 to eliminate @xmath332 .
this gives , changing variables from @xmath30 to @xmath324 : @xmath333 ^ 2 } { y-1 } -y \right)\ .\ ] ] the last term in the brackets is easily integrated .
for the remaining terms , we integrate by part and use ( [ k1 ] ) to replace @xmath334 by @xmath335 : @xmath336 this yields the final result @xmath337 we find : @xmath338 has a quadratic behavior around @xmath339 , similar to the 1-loop result , and corrects the value of the cusp . since we know the exact fixed point function up to a scale factor , we can now fix the scale by fitting the exact large @xmath47 behavior to @xmath29 where @xmath28 is the amplitude of the random field .
the general fixed point solution reads : @xmath340 where @xmath341 can be related to @xmath28 as : @xmath342 we need @xmath343 one can now express @xmath344 and thus compute , using ( [ ctd ] ) the universal amplitude ( [ lf109 ] ) associated to the mode @xmath120 in presence of a confining mass : @xmath345^{- \frac{1}{3 } } \sigma^{\frac{2}{3 } } \nonumber \ , \label{lf128}\end{aligned}\ ] ] where one has restored the factors @xmath346 absorbed in @xmath347 and @xmath348 .
expanding all factors in a series of @xmath5 one finds : @xmath349 the lowest order was obtained in ref .
@xcite and we have obtained here the next order corrections .
it is interesting to compare our result with the exact result in @xmath54 , which is @xcite : @xmath350 while the simple extrapolation setting @xmath351 of ( [ lf70 ] ) to one loop @xmath352 is very far off , to two loop it gives @xmath353 , surprisingly close to the exact result .
it was noted in ref .
@xcite that extrapolation of the 1-loop result could be considerably improved by not expanding ( [ lf127 ] ) in @xmath5 but instead directly setting @xmath351 ( with @xmath354 ) in ( [ lf127 ] ) . that gives @xmath355 , an underestimate already reasonably close from the exact result .
we extend this procedure to two loop by truncating the @xmath5 expansion of @xmath356 to second order in ( [ lf127 ] ) , and then set @xmath351 .
this yields @xmath357 , and the exact result is then halfway between @xmath358 and @xmath359 .
to summarize , our 2-loop corrections ( [ lf128 ] ) have the correct sign and order of magnitude to improve the agreement with the exact result in @xmath54 . the universal amplitude for the massless case ( [ fdmassless ] ) ( or @xmath360 ) is obtained from ( [ cd ] ) with @xmath361 from section [ sec : correlations ] as : @xmath362 \nonumber \\ & = & \epsilon^{1/3 } \left [ 3.52459 - 1.89994 \epsilon + o(\epsilon^2 ) \right ] \sigma^{2/3}\ , \qquad \end{aligned}\ ] ] and writing @xmath363 should provide a reasonable extrapolation to low dimensions .
finally , we recall that for random field disorder , this coefficient is different for different large scale boundary conditions .
the result for periodic boundary conditions can be obtained from formula ( [ ratfin ] ) . in ref .
@xcite , the 1-loop result was compared to the result of the gaussian variational method ( gvm ) .
it is instructive to pursue this comparison to two loops .
we get from @xcite : @xmath364^{- \frac{1}{3 } } \frac{1}{1 - \frac{\epsilon}{12 } } \sigma^{\frac{2}{3 } } \\ & = & \epsilon^{1/3 } ( 3.69054 - 0.894223 \epsilon + o(\epsilon^2 ) ) \sigma^{2/3 } \nonumber \\
\frac{c_{\mathrm{gvm } } ( d)}{\tilde c_{\mathrm{gvm}}(d ) } & = & \left ( 1 - \frac{\epsilon - 2 \zeta}{2 } \right)\!\ ! \left ( 1 - \frac{\epsilon - 2 \zeta}{4 } \right ) \frac{\pi \left(\epsilon - 2 \zeta\right)/2 } { \sin(\pi ( \epsilon - 2 \zeta)/2 ) } \nonumber \\ & = & 1 - \frac{\epsilon}{4 } + o(\epsilon^2)\end{aligned}\ ] ] where in the last line we have inserted @xmath3 and performed the @xmath5 expansion .
thus one finds , quite generally that @xmath365 .
as noted in @xcite to one loop the frg and the gvm give rather close amplitudes ( differing by about 5 per cent ) .
we see here that to two loop , i.e. next order in @xmath5 , the difference increases .
finally , @xmath366 and the coefficient remains rather close to the one in ( [ lf131a ] ) . at 1-loop order for non - periodic disorder.,title="fig : " ] there is a family of fixed points such that @xmath367 associated with @xmath368 these fixed points where found for infinite @xmath6 in any @xmath9 in ref .
@xcite ( we use the same notations ) .
they were studied to first order in @xmath5 for any @xmath6 in @xcite , and argued to be stable only for @xmath369 the value of the crossover to short range identified in @xcite as @xmath370 . here
, we have not studied these fixed points in detail but we note that the 2-loop corrections do not change @xmath19 , by the same discussion as for the random field case @xmath371
. they will however affect the amplitudes . for periodic @xmath27 as e.g. cdw
there is another fixed point of ( [ rgdisorder ] ) .
it is sufficient to study the case where the period is set to unity , all other cases are easily obtained using the reparametrization invariance of equation ( [ deltarescale ] ) .
no rescaling is possible in that direction , and thus the roughness exponent is @xmath372 the fixed - point function is then periodic , and can in the interval @xmath373 $ ] be expanded in a taylor - series in @xmath374 .
even more , the ansatz @xmath375 allows to satisfy the fixed - point equation ( [ rgdisorder ] ) to order @xmath376 and will presumably work to all orders . for a more general case of this see ref .
@xcite . to gain insight into the more general case ,
let us write the fixed point for ( [ rgdisorder ] ) with arbitrary @xmath203 : @xmath377 one can see on this solution that @xmath238 is the only value which avoids the appearance at two loops of the supercusp , i.e. a cusp in the potential correlator @xmath266 rather than in the force correlator @xmath378 .
the same discussion can be made on the the flow equation of @xmath379 by taking two derivatives of ( [ rgdisorder ] ) .
one finds that there is a priori an unstable direction corresponding to a uniform shift in @xmath380 . while this is natural in e.g. depinning ,
it is here forbidden by the potential nature of the problem which requires @xmath381 since in a potential environment , the integral of the force over one period must vanish .
this is indeed satisfied for the fixed point for @xmath382 @xmath383 only if @xmath238 : @xmath384 the values for depinning are obtained by setting @xmath385 : in that case , the problem becomes non - potential at large scales .
this fixed point implies for the amplitude of the zero mode in presence of an harmonic well , defined in ( [ lf109 ] ) , using ( [ ctd ] ) : @xmath386 in the other limit @xmath274 one obtains the amplitude ( using @xmath387 from section [ sec : correlations ] ) : @xmath388 note that we prove in section ( [ sec : correlations ] ) that this amplitude is _ independent _ of large scale boundary conditions , and is thus identical for e.g. periodic boundary conditions and in presence of a mass . as can be seen from ( [ ratfin ] )
this is a consequence of @xmath19 being zero .
this can be compared to the gvm method @xcite : @xmath389 with coefficients surprisingly close to the @xmath5-expansion .
it is interesting to compare predictions in @xmath13 .
we recall that we are studying a problem where the period is unity , the general case being obtained by a trivial rescaling in @xmath30 . since ( [ s1 ] ) has a poor behavior ( and so does ( [ lf138 ] ) which resums into ( [ lf136 ] ) ) , it is better to use instead ( [ quasi ] ) .
it was indeed noted in @xcite that the improved 1-loop prediction @xmath390 obtained by setting @xmath391 and ignoring the @xmath392 term in ( [ quasi ] ) yields a value rather close to the prediction of the gvm : @xmath393 including the 2-loop @xmath392 term now gives @xmath394 and @xmath395 for the two pades respectively .
this type of extrapolation makes the gvm and frg predictions get closer when including the 2-loop corrections . on the other hand , comparison of ( [ s1 ] ) and ( [ lf138 ] )
suggests that @xmath396 .
this is in reasonable agreement with the numerical results of middleton et al .
they obtained good evidence for the existence of the bragg glass ( i.e. its stability with respect to topological defects predicted in @xcite ) .
they measure directly the correlation ( [ fdmassless ] ) and obtain strong evidence for the behavior ( [ quasi ] ) ( as well as the correct correction to scaling behavior ) with @xmath397 ( their amplitude @xmath170 is twice our @xmath398 ) which lies in between the gvm and the 1-loop frg .
( more precisely two different discretizations gave @xmath399 and @xmath400 ) .
another interesting observable is the slow growth of displacements characteristic of the bragg glass : @xmath401 at large @xmath8 . performing the momentum integral from ( [ fdmassless ] ) ,
one obtains : @xmath402 \label{amp}\end{aligned}\ ] ] if one expands each factor in @xmath5 it yields : @xmath403 for comparison , the gvm gives @xmath404 here extrapolation directly setting @xmath391 in ( [ adeps ] ) looks possible , and yields @xmath405 to one loop increasing to @xmath406 to two loop . on the other hand , setting @xmath391 in ( [ amp ] ) yields instead @xmath407 to one loop decreasing to @xmath408 at two loops .
the gvm gives the result @xmath409 .
another interesting observable is : @xmath410 where @xmath128 is the linear system size . in ref .
@xcite it was assumed that @xmath411 thus in @xmath13 , @xmath412 yielding a value of @xmath398 consistent with the direct measurement of this quantity .. if @xmath30 were a _ gaussian _
variable with the same 2-point correlator , the exact result for extrema of logarithmically correlated gaussian variables predicted in @xcite yields : @xmath413 where @xmath414 , @xmath415 , @xmath416 and @xmath417 a fluctuating constant of order @xmath418 ( in their notations @xmath180 is our @xmath419 ) . inserting the value obtained numerically in @xcite for @xmath180 yields @xmath420 and
this is in reasonable agreement with the measured values @xmath422 quoted in @xcite . since deviations from gaussian
are not expected to be large , this agreement could probably be improved by using the above form of finite size corrections ( as was done in @xcite for a one dimensional version where much larger sizes had to be considered ) rather than the simpler form used in @xcite . ]
this was also done in @xcite where it was deduced from a measurement of @xmath419 that @xmath423 . , which , within a gaussian approximation for the distribution of @xmath30 , yields the decay @xmath424 , with the bragg glass exponent @xmath425 . ]
although this is a reasonable approximation , it is not exact .
indeed the quantity @xmath419 , contrary to @xmath398 , depends on the ( large scale ) boundary conditions .
it is of course universal , since it does not depend on small scale details .
its value can be computed e.g. for periodic boundary conditions and pinned zero mode , and depends on the whole finite size scaling function ( [ finitesizeg ] ) computed in section [ sec : correlations ] : @xmath426 as shown recently , @xmath427 fluctuates from sample to sample and the full distribution @xmath428 averaged over disorder realizations was computed for the depinning problem @xcite .
let us now consider the case of long range elasticity .
there are physical systems where the elastic energy does not scale with the square of the wave - vector @xmath193 as @xmath429 but as @xmath430 .
in this situation , the upper critical dimension is @xmath431 and we define : @xmath432 the most interesting case , a priori relevant to model a contact line is @xmath433 , thus @xmath434 . for calculational convenience , we choose the elastic energy to be @xmath435 this changes the free correlation to : @xmath436 the energy exponent in that case is : @xmath437 the changes are very similar to the case of ref .
@xcite so we summarize them here only briefly . the @xmath0-function is still given by ( [ rgdisorderunrescaled ] ) but with the integrals replaced by : @xmath438 and thus the @xmath0-function is given by ( [ rgdisorder ] ) with : @xmath439 ( see appendix f of ref .
and of course the relation ( [ resc ] ) between @xmath86 and @xmath440 is identical except that @xmath248 must be replaced by @xmath441 .
since @xmath442 is finite , the @xmath0-function is finite ; this is of course necessary for the theory to be renormalizable . for the cases of interest @xmath433 and @xmath443 , we find @xmath444 the exponent @xmath19 ( as a function of @xmath5 ) and the fixed point function is thus changed only at two loops .
let us now give the results in the cases of interest : the solution of ( [ rgdisorder ] ) with @xmath445 can be written , to second order in @xmath5 as : @xmath446 since the equation ( [ lf49 ] ) for @xmath447 is linear .
thus one has for any @xmath448 : @xmath449 for the case of most interest @xmath450 , @xmath451 one finds : @xmath452 and @xmath453
. it would thus be interesting to perform numerical simulations in @xmath15 for the directed polymer with lr elasticity .
this would be another non trivial test of the 2-loop corrections .
the 1-loop prediction is @xmath454 , significantly smaller than the roughness for sr elasticity @xmath53 .
the naive 2-loop result is ( setting @xmath391 ) , @xmath455 .
error bars are estimated by half the difference between the 1-loop and 2-loop results .
note that the bound @xmath304 implies @xmath456 in @xmath15 , already rather close to the 2-loop result .
the exponent is still @xmath457 and was indeed measured in experiments on an equilibrium contact line @xcite .
it would be of interest to measure the universal distributions there , such as the one defined in @xcite .
the fixed point function is given by ( [ k1 ] ) and ( [ fy ] ) upon replacing @xmath458 .
the amplitude of the zero mode in a well @xmath398 is now given by : @xmath459 and the amplitude of the massless propagator @xmath460 where @xmath461 is given in ( [ lf193 ] ) setting @xmath462 .
the fixed point becomes : @xmath463 for the periodic case , the universal amplitude reads : @xmath464 and @xmath465 setting @xmath466 in ( [ lf193 ] ) yields @xmath467 using @xmath468 , this gives @xmath469 which in the case of @xmath433 takes the simple form @xmath470
as we have seen above ambiguities arise in computing the effective action at the level of 2-loop diagrams if one uses a non - analytic action .
one can see that these arise even at the 1-loop level for correlations ( see below section [ sec : correlations ] ) . to resolve this issue ,
our strategy has been to use physics as a guide and require the theory to be renormalizable , potential and without supercusp .
this pointed to a specific assignment of values to the `` anomalous '' graphs .
the physical properties of the ensuing theory , studied in the previous section , were found to be quite reasonable . of course , one would like to have a better , more detailed justification of the used `` prescription '' .
although we do not know at present of a derivation of this theory from first principles , we have developed a set of observations and a number of rather natural and compelling `` rules '' which all lead to the same theory .
we describe below our successful efforts in that direction as well as some unsuccessful ones , which illustrate the difficulty of the problem .
a number of approaches can be explored to lift the ambiguities in the non - analytic theory .
we here give a list ; some of the methods will be detailed in the forthcoming sections .
: at @xmath49 previous wilson 1-loop frg analysis@xcite found that the effective action _ remains analytic _ in a boundary layer @xmath471 .
however , since the rescaled temperature ( [ rescaled ] ) flows to zero as @xmath472 as @xmath278 ( temperature being formally irrelevant ) all ( even ) derivatives of @xmath27 higher than second grow unboundedly as @xmath61 , for instance @xmath473 ( in terms of the zero temperature fixed point function ) . on a _
qualitative _ level one can thus see how finite @xmath70 diagrams such as @xmath152 in fig .
[ alldiag2loop ] yielding @xmath474 can build up `` anomalous '' terms in the @xmath0-function , hence confirming what is found here@xcite .
however , correctly and quantitatively accounting for higher loops is a non - trivial problem as stronger blow - up in @xmath475 seem to arise .
in fact each new loop brings two derivatives and a propagator , hence an additional factor @xmath476 . despite some recent progress , a quantitative finite - temperature approach which would reproduce and justify
the present @xmath5 expansion has proved difficult@xcite . not only for technical reasons , as methods using exact rg where found to be appropriate , but also for physical reasons , as an extension to non - zero @xmath70 must also handle low - lying thermal excitations in the system ( e.g. droplets ) .
a theory from first principles at @xmath49 is thus presently not available and will not be further addressed here .
all other methods use a _ non - analytic action_. : exact rg methods directly at @xmath39 have been studied to one loop @xcite and two loops@xcite . although it does yield interesting insights into the way to handle ambiguities ( see below ) , and confirm the present results , it suffers from basically the same problems as described here . : in this approach one attempts a direct evaluation of non - analytic averages ( e.g. in fully saturated diagrams ) .
for instance , expanding at each vertex the disorder @xmath477 in powers of @xmath478 using the proper non - analytic taylor expansion : @xmath479 one can try to compute directly all averages in vertex functions and correlations .
after performing a few wick contractions one typically ends up with averages involving sign functions or delta functions
. these can be computed _ in principle _ using the free gaussian measure .
for instance , using formulae such as : @xmath480 although promising at first sight , the results are disappointing .
averages over the thermal measure involve many changes of signs which kill all interesting divergences indicating that some physics is missing .
the method , briefly described in appendix [ sec : nona ] is thus not developed further
. a dynamical version of this method which is similar in spirit @xcite , did work for depinning , although there it simply identified with another method used below , the background field ( which , for depinning is @xmath481 see below ) .
* 4 ) calculation of @xmath482 with excluded vertices and symmetrization * : a valid , general and useful observation ( not limited to this method ) is that if one uses the _ excluded vertex _
@xmath483 then all wick contractions can be performed _ without ambiguities_.
the excluded vertex is as good as the non - excluded one since one can always add a constant @xmath484 to the action of the model ( [ action ] ) . thus one can compute _ without any ambiguity _ the effective action @xmath482 for an `` off - diagonal '' field configuration @xmath485 since then no vertex is ever evaluated at @xmath18 .
the drawback is that one ends up with expressions containing terms such as @xmath486 which superficially looks like a three replica term , but due to the exclusions , may in fact contain a 2-replica part which can in principle be recovered from the above by adding appropriate diagonal terms , using that @xmath68-replica parts are properly defined as _ free _ replica sums e.g. from a cumulant expansion .
the 2-replica part of ( [ 3rep ] ) thus naively is @xmath487 and one is again faced with the problem of assigning a value to @xmath488 .
the calculation with excluded vertices thus yields a sum of @xmath68-replica terms with @xmath489 and to project them onto the needed 2-replica part , one may need to continue these expressions to coinciding arguments @xmath490 .
the symmetrization method attempts to do that in the most `` natural '' way . using the permutation symmetry over replicas and the hypothesis of no supercusp
yields a rather systematic method of continuation .
surprisingly , it fails to yield a renormalizable theory at two loops .
we identified some difference with methods which do work , but the precise reason for the failure in terms of continuity properties remains unclear .
it may thus be that there is a way to make this method work but we have not found it .
being interesting in spirit this method is reported in some details in appendix [ sec : symm ] .
if one renounces to the projection onto 2-replica terms one can , in a certain sense , obtain renormalizability properties .
this generates an infinite number of different replica sums and seems to be not promising , too .
it is described in appendix [ sec : finitet ] .
we now come to methods which were found to work , and which will be described in detail in the next section . in all of them
one performs the wick contractions in some given order ( the order hopefully does not matter ) and uses at each stage some properties .
the fact that one can order the wick contractions stems from the identity , which we recall , for any set of mutually correlated gaussian variables @xmath491 : @xmath492 under very little analyticity assumption for @xmath493 , which can even be a distribution . at each stage
one can either use excluded or non - excluded vertices as is found more convenient .
* 5 ) elimination of sloops : * we found another method , which seems rather compelling , to determine the 2-replica part of terms such as ( [ 3rep ] ) .
it starts , as the previous one , by computing ( unambiguously ) diagrams with the excluded vertices .
then instead of symmetrization , one uses identities derived from the fact that diagrams with free replica sums and which contain sloops can not appear in a @xmath39 theory and can thus be set to zero .
further contracting such diagrams generates a set of identities which , remarkably , is sufficient to obtain unambiguously the 2-replica projection without any further assumption .
it works very nicely and produces a renormalizable theory , as we have checked up to three loops . in some sense
, it uses in a non - trivial way the constraint that we are working with a true @xmath39 theory .
this method is detailed below .
* 6 ) background field method : * this method is similar to method number 3 except that the vertex @xmath27 at point @xmath8 is evaluated at the field @xmath494 , then expanded in @xmath495 , which then are contracted in some order .
this amounts to compute the effective action in presence of a uniform background field which satisfies ( [ diag ] ) .
thanks to this uniform background and upon some rather weak assumptions , the ambiguities seem to disappear .
the method is explained below . * 7 ) recursive construction : * an efficient method is to construct diagrams recursively .
the idea is to identify in a first step parts of the diagram , which can be computed without ambiguity .
this is in general the 1-loop chain - diagram ( [ 1loop ] ) . in a second step ,
one treats the already calculated sub - diagrams as effective vertices . in general , these vertices have the same analyticity properties , namely are derivable twice , and then have a cusp .
( compare @xmath496 with @xmath497 ) . by construction , this method ensures renormalizability , at least as long as there is only one possible path .
however it is not more general than the demand of renormalizability diagram by diagram , discussed below . *
8) renormalizability diagram by diagram : * in section iii we have used a _ global _ renormalizability requirement : the 1-loop repeated counter - term being non - ambiguous one could fix all ambiguities of the divergent 2-loop corrections .
however , as will be discussed in @xcite , this global constraint appears insufficient at three loops to fix all ambiguities .
fortunately , one notes that renormalizability even gives a stronger constraint , namely renormalizability _
diagram by diagram_. the idea goes back to formal proofs of perturbative renormalizability in field - theory , see e.g.@xcite .
these methods define a subtraction operator @xmath498 .
graphically it can be constructed by drawing a box around each sub - divergence , which leads to a `` forest '' or `` nest '' of sub - diagrams ( the counter - terms in the usual language ) , which have to be subtracted , rendering the diagram `` finite '' .
the advantage of this procedure is that it explicitly assigns all counter - terms to a given diagram , which finally yields a proof of perturbative renormalizability .
if we demand that this proof goes through for the _ functional _ renormalization group , the counter - terms must necessarily have the same functional dependence on @xmath27 as the diagram itself . in general ,
the counter - terms are less ambiguous , and this procedure can thus be used to lift ambiguities in the calculation of the diagram itself . by construction
this procedure is very similar to the recursive construction discussed under point 7 .
it has some limitations though .
indeed , if one applies this procedure to the 3-loop calculation , one finds that it renders unique all but one ambiguous diagram , namely @xmath499{./figures/3loopi}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/3loopi}{\hspace*{0.8ex}}}}\ , \ ] ] which has no subdivergence , thus there are no counter - terms , which could lift the ambiguities .
thus this diagram must be computed directly and we found that it can be obtained unambiguously by the sloop elimination method@xcite . * 9 ) reparametrization invariance : * from standard field theory , one knows that renormalization group functions are not unique , but depend on the renormalization scheme .
only critical exponents are unique .
this is reflected in the freedom to reparametrize the coupling constant @xmath188 according @xmath500 where @xmath501 is a smooth function , which has to be invertible in the domain of validity of the rg @xmath0-function .
here we have chosen a scheme , namely defining @xmath27 from the exact zero momentum effective action , using dimensional regularization , and a mass .
one could explore the freedom in performing reparametrization . in the functional rg framework , reparametrizations are also functional , of the form @xmath502 ( u)\ .\ ] ] of course the new function @xmath503 does not have the same meaning as @xmath27 .
perturbatively this reads @xmath504 where @xmath505 is a functional of @xmath86 . for consistency , one has to demand that @xmath506 has the same analyticity properties as @xmath86 , at least at the fixed point @xmath507 , i.e. @xmath508 should as @xmath86 be twice differentiable and then have a cusp .
a specifically useful candidate is the 1-loop counter - term @xmath509 .
one can convince oneself , that by choosing the correct amplitude , one can eliminate all contributions of class a , in favor of contributions of class b. details can be found in @xcite .
apart from methods 3 and 4 which did not work for reasons which remain to be better understood , methods 2,5,6,7,8,9 were all found to give consistent result , making us confident that the resulting theory is sufficiently constrained by general arguments ( such as renormalizability ) to be uniquely identified .
let us now turn to actual calculations using these methods .
let us redo the calculation of section [ sec : twoloop ] using _ excluded vertices_. from now on we use sometimes the short - hand notations @xmath510 whenever confusion is not possible .
the resulting diagrammatics looks very different from the usual unexcluded one .
when making all four wick contractions of the 2-loop diagrams a , b and c in fig .
[ f : twoloop ] between three unsplitted vertices one now excludes all diagrams with saturated vertices , but instead has to allow for more than two connected components and for sloops .
the splitted excluded diagrams corresponding to classes a , b and c are given in fig .
there is an additional multiplicative coefficient @xmath511 in the combinatorics for each pair of unsplitted vertices ( say @xmath512 and @xmath513 ) linked by an internal line where @xmath514 propagators link @xmath515 , @xmath516 link @xmath517 , @xmath518 link @xmath519 , @xmath516 link @xmath520 .
( this is equivalent to assigning a color to each propagator ) .
is a 1-replica term , @xmath0 , @xmath521 , @xmath125 and @xmath32 are improper three replica terms and @xmath203 is an improper four replica term.,title="fig : " ] let us denote by @xmath522 the 2-loop contribution of all diagrams of class a to the effective action .
one finds : @xmath523 i_a \label{deltaa}\end{aligned}\ ] ] coming respectively and in the same order from graphs @xmath448 , @xmath0 , @xmath521 , @xmath125+@xmath32 ( they are equal ) and @xmath203 in fig .
the only graph common to excluded and free - sum diagrammatics is @xmath448 which is graph @xmath180 of fig.[static-2-loop - do ] , since all the other graphs in fig.[static-2-loop - do ] have saturated vertices .
similarly , the graphs of class b give a total contribution : @xmath524 i_1 ^ 2 \nonumber \\ & & \label{deltab}\end{aligned}\ ] ] coming respectively and in the same order from graphs @xmath525 , @xmath526 , @xmath527 , @xmath528 in fig .
again , the only graph common to excluded and free - sum diagrammatics is @xmath525 which is graph @xmath188 of fig . [ banana ] , since all the other graphs in fig . [ banana ] have saturated vertices . the contribution @xmath529 of the diagrams of class c is given in appendix [ sec : classc ] .
note that adding a tadpole does not alter the structure of the summations in the excluded - replica formalism , since a tadpole can never identify indices on different vertices .
this indicates that class c does not contain a 2-replica contribution , but starts with a 3-replica contribution ( times @xmath70 ) .
this is explained in more details in appendix [ sec : sloopbc ] one can first check that when @xmath27 is _ analytic _ one recovers correctly the same result as ( [ d22 ] ) setting the last ( anomalous ) term to zero . adding and subtracting the excluded terms in ( [ deltaa ] ) to build free replica sums ( using @xmath182 in that case ) , or equivalently lifting all exclusions but replacing everywhere @xmath530 and then expanding and selecting the 2-replica part , one finds the contributions @xmath531 similarly in ( [ deltab ] ) one obtains @xmath532 we now want to perform the same projection for a non - analytic @xmath27 .
the idea of the method is very simple .
let us consider the 1-loop functional diagram ( a ) in fig .
[ f : sloops ] which contains a sloop .
it is a three replica term proportional to the temperature . in a @xmath39 theory
such a diagram should not appear , so it can be identically set to zero : @xmath533 it is multiplied by @xmath534 , which we have not written .
we will also omit global multiplicative numerical factors .
projecting such terms to zero at any stage of further contractions is very natural in our present calculation ( and also e.g. in the exact rg approach , where terms are constructed recursively and such forbidden terms must be projected out ) .
it is valid only when ( i ) the summations over replicas are free ( ii ) the term inside the sum is non - ambiguous .
these conditions are met for any diagram with sloops , provided the vertices have at most two derivatives .
( one can in fact start from vertices which either have no derivative or exactly two . )
let us illustrate the procedure on an example .
we want to contract @xmath535 with a third vertex @xmath86 at point @xmath536 , i.e. we first write the product : @xmath537 where implicitly here and in the following the vertices are at points @xmath538 in that order .
we will contract the third vertex twice , once with the first and once with the second , i.e. look at the term proportional to @xmath539 .
note that since we will contract each vertex , we are always allowed to introduce excluded sums ( clearly the diagonal terms @xmath540 , @xmath541 or @xmath542 give zero , since @xmath543 and its two lowest derivatives at @xmath540 are field independent constants ) . performing the first contraction ( i.e. inserting @xmath544 multiplied by the exclusion factors @xmath545 yields ( up to a global factor of @xmath165 ) : @xmath546 \equiv 0 \ .\ ] ] similarly , the second contraction then yields ( up to a global factor of @xmath166 ) : @xmath547 \equiv 0 \nonumber \ .
\\ & & \label{id4}\end{aligned}\ ] ] this non - trivial identity tells us that the sum of all terms ( or diagrams ) generated upon contractions of diagram ( a ) of fig . [
f : sloops ] ( i.e. the 1-loop sloop - diagram equivalent to term @xmath535 in ( [ id1 ] ) ) with other vertices , must vanish .
stated differently : a sloop , as well as the sum of all its descendents vanishes .
note that this is _ not _ true for each single term , but only for the sum .
a property that we request from a proper @xmath68-replica term is that upon one self contraction it gives a @xmath548-replica term .
it may also give @xmath70 times a @xmath68-replica term ( a sloop ) but this is zero at @xmath39 , so we can continue to contract .
thus we have generated several non - trivial projection identities .
the starting one is that the 2-replica part of ( [ id1 ] ) is zero , since ( [ id1 ] ) is a proper 3-replica term .
thus , ( [ id2 ] ) prior to the exclusions , is a legitimate 5-replica term , and its 4-replica part is zero .
upon contracting once we obtain that the 3-replica part of ( [ id3 ] ) is zero .
the final contraction tells us that the 2-replica part of ( [ id4 ] ) is zero .
this is what is meant by the symbol `` @xmath549 '' above and the last identity is the one we now use
. indeed compare ( [ id4 ] ) with ( [ deltaa ] ) .
one notices that all terms apart from the first in ( [ deltaa ] ) appear in ( [ id4 ] ) , and with the same relative coefficients , apart from the third one of ( [ deltaa ] ) .
thus one can use ( [ id4 ] ) to simplify ( [ deltaa ] ) : @xmath550 i_a \ .\ ] ] the function @xmath227 , which appears in the last term , is continuous at @xmath18 .
it is thus obvious how to rewrite this expression using free summations and extract the 2-replica part @xmath551 i_a\ , \end{aligned}\ ] ] which coincides with the contribution of diagrams a in ( [ d22 ] ) with @xmath238 .
we can write diagrammatically the subtraction that has been performed @xmath552{./figures / lh}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / lh}{\hspace*{0.8ex}}}}-{\settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / lhsloop2}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / lhsloop2}{\hspace*{0.8ex } } } } \ , \ ] ] where the loop with the dashed line represents the sub - diagram with the sloop , i.e. the term ( [ id1 ] ) ( with in fact the same global coefficient ) .
the idea is of course that subtracting sloops is allowed since they formally vanish .
there are other possible identities , which are descendants of other sloops .
for instance a triangular sloop gives , by a similar calculation : @xmath553{./figures / lhsloop3}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / lhsloop3}{\hspace*{0.8ex } } } } & = & r''(0 ) \sum_{a \neq b } ( r'''_{ab})^2 + \sum_{a \neq b , a \neq c } r''(0 ) r'''_{ab } r'''_{ac } \nonumber \\ & & + \sum_{a \neq b , b \neq c } r''_{bc } ( r'''_{ab})^2 + \sum_{a \neq c , b \neq c , c \neq d } r'''_{ac } r'''_{bc } r''_{cd } \nonumber \ .
\\ & & \end{aligned}\ ] ] this however does not prove useful to simplify @xmath554 .
since the above method generates a large number of identities , one can wonder whether they are all compatible .
we have checked a large number of examples ( see the 3-loop calculations in @xcite ) and found no contradictions , although we have not attempted a general proof .
the diagrams b and c are computed in appendix [ sec : sloopbc ] .
one finds by the same procedure @xmath555 confirming our earlier results in section [ s : classb ] and [ s : classc ] . in the background method
, one computes @xmath64 $ ] to two loops for a uniform background @xmath30 such that @xmath556 for any @xmath557 .
we start from @xmath558 ^ 3 \right>_{\mathrm{1pi}}\ ] ] taylor expand in @xmath559 , and contract all the @xmath126 fields keeping only 1pi - diagrams .
this is certainly a correct formula for the uniform ( i.e. zero momentum ) effective action .
then one needs the small @xmath47-expansion of derivatives of @xmath86 , i.e.([exp ] ) as well as @xmath560 let us start from : @xmath561 we expand in @xmath126 and of course in diagrams a one must handle terms involving @xmath488 and in diagrams b terms proportional to @xmath562 .
let us start with diagrams a , which come from the following term in the taylor expansion : @xmath563 here and in the following , we will drop all combinatorial factors .
note that the expectation - values vanish at coinciding replicas so there is no need to specify the values of @xmath564 at @xmath540 .
let us perform the first @xmath565 contraction @xmath566 if we now perform a second @xmath565 contraction there is a @xmath567 term which is a sloop and thus should be discarded .
the @xmath568 terms build saturated vertices .
however the corresponding expectation values contain @xmath569 which is reasonably set to zero .
thus the first two contractions have been performed with no ambiguity leading to @xmath570 this term is no more ambiguous .
expanding as in ( [ r ] ) the potentially ambiguous part is @xmath571 clearly free of any ambiguity .
it yields the result ( [ delta2ar ] ) .
the question arises , whether the result may depend on the order .
we found that when first contracting @xmath565 and @xmath572 , one reproduces the result ( [ delta2ar ] ) .
however when one first contracts @xmath565 and @xmath573 ( in any order ) one encounters a problem , if one wants to contract @xmath573 again .
the intermediate result after the first two contractions is @xmath574 the next contraction between @xmath565 contains one term with a single @xmath575 .
one would like to argue that this term can be set to 0 . following
this procedure however leads to problems .
we therefore adopt the rule , that whenever one arrives at a single @xmath575 , one has to stop , and search for a different path .
note that this equivalently applies to the recursive constructions method . in 2-loop order ,
one can always find a path , which is unambiguous .
it seems to fail at 3-loop order ; at least we have not yet been able to calculate @xmath576{./figures/3loopi}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/3loopi}{\hspace*{0.8ex}}}}\ ] ] using any other than the sloop elimination method .
whether some refinement of the background method can be constructed there is an open question . for diagrams of class b one
expands as @xmath577 again no need to attribute a value to @xmath578 for @xmath579 since the summand vanishes there .
contract @xmath565 : @xmath580 contracting @xmath573 one gets @xmath581 contracting next @xmath565 the danger is the term @xmath582 yielding a saturated vertex in the middle .
but , again if one takes @xmath583 then one gets unambiguously @xmath584 the rest is straightforward .
the backgound method thus seems to work properly at two loop order . in section [ sec : list ] we have stated that renormalization diagram by diagram gives a method to lift the ambiguity of a given diagram , as long as it has sufficient sub - divergences .
this method is inspired by formal proofs of perturbative renormalizability ; the reader may consult @xcite for more details .
the key - ingredient is the subtraction operator @xmath585 , which acts on the effective action , i.e. all terms generated in perturbation - theory , which contribute to the renormalized @xmath86 , and which subtracts the divergences at a scale @xmath586 . at 1-loop order ,
the renormalized disorder @xmath587 at scale @xmath98 is symbolically ( with @xmath588 the bare disorder ) @xmath589{./figures/1loop}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/1loop}{\hspace*{0.8ex } } } } + \ldots \right]\nonumber\ , \ ] ] where of course the integral @xmath590{./figures/1loop}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/1loop}{\hspace*{0.8ex}}}}\renewcommand{{0.35}}{{{0.35}}}}$ ] depends on @xmath98 .
the operator @xmath591 rewrites this as a function of the _ renormalized _ disorder @xmath592 at scale @xmath593 @xmath594{./figures/1loop}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/1loop}{\hspace*{0.8ex } } } } + \dotsb \right]\nonumber \\ & : = & r_{\mu } + ( r''_{\mu } ) ^{2}\left ( { \settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/1loop}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/1loop}{\hspace*{0.8ex } } } } - \fbox{{\settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/1loop}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/1loop}{\hspace*{0.8ex } } } } } + \dotsb \right)\qquad~\end{aligned}\ ] ] here , the boxed diagram is defined as @xmath595{./figures/1loop}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/1loop}{\hspace*{0.8ex } } } } } = \left.{\settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/1loop}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/1loop}{\hspace*{0.8ex } } } } \right|_\mu \ .\ ] ] the idea behind this construction is that at any order in perturbation theory , any observable in the renormalized theory can be written as perturbative expansion in the _ bare _ diagrams , to which one applies @xmath585 .
@xmath585 reorganizes the perturbative expansion in terms of the _ renormalized _ diagrams .
the action of @xmath585 is to subtract divergencies , which graphically is denoted by drawing a box around each divergent diagram or sub - diagram , and to repeat this procedure recursively inside each box .
the second line of ( [ lf94 ] ) is manifestly finite , since it contains the diagram at scale @xmath98 minus the diagram at scale @xmath593 .
this is easily interpreted as the 1-loop contribution to the @xmath0-function .
the power of this method is not revealed before 2-loop order .
let us give the contribution from the hat - diagram ( class a ) : @xmath596{./figures / lh}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / lh}{\hspace*{0.8ex } } } } r''_{0 } ( r'''_{0})^{2}\ ] ] using @xmath585 , this is rewritten as @xmath597{./figures / lh}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / lh}{\hspace*{0.8ex } } } } \right ] = r '' ( r''')^{2 } \left[{\settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / lh}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / lh}{\hspace*{0.8ex}}}}-\,\fbox{{\settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / lh}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / lh}{\hspace*{0.8ex}}}}}\,- { \settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / lhsub}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / lhsub}{\hspace*{0.8ex}}}}+\,\fbox{{\settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / lhsub}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / lhsub}{\hspace*{0.8ex}}}}}\;\right]\end{aligned}\ ] ] note that not only the global divergence is subtracted , but also the sub - divergence in the bottom loop ; and finally the divergence which remains , after having subtracted the latter ( last term ) . note the factor of @xmath598 in front of the last diagram , which comes with the two ( nested ) boxes .
let us halt the discussion of the formal subtraction - operator at this point , and not prove that the procedure renders all expectation values finite ; this task is beyond the scope of this article , event hough it is not difficult to prove e.g. along the lines of @xcite , once the question of the ambiguities of a diagram are settled .
however let us discuss , what the subtraction procedure can contribute to the clarification of the amgibuities
. in standard field - theory , the main problem to handle is the cancellation of divergences , whereas the combinatorics of the vertices is usually straightforward .
this means that the sum of the integrals , represented by the diagrams in the brackets on the r.h.s . of ( [ lf256 ] )
are finite .
this ensures of course renormalizability , subject to the condition that all diagrams have the same functional dependence on @xmath86 . here
, the factor @xmath599 should more completely read @xmath600 for the first term , there was no problem .
however , we have seen that the last term was more difficult to obtain .
demand renormalizability diagram by diagram _ , all diagrams have to give the _ same _ factor ( [ rd1 ] ) .
thus , if _ at least one of them _ can be calculated without ambiguity , we have an unambiguous procedure to calculate _ all of them_. we now demonstrate that @xmath590{./figures / lhsub}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / lhsub}{\hspace*{0.8ex}}}}\renewcommand{{0.35}}{{{0.35}}}}$ ] is unambiguous . to this aim , we detail on the subtraction operator @xmath601 , whose action is represented by the box .
this box tells us to calculate the divergent part of the sub - diagram in the box , and to replace everything in the box by the counter - term , which here is @xmath602{./figures/1loop}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/1loop}{\hspace*{0.8ex}}}}}= i_{1}^{\mu } \left ( r '' ( u)^{2}-2 r '' ( u)r '' ( 0 ) \right)\ .\ ] ] in a second step , one has to calculate the remaining diagram , which is obtained by treating the box as a point , i.e. as a local vertex .
the idea is of course , that the sub - divergence comes from parts of the integral , where the distances in the box are _ much _ smaller than all remaining distances , such that this replacement is justified .
graphically this can be written as @xmath603{./figures / lhsub}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / lhsub}{\hspace*{0.8ex}}}}= i_{1}^{\mu } ~ \times { \settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/1loopwithctins}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/1loopwithctins}{\hspace*{0.8ex}}}}\ ] ] remains to calculate the rightmost term , i.e. to calculate the 1-loop diagram , from one vertex @xmath496 and a second vertex @xmath604 .
the result is in straightforward generalization of ( [ 1loop ] ) @xmath605{./figures/1loopwithctins}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/1loopwithctins}{\hspace*{0.8ex}}}}= { i_{1 } } \big [ r '' ( u ) v '' ( u)- r '' ( 0)v '' ( u)- r '' ( 0)v '' ( 0 ) \big]\ ] ] we need @xmath606 the omitted terms are proportional to @xmath607 , and contribute to class b. we could have avoided their appearance altogether , but this would have rendered the notation unnecessarily heavy .
the term which contributes to ( [ rd4 ] ) is @xmath608 .
it has the same analyticity properties as @xmath609 , especially can unambiguously be continued to @xmath18 , i.e. @xmath610 .
expression ( [ rd4 ] ) becomes @xmath611{./figures/1loopwithctins}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/1loopwithctins}{\hspace*{0.8ex}}}}\!\ !
= { i_{1 } } \big [ r '' ( u ) r ' '' ( u)^{2}- r '' ( 0)r ' '' ( u)^{2}- r '' ( u)r ' '' ( 0^{+})^{2 } \big]\ ] ] without any ambiguity . to summarize : using ideas of perturbative renormalizability diagram by diagram
, we have been able to compute unambiguously one of the terms in ( [ lf256 ] ) and can use this information to make the functional dependence of the whole expression unambiguous . if we were to chose any other prescription , a proof of perturbative renormalizability is doomed to fail , a scenario which we vehemently reject .
this method is very similar in spirit to the one of section [ rd7 ] .
there we had first calculated a subdiagram , and then treated the result as a new effective vertex .
this procedure can be made a prescription , which insures renormalizability and potentiality , since the 1-loop diagram insures the latter . only at 3-loop order
appears a new diagram , ( [ lf174 ] ) , which can not be handled that way , but the procedure , which is otherwise very economic , can handle again most diagrams at 4-loop order , using the new 3-loop diagram ( [ lf174 ] ) .
here we address the issue of the calculation of correlation functions .
we note that it has not been examined in detail in previous works on the @xmath39 frg .
usually correlations are obtained from tree diagrams using the proper or renormalized vertices from the polynomial expansion of the effective action .
thus in a standard theory one could check at this stage that correlation functions are rendered finite by the above counter - terms , compute them and obtain a universal answer . in a more conventional theory that would be more or less automatic here , as we point out
, it is not so easy . indeed , as we show below if one tries to compute even the simplest 2-point correlation at non - zero momentum one finds ambiguities already at one loop .
this is because the effective action ( the counter - term ) is non - analytic . again
the requirement of renormalizability and independence of short - scale details guide us toward a proper definition of the correlation functions that we can compute .
interestingly , this definition is very similar as the one obtained from an exact solution in the large @xmath6 limit in @xcite .
let us illustrate this on the 2-point function , and at the same time , derive the ( finite size ) scaling function for any elasticity ( not done in @xcite ) for massive and finite size scheme .
we want to compute the 2-point correlation function at @xmath39 . in fourier - representation
it is given by ( [ lf107 ] ) with : @xmath612 in terms of the quadratic part of the effective action , which reads at any @xmath70 @xmath613 i.e. by construction here @xmath168 gives the exact off - diagonal element of quadratic part of the effective action .
inverting the replica matrix gives the relation , exact to all orders : @xmath614 @xmath27 is exactly the function entering the @xmath243-function ( in the rescaled form @xmath615 ) . in the second line
we have inserted the fixed point form , which thus gives exactly the @xmath120 correlations in the small @xmath98 limit ( i.e. up to subdominant terms in @xmath616 ) which are bounded because of the small confining mass .
we now compute @xmath617 for arbitrary but small wave vector @xmath193 , and to one loop , i.e. to next order in @xmath5 .
one expects the scaling form ( [ lf108 ] ) and that the scaling function is independent of the short - scale uv details ( i.e. universal ) , if the theory is renormalizable .
it satisfies @xmath618 and , from scaling should satisfy @xmath619 at large @xmath536 . in @xmath38 one has @xmath620 and we want to obtain the scaling function to the next order , i.e. identify @xmath180 in @xmath621 .
let us use straight perturbation theory with @xmath204 , defined as in section ( [ sec : rgdis ] ) , including the 1-loop diagrams .
this amounts to attach two external legs to the 1-loop diagrams in fig .
[ oneloop ] , and use a non - analytic perturbatively from @xmath86 to a given order , then @xmath204 is not the original bare action ( which is analytic ) ; thus there is no contradiction in @xmath204 being non analytic . in a sense
, introducing @xmath204 is just a trick commonly used in field theory to express a closed equation for the flow of @xmath86 to the same order .
the ( perturbative ) exact rg method introduced in @xcite does that automatically without the need to introduce @xmath204 . ]
our result is : @xmath622 there is however an ambiguity in this calculation , i.e. again it is not obvious , a priori , how to interpret the @xmath623 which appear .
if one computes the one loop correction using ( [ exp ] ) , one must evaluate : @xmath624 there is no ambiguity , as the interaction term is analytic to this order . then performing the average amounts to take two derivatives @xmath625 of @xmath626 which
, to this order , is analytic . in this case
this is exactly the same calculation as for the repeated 1-loop counter - term .
however , the full expression ( [ lf186 ] ) integrated over @xmath627 is , itself , ambiguous .
interestingly , this simple ambiguity already to one loop has never been discussed previously .
let us first show that _ renormalizability _ fixes the form to be the one written in ( [ cr0 ] ) .
indeed , let us re - express ( [ lf185 ] ) in terms of the renormalized dimensionless disorder in ( [ do ] ) and ( [ d21 ] ) @xmath628 as discussed in section ( [ sec : rgdis ] ) , no ambiguity arises when taking two derivatives of ( [ do ] ) at @xmath323 , i.e. the 1-loop counter - term is unambiguous .
this gives @xmath629 { \nonumber}\end{aligned}\ ] ] thus the substitution ( [ subst ] ) acts as a counter - term which exactly subtracts the divergence as it should .
the result is finite , as required by renormalizability , only with the above choice ( [ cr0 ] )
. stated differently , the @xmath120 calculation of ( [ cr0 ] ) fixes the ambiguity .
we show below that the methods described in the previous section also allow to obtain this result unambiguously . before that , let us pursue the calculation of the scaling function . upon using ( [ resc ] ) and the fixed point equation ,
we obtain : @xmath630\ .\ ] ] apart from the dependence on @xmath19 , the calculation of the scaling function is very similar to the one given in @xcite .
we perform here a more general calculation which also contains the case of elasticity of arbitrary range @xmath631 and expand in @xmath632 . using that , in that case @xmath633^{-\epsilon/2 } \ .\qquad\end{aligned}\ ] ] defining the 1-loop value of @xmath634 , we obtain , to the same accuracy , the scaling function in the form ( @xmath635 ) : @xmath636 \right\ } \nonumber \\ & = & \frac1{(1 + z^2)^{\alpha } } \left\ { 1 + \frac{\epsilon}{2 } ( 1 - 2 \zeta_1 ) \frac{\gamma ( \alpha)}{\gamma ( \frac{\alpha}{2})^{2 } } \int_0^{1 } { { \mathrm{d}}}s\ , [ s ( 1-s ) ] ^{\frac{\alpha}{2}-1 } \ln ( 1 + s ( 1-s ) z^2 ) \right\ } \nonumber \\\end{aligned}\ ] ] ( we have used the variable transformation @xmath637 ) . to obtain @xmath180
, we need the large @xmath536 behavior of the scaling function : @xmath638 \bigg\}\ ] ] we want to match at large @xmath536 : @xmath639\ .\qquad \label{equiv}\end{aligned}\ ] ] the above result yields @xmath640\nonumber \\ & = & \left\ { \begin{array}{lcc } - 2 ( 1 - 2 \zeta_1 ) \ln 2 & ~\mbox{for}~&\alpha = 1\\ - ( 1 - 2 \zeta_1 ) & ~\mbox{for}~&\alpha = 2 \end{array } \right.\end{aligned}\ ] ] let us now present two additional methods to lift the ambiguity in the 1-loop correction to the 2-point function and recover ( [ lf185 ] )
. in the background method of section [ sec : background ] one performs this calculation in presence of a background field , i.e. considering that the field @xmath641 has a uniform background expectation value : @xmath642 with @xmath643 for all @xmath557 , and contracting the @xmath644
. then at @xmath39 the sign of any @xmath178 is determined , and the above ambiguity in ( [ lf186 ] ) is lifted ( contracting further the @xmath644 yields extra factors of @xmath70 and thus is not needed here ) . using the background method is physically natural as it amounts to compute correlations by adding a small external field which splits the degeneracies between ground states whenever they occur , as was also found in @xcite . on the other hand ,
performing the calculation in the absence of a background field , in perturbation theory directly of the non - analytic action yields a different result , detailed in appendix [ sec : nona ] , which appears to be inconsistent .
it presumably only captures correlations within a single well .
the second method is sloop elimination .
we want to compute contractions of : @xmath645 where the two disorder are at points @xmath536 and @xmath646 respectively .
let us restrict to the part proportional to @xmath647 which gives the @xmath193 dependent part of the two point function .
since @xmath58 is fixed , we need to extract the `` 0-replica part '' of the expression after contractions ( which will necessarily involve excluded vertices ) . starting by contracting twice the two @xmath86 s we get @xmath648\end{aligned}\ ] ] subtracting the sloop @xmath535 from ( [ id1 ] ) gives ( up to terms which do not depend on both @xmath646 and @xmath536 , and which thus disappear after the two remaining contractions ) @xmath649 contracting the external @xmath30 s with ( [ u2 ] )
we obtain ( restoring the correlation - functions ) @xmath650 the excluded sum can be rewritten as the sum minus the term with coinciding indices .
only the latter is a 0-replica term , which gives the desired result : @xmath651 this result can also be obtained writing directly the graphs with excluded vertices and eliminating the descendants of the sloop .
the frg method described here can also be applied to a system of finite size , with e.g. periodic boundary conditions @xmath652 , and zero mass , which are of interest for numerical simulations .
the momentum integrals in all diagrams are then replaced by discrete sums with @xmath281 , @xmath282 .
one must however be careful in specifying the mode @xmath120 , i.e.@xmath653 .
the simplest choice is to constrain @xmath654 in each disorder configuration , which we do for now . since the zero mode is forbidden to fluctuate sums over momentum in each internal line exclude @xmath120 .
one then finds that the 2-loop frg equation remains identical to ( [ rgdisorder ] ) , the only changes being that 1 .
@xmath655 has to be replaced by @xmath656 .
2 . @xmath657 in the definition of the rescaled disorder .
the 1-loop integral @xmath658 entering into the definition of the rescaled disorder has to be replaced by its homologue for periodic boundary conditions : @xmath659 used below . here and
below we use a prime to distinguish the different ir schemes .
as we have seen @xmath250 is , to dominant order , independent of the ir cutoff procedure .
thus we can now compute the 2-point function . following the same procedure as above ,
we find : @xmath660\ \end{aligned}\ ] ] with @xmath661 , and , for @xmath281 : @xmath662 thus one finds the finite size scaling function ( defined in ( [ finitesizeg ] ) ) @xmath663\ \end{aligned}\ ] ] as function of @xmath664 .
the asymptotic behaviour is @xmath665 \nonumber \\ & \qquad \stackrel{q \to \infty } { -\!\!\!\longrightarrow } \left [ 1 + b ' \epsilon + ( \epsilon - 2 \zeta ) \ln(q l ) \right ] \ .\end{aligned}\ ] ] which defines @xmath666 .
the corresponding equation ( [ equiv ] ) , when regularizing with a mass , holds . taking the difference between the two equations yields @xmath667\ ] ] where @xmath668 . to leading order in @xmath669 , @xmath670 , such that this difference takes the simpler form @xmath671 + \left[\tilde i ( 0){-}\tilde i ' ( 0 ) \right]\!\bigg\}\nonumber \\ & & + o ( \epsilon^{2})\label{bmbpsimp}\end{aligned}\ ] ] now observe that for large @xmath193 the first integral can be bounded by @xmath672 which is obtained by estimating the maximal difference of integral and discrete sum in each cell ( defined by the discreteness of the sum ) , and then integrating .
the difference @xmath673 is finite and can be evaluated in @xmath38 dimensions .
we need the following formulae @xmath674 where @xmath675 is the elliptic @xmath676-function .
this allows to write sum and integral as @xmath677 \qquad \\
\label{sumint2 } \int { { \mathrm{d}}}^{4 } n \frac{1}{(n^{2}+1)^{2 } } & = & \int_{0}^{\infty}{{\mathrm{d}}}s\ , s\,\frac{\pi^{2}}{s^{2 } } \,{{\mathrm{e}}}^{-s}\end{aligned}\ ] ] the difference in question is integrated numerically : @xmath678\nonumber \\ & = & -\frac{14.5019}{(2\pi)^{4 } } = -0.00930479\ .\end{aligned}\ ] ] we thus arrive at ( @xmath679 ) @xmath680 since the frg equation and fixed point value @xmath681 is universal to two loops , the final result for the amplitude ratio between periodic and massive boundary conditions is @xmath682 let us now show how to compute higher correlation - functions with no ambiguity using the sloop - method . let us illustrate the method on e.g. the 4-point function : @xmath683 the following class of diagrams contributes : @xmath684{0mm}{16mm}^{x , a}_{w , a}{\settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/4pointc}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/4pointc}{\hspace*{0.8ex } } } } \rule[-7.2mm]{0mm}{16mm}^{y , a}_{z , a}\ ] ] an arrow indicates contraction towards an external field , with position and replica - index as indicated .
the combinatorial factors are : @xmath685 from the 4 @xmath86 s .
@xmath686 , the prefactor of the @xmath86 s .
@xmath687 the possibilities to connect the @xmath30 s to the @xmath86 s . 3 for the ways to make the loop of @xmath86 s .
when contracting first the @xmath30 s , there is another @xmath688 for the possibilities , to attach the @xmath30 s to the two replicas of @xmath86 .
therefore only the factor of 3 remains , which is the combinatorial factor for ordering 4 points on an unoriented ring .
we start by contracting the four @xmath30 s with four @xmath86 s , schematically : @xmath689{0mm}{25.5mm}^{x , a}_{w , a}{\settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/4pointa}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/4pointa}{\hspace*{0.8ex } } } } \rule[-11.9mm]{0mm}{25.5mm}^{y , a}_{z , a}\ ] ] and then we perform the four other contractions .
since exclusions at each vertex can be introduced early on , the number of possibilities is not too high and one easily obtains : @xmath690 where all terms have to be summed over with excluded replicas at each vertices . due to the factors of @xmath691 with an odd power , it is not trivial how to project this expression onto the space of 0-replica terms to yield the desired expectation value ( as in the previous section @xmath58 is fixed and thus no free replica sum should remain in the final result ) . to perform this projection we will first simplify the above expression using sloops .
there are a number of possible sloops which can be subtracted .
the first one is obtained by starting from @xmath692{./figures/4pointd}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/4pointd}{\hspace*{0.8ex}}}}\ .\ ] ] it reads @xmath693 the next sloop is @xmath694{./figures/4pointe}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/4pointe}{\hspace*{0.8ex}}}}\ , \ ] ] generating @xmath695 the last needed sloop is @xmath696{./figures/4pointf}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/4pointf}{\hspace*{0.8ex}}}}\ , \ ] ] leading to @xmath697 the simplest combination out of @xmath153 , @xmath698 , @xmath699 and @xmath700 is @xmath701 this expression is unambiguous because only squares of @xmath702 appear and it is easily projected onto the 0-replica part @xmath703 e.g. one can replace @xmath704 in ( [ fp11 ] ) and use free summations .
other possible contributions are given on figure [ other4point ] .
however none of these diagrams contributes .
the reason is that they are all descendents of a sloop .
we start by noting that @xmath705{0mm}{9.1mm}^{x , a}_{w , a}{\settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/4pointg}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/4pointg}{\hspace*{0.8ex}}}}\sim \sum_{b}r''_{ab}\ ] ] is a true 1-replica term , i.e. a sloop . when constructing a diagram on figure [ other4point ] , _ each _ of the terms in the excluded replica formalism is proportional to ( [ fp12 ] ) , thus descendant of a sloop .
this means , that to any order in perturbation theory , at @xmath39 , no diagram contributes to a connected expectation value ( of a single replica ) , which has two lines parting from one @xmath86 towards external points .
thus the leading contribution in @xmath86 to the connected 4-point function , as determined by the sloop method is the 1-loop diagram @xmath706 if one expressed this result in terms of the force correlator @xmath707 we thus find that this expression is formally _ identical _ to the one that we obtained for the same four point function at the @xmath39 quasistatic depinning threshold ( equation 5.4 in @xcite ) .
this is quite remarkable given that the method of calculation there , i.e. via the non - analytic dynamical field theory , is very different .
of course the two physical situations are different and here one must insert the fixed point value for @xmath708 from the statics frg fixed point , while in the depinning calculation @xmath709 takes a different value at the fixed point . in both problems
the connected four point function starts at order @xmath710 . however in some cases
the difference appears only to the next order in @xmath5 .
for instance , we can conclude that the results of @xcite still hold here for the _ static random field _ to the lowest order in @xmath5 at which they were computed there ( of course one expects differences at next order in @xmath5 ) . on the other hand .
for the static random bond case , the result for the connected four point function will be different from depinning even at leading order in @xmath5 .
it can easily be obtained from the above formula following the lines of @xcite .
in this article we constructed the field theory for the statics of disordered elastic systems .
it is based on the functional renormalization group , for which we checked explicitly renormalizability up to two loops .
this continuum field theory is novel and highly unconventional : not only is the coupling constant a function , but more importantly this function , and the resulting effective action of the theory , are non - analytic at zero temperature , which requires a non - trivial extension of the usual diagrammatic formulation . in a first stage
, we showed that 2-loop diagrams , and _ in some cases even 1-loop diagrams _ , are at first sight ambiguous at @xmath39 .
left unaddressed , this finding by itself puts into question all previous studies of the problem .
indeed , nowhere in the literature the problem was adressed that even the 1-loop corrections to the most basic object in the theory , the 2-point function , are naively ambiguous in the @xmath39 theory . since the problem is controlled by a zero - temperature fixed point there is no way to avoid this issue .
an often invoked criticism states that the problems are due to the limit of @xmath711 replicas .
we would like to point out that even though we use replicas , we use them only as a tool in perturbative calculations , which could equally well be performed using supersymmetry , or at a much heavier cost , using disorder averaged graphs .
so replicas are certainly not at the root of any of the difficulties . instead
the latter originate from the physics of the system , i.e. the occasional occurence of quasi - degenerate minima , resulting in ambiguities sensible to the preparation of the system .
how to deal with this problem within a continuum field theory is an outstanding issue , and any progress in that direction is likely to shed light on other sectors of the theory of disordered systems and glasses .
the method we have propose to lift the apparent ambiguities is based on two constraints : ( a ) that the theory be renormalizable , i.e. yield universal predictions , and ( b ) that it respects the potentiality of the problem , i.e. the fact that all forces are derivatives of a potential . each of these physical requirements is sufficient to obtain the @xmath243-function at 2-loop order , and the 2-point function and roughness exponent to second order in @xmath5 .
next , we have proposed several more general , more powerful and mutually consistent methods to deal with these ambiguous graphs , which work even to higher number of loops and allow to compute correlation functions with more than two points .
we were then able to calculate from our theory the roughness exponents , as well as some universal amplitudes , for several universality classes to order @xmath4 . in all cases ,
the predictions improve the agreement with existing numerical and exact results , as compared to previous 1-loop treatments .
we also clarified the situation concerning the universality ( precise dependence on boundary conditions , independence on small scale details ) of various quantities .
another remarkable finding is that the 1-loop contribution to the 4-point function is formally identical to the one obtained via the dynamical calculation at depinning .
this hints to a general property that _ all _ 1-loop diagrams are undistinguishable in the statics and at depinning .
it would be extremely interesting to perform higher precision numerical simulations of the statics , and to determine not only exponents but universal amplitudes and scaling functions to test the predictions of our theory .
we strongly encourage such studies .
thus in this paper we have proposed an answer to the highly non - trivial issue of constructing a renormalizable field theory for disordered elastic systems .
contrarily to the closely related field theory of depinning , which we were able to build from first principles , we have not yet found a first principle derivation of the theory for the statics .
however , we have found that the theory is so highly constrained , and the results so encouraging , that we strongly believe that our construction of the field is unique .
it is after all , often the case in physics that the proper field theory is first identified by recurrence to higher physical principles such as renormalizability or symmetries , as is exemplified by the ginsburg - landau theory of superconductivity , for which only later a microscopic derivation was found , or gauge theories in particle physics .
the first observation is that one expects ( if decomposition in @xmath68-replica terms is to mean anything ) that one can write the ( local disorder part of the ) effective action as a sum over well defined @xmath68-replica terms in the form : @xmath712 & = & \sum_{p } \frac{1}{t^p } \gamma_p[u ] \nonumber \\ & = & \sum_{p } \frac{1}{p ! t^p } \sum_{a_1, ... a_p } f^{(p)}(u_{a_1}, .... u_{a_p)}\ , \qquad \label{lf195}\end{aligned}\ ] ] where the functions @xmath713 have full permutation symmetry .
the idea of the symmetrization method is that we also expect , even at @xmath39 that these functions @xmath713 should be _ continuous _ in their arguments when a number of them coincide .
this seems to be a rather weak and natural assumption .
physically these functions can be interpreted as the @xmath68-th connected cumulants of a renormalized disorder , i.e. a random potential @xmath714 in each environment .
discontinuity of the @xmath713 would mean that the @xmath714 would not be a continuous function .
this is not what one expects .
indeed discontinuity singularities ( the shocks ) are expected to occur only in the force @xmath715 as is clear from the study of the burgers equation ( see e.g.@xcite for the discussion of simple case , in the elastic manifold formulation the shocks corresponds to rare ground state degeneracies ) .
one thus expects @xmath714 to be a continuous function of @xmath30 .
a further and more stringent assumption , discussed above , is the absence of supercusp .
a supercusp would mean @xmath716 .
thus we assume that the non - analyticity in the effective action starts as @xmath717 . the usual interpretation .
indeed one could argue that if there are many small shocks they could build a supercusp .
for instance consider the non - trivial @xmath54 limit of the random field model , when @xmath718 has at large @xmath30 the statistics of a brownian motion .
then , in some definitions of a coarse grained disorder , e.g. such as used in @xcite where this model was solved exactly , @xmath719 is a continuous one dimensional brownian motion , thus with a infinite number of small shocks and indeed a supercusp .
however , in the present paper , @xmath719 is defined from the replicated effective action , and not from the action , and should possess - in that case - a weaker singularity . ]
is that there is a finite density of shocks and just counting how many shocks there is in a interval between @xmath30 and @xmath720 yields the @xmath721 non - analyticity in @xmath27 .
let us summarize the method before detailing actual calculations .
we thus define here the symmetrization method assuming no supercusp as a working hypothesis .
we then _ compute _ corrections to the ( local disorder part of the ) effective action up to a given order in powers of @xmath86 , with excluded vertices for any vector such that @xmath722 for @xmath557 , thus with no ambiguity .
this yields , as in section [ sec : unambdiag ] , sums over more than two replicas with exclusions .
these exclusions are not permutation symmetric so we first rewrite them in an explicitly permutation symmetric way which can be done with no ambiguity ( see below ) .
thus we have a sum of terms of the form @xmath723 where @xmath724 is a short - hand notation for @xmath725 for all @xmath726 , i.e. symmetrized exclusions .
each function @xmath22 is fully permutation symmetric , as indicated by the @xmath727 superscript .
next the non - trivial part is that we _
explicitly verify _ that these _ symmetrized _ corrections can indeed be continued to coinciding points unambiguously , e.g the limit @xmath728 exist and is independent of the direction of approach .
this in itself shows that the continuity discussed above seems to work .
the existence of a four replica term obliges us to also consider three coinciding points .
this is done by considering @xmath729 , i.e.symmetrizing the result of two coinciding points over @xmath730 and then taking @xmath731 .
we check explicitly that this again gives a function which can be continued unambiguously .
thus at first sight , it would appear as the ideal method to extract the functions @xmath713 above to order @xmath732 .
let us reconsider the diagrams of fig .
[ excl ] we first transform them in sum with symmetrized constraints . we illustrate this on diagram @xmath0 where the sum can be reorganized as : @xmath733 with clearly no ambiguity .
performing similar rearrangement on all the graphs of class a yields the sum of the graphs : @xmath734 now we use the property that has worked on all the examples needed here , namely that a symmetric continuous function on @xmath735 is continuous on @xmath736 . writing for any @xmath737 symmetric and continuous : @xmath738 and expanding yields , for the three and four replica sums : @xmath739 in shorthand notations such that @xmath740 .
this is just combinatorics .
for the three replica sums the procedure is straightforward , as symmetrization makes manifest the continuity .
one easily finds ( we drop an uninteresting single replica term ) : @xmath741 where in the first line we have applied ( [ lf201 ] ) to @xmath742 and so on ( we define @xmath743 as the sum over all permutations divided by @xmath744 ) .
for the 4-replica term we find that @xmath745 has the following limits ( in a symbolic form , omitting the free summations ) @xmath746 where at each step we had to symmetrize before taking coinciding point limits ( checking that this limit was unambiguous in each case ) .
the final result is found to be : @xmath747 the same procedure applied to the repeated counter - term confirms that it is unambiguous and give by ( [ 1-loop - rep - cc ] ) .
thus because of the ominous @xmath748 coefficient above , rather than the expected @xmath176 the theory , using this procedure , is not renormalizable
. diagrams of class b and c behave properly .
one finds with the same method their projections on the 2-replica part : @xmath749 note the @xmath562 which here is defined as @xmath750 since @xmath751 can be continued at zero .
one has , using the expressions given in appendix [ sec : classc ] : @xmath752 these graphs ( more precisely their contribution to 2-rep terms ) sum exactly to zero : @xmath753 in agreement with the result of the ambiguous diagrammatics in the case of an analytic function . to conclude , although promising at first sight this method is not satisfactory .
the projection defined here seems to fail to commute with further contractions .
for instance one can check that upon building diagrams a by contracting the subdiagram ( a ) in fig .
[ f : sloops ] onto a third vertex does give different answers if one first projects ( a ) or not . since ( a ) is the divergent subdiagram this spoils renormalizability .
since the initial assumptions of the method were rather weak and natural , it would be interesting to see whether this problem can be better understood in order to repair this method .
in this section we give some details on the method where one performs straight perturbation theory using a non - analytic disorder correlator @xmath754 in the action . expanding in @xmath754 ,
this involves computing gaussian averages of non - analytic functions , thus we start by giving a short list of formula useful for field theory calculation of this section .
one should keep in mind that these formula are equally useful for computing averages of non - analytic _ observable _ in a gaussian ( or more generally , analytic ) theory .
we start by deriving some auxiliary functions , then give a list of expectation values for non - analytic observables of a general gaussian measure .
we need @xmath755 integrating once over @xmath8 starting at 0 yields @xmath756 the r.h.s .
reduces in the limit of @xmath757 to @xmath758 , which gives a representation of @xmath759 @xmath760 by integrating once more , we obtain @xmath761 , by integrating more often . the result is @xmath762[0.8ex]{\rule{0.1ex}{2.75ex}\hskip0.2ex}}_{n } \ , \ ] ] where @xmath763[0.8ex]{\rule{0.1ex}{2.75ex}\hskip0.2ex}}_{n}$ ] means that one has to subtract the first @xmath59 taylor - coefficients of @xmath764 , such that @xmath763[0.8ex]{\rule{0.1ex}{2.75ex}\hskip0.2ex}}_{n}$ ] starts at order @xmath765 : @xmath766[0.8ex]{\rule{0.1ex}{2.75ex}\hskip0.2ex}}_{n } : = \sum_{\ell = n}^{\infty } \frac{[-(qx)^{2}]^{\ell}}{(2\ell ) ! } \ .\ ] ] we now study expectation values .
we use the measure @xmath767 from which the general case can be obtained by simple rescaling @xmath768 , @xmath769 .
let us give an explicit example ( we drop the convergence - generating factor @xmath770 since it will turn out to be superfluous . ) @xmath771 a more interesting example is @xmath772 another generally valid strategy is to use a path - integral .
we note the important formula @xmath773\nonumber\!\ ! \\ & & \end{aligned}\ ] ] an immediate consequence is @xmath774 \nonumber \\ & = & \frac{1}{2\pi}\int\limits_{-\infty}^{\infty}\!{{\mathrm{d}}}z\ , f(z\sqrt{1-t^{2}})\,\exp\!\left ( -z^{2 } \right ) \nonumber \\ & = & \frac{1}{\sqrt{2\pi } } \left < f ( x\sqrt{1-t^{2 } } ) \right > \label{lf221}\end{aligned}\ ] ] the very existence of the path - integral representation ( [ pathint ] ) also proves that wick s theorem remains valid .
let us give an example which can be checked by either using ( [ pathint ] ) or ( [ b4 ] ) : @xmath775 we finish our excursion by giving a list of useful formulas , which can be obtained by both methods : @xmath776 \\ \label{lf223 } \left < x y|y|\right > & = & 2 \sqrt{\frac{2}{\pi } } t \\
\label{xy|xy| } \left < x y |xy| \right > & = & \frac{2}{\pi } \left [ 3 t \sqrt{1-t^2 } + ( 1 + 2 t^2 ) \arcsin ( t)\right]\qquad \\
\left<|x y^3|\right > & = & \frac{2}{\pi } \left [ ( 2 + t^2 ) \sqrt{1-t^2 } + 3 t \arcsin ( t ) \right]\label{lf226}\end{aligned}\ ] ] let us consider the expansion of the two point function @xmath777 in powers of the disorder , where @xmath778 , with @xmath779 .
we want to evaluate these averages at @xmath39 with a non - analytic action @xmath754 .
we restrict ourselves to @xmath557 since at @xmath39 the result should be the same for @xmath540 , and we drop the subscript @xmath780 from now on . as mentioned above , the wick theorem still applies , thus we can first contract the external legs .
the term linear in @xmath781 yields the dimensional reduction result ( [ lf9 ] ) , thus we note @xmath782 and we find : @xmath783 up to @xmath784 terms . for peace of mind
one can introduce the restrictions @xmath785 in the first sum and @xmath786 in the second , but this turns out to be immaterial at the end .
we need only , in addition to ( [ exp ] ) : @xmath787 since higher order terms in @xmath30 yield higher powers of @xmath70 . using ( [ lf222 ] ) to evaluate gaussian averages
this yields : @xmath788 where we denote : @xmath789 note that the cross terms @xmath790 involve analytic averages can always be set formally to zero by a trivial additive random force contribution . ] and yield zero ( a remnant of dimensional reduction ) . also , to this order ,
no terms with negative powers of @xmath70 survive for @xmath60 ( see discussion below ) .
performing the combinatorics in the replica sums , we find for @xmath60 : @xmath791 \\
\phi_1(s ) & = 2 \phi_1\left(\frac{s}{2}\right ) - \phi_1(s ) \\ \phi_2(s ) & = - \frac{2}{3 } \phi_2(s ) + \frac{8}{3 } \phi_2\left(\frac{s}{2}\right ) - 2 \phi_2(0 ) \end{aligned}\ ] ] it is important for the following to note that cancellations occur in the small argument behavior of these functions , namely one has @xmath792 and @xmath793 .
in @xmath54 it simplifies ( setting @xmath794 and restoring the subscript ) to : @xmath795 with @xmath796 .
as such , this formula and ( [ res ] ) seem fine and it may even be possible to check them numerically in @xmath54 for large @xmath98 using a bare disorder with the proper non - analytic correlator @xmath754 . to obtain the asymptotic @xmath797 and large scale behavior in any @xmath9 , one must resum higher orders and use an rg procedure .
the question is whether the above formula ( [ res ] ) can be used in an rg treatment .
we found that this procedure does not work and we now explain why .
let us rewrite the result ( [ res ] ) , including the dimensional reduction term : @xmath798 } { ( q^2 + m^2)^2 } \nonumber \\
a_i(q ) & = & \int { { \mathrm{d}}}^d x\ , { { \mathrm{e}}}^{i q x } g(x)^2 \psi_i(x ) \nonumber \\ \psi_i(x ) & = & - \left(\frac{g(0)}{g(x ) } \right)^{\!2 } \phi_i\ !
\left(\frac{g(x)}{g(0)}\right)\end{aligned}\ ] ] with @xmath799 .
one notes that if @xmath800 were a constant equal to unity , one would recover the result ( [ lf185 ] ) obtained in section ( [ sec : correlations ] ) .
however , one easily sees that while @xmath801 approaches a constant as @xmath802 where @xmath803 is an ultraviolet cutoff , it decreases as @xmath804 at large @xmath8 , as a result of the above mentioned cancellations in the small argument behavior of the functions @xmath805 .
thus the infrared divergence responsible for all interesting anomalous dimensions in the 2-point function as the non - trivial value of @xmath19 is killed , and the method fails .
even more , the theory would not even be renormalizable .
we have performed a similar calculation in the dynamical field theory formulation of the equilibrium problem in the limit @xmath806 , using a non - analytic action . there
the method fails for very similar reasons . only at the depinning threshold
we were able to construct the dynamical theory as explained in @xcite .
one might suspect that one has to start with a somehow `` normal - ordered '' theory where self - contractions , i.e. terms proportional to @xmath807 are removed , since they _
never _ appear in the @xmath39 perturbation theory .
we have not been able to find such a formulation .
another problem with direct perturbation theory in a non - analytic action is that there is a priori no guarantee that it has a well defined @xmath39 limit .
let us illustrate this on a simple example in @xmath54 .
the following correlation has been computed exactly by a completely different method @xcite for the random field model in @xmath54 ( brownian motion plus quadratic energy landscape , @xmath808 indicates averages over all @xmath30 ) : @xmath809 a result which is also obtained by extrapolation from @xmath38 using the frg , as detailed in section [ ss : universal amplitude ] . on the other hand , the above perturbative method yields , expanding in @xmath28 : @xmath810 with @xmath811 . in the zero - temperature limit @xmath812 , which is ill behaved .
the absence of a well defined taylor expansion in the zero - temperature limit is of course a sign that the correct result ( [ lf92 ] ) is simply non - analytic in @xmath28 .
although this solvable example involves a correlator @xmath754 with a _ supercusp _ , it is possible that a similar problem occurs at higher orders ( three or higher ) in the expansion of the 2-point function in the case of the usual cusp non - analyticity .
there have been conflicting claims in the literature about this question @xcite , i.e. the presence of fractional powers at higher orders of the expansion in a non - analytic disorder , and it may be worth reexamining .
it is however important to note that , since the @xmath5-expansion proposed in the main text is not based on such a direct expansion , it _ does not _ yield fractional powers of @xmath5 , contrarily to what was conjectured in @xcite . finally , let us point out some properties of non - analytic _
observables_. let us study e.g. @xmath813 .
expansion in powers of @xmath781 yields a first order term @xmath814 .
this is the sign of non - analytic behavior and indeed it is easy to find that : @xmath815 + o(r^2)\nonumber \\ \label{abs}\end{aligned}\ ] ] where @xmath816 and @xmath817 .
the first term is obtained by noting that @xmath818 acts as a gaussian random force which can then be separated from the non - linear force , and the last term , evaluated using the above formula , is the only one which survives at @xmath39 to linear order in @xmath204 .
the formula ( [ abs ] ) is interesting as a starting point to compute universal ratio , such as @xmath819 or @xmath820 .
indeed one notes that for @xmath44 the integral in the term proportional to @xmath821 is infrared divergent at large @xmath324 .
this is left for future study .
in this appendix we give the expression of each of the diagrams of class c represented in fig .
( [ excl ] ) in the excluded ( nonambiguous ) diagrammatics .
one finds , including all combinatorial factors : @xmath822 with : @xmath823
let us consider the expression @xmath824 for the b diagrams in the excluded diagrammatics ( [ deltab ] ) .
let us start again from a single sloop ( [ id1 ] ) and ( [ id2 ] ) and contract this time between @xmath324 and @xmath536 twice to produce a diagram of type b. this yields : @xmath825{./figures / banana2sloop}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / banana2sloop}{\hspace*{0.8ex}}}}\nonumber = \frac{1}{2 } r''(0 ) \sum_{a \neq b } r''''_{ab } r''_{ab } \\ & & + \frac{1}{2 } r''(0 ) \sum_{a \neq b , b \neq c } r''''_{ab } r''_{bc } + \frac{1}{2 } \sum_{a \neq b , b \neq c } r''_{ab } r''''_{bc } r''_{bc } \nonumber \\ & & + \frac{1}{4 } \sum_{a \neq b , a \neq c , c \neq d } r''_{ab } r''''_{ac } r''_{cd } \nonumber \\ & & + \frac{1}{4 } \sum_{a \neq c , b \neq c , c \neq d } r''_{bc } r''''_{ac } r''_{cd } \equiv 0\end{aligned}\ ] ] the terms @xmath168 arise because the first vertex is not contracted in the process so one must separate the ( unambiguous ) diagonal part to obtain excluded sums .
if one subtracts this identity from ( [ deltab ] ) one finds that there remain some improper three replica term ( the improper four replica term however cancels ) .
this is because in the process of our last contractions we have generated new sloops , but , since replica were excluded they have to be extracted with care .
let us rewrite the two possible `` double sloop '' from unrestricted sums to restricted : @xmath826{./figures/2banana2sloopsa}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/2banana2sloopsa}{\hspace*{0.8ex } } } } = \frac{1}{2 } \sum_{a b c d } r''_{ac } r''''_{ab } r''_{ad } \nonumber \\ & & = \frac{1}{2 } \sum_{a \neq b , a \neq c , a \neq d } r''_{ac } r''''_{ab } r''_{ad } + \frac{1}{2 } \sum_{a \neq b } r''(0)^2 r''''_{ab } \nonumber \\ & & \qquad \qquad + \sum_{a \neq b } r''(0 ) r''''_{ab } r''_{ac } \label{lf249}\end{aligned}\ ] ] @xmath827{./figures/2banana2sloopsb}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/2banana2sloopsb}{\hspace*{0.8ex } } } } = \nonumber \frac{1}{2 } \sum_{a b c d } r''_{ac } r''''_{ab } r''_{bd } \\ & & = \frac{1}{2 } \sum_{a \neq b , a \neq c , b \neq d } r''_{ac } r''''_{ab } r''_{bd } + \frac{1}{2 } \sum_{a \neq b } r''(0)^2 r''''_{ab } \nonumber \\ & & \qquad + \sum_{a \neq b } r''(0 ) r''''_{ab } r''_{ac } \ .\end{aligned}\ ] ] in the process we have set to zero the terms @xmath828 since they are proper three and four replica terms . defining now : @xmath829{./figures/2banana2sloops}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/2banana2sloops}{\hspace*{0.8ex}}}}:=\frac{1}{2}\left [ { \settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/2banana2sloopsa}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/2banana2sloopsa}{\hspace*{0.8ex } } } } + { \settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/2banana2sloopsb}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/2banana2sloopsb}{\hspace*{0.8ex}}}}\right]\ .\ ] ] the simplest combination which allows to extract the 2-replica part is : @xmath830{./figures / banana}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / banana}{\hspace*{0.8ex}}}}-2 { \settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / banana2sloop}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures / banana2sloop}{\hspace*{0.8ex } } } } + { \settowidth{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/2banana2sloops}{\hspace*{0.8ex}}}\parbox{\bilderlength}{{\hspace*{0.8ex}}\includegraphics[scale={0.35}]{./figures/2banana2sloops}{\hspace*{0.8ex } } } } \nonumber \\ & & = \frac{1}{2 } \sum _ { a , b } r''''_{ab}\left(r''_{ab}-r''_{0 } \right)^{2}\ .\end{aligned}\ ] ] we now turn to graphs @xmath172 . the expression for @xmath831 is given as the sum of all contributions in appendix [ sec : classc ] . within the sloop method it gives immediately zero : @xmath832 .
this is because one can start by contracting the tadpole .
since this is a sloop it can be set to zero : @xmath833 upon further contractions , proceeding as in section [ sec : sloops ] , one obtains exactly that the sum of all graphs @xmath172 with excluded vertices is identically zero .
graphs @xmath172 sum to zero since they are all descendants of a sloop .
we will illustrate the universality of @xmath834 using a broad class of ir cutoff functions , namely a propagator : @xmath835 here we denote @xmath836 and we normalize @xmath837 ( consistent with fixing the elastic coefficient to unity ) .
we will show that @xmath838 independent of @xmath839 .
first we write : @xmath840 using the parameterization @xmath841 and @xmath842 one obtains : @xmath843 } \nonumber \\ & = & m^{-\epsilon } \left(\int_q e^{- q^2 } \right ) \gamma\left(\frac{\epsilon}{2 } \right ) \int_{x , x ' } \int_{-1/2}^{1/2 } { { \mathrm{d}}}\lambda \left[\frac{x + x'}{2 } + \lambda ( x - x ' ) \right]^{-\epsilon/2 } \label{xban}\end{aligned}\ ] ] the hat - diagram is @xmath844^{-d/2}{\nonumber}\\ & = & \left(\int_{q } { { \mathrm{e}}}^{-q^2}\right)^2 \int_{x_i } \int_{\alpha , \beta_1 , \beta_2 , \gamma > 0 } \gamma^{3-d } { { \mathrm{e}}}^{- m^2 ( x_1 \alpha + x_2 \beta_1 + x_2 ' \beta_2 + x_3 ) \gamma } ( \alpha + \beta_1 + \beta_2 + \alpha ( \beta_1 + \beta_2))^{-d/2 } { \nonumber}\\ & = & \left(\int_{q } { { \mathrm{e}}}^{-q^2}\right)^2 \gamma ( 4-d ) m^{- 2 \epsilon } j \ , \end{aligned}\ ] ] where we split the divergent integral @xmath845 in pieces , which are either finite or where the divergence can be calculated analytically : @xmath846 + x_3\right\}^{-\epsilon } \\
j_{1 } & = & \int_{x_i } \int_0^{\infty } { { \mathrm{d}}}\alpha \int_0^{1 } { { \mathrm{d}}}\beta \ , g(\alpha,\beta , x_i ) = \ln 2 + o(\epsilon ) \\ j_{2 } & = & \int_{x_i } \int_0^{\infty } { { \mathrm{d}}}\alpha \int_1^{\infty } { { \mathrm{d}}}\beta \left\ { g(\alpha,\beta , x_i ) - \frac{1}{(1+\alpha ) ^{2-\frac{\epsilon}{2 } } \beta^{1 + \frac{\epsilon}{2 } } } \int_{-1/2}^{1/2 } { { \mathrm{d}}}\lambda \left[\frac{x_2 + x_2'}{2 } + \lambda ( x_2-x_2')\right]^{-\epsilon } \right\}\nonumber\\ & = & - \ln 2 + o(\epsilon ) \\ j_3 & = & \int_0^{\infty } { { \mathrm{d}}}\alpha \int_1^{\infty } { { \mathrm{d}}}\beta\ , \frac{1}{(1+\alpha ) ^{2-\frac{\epsilon}{2 } } \beta^{1 + \frac{\epsilon}{2 } } } \int_{x_{2},x_{2}'}\int_{-1/2}^{1/2 } { { \mathrm{d}}}\lambda \left[\frac{x_2 + x_2'}{2 } + \lambda ( x_2-x_2 ' ) \right]^{-\epsilon}\end{aligned}\ ] ] this gives the final result for the hat - diagram @xmath847^{-\epsilon } { \nonumber}\\ & = & \left(\frac{1}{2 \epsilon^2 } + \frac{1}{4 \epsilon}+o\left(\epsilon^{0 } \right)\right ) \left(\epsilon i_1 \right)^2 \label{a.18 } \ , \end{aligned}\ ] ] where we have used that in the 1-loop integral ( [ xban ] ) @xmath848^{-\epsilon/2 } = \left(1+\frac{1}{2}\alpha\epsilon + o ( \epsilon^{2})\right)\ ] ] with @xmath448 depending on the regulating function @xmath849 and in the 2-loop integral ( [ a.18 ] ) @xmath850^{-\epsilon } = \left(1+\alpha\epsilon + o ( \epsilon^{2})\right ) = \left(1+\frac{1}{2}\alpha\epsilon + o ( \epsilon^{2})\right)^{2}\ ] ] with the _ same _ @xmath448 .
with excluded vertices.,title="fig : " ] in this section we give all 1-loop and 2-loop diagrams including finite @xmath70 , evaluated with the unambiguous diagrammatics , which have not been given in the text .
we use the unambiguous vertex @xmath851 , denote @xmath852 , @xmath853 etc .. the list of all uv - divergent diagrams up to two loops is given in fig . [
f : diverg ] .
we write their contribution to the effective action as @xmath854|_{u_x = u } & = & - \frac{1}{2 t^2 } r \\\label{lf238 } r & = & \sum_{ab } r_{ab } + \delta^{(1 ) } r + \delta^{(2 ) } r + \dots \ .\end{aligned}\ ] ] the total 1-loop contribution is @xmath855 the total 2-loop contribution is : @xmath856 where @xmath857 is given in ( [ deltaa ] ) , @xmath858 is given in ( [ deltab ] ) and @xmath859 i_t i_t \nonumber \\ & & \label{deltac}\end{aligned}\ ] ] while the finite @xmath70 diagrams are given by @xmath860 for an _ analytic _
@xmath86 one substitutes @xmath861 in the above formula and selects the 2-replica terms : @xmath862 i_1 \\\label{lf241 } \delta^{(2)}_b r & = & \frac{1}{2 } r '' '' ( r '' - r''(0))^2 i_1 ^ 2 \\\label{lf242 } \delta^{(2)}_a r & = & ( r '' - r''(0 ) ) ( r''')^2 i_a \\\label{lf243 } \delta^{(2)}_c r & = & 0 \\\label{lf244 } \delta^{(2)}_t r & = & \frac{1}{2 } t^2 r '' '' i_t^2 + \frac{1}{2 } t ( r''')^2 i_4 \nonumber \\ & & + \big[t r '' '' ( r '' - r''(0 ) ) - t r '' r''''(0)\big ] i_1 i_t\ .\nonumber \\ & & \label{lf245}\end{aligned}\ ] ] let us show that if one renounces to the projection onto 2-replica terms , one can still obtain some formal renormalizability property , but at the cost of introducing an unmanageable series of terms with more than two replicas . we show how to subtract divergences by adding counter - terms of similar form
let us discuss only @xmath39 . to cancel the 1-loop divergences we introduce the counter - term : @xmath863 the repeated counter - term is @xmath864 i_1 i_1^{\text{div}}\end{aligned}\ ] ] omitting all ( excluded ) sums .
one checks that @xmath865 using that @xmath866 .
thus there is some renormalizability property for @xmath86 .
one can thus define formally a @xmath0-function : @xmath867 ( \epsilon i_1 ) \nonumber \\ & & + \delta^{(2)}_a r \frac{\epsilon ( i_{a}-\frac{1}{2}i_{1}^{2})}{i_{a}}\end{aligned}\ ] ] @xmath86 however includes a series of terms each with excluded sums over @xmath68 replica .
thus to be consistent one should in principle include them from the start and pursue the method .
it is not clear that it can be closed in any way .
it is a pleasure to thank e. brzin , w. krauth and a. rosso , for stimulating discussions .
k.j.w . gratefully acknowledges financial support by the deutsche forschungsgemeinschaft ( heisenberg grant wi-1932/1 - 1 ) , and additional support from nsf under grant phy99 - 07949 .
thanks the kitp , and k.j.w .
thanks ens for hospitality during part of this work .
m. kardar , , in f. david , p. ginsparg and j. zinn - justin , editors , _ fluctuating geometries in statistical mechanics and field theory _ , _ _ volume _ lxii _ of _ les houches , cole dt de physique thorique 1994 _ , elsevier science , amsterdam , 1996 . | we study elastic systems such as interfaces or lattices , pinned by quenched disorder . to escape triviality as a result of `` dimensional reduction ''
, we use the functional renormalization group .
difficulties arise in the calculation of the renormalization group functions beyond 1-loop order . even worse , observables such as the 2-point correlation function exhibit the same problem already at 1-loop order .
these difficulties are due to the non - analyticity of the renormalized disorder correlator at zero temperature , which is inherent to the physics beyond the larkin length , characterized by many metastable states . as a result , 2-loop diagrams , which involve derivatives of the disorder correlator at the non - analytic point , are naively `` ambiguous '' .
we examine several routes out of this dilemma , which lead to a unique renormalizable field - theory at 2-loop order .
it is also the only theory consistent with the potentiality of the problem .
the @xmath0-function differs from previous work and the one at depinning by novel `` anomalous terms '' . for interfaces and random bond disorder
we find a roughness exponent @xmath1 , @xmath2 . for random field disorder
we find @xmath3 and compute universal amplitudes to order @xmath4 . for periodic systems we evaluate the universal amplitude of the 2-point function .
we also clarify the dependence of universal amplitudes on the boundary conditions at large scale .
all predictions are in good agreement with numerical and exact results , and an improvement over one loop .
finally we calculate higher correlation functions , which turn out to be equivalent to those at depinning to leading order in @xmath5 . |
quantization of vorticity is a remarkable manifestation of superfluid properties of bose - einstein condensates ( becs ) .
different kinds of vortex structures have been theoretically predicted and observed experimentally in atomic becs ( see e.g. @xcite and references therein ) : single vortex lines , vortex - antivortex pairs , vortex arrays , vortex rings , vortex knots , dimerons @xcite , three - dimensional vortex solitons @xcite , and solitary waves moving along a straight vortex line @xcite ( which are similar to `` hoop '' structures known in field theory @xcite ) .
it is of special interest to construct stable vortex - soliton states with complex structure , such as vortex knots @xcite , three - dimensional skyrmions , @xcite and hopfions .
a hopfion ( or hopf soliton ) is a topological soliton with two independent winding numbers : the first , @xmath0 , characterizes a horizontal circular vortex embedded into a three - dimensional soliton ; and the second , @xmath1 , corresponds to vorticity around the axis , perpendicular to this circle .
hopf solitons appear in many fields , including field theory , optics , ferromagnets , and semi- and superconductors .
hopfions have been predicted theoretically both in systems with multicomponent wave functions @xcite and in single - component bec @xcite .
multicharged ( @xmath2 , @xmath3 ) vortex structures have been demonstrated @xcite to be unstable in optical media .
stationary hopfion solutions with @xmath4 , @xmath5 of the single - component gross - pitaevskii - equation ( gpe ) have been obtained numerically in ref .
@xcite for a spherically - symmetric trapping potential . to the best of our knowledge
, neither a dynamical nor an energetic stability analysis has been performed previously for hopfions in trapped becs .
very recently @xcite hopfions have been demonstrated to be stable in a single - component bec without linear trapping potential but with a repulsive non - linearity , which grows fast enough from the center to the periphery of the trap . in our recent work
@xcite we suggested an experimentally feasible trapping configuration that can be used to create , stabilize , and manipulate a vortex ring in a controllable and nondestructive manner . in this paper
we use a similar trapping potential and a rotating condensate which allows the formation of a stable hopfion .
we demonstrate both energetic and dynamical stability of the hopf soliton for realistic experimental parameters .
the properties of an atomic bec close to thermodynamic equilibrium can be accurately described in mean - field approximation by a dissipative gross - pitaevskii equation ( dgpe ) , which takes the following form in a rotating frame @xcite , @xmath6\psi(\textbf{r},t).\end{aligned}\ ] ] here , @xmath7 is a phenomenological dissipation parameter , @xmath8 the rotation rate , @xmath9 is the laplace operator , @xmath10 is the angular momentum in @xmath11-direction , @xmath12 is the coupling strength , @xmath13 is the chemical potential of the condensate , @xmath14 is the mass of the atom , and @xmath15 is the @xmath16-wave scattering length .
we assume that the wave function is normalized to the number of atoms @xmath17 in the absence of dissipation ( @xmath18 ) , the particle number @xmath19 as well as the total energy @xmath20 are integrals of motion . in our model
the external trapping potential @xmath21 is created by an oblate harmonic trap @xmath22 and two blue - detuned laser beams : a radial laguerre - gaussian beam lg@xmath23 and an elliptic highly anisotropic `` sheet '' beam creating a tight repulsive potential in @xmath11-dimension : @xmath24 } + v_z e^{-(z / z_{\rm t})^2},\ ] ] where @xmath25 , @xmath26 is the effective width of the sheet - beam , @xmath27 is the radial coordinate of the trap minimum , and @xmath28 is the topological charge of the red - detuned lg@xmath23 optical beam .
let us consider a bec cloud with @xmath29 of @xmath30rb atoms ( @xmath31 nm ) in an oblate harmonic trap with @xmath32 hz , @xmath33 hz , and a characteristic oscillatory length @xmath34 m .
the tube beam has parameters @xmath35 , @xmath36 m , and @xmath37khz , the sheet beam has parameters @xmath38 m , and @xmath39khz , where @xmath40 is the planck s constant .
this trap geometry is similar to that used in ref .
@xcite modified slightly to increase the stability of the vortex line by squeezing the trap in the @xmath11 direction and shifting the toroidal void position farther from the center .
such a trap configuration is supposed to naturally accommodate the complex core structure of a hopfion ( see fig .
[ torus ] ) . in what follows we use dimensionless units for
the spatial coordinates @xmath41 , for time @xmath42 , and for other quantities @xmath43 ,
@xmath44 , @xmath45 , and @xmath46 . the dgpe then becomes @xmath47\psi(\textbf{r},t),\end{aligned}\ ] ] where @xmath48 .
the trapping potential in dimensionless units takes the form @xmath49 with the parameters @xmath50 , @xmath51 , @xmath52 , @xmath53 , @xmath54 . for dynamical calculations
the phenomenological dissipation parameter @xmath55 was chosen to be constant , though general results do not qualitatively depend on this value .
if we aim to study the dependence of the total energy ( [ e3d ] ) on the rotation rate @xmath8 then it is convenient to split it into two parts @xmath56 the first part @xmath57 is the energy in the absence of rotation @xmath58d\mathbf{r},\ ] ] and the second part @xmath59 depends on the average of the angular momentum @xmath60 we obtain the bec ground - state @xmath61 by imaginary time propagation @xcite from eq .
( [ gpedl ] ) with @xmath62 and then compute its energy @xmath63 from eq .
( [ e3d ] ) .
then , in order to model a hopfion state characterized by topological charges @xmath0 and @xmath1 , we imprint on the ground state an @xmath0-charged vortex ring with coordinates @xmath64 and an @xmath1-charged vortex line shifted by @xmath65 off the center .
@xmath66 with @xmath67 , @xmath68 , @xmath25 , @xmath69 , and @xmath70 .
the width of the vortex core is estimated by the healing length @xmath71 at the peak density of the condensate .
if the vortex line is straight and parallel to the @xmath11-axis and the vortex ring is parallel to the @xmath72-plane , the superflows of the vortex ring and line are moving in perpendicular directions .
additionally , the density perturbations caused by the vortex cores are tightly localized and do not overlap in the regions of interest .
therefore , the contributions to the total energy from the vortex ring and vortex line are independent and can be analysed separately , @xmath73 and we take into account that the vortex ring in the @xmath74plane does not contribute to the angular momentum in @xmath11-direction
. the energy analysis of the vortex ring is performed by putting @xmath5 and @xmath75 in ( [ psiimprint ] ) .
then the energy @xmath76 depends only on @xmath77 and @xmath78 ( see fig . [ energy_ring ] ) .
the vortex ring nucleation energy shows a minimum localized close to the density minimum at the crossing of the `` tube '' and `` sheet '' beams .
it is worth noticing that the existence of an energy minimum does not necessarily correlate with the existence of a density minimum in the ground state .
figure [ energy_ring2 ] shows that for certain intensities of the `` tube '' beam an off - center minimum exists in the vortex ring nucleation energy but not in the ground - state density .
( @xmath79 blue solid line , @xmath80 red dashed line , @xmath81 green dotted line ) . ] the energy associated with the vortex line is obtained by putting @xmath82 and @xmath4 in eq .
( [ psiimprint ] ) .
this energy @xmath83 depends only on @xmath8 and @xmath65 . in analogy to the analysis of fetter @xcite we can define two characteristic rotation speeds .
the first , @xmath84 , is chosen such that @xmath85 at @xmath86 .
it means that for @xmath87 a state with the vortex at the origin is energetically preferable over a state without vortex .
this rotation rate can be defined as @xmath88 note that @xmath89 .
therefore , in order to obtain this critical velocity only @xmath90 has to be calculated .
the second characteristic rotation rate @xmath91 corresponds to a vanishing curvature of @xmath92 at @xmath86 : @xmath93 this rotation rate defines the region @xmath94 where the energy shows a minimum at @xmath86 and the vortex line at the origin can be stable ( or metastable ) . due to our specific trap configuration
there might be another energy minimum for the vortex line which is close to the peak intensity of the `` tube '' beam .
but this minimum is not relevant for the stability of a hopfion as it correlates with a stable position of the vortex ring . as the line and the ring
have to be spatially separated , we demand the existence of a minimum at the origin as a situation favourable for a stable hopfion . as is seen from fig .
[ energy_line ] , in the suggested trap configuration , a vortex state can be energetically preferable without being stable at the center of the trap .
this means that if we need to stabilize the vortex core at the axis , the rotation rate has to be considerably higher than @xmath84 .
therefore @xmath95 , which is opposite to the case of a spherical condensate studied in ref .
@xcite . as a result
we consider the obtained value @xmath96 as a minimal rotation rate at which the vortex line can be stable at the center of the trap and , therefore , at which we can also expect the overall stability of the hopfion complex .
for the solution of the time - dependent gpe we use a split - step fourier transform method @xcite .
the initial state for the simulations takes the same form eq .
( [ psiimprint ] ) as in the energetic stability analysis of the previous section , i.e. a stationary state with imprinted vortex ring and line ( @xmath5 , @xmath4 ) .
their initial positions are considerably displaced from their equilibrium ( by up to @xmath97 ) .
we study the dynamical stability of this hopfion solution for a wide range of rotation rates @xmath8 .
we observe two different instability scenarios of the hopfion dynamics .
if the rotation rate is low or the initial position of the vortex line is too far away from equilibrium , the system is unstable with respect to reconnection of the line and ring cores . as a consequence ,
several line vortices are created which , in turn , can be stable and form a lattice at the off - center energy minimum ( see fig . [ unstable1_dynamics ] ) . .
in the initial state @xmath5 , @xmath4 .
reconnection happens at @xmath98 s. the four vortex lines in the final state are stable . ] the second instability scenario is observed at high rotation rates and is independent of the initial displacement of the vortices .
due to a low energy barrier ( see fig . [ energy_line ] ) , additional vortices can be nucleated at the outer edge of the bec cloud and then penetrate into the condensate destroying the ring vortex and again forming a regular vortex chain at the off - center energy minimum ( see fig . [ unstable2_dynamics ] ) . .
in the initial state @xmath5 , @xmath4 . the hopfion complex itself is stable , but starting from @xmath99 s vortices from the outside start to penetrate the cloud eventually destroying the vortex ring . ]
it is however possible to choose a rotation rate that is high enough for stabilization of the vortex line at the center but at the same time low enough that additional vortices are not generated at the outer edge of the condensate
. then the vortex hopfion may be dynamically stable for experimentally traceable time intervals which are comparable to the life - time of the condensate ( see fig . [ stable_dynamics ] ) . for a strong initial displacement of the vortex line of @xmath97 we found a region of stability between @xmath100 .
it is worth noticing that the vortex line relaxes to a position slightly off the center of the trap even though the energy minimum is located there .
however , the energetic stability analysis does not take into account bending of the vortex cores , which are energetically preferred over straight vortices .
this may also explain the fact that the vortex line is dynamically stabilized at a rotation rate lower than predicted by the energetic analysis . .
in the initial state @xmath5 , @xmath4 .
shapes of the vortex cores persist through all the time of simulation . ]
it is also instructive to analyze the dynamics of hopfions with a topological charge of the vortex line @xmath101 . as multiply charged vortices in bec are essentially unstable , the vortex line splits into a tangled complex of several singly charged vortices ( see fig .
[ m2_dynamics ] ) .
because these vortices repel each other they can not all be simultaneously stabilized at the center of the trap .
they rather relax to the energy minimum at the peak intensity of the `` tube '' beam .
this inevitably leads to reconnection with the vortex ring and the observed dynamical scenario is similar to the instability scenario of singly charged vortex lines at low rotation rates ( compare figs .
[ unstable1_dynamics ] and [ m2_dynamics ] ) .
, @xmath102 ) for the rotation rate @xmath103 . ]
similarly , the hopfion with a multicharged vortex ring ( @xmath3 ) can not be stabilized in a suggested trap geometry ( see fig . [ s2_dynamics ] ) .
the multi - charged vortex ring splits into two leapfrogging singly - charged rings , that spiral around the energy minimum drifting away from it until they are destroyed by growing kelvin waves .
it is remarkable that in sharp contrast to studies of multicharged ( @xmath3 ) hopfions in optical media @xcite we do not observe the formation of vortex knots in our simulations of the dissipative dynamics of hopf solitons in bec .
, @xmath4 ) for the rotation rate @xmath103 . ]
the destruction of the hopfion in the proposed setup always leads to formation of a vortex chain along the off - center energy minimum .
the number of vortices in this chain is defined not only by the rotation rate but also by the total energy released by destruction of the hopfion complex ( note a six - vortex final state in fig .
[ s2_dynamics ] ) .
the main aim of the present work is a demonstration that hopfions can be stabilized in experimentally feasible one - component trapped becs .
suitable parameters sets for the stabilization of hopfions depend on the specific experimental realization of bec .
we observed that hopfions are more stable in highly oblate traps since vortex line bending and tilting are suppressed under these conditions @xcite .
furthermore , imprinting the vortex line closer to the axis extends the stability region towards lower rotation rates .
we show that hopf solitons can be energetically and dynamically stable in a rotating trapped atomic bec . in the framework of a dissipative mean - field model we investigate different dynamical regimes of hopfion stability .
it turns out that within a certain interval of bec rotation velocities a hopfion may be stable . for rotation rates lower than this interval
the vortex line and ring deform and finally reconnect . for over - critical angular velocities , vortex lines are nucleated at the outer edge of the condensate and then reconnect with the vortex ring .
multiply charged vortex lines split into singly charged vortex lines , which then reconnect with the ring component .
decay of the hopfion leads to the formation of a vortex lattice .
the kinetic energy of the ring component is transferred via reconnection to the vortex lattice . as is well known
, the experimental detection of vortex rings is much more challenging than observation of the holes in the atomic cloud produced by vortex lines . by counting
the number of vortex lines in the vortex lattice one obtains clear experimental evidence of the presence of the vortex ring component .
we hope that the results presented in this paper will stimulate experiments to observe hopf solitons in atomic becs .
moreover , investigation of the hopfion in the well controlled environment of atomic physics could help to elucidate the properties of these fascinating topological structures in other fields including superconductors , ferromagnets , and quantum field theory .
ac and ay thank the physikalisch - technische bundesanstalt ( ptb ) for warm hospitality during their stay in braunschweig . ac
also thanks the national university of kiev for financial support given by the program `` 100 + 100 + 100 '' .
ay acknowledges support from project 1/30 - 2015 `` dynamics and topological structures in bose - einstein condensates of ultracold gases '' of the knu branch target training at the nas of ukraine .
s.v . is grateful for the support of this work to the german academic exchange service ( daad ) , grant no .
91563279 ( scholarship programme `` research stays for university academics and scientists , 2015 '' ) and the swiss national science foundation grant scope iz 7370 - 152581 . | we reveal that hopf solitons can be stabilized in rotating atomic bose - einstein condensates .
the hopfion is a matter - wave vortex complex which carries two independent winding numbers .
such a topological solitonic structure results from a superfluid flow of atoms simultaneously quantized in poloidal and toroidal directions . in the framework of a dissipative mean - field model
we observe different unstable evolution scenarios of the hopfions .
we demonstrate energetic and dynamical stability of the hopf solitons under experimentally feasible conditions . |
ngc 4038/9 (= arp 244 ) , aptly nicknamed `` the antennae '' , is one of the best - studied examples a galactic collision .
the long tails which distinguish this system are emblematic of violent tidal interactions between disk galaxies of similar mass ( toomre & toomre 1972 , schweizer 1978 ) .
tidal tails originate from the outer regions of galactic disks and are often rich in neutral atomic hydrogen .
these features develop kinematically and their velocity fields bear the imprint of the encounter process , so they provide crucial information for dynamical modeling of interacting galaxies .
we therefore targeted ngc 4038/9 for high - spatial and velocity resolution h@xmath0i observations
( full - width at half maximum , fwhm , as fine as @xmath7 5 km s@xmath8 ) in order to obtain the data needed to constrain future numerical simulations of this system .
ngc 4038/9 was the first major merger to be mapped in the 21-cm line of neutral hydrogen .
the presence of significant quantities of neutral hydrogen within both the main bodies and the tails of ngc 4038/9 was established from single - dish measurements ( peterson & shostak 1974 , huchtmeier & bohnenstengel 1975 ) .
subsequently , ngc 4038/9 has been the subject of h@xmath0i synthesis mapping observations using the westerbork synthesis radio telescope ( wsrt ; van der hulst 1979a ) , the very large array ( vla ; mahoney , burke & van der hulst 1987 ) and the australian telescope compact array ( atca ; gordon , koribalski & jones 2001 ) .
the observations reported in this paper provide a more detailed mapping of the tidal features than afforded by previously published h@xmath0i
maps , representing an improvement in spatial and velocity resolution by over a factor of two .
other spectral line studies have mapped the kinematics of ngc 4038/9 using co spectral lines ( stanford et al .
1990 , wilson et al .
2000 , gao et al .
2001 , zhu 2001 ) and optical emission lines ( burbidge & burbidge 1966 , ruben , ford & dodorico 1970 , amram et al .
1992 ) . however , the molecular and ionized gas is confined to the inner disk regions , and their kinematics do not constrain the encounter geometry as well as that of the tidal tails . at a heliocentric velocity of 1630 km
s@xmath8 ( corresponding to a velocity relative to the local group of 1440 km s@xmath8 ) , the antennae is one of the nearest on - going major mergers . to be consistent with the majority of the recent work on ngc 4038/9
, we derive physical quantities by adopting a hubble constant of 75 km s@xmath8 mpc@xmath8 , yielding a distance of 19.2 mpc .
however , a recent analysis of @xmath9 @xmath10 photometry of individual stars within the southern tidal tail of ngc 4038/9 by saviane , rich & hibbard ( 2001 ) suggests a much smaller distance of 13.80.5 mpc .
this distance is derived by identifying luminous red stars within the tail with the red giant branch of an old , metal poor population .
there are some caveats to this derivation , for which we refer the reader to saviane et al .
, so for the purposes of the present work , we will use the more commonly accepted value of 19.2 mpc . at this distance ,
the tidal tails of ngc 4038/9 extend some 65 kpc in radius and measure @xmath11 kpc from end to end .
the bodies of the galaxies are sites of extensive star formation , producing an ir luminosity of @xmath12 and an ir to blue luminosity ratio of @xmath13 .
long - wavelength studies from the nir to radio suggest that this luminosity is powered by an active system - wide starburst , with no indication of a significant contribution by an agn ( hummel & van der hulst 1986 , vigroux et al .
1996 , fischer et al .
1996 , kunze et al .
1996 , mirabel et al .
1998 , nikola , t. et al .
1998 , laurent et al . 2000 , haas et al .
2000 , xu et al .
2000 , neff & ulvestad 2000 , mengel et al .
while the ir luminosity is an order of magnitude lower than the most active star forming mergers ( the so - called ultraluminous infrared galaxies ) , it is still a factor of @xmath145 higher than non - interacting galaxy pairs ( e.g. kennicutt et al .
1987 , bushouse 1987 , bushouse , lamb & werner 1988 ) , with an inferred star formation rate ( sfr ) of @xmath15 ( evans , harper & helou 1997 ; zhang , fall & whitmore 2001 ) .
ngc 4038/9 is the nearest system with an identified dwarf - galaxy sized concentration of gas , light and young stars embedded within a luminous tidal tail ( schweizer 1978 , hereafter s78 ; mirabel , dottori , & lutz 1992 , hereafter mdl92 ) .
such concentrations have been known for quite some time ( e.g. zwicky 1956 ) , but have only recently received detailed attention .
several numerical studies lend support to the hypothesis that such systems are self - gravitating and may evolve into independent dwarf galaxies ( barnes & hernquist 1992 , elmegreen , kaufmann & thomasson 1993 ) , but the supporting observational evidence is primarily circumstantial .
these observations show that the optical condensations within tidal tails contain many young stars ( s78 , mdl92 , hunsberger , charlton , & zaritsky 1998 , weilbacher et al .
2000 , iglesias - p ' aramo & v ' ilchez 2001 , saviane , rich & hibbard 2001 ) and have global properties , such as size , luminosity , h@xmath0i mass , h@xmath0ivelocity dispersion , and co content , in common with dwarf irregular galaxies ( mirabel et al .
1992 ; hibbard et al . 1994 ; smith & higdon 1994 ; duc & mirabel 1994 , duc et al .
1997 , 2000 ; smith & struck 2001 ; braine et al . 2000 , 2001 ) .
recently , some tidal dwarfs have even been found to harbor `` super star clusters '' ( knierman et al .
2001 ) . for these reasons
, such concentrations are often referred to as `` tidal dwarf galaxies '' ( hereafter tdgs ) .
however , strong kinematic evidence that these concentrations are indeed self - gravitating is lacking , and their evolution into dwarfs is therefore still in question ( see hibbard , barnes , & van der hulst 2001 , hereafter paper ii , for more details ) .
another objective of the present observations is therefore to search for the kinematic signature of a self - gravitating mass in the vicinity of the putative tdg .
the paper is organized in the following manner . in
2 we present details on the observations and data reduction procedures . in
3 we describe in turn the morphology and kinematics of the tidal tails , sub - structure within the tails , the region in the vicinity of the tdg , the inner disks , and the companion galaxy . in 4 we try to explain various aspects of the h@xmath0i tidal morphology , and look for evidence for a kinematically distinct entity at the end of the southern tail .
our conclusions are presented in 5 .
the antennae was observed in two separate array configuration of the vla radio interferometer : the most compact configuration ( 1 km d - array ) in june of 1996 , and the hybrid cnb - array in june of 1997 . for the hybrid array , the antennae along the e and w arms of the array are at the standard c - array ( 3 km ) stations while those of the n arm are at the more extended b - array ( 9 km ) stations .
this compensates for the foreshortening of baselines to the north arm antennae that occurs for sources at low declination .
the short baselines of the d - array provide the highest sensitivity to extended gas , while the longer baselines of the cnb - array allow resolutions as fine as @xmath16 .
ngc 4038/9 was observed for a total of 3.5 hours in the d - array and 8 hours in the cnb - array .
the details of these observations are tabulated in table [ tab : hiobs ] .
we observed using the 21-cm spectral line mode with the system tuned to a central frequency corresponding to a heliocentric velocity of 1630 km s@xmath8 , the central velocity of the existing h@xmath0i synthesis observations ( mahoney et al .
1987 ) . since the atomic gas within tidal tails typically has narrow linewidths ( @xmath17 510 km s@xmath8 ; hibbard et al .
1994 , hibbard & van gorkom 1996 ) , the correlator mode was chosen to give the smallest channel spacing while still covering the velocity spread of existing h@xmath0i observations ( 13601860 km s@xmath8 ; van der hulst 1979a , mahoney et al .
new broader - bandwidth h@xmath0i ( gordon et al .
2001 ) and co ( gao et al .
2001 ) observations reveal that there is both atomic and molecular gas over the velocity range from 13401945 km s@xmath8 .
we discuss the implications of this result below . using on - line hanning smoothing ,
the chosen set - up provides 127 independent velocity channels over the 3.125 mhz bandwidth , resulting in a channel spacing of 5.21 km s@xmath8 .
the pointing center was chosen to place the ends of the optical tails at equal distances from the phase center .
this position is @xmath18 south of the main bodies , and places the ends of the optical tails at a radius of @xmath19 , which is approximately the 75% sensitivity point of the primary beam ( see figure [ fig : hidss ] ) .
the data were calibrated , mapped , and cleaned " using standard techniques and procedures in the astronomical image processing system ( aips ) ( see e.g. rupen 1999 ) .
the data from the separate arrays were calibrated and bandpass corrected independently , after which they were combined in the uv plane to produce the c+d dataset .
this latter dataset is used exclusively throughout this paper .
the final calibration of this dataset was achieved by three iterations of a phase - only self - calibration .
the resulting continuum map is very similar to others that have appeared in the literature ( e.g. hummel & van der hulst 1986 ; neff & ulvestad 2000 , gordon et al .
2001 ) , and is not shown here .
the continuum was subtracted from the velocity cube by fitting a first - order polynomial to the visibilities from the line free channels on either end of the bandpass .
this is done in an iterative manner in order to determine which channels are free of line emission .
the resulting fit was made using channels 1120 and 104111 of the 127 channel cube , corresponding to the velocities 18591911 km s@xmath8 and 13861423 km s@xmath8 .
as mentioned above , cold gas has been found within both of these velocity ranges by gordon et al .
( 2001 ) and gao et al.(2001 ) .
the gas at these velocities is of low surface brightness and totally confined to either the disk of ngc 4039 or the disk overlap region .
therefore , using this continuum range will not affect any of the measured properties of the tidal tails , which is our primary interest in this paper .
however , the h@xmath0i flux we measure for the central region will underestimate the true flux , and the kinematics of this region will not be accurately mapped ( although the fact that the line kinematics are weighted by intensity will somewhat mitigate the effects of the missing velocity information ) .
the continuum subtracted data were mapped using several values of the robust weighting parameter , @xmath20 ( briggs 1995 ) .
this parameter can be varied to emphasize either the outer ( @xmath21 ) or the inner ( @xmath22 ) regions of the uv plane ( i.e. , either small or large spatial scales ) . in practical terms
, a lower @xmath20 gives a finer spatial resolution at the expense of surface brightness sensitivity , while a larger @xmath20 gives a higher surface brightness sensitivity at a slightly coarser spatial resolution .
a high resolution data cube was made with @xmath23 , providing a surface brightness rms sensitivity of 1.3 mjy beam@xmath8 at a resolution of @xmath24 .
this corresponds to a single channel column density limit ( 2.5@xmath25 per channel ) of 2.2 averaged over the beam width of [email protected] kpc@xmath27 .
a more sensitive intermediate resolution data cube , made with @xmath28 , provides a resolution of @xmath29 ( [email protected] kpc@xmath27 ) and a column density limit of 4 . to further increase sensitivity to extended low column density gas ,
a low resolution data cube was made by convolving the @xmath28 data cube to a resolution of @xmath30 ( 3.7 kpc ) , reaching a detection limit of 1.2 .
these data will be referred to in the following as the high , intermediate , and low resolution data , respectively .
these parameters are summarized in table [ tab : hiobs ] .
the resulting per channel noise is close to theoretical once the higher system temperature due to the low declination of the source is taken into consideration .
visualization of the resulting data cubes and inter - comparison with the optical and nir data was greatly facilitated by using the karma visualization package ( gooch 1995 ) .
besides allowing multiple images to be interactively compared , this package also allows a full 3-dimensional rendering of datacubes , which is a particularly powerful tool for disturbed systems . after a careful viewing ,
the data cubes were integrated over the velocity axis to produce the moment maps . in order to suppress the effects of noise ,
only data above a fixed threshold are used in the moment summation ( the `` cutoff technique '' described by bosma , 1981 ) using the aips task momnt .
this task applies a user specified threshold to a version of the data cube smoothed in both space and velocity .
if a pixel in the smoothed data cube passes the threshold , the corresponding pixel in the un - smoothed data is included in the moment analysis .
this procedure favors low level emission that is extended in velocity and/or space over emission which is isolated .
the output from this analysis is an integrated intensity map ( zeroth moment ) , an intensity - weighted velocity map ( first moment ) , and an intensity - weighted velocity dispersion map ( second moment ) .
it should be noted that the first and second moment maps give an accurate representation of the mean h@xmath0i velocity and line - of - sight velocity dispersion at a location only if the line profiles are single - peaked .
the final zeroth moment maps were corrected for the primary beam attenuation before measuring h@xmath0i fluxes .
we measure h@xmath0i fluxes from an integrated intensity map constructed to match the new atca h@xmath0i observations of gordon et al .
( 2001 ) , since those authors were able to do a proper line - free continuum subtraction . for the moment analysis ,
this involved applying a threshold of 2.1 mjy beam@xmath8 to the data after smoothing with a 3@xmath263 pixel boxcar spatial filter and a five - channel hanning filter .
our resulting flux measurements are given in table [ tab : global ] .
the errors attached to the flux measurements have been determined from the single channel noise by properly taking into account the number of independent channels and beams over which the flux was integrated .
it should be noted that the error calculation does not include systematic errors , such as those introduced by poor continuum subtraction or incomplete cleaning .
the supporting optical observations were obtained during two nights of broad - band @xmath31 imaging on the ctio 0.9 m with the tek512 ccd in april of 1991 and one night of near - infrared ( nir ) @xmath32-band imaging on the uh @xmath33 with the quirc detector in january of 1995 .
the details of these runs are given in table [ tab : optobs ] .
for the optical run the conditions were photometric and the seeing was @xmath34 .
the f/13.5 re - imaging optics were used , giving a plate scale of @xmath35 pixel@xmath8 and a field of view of 3.8@xmath36 .
the data were calibrated via observations of standards in the graham e - regions ( graham 1982 ) observed on the same nights , with zeropoint errors ( 1@xmath25 ) of 0.04@xmath37 in @xmath38 and 0.02@xmath37 in both @xmath39 and @xmath20 . for the nir run
, the data were obtained through a @xmath32 filter ( @xmath40 m , @xmath41 m ; wainscoat & cowie 1992 ; hereafter referred to simply as @xmath42 ) .
the f/10 re - imaging optics were used , resulting in a plate scale of @xmath43 pixel@xmath8 and a field of view of 32 .
the seeing was @xmath34 .
these observations consist of three 120 sec target - sky pairs , with the ccd dithered by 15 between on - source positions .
the nir data are uncalibrated .
the combined images were transformed to the world coordinate system ( and thereby to the same reference frame as the radio images ) by registering to an image of the same area extracted from the digitized sky survey ( dss ) .
the registration was accomplished by referencing the location of @xmath14 20 stars in common on both images using the koords program in the karma package ( gooch 1995 ) to perform a non - linear least squares fit for the transformation equations .
the plate solution so found is accurate to a fraction of a pixel , and the overall registration should be as good as that of the southern portion of the guide star catalog , which is estimated to be @xmath44 ( taff et al.1990 ) .
deep optical images were constructed by applying a 9@xmath269 pixel boxcar median to the pixels with the lowest light levels , achieving limiting surface brightnesses of @xmath45=27.0 mag arcsec@xmath46 and @xmath47=26.5 mag arcsec@xmath46 .
color maps in @xmath48 and @xmath49 were made after convolving the images to a common resolution , and using only pixels with a signal - to - noise greater than five in both bands .
these maps will be shown here , but a full color analysis is deferred to a later paper ( evans et al . , in prep . ) .
finally , a smoothed @xmath38-band image was made by replacing stars with background values and convolving to the resolution of the intermediate resolution h@xmath0i data .
this image is used to evaluate the h@xmath0i mass to blue light ratio ( @xmath50 ) of various regions .
the large scale distribution of the h@xmath0i is shown in figure [ fig : hidss ] , where the left panel shows the low resolution h@xmath0i contoured upon the dss image , while the right panel shows the intermediate resolution h@xmath0i in greyscales with contours from the smoothed @xmath38-band superimposed .
our new observations delineate the distribution of tidal h@xmath0i much more clearly than earlier h@xmath0iobservations , showing that the gas in the northern tail extends further than previously known , and revealing a small h@xmath0i - rich disk companion to the antennae lying just beyond the southern tail .
these results are confirmed by the atca observations by gordon et al .
( 2001 ) . using the nasa extragalactic database ( ned ) ,
this latter object is identified with the galaxy eso 572g045 .
we break the h@xmath0i emission into four separate components : the southern tail , the northern tail , the inner disks , and the new dwarf companion eso 572g045 .
the h@xmath0i properties of each of these components are listed in table [ tab : global ] .
the large scale distribution and kinematics of the h@xmath0i derived from the intermediate resolution data are shown in figure [ fig : allmos ] .
the upper left panel shows a false color representation of the data , with the h@xmath0i shown in blue and starlight in green and white .
the upper right panel of fig .
[ fig : allmos ] shows the @xmath48 color map with @xmath38-band surface brightness contours superimposed .
the lower left panel shows a color representation of the intensity weighted h@xmath0ivelocity field with iso - velocity contours superimposed .
finally , the lower right panel shows a color representation of the h@xmath0i velocity dispersion , again with h@xmath0i column density contours superimposed .
as revealed by the wsrt observations of van der hulst ( 1979a ) , the majority of the h@xmath0i is associated with ngc 4038 and the southern tail , with much less gas directly associated with ngc 4039 and the northern tail .
we detect tidal h@xmath0i emission over a velocity range of 1420 1850 km s@xmath8 , similar to that found in the synthesis observations of mahoney et al .
( 1987 ) , but less than found in the broader bandwidth observations of gordon et al .
we measure a total h@xmath0i flux of at least 54.50.5 jy km s@xmath8 for ngc 4038/9 , corresponding to a h@xmath0i mass of @xmath514.7 ( table [ tab : global ] ) .
this is not very different from the total flux measurement of 57.1 jy km s@xmath8 measured at the atca by gordon et al .
however , when we divide the emission among the different components we do find differences .
we measure a total flux of 49.7 jy km s@xmath8 associated with the main disks and southern tail , compared to the 55.1 jy km
s@xmath8 measured by gordon et al .. examining the channel maps of gordon et al .
, we find a mean flux level of @xmath14 10 mjy beam@xmath8 ( for a 40@xmath52 beam ) over the velocity range which we used for continuum subtraction .
this suggest that , due to our improper continuum subtraction , we miss a total of 6 jy km s@xmath8 over the entire 600 km s@xmath8 velocity range mapped by gordon et al . , which explains the different disk flux measurements .
we also find differences for the northern tail and companion galaxy , where we recover over twice the flux found by gordon et al ..
we next present details of the observations for each of the components individually , as well as for the tidal dwarf candidate(s ) within the southern tail . the majority of the the tidal gas is associated with the southern tail .
this tail extends to a projected radius of @xmath53 ( 65 kpc ) , and contains 2.8 of h@xmath0i along its entire length .
the column density slowly decreases with increasing distance along the tail , reaches a minimum , and increases again near its end .
perpendicular to the tail , the gas is distributed more broadly than the optical light , with a notable extension running along the outer ( southern ) edge of the tail .
there is no optical counterpart to this outer extension at our limiting surface brightness of @xmath54 27 mag arcsec@xmath46 ( see fig . [
fig : hidss ] ) , with a resulting h@xmath0i gas - to - light ratio of @xmath50@xmath51 1.5 @xmath55 , compared to @xmath140.5 @xmath55 on the optical tail .
the gas within the southern tail shows considerable structure on the scale of one beamwidth ( @xmath14 12 kpc ) .
this is further illustrated in figure [ fig : posvel ] , which displays three orthogonal projections of the intermediate resolution h@xmath0i datacube .
the upper left panel shows the sky view , @xmath56 , and to either side of this are the two orthogonal position - velocity profiles , @xmath57 ( bottom ) and @xmath58 ( right ) .
the @xmath59 view is displayed using a `` maximum voxel '' function .
this function assigns to each pixel the maximum intensity found along the velocity axis .
the position - velocity plots are constructed in the more traditional manner , summing the emission along one of the spatial dimensions of the datacube , either right ascension or declination .
we will use these plots below when examining the tail kinematics ( [ sec : tailkin ] ) the maximum voxel display emphasizes regions where the h@xmath0i is cold and dense ( maximum number of h@xmath0i atoms per unit velocity ) . as a result , it emphasizes the dynamically cold tidal tails .
note especially how the disk h@xmath0i , which has the highest integrated h@xmath0icolumn density , appears fainter in fig .
[ fig : posvel ] than in the zeroth moment maps ( fig .
[ fig : hidss ] ) .
this is because the disk gas is spread over a very broad range of velocities .
[ fig : posvel ] reveals a number of dense knots within the southern tail , especially toward its end .
the maximum voxel display ensures that this is not due to a line - of - sight integration effect where the tail bends back along our line - of - sight .
notice also that similar dense knots are not found within the northern tail .
the southern tail also exhibits an interesting parallel or `` bifurcated '' structure that starts where the southern tail begins to bend westward , and joins back together just before the location of the star forming regions identified by mirabel et al .
( see fig .
[ fig : tdgmos ] for the precise location of the star forming regions ) . the tail structure and kinematics
are further illustrated in the channel maps , shown in figure [ fig : allchan ] ( low resolution data after hanning smoothing by a factor of two in velocity to @xmath60=10.6 km s@xmath8 , contoured upon the optical ) and figure [ fig : greychan ] ( intermediate resolution data ; @xmath61=5.2 km s@xmath8 , shown as a greyscale ) .
only channels containing emission from the tidal tails are shown in these figures , as the disk emission is spread over a much broader range of velocities and are shown separately below ( [ sec : disk ] ) .
[ fig : allchan ] gives a more complete mapping of the gas within the tidal tails and shows its relationship to the underlying starlight , while fig .
[ fig : greychan ] allows a more detailed view of the structure of the gas within the tails . the greyscale representation of the channel maps in fig .
[ fig : greychan ] clearly shows the `` bifurcated '' morphology of the southern tail mentioned above .
the bifurcation is particularly apparent in the channels between 1760 and 1734 km s@xmath8 .
the two filaments merge together in a sideways `` v '' shape in the vicinity of the star forming regions identified by mdl92 ( panel at 1734 km s@xmath8 in fig .
[ fig : greychan ] ) .
the outer ( southern ) filament is associated with the previously described high @xmath50 material that lies off of the optical tail ( see also panels at 1729 1772 km s@xmath8 in fig .
[ fig : allchan ] ) .
the inner ( northern ) filament projects onto the optical tail and has a higher characteristic column density . the low resolution moment map ( fig .
[ fig : hidss ] ) shows that the northern h@xmath0i tail extends beyond the end of the optical tail , to a projected radius of 73 ( 40 kpc ) from the center of ngc 4039 .
this extension is confirmed by the atca observations of gordon et al .
we detect a total of 4.2 of h@xmath0i associated with the northern tail , which is much more than found in any of the previous observations .
the northern tail is redder ( fig .
[ fig : allmos]b ) , less luminous , and has a lower relative h@xmath0icontent than the southern tail ( table [ tab : global ] ) .
we also notice that the gas in the northern tail is relegated to the outer half of the optical tail and in a purely gaseous extension beyond this . in particular , there is a distinct gap in the h@xmath0idistribution from a projected radius of 25 kpc back to where the tail connects onto ngc 4039 .
the difference in the h@xmath0i content of the tails appears to be mirrored in the quantities of h@xmath0i within the disks associated with either tail ( with ngc 4038 being gas rich and ngc 4039 being gas poor ; see [ sec : disk ] below ) .
these differences and the lack of h@xmath0i at the base of the northern tail were already noted by van der hulst ( 1979a ) , who attributed them to a difference in the h@xmath0i content of the parent galaxies .
we will return to this point in [ sec : ionize ] velocities along the northern tail show a regular gradient from 1610 km s@xmath8 at the base of the h@xmath0i to 1540 km s@xmath8 at the end of optical tail , while the velocities along the southern tail range from 1590 , where the tail connects onto the northern disk , to a peak of 1770 km s@xmath8 midway along the tail , declining to 1610 km s@xmath8 toward the tip .
this kinematic continuity is convincingly illustrated in the position - velocity plots in fig .
[ fig : posvel ] , which also demonstrates that the tidal gas is quite dynamically cold . the smooth run of velocities along the tails is already apparent from the westerbork data of van der hulst ( 1979a ) , and is expected from tidal interaction models ( e.g. barnes 1988 , hibbard & mihos 1995 ) .
the well - behaved kinematics enable us to infer the approximate tail geometry .
for example , along the southern tail the h@xmath0i column density is lowest at the regions of the most extreme redshift , and increases to either side of this .
this indicates that this region of the tail lies perpendicular to our line - of - sight , with its motion directed most nearly toward us . at this location ,
the tail has a minimum projected thickness and hence lowest h@xmath0i column density .
to either side of this the tail curves toward our line - of - sight , resulting in a larger pathlength through the gas - rich tidal material and hence a higher projected column density .
the sharper velocity gradients to either side reflect the slewing of the velocity vectors as the tail bends away from us .
similarly , the sharp up - turn at the end of the southern tail in both space and velocity suggests that this region curves away from us . we will address the implications of this geometry for any embedded tidal dwarf galaxy in [ sec : tdg ]
since tails are kinematically expanding structures ( toomre & toomre 1972 , barnes 1988 ) , the relative velocities of the two tails indicate that the southern tail is moving away from us and connects back to ngc 4038 from behind while the northern tail is swinging slightly toward us and connects back to ngc 4039 from the front .
this geometry is supported by the lack of a dust absorption feature associated with the southern tail as it crosses from the south back to ngc 4038 .
in fact , the optical colors of these regions are quite blue ( fig .
[ fig : allmos]b ) , whereas we would expect red colors from dust absorption if this gas - rich material lies between us and the disk . as shown directly by van der
hulst ( 1979a ) , these kinematics are consistent with the prograde spin geometry derived by the early numerical models of this system by toomre & toomre ( 1972 ; see also barnes 1988 ; mihos , bothun & richstone 1993 ) .
following the southern tail kinematics back to its progenitor disk ( ngc 4038 , to the north ) , we see that the velocities become blueshifted toward the base of the tail ( fig .
[ fig : posvel ] ) .
since the predominantly redshifted velocities of the tail constrain it to be swinging away from us and connect back to the disk of ngc 4038 from behind , the blueshifted velocities at the base indicate material that is streaming back toward the disk from the tidal tail ( hibbard et al .
such streaming is a natural consequence of such encounters ( e.g. barnes 1988 , hernquist & spergel 1992 , hibbard & mihos 1995 ) , and arises because the tail material at smaller radii is more tightly bound than material further out . as a result , this material reaches its orbital apocenter more quickly and subsequently falls back toward the inner regions .
we can use the maximum projected tail length of 65 kpc and the disk rotation speed inferred from the fabry - perot observations of amram et al .
1992 ( 155 km s@xmath8 , corrected for disk inclination ) , to infer an interaction age ( measured from orbital periapse , when the tails were launched ) of 420 myr .
this is in excellent agreement with the best - match time of 450 myr from the numerical model of ngc 4038/9 run by barnes ( 1988 ) .
it is also remarkably close to the age of the population of 500 myr old globular clusters found by whitmore et al.(1999 ; see also fritze-v.alvensleben 1998 ) , supporting the hypothesis that this gc population formed around the time when the tails were first ejected ( whitmore et al .
1999 ) . from fig .
[ fig : allmos]d we note that the velocity dispersion within the tidal tails is actually quite low : except for the base of the southern tail , the h@xmath0i dispersion is less than 16 km s@xmath8 , and the mean of 7 km s@xmath8 is typical of values measured in undisturbed disk galaxies .
this is true of long - tailed mergers in general ( hibbard & van gorkom 1996 , hibbard & yun 1999a ) .
the dispersion near the base of the southern tail has higher characteristic values of 1325 km s@xmath8 .
these regions likely lie along a steeper gradient in the potential , and are also regions where the line - of - sight velocities are changing rapidly ( fig .
[ fig : allmos]c ) .
these effects will lead to larger real and apparent velocity gradients , increasing the single - beam dispersion .
the improved resolution of the vla observations allows us to directly address the behavior of the atomic gas in the vicinity of the putative tidal dwarf galaxy(s ) identified near the end of the southern tail of the antennae .
we show a close - up of this area in figure [ fig : tdgmos ] , which has the same arrangement as the panels in fig .
[ fig : allmos ] .
schweizer ( 1978 ) was the first to identify four h@xmath0ii regions and blue @xmath62 colors associated with the end of the southern tail .
he also identified a patch of low surface brightness material bending sharply to the north after the end of the optical tail , which he suggested was a separate dwarf stellar system .
we indicate the approximate location of this candidate tdg in fig .
[ fig : tdgmos ] using a thick yellow contour drawn at approximately one half of the local peak h@xmath0i column density after background subtraction and labeled tdg [ s78 ] .
schweizer noted that in the low - resolution h@xmath0i map of van der hulst ( 1979 ) , the h@xmath0i appears to be more closely concentrated in the low surface brightness extension of the tail , and hypothesized that the dwarf may have been created during the interaction , as envisioned by zwicky ( 1956 ) .
the end of the tail was subsequently studied by mirabel et al .
( 1992 ) , who concentrated on what they called a `` detached condensation of gas and stars at the tip of a tidal tail '' , consisting of a twisted stellar bar embedded in an envelope of diffuse optical emission and containing three emission line complexes .
we indicate the approximate location of this candidate tdg in fig .
[ fig : tdgmos ] using a thick white contour drawn at approximately one half of the local peak h@xmath0icolumn density after background subtraction and labeled tdg [ mdl92 ] .
the three star forming regions identified by mdl92 are labeled as regions i iii in fig .
[ fig : tdgmos]a , and are represented by black or white dots in the remaining panels .
while all subsequent observers have referred to the region identified by mdl92 when discussing the putative tdg , we will address the dynamical nature of both of these regions .
this is not a trivial task , as the precise boundaries of the tdg candidates have never been explicitly defined .
our h@xmath0i observations ( as well as those by mahoney et al .
1987 and gordon et al .
2001 ) are a vast improvement on the observations of van der hulst ( 1979 ) , and show that the tail is a continuous kinematic structure .
specifically , there is no `` detached '' region which can readily be identified as a distinct entity .
further , the velocity field varies smoothly through the region containing the high gas column densities and star forming regions without a significant twist or kink .
the strongest velocity gradient corresponds with the region where the tail turns abruptly to the north ( compressed contours near the center of the map in fig .
[ fig : tdgmos]c , with a similar gradient just to the east of this ) .
this smooth gradient suggests that the velocity field in this region is dominated by projection effects , with the tail bending away from us back into the plane of the sky ; as a result the velocity vector slews from pointing toward us to pointing away from us .
therefore , we find no clear kinematic evidence in the velocity field for any dynamically distinct entities .
it is possible that the strong tidal gradient may mask the kinematic signature of any mass concentration . since there are no well defined boundaries to the dwarf candidates , we calculate the h@xmath0i mass and blue luminosity in their vicinity by summing the emission within successive circular apertures centered on the local h@xmath0i column density peaks .
the maximum aperture radius is set by the half - width of the tail
. a mean background is subtracted , and the h@xmath0i flux is summed within each aperture of radius @xmath20 .
the optical luminosity is measured from the smoothed star - subtracted @xmath38-band within the same apertures .
the results of these calculations are shown in fig .
[ fig : sumiring ] .
the left panel shows the result for tdg [ mdl92 ] and the right panel shows the result for tdg [ s78 ] . in these plots ,
the solid and dashed lines show how the h@xmath0i mass and blue luminosity grow as a function of the aperture radius @xmath20 . in table
[ tab : tdg ] we list the total h@xmath0i mass and optical luminosity measured by these curves for each dwarf candidate , but since there are no distinct boundaries for either region , these values should be considered very rough guides .
the strong velocity gradients mentioned above give rise to the regions with the highest h@xmath0i velocity dispersions in fig .
[ fig : tdgmos]d , as gas with different space velocities fall within the same beam . aside from these regions , there are other local peaks in the gas velocity dispersion which might provide possible evidence for a mass concentration : between regions i & ii and region iii the h@xmath0i velocity dispersion increases to @xmath1412 km s@xmath8 vs. an average of 67 km s@xmath8 along the tail .
similar signatures are seen at the location of tidal dwarf candidates within the optical tails of the merger remnants ngc 7252 ( hibbard et al .
1994 ) and ngc 3921 ( hibbard & van gorkom 1993 , 1996 ) . however , it is noteworthy that the three h@xmath0iiregions actually fall on the edges of h@xmath0i maxima .
similar signatures are seen near giant h@xmath0ii regions in dwarf galaxies ( e.g. stewart et al.2000 , walter & brinks 1999 ) , where the increased h@xmath0i linewidth is due instead to kinetic agitation of the gas from energy deposited by young stars , sne and stellar winds ( e.g. stewart et al .
2000 , yang et al.1996 , tenorio - tagle & bodenheimer 1988 and references therein ) rather than the gravitational effects of a mass concentration .
there is also an increased dispersion to the north of tdg [ s78 ] ( dispersion increases to 15.6 km s@xmath8 over the surrounding value of 11 km s@xmath8 ) .
however , an examination of the line profiles shows that the increased dispersion comes from low - level emission spread over many channels .
adjacent regions have this broad low - level component , but also have a brighter narrower component which leads to a lower intensity - weighted line - width .
so we do not believe that this is a signature of a mass concentration at this location . since the moment maps are susceptible to line - of - sight integration effects , we turn to the velocity cubes directly for further insight . from the channel maps , we find that the gas near the tdg obtains the highest gaseous phase - space densities in this system ( i.e. density enhancements localized in space and velocity ) .
in fact , this region has a higher h@xmath0i column density per unit km s@xmath8 than any other region in this system , including regions within the disk or overlap region ( see also fig . [
fig : posvel]a ) .
the peak of 21 mjy beam@xmath8 in a single velocity channel of width 5.2 km s@xmath8 corresponds to a single channel column density of 3.8 , and the integrated column density of 1.3 represents an enhancement of 6 over the mean tail density .
the high - resolution data shows an even higher single - channel peak of 6.4 .
a closer examination shows that there are actually two dense knots in this region .
these are seen in the channel maps from the intermediate resolution datacube , plotted at full velocity resolution in fig .
[ fig : tdgchan ] .
the densest concentration appears at @xmath39=1708 km s@xmath8 and lies just north of star forming region iii .
the second concentration encompasses star forming regions i & ii and appears at @xmath39=17191729 km s@xmath8 .
such concentrations suggest regions of gaseous dissipation .
while this is not unequivocal evidence of a self - gravitating dwarf - sized object , since increased gaseous dissipation is expected in regions of recent star formation , it does indicate that there is something unique about this region of the tail .
the channel maps also show the bifurcated structure mentioned in [ sec : stail ] .
it is very interesting that the two parallel filaments join just east of the location of the star forming regions associated with the tdg ( the v "- shaped feature seen in the panel at 1734 km s@xmath8 of fig .
[ fig : tdgchan ] ) .
again , these observations suggest that there is something special about the gas and tail geometry at the location of the star forming regions .
the parallel filaments can be seen separately in the channels from v=17551724 km s@xmath8 .
the northern filament quite clearly lies along the optical tail , while the southern filament is displaced by about 8 kpc to the south .
the density peaks are of similar magnitude in either filament , but only at the location of the putative tdg is there a corresponding increase seen in the optical light underlying the gas peaks .
the gas concentration associated with region iii appears to lie within the northern filament , while it is difficult to say whether the concentration associated with regions i & ii are associated with one filament or the other . at velocities below 1693 km
s@xmath8 , it appears that only the southern filament continues on to the gas rich `` hook '' at the end of the tail .
[ fig : allmos]b shows that the blue tail colors discovered by s78 are not restricted just to the high column density gas near either tdg , but represents a trend of increasing blueness with distance along the tail .
we find that while the outer disks and southern tail have similar @xmath48 colors , the end of the southern tail is bluer in @xmath63 than @xmath64 while the reverse holds for the outer disks .
this suggests that the tail material is younger and/or more metal poor than the outer disk material ( schombert et al . 1990 ,
weilbacher et al .
2000 ) , although we postpone to a later paper a full color analysis ( evans et al .
2001 ) .
[ fig : diskmos ] shows an overlay of the high resolution h@xmath0icontours upon a false - color representation of the @xmath38-band image ( fig .
[ fig : diskmos]a ) , of the @xmath49 color map ( fig .
[ fig : diskmos]b ) , of the h@xmath0i velocity dispersion map ( fig .
[ fig : diskmos]d ) , and of the nir @xmath42-band image ( fig .
[ fig : diskmos]e ) . in fig .
[ fig : diskmos]c we show a color representation of the h@xmath0i velocity field with isovelocity contours drawn at intervals of 5 km s@xmath8 . finally , in fig . [
fig : diskmos]f we show the integrated co map of wilson et al .
( 2000 ) contoured in green upon a greyscale representation of the h@xmath0i column density .
the latter image gives a better idea of the local minima and maxima in the h@xmath0idistribution , as these are not always self - evident in the contour map . in this panel
red crosses indicate the location of the super star clusters " ( sscs ) identified in the hst image of whitmore et al .
( 1995 , from their table 1 ) . in fig .
[ fig : diskmos]b dark regions represent the reddest @xmath49 colors , generally indicating regions of high dust extinction , while light regions represent the bluest colors , generally indicating unobscured massive young stars .
comparison of figs .
[ fig : diskmos]b & f shows that the co is closely associated with the red features in the @xmath49 map , so that the co traces the dust distribution . figs .
[ fig : diskmos]b & f show that on a global scale , the h@xmath0idistribution traces both the regions containing the most star clusters and the regions with the highest dust content , e.g. the star forming ring in ngc 4038 and the star forming half of the disk of ngc 4039 ( to the northeast of the nucleus ) , and the dusty region between the two galaxies and to the southwest of ngc 4039 . on finer scales ( i.e. of order a few kpc ) , the h@xmath0i is not particularly associated with the bluest star clusters or reddest dust concentrations .
most of the sscs in fig .
[ fig : diskmos]f actually appear to lie along steep gradients in the h@xmath0i and co distributions .
the molecular gas peaks are often displaced from the @xmath65 peaks and many of the sscs actually lie in the transition zone between the two gas phases .
where there are no displaced co peaks , the sscs still lie predominantly along steep gradients in the h@xmath0i distribution .
the nir image in figs .
[ fig : diskmos]b does not show additional light peaks in the regions of highest h@xmath0i column density ( recall that dust extinction is 10 times weaker in the nir compared to the optical ) , so we conclude that the occurrence of sscs along gradients in the h@xmath0i distribution is not due to dust associated with the h@xmath0i obscuring clusters at the regions of highest h@xmath0i column density .
rather , we suspect that it is due to young stars ionizing or sweeping the h@xmath0i on the one hand , and the conversion of h@xmath0i to molecular form on the other .
we refer the reader to zhang , fall & whitmore ( 2001 ) for a much more thorough discussion of the relationship between ssc s of various ages and the h@xmath0i , radio continuum , co , x - ray , fir , mir , @xmath66 and optical morphology and the h@xmath0i and co kinematics .
the lower h@xmath0i contours in fig .
[ fig : diskmos ] are very close together , with the h@xmath0i column density falling off sharply below 4 .
the western part of the disk of ngc 4038 is largely devoid of h@xmath0i , as is most of the extended disk of ngc 4039 .
examination of the lower resolution h@xmath0i maps ( figs .
[ fig : hidss ] , [ fig : allmos]a ) shows that this is true even at lower h@xmath0i column densities .
probably not coincidently , these regions lack current star forming regions .
the intensity weighted h@xmath0i velocities of ngc 4038/9 ( fig . [ fig : diskmos]c ) shows a very disturbed velocity field , similar to those mapped in h@xmath0i ( gordon et al .
2001 ) h@xmath67 ( burbidge & burbidge 1966 , rubin et al . 1970 ,
amram et al .
1992 ) and in co ( stanford et al . 1990 ,
wilson et al .
2000 , zhu 2001 ) .
in particular , the disk of ngc 4038 has a north - south gradient spanning @xmath14 250 km s@xmath8 , while ngc 4039 has a northeast - southwest gradient spanning a similar range ( see also fig . [
fig : posvel ] ) .
the general sense of motion north blueshifted , south redshifted agrees with the sense of rotation inferred from the tidal tail .
however , because of the extremely disturbed nature of both disks , the kinematics of these regions are not simply rotational .
it is likely that there are strong streaming motions along the star forming loop and along the overlap region connecting the two disks . in figure
[ fig : diskchan ] we present the disk channel maps , made from the intermediate resolution data after hanning smoothing by a factor of four in velocity ( @xmath61=20.4 km s@xmath8 ) . instead of a normal rotational pattern ,
the redshifted gas concentration to the sw of ngc 4039 shows very little spatial gradient over the velocity range v=18361708 km s@xmath8 , while the blueshifted gas to the northeast of ngc 4039 has very little spatial gradient over the velocity range v=15791472 km s@xmath8 .
the gas not associated with either of these two regions appears to flow from the ne at v=1558 km s@xmath8 to the sw at v=1665 km s@xmath8 .
this is likely material which is flowing along the overlap region between the two galaxies
. the h@xmath0i in the northeast part of ngc 4039 may be material which is being accreted from ngc 4038 along the bridge connecting the two systems .
we also notice that the disk gas connects smoothly to the gas associated with the base of the southern tail ( channels at v=14721686 km s@xmath8 ) . from the tail kinematics ( [ sec : tailkin ] ) we concluded that the blueshifted velocities associated with the base of the tail require that this gas streams back onto the outer disk of ngc 4038 from the tail . in fig .
[ fig : diskmos]d we find a region of high velocity dispersion along the northern loop of gas ( dispersion of 28 km s@xmath8 vs 22 km s@xmath8 at adjacent regions ) .
we suggest that this increase is associated with gas falling back from the tail mixing along the line - of - sight with gas associated with the disk . as part of this observation
we also detect h@xmath0i associated with the scd dwarf galaxy eso 572g045 just to the southwest of the tip of the southern tail ( 5.7@xmath36 or 32 kpc from the tdg ) .
this system lies at a projected distance of 90 kpc from the antennae .
this galaxy is very flattened , with an axis ratio near 10 , associating it with the class of `` superthin '' galaxies ( goad & roberts 1981 ) , and qualifying it for membership in the `` flat galaxy catalogue '' of karachentsev et al .
it is curious that a similar superthin galaxy was also serendipitously discovered by duc et al . (
2000 ) in their hi mapping of the interacting system ngc 2992/3 , another interacting pair with a candidate tidal dwarf within a gas - rich tidal tail .
contours of the h@xmath0i column density from the various resolution data cubes are shown on the optical image from the dss in figure [ fig : esomom0 ] .
we detect a total 3.7 of h@xmath0i spread over 100 km s@xmath8 and extending to a radius of 6 kpc ( i.e. , twice as far as the optical light ) .
the galaxy is nearly edge - on and shows a linear velocity gradient .
assuming an edge - on orientation and circular velocities ( with @xmath68 = 50 km s@xmath8 ) , this suggests a dynamical mass ( @xmath69 ) of 3.5 out to 6 kpc . using the optical properties given in ned ( major and minor axis of @xmath70 and @xmath7116.70 mag )
, we derive dimensions of 5.9 kpc @xmath26 0.6 kpc and a luminosity of @xmath72 .
this system is quite gas rich ( @xmath73 ) , with global properties similar to those of other superthin galaxies ( matthews et al .
1999 ) .
the low resolution h@xmath0i map in fig .
[ fig : esomom0]a clearly shows some low column density gas extending to the east .
since this extension is one - sided in nature , it is more indicative of ram pressure effects than tidal effects .
the very thin and flat disk of eso 572g045 also argues against significant tidal forces acting on it .
this suggests that there may be an extended warm or hot gaseous halo around ngc 4038/9 that is affecting the diffuse cold ism of eso 572g045 ( e.g. moore & davies 1994 ) .
the wealth of information provided by these observations is a testament to the utility of h@xmath0i spectral line mapping of disturbed systems .
while in some sense the present data are just the beginning point of many more involved investigations , particularly for dynamical modeling of this system , they also lead to many interesting results on their own , which we discuss in this section .
the difference between the northern and southern tail in both optical structure and h@xmath0i content was already noted by van der hulst ( 1979a ) , who attributed it to a difference in the h@xmath0i content of the parent galaxies .
the lack of h@xmath0i within the main disk ngc 4039 ( i.e. the progenitor of the northern tail ) was seen as support for this interpretation .
however , we now feel that the observations are at odds with this explanation . for
while it is true that the outermost regions of tidal tails should originate from the outermost regions of the progenitor disks , it is not true that the outermost regions of the progenitor disks end up _ only _ at the outermost regions of tidal tails .
in fact , this gas rich material should extend all the way back to the host system .
this fact is apparent already in the simple models and illustrations of toomre & toomre 1972 ( see their fig . 2 ) .
this reflects the fact that tails are not linear structures , but actually two dimensional ribbons " twisting through space , with the outer disk material forming a sheath " around the inner disk material ( mihos 2001 ) . as a result
, the gas rich outer regions should extend along the entire length of the tidal feature back toward the progenitor disk .
it therefore seems that a low h@xmath0i gas content alone can not explain the gas - rich outer tail and gas - poor inner tail . in a recent paper , mihos ( 2001 ) describes a kinematically decoupling between the gaseous and stellar components of a tail that occur during tail formation .
however , this decoupling leads to a displacement between the gaseous and stellar tidal features , and does not totally remove large quantities of gas from along radial segments of the tail .
it is unlikely , therefore , that this mechanism explains the present observations .
hibbard , vacca & yun ( 2000 ) investigate several possible mechanisms for creating differences between h@xmath0i and optical tidal morphologies .
two plausible mechanisms are ram - pressure stripping of the tidal gas from an expanding superwind and ionization of the gas by uv photons escaping from the starburst . while there is some evidence for a nuclear outflow in ngc 4038/9 ( read , ponman , & wolstencroft 1995 , sansom et al .
1996 , fabbiano , schweizer & mackie 1997 , lipari et al.2001 , fabbiano , zezas & murray 2001 ) , the soft x - ray morphology ( read et al .
1995 , fabbiano et al .
2001 ) does not suggest that this wind is directed toward the northern tail , so we concentrate instead on possible ionization effects .
the situation explored by hibbard et al .
( 2000 ) is that the starburst has cleared enough dust and gas from the inner regions to provide direct unobscured sightlines to the tails . by equating the surface ionization rate with the recombination rate , they find that the gas should be ionized out to a distance given by : @xmath74 where @xmath65 and @xmath75 are the column density and thickness of the h@xmath0ilayer being ionized , @xmath76 is the fraction of ionized photons emerging from the starburst , and we have made the standard assumption that most of the starburst luminosity is emitted in the far infrared ( see hibbard et al .
2000 for details ) .
the fiducial values used in eqn . [ eq : rion ] are appropriate for ngc 4038/9 , and we thus find that the starburst may well be capable of ionizing gas out to the h@xmath0i base of the northern tail ( 25 kpc projected radius ) , providing the tail has a clear sightline to the ionizing stars and that @xmath76 is not much smaller than 0.05 .
evidence that @xmath76 is at least as large as this is provided by numerical work by dove , shull & ferrara ( 2000 ) , who derive @xmath77 for normal galaxies , and suggest that starbursts should have significantly higher values .
hurwitz , jelinsky , & dixon ( 1997 ) find upper limits to the escape fraction of @xmath78 photons of 357% for four ir luminous starbursts , while bland - hawthorn & maloney ( 1999 ) calculate @xmath79 for the milky way ( see discussion in bland - hawthorn & putman 2001 ) .
so the question becomes , is it likely that the northern tail of ngc 4039 has a relatively clear view of the star - forming disk , while the southern tail does not ? we believe the answer to this question is yes .
the majority of the unobscured massive star forming regions in ngc 4038/9 are associated with the disk of ngc 4038 and the material along the bridge ( whitmore & schweizer 1995 , whitmore et al .
since tidal tails are spun off close to the spin plane of the progenitor ( toomre & toomre 1972 ) , the southern tail will not have a clear view to the star forming regions located within the disk ngc 4038 , and the body of ngc 4039 would block ionization originating from the bridge star forming regions .
the body of ngc 4039 , on the other hand , appears to have a perfect view of the star forming disk of ngc 4038 and the northern tail should have a view to the backside of this disk .
this prediction can be tested by numerical models to constrain the space geometry of the tail .
we suggest that ngc 4039 had a normal h@xmath0i distribution , but that much of this gas has been photoionized by the starburst .
we believe a similar effect is responsible for the lack of h@xmath0i at smaller radii in the otherwise gas - rich tidal tails of the ngc 7252 , arp 105 , and arp 299 systems ( hibbard , vacca & yun 2000 ) .
if this is indeed the case , then the ionized gas should be visible via its recombination radiation .
using the equation given in hibbard et
al.(2000 ) , the expected emission measure of this radiation should be @xmath80 for a gas column density of 2 and thickness of 5 kpc .
this is within the capabilities of modern ccd detectors ( e.g. donahue et al .
1995 ; hoopes , walterbos & rand 1999 ) . as mentioned above
, tails are two dimensional ribbons " twisting through space .
it is possible that a lateral twist may cause the outer edge of the ribbon to lie in a different plane from the inner regions , i.e. for the most gas - rich ( highest @xmath50 ) regions to appear in a different plane from the less gas - rich but optically brighter regions ( mihos 2001 ) . in hibbard & yun ( 1999a )
we suggest that this this effect may be exacerbated by a pre - existing warp in the progenitor disk , providing even more marked offsets .
this provides a simple explanation for the gas - rich but optically faint outer edge of the southern tail .
similar bifurcated tails are seen in several systems ( e.g. m81 , van der hulst 1979b , yun et al . 1994 ; ngc 3921 , hibbard & van gorkom 1996 ; ngc 2535/6 , kaufman et al .
1997 ; arp 299 hibbard & yun 1999a ) .
what is intriguing is that the two filaments appear to join back together just at the location of the star forming regions associated with the candidate tidal dwarf tdg [ mdl92 ] ( the `` v''-shape seen in figs .
[ fig : greychan ] & [ fig : tdgchan ] and discussed in
[ sec : stail ] & [ sec : tdg ] ) .
we do not know whether this is a coincidence or an important clue .
further insight into this question will have to await a more detailed numerical study of this system .
we wish to take advantage of the higher spatial and velocity resolution of our h@xmath0i observations to examine the dynamical nature of material in the vicinity of the tdg candidates .
this is a difficult task , since there are no distinct objects which can be defined unambiguously .
further , if there are objects within the tail , they are clearly not in isolation , and the kinematics may not be due to the local mass concentration .
all of these factors make a clear dynamical analysis fraught with uncertainties .
still , the present data are the best available , in terms of both spatial and kinematic resolution , and this is the nearest system with a candidate tidal dwarf , so we will proceed with all due caution . the first hypothesis we will test is whether there are concentrations of gas and light centered on the gas peaks with enough mass in luminous matter alone to account for the measured h@xmath0i line - width . to this end , we compare the virial mass inferred from the h@xmath0i linewidth ( @xmath81 ) to the total luminous mass inferred from the h@xmath0i flux and the optical luminosity ( @xmath82 ) .
a value of @xmath83 is sufficient for a region to be bound , while a value of @xmath84 is required for the region to be virialized . to calculate the virial mass
, we must make some assumptions about the 3-dimensional structure of any mass concentration .
for the sake of simplicity , we assume that there is an isotropic spherical mass concentration of constant density .
for this situation , the virial mass is given by @xmath85 , where @xmath86 is the gravitational constant , @xmath25 is the one - dimensional velocity dispersion , @xmath87 is the projected half - light radius , and @xmath88 is a geometric factor ( binney & tremaine 1987 ) @xmath89 for the adopted geometry .
we take the velocity dispersion at the location of the peak column density , and use the deconvolved half - light radius found from fitting a gaussian plus constant background to the h@xmath0imass distribution .
the luminous mass is calculated from the observed mass and optical luminosity following the prescription given in paper ii .
in particular , the possible contribution to the luminous mass by stars is estimated by adopting a range of stellar mass - to - light ratios ( @xmath90 ) representing different star formation histories ( sfhs ) .
the gas mass is derived from the h@xmath0i mass by multiplying by a factor of 1.36 to take into account a primordial abundance of helium .
no correction is made for the presence of molecular gas since recent co observations at the location of the tdg by gao et al .
( 2001 ) and braine et al .
( 2001 ) find only trace amounts of molecular gas in the vicinity of the tidal dwarf ( 4 ) .
we evaluate three different stellar mass - to - light ratios , which span a range of star formation histories .
we evaluate @xmath91 0 @xmath55(i.e . ,
@xmath92 , also appropriate for a young instantaneous burst ) ; @xmath91 2 @xmath55 , corresponding to a constant sfh at an age of 10 gyr , and @xmath91 5 @xmath55 , representing an exponentially decreasing sfh with a time constant of 4 gyr at an age of 10 gyr ( see paper ii for more details ) .
the results of the luminous to virial mass calculation are given in table [ tab : tdg ] .
the ratios of the luminous to virial mass given in table [ tab : tdg ] are less than one , suggesting that neither mass concentration is virialized .
however , they do not rule out the presence of a bound object .
they do imply that there is not a sufficient gas density for the regions to be bound by gas alone ( i.e. , @xmath93 for @xmath91 0 @xmath55 ) .
this contradicts the conclusion reached for this same region by braine et al .
the reason for this discrepancy is that braine et al .
use the much narrow co linewidth to calculate a dynamical mass , yet include both h@xmath0i and stellar light in their luminous mass calculation ( as mentioned above , the mass in molecular mass inferred from the co line flux is negligible compared to the h@xmath0i ) .
the fact that the h@xmath0i linewidth is much broader than the co linewidth argues that a much larger mass concentration is required to bind it , and the braine analysis is inappropriate .
this point is discussed in more detail in paper ii .
a further conclusion from the above analysis is that if the dwarf candidates _ are _ bound , they require a significant contribution to the total mass from evolved stars , or they are bound by dark matter . in this case , the calculated values of @xmath94 are valid _ only _ if the h@xmath0i line - width also characterizes the motion of the stars ( and/or dark matter ) .
this is probably not the case , since in normal disks the older stellar populations ( i.e. , those with higher @xmath95 ratios ) have a significantly higher dispersion than the gas , by factors of 23 ( van der kruit 1988 ) , which would lead to a virial mass estimate a factor of 410 higher and @xmath94 ratios correspondingly smaller .
young stellar populations have a velocity dispersions similar to the cold gas , but also have smaller @xmath95 ratios .
a measurement of the stellar kinematics and a better estimate of the stellar mass - to - light ratio is needed to be more conclusive .
finally , we note that the expected circular velocity given by @xmath96 is 2030 km s@xmath8 .
[ fig : tdgmos]c shows no sign of rotation at this level .
we conclude that either tdg candidate may be , but need not be , self - gravitating .
this virial analysis includes many caveats .
primarily , the above analysis implies that we know the ratio @xmath94 to within a factor of two , and in truth we do not know either quantity to this precision . further , unless the bound regions are much smaller than the width of the tail , the geometry is unlikely to be spherical , and the velocities are unlikely to be isotropic .
finally , any entrained objects are clearly not in isolation , complicating the virial analysis .
we will be addressing these uncertainties using simulated observations of bound objects within n - body tails ( kohring , hibbard & barnes , in preparation ) .
the above analysis depends on a specific geometry and the details of the gaussian fit to the gas distribution . a more common but even less clear analysis uses a dynamical mass indicator @xmath97 , where @xmath98 is the fwhm linewidth ,
@xmath20 is a radius , and @xmath86 is the gravitational constant ( e.g. braine et al .
this quantity has the dimensions of mass , but it is not obvious what mass it defines , especially when the radius used is not physically motivated .
however , we can use the quantity to see how taking larger or smaller region in the vicinity of either dwarf candidate might affect our conclusions . specifically ,
we evaluate how the ratio of the luminous mass to this dynamical mass indicator changes with radius using the circular apertures defined in [ sec : tdg ] and the light curves given in fig .
[ fig : sumiring ] . in this case
, @xmath98 is 2.35 times the one - dimensional velocity dispersion , and @xmath20 is the deconvolved aperture radius .
the luminous mass is calculated from the h@xmath0i mass and optical luminosity as above .
the results of this exercise is shown in fig .
[ fig : mlummdyn ] .
three curves are drawn , each one representing a different stellar mass - to - light ratio : the solid curve shows the result for @xmath99 @xmath55 , the dotted curve shows the results for @xmath100 @xmath55 , and the dashed line shows the results for @xmath101 @xmath55 .
we see that this ratio decreases with larger apertures .
therefore , taking a larger radius will not change the conclusions reached above .
taking a smaller radius might turn up objects with a larger luminous to dynamical mass , but in this case they would have a much smaller mass scale than is usually assumed . in conclusion
, we find no kinematic signature of a distinct dynamical entity within the southern tail , but any signature might be masked by the strong geometric velocity gradients in this region . if there are bound mass concentrations , they require a significant contribution from either evolved stars or dark matter to bind them . if the lower distance of 13.8 mpc suggested by saviane et al .
( 2001 ) is correct , then the luminous mass decreases by a factor of 2 while the virial mass decreases by a factor of 1.4 , so the luminous to virial mass estimates all decrease by a factor of 1.4 , making it less likely that either tdg candidate is a bound object .
additionally , if there are bound regions , the mass scale would be more typical of dwarf spheroidal galaxies than dwarf irregulars , and the gaseous concentrations may have more relevance to remnant streams around galaxies than the formation of dwarf irregulars ( kroupa 1998 ) . if there are dwarf - galaxy sized concentrations forming within the tail , then the present observations appear inconsistent with the tidal dwarf formation scenario suggested by elmegreen , kaufmann & thomasson ( 1992 ) . in their scenario
the velocity dispersion of tidal gas is increased over the pre - encounter value due to the interaction , increasing the jeans mass in the tidal gas and leading to the formation of dwarf irr sized objects . as noted in [ sec : tailkin ] , the velocity dispersion within the tidal tails of ngc 4038/9 is typical of values measured in undisturbed disk galaxies .
this is true of long - tailed mergers in general ( hibbard & van gorkom 1996 , hibbard & yun 1999a ) and agrees with the expectation from numerical models . the tails form from the dynamically coldest material on the side of the disk opposite orbital periapses , and expand kinematically thereafter .
tails , particularly those of low - inclination prograde encounters , should not and appear not to be significantly kinematically heated by the encounter . and
while it is true that an increased velocity dispersion will suppress the formation of low mass objects , it does not follow that it should lead to the collapse of higher mass objects ; it simply means that a higher mass is required in order for the concentration to collapse .
therefore , the model of tidal dwarf formation suggested by elmegreen , kaufmann & thomasson ( 1992 ) does not appear to apply here .
a last point that we wish to make is that ngc 4038/9 is clearly an on - going merger experiencing a massive young starburst which is spread throughout the disk of ngc 4038 and the overlap region . and
yet , the long tails of ngc 4038/9 suggest that the interaction began in earnest of order 400 myr previously .
this is very similar to the situation in the infrared luminous merger arp 299 ( hibbard & yun 1999a ) .
additionally , the proximity of ngc 4038/9 have allowed the identification of a population of sscs with ages of @xmath14 500 myr ( whitmore et al .
1999 ) spread throughout the disk region .
clearly , then , merger induced extra - nuclear star - formation began around the time that the tails were launched , has continued until today , and will likely continue until the merger is complete . therefore , the merger induced
burst " population will be broadly spread throughout the remnant , both spatially and temporarily .
while a significant amount of the present gas supply may find its way into the innermost regions of the remnant , it is not clear how much of this gas will be converted into stars .
much of this gas may be expelled in starburst driven superwind ( heckman et al . 1999 ,
hibbard & yun 1999b ) . either way
, the expected signature of the evolved merger remnant will not be as pronounced as is often assumed ( e.g. , mihos & hernquist 1994 , bekki & shioya 1997 , silva & bothun 1998a , b ) . instead of a single epoch
metal - enriched population confined to the innermost radii of the remnant and surrounded by a uniformly - old metal poor population , there will be stars with an age range of @xmath14 1 gyr spread throughout remnant body , with perhaps an inner population of stars formed from whatever gas is not blown out .
* we have mapped the h@xmath0i in the classical double - tailed merger , ngc 4038/9 , `` the antennae '' , with unprecedented spatial and velocity resolution . in agreement with other observers ( van der hulst 1979a , mahoney et al .
1987 , gordon et al .
2001 ) , we find the northern tail to be gas - rich at its tip but gas - poor at its base , the southern tail to be gas rich along its entire length and with a significant enhancement in the gas concentration in the vicinity of the tdg .
the tail kinematics are broadly consistent with the interaction geometry inferred from numerical simulations .
* we suggest that the lack of h@xmath0i at the base of the northern tail is due to photoionization by uv photons escaping the disk starburst region .
if so , the ionized gas should be readily detectable , with and expected emission measure of order 0.8 @xmath102 pc . * the h@xmath0i velocity field at the end of the southern tail
is dominated by strong velocity gradients which suggest that at this location the tail is bending away from us .
the tail velocity gradients may mask the kinematic signature of any self - gravitating condensation in this region . *
it is not clear whether the tdg candidates identified by schweizer ( 1978 ) and mirabel et al .
( 1992 ) are self - gravitating .
the observed kinematics suggest that there is not enough mass in gas alone to account for the h@xmath0i linewidths of the regions we have delineated .
further insight requires a measurement of the stellar velocity dispersion and tighter constraints on the stellar mass - to - light ratio . * whether or not the tdg candidate identified by mirabel et al .
( 1992 ) is a self - gravitating entity , it is clear that this concentration of gas , stars , and star forming regions in the tail is unique . at this location the h@xmath0i column densities per channel are higher than anywhere else in the system , including within the main disks .
we believe a clue to the origin of this concentration is that it occurs just after the the `` v''-shape where the two filaments connect back to each other occurs just before the location of of the tdg . *
these observations reveal that the `` superthin '' galaxy eso 572g045 is a companion to ngc 4038/9 .
our low resolution data show a low column density h@xmath0i extension to the east , suggesting that this system may be experiencing ram pressure as it orbits ngc 4038/9 .
we thank baerbel koribalski for conversations and sharing of the gordon et al .
atca results prior to publication , bill vacca for running population synthesis models for us , christine wilson and vassilis charmandaris for sharing their co map and allowing us to reproduce it in fig .
[ fig : diskmos]f , and the referee , pierre - alain duc , for a helpful referees report .
we also thank ed fomalont and juan uson for advice on self - calibration and mapping , raja guhrathakurta for help with the optical data and reductions , pat smiley at nrao for help with the color figures , and francois schweizer and jacqueline van gorkom for helpful discussions .
the karma visualization package is highly recommended and is available at _ http://www.atnf.csiro.au / karma/_. amram , p. , marcelin , m. , boulesteix , j. , & le coarer , e. 1992 , a&a , 266 , 106 barnes , j.e .
1988 , apj , 331 , 699 barnes , j. e. , & hernquist , l. 1992 , nature , 360 , 715 bekki , k. , & shioya , y. 1997 , apj , 478 , l17 binney , j. , & tremaine s. 1987 , `` galactic dynamics '' , ( princeton : princeton university press ) .
bland - hawthorn , j. , & maloney , p.r . 1999 ,
apj , 510 , 33 bland - hawthorn , j. , & putman , m. 2001 , in asp conf .
240 , gas and galaxy evolution , eds .
hibbard , m.p .
rupen & j.h .
van gorkom ( asp , san francisco ) , 369 bosma , a. 1981 , aj , 86 , 1791 * * braine , j. , duc , p .- a . ,
lisenfeld , u. , leon , s. , vallejo , o. , charmandaris , v. & brinks , e. 2001 , a&a , in press .
braine , j. , lisenfeld , u. , duc , p .- a . , & leon , s. , 2000 , nature , 403 , 867 briggs , d. 1995 , phd thesis , nmimt burbidge , e. m. , & burbidge , g. r. 1966 , apj , 145 , 661 bushouse , h. a. 1987 , apj , 320 , 49 bushouse , h. a. , lamb , s. a. , & werner , m. w. 1988 , apj , 335 , 74 dove , j. b. & shull , j. m. & ferrara , a. 2000 , apj , 531 , 846 donahue , m. , aldering , g. , & stocke , j. t. 1995 , apj , 450 , l45 duc , p. -a . , mirabel , i. f. 1994 , a&a , 289 , 83 duc , p. -a . , brinks , e. , wink , j. e. , & mirabel , i. f. 1997 , a&a , 326 , 537 duc , p. -a . ,
brinks , e. , springel , v. , pichardo , b. , weilbacher , p. , & mirabel , i.f . , 2000 ,
aj , 120 , 1238 elmegreen , b. , kaufmann , m. , & thomasson , m. 1993 , apj , 412 , 90 evans , r. , harper , a. , & helou , g. 1997 , in extragalactic astronomy in the infrared , eds . g.a .
mamon , t.x .
thuan & j.t .
thanh van .
( paris : ed .
frontieres ) , 143 * * * evans , r. et al .
2001 , in preparation fabbiano , g. , schweizer , f. , & mackie , g. 1997 , apj , 478 , 542 fabbiano , g. , zezas , a. , & murray , s.s .
2001 , apj , 554 , 1035 fischer , j. et al .
1996 , a&a , 315 , l97 fritze-v.alvensleben , u. 1998 , a&a , 336 , 83 gao , y. , lo , k. y. , lee , s .- w . , & lee , t .- h .
2001 , apj , 548 , 172 goad , j. w. , & roberts , m. s. 1981 , apj , 250 , 79 gooch , r.e .
, 1995 , in asp conf .
series vol . 101 , astronomical data analysis software and systems v , eds g.h .
jacoby & j. barnes ( asp , sf ) , 80 gordon , s. , koribalski , b. , & jones , k. 2001 , mnras , 326 , 578 graham , j. a. 1982 , pasp , 94 , 244 haas , m. , klaas , u. , coulson , i. , thommes , e. , & xu , c. 2000 , a&a , 356 , l83 heckman , t. m. , armus , l. , weaver , k. & wang , j. 1999 , apj , 517 , 130 hernquist , l. , & spergel , d. n. 1992 , apj , 399 , l117 hibbard , j. e. , & van gorkom , j. h. 1996 , aj , 111 , 655 hibbard , j. e. , & van gorkom , j. h. 1993 , in asp conf . series , vol .
48 , the globular cluster galaxy connection , eds .
g. h. smith and j. p. brodie ( asp , san francisco ) , 619 hibbard , j. e. , guhathakurta , p. , van gorkom , j. h. , & schweizer , f. 1994 , aj , 107 , 67 hibbard , j.e . , & mihos , j.c .
1995 , aj , 110 , 140 .
hibbard , j. e. , vacca , w. d. , & yun , m. s. 2000 , aj , 119 , 1130 .
hibbard , j. e. , & yun , m. s. 1999a , aj , 118 , 162 hibbard , j. e. , & yun , m. s. 1999b , apj , 522 , l93 hoopes , c. , walterbos , r. & rand , r. 1999 , apj , 522 , 669 huchtmeier , w. k. , & bohnenstengel , h .- d .
1975 , a&a , 41 , 477 hummel , e. , & van der hulst , j. m. 1986 , a&a , 155 , 151 hunsberger , s. , charlton , j. , & zaritsky , d. 1996 , apj , 462 , 50 hunsberger , s. , charlton , j. , & zaritsky , d. 1998 , apj , 505 , 536 hurwitz , m. , jelinsky , p. , & dixon , w. 1997 , apj , 481 , l31 iglesias - p ' aramo , j. , & v ' ilchez , j. m. 2001 , apj , 550 , 204 karachentsev , i. d. , karachentseva , v. e. , & parnovskij , s. l. 1993 , _ astronomische nachrichten _ , 314 , 97 kaufman , m. , brinks , e. , elmegreen , d. m. , thomasson , m. , elmegreen , b. g. , struck , c. & klaric , m. 1997 , aj , 114 , 2323 kennicutt , r. c. jr . 1989 .
apj , 344 , 685 kennicutt , r. c. jr .
, keel , w. c. , van der hulst , j. m. , hummel , e. , roettiger , k. a. , 1987 , aj , 93 , 1011 * * knierman , k. a. , gallagher , s. c. , charlton , j. c. , hunsberger , s. d. , whitmore , b. , kundu , a. , hibbard , j. e. , & zaritsky , d. 2001 , aj , in press kroupa , p. 1998 , mnras , 300 , 200 kunze , d. et al . 1996 ,
a&a , 315 , l101 laurent , o. , mirabel , i.f . ,
charmandaris , v. , gallais , p. , madden , s.c . , sauvage , m. , vigroux , l. , & cesarsky , c. 2000 , a&a , 359 , 887 * * lipari , s. , sanders , d. , terlevich , r. , veilleux , s. , diaz , r. , taniguchi , y. , zheng , w. , kim , d. , tsvetanov , z. , carranza , g. , & dottori , h. 2001 , apj , submitted ( astro - ph/0007316 ) mahoney , j. m. , burke , b. f. , van der hulst , j. m. 1987 , in iau symp .
117 , dark matter in the universe , eds .
j. kormendy , g.r .
knapp ( reidel , dordrecht ) , 94 matthews , l. d. , gallagher , j. s. iii , & van driel , w. 1999 , aj , 118 , 2751 ( superthins ) mengel , s. , lehnert , m. d. , thatte , n. , tacconi - garman , l. e. , & genzel , r. 2001 , apj , 550 , 280 mihos , j. c. 2001 , apj , 550 , 94 mihos , j. c. , bothun , g. d. , richstone , d. o. 1993 , apj , 418 , 82 mihos , j. c. , hernquist , l. 1994 , apj , 437 , l47 mirabel , i. f. , dottori , h. , & lutz , d. 1992 , a&a , 256 , l19 mirabel , i. f. , vigroux , l. , charmandaris , v. , sauvage , m. , gallais , p. , tran , d. , cesarsky , c. , madden , s. c. , & duc , p .- a . 1998 , a&a , 333 , l1 moore , b. , & davis , m. 1994 , mnras , 270 , 209 neff & ulvestad 2000 , aj , 120 , 670 nikola , t. , genzel , r. , herrmann , f. , madden , s.c . ,
poglitsch , a. , geis , n. , townes , c.h . , & stacey , g.j .
1998 , apj , 504 , 749 peterson , s. d. , & shostak , g. s. 1974 , aj , 79 , 767 read , a. m. , ponman , t. j. , & wolstencroft 1995 , mnras , 277 , 397 rubin , v. c. , ford , w. k. jr , & dodorico , s. 1970 , apj , 160 , 801 rupen , m. p. 1999
, in _ synthesis imaging in radio astronomy _ , asp conf .
series vol .
g. b. taylor , c. l. carilli and r. a. perley ( asp , san francisco ) , 229 sansom , a. e. , dotani , t. , okada , k. , yamashita , a. , & fabbiano , g. 1996 , mnras , 281 , 48 * * * saviane , i. , rich , r.m.r . ,
& hibbard , j.e .
2001 , in preparation schombert , j. m. , wallin , j. f. , & struck - marcell , c. 1990 , aj , 99 , 497 schweizer , f. 1978 , the structure and properties of nearby galaxies , iau symp .
77 , edited by e. m. berkhuijsen and r. wielebinski ( reidel , dordrecht ) , 279 schweizer , f. 1998 , in saas - fee advanced course no . 26 , galaxies : interactions and induced star formation , r. c. kennicutt jr . , f. schweizer and j. e. barnes ( springer , berlin ) , 105 silva , d. r. , & bothun , g. d. 1998a , aj , 116 , 85 silva , d. r. , & bothun , g. d. 1998b , aj , 116 , 2793 smith , b. j. , & higdon , j. l. 1994 , aj , 108 , 837 smith , b. j. , & struck , c. 2001 , apj , 121 , 710 stanford , s. a. , sargent , a. i. , sanders , d. b. , & scoville , n. z. 1990 , apj , 349 , 492 stewart , s. g. , fanelli , m. n. , byrd , g. g. , hill , j. k. , westpfahl , d. j. , cheng , k .- p . , oconnell , r. w. , roberts , m. s. , neff , s. g. , smith , a. m. , & stecher , t. p. 2000 , apj , 529 , 201 taff , l. g. , lattanzi , m. g. , bucciarelli , b. , gilmozzi , r. , mclean , b. j. , jenkner , h. , laidler , v. g. , lasker , b. m. , shara , m. m. , & sturch , c. r. 1990 , apj , 353 , l45 tenorio - tagle , g. , bodenheimer , p. 1988
, ara&a , 26 , 145 toomre , a. , & toomre , j. 1972 , apj , 178 , 623 van der hulst , j. m. 1979a , a&a , 155 , 151 van der hulst , j. m. 1979b , a&a , 75 , 97 van der kruit , p. c. 1988 , a&a , 192 , 117 vigroux , l. et al .
1996 , a&a , 314 , l93 wainscoat , r. j. , & cowie , l. l. 1992 , aj , 103 , 332 walter , f. , & brinks , e. 1999 , aj , 118 , 273 weilbacher , p. m. , duc , p .- a . , fritze-v.alvensleben , u. , martin , p. , & fricke , k. j. 2000 , a&a , 358 , 819 whitmore , b. c. , & schweizer , f. 1995 , aj , 109 , 960 whitmore , b. c. , zhang , q. , leitherer , c. , fall , s. m. , schweizer , f. , miller , b. w. , 1999 , aj , 118 , 1551 wilson , c. d. , scoville , n. , madden , s. c. , & charmandaris , v. 2000 , apj , 542 , 120 xu , c. , gao , y. , mazzarella , j. , lu , n. , sulentic , j. w. , & domingue , d.l .
2000 , apj , 546 , 644 yang , h. , chu , y .- h . ,
skillman , e. d. , & terlevich , r. , 1996 , aj , 112 , 146 yun , m. s. 1999 , in iau symp .
186 , galaxy interactions at low and high redshift , ed .
d. sanders & j. barnes ( city : publisher ) , 81 yun , m. s. , ho , p. t. p. , & lo , k. y. 1994 , nature , 372 , 530 * * zhang , q. , fall , s. m. , & whitmore 2001 , apj , in press ( astro - ph/0105174 ) zhu , m. 2001 , ph.d .
thesis , university of toronto zwicky , f. 1956 , _ ergebnisse der exakten naturwissenschaften _
, 29 , 344 llll date & & 1996 jun 3 & 1997 jun 9 array & & d & cnb time on source ( hrs ) & & 2.5 & 6.5 phase center ( @xmath67 , @xmath103 , j2000 ) & & 12@xmath104 01@xmath37 52@xmath105 & 18@xmath106 55@xmath36 40@xmath52central velocity ( heliocentric ) & & 2 1630 km s@xmath8 primary beam ( fwhm ) & & 2 30@xmath36 phase calibrator & & 2 1130148 flux calibrator & & 2 3c286 correlator mode & & 2 1a , on - line hanning bandwidth ( mhz ) & & 2 3.125 number of channels & & 2 128 channel separation ( km s@xmath8 ) & & 2 5.21 data set & high resolution & medium resolution & low resolution robust"@xmath107 parameter & -1 & + 1 & + 1 , convolved to 40@xmath52 synthesized beam major axis @xmath26 minor axis ( fwhm ) & @xmath108 & @xmath109 & @xmath110 position angle ( east of north ) & 76@xmath106 & 24@xmath106 & 0@xmath106 noise level ( 1@xmath25 ) flux density ( mjy beam@xmath8 ) & 1.3 & 0.9 & 1.3 column density ( beam@xmath8 ch@xmath8 ) & 8.8 & 1.6 & 0.47 brightness temperature ( k beam@xmath8 ) & 9.3 & 1.7 & 0.49 [ tab : hiobs ] lll date & 1991 apr 17/18 & 1995 jan 22 telescope & ctio 0.9 m & uh 88@xmath52 camera & tek 512 & quirc readout mode & unbinned & unbinned focal ratio & f/13.5 & f/7.5 field of view : & & single ccd frame & @xmath111 & @xmath112 final image & 17@xmath36@xmath2617@xmath36 & 4@xmath36@xmath264@xmath36 pixel size & @xmath113 & @xmath114 filters & @xmath115 & @xmath32 ( 2.15 @xmath116 m ) seeing & @xmath117 & @xmath118 sky brightness ( mag arcsec@xmath46 ) & 22.1 , 21.3 , 20.6 & @xmath119 sky noise , binning 1@xmath261 ( mag arcsec@xmath46 ) & 25.5 , 25.3 , 25.0 & @xmath120 surface brightness limit , binning 9@xmath269 ( mag arcsec@xmath46 ) & 27.0 , 26.8 , 26.5 & effective exposure time & @xmath142@xmath26600@xmath0s & 3@xmath26120@xmath0s [ tab : optobs ] lrrrrrr disk@xmath121 & @xmath5117.20.16 & @xmath122 & @xmath5114201850 & 9 & @xmath123 & @xmath510.06 s tail & 32.50.23 & @xmath124 & 15901770 & 65 & @xmath125 & 0.8 n tail & 4.80.10 & @xmath126 & 15401600 & 40 & @xmath127 & 0.3 total ngc 4038/9 & @xmath5154.50.49 & @xmath128 & @xmath5114201850 & 110 & @xmath129 & @xmath510.2 eso 572g045 & 4.30.10 & @xmath130 & 16401755 & 88 & @xmath131 & 1.9 [ tab : global ] lrr right ascension@xmath107 & 12@xmath104 01@xmath37 25.7@xmath105 & 12@xmath104 01@xmath37 22.1@xmath105 declination@xmath107 & 19@xmath106 00@xmath36 42@xmath52 & 19@xmath106 00@xmath36 12@xmath52background@xmath107 + @xmath65 ( atoms cm@xmath46 ) & 3.8 & 2.9 + @xmath45 ( mag arcsec@xmath46 & 26.0 & 26.3 + half - light radius@xmath107 , @xmath87 ( deconvolved ) & 3.2 kpc & 3.5 kpc + h@xmath0i velocity dispersion@xmath133 , @xmath25 & 13.4 km s@xmath8 & 10.5 km s@xmath8 + @xmath134 & 2.4 & 1.7 + @xmath135 & 8.7 & 5.3 + @xmath81 & 1.1 & 7.4 + @xmath136 $ ] & 0.3 & 0.3 + @xmath137 $ ] & 0.4 & 0.5 + @xmath138 $ ] & 0.7 & 0.7 + [ tab : tdg ] | we present new vla c+d - array h@xmath0i observations and optical and nir imaging of the well known interacting system ngc 4038/9 , `` the antennae '' . at low spatial resolution ( @xmath1 ) ,
the radio data reach a limiting column density of @xmath2 @xmath3 , providing significantly deeper mapping of the tidal features than afforded by earlier h@xmath0i maps . at relatively high spatial resolution ( @xmath4 ) ,
the radio data reveal a wealth of gaseous sub - structure both within the main bodies of the galaxies and along the tidal tails . in agreement with previous h@xmath0i studies , we find that the northern tail has h@xmath0i along its outer length , but none along its base .
we suggest that the h@xmath0i at the base of this tail has been ionized by massive stars in the disk of ngc 4038 .
the gas in the southern tail has a bifurcated structure , with one filament lying along the optical tail and another running parallel to it but with no optical counterpart .
the two filaments join just before the location of several star forming regions near the end of the tail .
the h@xmath0ivelocity field at the end of the tail is dominated by strong velocity gradients which suggest that at this location the tail is bending away from us .
we delineate and examine two regions within the tail previously identified as possible sites of a so - called `` tidal dwarf galaxy '' condensing out of the expanding tidal material .
the tail velocity gradients mask any clear kinematic signature of a self - gravitating condensation in this region .
a dynamical analysis suggest that there is not enough mass in gas alone for either of these regions to be self - gravitating .
conversely , if they are bound they require a significant contribution to their dynamical mass from evolved stars or dark matter .
even if there are no distinct dynamical tidal entities , it is clear that there is a unique concentration of gas , stars and star forming regions within the southern tail : the h@xmath0ichannel maps show clear evidence for a significant condensation near the tail star forming regions , with the single - channel h@xmath0i column densities higher than anywhere else in the system , including within the main disks .
finally , the data reveal h@xmath0i emission associated with the edge - on superthin " scd galaxy eso 572g045 that lies just beyond the southern tidal tail of the antennae , showing it to be a companion system .
= cmcsc10 at 12pt = cmr12 = cmr10 # 1@xmath5 # 1@xmath6 |
it is now well established that organic conductors @xmath0-(bedt - ttf)@xmath1i@xmath2 @xcite , exhibit a zero gap - state ( zgs ) characteristic of the presence of massless dirac electrons close to the fermi energy @xcite .
however , compared to graphene , the dirac cones are strongly tilted resulting in a strong anisotropy of the dirac electron velocity components .
furthermore , the fact that @xmath0-(bedt - ttf)@xmath1i@xmath2 has a unit cell composed of four inequivalent molecules leads to some unique physical properties associated to this zgs .
in particular , as revealed by recent nmr measurements , the anisotropy of the dirac particles appears strongly related to their local spectral weight on the four inequivalent molecules per unit cell ; such that the smallest ( largest ) velocity dirac particles have their largest weight on the electron rich ( poor ) molecule site @xcite . from a theoretical point of view
, the electronic properties of @xmath0-(bedt - ttf)@xmath1i@xmath2 are usually described by a tight - binding model with seven nearest - neighbor transfer integrals @xmath12 between the four molecules a , a , b , and c of the unit cell . as illustrated on figure fig .
[ fig : structure ] , this model presents an inversion symmetry with inversion centers located on b , c and the middle point between a and a.@xcite .
thank s to this inversion symmetry , recent studies have used fu - kane topological argument in order to determine quantitatively the existence domain of dirac electrons from the knowledge of the energy and inversion parity eigenvalues at the four time reversal invariant momenta ( trim ) . despite their usefulness , these studies do not allow to distinguish zgs systems in which dirac electrons are at the fermi energy ( with a vanishing density of states at the fermi energy ) from metallic systems in which there is an indirect overlapp between valence and conduction band such that the dirac electrons are below ( or above ) the fermi energy .
the purpose of the present work is to clarify the relative role played by the the inhomogeneity in the moduli and in the signs of the seven transfer integrals on the appearance and stability of dirac electrons with zgs .
in fact as explained in more details below , for given moduli @xmath4 , the possibility of having dirac electron with a zgs at @xmath5 ( or @xmath6 ) filling strongly depends on the specific configurations of signs of the seven transfer integral .
more precisely it is possible to classify the sign configurations into essentially four classes determined by and @xmath7 . using extensive numerics ,
we show that for weak , large and no inhomogeneity in the moduli , the class @xmath8 is most favorable to find dirac electrons with zgs at @xmath5 ( or @xmath6 ) filling .
the classes @xmath10 and @xmath11 constitute equivalent but less favorable situation for the occurrence of dirac electrons with zgs at @xmath5 ( or @xmath6 ) filling .
finally , the last class is @xmath9 and for the range of inhomogeneity that was explored no zgs was obtained for either @xmath6 or @xmath5 filling .
the paper is organized as follows . in
2 we present the model and the main quantities of interest . in
3 we first introduce the chirality patterns and their properties and then using properties established in appendix a we explained the relation between the chirality patterns and the energy band structure and local density of states . in
4 we present our extensive numerical studies of the existence of zgs at @xmath5 filling for the three pertinent classes of chirality patterns . in 5 we summarized our main results .
-(bedt - ttf)@xmath1i@xmath2 , with four molecules a , a , b , and c per unit cell ( dotted square ) .
inversion centers are located on sites b , c , and the middle point between a and a.@xcite .
the seven nearest - neighbor transfer integrals are shown by @xmath13 .
the chirality of four distinct triangular plaquettes is shown by @xmath14 , @xmath15 , @xmath16 , and @xmath17 .
, width=264 ]
the starting point is the @xmath18 bloch hamiltonian matrix @xmath19 corresponding to the real space tight - binding model shown in fig . [
fig : structure ] : @xmath20 with @xmath21 , @xmath22 , @xmath23 and @xmath24 ( the lattice constant is taken as unity ) .
the boch basis @xmath25 ( @xmath26 ) was chosen such that the hamiltonian matrix is bloch periodic @xmath27 with @xmath28 a reciprocal lattice vector . in this bloch basis , the energy eigenstate @xmath29 of the @xmath30 band is represented by a four components vector @xmath31 such that @xmath32 where by convention @xmath33 .
the local density of states ( ldos ) on each site is then obtained as ( @xmath26 ) @xmath34 for @xmath5 filling , the existence of a pair of dirac points at @xmath35 implies @xmath36 with @xmath37 ; the existence of a zgs further necessitates that @xmath38 for any @xmath39 .
in that situation the ldos verify @xmath40 for @xmath41 near @xmath42 where @xmath43 is the site dependent angular averaged fermi velocity .
for fixed moduli ( @xmath44 ) , the energy bands @xmath45 and the eigenfunction components @xmath46 have the freedom of the sign of the seven hopping parameters , as a result one may expected to obtain @xmath47 distinct ldos @xmath48 . as explained in [ ldos ] , the number of distinct ldos @xmath48 is entirely determined by the number of patterns of the four chiralities @xmath49 ( @xmath50 ) @xmath51 where each chirality @xmath52 encodes the sign of the product of the transfer integrals for each of the four distinct triangular plaquette that compose the unit cell . the are in total @xmath53 chirality patterns @xmath54 which are summarized in table [ table_1 ] .
interestingly it appears that the chirality patterns are not all _ independent _ , more precisely one can essentially define five classes of patterns with independent real space _ topology _ : @xmath55 , @xmath56 , @xmath57 , @xmath58 , @xmath59 . as an illustration of their distinct _ topology _ , the chirality patterns associated to @xmath60 are represented on fig .
[ patterns](a), .. (e ) .
each class is characterized by the fact that it is globally invariant under the following three transformations : ( i ) chirality conjugaison @xmath61 : @xmath62 , ( ii ) mirror along y - axis chains @xmath63 : @xmath64 , ( iii ) translation by half unit along y - axis @xmath65 : @xmath66 .
as example one easily verifies the identities @xmath67=m_y [ a_7]$ ] , @xmath68 $ ] and @xmath69 $ ] . note that for patterns @xmath70 in classes @xmath71 one has @xmath72=t_y [ s_i]$ ] ; in that sens it is more meaningful to define a single class @xmath73 .
in fact it then appears that the four classes @xmath74 are equivalent to classify the @xmath75 chirality patterns in terms of the two effective chiralities @xmath76 such that @xmath77 with @xmath78 , @xmath79 , @xmath80 and @xmath81 .
we terminate this section by listing other properties relating chirality patterns and ldos .
+ @xmath82 : since there are only @xmath83 patterns , it implies that there are @xmath84 ( out of @xmath47 ) configurations of signs of the seven transfer integrals that leads to the same pattern @xmath54 and to the same ldos @xmath48 ( see [ appendix_a ] ) . in other words , without loss of generality , it is always possible to fix arbitrarily the sign of three out of the seven transfer integrals and hereafter the convention @xmath85 is adopted in all numerical computations .
+ @xmath86 : from the properties of the moments of the ldos ( [ ldos ] ) , it is straightforward to show that chirality conjugation @xmath61 is associated to @xmath87 . more physically it means that if a pattern exhibits a zgs at @xmath6 ( resp .
@xmath88 , @xmath5 ) filling then its chirality conjugate pattern exhibits a zgs at @xmath5 ( resp .
@xmath88 , @xmath6 ) filling .
+ @xmath89 : in the cases ( @xmath90 ) , using eq.([a6 ] ) it easily shown that for any pattern @xmath54 , its @xmath63 ( or @xmath65 ) transformed pattern has the same dos . as a consequence of properties ( @xmath91 ) , for the cases ( @xmath90 )
it is thus sufficient to study the zgs and ldos properties of patterns @xmath60 to deduce the properties of all other patterns .
lastly , we note that all previous numerical studies of dirac points and zgs in @xmath0-(bedt - ttf)@xmath1i@xmath2 corresponds to the pattern @xmath92 .
cccccccccccccccccc & @xmath93 & @xmath94 & @xmath92 & @xmath95 & @xmath96 & @xmath97 & @xmath98 & @xmath99 & @xmath100 & @xmath101 & @xmath102 & @xmath103 & @xmath104 & @xmath105 & @xmath106 & @xmath107 + @xmath14 & - & - & - & - & + & + & + & + & -&-&-&-&+&+&+&+ + @xmath15 & - & - & + & + & - & - & + & + & -&-&+&+&-&-&+&+ + @xmath16 & - & + & - & + & - & + & - & + & -&+&-&+&-&+&-&+ + @xmath17 & + & - & - & + & - & + & + & - & -&+&+&-&+&-&-&+ + [ table_1 ] [ cols="^,^,^ " , ]
in the following , we concentrate our numerics on the study of zgs at @xmath5 filling for patterns @xmath108 of classes @xmath109 respectively .
we stress that patterns @xmath110 present very similar properties as @xmath108 .
we discard patterns of class @xmath111 because for the range of parameters that we have studied we never found zgs at either @xmath6 or @xmath5 filling for these patterns . as a first illustration of the importance of the chirality pattern on the appearance of zgs , fig .
[ ldoshomo ] represents respectively the ldos @xmath48 ( @xmath113 ) for patterns @xmath114 in the case of homogeneous moduli @xmath112 . on fig .
[ ldoshomo ] , for each pattern , the energies corresponding to @xmath6 and @xmath5 fillings are indicated by vertical arrows . for pattern @xmath92 ,
the @xmath115 sites have strongly different ldos ; in addition at @xmath6 and @xmath5 fillings the system is metallic , moreover there is a strong 1d like vanhove singularity for @xmath116 at @xmath5 filling . for pattern @xmath98 ,
the ldos verify @xmath117 ; in addition there is a _ wide _ zgs at @xmath5 filling and a metallic state at @xmath6 filling with a strong vanhove singularity for @xmath118 . for pattern @xmath105 ,
the ldos verify @xmath119 moreover there is a _
narrow _ zgs both for @xmath6 and @xmath5 fillings whereas there is a vanhove singularity at @xmath88 filling . from these results
, one can already anticipate that pattern @xmath92 which shows a vanhove singulary at @xmath5 filling may thus require some peculiar and probably strong anisotropy in the moduli in order to exhibit a zgs state at this filling . in a similar manner
, one can anticipate that the zgs at @xmath5 filling obtained for patterns @xmath98 and @xmath105 may have distinct robustness against the introduction of an anisotropy in the moduli .
more precisely the zgs of pattern @xmath98 which exhibits a linear ldos over wide energy range is anticipated to be more robust than the zgs of pattern @xmath105 which is relatively narrow in energy . in the case of homogeneous modulus @xmath112 , @xmath120 black line , @xmath121 red line and @xmath122 blue line .
] as a first illustration of the combined role of chirality pattern and inhomogeneity in the moduli , fig [ ldosbedt ] represents the ldos @xmath48 ( @xmath113 ) for patterns @xmath114 in the case of bedt moduli at pressure @xmath123 kbars ( @xmath124,@xmath125,@xmath126,@xmath127,@xmath128,@xmath129 ) = @xmath130@xcite , for which it is known that pattern @xmath92 exhibits a zgs at @xmath5 filling .
the value of @xmath131 is taken such that the second moment of the dos is the same as for homogeneous modulus . at first sight ,
the main consequence of the strong inhomogeneity in the moduli is that the shape of the ldos are now qualitatively much more similar for all patterns .
in particular , now all patterns exhibit a zgs at @xmath5 filling , however for pattern @xmath92 the zgs is very narrow in energy and with strong vanhove singularities very near the dirac point indicating a strong tilt of the dirac cones .
note that for pattern @xmath105 there is no more a zgs at @xmath6 filling where it is now metallic with a strong vanhove singularity .
overall , these results seem to indicate that chirality patterns might play a weaker role for strong inhomogeneity . in the case of bedt moduli at pressure @xmath123 kbars ( @xmath124,@xmath125,@xmath126,@xmath127,@xmath128,@xmath129 )
= @xmath130 @xcite .
@xmath120 black line , @xmath121 red line and @xmath122 blue line . ] the combined roles of chirality pattern and inhomogeneity of moduli on the existence and stability of zgs at @xmath5 fillings are now examined more systematically for the three patterns @xmath114 . to characterize the inhomogeneity of the moduli it is convenient to define the following six dimensionless parameters : @xmath132 the quantities @xmath133 characterize the _ dimerization _ along the chains ( aa ) and the _ trimerization _ along the two effective chains ( baca ) .
the ratio @xmath134 quantify essentially the relative anisotropy between the different chains as compared to chain ( bc ) . within this parametrization ,
the case of homogeneous moduli corresponds to @xmath135 and @xmath136 , whereas the case of bedt moduli at @xmath123 corresponds to @xmath137 . in the following , for a given set of parameters @xmath138 ,
the value of @xmath139 is determined so as to keep the second moment of the dos equal to that of the homogeneous case ; in that way the total bandwidth is roughly constant : @xmath140}$ ] with the convention @xmath85 , the seven hopping parameters read : @xmath141 for a given chirality pattern and a given set of parameters ( @xmath142,@xmath143 ) , to characterize the existence of a zgs at @xmath5 filling we proceed as follows . in a first step
the energy eigenvalues @xmath144 are computed over a regular mesh of @xmath145 k - points in the bz such that the maximum @xmath146 and minimum @xmath147 of each band are determined . in a second step a zgs at filling @xmath5
is obtained whenever the following three conditions are simultaneously verfied : @xmath148 : @xmath149 ; @xmath150 : @xmath151 ; @xmath152 : @xmath153 where in practice @xmath154 was taken . if either @xmath148 or @xmath150 are not verified or if @xmath155 the system is considered as metallic , whereas it is considered as gapped if @xmath156 .
note that taking a smaller value for @xmath157 will a priori decrease the number of configuration which can be considered as metallic or zgs . for each of the three patterns ( @xmath158 ) , the numerical determination of possible zgs at @xmath5 fillings
is done systematically for parameters @xmath159 and for six distinct dimerization configurations @xmath160 as given in table [ table_2 ] ; such that configuration i is equivalent to no dimerization , ii to weak and vi to strong dimerization . for a given pattern and a given dimerization configuration the existence of zgs
is thus explored for @xmath161 distinct configurations of parameters @xmath142 .
ccccccc & i & ii & iii & iv & v & vi + @xmath162 & 0 & 0.25 & 0.25 & 0.5 & 0.75&0.75 + @xmath163 & 0 & 0.25 & 0.75 & 0.5 & 0.25&0.75 + @xmath164 & 0 & 0.25 & 0.25 & 0.5 & 0.75&0.75 + [ table_2 ] the first statistic fig.[histo3 ] represents the histogram distribution of the number of gapped , zgs and metallic configurations for patterns ( @xmath158 ) at @xmath5 filling and for the six distinct dimerization configurations ( i , ii, ...
the common and expected feature is that the number of gapped configurations ( black bars ) increases with the strength of the dimerization along the different chains ; for all patterns it eventually becomes the dominant phase for strong dimerization ( vi ) . despite this common feature ,
we find it relatively surprising that for pattern @xmath105 , the number of gapped configurations is always relatively large even for low or no dimerization .
as far as the number of metallic configurations ( blue bars ) is concerned , its behaviour is very different for the three distinct patterns ( @xmath158 ) . for pattern @xmath92
, it represents the dominant phase for small dimerization ( i ) and then it steadily decreases with increasing dimerization and it becomes the rarest phase for large dimerization ( vi ) ; this type behaviour is somehow as expected . for pattern @xmath98 , whatever the dimerization , the number of metallic configurations is surprisingly always very small in proportion such that roughly speaking this pattern exhibits either a zgs or a gapped phase . for pattern @xmath105 ,
the number of metallic configurations slowly decreases when dimerization is increased and it is always below that of gapped and zgs phases excepted for dimerization ( iii ) where it represents the dominant phase .
overall , we find it quite remarkable that for most dimerization configurations and for the three patterns ( @xmath158 ) , the number of metallic configurations remains always smaller than the number of zgs phases .
in fact , regarding the number of zgs configurations , the three patterns ( @xmath158 ) show distinctive behaviours . for patterns ( @xmath92,@xmath105 ) the number of zgs configurations shows no clear tendency as a function of dimerization strength ; for each dimerization configuration it constitutes around @xmath165 of the 9261 anisotropy configurations , slightly higher for @xmath105 and slightly lower for @xmath92 .
more remarkably , for pattern @xmath98 , even if the number of zgs configurations decreases steadily when the dimerization strength is increased , it still constitutes the largely dominant phase over the entire studied range of dimerization . in summary , the main message to retain from fig.[histo3 ] is that over the range of anisotropy and dimerization that we have considered , the pattern @xmath98 is by far the most favorable for the appearance of stable zgs at @xmath5 filling whereas patterns @xmath166 are more or less similarly favorable . , @xmath98 , @xmath105 at @xmath5 filling for the six distinct dimerization configurations ( i , ii, ...
vi ) . ]
the previous histograms mainly provide quantitative information of the role played by the dimerization parameters @xmath160 on the appearance of zgs at @xmath5 filling , however they do not allow to understand the role played by the anisotropy parameters @xmath142 . in order to quantify the role played by the anisotropy between the different chains , for each dimerization configuration we have computed both the average value and the variance of the anisotropy for configurations @xmath142 that lead to zgs at @xmath5 filling for patterns @xmath92 , @xmath98 , @xmath105 .
the results are summarized on fig .
[ zazbzc3 ] . on this figure
the grey region corresponds to the case of a uniform distribution of zgs configurations over the explored interval of anisotropy : @xmath167 $ ] ; we reminds that no anisotropy correspond to @xmath168 . for pattern @xmath92 , for no or weak dimerization
there is a strong deviation from the uniform distribution , more precisely the more favorable anisotropy configurations correspond to large value of @xmath169 and small value of @xmath170 . by increasing the dimerization the average values of @xmath134
enter the region corresponding to a uniform distribution .
for pattern @xmath105 the behaviour is somewhat symmetric to that of pattern @xmath92 , for small dimerization small value of @xmath169 and larger value of @xmath170 are more favorable to obtain a zgs at @xmath5 filling . finally for pattern
@xmath98 the distribution of zgs is almost independent on both the anisotropy and dimerization .
the main message to retain from this fig .
[ zazbzc3 ] , is that the larger the dimerization the less important is the role of the anisotropy ; in other words the strength of the dimerization along the different chains seems a more pertinent parameter than the anisotropy between the chains .
( red ) , @xmath171 ( black ) @xmath172 ( blue ) versus dimerization configurations ( i , ii, ... vi ) for the zgs configurations found for patterns @xmath92 , @xmath98 , @xmath102 at @xmath5 filling .
the errorbars encode the meansquare fluctuations . for comparison
, the grey region corresponds to the case of a uniform distribution of zgs configurations over the interval @xmath167 $ ] . ] as a last characterization of the zgs properties , the following fig .
[ distria3 ] , represents the evolution of the k - space distribution of the dirac points associated to the zgs at @xmath5 for pattern @xmath92 as function of the dimerization . figs .
[ distria7],[distris6 ] represent this same quantity but for pattern @xmath98 and @xmath105 respectively .
the most striking feature on these figures is that the case of no dimerization ( i ) is very different from all the other cases .
in fact , in the absence of dimerization the distribution of dirac points associated to zgs is much less spreaded over the brillouin zone , moreover it differs strongly from one pattern to the other . by increasing the dimerization , the k - space distribution of zgs s dirac points for patterns @xmath92 and @xmath98 become more and more similar , in fact they seem to only differ by number of dirac points ( intensity ) as already seen in their corresponding histogram distribution . for sufficiently large dimerization ( v , vi ) ,
the k - space distribution of zgs s dirac points for all patterns are very similar .
all these features are further indications that the larger the dimerization the less important is the role played by the form of the chirality pattern . at @xmath5
filling for the six distinct dimerization configurations ( i , ii, ... vi).,width=453 ] at @xmath5 filling for the six distinct dimerization configurations ( i , ii, ... vi).,width=453 ] at @xmath5 filling for the six distinct dimerization configurations ( i , ii, ... vi).,width=453 ]
in summary , we have examined the existence of zgs at @xmath5 filling for a tight binding model with seven nearest neighbors transfer integrals ( @xmath3 ) between the four molecules of the unit cell , as in @xmath0-type organic conductor .
we have systematically studied the combined role of the sign and moduli of the transfer integrals on the appearance of zgs at @xmath5 filling .
we have shown that for given moduli , the are only 16 independent band structures @xmath173 . each band structure
is characterized by a chirality pattern @xmath174 , where each chirality encodes the sign of the product of three transfer integrals that compose one of the four distinct triangular plaquette of the unit cell ( @xmath175 ) . in term of this classification
, it appears that all previous works on @xmath0-(bedt - ttf)@xmath1i@xmath2 corresponds to pattern @xmath92 with @xmath176 .
as far as the properties of the ldos and zgs at @xmath5 ( resp .
@xmath6 ) filling are concerned , the 16 patterns can be further classified into only four different classes determined by two effective chiralities @xmath177 and @xmath7 . using extensive numerics
we have established that for both large , or no inhomogeneity in the moduli @xmath4 , the patterns @xmath178 ( resp .
@xmath179 ) of class are the most favorable patterns for the existence of zgs at @xmath5 ( resp .
@xmath6 ) filling . at the opposite , for patterns ( @xmath180 ) of class @xmath181(+,+ ) we never found any zgs at either @xmath6 or @xmath5 filling .
finally , patterns ( @xmath182 ) ( resp .
@xmath183 ) and @xmath184 ( resp .
@xmath185 ) of class and present an intermediate situation ; they allow for zgs at @xmath5 ( resp .
@xmath6 ) filling but are much less favorable that patterns of class .
we note however that the role played by the chirality pattern is dependent on strength of the dimerization in the different effective chains formed by the transfer integrals .
more precisely for weak or no dimerization the form of the chirality pattern strongly determines the probability to find a zgs at @xmath6 or @xmath5 filling whereas for strong dimerization the possibility of having a zgs become more or less equivalent for the three classes .
from that perspective we note that the usual transfer integral parameters taken to modelize @xmath0-(bedt - ttf)@xmath1i@xmath2 are in the strong dimerization regime . on a larger perspective , we note that for bipartite lattice ( like square lattice ) when all plaquette chiralities are negative it corresponds to the so called uniform @xmath186-flux phase for which it is well known that there is a zgs at @xmath88 filling .
our results suggest that on a non - bipartite lattice like the triangular lattice , it is necessary to have a fractional density @xmath187 of @xmath186-flux plaquettes to obtain a zgs at some filling @xmath188 ; the difficulty in such a situation is that there appears many non equivalent chirality patterns .
consider the @xmath0-bedt tight binding model with only nearest - neighbor hopping on the triangular lattice @xmath189 , \label{eq : transfer_1 } \ ] ] with @xmath190 .
the one particle local greens function on a site @xmath191 is @xmath192 .
it can be expanded as @xmath193 the local density of states is obtained from the identity @xmath194 , which allows to write @xmath195 this last identity means that the @xmath196 moment @xmath197 of the local density of states on site @xmath191 is equal to sum over all closed path of @xmath198 steps that start from @xmath191 and return to @xmath191 .
the key property is that each path contributes with a weight proportionnal to the product of moduli of all the hoppings along the path and with a sign given by the product of the chiralities of all the triangular plaquettes encircled by the path . for the tight binding model of @xmath0-bedt , the moment for each site up to @xmath199
are easily obtained as : @xmath200 more generally , odd and even moment have the generic form : @xmath201 where @xmath202 depend only on the moduli @xmath203 . for @xmath204 , @xmath205 and @xmath206 ,
one easily obtains that the global moments @xmath207 verify @xmath208
cccccccccc @xmath92 & ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) & ( 6 ) & ( 7 ) & ( 8) + @xmath210 & + & + & + & + & - & - & - & - + @xmath211 & + & + & - & - & + & + & - & - + @xmath212 & - & - & + & + & - & - & + & + + @xmath213 & + & - & + & - & + & - & + & - + @xmath214 & - & + & - & + & + & - & + & - + @xmath215 & + & - & - & + & - & + & + & - + @xmath216 & - & + & + & - & - & + & + & - + [ table_a3 ] | dirac electrons with a zero - gap state ( zgs ) in organic conductor @xmath0-(bedt - ttf)@xmath1i@xmath2 result from a fine tuning of the seven nearest neighbors transfer integrals ( @xmath3 ) between the four molecules of the unit cell . in this work
we show that for given moduli @xmath4 , the possibility of having dirac electron with a zgs at @xmath5 ( or @xmath6 ) filling strongly depends on the specific configurations of signs of the seven transfer integral .
more precisely it is possible to classify the sign configurations into essentially four classes determined by and @xmath7 . using extended numerics
, we show that for inhomogeneity in the moduli , the class @xmath8 is the most favorable to find dirac electrons with zgs at @xmath5 ( or @xmath6 ) filling . for the class @xmath9 we never found any zgs at either @xmath6 or @xmath5 filling . the last two classes @xmath10 and @xmath11 an intermediate situation ; they allow for zgs at @xmath5 ( resp .
@xmath6 ) filling but are much less favorable than class . as a matter of fact , all previous numerical studies of dirac electrons and zgs in @xmath0-(bedt - ttf)@xmath1i@xmath2 correspond to class @xmath11 . |
we consider the kadomtsev - petviashvili ( kp ) equation @xmath0 in which @xmath1 depends upon the spatial variables @xmath2 , and the temporal variable @xmath3 , and @xmath4 is equal to either @xmath5 or @xmath6 .
this equation is a generalization to two spatial dimensions of the well - known korteweg - de vries ( kdv ) equation @xmath7 and has been derived as a model in the study of the transverse stability of the line solitary waves , i.e. , the stability of the solitary waves of the kdv equation with respect to two - dimensional perturbations @xcite . in particular , it arises as a model for long waves in the water - wave problem , in the absence of surface tension when @xmath8 , and in the presence of a strong surface tension when @xmath9 . the equation with @xmath8 ( negative dispersion )
is also called the kp - ii equation , whereas the one with @xmath9 ( positive dispersion ) is called the kp - i equation .
it turns out that the stability properties of the line solitary waves strongly depend upon the sign of @xmath4 : when @xmath8 the line solitary waves are transversely stable , and when @xmath8 they are unstable @xcite .
in addition to the well - known solitary waves , the kdv equation possesses a family of periodic traveling waves . in this paper
we are interested in the stability of these periodic waves as solutions of the kp equation , i.e. , their transverse stability . as solutions of the kdv equation ,
the stability of these periodic traveling waves is quite well - understood , though the results are less complete than those for solitary waves .
in contrast to the case of localized solutions , for periodic waves of dispersive equations there are mainly to types of stability results : nonlinear ( orbital ) stability with respect to periodic perturbations which have the same period as the traveling wave , and spectral stability with respect to non - periodic ( localized or bounded ) perturbations .
the question of nonlinear stability with respect to non - periodic perturbations is widely open , so far . for the kdv
, it has been recently shown that the periodic traveling waves are orbitally stable with respect to periodic perturbations @xcite , and spectrally stable with respect to non - periodic perturbations @xcite . very recently , johnson and zumbrun @xcite derived an instability criterion for the transverse instability of periodic waves for the generalized kp equation , with respect to perturbations which are periodic in the direction of propagation and have long wavelengths in the transverse direction .
in contrast , we restrict here to the case of small periodic waves , and consider the question of spectral stability for more general perturbations for the kp - i equation , and for the same type of perturbations for the kp - ii equation .
a particularity of the kp equation is that the spectral problem can not be directly formulated in terms of the spectrum of a linear operator , because of the mixed time - space derivative @xmath10 appearing in .
we present a functional set - up which leads to such a formulation , for both periodic and non - periodic perturbations . for periodic perturbations ,
we use the invertibility of the operator @xmath11 when restricted to a space of periodic functions with zero mean , and for non - periodic perturbations we rely upon floquet theory for differential operators with periodic coefficients and the invertibility of the operator @xmath12 . for the analysis of the spectra of the resulting operators we rely upon several tools from the perturbation theory for linear operators , which turn out to be suitable for the analysis of spectra of differential operators with periodic coefficients @xcite .
in addition , we use the decomposition of these operators in the product of a skew - adjoint with a self - adjoint operator , and a counting result which relates the number of unstable eigenvalues of the product operator to the number of negative eigenvalues of the self - adjoint operator @xcite .
we focus on the kp - i equation , for which we show that the small periodic waves are transversely unstable , with respect to both types of perturbations . for the kp - ii equation ,
a complete analysis of the spectra turns out to be more difficult , because of the different behavior of the dispersion relation . restricting to perturbations which are periodic in the direction of propagation and have long wavelengths in the transverse direction ,
we show that the small periodic waves are spectrally stable with respect to such perturbations .
in particular , these results show that the periodic waves have the same stability properties as the solitary waves , but under the restriction to long wavelength transverse perturbations for the kp - ii equation .
the paper is organized as follows . in section [ s:2 ]
, we describe the family of the one - dimensional periodic traveling waves of the kp equation and give a parameterization of the small waves . the spectral stability problem is first formulated in section [ s:3 ] , and then reformulated for periodic perturbations in section [ ss : specper ] , and for non - periodic perturbations in section [ ss : bloch ] .
the results in these latter two sections are presented for the kp - i equation , but it is easily seen that the same results hold for the kp - ii equation .
we discuss the spectra of the resulting linear operators , and show the transverse spectral instability of the small periodic waves of the kp - i equation for periodic perturbations in section [ s:4 ] , and for non - periodic perturbations in section [ s:5 ] . in section [ s:6 ] , we prove the transverse spectral stability result for the periodic waves of the kp - ii equation mentioned above , and we conclude with a brief discussion in section [ s:7 ] .
one - dimensional traveling waves of the kp equation are solutions of the form @xmath13 where @xmath14 is the speed of propagation , and @xmath15 satisfies the ode @xmath16 integrating this equation twice , and writing @xmath17 instead of @xmath18 , we obtain the second order ode @xmath19 in which @xmath20 and @xmath21 are arbitrary constants .
since we are interested in periodic solutions , we can set @xmath22 and the equation becomes @xmath23 [ [ symmetries ] ] symmetries + + + + + + + + + + the kp equation possesses the _
galilean invariance _
@xmath24 which leads to the invariance of the traveling - wave equation under the transformation @xmath25 as a consequence , we can restrict to positive speeds @xmath26 , and to @xmath27 .
in addition , the kp equation possesses the _ scaling invariance _
@xmath28 which allows to restrict to the speed @xmath29 .
[ [ small - periodic - waves ] ] small periodic waves + + + + + + + + + + + + + + + + + + + + restricting to @xmath29 and @xmath27 , the equation becomes @xmath30 in the phase - plane @xmath31 , this equation possesses the equilibria @xmath32 and @xmath33 , which are a center and a saddle , respectively . the center @xmath32 is surrounded by a one - parameter family of periodic orbits , with periods @xmath34 . as @xmath35 , the periodic orbits shrink to the equilibrium @xmath32 , whereas for @xmath36 the periodic orbits tend to a homoclinic orbit connecting the saddle equilibrium @xmath33 to itself .
these periodic orbits are symmetric with respect to the @xmath15-axis in the phase - plane @xmath31 , so that to each periodic orbit corresponds a periodic solution of which is even in @xmath17 , and unique up to translations in @xmath17 .
we are interested here in the periodic orbits close to the center equilibrium @xmath32 , which correspond to small periodic traveling waves of the kp equation . looking for solutions of of the form @xmath37 , where @xmath38 is
the wavenumber and @xmath39 is a @xmath40-periodic function , a direct calculation shows that the one - parameter family of small periodic solutions of , which are even in @xmath17 , is given by @xmath41 where @xmath42 the parameter @xmath43 is small and corresponds to the fourier mode @xmath5 in the fourier expansion of the @xmath40-periodic function @xmath44 , @xmath45 the function @xmath44 is even in @xmath46 , and notice the symmetry @xmath47 taking into account the translation invariance we obtain a two - parameter family of periodic solutions of , and including the parameters @xmath20 and @xmath48 , we find a four parameter family of one - dimensional periodic traveling waves of the kp equation .
consider the one - dimensional periodic traveling waves of the kp equation found in section [ s:2 ] , @xmath49 we introduce the scaling @xmath50 which transforms the kp equation into @xmath51 where we have dropped the tilde for notational simplicity .
then @xmath44 is a stationary solution of , and our goal is to study its spectral stability . following the standard approach to spectral stability , we linearize about the stationary solution @xmath44 and obtain the linear evolution equation @xmath52 this equation has coefficients depending upon @xmath46 , only , so that we make the ansatz , @xmath53 which leads to the equation @xmath54 the left hand side of this equation defines the linear differential operator @xmath55 and the spectral stability problem is concerned with the invertibility of this operator for @xmath56 and @xmath57 : the periodic wave is spectrally stable if this operator is invertible for any @xmath56 with @xmath58 , and unstable otherwise .
the type of the allowed perturbations is determined by the choice of the function space and the values of @xmath59 .
[ [ one - dimensional - perturbations ] ] one - dimensional perturbations + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the spectral stability problem for one - dimensional perturbations , i.e. , perturbations which are independent of @xmath60 , corresponds to the case @xmath61 , when @xmath62 the linear operator @xmath63 in this decomposition is in fact the linear operator arising in the spectral stability problem for the kdv equation . as solutions of the kdv equation , the periodic waves @xmath44 are spectrally stable with respect to perturbations which are periodic in @xmath46 , but also with respect to perturbations which are localized or bounded in @xmath46 , i.e. , the linear operator @xmath63 is invertible , for any @xmath56 , @xmath58 , in @xmath64 , and also @xmath65 or @xmath66 ( e.g. , see @xcite ) . on the other hand ,
the operator @xmath67 is not invertible in these spaces , so that the linear operator @xmath68 is not invertible , for any @xmath56 , @xmath58 .
this ill - posedness of the spectral stability problem for one - dimensional perturbations shows that for such perturbations we have to use the operator @xmath63 instead of @xmath68 , i.e. , the kp operator @xmath69 should be replaced by the kdv operator @xmath70 , when @xmath61 .
[ [ two - dimensional - perturbations ] ] two - dimensional perturbations + + + + + + + + + + + + + + + + + + + + + + + + + + + + + truly two - dimensional perturbations correspond to values @xmath71 .
we consider three types of such perturbations : perturbations which are periodic in @xmath46 , perturbations which are localized in @xmath46 , and perturbations which are bounded in @xmath46 .
the type of these perturbations is determined by choice of the function space in which acts the linear operator @xmath72 : 1 .
@xmath72 is considered in @xmath64 , with domain @xmath73 for perturbations which are periodic in @xmath46 ; 2 .
@xmath72 is considered in @xmath65 , with domain @xmath74 , for perturbations which are localized in @xmath46 ; 3 .
@xmath72 is considered in @xmath66 , with domain @xmath75 , for perturbations which are bounded in @xmath46 .
[ [ spectral - stability ] ] spectral stability + + + + + + + + + + + + + + + + + + summarizing , we give the following definition . 1 .
we say that the periodic wave @xmath44 is _ spectrally stable in one dimension _ _ with respect to periodic perturbations _
_ localized or bounded perturbations _ ) , if the kdv operator @xmath63 acting in @xmath64 ( resp .
@xmath65 or @xmath66 ) with domain @xmath76 ( resp .
@xmath77 or @xmath78 ) is invertible , for any @xmath56 , @xmath58 .
we say that the periodic wave @xmath44 is _ transversely spectrally stable _ _ with respect to periodic perturbations _
_ localized or bounded perturbations _ ) , if it is spectrally stable in one dimension , and the kp operator @xmath72 acting in @xmath64 ( resp .
@xmath65 or @xmath66 ) with domain @xmath79 ( resp .
@xmath74 or @xmath75 ) is invertible , for any @xmath56 , @xmath58 and any @xmath71 ; the results in @xcite show that the periodic waves @xmath44 are spectrally stable in one dimension .
we discuss their transverse spectral stability in the following sections .
it turns out that the stability properties strongly depend upon the sign of @xmath4 .
we focus on the case @xmath9 , i.e. , the kp - i equation , in sections [ s:4 ] and [ s:5 ] , and then briefly discuss the case @xmath8 , i.e. , the kp - ii equation , in section [ s:6 ] .
notice that depending upon the values of @xmath59 we may distinguish three different regimes : _ short wavelength transverse perturbations _
, when @xmath80 , _ long wavelength transverse perturbations _ , when @xmath81 , and _ finite wavelength transverse perturbations _ , otherwise .
the results in @xcite concern the transverse spectral stability with respect to periodic perturbations in the regime of long wavelength transverse perturbations .
in this section we study the transverse spectral stability of the periodic waves @xmath44 with respect to periodic perturbations , for the kp - i equation , i.e. , when @xmath9 .
more precisely , we study the invertibility of the operator @xmath82 acting in @xmath64 with domain @xmath79 , for @xmath56 , @xmath58 , and @xmath57 , @xmath71 .
we show that the question of the invertibility of the operator @xmath72 is equivalent to the study of the spectrum of the linear operator @xmath83 acting in the space @xmath84 of square - integrable functions on @xmath85 with of zero - mean , with domain @xmath86 . here
@xmath87 is the inverse of the restriction of @xmath67 to the subspace @xmath88 .
assume that @xmath56 and @xmath57 , @xmath71 .
then the linear operator @xmath72 acting in @xmath64 with domain @xmath79 is invertible if and only if its restriction to the subspace @xmath88 is an invertible operator .
first , notice that the subspace @xmath89 is invariant under the action of @xmath72 , since @xmath90 for any @xmath91 .
next , the operator @xmath72 has compact resolvent , since @xmath79 is compactly embedded in @xmath64 .
consequently , the spectrum of @xmath69 consists of isolated eigenvalues with finite algebraic multiplicity .
in particular , @xmath69 is invertible if and only if @xmath92 is an eigenvalue of @xmath69 , i.e. , if and only if there exists @xmath93 , @xmath94 , such that @xmath95 since @xmath71 , we conclude that @xmath96 , so that any eigenfunction @xmath93 associated to the eigenvalue @xmath92 belongs to @xmath88 .
this implies that @xmath92 is an eigenvalue of @xmath72 if and only if @xmath92 is an eigenvalue of the restriction of @xmath72 to @xmath88 , which proves the lemma .
since the operator @xmath67 acting in @xmath88 with domain @xmath97 is invertible , the following result is an immediate consequence of the above lemma .
[ c : inv ] assume that @xmath56 and @xmath57 , @xmath71 .
then the linear operator @xmath72 acting in @xmath64 with domain @xmath79 is invertible if and only if @xmath98 belongs to the spectrum of the operator @xmath99 acting in @xmath88 with domain @xmath86 .
our problem consists now in the study of the spectrum of the operator @xmath99 acting in @xmath88 with domain @xmath86 , for @xmath57 , @xmath71 .
this operator has compact resolvent , just as @xmath69 , so that its spectrum consists of isolated eigenvalues with finite algebraic multiplicity .
in addition , the spectrum has the following symmetry property .
[ l : syms ] the spectrum of @xmath100 is symmetric with respect to both the real and the imaginary axis .
first , the spectrum of @xmath99 is symmetric with respect to the real axis , since @xmath99 is a real operator .
next , consider the reflection @xmath101 defined by @xmath102 and notice that @xmath99 anti - commutes with @xmath101 , @xmath103 where we have used the fact that @xmath44 is an even function . if @xmath98 is an eigenvalue of @xmath99 with associated eigenvector @xmath104
, @xmath105 then @xmath106 consequently , @xmath107 is an eigenvalue of @xmath100 .
this implies that the spectrum of @xmath99 is also symmetric with respect to the origin , and completes the proof .
the analysis of the spectrum of @xmath99 in the next sections relies upon two main tools : * the decomposition of @xmath100 into the product of a skew - adjoint and a self - adjoint operator : @xmath108 which leads to a simple characterization of the unstable eigenvalues of @xmath99 ; * perturbation arguments for linear operators : we regard @xmath100 as a perturbation , for small @xmath43 , of the operator with constant coefficients @xmath109 obtained by setting @xmath110 .
the main result is the following theorem showing the spectral instability of the periodic waves @xmath44 , for @xmath43 sufficiently small .
the instability is due to a pair of real eigenvalues , with opposite signs , which is found in the long wavelength regime , for @xmath111 .
[ t : per ] for any @xmath43 sufficiently small , there exists @xmath112 , such that 1 . for any @xmath113 ,
the spectrum of @xmath100 is purely imaginary ; 2 . for any @xmath114 ,
the spectrum of @xmath100 is purely imaginary , except for a pair of simple real eigenvalues , with opposite signs .
consider the decomposition of the linear operator @xmath99 .
the operators @xmath115 and @xmath116 in this decomposition are skew - adjoint and self - adjoint , respectively , as operators acting in @xmath64 . in @xcite
it has been shown , under some general assumptions , that for an operator @xmath117 , where @xmath118 and @xmath119 are skew - adjoint and self - adjoint , respectively , the number of unstable eigenvalues , counted with multiplicities , is not larger than the number of negative eigenvalues of the operator @xmath119 . in particular ,
if @xmath119 has no negative eigenvalues , then @xmath120 has no unstable eigenvalues . for the operators in we can not directly apply this result , since the subspace @xmath88 on which acts @xmath99 is not an invariant subspace for the operator @xmath116
nevertheless , we can use part of the arguments in @xcite to prove the following result .
[ l : uev ] assume that @xmath98 is an eigenvalue of the operator @xmath100 acting in @xmath88 , and that @xmath104 is an associated eigenvector . if @xmath121 , then @xmath122 where @xmath123 is the usual scalar product in @xmath64 .
recall that the operator @xmath67 acting on @xmath88 is skew - adjoint and invertible .
then , following the proof of ( * ? ? ?
* lemma 2.7 ) , we find @xmath124 and similarly , @xmath125 since @xmath116 is self - adjoint , we conclude that @xmath126 so that @xmath127 , since @xmath121 .
[ c : imspec ] assume that there exists a positive constant @xmath48 such that @xmath128 then the spectrum of @xmath99 is purely imaginary . assuming that there exists an eigenvalue @xmath98 of @xmath100 with @xmath121 , then any associated eigenvector @xmath129 satisfies @xmath130 , by hypothesis .
this contradicts the result in lemma [ l : uev ] , and proves the corollary .
we start the analysis of the spectrum of @xmath99 with the values of @xmath59 away from the origin , @xmath131 , for some @xmath132 , i.e. , finite and short wavelength transverse perturbations .
we use the result in corollary [ c : imspec ] to show that the spectrum of @xmath99 is purely imaginary for such values of @xmath59 , provided @xmath43 is sufficiently small .
[ l : short ] assume that @xmath132 .
there exists @xmath133 , such that the spectrum of @xmath99 is purely imaginary , for any @xmath59 and @xmath43 satisfying @xmath131 and @xmath134 . according to the corollary [ c : imspec ]
it is enough to show that @xmath135 for some @xmath136 .
we write @xmath137 the operator @xmath138 has constant coefficients , and using fourier series we find that its spectrum in @xmath88 is given by @xmath139 and satisfies @xmath140 for any @xmath59 , @xmath131 . as a consequence , we have that @xmath141 the linear operator @xmath142 is bounded in @xmath64 , and a direct calculation shows that @xmath143 for some @xmath144 , and any @xmath43 sufficiently small . together with , this implies that @xmath145 which proves , for any @xmath59 , @xmath131 , provided @xmath43 is sufficiently small .
we consider now the spectrum of @xmath99 for small @xmath59 .
notice that the arguments in the proof of lemma [ l : short ] do not work when @xmath59 is small , since the constant @xmath146 in tends to zero , as @xmath147 . here , we regard @xmath99 as a perturbation of the operator with constant coefficients @xmath148 acting in @xmath88 .
a direct calculation shows that the spectrum of @xmath149 is given by @xmath150 in particular , zero is a double eigenvalue of @xmath149 , and the remaining eigenvalues are all simple , purely imaginary , and located outside the open ball @xmath151 of radius @xmath152 centered at the origin .
[ l : spdec ] the following properties hold , for any @xmath59 and @xmath43 sufficiently small . 1 .
the spectrum of @xmath99 decomposes as @xmath153 with @xmath154 2 .
the spectral projection @xmath155 associated with @xmath156 satisfies @xmath157 .
the spectral subspace @xmath158 is two dimensional . _
( i ) _ consider @xmath159 .
then @xmath98 belongs to the resolvent set of @xmath160 , i.e. , the operator @xmath161 is invertible .
we write @xmath162 where @xmath163 .
a straightforward calculation shows that there exist positive constants @xmath146 and @xmath164 such that @xmath165 for any @xmath159 , and any @xmath59 and @xmath43 sufficiently small .
consequently , @xmath166 for any @xmath159 , provided @xmath59 and @xmath43 are sufficiently small , so that the operator @xmath167 is invertible . consequently , @xmath168 is invertible , for any @xmath169 , which proves the first part of the lemma . _
( ii ) _ the spectral projection @xmath155 can be computed with the help of the dunford integral formula @xmath170 where , according to the previous arguments , @xmath171 consequently , @xmath172 and since @xmath173 for any @xmath174 , and any @xmath59 and @xmath43 sufficiently small , we conclude that @xmath157 . _ ( iii ) _ the last part of the lemma is an immediate consequence of the estimate in ( ii ) ( e.g. , see ( * ? ? ?
* lemma b.1 ) ) .
next , we use the result in lemma [ l : uev ] , to locate the eigenvalues in @xmath175 , and then we determine the two eigenvalues in @xmath176 , by computing an expansion of these eigenvalues for small @xmath59 and @xmath43 . these results complete the proof of theorem [ t : per ] .
[ l : long1 ] consider the decomposition of the spectrum of @xmath99 in lemma [ l : spdec ] .
then @xmath177 , for any @xmath59 and @xmath43 sufficiently small .
consider the restriction @xmath178 of @xmath100 to the spectral subspace @xmath179 , so that @xmath180 assume that @xmath98 is an eigenvalue of @xmath178 , and that @xmath129 is an associated eigenvector .
in particular , @xmath181 .
we use the result in lemma [ l : uev ] to show that @xmath182 .
recall that @xmath183 , and write @xmath184 so that @xmath185 is a bounded operator on @xmath64 , with norm @xmath186 for some positive constant @xmath146 , and any @xmath59 and @xmath43 sufficiently small .
we set @xmath187 , and then the property in lemma [ l : spdec ] ( ii ) shows that @xmath188 for some positive constant @xmath164 , and any @xmath59 and @xmath43 sufficiently small . we compute @xmath189 since the spectrum of the restriction of @xmath190 to @xmath191 consists of the eigenvalues @xmath192 we have that @xmath193 and taking into account the estimates and , we conclude that @xmath194 for some positive constant @xmath195 , and any @xmath59 and @xmath43 sufficiently small . consequently , @xmath196 , provided @xmath59 and @xmath43 are sufficiently small , so that @xmath182 , by lemma [ l : uev ] .
[ l : long0 ] assume that @xmath59 and @xmath43 are sufficiently small .
there exists @xmath112 , such that the two eigenvalues in @xmath197 are purely imaginary , if @xmath113 , and are real , one positive and one negative , if @xmath114 .
the eigenvalues in @xmath198 are the eigenvalues of the restriction of @xmath100 to the two - dimensional spectral subspace @xmath199 .
we determine the location of these eigenvalues by computing successively a basis of @xmath200 , the @xmath201 matrix representing the action of @xmath99 on this basis , and the eigenvalues of this matrix .
we start with the computation for @xmath110 .
then @xmath202 is an operator with constant coefficients , and @xmath203 the associated eigenvectors are @xmath204 and @xmath205 , and we choose @xmath206 as basis of the corresponding spectral subspace . since @xmath207 the @xmath201 matrix representing the action of @xmath208 on this basis is given by @xmath209 next , recall that @xmath99 anti - commutes with the reflection @xmath101 defined by .
since @xmath210 the basis @xmath211 can be extended to a basis @xmath212 , for @xmath213 , with the same property , i.e. , such that @xmath214 and @xmath215 are odd and even in @xmath46 , respectively .
in this basis the @xmath201 matrix representing the action of @xmath99 is of the form @xmath216 in order to compute the terms of order @xmath43 we take @xmath61 .
the derivative @xmath217 of the periodic wave always belong to the kernel of @xmath218 , @xmath219 due to the invariance of the kp equation under translations in @xmath17 , and @xmath220 is an odd function , with expansion @xmath221 then we choose @xmath222 as a first vector of the basis , which is compatible with the choice for @xmath110 , and satisfies @xmath223 . for the second vector we set @xmath224 and @xmath225
then a direct calculation gives @xmath226 so that @xmath227 together with the expression of @xmath228 , this shows that @xmath229 the two eigenvalues of @xmath230 , which are also the eigenvalues in @xmath176 , are roots of the characteristic polynomial @xmath231 furthermore , since @xmath232 , the two roots of this polynomial are the same for @xmath43 and @xmath233 , and we conclude that @xmath234 consequently , for any @xmath43 sufficiently small there exists a value @xmath235 such that the two eigenvalues are purely imaginary when @xmath236 , and real , with opposite signs when @xmath114 .
this proves the lemma .
in this section we study the transverse spectral stability of the periodic waves @xmath44 of the kp - i equation , with respect to perturbations which are localized or bounded in @xmath46 , i.e. , we study the invertibility of the operator @xmath82 acting in @xmath65 or @xmath66 , for @xmath56 , @xmath58 , and @xmath57 , @xmath71 .
in contrast to the case of perturbations which are periodic in @xmath46 discussed in the previous section , for perturbations which are localized or bounded , the linear operator @xmath72 acting in @xmath65 or @xmath66 has continuous spectrum . a convenient way to treat
this situation is with the help of the so - called bloch - wave decomposition , which leads to the following result .
[ l : bloch ] the linear operator @xmath72 is invertible , in either @xmath65 or @xmath66 , if and only if the linear operators @xmath237 acting in @xmath64 with domain @xmath79 are invertible , for any @xmath238 $ ] .
the proof relies upon floquet theory .
we refer , e.g. , to ( * ? ? ?
* proposition 1.1 ( ii)-(iii ) ) for a detailed proof in similar situation .
the key property of the operators @xmath239 in this lemma is that they have only point spectrum , since they act in @xmath64 with compactly embedded domain @xmath79 . when @xmath240 we recover the case of periodic perturbations studied in the previous section , so that we restrict now to @xmath241
. then the operator @xmath242 is invertible in @xmath64 , property which leads to the following result .
assume that @xmath238 $ ] and @xmath241 .
then the linear operator @xmath243 is invertible in @xmath64 if and only if @xmath98 belongs to the spectrum of the operator @xmath244 acting in @xmath64 with domain @xmath76 .
we point out that the operator @xmath245 becomes singular , as @xmath246 , so that the results found by replacing the study of the invertibility of @xmath239 by the study of the spectrum of @xmath247 are not uniform for small @xmath248 , i.e. , the set of @xmath43 for which these results are valid depends upon @xmath248 .
we restrict here to this latter study which turns out to be simpler , and sufficient to detect the values of @xmath59 and @xmath248 where instabilities occur . as in the previous section ,
the linear operator @xmath249 decomposes as @xmath250 in which @xmath242 and @xmath251 are skew - adjoint and self - adjoint operators , respectively . as a consequence
, the result in corollary [ c : imspec ] also holds for these operators .
however , in contrast to the previous section , here we can also directly apply the result in ( * ? ? ? * theorem 2.13 ) showing that the numbers @xmath252 of negative eigenvalues of @xmath253 ( counted with multiplicities ) and @xmath254 of unstable eigenvalues of @xmath247 ( counted with multiplicities ) , satisfy the inequality @xmath255 provided @xmath251 is invertible .
in fact the result in ( * ? ? ? * theorem 2.13 ) is more precise , but this inequality is enough for our purposes , since we shall only use this result in the case @xmath256 .
actually , in this situation the inequality also follows from corollary [ c : imspec ] . indeed ,
if @xmath257 and @xmath251 is invertible , then the spectrum of @xmath251 is strictly positive .
since @xmath251 is self - adjoint this implies the inequality , so that the spectrum of @xmath247 is purely imaginary .
consequently , @xmath258 , which proves the inequality .
another consequence of the decomposition is that the spectrum of @xmath259 is symmetric with respect to the imaginary axis , just as the spectrum of @xmath99 in the previous section , but it is no longer symmetric with respect to the real axis , since the operator @xmath247 is not real .
instead , we have the following result .
[ l : spsym2 ] assume that @xmath238 $ ] and @xmath241 .
then the spectrum @xmath260 of @xmath261 is symmetric with respect to the imaginary axis , and @xmath262 .
the first property is a consequence of the decomposition and of the result in ( * ? ? ?
* proposition 2.5 ) .
the second property is due to the equality @xmath263 where @xmath101 is the reflection symmetry defined by .
we determine now the spectrum of the operator @xmath259 . as a consequence of the second property in lemma [ l : spsym2 ] we restrict to @xmath264 $ ] .
the main result is the following theorem showing the transverse spectral instability of the periodic waves .
[ t : nonper ] assume that @xmath264 $ ] and set @xmath265 .
for any @xmath43 sufficiently small , there exists @xmath266 such that 1 . for @xmath267 ,
the spectrum of @xmath268 is purely imaginary ; 2 . for @xmath269 , the spectrum of @xmath268 is purely imaginary , except for a pair of complex eigenvalues with opposite nonzero real parts .
the remainder of this section is occupied by the proof of this theorem .
we rely upon the decomposition , the result in corollary [ c : imspec ] and the inequality , and perturbation arguments for linear operators , just as in section [ s:4 ] .
[ [ spectrum - of - boldsymbolmathcal - l_0ellgamma - and - consequences ] ] spectrum of @xmath270 and consequences + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the linear operator @xmath251 is a small bounded perturbation of @xmath271 using fourier series we find @xmath272 which shows that * the eigenvalues corresponding to the fourier modes @xmath273 are all positive , @xmath274 * the eigenvalue @xmath275 corresponding to the fourier mode @xmath276 is positive when @xmath277 , where @xmath278 , it is zero when @xmath279 , and it is negative when @xmath280 ; * the eigenvalue @xmath281 corresponding to the fourier mode @xmath282 is positive when @xmath283 , where @xmath284 , it is zero when @xmath285 , and it is negative when @xmath286 ; ( see also figure [ f : disp1 ] ( a ) ) . here
@xmath287 for @xmath288 and @xmath289 ( from left to right ) .
the eigenvalues of @xmath290 are found by taking @xmath291 , @xmath292 .
( b ) graphs of the dispersion relation @xmath293 for the same values of @xmath59 .
the imaginary parts of the eigenvalues of @xmath294 are found by taking @xmath291 , @xmath292 .
notice that the zeros of the two maps are the same._,title="fig:",width=132 ] for @xmath288 and @xmath289 ( from left to right ) .
the eigenvalues of @xmath290 are found by taking @xmath291 , @xmath292 .
( b ) graphs of the dispersion relation @xmath293 for the same values of @xmath59 .
the imaginary parts of the eigenvalues of @xmath294 are found by taking @xmath291 , @xmath292 .
notice that the zeros of the two maps are the same._,title="fig:",width=132 ] for @xmath288 and @xmath289 ( from left to right ) .
the eigenvalues of @xmath290 are found by taking @xmath291 , @xmath292 .
( b ) graphs of the dispersion relation @xmath293 for the same values of @xmath59 .
the imaginary parts of the eigenvalues of @xmath294 are found by taking @xmath291 , @xmath292 .
notice that the zeros of the two maps are the same._,title="fig:",width=132 ] for @xmath288 and @xmath289 ( from left to right ) .
the eigenvalues of @xmath290 are found by taking @xmath291 , @xmath292 .
( b ) graphs of the dispersion relation @xmath293 for the same values of @xmath59 .
the imaginary parts of the eigenvalues of @xmath294 are found by taking @xmath291 , @xmath292 .
notice that the zeros of the two maps are the same._,title="fig:",width=132 ] _ ( a ) ( b ) _
[ f : disp1 ] for any @xmath295 , so that the unperturbed operator has positive spectrum for @xmath277 , one negative eigenvalue if @xmath296 , and two negative eigenvalues if @xmath297 . the following result is an immediate consequence of these properties .
assume that @xmath264 $ ] .
for any @xmath298 there exists @xmath133 , such that the spectrum of @xmath259 is purely imaginary , for any @xmath59 and @xmath43 satisfying @xmath299 and @xmath134 . the properties above show that the eigenvalues of the operator @xmath290 are strictly positive , @xmath300 , for some @xmath301 , when @xmath299 . since @xmath251 is a small bounded perturbation of @xmath290 a standard perturbation argument shows that the eigenvalues of @xmath253 remain strictly positive , provided @xmath43 is sufficiently small . in particular , @xmath251 is invertible and @xmath302 , so that @xmath303 , according to , which proves the lemma .
it remains to determine the spectrum of @xmath247 for @xmath304 .
we proceed as in section [ ss : longper ] , and decompose the spectrum of @xmath259 into @xmath305 containing a minimal number of eigenvalues , and @xmath306 for which we argue as in lemma [ l : long1 ] to show that it is purely imaginary .
more precisely , @xmath306 is such that the restriction of @xmath251 to the corresponding spectral subspace is positive definite , i.e. , it satisfies the inequality .
[ [ spectrum - of - boldsymbolmathcal - a_0ellgamma - and - decomposition ] ] spectrum of @xmath307 and decomposition + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we begin by decomposing the spectrum of the unperturbed operator @xmath294 in such a way that this decomposition persists for sufficiently small @xmath43 . using fourier series again we find @xmath308 the location on the imaginary axis of these eigenvalues can be viewed with the help of the dispersion relation @xmath309 ( see figure [ f : disp1 ] ( b ) ) .
notice that the eigenvalues of @xmath294 which correspond to negative eigenvalues of @xmath290 are the ones for @xmath276 , if @xmath296 , and for @xmath276 and @xmath282 , if @xmath297 .
as a consequence of corollary [ c : imspec ] there are these eigenvalues which may lead to instabilities when perturbing the operator @xmath310 .
therefore , the idea of the spectral decomposition is to separate these eigenvalues from the remaining ( infinitely many ) eigenvalues , which correspond to positive eigenvalues of @xmath290 . of course this separation is possible as long as there are no collisions with other eigenvalues .
looking for such collisions we find that * for @xmath296 , the eigenvalue @xmath311 collides with @xmath312 , when @xmath313 and it is simple for the other values of @xmath59 ; * for @xmath314 , the two eigenvalues @xmath311 and @xmath312 are simple .
this leads to the following result .
[ l : spa0 ] assume that @xmath264 $ ] .
there exist @xmath298 and @xmath136 such that 1 . for any @xmath59 satisfying @xmath315
, the spectrum of @xmath294 decomposes as @xmath316 with @xmath317 ; 2 . for any @xmath59 satisfying @xmath318
, the spectrum of @xmath294 decomposes as @xmath319 with @xmath320 .
perturbation arguments show that this decomposition persists for the operator @xmath247 , for sufficiently small @xmath43 , and we argue as in section [ ss : longper ] to locate the spectrum of @xmath247 .
[ [ spectrum - of - boldsymbolmathcal - a_aellgamma - for - boldsymbolell_c2varepsilon_ell2ell_-2varepsilon _ ] ] spectrum of @xmath321 for @xmath322 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we prove the following result showing that the spectrum of @xmath268 is purely imaginary for such values of @xmath59 , provided @xmath43 is sufficiently small .
[ l : lc1 ] assume that @xmath264 $ ] .
there exist @xmath298 and @xmath133 such that the spectrum of @xmath268 is purely imaginary , for any @xmath59 and @xmath43 satisfying @xmath315 and @xmath134 .
the operator @xmath323 is a small relatively bounded perturbation of the operator @xmath324 , and there exists @xmath144 such that @xmath325 for any @xmath326 and @xmath43 sufficiently small .
then by arguing as in the proof of lemma [ l : spdec ] , taking into account the spectral decomposition in lemma [ l : spa0 ] , we conclude that the spectrum of @xmath323 decomposes as @xmath327 where @xmath328 and @xmath329 have the following properties 1 .
there exist two open balls @xmath330 centered at @xmath311 such that @xmath331 2 . the spectral projection @xmath332 associated with @xmath333 satisfies @xmath334 ; 3 .
the spectral subspace @xmath335 is one dimensional ; ( see also lemma [ l : spdec ] ) .
this decomposition holds when @xmath315 and @xmath134 , for @xmath336 and @xmath337 are sufficiently small .
next , the eigenvalues in @xmath338 correspond to positive eigenvalues of @xmath339 , so that we can argue as in the proof of lemma [ l : long1 ] to conclude that the eigenvalues in @xmath306 are purely imaginary .
finally , the property ( iii ) shows that @xmath305 consists of one simple eigenvalue .
this eigenvalue is the unique eigenvalue of @xmath247 in the open ball @xmath340 , and it is purely imaginary since the spectrum of @xmath259 is symmetric with respect to the imaginary axis .
this completes the proof of the lemma .
[ [ spectrum - of - boldsymbolmathcal - a_aellgamma - for - boldsymbol0ell2leqslantell_c2varepsilon _ ] ] spectrum of @xmath321 for @xmath341 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + finally , we prove the following result showing that there is a pair of unstable eigenvalues of @xmath247 , for @xmath59 close to the value @xmath342 at which the two eigenvalues of @xmath294 corresponding to the fourier modes @xmath276 and @xmath282 collide .
this result completes the proof of theorem [ t : nonper ] .
[ l : lc2 ] assume that @xmath264 $ ] .
there exist positive constants @xmath336 , @xmath337 , and @xmath343 such that the spectrum of @xmath268 is purely imaginary , if @xmath344 and @xmath345 .
if @xmath346 then the spectrum of @xmath268 is purely imaginary , except for a pair of complex eigenvalues with opposite nonzero real parts . as in the previous case , using the spectral decomposition for @xmath110 in lemma [ l : spa0 ] , and the arguments in the proof of lemmas [ l : spdec ] and [ l : long1 ] we conclude that the spectrum of @xmath323 decomposes as @xmath327 where @xmath347 is purely imaginary , and @xmath328 consists now of two eigenvalues which are the continuation for small @xmath43 of the eigenvalues @xmath311 and @xmath348 .
this decomposition holds for any @xmath59 and @xmath43 satisfying @xmath344 and @xmath134 , with @xmath336 and @xmath337 sufficiently small .
consequently , it remains to locate the two eigenvalues in @xmath328 .
the two eigenvalues @xmath311 and @xmath348 are simple and the distance between them is strictly positive @xmath349 for any @xmath59 outside a neighborhood of @xmath342 , where the eigenvalues collide , and such that @xmath344 .
a standard perturbation argument then shows that the continuation of this eigenvalues for sufficiently small @xmath43 is a pair of simple eigenvalues of @xmath323 .
each of these eigenvalues of @xmath323 is purely imaginary , since the spectrum of @xmath323 is symmetric with respect to the imaginary axis , just as in the proof of lemma [ l : lc1 ] .
this shows that the spectrum of @xmath323 is purely imaginary for any @xmath59 outside a neighborhood of @xmath342 . in order to locate @xmath328 for @xmath59 close to @xmath342
, we proceed as in the proof of lemma [ l : long0 ] .
we compute successively a basis for the two - dimensional spectral subspace associated with @xmath328 , the @xmath201 matrix @xmath350 representing the action of @xmath247 on this basis , and the eigenvalues of this matrix . at @xmath110
we choose as basis the two eigenvectors associated with the eigenvalues @xmath348 and @xmath311 , @xmath351 and then the @xmath201 matrix representing the action of @xmath294 on this basis is given by @xmath352 next , notice that the operator @xmath247 anti - commutes with the conjugation - reflection symmetry @xmath353 defined by @xmath354 since @xmath355 the basis @xmath356 can be extended to a basis @xmath357 , for @xmath213 , with the same property . in this basis ,
the @xmath201 matrix @xmath350 anti - commutes with the conjugation ( the map induced by @xmath353 ) , i.e. , @xmath358 , which implies that the coefficients of @xmath350 are purely imaginary . in order to compute the terms of order @xmath43
, we take @xmath359 , and proceed as in the computation of the vector @xmath360 in the proof of lemma [ l : long0 ] .
then at order @xmath43 we find @xmath361 together with the expression of @xmath362 , this shows that @xmath363 where @xmath364 .
recall that @xmath365 then looking for eigenvalues @xmath98 of the form @xmath366 we find that @xmath367 is root of the polynomial @xmath368 this polynomial has real coefficients , since the matrix @xmath350 is purely imaginary . here
we have also taken in account the fact that these coefficients are even in @xmath43 , due to the equality @xmath232 , which implies that the two roots of the polynomial are the same for @xmath43 and @xmath233 .
a direct computation shows that the discriminant of this polynomial is @xmath369 and for any @xmath43 sufficiently small there exists @xmath370 such that the two roots of the polynomial are real when @xmath371 , and complex otherwise .
this implies that the two eigenvalues of @xmath247 are purely imaginary when @xmath345 , and complex , with opposite nonzero real parts when @xmath372 , which proves the lemma .
in this section we briefly discuss the kp - ii equation , when @xmath8 in equation .
the functional set - up developed in section [ ss : specper ] for periodic perturbations , and in section [ ss : bloch ] for non - periodic perturbations remains valid for the kp - ii equation .
however , it turns out that locating the spectra of the resulting operators is more complicated , because of the qualitatively different behavior of the dispersion relation .
we shall restrict here to the case of periodic perturbations , when we have to determine the spectrum of the operator @xmath373 acting in @xmath88 with domain @xmath86 ( see corollary [ c : inv ] in section [ ss : specper ] ) .
we begin by analyzing the spectra of the unperturbed operators @xmath138 and @xmath208 .
[ [ spectrum - of - boldsymbolmathcal - l_0ell ] ] spectrum of @xmath374 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + using fourier series we find that the spectrum of the operator @xmath138 acting in @xmath88 is given by @xmath375 ( see also figure [ f : disp2 ] ( a ) ) . for @xmath289 .
the eigenvalues of @xmath138 are found by taking @xmath376 , @xmath377 .
( b ) graph of the dispersion relation @xmath378 for @xmath289 .
the imaginary parts of the eigenvalues of @xmath208 are found by taking @xmath376 , @xmath377 .
notice that the zeros of the two maps are the same._,title="fig:",width=132 ] for @xmath289 .
the eigenvalues of @xmath138 are found by taking @xmath376 , @xmath377 .
( b ) graph of the dispersion relation @xmath378 for @xmath289 .
the imaginary parts of the eigenvalues of @xmath208 are found by taking @xmath376 , @xmath377 .
notice that the zeros of the two maps are the same._,title="fig:",width=132 ] _ ( a ) ( b ) _
[ f : disp2 ] in contrast to the kp - i equation , the spectrum of @xmath138 contains now negative eigenvalues , and the number of these eigenvalues increases with @xmath59 .
therefore we can not use the result in corollary [ c : imspec ] to exclude unstable eigenvalues of @xmath99 for certain values of @xmath59 , e.g. , as in lemma [ l : short ] .
however , for a given @xmath59 we can use this corollary to reduce the number of potentially unstable eigenvalues to a finite number , e.g. , as in lemmas [ l : long1 ] , [ l : lc1 ] , or [ l : lc2 ] , but this number tends to @xmath379 as @xmath380 .
[ [ spectrum - of - boldsymbolmathcal - a_0ell ] ] spectrum of @xmath381 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the spectrum of the operator @xmath208 acting in @xmath88 is given by @xmath382 ( see also figure [ f : disp2 ] ( b ) ) .
notice that the dispersion relation @xmath383 is monotonically decreasing on @xmath384 $ ] and @xmath385 , so that colliding eigenvalues correspond to fourier modes with opposite signs .
a direct calculation then shows that for any @xmath386 the eigenvalues corresponding to the fourier modes @xmath387 and @xmath388 collide when @xmath389 moreover , the corresponding eigenvalues of @xmath138 have opposite signs , so that any of these collisions may lead to unstable eigenvalues of the operator @xmath99 .
we can use the arguments presented in the previous sections to reduce the spectral analysis to that of the continuation , for small @xmath43 , of the two colliding eigenvalues , but an additional difficulty comes now from the fact that there are infinitely many values of @xmath59 where two potentially unstable eigenvalues collide .
furthermore , it turns out that in order to determine the location of these two eigenvalues , we need to compute taylor expansions up to order @xmath390 for the resulting @xmath201 matrix .
consequently , the complete spectral analysis seems to be much more difficult in this case .
however , we can easily conclude in the case @xmath391 , which corresponds to values of @xmath59 close to @xmath392 .
[ [ long - wavelength - transverse - perturbations ] ] long wavelength transverse perturbations + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + restricting to small values of @xmath59 , i.e. , long wavelength transverse perturbations , we can argue as in section [ ss : longper ] and determine the spectrum of @xmath99 .
in contrast to the kp - i equation , now we show that the spectrum of @xmath99 is purely imaginary when @xmath59 and @xmath43 sufficiently small , i.e. , the small periodic waves of the kp - ii equation are spectrally stable with respect to long wavelength transverse perturbations .
the spectrum of @xmath99 is purely imaginary , for any @xmath59 and @xmath43 sufficiently small .
it is enough to replace @xmath393 by @xmath394 in the proofs given in section [ ss : longper ] .
then the results in lemmas [ l : spdec ] and [ l : long1 ] remain the same , the only change appears in the result of the calculation of the eigenvalues of the @xmath201 matrix @xmath230 in lemma [ l : long0 ] . upon replacing @xmath393 by @xmath394
we find that the two eigenvalues satisfy @xmath395 consequently , @xmath396 , for @xmath59 and @xmath43 sufficiently small , which shows that the two eigenvalues are purely imaginary , so that the spectrum of @xmath99 is purely imaginary .
we have presented a functional set - up for the study of the transverse spectral stability of the one - dimensional periodic waves of the kp equation . relying upon perturbation arguments for linear operators we have determined the spectra of the operators arising in the spectral problems for the waves of small amplitude of the kp - i equation , and showed that these waves are transversely unstable with respect to perturbations which are either periodic or non - periodic ( localized or bounded ) in the direction of propagation . for the kp - ii equation ,
the study turns out to be more complicated , due to the ` negative dispersion ' , which leads to a great number of ` potentially unstable ' modes . in this case
it is only shown that the periodic waves are transversely spectrally stable with respect to perturbations which are periodic in the direction of propagation and have long wavelengths in the transverse direction .
it seems much more difficult to study transverse perturbations in the finite and short wavelength regimes , with this type of arguments , since the location of the potentially unstable eigenvalues is determined by terms of arbitrary high order in the expansion of these eigenvalues for small @xmath43 . in these regimes
a numerical computation of the spectrum , using for instance the package described in @xcite , may help to locate the spectrum , and so get more insight in the stability properties of these waves .
this type of approach to transverse spectral stability of periodic waves can be adapted to other two - dimensional dispersive models , e.g. , generalized kp equations , boussinesq systems , equations of schrdinger type .
however , one has to restrict to periodic waves of small amplitude , since we strongly rely upon perturbation arguments for linear operators . for large waves , we mention the approach in @xcite which leads to instability criteria , and in the particular case of integrable equations , so also for the kp equation
, one could take advantage of the integrability and explicitly compute the spectra of the linear operators ( e.g. , see @xcite for such an approach in the case of the kdv equation ) .
finally , we point out that these results concern only the spectral stability of the periodic waves , the question of nonlinear stability being widely open .
nevertheless , we expect that the recent methods developed in @xcite for the transverse nonlinear instability of the solitary waves , can be adapted to periodic waves .
in particular , these would lead to a transverse nonlinear instability result for the periodic waves of the kp - i equation . | the kadomtsev - petviashvili ( kp ) equation possesses a four - parameter family of one - dimensional periodic traveling waves .
we study the spectral stability of the waves with small amplitude with respect to two - dimensional perturbations which are either periodic in the direction of propagation , with the same period as the one - dimensional traveling wave , or non - periodic ( localized or bounded ) .
we focus on the so - called kp - i equation ( positive dispersion case ) , for which we show that these periodic waves are unstable with respect to both types of perturbations .
finally , we briefly discuss the kp - ii equation , for which we show that these periodic waves are spectrally stable with respect to perturbations which are periodic in the direction of propagation , and have long wavelengths in the transverse direction .
* transverse spectral stability of small periodic + traveling waves for the kp equation * + * mariana haragus * + laboratoire de mathmatiques , universit de franche - comt + 16 route de gray , 25030 besanon , france + [email protected] * running head : * periodic waves of the kp equation * keywords : * kp equation , periodic traveling waves , spectral stability this work was partially supported by the agence nationale de la recherche through grant anr prefered . |
few - particle systems with resonant interactions ( characterized by a large two - particle scattering length @xmath0 greatly exceeding the interaction range ) exhibit universal behaviour .
their properties are independent of the short - range interaction details .
the theoretical study of such systems was initiated by v. efimov in 1970 who predicted the existence of weakly bound three - particle states ( efimov trimers ) with asymptotic discrete scaling symmetry @xcite .
the interest in the universal few - particle systems raised after the efimov s prediction got confirmed in the cold - atom physics experiments @xcite . in the present work we summarize some recent theoretical developments for the universal system of four identical bosons .
the existence of two tetramers below each efimov trimer was predicted in refs .
@xcite while their energies and widths were determined very precisely in refs .
the resulting four - boson energy ( @xmath1 ) spectrum as function of @xmath2 is schematically shown in fig .
[ fig : efi ] ( see refs .
@xcite for quantitative relations between @xmath0 and @xmath1 that , for a better visualisation , are not preserved here ) .
the above - mentioned tetramers are labeled as @xmath3 where @xmath4 refers to the associated efimov trimer and @xmath5 . in the present work we focus on a different type of tetramers , namely , the ones having an approximate dimer - atom - atom structure and labeled as @xmath6 .
their existence was predicted in ref .
@xcite as a consequence of the three - body efimov effect in the three - body system made off a dimer and two atoms . in a special regime near the efimov trimer intersection with the atom - dimer threshold
the atom - dimer scattering length becomes arbitrarily large and greatly exceeds the dimer size . under these circumstances
one may expect to mimic some properties of the few - boson systems using a model that considers the dimer as a pointlike particle .
then in the effective three - body system consisting of a dimer and two atoms there are two atom - dimer pairs with infinite two - body scattering length .
in such a three - body model of the four - boson system the three - body efimov effect occurs , however , with a very large discrete scaling factor @xcite .
the dimer - atom - atom structure of the resulting efimov - like states is only an approximation , and we present the first successful attempt @xcite to describe them rigorously as four - boson states . since they lie in the continuum , we use scattering equations for the four - particle transition operators . for a given efimov trimer
the universal properties of the lowest associated tetramer @xmath7 are determined and its impact on the collisions in the four - boson system is discussed .
furthermore , using the same technical framework we study the recombination processes in the four - boson system , i.e. , the four - boson recombination into atom plus trimer @xcite and dimer - atom - atom recombination into atom plus trimer or dimer plus dimer @xcite .
we employ alt , grassberger , and sandhas ( ags ) equations @xcite for the transition operators that provide an exact description of the four - particle scattering process . in the case of four identical particles
there are only two distinct two - cluster partitions , one of @xmath8 type and one of @xmath9 type .
we choose those partitions to be ( 12,3)4 and ( 12)(34 ) and denote them in the following by @xmath10 and @xmath11 , respectively .
the corresponding transition operators @xmath12 for the system of four identical bosons obey symmetrized ags equations [ eq : u ] @xmath13 here @xmath14 is the free green s function of the four - particle system with energy @xmath1 and kinetic energy operator @xmath15 , the two - particle transition matrix @xmath16 acting within the pair ( 12 ) is derived from the corresponding potential @xmath17 , and the symmetrized operators for the 3 + 1 and 2 + 2 subsystems are obtained from the integral equations @xmath18 the employed basis states have to be symmetric under exchange of two particles in subsystem ( 12 ) for the @xmath8 partition and in ( 12 ) and ( 34 ) for the @xmath9 partition .
the correct symmetry of the four - boson system is ensured by the operators @xmath19 , @xmath20 , and @xmath21 where @xmath22 is the permutation operator of particles @xmath0 and @xmath23 .
all observables for two - cluster reactions are determined by the transition amplitudes @xmath24 obtained @xcite as on - shell matrix elements of the ags operators ; the weight factors @xmath25 with values @xmath26 , @xmath27 , @xmath28 , and @xmath29 arise due to the symmetrization @xcite .
the matrix elements are calculated between the faddeev components @xmath30 of the corresponding initial / final atom - trimer or dimer - dimer states @xmath31 .
the amplitude for the three - cluster breakup of the initial two - cluster state is given @xcite by @xmath32 @xmath33 and @xmath34 and the dimer - atom - atom three - cluster channel state @xmath35 .
due to the time - reversal symmetry the three - cluster recombination into a two - cluster state is described by the same amplitude , i.e. , @xmath36 .
the amplitude for the four - cluster breakup can be obtained @xcite as @xmath37 t \ , g_0 u_1 g_0 \ , t \ , g_0 \ , { \mathcal{u}}_{1\alpha } \\ { } & + ( 1+p_2 ) t \ , g_0 u_2 g_0 \ ,
, g_0 \ , { \mathcal{u}}_{2\alpha } \ } \end{split}\ ] ] with @xmath38 and @xmath39 and the nonsymmetrized four - boson free - channel state @xmath40 .
again , due to the time - reversal symmetry the amplitude for the four - boson recombination into a two - cluster state is given by @xmath41 .
we solve the ags equations in the momentum - space partial - wave framework ; the technical details can be found in refs . @xcite .
to obtain universal results we consider reactions involving highly excited efimov trimers where the finite - range effects become negligible ; typically , @xmath42 .
the @xmath7 tetramer emerges at the @xmath4th atom - trimer threshold at the two - boson scattering length @xmath43 with @xmath44 corresponding to the intersection of atom - trimer and atom - atom - dimer thresholds . as shown in ref .
@xcite , this tetramer stays very close to the @xmath4th atom - trimer threshold and decays through the atom - atom - dimer threshold at @xmath45 .
the higher tetramers @xmath46 exist only extremely close to @xmath47 and can not be resolved in our four - boson calculations ; estimations based on the three - body model are given in ref .
@xcite .
the intersections of the tetramers with the @xmath4th atom - trimer threshold manifest themselves most prominently in the collisions of atoms and trimers at vanishing relative kinetic energy leading to the resonant behaviour of the corresponding atom - trimer ( @xmath4th ) scattering length @xmath48 .
it is shown in fig .
[ fig : a ] as a function of the two - boson scattering length @xmath0 .
here we use @xmath49 to build universal dimensionless ratios ; this special value of @xmath0 corresponds to the intersection of the @xmath4th atom - trimer and dimer - dimer thresholds .
the @xmath48 resonance at @xmath50 is due to the @xmath7 tetramer .
the increase of @xmath51 shown in fig .
[ fig : a ] near @xmath52 is not related to the tetramers ; it is due to the scaling of the elastic cross section with the spatial size of the trimer that grows with decreasing binding relative to the atom - dimer threshold .
the @xmath48 resonances at @xmath53 and @xmath54 correspond to the disappearance and reappearance of the @xmath55 tetramer as shown in fig .
[ fig : efi ] .
the explanation for such a spectacular behaviour of the @xmath55 tetramer may be the following : increasing @xmath2 , i.e. , increasing the strength of the two - boson interaction , the binding energies increase for all few - boson states but at different rates .
the increase is faster for the dimer , slower for the trimers , and even slower for the tetramers .
this is why the @xmath4th trimer decays through the atom - dimer threshold at @xmath47 , and the @xmath55 tetramer , having a structure of an atom weakly attached to the trimer , decays through the atom - trimer threshold at @xmath56 , becoming an inelastic virtual state ( ivs ) .
however , when the the @xmath4th atom - trimer and dimer - dimer thresholds are close to each other , the interference of the atom - trimer and dimer - dimer configurations in the @xmath55 tetramer yield additional binding such that the tetramer becomes an unstable bound state until it finally decays through the dimer - dimer threshold @xcite .
the number of the four - atom recombination events in a non - degenerated atomic gas per volume and time is @xmath57 where @xmath58 is the density of atoms and @xmath59 is the four - atom recombination rate .
its relation to the breakup / recombination amplitude is given in ref .
the zero - temperature limit of the four - atom recombination rate @xmath60 is shown in fig .
[ fig : k4 ] as a function of the two - boson scattering length .
the special value of @xmath61 corresponds to the @xmath4th efimov trimer being at the three free atom threshold .
the two peaks of @xmath60 at @xmath62 and @xmath63 correspond to the @xmath64 and @xmath55 tetramers being at the four free atom threshold , respectively .
this is qualitatively consistent with the results of ref .
@xcite .
the results for the dimer - atom - atom recombination reactions leading to the atom - trimer or dimer - dimer final states are presented in ref . @xcite .
we studied the dimer - atom - atom three - body efimov effect using rigorous four - particle scattering equations for the transition operators that were solved in the momentum - space framework .
we determined the properties of the deepest resulting tetramers and demonstrated that in ultracold atom - trimer collisions they lead to resonant enhancement of elastic and inelastic reactions .
furthermore , we considered three- and four - cluster recombination reactions in the universal four - boson system and presented results for the ultracold four - boson recombination . | recent theoretical developments in the four - boson system with resonant interactions are described .
momentum - space scattering equations for the four - particle transition operators are used .
the properties of unstable tetramers with approximate dimer - atom - atom structure are determined .
in addition , the three- and four - cluster recombination processes in the four - boson system are studied . |
the spectral properties of hypoelliptic operators and the problem of return to equilibrium for the associated parabolic equations are topics of current interest .
such operators arise naturally in non - equilibrium statistical mechanics as the generators of markov processes which model the dynamics of classical open systems @xcite .
a standard example of hypoelliptic equation is the fokker - planck equation @xcite , @xmath0 this equation is the forward kolmogorov equation for the langevin stochastic differential equation @xmath1 where @xmath2 denotes a standard brownian motion in @xmath3 , @xmath4 the friction coefficient and @xmath5 the inverse temperature .
several properties of the fokker - planck equation have been studied including spectral properties and convergence to equilibrium @xcite , homogenization @xcite , short time asymptotics @xcite and semiclassical limits @xcite .
related problems have been studied for the fokker - planck equation corresponding to the generalized langevin equation under a markovian approximation @xcite .
recently , the theory of hypocoercivity @xcite has been developed for studying the convergence to equilibrium of kinetic equations .
this theory applies to evolution equations of the form @xmath6 where @xmath7 are some operators defined on an appropriate hilbert space @xmath8 , which is usually the @xmath9 space weighted by the invariant measure of the underlying markov process , with @xmath10 being skew - adjoint in this space , @xmath11 . under appropriate assumptions on the commutators between the operators @xmath7 and higher order commutators , it is possible to prove the exponential return to equilibrium for these evolution equations . regarding the fokker - planck operator , quantitative estimates on the rate of convergence
can be obtained via the analysis of the spectral gap .
however , it is not possible in general to determine exactly the spectrum and one of the few instances when this is possible is the case of quadratic potentials @xcite . in the present work ,
we consider evolution equations associated with general quadratic operators @xmath12 and address the problem of the exponential return to equilibrium for these systems .
quadratic operators are pseudodifferential operators defined in the weyl quantization @xmath13 by symbols @xmath14 , with @xmath15 and @xmath16 , which are complex - valued quadratic forms @xmath17 these operators are differential operators with simple and fully explicit expression .
indeed , the weyl quantization of the quadratic symbol @xmath18 , with @xmath19 , @xmath20 , is the differential operator @xmath21 notice that quadratic operators are non - selfadjoint operators and that those with symbols having non - negative real parts are accretive . since the classical work by j. sjstrand @xcite , a complete description for the spectrum of elliptic quadratic operators is known and has played an important rle in the analysis of partial differential operators with double characteristics .
elliptic quadratic operators are quadratic operators whose symbols satisfy the ellipticity condition @xmath22 in a recent work @xcite , the spectral properties of non - elliptic quadratic operators , that is operators whose symbols may fail to satisfy the ellipticity condition on the whole phase space @xmath23 , were investigated .
for any quadratic operator whose symbol has a real part with a sign , say here a symbol with non - negative real part @xmath24 it was pointed out the existence of a particular linear subvector space @xmath25 in the phase space @xmath23 , @xmath26 , intrinsically associated to the symbol @xmath27 and called singular space , which plays a basic rle in the understanding of the properties of this non - elliptic quadratic operator .
in particular , this work @xcite ( theorem 1.2.2 ) gives a complete description for the spectrum of any non - elliptic quadratic operator @xmath28 whose symbol @xmath27 has a non - negative real part , @xmath29 ; and satisfies an assumption of partial ellipticity along its singular space @xmath25 , @xmath30 under those assumptions , the spectrum of the quadratic operator @xmath28 is showed to be composed of a countable number of eigenvalues with finite multiplicity and the structure of the spectrum is similar to the one known for elliptic quadratic operators @xcite .
this condition of partial ellipticity is weaker than the condition of ellipticity in general , @xmath31 , and allows to deal with more degenerate situations .
an important class of quadratic operators satisfying condition ( [ sm2 ] ) are those with zero singular space @xmath32 . in this case , the condition of partial ellipticity trivially holds and these differential operators have been showed to be hypoelliptic @xcite ( theorem 1.2.1 ) .
more precisely , they enjoy specific subelliptic properties with a loss of derivatives with respect to the elliptic case depending directly on the structure of their singular spaces .
we refer the reader to @xcite for a complete presentation of these subelliptic properties and examples of such subelliptic quadratic operators . in the present work
, we shall consider this class of hypoelliptic operators with zero singular space and study further the structure of the bottom of their spectra .
more specifically , we shall see that the first eigenvalue in the bottom of their spectra has always algebraic multiplicity one with an eigenspace spanned by a ground state of exponential type .
we shall also give an explicit formula for the spectral gap which is computable via a simple algebraic calculation ; and finally answer the question of long time behavior of the associated evolution equations by proving the property of exponential return to equilibrium for these quadratic systems .
let us finally mention that the singular space theory @xcite apply in various settings and that certain applications to the study of chains of oscillators and the generalized langevin equation are discussed at the end of this article .
we begin by recalling miscellaneous facts and notations about quadratic operators . in all the following
, @xmath33 stands for a complex - valued quadratic form with a non - negative real part @xmath34 we know from @xcite ( p.425 - 426 ) that the maximal closed realization of the operator @xmath28 , i.e. , the operator on @xmath35 with domain @xmath36 coincides with the graph closure of its restriction to @xmath37 , @xmath38 and that this operator is maximally accretive @xmath39 associated to the quadratic symbol @xmath27 is the numerical range @xmath40 defined as the closure in the complex plane of all its values , @xmath41 the hamilton map @xmath42 associated to the quadratic form @xmath27 is the unique map defined by the identity @xmath43 where @xmath44 stands for the polarized form associated to the quadratic form @xmath27 and @xmath45 is the canonical symplectic form on @xmath23 , @xmath46 it directly follows from the definition of the hamilton map @xmath47 that its real part and its imaginary part @xmath48 are the hamilton maps associated to the quadratic forms @xmath49 and @xmath50 , respectively .
one can notice from ( [ 10 ] ) that a hamilton map is always skew - symmetric with respect to @xmath45 .
this is just a consequence of the properties of skew - symmetry of the symplectic form and symmetry of the polarized form @xmath51 associated to the quadratic symbol @xmath27 is the singular space @xmath25 defined in @xcite as the subvector space in the phase space equal to the following intersection of kernels @xmath52\big ) \cap { \mathbb{r}}^{2n},\ ] ] where the notations @xmath53 and @xmath54 stand respectively for the real and imaginary parts of the hamilton map associated to @xmath27 .
notice that the cayley - hamilton theorem applied to @xmath54 shows @xmath55 where @xmath56 is the vector space spanned by the vectors @xmath57 , ... , @xmath58 .
the singular space is therefore equal to the finite intersection of the kernels @xmath59\big ) \cap { \mathbb{r}}^{2n}.\ ] ] as mentioned above , when the quadratic symbol @xmath27 satisfies a condition of partial ellipticity along its singular space @xmath25 , @xmath60 theorem 1.2.2 in @xcite gives a complete description for the spectrum of the quadratic operator @xmath28 which is only composed of eigenvalues with finite algebraic multiplicity @xmath61 where @xmath47 is the hamilton map associated to the quadratic form @xmath27 , @xmath62 is the dimension of the space of generalized eigenvectors of @xmath47 in @xmath63 belonging to the eigenvalue @xmath64 , @xmath65 in the present paper , we study evolution problems associated to accretive quadratic operators @xmath28 with zero singular space @xmath66 notice that in this case @xmath67 and the condition of partial ellipticity along the singular space trivially holds .
the spectrum is then reduces to @xmath68 define @xmath69 and @xmath70 our first result shows that the first eigenvalue in the bottom of the spectrum @xmath71 has always algebraic multiplicity one with an eigenspace spanned by a ground state of exponential type and that the spectral gap is exactly given by the positive rate @xmath72 . [ th1 ] let @xmath28 be a quadratic operator whose symbol @xmath33 is a complex - valued quadratic form with a non - negative real part @xmath73 and a zero singular space @xmath32 .
then , the first eigenvalue in the bottom of the spectrum defined in _ ( [ vp0 ] ) _
has algebraic multiplicity one and the eigenspace @xmath74 is spanned by a ground state of exponential type @xmath75 where @xmath76 is a complex - valued quadratic form on @xmath77 whose real part is positive definite .
moreover , the spectral gap of the operator @xmath28 is exactly given by the positive rate @xmath72 defined in _ ( [ rate ] ) _ , @xmath78 let us underline that operators satisfying the assumptions of theorem [ th1 ] are not elliptic in general and that their symbol may fail to satisfy the ellipticity condition @xmath79 however , the assumption about the singular space @xmath32 ensures that such an operator enjoys noticeable subelliptic ( hypoelliptic ) properties @xmath80 with @xmath81 and a loss of derivatives with respect to the elliptic case @xmath82 . more specifically ,
consider @xmath33 a complex - valued quadratic form with a non - negative real part @xmath83 and zero singular space @xmath32 . according to the definition of the singular space ( [ h1bis ] )
, there exists a smallest integer @xmath84 such that @xmath85\big ) \cap { \mathbb{r}}^{2n}=\{0\},\ ] ] where @xmath47 stands for the hamilton map of @xmath27 .
theorem 1.2.1 in @xcite shows that the quadratic operator fulfills a global subelliptic estimate with loss of @xmath86 derivatives with respect to the elliptic case , that is , there exists @xmath87 such that for all @xmath88 , @xmath89 with @xmath90 those subelliptic estimates are a key argument for getting sharp resolvent estimates and proving the property of exponential return to equilibrium .
next result proves this property of exponential return to equilibrium and provides an exact formula for the optimal rate of convergence : [ th2 ] let @xmath28 be a quadratic operator satisfying the assumptions of theorem _ [ th1]_. by using the notations introduced in _ ( [ vp0 ] ) _ and _ ( [ rate ] ) _ , we consider the operator @xmath91 then , for all @xmath92 , there exists a positive constant @xmath87 such that @xmath93 where @xmath94 is the rank - one spectral projection associated with the simple eigenvalue zero of the operator @xmath95 and @xmath96 stands for the norm of bounded operators on @xmath35 .
let @xmath28 be a quadratic operator satisfying to the assumptions of theorems [ th1 ] or [ th2 ] , and @xmath71 be the first eigenvalue in the bottom of the spectrum .
theorem [ th1 ] shows that @xmath97 where @xmath47 stands for the hamilton map of the quadratic form @xmath27 .
we shall now consider the specific case when this quadratic operator is real in the sense that @xmath98 is a real - valued function whenever @xmath99 is a real - valued function .
we begin by checking that this assumption ensures that @xmath71 is real and that the quadratic form @xmath76 defining the ground state @xmath100 is then positive definite . indeed , by using that the operator is real and passing to the complex conjugate in the equation @xmath101 we obtain that @xmath102 it shows that @xmath103 is also an eigenvalue for the operator @xmath28 .
since by theorem [ th1 ] , we have @xmath104 with @xmath72 , @xmath105 implies that the eigenvalue @xmath71 is real .
furthermore , notice that @xmath106 and @xmath107 are therefore two eigenvectors associated to the eigenvalue @xmath71 .
we deduce from theorem [ th1 ] that there exists @xmath108 , not both equal to zero such that for any @xmath109 , @xmath110 recalling that @xmath76 is a quadratic form , we obtain with @xmath111 that @xmath112 , which implies successively that @xmath113 , @xmath114 , @xmath115 , @xmath116 , @xmath117 , if @xmath118 stands for the imaginary part of the quadratic form @xmath76 .
this shows that the quadratic form @xmath118 is constant @xmath119 and therefore identically equal to zero .
the quadratic form @xmath76 is therefore a positive definite quadratic form . on the other hand ,
notice that the adjoint operator ( see @xcite , p.426 ) is actually given by the quadratic operator @xmath120 whose symbol is the complex conjugate symbol of @xmath27 .
it is therefore a complex - valued quadratic form with non - negative real part and a zero singular space for which theorem [ th1 ] applies .
then , @xmath71 is also the first eigenvalue in the bottom of the spectrum for the quadratic operator @xmath121 , and we know from theorem [ th1 ] that the eigenspaces associated with the eigenvalue @xmath71 for both operators @xmath28 and @xmath121 are one - dimensional subvector spaces with ground states of exponential type .
we shall assume further that the two operators have same ground state @xmath122 with @xmath123 , @xmath109 ; where @xmath76 is a positive definite quadratic form on @xmath77 .
under those assumptions , one can specify further the result given in theorem [ th2 ] : [ th3 ] let @xmath28 be a quadratic operator satisfying the assumptions of theorem _ [ th1]_. assume that this quadratic operator is real and satisfies _ ( [ y11])_. using the notations introduced in _ ( [ vp0 ] ) _ and _ ( [ rate ] ) _ , we consider the operator @xmath91 then , for all @xmath92 , there exists a positive constant @xmath87 such that @xmath124 where @xmath125 is the @xmath35 scalar product of @xmath99 and @xmath126 , @xmath127 we finally recall that the singular space theory @xcite was mainly inspired by the seminal work on the fokker - planck equation @xcite and aims at extending to wider classes of doubly characteristic operators some results proved for fokker - planck operators in @xcite .
after preliminary advances in this direction in @xcite , the main algebraic fact discovered in @xcite was to notice that the symbols of operators satisfying the assumptions of theorem [ th1 ] have real parts which may not be positive definite in general ; however , when the real part of these quadratic symbols are averaged by the flow of the hamilton vector field associated to their imaginary parts @xmath128 , @xmath129 with @xmath130 , quadratic forms @xmath131 become positive definite for any @xmath132 .
this averaging condition ( [ hi1 ] ) is a key property in the works @xcite .
one of the main interest of the singular space theory is to provide an easily computable algebraic condition on symbols of general quadratic operators for checking that the averaging property ( [ hi1 ] ) holds .
the remainder of this article is organized as follows .
next section is devoted to the proofs of theorem [ th1 ] , [ th2 ] and [ th3 ] whereas various applications of these results to the study of chains of oscillators and the generalized langevin equation are discussed in section [ applications ]
. * acknowledgement .
* the third author would like to express his gratefulness to johannes sjstrand for enriching remarks and comments on the present work .
this section is devoted to the proofs of theorems [ th1 ] , [ th2 ] and [ th3 ] .
we begin by noticing that the set @xmath133 is non - empty .
indeed , the work @xcite gives an alternative description of the singular space in terms of eigenspaces of the hamilton map @xmath47 associated to its real eigenvalues . more specifically , when @xmath27 is a complex - valued quadratic form with a non - negative real part fulfilling the partial ellipticity condition on its singular space @xmath134 which is obviously satisfied when @xmath32 , then the set of real eigenvalues of the hamilton map @xmath47 can be written as @xmath135 with @xmath136 and @xmath137 if @xmath138 . the singular space is then ( @xcite , p.807 ) the direct sum of the symplectically orthogonal spaces @xmath139 where @xmath140 is the symplectic space @xmath141 recalling from proposition 4.4 in @xcite that the subvector space @xmath142 is the complexification of its intersection with @xmath23 , condition @xmath32 therefore implies that the hamilton map @xmath47 has no real eigenvalue @xmath143 on the other hand , we know from @xcite ( lemma 21.5.2 ) that @xmath144 the space of generalized eigenvectors of @xmath47 belonging to the eigenvalue @xmath64 is the dual with respect to the symplectic form of the space of generalized eigenvectors @xmath145 belonging to the eigenvalue @xmath146 .
it follows that @xmath147 this implies that the set @xmath148 is non - empty ; and that the quantities @xmath149 and @xmath72 are well - defined in ( [ vp0 ] ) and ( [ rate ] ) . since from ( [ sm6 ] ) the spectrum of the operator @xmath28 is only composed of the following eigenvalues with finite algebraic multiplicity @xmath150 we notice that the first eigenvalue in the bottom of the spectrum is equal to @xmath151 and we easily identify the spectral gap to be the positive rate @xmath152 it remains to check that this eigenvalue @xmath71 has algebraic multiplicity one with an eigenspace @xmath153 spanned by a ground state of exponential type @xmath75 where @xmath76 is a complex - valued quadratic form on @xmath77 whose real part is positive definite . in order to do so , we shall first consider the case when the real part of the quadratic symbol @xmath154 is positive definite . in this specific case ,
the symbol @xmath27 is elliptic on the whole phase space @xmath23 , @xmath155 and one can refer to the original work by j. sjstrand on elliptic quadratic differential operators @xcite . following this fundamental work on which relies the analysis led in @xcite
, we consider the positive lagrangian plane ( proposition 3.3 in @xcite ) , @xmath156 where @xmath144 is the space of generalized eigenvectors belonging to the eigenvalue @xmath157 of the hamilton map @xmath47 associated to the quadratic symbol @xmath27 .
we recall that a complex subspace @xmath158 of dimension @xmath159 is called lagrangian if the canonical symplectic form @xmath45 vanishes identically on @xmath160 .
furthermore , a lagrangian plane @xmath158 is said to be positive if @xmath161 for all @xmath162 according to ( 3.5 ) in @xcite ( p.95 ) , there exists a symmetric matrix @xmath163 whose imaginary part @xmath164 is positive definite such that @xmath165 define the two quadratic forms @xmath166 notice that the real part of the quadratic form @xmath76 is positive definite . keeping on following the analysis led in @xcite ( p.98 ) , on
can then use a jordan decomposition for the hamilton map @xmath47 in order to find a canonical transformation @xmath167 of @xmath63 such that @xmath168 with @xmath169 .
here , @xmath170 , with @xmath171 , @xmath172 and @xmath173 ; stands for a relabeling of the standard real symplectic coordinates @xmath174 ; and @xmath175 with @xmath176 .
we recall that @xmath62 stands for the dimension of @xmath144 the space of generalized eigenvectors associated to the eigenvalue @xmath64 .
this reduction is proved in @xcite ( p.98 ) in ( 3.15 ) and ( 3.16 ) .
consider the two quadratic operators @xmath177 and @xmath178 notice that these two operators commute and that @xmath179 proposition 3.10 in @xcite shows that every monomial is a generalized eigenvector of the operator @xmath180 where @xmath181 stands for the space of complex polynomials regarded as functions on @xmath77 .
more specifically , the spectrum of the operator @xmath182 acting on the space @xmath181 is exactly given @xmath183 proposition 3.10 in @xcite also shows that the space of generalized eigenvectors @xmath184 associated to an eigenvalue @xmath185 is of finite dimension @xmath186 inspecting the proof of proposition 3.10 in @xcite shows that @xmath187 for any homogeneous polynomial of degree @xmath188 in the variables @xmath189 , ... , @xmath190 .
this implies that the monomials @xmath191 constitute a basis of eigenvectors for the operator @xmath192 acting on @xmath181 , @xmath193 according to this description , we notice that the first eigenvalue in the bottom of the spectrum for the operator @xmath192 acting on the space @xmath181 is the value @xmath71 defined in ( [ vp0 ] ) .
furthermore , this eigenvalue @xmath71 has algebraic multiplicity 1 .
indeed , all the non - negative integers @xmath188 must necessarily be zero and the only possible eigenvector associated to @xmath71 is up to a non - zero constant the constant polynomial @xmath194 . keeping on following the proof of proposition 3.10 in @xcite
, one may notice that the operator @xmath195 is nilpotent on any space of generalized eigenvectors @xmath196 with @xmath197 ; for the operator @xmath192 seen as acting on the space @xmath181 .
this proves in particular that @xmath71 is an eigenvalue with algebraic multiplicity 1 for the operator @xmath180 with eigenvector equal to the constant polynomial 1 .
according to @xcite ( p.100 ) , there exists a bijection @xmath198 such that @xmath199 more specifically , according to ( 3.22 ) , ( 3.23 ) , ( 3.24 ) and ( 3.25 ) in @xcite , this mapping may be written as @xmath200 with some operators @xmath201 @xmath202 and @xmath203 where @xmath204 is the fourier transformation , @xmath205 a quadratic form and @xmath206 an invertible complex matrix .
proposition 3.11 in @xcite then establishes that the spectrum of the operator @xmath207 is also exactly given by ( [ sach1 ] ) ; and that for every eigenvalue the corresponding space of generalized eigenfunctions is finite dimensional .
more specifically , we deduce from ( [ sach2 ] ) , ( [ sach13 ] ) , ( [ sach3 ] ) , ( [ sach4 ] ) and ( [ sach5 ] ) that @xmath71 is an eigenvalue with algebraic multiplicity 1 for the operator @xmath207 with associated eigenvector @xmath208 .
an explicit computation shows that @xmath209 with @xmath210 . as a final step in the proof of theorem 3.5 , it is proved in @xcite ( p.102 ) that for any @xmath197
, the space of generalized eigenvectors associated to the eigenvalue @xmath185 for the operator @xmath207 exactly coincides with the space of generalized eigenvectors associated to the eigenvalue @xmath185 for the operator @xmath211 that is @xmath212 this implies that when the real part of the quadratic symbol @xmath154 is positive definite , @xmath71 the first eigenvalue in the bottom of the spectrum of the operator @xmath28 acting on @xmath35 has algebraic multiplicity 1 with a ground state of exponential type @xmath213 with @xmath214 where @xmath76 is a complex - valued quadratic form on @xmath77 , whose real part is positive definite .
this proves theorem [ th1 ] in the special case when the real part of the quadratic symbol @xmath154 is positive definite .
let us now consider the general case when the quadratic symbol @xmath33 is a complex - valued quadratic form with a non - negative real part @xmath29 and a zero singular space @xmath32 .
we shall now inspect the analysis led in @xcite in order to prove theorem [ th1 ] in the general case .
notice that in this general case the real part of the quadratic symbol @xmath154 is not necessarily positive definite .
however , as noticed in @xcite ( p.809 - 810 ) ( see also proposition 2.0.1 in @xcite ) and mentioned above , the fact that the singular space @xmath25 is reduced to zero implies that the average of the real part of the quadratic form @xmath27 by the flow generated by the hamilton vector field of its imaginary part @xmath128 , @xmath215 is positive definite for any @xmath132 .
we shall now inspect more specifically the proof of proposition 3.1.1 in @xcite .
notice that we consider in @xcite complex - valued quadratic symbols with non - positive real parts .
one can therefore apply all this analysis when considering the symbol @xmath216 instead of @xmath27 and this accounts for sign differences when referring to results proved in @xcite .
following @xcite ( p.826 ) , we shall discuss certain ir - deformations of the real phase space @xmath23 , where the averaging procedure along the flow defined by the hamilton vector field of @xmath50 plays an important rle . to that end , let @xmath217 be a real - valued quadratic form on @xmath23 satisfying @xmath218 as in @xcite , we solve ( [ eq6 ] ) by setting @xmath219 where @xmath220 and @xmath221 is the odd function given by @xmath222 let us notice that @xmath223 and @xmath224 have a jump of size @xmath194 at the origin .
associated with @xmath225 is a linear ir - manifold , defined for @xmath226 , with @xmath227 small enough @xmath228 where @xmath229 stands for the flow generated by the linear hamilton vector field @xmath230 taken at time 1 . keeping on following @xcite ( p.827 ) ( see ( 3.1.23 ) ) , we consider the quadratic symbol @xmath231 whose real part is easily seen to be equal to @xmath232 we deduce from ( [ eq6 ] ) and the facts that the quadratic form @xmath233 is positive definite and @xmath29 , @xmath234 for @xmath235 and @xmath236 .
we proved in @xcite ( lemma 3.1.1 ) that for any @xmath237 , with @xmath227 sufficiently small , the spectrum of the quadratic operator @xmath238 is actually independent of the parameter @xmath237 , @xmath239 where @xmath47 is the hamilton map associated to the quadratic form @xmath27 , @xmath62 is the dimension of the space of generalized eigenvectors of @xmath47 in @xmath63 belonging to the eigenvalue @xmath64 .
the independence of the spectrum of the operator @xmath238 with respect to the parameter @xmath237 , is linked to the fact that the hamilton map @xmath240 of the quadratic form @xmath241 is isospectral for any @xmath242 with the hamilton map @xmath47 of the quadratic form @xmath27 , since the symbols @xmath27 and @xmath241 are related by a canonical transformation . as a final step in @xcite , we proved in ( p.830 - 831 ) that if the positive constant @xmath227 is chosen sufficiently small then the spectrum of the operator @xmath28 is equal to the spectrum of the operator @xmath238 for any @xmath237
. furthermore , eigenvectors and generalized eigenvectors agree , see @xcite ( p.830 - 831 ) and @xcite ( proposition 2.1 ) .
this implies that one can deduce the result of theorem [ th1 ] in the general case when the quadratic symbol has a non - negative real part @xmath29 and a zero singular space @xmath32 from the one proved previously in the particular case when the real part of the symbol is positive definite .
this ends the proof of theorem [ th1 ] .
we consider @xmath33 a complex - valued quadratic form with a non - negative real part @xmath83 and zero singular space @xmath32 .
according to the definition of the singular space ( [ h1bis ] ) , there exists a smallest integer @xmath84 such that @xmath243\big ) \cap { \mathbb{r}}^{2n}=\{0\},\ ] ] where @xmath47 stands for the hamilton map of @xmath27 ; and theorem 1.2.1 in @xcite shows that the operator @xmath28 fulfills a global subelliptic estimate with loss of @xmath86 derivatives with respect to the elliptic case , that is , there exists @xmath87 such that for all @xmath88 , @xmath244 with @xmath90 by using the proof given for establishing theorem 1.2.1 in @xcite , one can actually directly obtain the more general estimate that there exists a constant @xmath87 such that for all @xmath88 and @xmath245 , @xmath246 to prove this fact , it is actually sufficient to notice that one can replace the operator @xmath247 by the operator @xmath248 in the estimate ( 2.10 ) in @xcite .
indeed , we recall that real hamiltonians get quantized in the wick quantization by formally selfadjoint operators on @xmath9 .
since the weight function @xmath249 given by proposition 2.0.1 in @xcite is real - valued and @xmath250 a positive constant , we have the equality @xmath251 which justifies that one can actually replace the operator @xmath247 by the operator @xmath248 in the estimate ( 2.10 ) in @xcite .
then , by noticing that @xmath252 one can also replace the operator @xmath247 by the operator @xmath248 in the estimate preceding the estimate ( 2.19 ) in @xcite
. finally , after having done these two slight modifications , one can then use exactly the same proof as the one given for proving theorem 1.2.1 in @xcite in order to establish the estimate ( [ est2 ] ) .
we now need to recall few facts about the relationship between pseudo - differential calculus and functional calculus when using the sobolev scale @xmath253 , @xmath254 ; defined by the operator @xmath255 in order to do so , it will be convenient to directly refer to the presentation given by f. hrau and f. nier in @xcite ( appendix a , p.205 ) .
our study corresponds to the specific case when the potential @xmath256 in @xcite is taken quadratic @xmath257 define @xmath258 and @xmath259 by referring to @xcite ( proposition a.4 ) , we notice that , for any @xmath254 , the operator @xmath253 is a pseudo - differential operator whose weyl symbol belongs to the symbol class @xmath260 composed by @xmath261 functions satisfying @xmath262 furthermore , for any @xmath254 , the domain of the operator @xmath253 is equal to the space @xmath263 with equivalence of the two norms @xmath264 and @xmath265 .
it follows from ( [ est2 ] ) that there exists a constant @xmath87 such that for all @xmath88 and @xmath245 , @xmath266 consider now the metric @xmath267 and weight @xmath268 .
we recall for instance from @xcite ( lemma 2.2.18 and section 2.2 ) that this metric is admissible with gain @xmath269 .
any quadratic form is obviously a first order symbol in the symbolic calculus associated to this metric @xmath270 one can then directly deduce from the fefferman - phong inequality ( see for instance theorem 2.5.5 in @xcite ) that there exists a positive @xmath87 such that for all @xmath271 , @xmath272 with @xmath273 notice from theorem [ th1 ] together with ( [ vp0 ] ) and ( [ mi1 ] ) that @xmath274 one can apply the abstract functional analysis led in ( * ? ? ?
6.1 ) in order to obtain a specific control of the resolvent of the operator @xmath28 in particular regions of the resolvent set . more specifically , following the analysis in @xcite
, we deduce from the hypoelliptic estimates ( [ est3 ] ) and ( [ est4 ] ) that there exists some positive constants @xmath275 and @xmath206 such that @xmath276 and @xmath277 for all @xmath278 such that @xmath279 and @xmath280 . the property of exponential return to equilibrium may deduced from those resolvent estimates . indeed ,
theorem [ th2 ] is a consequence of the abstract analysis led in ( * ? ? ?
* theorem 6.1 ) or @xcite ( section 12 , p.754 - 756 ) .
following more specifically the presentation given in @xcite ( section 12 ) , we notice from ( [ y1 ] ) and ( [ dl3 ] ) that under the assumptions of theorem [ th2 ] , the quadratic operator @xmath28 is a closed densely defined unbounded operator acting on @xmath35 fulfilling the assumptions ( 12.1 ) and ( 12.2 ) in @xcite ( p.754 ) with @xmath281 define for any @xmath282 , @xmath283 where @xmath4 is a contour to the left of the spectrum that outside a compact set coincides with the curve @xmath284 with @xmath87 and @xmath4 oriented in the direction of decreasing @xmath285 . according to ( [ y1 ] ) and ( [ dl3 ] ) , this integral is convergent and defines a bounded operator depending smoothly on the parameter @xmath282 .
it is then proved in @xcite that this integral operator is actually the semigroup @xmath286 generated by the accretive quadratic operator @xmath28 , @xmath287 let @xmath92 with @xmath288 the positive rate given by theorem [ th1 ] . keeping on following the analysis led in @xcite ( p.755 ) in the particular case when @xmath289 , we introduce two contours @xmath4 and @xmath290 .
both contours are given by the curve @xmath284 in the region where @xmath291 , with @xmath292 .
in the region , where @xmath293 , the contour @xmath4 is given by the equation @xmath294 ; while @xmath290 joins the two points @xmath295 further to the left so that @xmath290 is entirely to the left of the spectrum of the quadratic operator @xmath28 , while @xmath4 will only have the first eigenvalue in the bottom of the spectrum @xmath71 given by theorem [ th1 ] to its left , since @xmath92 .
let @xmath296 denote the vertical part of the contour @xmath4 in the region where @xmath294 and let @xmath297 denote the part of @xmath4 in the region where @xmath298 . on the exterior piece @xmath297 , we have the resolvent estimate @xmath299 whereas on the interior piece @xmath296 , we have @xmath300 since the compact contour @xmath296 does not intersect the spectrum of the operator @xmath28 .
according to theorem [ th1 ] , @xmath71 is an eigenvalue with algebraic multiplicity 1 for the operator @xmath28 .
arguing as in ( 12.13 ) in the article @xcite , we may write that @xmath301 where @xmath302 stands for the rank - one spectral projection associated with the eigenvalue @xmath71 for the operator @xmath28 . using the decomposition @xmath303 together with ( [ y3 ] ) and ( [ y4 ] ) , one can then estimate from the above the two parts of the integral @xmath304 as follows @xmath305 @xmath306 uniformly for @xmath307 .
since the accretive operator @xmath28 defines a contraction semigroup , we deduce from ( [ y5 ] ) , ( [ y6 ] ) and the previous estimates that @xmath308 uniformly for @xmath309 . recalling that @xmath292 and considering now the operator @xmath310 we directly deduce from ( [ y7 ] ) that there exists a positive constant @xmath87 such that @xmath93 where @xmath94 is the rank - one spectral projection associated with the simple eigenvalue zero of the operator @xmath95 .
let @xmath28 be a quadratic operator satisfying the assumptions of theorem [ th1 ] .
assume furthermore that this quadratic operator is real and satisfies ( [ y11 ] ) . as noticed in the discussion preceding the statement of theorem [ th3 ] , the lowest eigenvalue @xmath71 is necessarily real and the quadratic form @xmath76 defining the ground state @xmath100 is positive definite .
it remains to check that the rank - one spectral projection @xmath94 associated with the simple eigenvalue zero of the operator @xmath310 provided by theorem [ th2 ] is orthogonal @xmath311 in order to deduce theorem [ th3 ] from theorem [ th2 ] .
this property is a direct consequence of the assumption @xmath312 indeed , notice from ( 6.52 ) in ( * ? ? ? *
iii.6 ) that the adjoint of the spectral projection @xmath94 associated with the eigenvalue zero for the operator @xmath95 is also a spectral projection for the operator @xmath313 , @xmath314 with @xmath315 being the mirror image of @xmath316 with respect to the real axis and @xmath316 a circular contour centered in the eigenvalue @xmath317 with a sufficiently small radius .
notice that in the integrals ( [ y14 ] ) both contours @xmath316 and @xmath315 are taken in the positive direction and that @xmath318 is actually the spectral projection associated with the eigenvalue zero for the operator @xmath313 . since @xmath94 and @xmath318 are both projections , one may decompose the @xmath35 space in two ( possibly different ) direct sums @xmath319 it follows from ( [ yy11 ] ) that @xmath320 indeed , theorem [ th1 ] applies for both quadratic operators @xmath28 and @xmath121 ; and the eigenvalue zero has therefore algebraic multiplicity 1 for both operators @xmath95 and @xmath313 .
furthermore , since @xmath321 we conclude that the spectral projection @xmath94 is orthogonal @xmath322 this ends the proof of theorem [ th3 ] .
in this section , several applications of the singular space theory are discussed . as pointed out in @xcite , a noticeable example of quadratic operator with zero singular space
is the kramers - fokker - planck operator considered on the unweighted @xmath323 space @xmath324 with a quadratic potential @xmath325 indeed , this operator writes as @xmath326 with a symbol @xmath327 which is a non - elliptic complex - valued quadratic form whose real part is non - negative .
simple algebraic computations show that the associated hamilton map @xmath328 is @xmath329 and that the singular space is equal to zero @xmath32 .
more precisely , the integer @xmath330 appearing in ( [ er133 ] ) is equal to 1 : @xmath331 now a computation shows that @xmath64 is an eigenvalue of @xmath332 precisely when @xmath333 and we easily see that when @xmath334 , the eigenvalues @xmath157 of @xmath332 with @xmath335 are given by @xmath336 when @xmath337 , we get the eigenvalues @xmath338 according to ( [ sm6 ] ) , the spectrum of the operator @xmath339 is given by @xmath340 in particular , when @xmath337 , we observe that the lowest eigenvalue of the spectrum of the operator @xmath339 is @xmath341 whereas when @xmath334 , this lowest eigenvalue is equal to @xmath342 the spectral gap @xmath72 appearing in theorem [ th1 ] is respectively equal to @xmath343 when @xmath344 notice that the lowest eigenvalue is always real .
this is consistent with the fact that the quadratic operator @xmath339 is real .
further calculations allow to determine explicitly the ground state @xmath345 when @xmath337 , we obtain the usual maxwellian @xmath346 whereas when @xmath334 , the ground state is given by @xmath347 notice further that the assumption @xmath348 holds true only if @xmath337 .
theorems [ th1 ] and [ th2 ] therefore apply whenever @xmath349 , whereas theorem [ th3 ] only applies when @xmath337 .
this application comes from the series of works @xcite and was studied in @xcite ( section 6 ) .
we shall begin by recalling the setting of the analysis @xcite and explain how the results obtained there relate with the present work .
this example is a model describing a chain of two oscillators coupled with two heat baths at each side .
the particles are described by their respective position and velocity @xmath350 . for each oscillator @xmath351 ,
the particles are submitted to an external force derived from a real - valued potential @xmath352 and a coupling between the two oscillators derived from a real - valued potential @xmath353 .
we write @xmath256 the full potential @xmath354 @xmath355 the velocities and @xmath356 the variables describing the state of the particles in each of the heat baths . in each bath , the particles are submitted to a coupling with the nearest oscillator , a friction force given by the friction coefficient @xmath4 and a thermal diffusion at temperature @xmath357 , @xmath351 .
we denote by @xmath358 , @xmath351 , two @xmath359-dimensional brownian motions with mean 0 and variance 1 ; and write @xmath360 .
the system of equations describing this model reads as ( see @xcite ) , @xmath361 following @xcite and setting @xmath362 , @xmath363 , the corresponding equation for the density of particles becomes @xmath364 define @xmath365 and @xmath366 with @xmath367 and @xmath87 .
one may check that the function @xmath368 , with @xmath369 , is a maxwellian of the process when the temperatures are the same .
considering now the general case when temperatures may be different , the function @xmath368 is used to define the weighted space @xmath370 . in order to work in the flat @xmath9 space
, we change the unknown by setting @xmath371 . the new equation for the unknown @xmath99 reads as @xmath372 we consider the case when external potentials are quadratic .
for simplicity , we may assume that @xmath289 , @xmath373 , @xmath374 and take @xmath375 with @xmath376 . equation ( [ osc3 ] ) writes as @xmath377 where @xmath378 is the quadratic operator with symbol @xmath379,\end{gathered}\ ] ] with @xmath380 the notations @xmath381 standing respectively for the dual variables of @xmath382 . the condition @xmath383 appearing in @xcite ( section 6 ) exactly ensures that this quadratic symbol has a non - negative real part @xmath29 .
notice that only the case with identical temperatures @xmath369 is discussed in @xcite .
more precisely , the authors mention that no maxwellian is known in the case of different temperatures and that it prevents them from finding any supersymmetric structure . here
, we consider the general case with possibly different temperatures @xmath384 for the quadratic potentials defined in ( [ osc10 ] ) and assume that @xmath385 direct algebraic computations using the definition of hamilton maps and the explicit expressions @xmath386 @xmath387 show that @xmath388 with @xmath389 @xmath390 and @xmath391 with @xmath392 @xmath393 @xmath394 @xmath395 it follows that @xmath396 @xmath397 and @xmath398 when the condition @xmath399 holds , the singular space reduces to @xmath400 after intersecting exactly @xmath401 kernels with here @xmath402 .
notice that condition ( [ osc4 ] ) corresponds exactly to the assumption @xmath403 is a morse function required in @xcite ( lemma 6.1 ) to ensure that the needed dynamical conditions hold .
the other conditions @xmath404 , when @xmath405 , for @xmath406 ; and @xmath407 when @xmath408 , are also satisfied for quadratic potentials fulfilling condition ( [ osc4 ] ) . when this condition holds , theorems [ th1 ] and [ th2 ] applies .
this shows in particular that the first eigenvalue ) ] in the bottom of the spectrum @xmath71 has algebraic multiplicity one with an eigenspace @xmath409 spanned by a ground state of exponential type @xmath410 where @xmath76 is a complex - valued quadratic form on @xmath411 whose real part is positive definite .
notice that the operator @xmath412 is real .
the discussion preceding the statement of theorem [ th3 ] then shows that the quadratic form @xmath76 is positive definite .
this proves the existence of a maxwellian @xmath413 in the general case when the temperatures may be different @xmath384 . it would be interesting to push further the computations in order to derive the exact expressions for the eigenvalue @xmath71 , the spectral gap @xmath288 and the maxwellian @xmath413 .
another question of interest would be to investigate if theorem [ th3 ] applies by checking the validity of condition ( [ y11 ] ) .
those questions are left for a future work . from now
, we may only notice from ( [ y1 ] ) and ( [ dl3 ] ) that the following resolvent estimate holds @xmath414 @xmath415 for all @xmath278 such that @xmath279 and @xmath416 , since the integer @xmath330 is here equal to 2 .
the spectrum of the fokker - planck operator associated to a linear stochastic differential equation in @xmath417 has been calculated in @xcite , see also @xcite . in this section
we consider a particular system of linear stochastic differential equations that is obtained as a finite - dimensional markovian approximation of the non - markovian generalized langevin equation ( gle ) in @xmath3 , @xmath418 where @xmath419 is a smooth confining potential and @xmath420 a mean zero stationary gaussian process with autocorrelation function @xmath421 , in accordance to the fluctuation - dissipation theorem @xmath422 here @xmath423 stands for the inverse temperature and @xmath424 for the identity matrix .
the gle equation together with the fluctuation - dissipation theorem appear in various settings such as surface diffusion @xcite and polymer dynamics @xcite .
this equation also serves as one of the standard models of non - equilibrium statistical mechanics describing the dynamics of a ` small ' hamiltonian system ( the distinguished particle ) coupled to one or more heat baths which are modelled as linear wave equations with initial conditions distributed according to appropriate gibbs measures @xcite . for further details ,
see @xcite and the references therein . for simplicity , we will consider the case @xmath374 . as in @xcite , we study the case when the memory kernel is a sum of exponentials @xmath425 with @xmath426 , @xmath136 , @xmath427 . equation ( [ gle ] ) is then equivalent to the dynamics of the following system @xmath428 with @xmath429 , @xmath430 and @xmath431 being independent standard brownian motions .
we consider the specific case when @xmath256 is a non - degenerate quadratic potential @xmath432 with @xmath433 .
the density of the invariant measure associated to the dynamics ( [ system1 ] ) is then given by @xmath434 with @xmath435 being the normalization constant and @xmath436 the euclidean norm on @xmath437 .
this process is markovian whose generator @xmath438 is an hypoelliptic operator acting on @xmath439 , with @xmath9-adjoint @xmath440 acting on @xmath441 . by using the transformation @xmath442 in order to work with an operator acting on the unweighted @xmath443 space @xmath444 and setting @xmath445 we notice that @xmath446 is a quadratic operator , whose symbol is given by @xmath447 this quadratic symbol
has a non - negative real part @xmath448 and a direct computation shows that its hamilton map @xmath449 is given by @xmath450 , with @xmath451 @xmath452 it directly follows that @xmath453 @xmath454 @xmath455 this proves that the singular space associated to the symbol @xmath27 reduces to @xmath400 after intersecting exactly @xmath401 kernels with here @xmath456 . furthermore , notice that the quadratic operator @xmath457 is real .
setting @xmath458 and @xmath459 a direct computation shows @xmath460 on the other hand , notice that @xmath461 since @xmath462u , u)_{l^2}\ ] ] and @xmath463u , u)_{l^2}=\frac{1}{2}\sum_{j=1}^m \alpha_j\|u\|_{l^2}^2=\mu_0\|u\|_{l^2}^2.\ ] ] here the notation @xmath464 $ ] stands for the commutator of the two operators @xmath465 and @xmath466 .
this implies that @xmath71 is necessarily the lowest eigenvalue in the spectrum of the quadratic operator @xmath467 and that @xmath468 is the associated ground state appearing in theorem [ th1 ] .
we notice further that all the assumptions of theorems [ th1 ] , [ th2 ] and [ th3 ] hold true .
thus , these three theorems apply .
this proves the property of exponential return to equilibrium for this finite - dimensional markovian approximation of the non - markovian generalized langevin equation with quadratic potential .
here , theorem [ th1 ] provides an exact formula for the rate of convergence that may be computed explicitly @xmath469 for simplicity , consider the specific case when @xmath470 and denote @xmath157 and @xmath471 for the parameters @xmath472 and @xmath473 .
an explicit computation of the characteristic polynomial of @xmath47 shows that @xmath474 this is a polynomial equation of the third degree in the variable @xmath475 which may be solved explicitly . refraining from using cardan formulas for giving an explicit , but rather complicated , formula for the rate of convergence @xmath288
, we may instead try to compute numerically this rate in order to study its dependence on the parameters @xmath471 and @xmath157 . a limited attempt in this direction is provided by the following calculations where
the rate @xmath288 is numerically computed for different values of the parameters @xmath157 and @xmath471 in the case when @xmath476 .
notice in particular that the spectral gap seems to be maximized along a curve of the type @xmath477 , with @xmath478 a positive constant . on the second figure ,
the spectral gap is computed as a function of the parameter @xmath157 while keeping the friction coefficient @xmath479 fixed .
we observe that the spectral gap becomes almost constant for a large range of value s of @xmath157 .
this is in accordance with the numerical observations reported in @xcite
. it would be interesting to maximize the spectral gap , i.e. to optimize the rate of convergence to equilibrium , by choosing appropriately the parameters @xmath480 and @xmath481 in .
this question , together with possible applications to markov chain monte carlo ( mcmc ) techniques will be studied in future work . | we study the problem of convergence to equilibrium for evolution equations associated to general quadratic operators .
quadratic operators are non - selfadjoint differential operators with complex - valued quadratic symbols . under appropriate assumptions , a complete description of the spectrum of such operators
is given and the exponential return to equilibrium with sharp estimates on the rate of convergence is proven . some applications to the study of chains of oscillators and the generalized langevin equation are given
. [ section ] [ section ] [ section ] [ section ] |
gravitational topological objects may have been formed in the early universe during its phase expansion due to spontaneously symmetry braking @xcite .
depending on the topology of the vacuum manifold , @xmath2 , these are domain walls , strings , monopoles and texture , corresponding to the homotopy groups @xmath3 , @xmath4 , @xmath5 and @xmath6 , respectively .
global monopoles are heavy topological objects formed in the phase transition of a system composed by a self - coupling iso - triplet scalar field , @xmath7 , whose original global @xmath8 symmetry is spontaneously broken to @xmath9 .
the scalar matter field plays the role of an order parameter which outside the monopole s core acquires a non - vanishing value . the global monopole was first introduced by sokolov and starobinsky in @xcite , and its gravitational effects have been analyzed by barriola and vilenkin @xcite . in @xcite , has been shown that for points far away from the monopole s center the geometry of the spacetime can be given by the following line element : @xmath10 where the parameter @xmath11 , smaller than unity , depends on the energy scale @xmath12 where the global symmetry is spontaneously broken .
the spacetime described by ( [ gm ] ) has a non - vanishing scalar curvature , @xmath13 , and the solid angle of a sphere of unit radius is smaller than the usual one , presenting in this way , a solid angle deficit : @xmath14 .
although the geometric properties of the spacetime outside the monopole are very well understood , the analysis of the metric tensor in the region inside requires the complete knowledge of the energy - momentum tensor associated with the scalar field , @xmath15 , which on the other hand depends on the knowledge of the components of the metric tensor , providing , in this way , a non solvable integral equation @xcite . because of this fact , many interesting investigations of physical effects associated with global monopole consider this object as a point - like defect . in this way
, it is considered that the geometry of the whole spacetime is described by ( [ gm ] ) ; however , adopting this model , calculations of vacuum polarization effects , for example , present divergences on the monopole s core .
one of most interesting phenomenon associated with the gravitational topological defects , is related with the induced self - energy on an electric charged particle placed at rest in their neighborhood .
this effect has been analysed in an idealized cosmic string spacetime by linet @xcite and smith @xcite , independently , and in the spacetime of a point - like global monopole in @xcite . in these analysis , the induced self - energies present divergences on the respective defects core . in order to avoid this problem , a more realistic model for the defects should take into account a inner structure for them . as to cosmic string
, two different models have been adopted to describe the geometry inside it : the ballpoint - pen model proposed independently by gott and hiscock @xcite , replaces the conical singularity at the string axis by a constant curvature spacetime in the interior region , and flower - pot model @xcite , presents the curvature concentrated on a ring with the spacetime inside the string been flat .
khusnutdinov and bezerra in @xcite , revisited the induced electrostatic self - energy problem considering the hiscock and gott model for the region inside the string . recently
the electrostatic self - energy problem in the context of global monopole has been analyzed considering the flower - pot model @xcite for the region inside . in order to complete the analysis of electrostatic self - energy in the global monopole spacetime ,
the propose of this paper is to consider the ballpoint - pen model for the region inside .
this paper is organized as follows : in section [ sect1 ] is presented the model adopted to describe the spacetime in the region inside the global monopole and some its geometric properties . in section [ sect2 ] , we write the maxwell equations in the corresponding spacetime , and derive the green function for points outside and inside to the monopole s core . in section [ sect3 ]
we calculate the corresponding renormalized induced electrostatic self - energy , and give its behavior in various asymptotic regions of the parameters . in section [ sect4 ]
we present our conclusions and more relevant remarks .
the simplest model which gives rise to a global monopole has been proposed by barriola and vilenkin @xcite , and is described by a lagrangian whose original @xmath8 global gauge symmetry is spontaneously broken to @xmath9 .
the influence of this object on the geometry of the spacetime can be analysed by coupling the energy - momentum tensor associated with this matter field with einstein equations .
barriola and vilenkin have shown that for points very far away form the monopole s center , the spacetime is described by the line element ( [ gm ] ) . although there are no analytical solutions for the metric tensors in the inner region , harari and loust @xcite presented a simplified version for the geometry in this region .
they have shown that it can be represented by a de sitter spacetime , whose constant scalar curvature is proportional to the energy scale @xmath12 where the symmetry is broken .
although this model shares the main features of the exact solution , the analysis of the three - dimensional green function in the region inside the monopole is a very difficult task . another model which also shares some features with the exact solution and allows us to obtain closed solutions for this function is the ballpoint - pen model .
this model is given as shown below : * for the region outside the monopole the metric of the spacetime is given by ( [ gm ] ) , where the radial coordinates is defined in the interval @xmath16 , being @xmath17 the radius of the monopole . * for inner region
the metric tensor is given by the following line element @xmath18 with the new radial coordinate defined in the interval @xmath19 $ ] .
the inner solution can match the exterior solution without the necessity to include an infinitely thin shell at the boundary , provided the junction conditions below : @xmath20 in the inner region , the non zero components of the riemann and ricci tensors and scalar curvature are : @xmath21 it is possible to describe the metric tensor in the inner region by using the radial @xmath22 coordinate through the relation @xmath23 in this case , the line element ( [ gm1 ] ) can be written by @xmath24 where @xmath25
with the objective to construct the green function associated with an electric charged particle at rest in the spacetime of a global monopole , we write down the maxwell equation in an arbitrary curved spacetime @xmath26 with @xmath27 where @xmath28 and @xmath29 are the four - vector potential and current , respectively . for a point - like particle at rest with
coordinates @xmath30 , the static four - vector current and potential read : @xmath31 and @xmath32 .
the only nontrivial component of ( [ meq ] ) is for @xmath33 , with @xmath34being @xmath35 is the charge of the particle .
the corresponding maxwell equation is written as @xmath36 where the three - dimensional laplacian operator in coordinates system given in ( [ gm2 ] ) is @xmath37a^{0 } \ , \end{aligned}\ ] ] being @xmath38 the angular momentum operator .
moreover in this coordinate @xmath39 .
the green function associated with the respective operator is obtained by substituting @xmath40 into the above equations .
so , this non - homogeneous differential equation reads : @xmath41 g(\vec{r},\vec{r}')=-\frac{\delta(r - r')\delta(\theta-\theta')\delta(\varphi-\varphi')}{\alpha^2\sin\theta } \ .\end{aligned}\ ] ] taking into account the spherical symmetry of the problem , we may present the green function as the expansion @xmath42 with @xmath43 being the ordinary spherical harmonics . substituting ( [ green - b ] ) into ( [ green - a ] ) and using the well known closure relation for the spherical harmonics , we arrive at the following differential equation for the radial function : @xmath44g_l(r , r')= -\frac{\delta(r - r')}{\alpha^{2 } } \ .\end{aligned}\ ] ] the junction of the first radial derivative at @xmath45 is obtained by integrating the above equation about this point : @xmath46 the radial green function is calculated by the standard method : @xmath47 where @xmath48 and @xmath49 are the two linearly independent solutions of the homogeneous equation corresponding to ( [ gr ] ) .
we assume that @xmath48 is regular at the core center , @xmath49 goes to zero at infinity , and that these solutions are normalized by the wronskian relation @xmath50moreover , both solutions must be regular with their first radial derivative regular at the boundary , i.e. , @xmath51 for @xmath52 . in the region outside the core
the linearly independent solutions to the corresponding homogeneous equation are the functions @xmath53 and @xmath54 , where @xmath55 and for the region inside , in coordinates given in ( [ gm2 ] ) , the regular and singular solutions are @xmath56 and @xmath57 , respectively , with @xmath58 , being @xmath59 the associated legendre functions .
now taking into account ( [ f1 ] ) , ( [ wronin ] ) and ( [ f2 ] ) , the radial green functions are given by : * for the charged particle outside the monopole s core , @xmath60 , we have : @xmath61 and @xmath62 \ , \ { \rm for } \ \
r\geq r_0 \ \end{aligned}\ ] ] with @xmath63 * for the charged particle inside the monopole s core , @xmath64 , we have : @xmath65 \ , \
{ \rm for } \ \ r\leq r_0\end{aligned}\ ] ] and @xmath66 \ , \ { \rm for } \
\ r\geq r_0 \ \end{aligned}\ ] ] with ) and ( [ g4 ] ) , we have used the corresponding wronskian for the associated legendre functions @xcite : @xmath67=\frac{2\sin\mu\pi}{\pi(1-x^2)}$ ] . ]
@xmath68 and @xmath69 in these formulas we have adopted @xmath70 and @xmath71 , with the same convention for @xmath72 and @xmath73 .
first , let us consider the case where the charged particle is outside the monopole s core . substituting ( [ g2 ] ) into ( [ green - b ] )
, we have @xmath74 where the first term on the right hand side of ( [ ap ] ) is the green function for the geometry of a point - like global monopole , @xmath75 , and the second , @xmath76 , is induced by the non - trivial structure of the core .
@xmath77 is the angle between the directions @xmath78 and @xmath79 . for the particle inside the monopole s core ,
the green function is given by substituting ( [ g3 ] ) into ( [ green - b ] ) , the result is : @xmath80 being @xmath81 also , this function presents two contributions : the first one , @xmath82 , is the green function in the background geometry described by the line element ( [ gm1 ] ) using the external radial coordinate , and the second , @xmath83 , is due to the geometry of global monopole for @xmath84 .
according to @xcite , the induced electrostatic self - energy associated with the electric charge is given by @xmath85 where @xmath86 is the renormalized green function defined as @xmath87 being @xmath88 the hadamard function .
the general expression for the hadamard function in a three - dimensional space is @xmath89 where @xmath90 , the van vleck - morette determinant , is given by @xmath91 and @xmath92 the geodesic interval between @xmath93 and @xmath94 . because the self - energy depends only on the radial coordinates ,
the respective induced self - force is given by @xmath95 in the rest of this section we shall calculate the induced electrostatic self - energy for the two cases specified : @xmath96 for the charge outside the monopole , and @xmath97 for the charge inside it .
as we have seen , for the region outside the monopole s core the green function to be considered in the calculation of the self - energy is given by ( [ ap ] ) . according to ( [ se ] )
, the induced electrostatic self - energy is given by taking the coincidence limit of the renormalized green function .
we observe that for points with @xmath84 , the core induced part of this function is finite in the coincidence limit , and that the divergent contribution appears only in the point - like part only .
so , in the renormalization procedure the only part to be considered is @xmath98 : @xmath99 \ .\end{aligned}\ ] ] in the coincidence limit , let us first take @xmath100 , so , for points along the radial distance , the singular part of the hadamard function reads , @xmath101 now , by using @xmath102 given in ( [ ap ] ) , we have : @xmath103 , \label{gmren1}\ ] ] where @xmath104 . in order to evaluate the limit on the right hand side of the above equation
, we note that @xmath105 so , on the basis of this relation , replacing in ( [ gmren1 ] ) @xmath106 by the first term in the brackets in ( [ rellim ] ) , we find @xmath107 in the above expression we have introduced the notation @xmath108 \ .
\label{falfa}\ ] ] the function @xmath109 is positive ( negative ) for @xmath110 ( @xmath111 ) and , hence , the corresponding self - force is repulsive ( attractive ) . developing a series expansion in the parameter @xmath112
, we can see that @xmath113 where @xmath114 are the bernoulli numbers .
the leading term in the expression on the right is @xmath115 . for large values
@xmath116 the main contribution into the series in ( [ falfa ] ) comes from large values @xmath117 . replacing the summation by the integration we can see that in the limit @xmath118 the function @xmath109 tends to the limiting value @xmath119 . for small values @xmath116 ,
@xmath120 , the main contribution comes from the term @xmath121 and one has @xmath122 .
now combining formulas ( [ se ] ) , ( [ ap ] ) and ( [ gmren2 ] ) , the induced electrostatic self - energy reads @xmath123 the second term on the right hand side of the above expression if positive for @xmath124 and negative for @xmath125 , consequently , according to the properties discussed for the function @xmath126 , the corresponding self - force given by @xmath127 is repulsive for the first case and attractive for the second one . for large distance from the monopole s core ,
the main contribution associated with the induced core part comes from the @xmath128 component ( the @xmath121 component vanishes ) , so we have @xmath129 \ .\end{aligned}\ ] ] there is a suppressed factor , @xmath130 , in the core - induced contribution .
the core - induced part is divergent at the boundary , @xmath131 . in order to verify this singular behavior
it is sufficient to analyse this quantity for large values of @xmath117 . the asymptotic expression for ( [ gamma ] )
can be obtained by using the relation between legendre function and hypergeometric one @xcite , @xmath132 for the case under consideration @xmath133 and @xmath134 .
for large value of @xmath117 the leading term of the corresponding hypergeometric function is the unity ( see @xcite ) , and we can write @xmath135 on the other hand @xmath136 .
so after some additional steps we find @xmath137 and we can see that the electrostatic self - energy is dominated by the core - induced part . the analysis of the behavior of the complete self - interaction can only be provided numerically .
so , in in figure [ fig1 ] we have plotted @xmath138 as function of the parameter @xmath116 and @xmath139 . in order to develop the analysis of the self - energy in the region inside the monopole , we should consider the corresponding green function given in ( [ bl ] ) .
as in the previous analysis , the core - induced contribution of this function is finite at the coincidence limit for @xmath140 . the only divergent contribution in this limit comes from @xmath141 .
the hadmard function , in the region inside the monopole , needed to renormalize this green function can be given in a simple way by using the coordinate system defined in ( [ gm1 ] ) .
taking the coincidence limit in the angular variables first , this function reads : @xmath142 because we have obtained the green function ( [ bl ] ) in the coordinate system defined in ( [ gm2 ] ) , it is more convenient to express the above hadamard function in this system .
so by using ( [ rela ] ) the singular behavior of ( [ hada ] ) is @xmath143 finally the renormalized green function is given by @xmath144 \
.\end{aligned}\ ] ] so developing some intermediate steps , we obtain the following expression for @xmath145 : we present the explicit calculation adopted to obtain ( [ gr0 ] ) . ]
@xmath146 where @xmath147 \end{aligned}\ ] ] and @xmath148 now we are in position to write the electrostatic self - energy for a charged particle inside the monopole s core : @xmath149 as in the case of exterior region , the self - energy is positive for @xmath150 and negative for @xmath151 .
the corresponding self - force is obtained by using ( [ sfo ] ) , and is repulsive with respect to the boundary of the monopole core in the first case and attractive in the second case . near the core s center
the most relevant contribution comes from the @xmath121 component of the core - induced contribution . in this region
the self - energy behaves as @xmath152 also the core - induced term diverges at the at the boundary . in order to see that let us analyse this terms for large value of @xmath117 . in this limit
we can verify that @xmath153 and @xmath154 substituting these expressions and developing some additional steps we find @xmath155 finally in figure [ fig2 ] we present the complete behavior of @xmath138 as function of the parameter @xmath116 and @xmath156 .
the induced electrostatic self - energy associated with a charged particle placed at rest in the global monopole spacetime , considering it as a point - like topological defect , is divergent at the center of the object @xcite . in principle
this undesirable result can be avoided by considering a more realistic model for its core . with this objective in mind , in @xcite
we investigated the influence of a non - trivial structure for the monopole s core on the self - energy considering the flower - pot model for the region inside .
for this case the corresponding induced self - energy is expressed in term of two distinct contributions : one coming from the geometry background of the corresponding spacetime , @xmath157 , and the other induced by the boundary , @xmath158 . in order to continue in the same line of investigation , in this paper we decided to revisit the induced electrostatic self - interaction problem , considering at this time a more sophisticated model for the inner structure of the monopole : the ballpoint - pen model .
analogously with happens in the flower - pot model , the renormalized electrostatic self - energy presents two distinct contributions : @xmath157 and @xmath158 .
as it was mentioned the introduction , there is no exact solution for the field equations in the region inside the global monopole .
the de sitter model proposed in @xcite describes the geometry for points very close to the monopole s center . as to the model adopted in this paper , like the de sitter one
, it presents a constant scalar curvature .
moreover , we can verify that it is an exact solution for the einstein equation for the matter field energy - momentum tensor given bellow : @xmath159 admitting this model , were able to obtain the renormalized self - interaction in a closed form , and analyzed it for different positions of the charge , and in various asymptotic regions of the parameters . specifically as to the self - energy at the monopole s center , we found a finite result . combining this result
with the previous one found in @xcite , we can see that attributing a non - trivial structure for the monopole , the behavior of the self - energy changes drastically , mainly near the boundary and the monopole s center .
before to finish this section , we would like to mention that , although the contributions due to the geometry background and boundary , are different for the flower - pot and ballpoint - pen models , we observe that they share some similarities : the respective induced self - forces are repulsive with respect to the boundary for @xmath150 and attractive for @xmath151 , the self - energy is finite at the monopole s center , and the core - induced contributions are dominant near the boundaries .
in fact they are logarithmically divergent , presenting exactly the same expressions for the outer regions .
db thanks capes for financial support .
erbm thanks cnpq . for partial financial support , fapesq - pb / cnpq .
( pronex ) and fapes - es / cnpq .
( pronex ) .
the expression for the green function @xmath160 is @xmath161 by using ( [ legend ] ) , we can express the term inside the summation by @xmath162 with @xmath163 being @xmath73 and @xmath72 given in ( [ x ] ) . in order to observe the divergent contribution of @xmath164 ,
let us analyse ( [ y ] ) in the limit of large value for @xmath117 .
the corresponding behavior is @xmath165 the contributions of the first and second terms above in the summation in ( [ g0 ] ) are , respectively , @xmath166 and @xmath167 . however the second contribution provide a vanishing result in the coincidence limit
. considering @xmath168 , @xmath169 and consequently @xmath170 finally the contribution due to this term in ( [ g0 ] ) is @xmath171 so the renormalized green function is given by @xmath172 \ , \end{aligned}\ ] ] which provide the results ( [ gr0 ] ) , ( [ gr1 ] ) and ( [ gr2 ] ) .
99 t. w. kibble , j. phys .
a. * 9 * , 1387 ( 1976 ) .
a. vilenkin and e. p. s. shellard , _ cosmic strings and other topological defects _
( cambridge university press , cambridge , england , 1994 ) .
d. d. sokolov and a. a. starobinsky , dokl .
nauk ussr .
* 234 * , 1043 ( 1977 ) [ sov .
- doklady * 22 * , 312 ( 1977 ) ] .
m. barriola and a. vilenkin , phys .
lett . * 63 * , 341 ( 1989 ) .
e. r. bezerra de mello , braz .
* 31 * , 211 ( 2001 ) .
f. d. mazzitelli and c. o. lousto , phys .
d * 43 * , 468 ( 1991 ) .
e. r. bezerra de mello , v. b. bezerra and n. r. khusnutdinov , phys .
d * 60 * , 063506 ( 1999 ) .
b. linet , phys .
d * 33 * , 1833 ( 1986 ) .
smith , in _ the formation and evolution of cosmic strings _ , proceedings of the cambridge workshop , cambridge , england , 1989 , edited by g.w .
gibbons , s.w . hawking , and t. vachaspati ( cambridge university press , cambridge , england , 1990 ) .
bezerra de mello and c. furtado , phys .
d * 56 * , 1345 ( 1997 )
. j. r. gott , astrophys j. * 288 * , 422 ( 1985 ) ; w. a. hiscock , phys . rev .
d * 31 * , 3288 ( 1985 ) .
b. allen and a.c .
ottewill , phys .
d * 42 * , 2669 ( 1990 ) .
n.r . khusnutdinov and v.b .
bezerra , phys .
d * 64 * , 083506 ( 2001 ) .
e. r. bezerra de mello and a. a. saharian , class .
quantum grav .
* 24 * , 2389 ( 2007 ) .
e. r. bezerra de mello and a. a. saharian , jhep * 610 * , 049 ( 2006 ) .
e. r. bezerra de mello and a. a. saharian , phys .
d * 75 * , 065019 ( 2007 ) .
d. harari and c. loust , phys .
d * 42 * , 2626 ( 1990 ) .
p. candelas and d. j. raine , phys .
d * 12 * , 965 ( 1975 ) .
n. d. birrell and p. c. w. davis , _ quantum fields in curved space _ ( cambridge university press , 1982 ) . i. s. gradshteyn and i. m. ryzhik , _ table of integrals , series and products _
( academic press , new york , 1980 ) . m. abramowitz and i.a .
stegun , _ handbook of mathematical functions _ ( national bureau of standards , washington dc , 1964 ) . | in this work we analise the electrostatic self - energy and self - force on a point - like electric charged particle induced by a global monopole spacetime considering a inner structure to it . in order to develop this analysis
we calculate the three - dimensional green function associated with this physical system .
we explicitly show that for points inside and outside the monopole s core the self - energy presents two distinct contributions .
the first is induced by the geometry associated with the spacetime under consideration , and the second one is a correction due to the non - vanishing inner structure attributed to it . considering specifically the ballpoint - pen model for the region inside ,
we were able to obtain exact expressions for the self - energies in the regions outside and inside the monopole s core .
+ pacs numbers : @xmath0 , @xmath1 |
the most promising anticipated source for the united states laser interferometer gravitational - wave observatory ( ligo ) @xcite , or its french / italian counterpart virgo @xcite , is the radiation emitted during the final moments of inspiral before the coalescence of a neutron star - neutron star ( ns - ns ) binary system @xcite .
the instruments operating in both the ligo and virgo facilities will evolve over time , eventually becoming sensitive to neutron star binary inspirals at distances approaching 2 gpc @xcite .
binary inspiral observations in the ligo or virgo detectors will be characterized by their signal strength and `` chirp mass '' ( a combination of the binary s component masses and cosmological redshift ) .
the distribution of observed inspirals with signal strength and chirp mass depends on cosmological parameters that describe our universe ( hubble constant , deceleration parameter , density parameter ) , the distribution of neutron star masses in binary systems , the overall density of coalescing binaries , and the properties of the detector . in this paper
i explore the binary inspiral event distribution ( with signal strength and chirp mass ) in the ligo and virgo detectors for different cosmological models .
in addition to their value as quantitative expectations of what ligo and virgo can expect to observe , these _ a priori _ distributions will play a pivotal role both in the construction of templates for detecting binary inspirals and in the interpretation of the observations .
the distributions presented here , evaluated for our preconceived notion of the binary inspiral rate and cosmological parameters , are the _ prior probabilities _ required to form the likelihood function from the observed detector output @xcite .
additionally , the observed inspirals will be a sample drawn from a particular cosmological model characterized by @xmath10 , @xmath11 , @xmath12 , neutron star mass distribution , and evolution characteristic of our own universe . by comparing the observed distributions to the ones described here
we can measure those properties of our own universe these cosmological tests are analogous to the number - count tests of classical cosmology , which , in their simplest form , involve observing the distribution of a source population as a function of apparent luminosity or redshift .
the first suggestion that binary inspiral number counts be used to measure interesting cosmological parameters was made by finn and chernoff @xcite ( although they did not use the language usually associated with this technique of classical astronomy ) . using monte carlo simulations they demonstrated that the distribution of inspiral events with signal strength and chirp mass could be used to measure the hubble constant . in this work
i provide a more general , explicit , and complete exposition of the properties of binary inspiral observation catalogs .
the cosmological implications of gravitational wave observations of binary inspiral were first recognized by schutz @xcite .
he pointed out that each inspiraling binary is a standard candle in the sense that , if observed in three independent interferometers , its luminosity distance can be determined from the observed detector response .
if an observed inspiral is associated with one of the several galaxy clusters that reside in its positional error box ( whose determination also requires three interferometers ) , and if the redshifts of those clusters are determined optically , then observation of several inspiraling binaries would lead to a statistical determination of the hubble constant that is independent of the cosmic distance ladder and the uncertainties that lurk therein .
markovi @xcite proposed a variation on the general theme introduced by schutz : he observed that known neutron star masses were all close to @xmath13 and that , in any event , there is a _ maximum _
neutron star mass .
the observed chirp mass is a function of the mass of the binary s two components and its redshift . assuming that the mass distribution in neutron star binaries does not evolve significantly over the range of binary inspiral observations , examination of the chirp mass distribution in binary systems at fixed luminosity distance
would reveal the corresponding redshift .
thus , gravitational radiation observations alone might suffice to determine the hubble constant .
unfortunately , detailed calculations show that , even for the most advanced ligo and virgo detectors that have been discussed , the fractional uncertainty in the measured luminosity distance will be of order unity for events seen more frequently than thrice per year ( _ i.e. , _ for events at distances greater than approximately @xmath14 ) @xcite , and the angular position error boxes for these events are likewise large ( on order @xmath15 @xcite ) . consequently ,
cosmological tests that rely on accurate and precise measurements of the distance and position of inspiraling binaries using ligo and virgo are not promising .
in contrast , the cosmological tests discussed here and in @xcite require only gravitational wave observation in a single interferometer . furthermore
, advanced ligo detectors can expect to observe approximately 50 ns - ns binary inspiral events per year , from distances up to @xmath4 , whose signal strength can be measured to better than 10% and whose chirp mass can be measured to better than @xmath16 @xcite .
the rate , depth , accuracy and precision of these single interferometer observations suggest that cosmological tests based on the distribution of observed events with signal strength and chirp mass have great promise . in this paper
i calculate the expected properties of a catalog of binary inspiral observations made by a single interferometric gravitational wave detector .
a catalog is presumed to contain a record of all binaries that coalesced during the observation period and whose inspiral signal - to - noise ratio @xmath0 was greater than the catalog limit @xmath2 .
the catalog properties depend on the coalescence rate density and the binary system component mass distribution , both of which may vary with redshift .
notation for these and a discussion of the observations that bear on them is the topic of section [ sec : dn / dtdmdv ] . in section [ sec : snr ]
i discuss the signal - to - noise ratio @xmath0 as a measure of signal strength .
the signal - to - noise ratio of a particular inspiraling binary depends on the binary s intrinsic properties , its distance from and orientation with respect to the detector , and the detector s intrinsic properties .
the way in which @xmath0 depends on the signal and detector properties suggests a useful measure of the detector bandwidth , which i discuss here as well .
finally , section [ sec : snr ] concludes by specializing the discussion to the specific properties of the proposed ligo and virgo gravitational radiation detectors .
the important properties of a binary inspiral observation catalog are the distribution of cataloged events with @xmath0 and the binary system `` chirp mass '' @xmath1 .
these distributions depend on the cosmological model , which includes the evolution with redshift of the neutron star mass distribution and the coalescing binary number density . in section [ sec : discussion ]
i give general expression for these distributions and discuss how they may be used together with actual observations to test cosmological models .
also in this section i give expressions for the _ catalog depth _ ( the maximum redshift of a binary system that can have @xmath0 greater than @xmath2 ) and the total rate that proposed interferometers can expect to observe inspiraling binaries with @xmath0 greater than @xmath2 .
all of these general expressions are evaluated explicitly for relevant matter - dominated friedmann - robertson - walker cosmological models and a simple model of the neutron star mass distribution .
finally , i summarize my conclusions in section [ sec : conclusions ] .
the signal - to - noise ratio @xmath0 of a binary inspiral in a ligo - like interferometer depends on the relationship between the binary and the detector ( _ i.e. , _ orientation , distance and redshift ) and also on certain intrinsic properties of the system ( _ i.e. , _ component masses and spins ) . of these intrinsic properties , the _ intrinsic chirp mass , _
@xmath17 where @xmath18 and @xmath19 are the binary s reduced and total mass , plays the most important role : all the other intrinsic properties offer only small corrections to @xmath0 .
gravitational - wave detectors like ligo or virgo do not measure @xmath20 ; instead , they measure @xmath21 where @xmath22 is the system s redshift with respect to the detector . to distinguish between @xmath1 , which involves the system s redshift , and @xmath23 , which depends only on the binary s intrinsic properties , i refer to the former as the _ observed _ chirp mass , or simply the chirp mass , and the latter as the _ intrinsic _ chirp mass . in order to describe the binaries included in a signal - to - noise limited catalog
we must first describe the coalescing binary distribution in space and in @xmath23 .
the notation i use to describe this distribution is defined in section [ sec : caln ] , while in section [ sec : nconstraints ] i discuss what is known about the distribution from present - day observations .
assume that coalescing binaries are distributed homogeneously and isotropically with the cosmological fluid and define the binary coalescence _ local specific rate density _
@xmath24 by @xmath25 where @xmath26 is a _ co - moving _ cosmological fluid volume element and @xmath27 is a proper time interval measured in the fluid rest frame .
the total co - moving rate density on the surface of homogeneity at redshift @xmath22 is thus @xmath28 define the ratio of the total co - moving rate density at a redshift @xmath22 [ @xmath29 to that at the present epoch [ @xmath30 by @xmath31 : @xmath32 the distribution of coalescing binaries with intrinsic chirp mass @xmath23 on the surface at redshift @xmath22 is @xmath33 where , by construction , @xmath34 the homogeneous and isotropic local specific rate density can thus be written @xmath35 since we have defined @xmath24 and @xmath36 in terms of the co - moving volume element and intrinsic chirp mass , in the absence of evolution @xmath31 is unity and @xmath37 is independent of @xmath22 .
additionally , for infinitesimal @xmath38 , @xmath39 can be interpreted as the probability that a randomly chosen binary on the surface at redshift @xmath22 has intrinsic chirp mass in the range @xmath23 to @xmath40 .
cosmological tests that depend on the observed _ distribution _ of inspirals with @xmath0 and/or @xmath1 do not depend on @xmath41 .
nevertheless , it is necessary to know @xmath41 in order to estimate how long it will take to accumulate a catalog of observations large enough that such tests will give meaningful results . the best current estimate of the ns - ns binary coalescence rate density at the current epoch is @xmath42 , where @xmath43 is the hubble constant measured in units of @xmath44 @xcite .
this estimates relies on the 3 observed binary pulsar systems that will coalesce in less than a hubble time ( psrs 1913 + 16 , 1534 + 12 , and 2127 + 11c @xcite ) .
since the number of observed systems is small the actual rate is quite uncertain : phinney @xcite has estimated that , while unlikely , a rate two orders of magnitude higher or lower could be reconciled with current observations . black hole - black hole ( bh - bh ) and bh - ns binaries are believed to form at rates comparable to ns - ns binaries ; however , the masses of these systems and the fraction that merge in less than a hubble time are entirely uncertain @xcite .
the intrinsic chirp mass distribution @xmath47 depends on the binary system component mass distribution on the slice of homogeneity at redshift @xmath22 .
denote a binary s component masses as @xmath48 and @xmath49 and write their joint probability density on a surface of redshift @xmath22 as @xmath50 .
the intrinsic chirp mass distribution @xmath47 on that slice is then @xmath51 the determination of @xmath47 thus reduces to finding @xmath50 .
both theoretical and observational evidence suggest that the neutron star mass distribution is narrow @xcite .
a simple model of the mass distribution has the component masses in a binary uncorrelated and uniformly distributed between upper and lower bounds @xmath52 and @xmath53 ; then @xmath54 and @xmath55 \over \left(m_u - m_l\right)^2 } .
\label{eqn : p(m0)}\end{aligned}\ ] ] where @xmath53 and @xmath52 may depend on @xmath22 .
the probability density @xmath56 is maximized when @xmath57 .
the limited observations of the neutron star mass distribution in binary pulsar systems provide independent 95% confidence intervals for @xmath53 and @xmath52 @xcite : @xmath58 the most likely values of @xmath53 and @xmath52 are @xmath59 over this narrow range @xmath56 is , to an excellent approximation , piecewise linear : @xmath60 where @xmath61 at present we can observe only nearby binary pulsar systems ; consequently , there are no observations that bear directly on the variation of @xmath46 or @xmath62 with @xmath22 .
theoretical studies suggest that the initial mass of neutron stars formed by stellar core collapse do not vary significantly with the progenitor mass or composition @xcite .
after formation the mass may evolve owing to accretion from a companion ; however , in any event it is not likely that either @xmath37 or @xmath31 vary with @xmath22 more rapidly than do galaxies . in section [ sec : discussion ] i provide general expressions for and detailed examples of the expected distribution of events in a catalog of binary inspiral events ; for the detailed examples i neglect evolution in @xmath46 and @xmath62 entirely . as shown in section [ sec : depth ] , advanced ligo and virgo detectors will observe ns - ns binaries from redshifts not expected to exceed @xmath63 and with the preponderance of events arising from @xmath64 ; consequently , neglect of evolution is not an unreasonable approximation .
the signal - to - noise ratio measures the signal amplitude in terms of a detector s noise properties . in subsection
[ sec : snr / subsec : intro ] i define the signal - to - noise ratio @xmath0 and discuss the subtle issue of how @xmath0 is estimated , but not determined , by observation .
the functional form of @xmath0 depends on the detector response to binary inspiral radiation , which i describe in subsection [ sec : response ] . in subsection
[ subsec : snr ] i give the binary inspiral signal - to - noise ratio in terms of the same parameters that characterize the detector response .
the form of @xmath0 suggests a natural definition of a detector s effective bandwidth for binary inspiral observations ; i discuss this bandwidth function in subsection [ subsec : snr ] as well .
the relative orientation of the detector and binary is described by a function of four angles .
while these angles can not be measured by observations in a single interferometer , important properties of the angular orientation function can be calculated independently of the particulars of the binary or the detector .
these properties play an important role in determining the binary inspiral catalog properties and interpreting individual observations .
i discuss the angular orientation function in subsection [ sec : ptheta ] .
finally , in subsection [ sec : ligo / virgo ] i specialize the discussion of the signal - to - noise to the proposed ligo and virgo gravitational radiation detectors .
let @xmath65 be the detector response to a gravitational radiation signal from any source .
if the detector noise is gaussian with one - sided power spectral density @xmath66 then the signal - to - noise ratio @xmath67 is _ defined _ to be @xmath68 where @xmath69 is the fourier transform of the detector response , @xmath70 and @xmath71 is its complex conjugate @xcite .
an observation of a gravitational wave signal in a noisy detector entails a measurement of the signal properties in the presence of a particular instance of the detector noise .
analysis of the detector output results in an _ estimate _ of @xmath0 as well as other parameters that describe @xmath65 .
_ throughout this work i use @xmath0 to mean the actual signal - to - noise ratio , as defined by equation [ eqn : snr ] , and not the estimate that arises in an observation . _ in section [ sec : discussion ] i find the distribution of sources with @xmath0 and @xmath1 in different cosmological models . by comparing these distributions with the observed one we can determine the model that best describes our own universe . in making that comparison it is critical to distinguish between @xmath0 and @xmath1 ( as _ defined _ by equations [ eqn : snr ] and [ eqn : m ] ) and the _ estimates _ of @xmath0 and @xmath1 that results from observations made in a detector . _
the estimate that results from an observation is a probability distribution @xmath72 for the parameters that describe the signal _ in the case of binary inspiral in a single interferometer , these include @xmath0 and the chirp mass @xmath1 ( see sec .
[ sec : response ] below ) .
the probability distribution associated with an observation is generally not reported ; instead , what is reported most often is a set of _ estimators _ that characterize the distribution and its moments . among the most popular
is the the maximum likelihood estimator , which is the set of parameter values that maximize @xmath72 .
estimators are summaries of the distribution @xmath72 and their utility depends on how accurately they are able to represent it . when @xmath72 is very sharply peaked ( _ i.e. , _ there is little uncertainty in the measurement ) then it may be approximated by a @xmath73-function and summarized accurately by the maximum likelihood estimator . when @xmath72 is sharply peaked but with a not insignificant width , then it may be approximated near its peak by a gaussian and accurately represented by the maximum likelihood estimators and their covariance .
when the distribution is not sharply peaked , however , then the more general uncertainties reflected in the detailed structure of @xmath72 play an essential role in the observation s interpretation and no summary of @xmath72 is especially useful . that a small set of estimators built from @xmath72 for a particular observation do not provide a useful summary does not mean that the observation itself is unreliable or uninformative ; rather , it means only that greater care must be taken in its interpretation .
finn ( * ? ? ?
( c ) ) gives an example of how the the probability distribution resulting from an observation should be used in the interpretation of neutron star mass observations in binary pulsar systems ; a further discussion of this point in the context of cosmological tests using binary inspiral observation catalogs is part of a work in preparation .
the detector response @xmath65 to the gravitational radiation from an inspiraling binary system depends on the distance and relative orientation of the source and the detector , as well as on certain intrinsic properties of the binary .
the relative orientation of the source and the detector is described by four angles : two ( @xmath74 and @xmath75 ) describe the direction to the binary relative to the detector , and two ( @xmath76 and @xmath77 ) describe the binary s orientation relative to the line - of - sight between it and the detector . to describe @xmath74 and @xmath75 ,
consider a single interferometric gravitational wave detector whose arms form a right angle .
let the arms themselves determine the @xmath78 and @xmath79 axes of a right - handed cartesian coordinate system with the @xmath80-axis pointing skyward . in this coordinate system the gravitational waves from an inspiraling binary
arrive from the direction @xmath81 , which can be defined in terms of the spherical coordinates @xmath74 and @xmath75 in the usual way : @xmath82 to describe @xmath76 and @xmath77 , let @xmath83 represent the total angular momentum of the binary system .
the detector response @xmath65 depends on the _ inclination angle _
@xmath76 between @xmath83 and @xmath81 , @xmath84 and the orientation @xmath77 of the angular momentum about @xmath81 , @xmath85 } .\ ] ] the conventions for the orientation angles @xmath74 , @xmath75 , @xmath76 , and @xmath77 described here are the same as those used in @xcite ( note , however , that the _ description _ of the angles in @xcite is incorrect ) . at newtonian ( _ i.e. , _ quadrupole formula ) order the only intrinsic property of the binary system that affects
the waveform is @xmath20 . at this order
the detector response ( a dimensionless strain ) to the binary inspiral signal is @xmath86 & \text{for $ t < t'$},\\ 0&\text{for $ t > t'$ } , \end{array } \right .
\label{eqn : response}\ ] ] where @xmath87 is a constant phase , @xmath88^{1/2},\\ f_{{}+ { } } & \equiv & { 1\over2}\left(1+\cos^2\theta\right)\cos2\phi\cos2\psi \nonumber\\ & & \qquad{}- \cos\theta\sin2\phi\sin2\psi , \\
f_\times & \equiv & { 1\over2}\left(1+\cos^2\theta\right)\cos2\phi\sin2\psi \nonumber\\ & & \qquad{}+ \cos\theta\sin2\phi\cos2\psi , \\ f & = & { 1\over\pi\cal m}\left[{5\over 256}{{\cal m}\over t - t}\right]^{3/8},\\ \phi & \equiv & 2\pi\int_t^t
f(t ) dt = -2\left({t - t\over5{\cal m}}\right)^{5/8 } , \end{aligned}\ ] ] @xmath89 is the source s luminosity distance , @xmath90 is the moment when the inspiral waveform terminates ( either because the binary components have coalesced or because the orbital evolution is no longer adiabatic ) , and @xmath91 would be the moment of coalescence if the two components of the binary system were treated as point masses in the quadrupole approximation ( the difference between @xmath92 and @xmath90 is small but not negligible ) .
note that dependence of the response on the angles @xmath74 , @xmath75 , @xmath76 , and @xmath77 is contained entirely in the orientation function @xmath93 , which does not depend on any other properties of the binary or the interferometric detector .
this important point will be discussed further in subsection [ sec : ptheta ] .
post - newtonian corrections do not contribute significantly to the signal - to - noise ratio for solar - mass binary inspiral in the ligo and virgo interferometers .
for symmetric binaries ( _ i.e. , _ those with equal or near equal mass components ) the first post - newtonian correction is proportional to @xmath94 , where @xmath19 is the binary s total mass and @xmath95 the component separation .
advanced ligo and virgo interferometers are expected to be most sensitive to binary inspiral radiation in the bandwidth 20@xmath96 ( see sec .
[ sec : bandwidth ] ) and , in this bandwidth , @xmath97 for solar mass binaries .
post - newtonian effects are more important for more massive binaries , since at fixed quadrupole radiation frequency @xmath94 is greater for greater @xmath19 : @xmath98 post - newtonian effects are more important for binaries with extreme mass ratios : in general , the first post - newtonian correction is proportional to @xmath99 , where @xmath100 is the component mass difference
. finally , spin - orbit coupling in systems whose components have large spin angular momenta can lead to orbital precession and a waveform modulation which can affect @xmath0 significantly @xcite . that the quadrupole waveform can be used to _ estimate _
@xmath0 for an inspiraling binary does not mean it can also be used to _ detect _ a binary inspiral signal in the detector output .
detection , or estimation of signal parameters from observation , involves the comparison of the detector output with a model of the detector response to the radiation .
`` small '' differences between the actual and modeled signal evolution , particularly its phase , can have a large effect on the measured value of the signal parameters , including the signal - to - noise ratio estimated from the observations @xcite .
the signal - to - noise ratio @xmath0 ( eq . [ eqn : snr ] ) corresponding to the detector response @xmath65 ( eq . [ eqn : response ] ) is @xcite @xmath101 where @xmath102 in equation [ eqn : rho ] @xmath93 , @xmath103 and @xmath89 depend on the particular binary system under consideration and @xmath104 is a characteristic distance that depends only on the detector s noise power spectral density @xmath66 .
the dimensionless function @xmath105 , which also depends only on the detector s noise spectrum , increases monotonically from 0 to 1 .
its argument @xmath106 is the redshifted instantaneous orbital frequency when the inspiral terminates ( at @xmath107 ) either because the compact components have coalesced or because the orbital evolution is no longer adiabatic and coalescence is imminent .
the characteristic distance @xmath104 and the function @xmath105 describe different aspects of an interferometric detectors sensitivity to binary inspiral gravitational radiation . for a fixed binary , the larger @xmath104 , the greater @xmath0 ; for fixed @xmath0 , the larger @xmath104 , the farther the detector can `` see . ''
decreasing the noise power in any band increases @xmath104 , but ( owing to the factor @xmath108 in the integrand of the expression for @xmath109 ) improvements at low frequency are more effective than those high frequency .
the function @xmath105 reflects the overlap of the signal power with the detector bandwidth .
the orbit of an inspiraling compact binary evolves adiabatically owing to gravitational radiation emission until an innermost circular orbit ( ico ) is reached at an instantaneous orbital frequency @xmath110 . at the ico the orbit evolves on a dynamic timescale and the components coalesce quickly .
thus , the adiabatic inspiral waveform ends when the orbital frequency reaches @xmath110 . ] . the quadrupole radiation frequency observed at the detector when the orbital frequency is @xmath110 is @xmath111 and the inspiral signal termination
is represented in the detector response by the cut - off at @xmath112 ( see eqs . [
eqn : response ] and [ eqn : fmax0 ] ) .
since the binary orbit evolves adiabatically from low frequencies to @xmath110 , the observed quadrupole radiation spectrum has significant power only up to frequency @xmath113 . if @xmath113 is very much greater than the frequency where the detector noise is minimized , than the signal power bandwidth overlaps completely with the detector bandwidth and @xmath114 .
on the other hand , if @xmath113 is much less than the frequency where the detector noise is minimized , than the overlap of the signal power and detector bandwidths is negligible , @xmath115 and @xmath116 . the orbital frequency at the transition of the binary orbit from adiabatic inspiral to plunge and
coalescence has been studied using high - order post - newtonian methods @xcite . for symmetric binaries ( _ i.e. , _ those with equal - mass components ) the instantaneous redshifted instantaneous orbital frequency of the ico is given by [ eqn : fmax ] @xmath117 where @xmath19 is the binary s total mass .
more generally , for binaries with compact components @xmath110 depends inversely on @xmath19 and directly on a function of the dimensionless ratio @xmath118 ; kidder @xcite shows @xmath119 for binary systems of arbitrary mass asymmetry . for a ns - ns binary
the component s proper separation at the ico is greater than a neutron star diameter ; so , coalescence occurs after the transition from inspiral to plunge @xcite .
tidal dissipation is important in determining the inspiral rate only in the last few orbits before contact @xcite ; consequently , equation [ eqn : fmax ] is applicable for ns - ns binaries and should be applicable for black hole binaries as well . the probability that a signal with detector response @xmath65 is present in the output @xmath120 of a noisy detector is related to [ eqn : prob ] @xmath121 where @xmath122 the actual signal , and consequently the detector response @xmath123 , may be difficult or impossible to evaluate exactly ( this is certainly the case for binary inspiral ) ; so , we would like to be able to use an approximate model @xmath124 in lieu of the actual detector response @xmath123 in equations [ eqn : prob ] . from equations [ eqn : prob ]
it is clear that any approximate model @xmath124 may be used as long as it matches @xmath125 closely wherever @xmath126 is relatively large .
this latter quantity is just the fraction of @xmath67 contributed by the signal power at frequency @xmath127 and is proportional to @xmath128 .
this suggests that we define the _ bandwidth function _
@xmath129 which is the fraction of the signal - to - noise ratio contributed by signal power at frequency @xmath127 in a logarithmic frequency bandwidth
. _ an approximate detector response model @xmath130 may be used instead of the actual response @xmath69 as long as @xmath130 accurately reflects @xmath69 wherever @xmath131 is large . _ knowing where and how @xmath130 needs to be accurate can simplify greatly the construction of approximate templates for data analysis ; for this purpose @xmath131 should prove a useful guide .
the signal - to - noise ratio of an inspiraling binary in a single interferometric detector depends on the relative orientation of the source and the detector through the angles @xmath74 , @xmath75 , @xmath76 , and @xmath77 .
the dependence of @xmath0 on these angles is confined to the angular orientation function @xmath93 , which is independent of all other properties of the binary system . from observations of binary inspiral in a single interferometer
we can measure @xmath0 and @xmath1 but not @xmath93 . even though @xmath93 can not be measured , because it depends only on @xmath74 , @xmath75 , @xmath76 and @xmath77 we actually know quite a bit about it : since , averaged over many binaries , @xmath132 , @xmath133 , @xmath134 , and @xmath135 are all uncorrelated and uniformly distributed over the range @xmath136 we know the probability that @xmath93 takes on any value .
this probability distribution is found numerically in @xcite ; here i note only that @xmath137 and that to an excellent approximation @xmath138 to determine the binary coalescence rate @xmath0 greater than @xmath2 we need to know @xmath24 and also how the signal - to - noise ratio of binaries with intrinsic chirp mass @xmath23 on a surface at redshift @xmath22 is distributed .
this latter distribution , @xmath139 , is related to @xmath140 : @xmath141 \left.{\partial\theta\over\partial\rho}\right|_{{\cal m}_0,z } \nonumber\\ & = & p_\theta\left [ { \rho\over8}{d_l(z)\over r_0\left(1+z\right)^{5/6 } } \left(1.2\,\text{m}_\odot\over{\cal m}_0\right)^{5/6 } \right ] \nonumber\\ & & \qquad{}\times { d_l\left(1.2\,\text{m}_\odot/{\cal m}_0\right)^{5/6}\over 8 r_0\left(1+z\right)^{5/6 } } , \label{eqn : p(rho)}\end{aligned}\ ] ] where @xmath142 represents the cosmological model ( @xmath10 , @xmath11 , @xmath12 , @xmath62 , @xmath46 ) and @xmath143 the detector ( _ i.e. , _ @xmath104 and @xmath105 ) .
ligo will consist of three interferometers : one in livingston , louisiana and two in hanford , washington .
the louisiana interferometer and one of the washington interferometers will have 4 km armlengths ; the second washington interferometer will have a 2 km armlength and share the same vacuum system as the 4 km interferometer .
for the proposed ligo interferometers as described in @xcite and modeled in @xcite , @xmath104 ranges from @xmath144 ( for `` initial '' interferometers ) to @xmath145 ( for `` advanced '' interferometers ) for initial ligo interferometers , but leaves it unchanged for advanced ones.[fnote : noise ] ] .
as the ligo detectors develop , incremental improvements will increase @xmath104 .
the orientation of the washington and louisiana interferometers were chosen to be as close to parallel as possible ; consequently , a simple approximation treats the ligo _ detector _
( all three interferometers operating in `` triple - coincidence '' ) as a single interferometer of arm length @xmath146^{1/2}\text{km } = 6\,\text{km.}\ ] ] for this `` super - interferometer '' @xmath104 ranges from @xmath147 to @xmath148 .
here and below , reference to the ligo _ detector _ refers to the three interferometers operating as a single detector , while reference to a ligo _
_ indicates one of the 4 km interferometers operating in isolation .
figure [ fig : zeta ] shows @xmath105 for early ligo and virgo interferometers and also more advanced ligo interferometers as have been discussed in the literature @xcite .
the solid curve shows @xmath105 representative of advanced interferometers , while the dashed and dotted curves show @xmath105 representative of initial ligo and virgo interferometers .
as the ligo detector evolves , @xmath104 will increase from approximately @xmath149 toward @xmath150 , and @xmath105 will evolve from the dashed curve toward the solid curve . setting aside the overall sensitivity gain with increasing @xmath104 ,
the evolution of @xmath105 signifies an increasing _ relative _ sensitivity to systems with small @xmath106 : _
i.e. , _ systems with larger total masses and/or redshifts .
virgo will consist of a single 3 km interferometer .
the target noise curve for the initial operation of the virgo interferometer is described in @xcite ; for this interferometer @xmath104 is @xmath144 . like the ligo interferometers , incremental improvements in virgo will increase @xmath104 and it is reasonable to assume that these improvements could raise @xmath104 for virgo to approximately @xmath151 .
the dotted curve in figure [ fig : zeta ] shows @xmath105 for the initial virgo interferometers .
note how , even though @xmath104 is the same for both the initial ligo and virgo interferometers , the relative sensitivity of these two detectors to signals at low and high frequency corresponding to more or less massive binaries is markedly different .
improvements in the virgo interferometer can be expected to evolve @xmath105 toward the solid curve as well .
note that the initial virgo interferometers are expected to be relatively more sensitive to massive ( _ i.e. , _ low @xmath106 ) binaries than are the initial ligo interferometers . for neutron star binaries @xmath152 ( eq .
[ eqn : fmax ] ) and @xmath114 in any of the proposed ligo and virgo interferometers ( see fig . [
fig : zeta ] ) .
figure [ fig : bwidth ] shows the bandwidth function for neutron star binary observations in early virgo ( dotted curve ) and ligo ( dashed curve ) interferometers and also more advanced ligo interferometers ( solid curve ) .
it is clear that the approximate detector response models for binary inspiral observations must evolve with the interferometers : in the early interferometers the detector response models will need to be quite accurate at high frequencies ( @xmath153 ) but not at low ones ; on the other hand , in more advanced interferometers these models will need to be accurate at low frequencies ( @xmath154 ) and , to obtain this low frequency accuracy , the high frequency performance may be sacrificed at no cost to the signal detectability .
consider a catalog of binary inspiral events with signal - to - noise ratio @xmath0 greater than a threshold @xmath2 .
suppose that for each event in the catalog @xmath0 and @xmath1 are known .
how do we make use of the catalog to test cosmological models against cataloged observations ?
the simplest test involves the distribution of events with @xmath0 .
as the catalog limit @xmath2 decreases , sources at increasingly larger distance become members .
the number of added sources depends on the increase of the spatial volume , the density of sources at that distance , and ( since these sources are _ events _ that occur at a given _ rate _ ) the cosmological redshift .
thus , adopting a cosmological model @xmath155 implies an expected distribution @xmath156 for the catalog events taken with detector @xmath157 .
denote the cataloged binary inspiral signal - to - noise ratio observations by @xmath158 .
suppose that , before studying these observations , we have reason or are otherwise prejudiced to believe the probability that @xmath142 is the correct cosmological model is @xmath159 .
using bayes law of conditional probabilities , the _ a posteriori _ probability that we assign to model @xmath142 after considering the observations is @xmath160 where @xmath161 is the signal - to - noise ratio of the @xmath162 catalog observation and @xmath163 represents the detector configuration ( @xmath104 , @xmath105 ) when the observation was made and the cosmological model .
this is not a problem in the interpretation of a catalog as long as the detector properties are properly associated with each cataloged observation . ] .
this test is exactly analogous to number - flux cosmological tests using distant galaxies .
more subtle tests involve other distributions of cataloged events .
for example , each source in the catalog is characterized by its observed chirp mass @xmath1 .
the observed chirp mass depends on both the intrinsic chirp mass @xmath23 and the redshift @xmath22 ( see eq .
[ eqn : m ] ) ; consequently , the distribution of events in the catalog with @xmath1 depends on the cosmological model .
denote the cataloged chirp mass observations by @xmath164 . adopting a cosmological model @xmath142 implies an expected distribution @xmath165 . as before , if we initially favor model @xmath155 with probability @xmath159 , then after studying the observations the probability we ascribe to model @xmath155 is @xmath166 where @xmath167 represents the @xmath162 cataloged chirp mass observation .
of the four parameters that describe , at quadrupole order , binary inspiral observed in a single interferometer , only @xmath0 and @xmath1 convey astrophysically interesting information .
the distributions @xmath168 and @xmath169 are each integrals over the distribution @xmath170 that completely characterizes the catalog : @xmath171 these integrals are _ summaries _ of the catalog contents : as such , they are less informative than @xmath172 .
the most sensitive test that we can make using the catalog involves not a summary , but the full information available : given the observations @xmath173 , the probability that cosmological model @xmath155 describes our universe is @xmath174 cosmological tests based on the summary distributions are still useful to make .
summary distribution often depend only weakly or not at all on some of the parameters of the model @xmath155 ; in that case , the effective dimensionality of @xmath155 is reduced in the summary test and we may be able to distinguish more closely among cosmological models described by the remaining parameters than with a test using the full distribution .
this is particularly true when the number of observations in the catalog is small . in this section
i give general expressions for the distributions @xmath170 and the summary distributions @xmath168 and @xmath175 that can be evaluated in any homogeneous and isotropic cosmological model ; in addition , i evaluate the summary distributions explicitly for matter - dominated friedmann - robertson - walker ( frw ) models .
the essential properties of frw cosmological models are summarized in section [ sec : frw ] .
in addition to the distributions @xmath176 , @xmath168 and @xmath169 are other catalog properties of intrinsic interest . in a given cosmological model
there is a distance beyond which no binary inspiral of fixed intrinsic chirp mass will have signal - to - noise ratio greater than the catalog limit @xmath2 , and i calculate this _ catalog depth _ @xmath177 in section [ sec : depth ] .
a catalog takes time to build , and , in a given period of time , the size of a binary inspiral catalog is limited by the rate @xmath178 at which binaries coalesce with @xmath0 greater than the catalog threshold @xmath2 . in section [ sec : rate ]
i calculate both the expected rate of binary inspiral observations and the distribution @xmath168 for advanced ligo interferometers and describe how these depend on the cosmological model @xmath155 .
finally , in section [ sec : spectrum ] i calculate the distribution @xmath179 and describe how it depends on @xmath155 .
specific examples of the catalog distributions , catalog depth , and rate of binary inspiral observations made in this section are in the context of matter - dominated friedmann - robertson - walker ( frw ) cosmological models .
the robertson - walker spacetime metric has the line element @xmath180,\ ] ] where @xmath181 is the usual ( dimensioned ) scale factor , @xmath95 is a dimensionless parameter related to the area of spheres of constant radius , and @xmath182 is @xmath183 , @xmath184 , or @xmath185 depending on whether the spatial geometry of the slices of homogeneity has positive , negative , or zero curvature ( _ i.e. , _ is closed , open , or flat ) . in matter dominated frw models ( which have vanishing cosmological constant ) the co - moving radial coordinate @xmath95 and the luminosity distance @xmath89 can be written in terms of the redshift explicitly : [ eqn : frw ] @xmath186\over a_0 h_0 q_0 ^ 2\left(1+z\right ) } , \label{eqn : r } \\ d_l & = & a_0\left(1+z\right ) r , \label{eqn : dl}\end{aligned}\ ] ] where @xmath187 , @xmath10 , and @xmath11 are the scale factor , expansion rate , and curvature parameter ( deceleration parameter ) at the present epoch ( * ? ? ?
* eq . 15.3.24 ) .
the signal - to - noise ratio of an inspiraling neutron star binary system with intrinsic chirp mass @xmath23 is @xmath188 since @xmath93 is between @xmath185 and @xmath189 , even an optimally oriented binary system has @xmath0 less than @xmath2 when @xmath22 is greater than @xmath177 , where @xmath177 satisfies @xmath190 evaluation of @xmath177 requires @xmath191 , which depends on the details of the detector s noise power spectral density ( through @xmath105 ) as well as the binary s component masses and redshift ( through @xmath106 ) . for advanced interferometers @xmath192 as long as @xmath193 ( see fig . [
fig : zeta ] ) while @xmath194 $ ] for symmetric binary inspiral ( see eq .
[ eqn : fmax ] ) .
consequently , we can approximate @xmath192 as long as @xmath195 . for small @xmath177 we can approximate @xmath196 ; then @xmath197 which is much less than @xmath198 .
thus , the approximation @xmath192 is a good one for binary systems with solar mass components , but not for binaries whose components are on order @xmath199 . for a specific example , focus attention first on solar mass component binaries .
then @xmath192 for any of the proposed ligo or virgo interferometers ( at the end of this subsection i return to consider briefly the case of more massive binary systems where this is not true ) . in an einstein - desitter ( @xmath200 )
cosmological model @xmath177 is then given explicitly by [ eqn : z0eds ] @xmath201 ^ 6 - 1,\ ] ] where @xmath202 @xmath43 is the hubble parameter ( @xmath10 in units of @xmath203 ) , and @xmath204^{1/3}.\ ] ] more generally ( @xmath205 but @xmath105 still unity ) @xmath177 satisfies [ eqn : z0 ] @xmath206 where @xmath207 is a root of @xmath208 the appropriate root of this equation can be found as a power series for small @xmath209 : @xmath210 this truncated expansion is accurate to better than 0.2% for @xmath211 and @xmath212 ; for @xmath213 and @xmath214 the accuracy is better than 0.01% .
an asymptotic expansion for @xmath177 valid for large @xmath209 ( but @xmath105 still @xmath215 ) is @xmath216 this asymptotic expansion is accurate to better than 1% for @xmath217 and @xmath218 .
figure [ fig : z0short ] shows @xmath177 as a function of @xmath43 for neutron star binaries with @xmath219 ( corresponding to @xmath220 neutron stars ) , @xmath221 , @xmath222 , and three different values of @xmath11 corresponding to an open ( @xmath223 , dotted curve ) , flat ( @xmath200 , solid curve ) and closed ( @xmath224 , dashed curve ) cosmological model . the redshift of the most distant such source is less than @xmath225 for @xmath226 and does not exceed @xmath227 for @xmath228 .
the sensitivity of @xmath177 to @xmath11 is modest but significant : inspection of equation [ eqn : z0 ] shows that @xmath229 + { \cal o}\left[\alpha_0 ^ 3,\left(2q_0 - 1\right)^2\right],\ ] ] where @xmath230 for advanced ligo interferometers . for open spatial geometries ( @xmath231 ) the most distant sources are at smaller redshifts than in closed spatial geometries ( where @xmath232 ) . in luminosity distance ,
the sample depth is @xmath233.\end{aligned}\ ] ] figure [ fig : dl ] shows @xmath234 for the same cases ( @xmath43 , @xmath11 ) as figure [ fig : z0short ] shows @xmath177 .
advanced ligo interferometers may observe neutron star binaries with @xmath0 greater than @xmath235 at luminosity distances of order 2 gpc . for more massive binaries @xmath191
is substantially less than unity at @xmath177 and the approximation @xmath192 is no longer valid .
figure [ fig : z0long ] shows @xmath177 ( given implicitly by eq . [ eqn : z0implicit ] ) for an advanced ligo detector ( @xmath236 ) and a range of binary systems in two distinct einstein - de sitter ( @xmath200 ) cosmological models .
one pair of curves show @xmath177 for symmetric binaries consisting of two ( _ e.g. _ ) black holes each of mass @xmath237 , while the second pair of curves shows @xmath177 for asymmetric binaries , consisting of ( _ e.g. , _ ) a black hole of mass @xmath237 and a neutron star of mass @xmath13 .
note how , in _ all _ cases , @xmath177 does not increase without bound as @xmath237 increases ; instead , it has a maximum value for some @xmath237 and decreases as @xmath237 increases further .
the finite bandwidth of the detector and its overlap with the signal power , as represented by @xmath105 , determines the maximum redshift at which a detector can observe binary inspiral .
if @xmath105 were constant then @xmath177 would increase monotonically with @xmath23 ( see eq .
[ eqn : z0implicit ] ) ; however , @xmath105 is not constant and as @xmath1 increases @xmath106 decreases both because @xmath110 is decreasing and also because @xmath177 is increasing ( @xmath238 ; see eq .
[ eqn : fmax ] ) . eventually @xmath105 begins to decrease with increasing @xmath23 or decreasing @xmath106 ( for advanced ligo interferometers , @xmath105 remains unity until @xmath239 ; see fig . [
fig : zeta ] ) . once @xmath105 begins to decrease so to will @xmath177 . for
sufficiently large @xmath23 even a local binary ( @xmath240 ) will coalesce before the gravitational radiation signal enters the detector bandwidth .
thus , decreasing the detector noise at low frequencies has a dual effect : by increasing @xmath104 it increases the overall detector sensitivity , and by decreasing the frequency below which @xmath241 it further boosts the sensitivity of the interferometer to more massive and more distant binary systems .
advanced ligo detectors are expected to be most sensitive to inspiraling binaries with two @xmath9 black hole components , which will be observable with @xmath0 greater than 8 at redshifts greater than 1.5 .
the distribution of catalog events with @xmath0 is given by [ eqn : p(rho|i ) ] @xmath242 where @xmath243 and @xmath244 is given in equation [ eqn : p(rho ) ] .
since @xmath245 depends on @xmath0 only through @xmath244 , the integral over @xmath0 in equation [ eqn : dn / dt ] can be evaluated explicitly .
first note that @xmath246 where @xmath247 the probability density @xmath244 is a conditional one : it depends on @xmath23 , @xmath22 , and the cosmological model .
in contrast , the distribution @xmath140 of @xmath93 is universal : it is independent of all the specific properties of the binary or the detector .
@xmath248 is universal in this same way and can be evaluated with @xmath140 . in terms of @xmath249 the total rate of inspirals with @xmath0 greater than @xmath2
is [ eqn : dn / dt ] @xmath250 where @xmath251 \label{eqn : dn / dtdm0}\end{aligned}\ ] ] and @xmath252 equation [ eqn : dn / dt ] for @xmath178 , which is also the total rate of binary inspiral with @xmath253 , and equations [ eqn : p(rho|i ) ] for @xmath168 are valid for any homogeneous and isotropic cosmological model .
turn now to the specific example of neutron star binaries in matter - dominated frw cosmological models , so that @xmath192 and @xmath95 and @xmath89 are given by equations [ eqn : frw ] .
assume also that evolution in the binary population is unimportant over redshifts @xmath254 ; then @xmath31 is unity and @xmath37 is independent of @xmath22 . finally , use the approximation given above for @xmath140 ( eq . [ eqn : ptheta ] ) to evaluate @xmath249 : @xmath255 the integral for @xmath256 ( eq . [ eqn : dn / dtdm0 ] ) can then be evaluated as a power series in small @xmath209 : [ eqn : approx - dn / dtdm0 ] @xmath257 where @xmath258\label{eqn:(dn / dt)0}\end{aligned}\ ] ] and @xmath209 is given by equation [ eqn : alpha0 ] .
similarly , @xmath259 is given by [ eqn : approx - dn / dtdrhodm0 ] @xmath260 where @xmath261\end{aligned}\ ] ] and @xmath262 equations [ eqn : approx - dn / dtdm0 ] and [ eqn : approx - dn / dtdrhodm0 ] for @xmath256 and @xmath263 can be integrated over @xmath23 to find @xmath178 and @xmath245 , and thus @xmath168 .
observational and theoretical evidence suggests that the ns - ns binary intrinsic chirp mass distribution is sharply peaked ( see sec .
[ sec : nconstraints ] ) . in the most extreme limit
that all binary systems have intrinsic chirp mass @xmath23 , @xmath62 is a @xmath73-function in @xmath23 and @xmath264 figure [ fig : dn / dt ] shows @xmath265 as a function of @xmath43 in an einstein - desitter cosmological model ( @xmath200 ) for a reasonable signal - to - noise ratio limit in an advanced ligo detector ( @xmath221 and @xmath236 ) , typical neutron star masses ( @xmath266 ) , and two different coalescence rate densities .
the solid curve shows @xmath265 for the `` best guess '' coalescence rate density , @xmath267 ( see sec .
[ sec : n0 ] ) , which is proportional to @xmath43 ; the dashed curve shows @xmath265 for a constant coalescence rate density @xmath268 .
the dashed curve shows how @xmath265 scales for constant @xmath41 in advanced interferometers with interesting @xmath43 .
present estimates of @xmath41 suggest that , if @xmath269 , advanced ligo detectors can expect to observe a little more than 1 neutron star binary inspiral event per week .
the distribution @xmath168 is also sensitive to @xmath43 .
let @xmath270 represent a flat and static cosmological model .
then @xmath271 the first order correction ( in @xmath272 ) owing to the expansion of the universe depends only on the rate of expansion ( @xmath10 ) ; corrections owing to the curvature of space ( @xmath273 ) enter only at second order in @xmath272 . since , for ns - ns binaries and reasonable limiting signal - to - noise @xmath274 is less than @xmath275 even for advanced interferometer designs , cosmological tests that focus only on the observed distribution of @xmath0 will be insensitive to @xmath11 even though the redshift to the most distant sources in the catalog is large ( this weak dependence on @xmath11 is characteristic of number - flux cosmological tests ( see , _ e.g. , _ ( * ? ? ? * pg .
798 ) ) . on the other hand , for advanced detectors and interesting cosmological models ,
the distribution @xmath168 is sensitive to @xmath43 at the 10% level , which makes possible a measurement of @xmath43 from the observations @xmath158 alone .
figure [ fig : p(rho|rho0 ) ] shows @xmath276 for two einstein - desitter cosmological models that differ only by @xmath43 ( in both cases @xmath266 ) .
the general trend is that with increasing @xmath43 inspiral events are shifted toward larger @xmath0 compared to the distribution in a static universe . in a less extreme limit assume
that neutron star masses are bounded above by @xmath277 and below by @xmath53 , that between these bounds their masses are uniformly distributed , and that the component masses of an inspiraling binary system are uncorrelated .
then @xmath278 is given approximately by equation [ eqn : approx - p(m0 ) ] , and @xmath279\\ { dn\over dt } & = & \left({128\over21}\pi r_0 ^ 3\dot{n}_0\right ) \left(8\over\rho_0\right)^3 \left(m_0\over1.2\text{m}_\odot\right)^{5/2 } \left[\xi_{5/2 } - { 25\over9}\bar{\alpha}_0\xi_{10/3 } + { 2\over45}\left(6q_0 + 85\right)\bar{\alpha}^2_0\xi_{25/6}\right.\nonumber\\ & & \qquad\left .
{ } - { 224\over33}q_0\bar{\alpha}^3_0\xi_{5 } + { 20\over8019}\left(1188q_0 ^ 2 + 9108q_0 - 4123\right)\bar{\alpha}^4_0\xi_{35/6 } + { \cal o}\left(\bar{\alpha}^5_0\right ) \right]\end{aligned}\ ] ] where @xmath280 \over \left(n+2\right)\left(n+1\right)\left(x_h - x_l\right ) \left(x_h-1\right)\left(1-x_l\right)}\end{aligned}\ ] ] and @xmath281 over the range of @xmath282 relevant for neutron star binary inspiral observations in ligo - like interferometric detectors , @xmath178 and @xmath245 are only weakly dependent on @xmath53 and @xmath277 for constant @xmath283 : taking @xmath284 we find @xmath285 .\end{aligned}\ ] ] observations suggest that @xmath286 is small @xcite , in which case the distribution of @xmath0 in ligo observations is insensitive to the finite neutron star mass range . in a homogeneous and isotropic cosmology
the rate at which binary inspiral signals corresponding to chirp mass @xmath1 and @xmath0 greater than @xmath2 are observed is @xmath288 .
\label{eqn : dn / dtdm}\end{aligned}\ ] ] the catalog s _ a priori _ chirp mass distribution is thus @xmath289 where @xmath178 is given by equation [ eqn : dn / dt ] and @xmath155 represents assumptions regarding the cosmology ( @xmath10 , @xmath11 , @xmath12 ) , evolution and the distribution of neutron star masses ( @xmath31 , @xmath37 ) , and other , unenumerated , model assumptions . as discussed in section [ sec : nconstraints ] , observational and theoretical evidence suggest that the ns - ns binary intrinsic chirp mass distribution is sharply peaked . neglecting evolution and taking the intrinsic chirp mass of all binary systems to be a constant @xmath23 , @xmath290 , \label{eqn : diff - explicit}\end{aligned}\ ] ] where @xmath291 figure [ fig : p(m|h , q0 ) ]
shows the distribution @xmath175 for catalogs with @xmath0 greater than 8 compiled by an advanced ligo detector ( @xmath236 ) in several matter dominated frw cosmological models .
the intrinsic chirp mass of all systems is assumed to be @xmath292 .
six different models are shown , exploring two different @xmath43 ( @xmath7 and @xmath293 ) and three different @xmath11 ( @xmath294 , @xmath225 and @xmath295 ) . the closely spaced curves with co - located extrema are of the same @xmath43 and differ only in @xmath11 .
note the strong dependence of @xmath175 on @xmath43 and the weaker , but still significant , dependence on @xmath11 : the dotted and solid curves correspond to flat cosmological models ( @xmath200 ) , the long - dash and dot - long - dash curves correspond to open cosmological models ( @xmath223 ) , and the short - dash and dot - short - dash curves correspond to the closed models ( @xmath224 ) .
in general the smaller @xmath11 , the more compressed the spectrum and the smaller the tail at large @xmath296 .
since the abscissa @xmath103 is related to redshift according to @xmath297 figure [ fig : p(m|h , q0 ) ] also shows the redshift of the preponderance of catalog events . for @xmath6 most events
are at a redshift of 9% , while for @xmath298 most events are at a redshift of 14% . more generally , neutron star masses are not all identical ; correspondingly , @xmath299 is not as simple as a @xmath73-function . modeling the neutron star mass distribution as
uniform between lower bound @xmath53 and upper bound @xmath277 leads to the intrinsic chirp mass distribution @xmath299 given by equation [ eqn : p(m0 ) ] ; the corresponding spectrum @xmath175 can be determined through equations [ eqn : dn / dtdm ] and [ eqn : p(m|i ) ] .
figure [ fig : p(m|ml , mh ) ] shows the spectrum @xmath179 for four different matter - dominated einstein - desitter cosmological models corresponding to two different @xmath43 ( @xmath7 and @xmath293 ) and two different neutron star mass distributions : one set of curves corresponds to the assumption that @xmath219 for all binary systems , while in the second set the binary component masses are assumed to be uniformly distributed between @xmath300 and @xmath301 . for all models shown @xmath236 and @xmath221 . as @xmath53 and @xmath277 approach @xmath283 , @xmath299 approaches @xmath302 .
the dependence of @xmath287 on @xmath11 , shown in figure [ fig : p(m|h , q0 ) ] , is similar but not identical to its dependence on @xmath303 ( for constant @xmath283 ) : variations in @xmath11 shift the spectrum s large @xmath1 tail , while increasing @xmath303 increases the spectrum s overall breadth .
observations of binary inspiral in a single interferometric gravitational wave detector can be cataloged according to signal strength ( as measured by signal - to - noise ratio @xmath0 ) and chirp mass @xmath1 .
the distribution of events in a catalog composed of observations with @xmath0 greater than a threshold @xmath2 depends on the hubble expansion , deceleration parameter , and cosmological constant , as well as the distribution of component masses in binary systems and evolutionary effects ( though for neutron star binary observations evolution is not expected to be important ) . in this paper
i find general expressions for the distribution with @xmath0 and @xmath1 of cataloged events , valid in any homogeneous and isotropic cosmological model ; i also evaluate those distributions explicitly for matter - dominated friedmann - robertson - walker models and simple models of the neutron star mass distribution .
these distribution have two immediate , practical uses in gravitational - wave data analysis for interferometric detectors : first , when evaluated for with the cosmological parameters reflecting our current best understanding of the universe , they are the prior probabilities which , together with the matched - filtered detector output , form the likelihood function and determine the posterior probability that an inspiral has been detected ; second , when compared with the observed distribution in @xmath0 and @xmath1 of many separate binary inspiral observations , they are used to infer new and more informed estimates for the cosmological parameters that describe the universe . the signal - to - noise ratio @xmath0 of a binary inspiral event is a detector - dependent measure of the signal strength from a radiation source that is estimated in the course of making an observation .
the normalization of @xmath0 depends on the detector s noise power spectral density @xmath66 . in interferometric gravitational wave detectors like ligo and
virgo the normalization involves a characteristic distance @xmath104 and a function of the detector bandwidth @xmath105 , both of which are depend on @xmath66 .
the characteristic distance @xmath104 gives an overall sense of the depth to which the detector can `` see '' binary systems whose radiation traverses the detector bandwidth , which is determined by @xmath105 .
advanced ligo interferometers are expected to be most sensitive to binary inspiral radiation in the bandwidth 30@xmath96 : over 90% of the signal - to - noise ratio is contributed by signal power in this narrow band . when searching the output of a detector for the signal from a source an accurate model of the detector response is needed .
the model need not be accurate over the entire frequency domain , but only over that part of the domain where the signal power overlaps with the detector bandwidth .
the dependence of @xmath0 on @xmath105 suggests the definition of a _ bandwidth function _ @xmath131 for binary inspiral observations that will be useful for determining over what range approximate templates describing the detector response need be accurate .
cosmological tests based on catalogs of binary inspiral observations with @xmath0 greater than a threshold @xmath2 depend on the distribution of cataloged events with @xmath0 and @xmath1
. tests can make use of all the information available in the catalog or properly constructed summaries of the cataloged events .
the sensitivity of the test depends on how the expected catalog distributions change with changing cosmological models . for advanced ligo detectors , the most distant neutron star binary inspiral events with signal - to - noise ratio greater than @xmath235 will arise from distances not exceeding approximately @xmath4 , corresponding to a redshift of @xmath5 ( 0.26 ) for @xmath6 ( @xmath7 ) .
the depth is only weakly dependent on the range of neutron star masses or the deceleration parameter .
as the binary system mass increases so does the distance it can be seen , up to a limit : in a matter dominated einstein - desitter cosmological model with @xmath6 ( @xmath7 ) the limit is approximately @xmath8 ( 1.7 ) for binaries consisting of approximately @xmath9 black holes .
the distribution of catalog events with @xmath0 depends primarily on @xmath43 and is only very weakly sensitive to either @xmath11 or the range of neutron star masses ; however , the chirp mass spectrum ( _ i.e. , _ the distribution of catalog events with @xmath1 ) is very sensitive to @xmath43 and reasonably sensitive to both @xmath11 and the neutron star mass range .
this suggests that the spectrum is an especially powerful tool for cosmological measurements .
the dependence of the spectrum on @xmath11 and the neutron star mass distribution is similar ( though not identical ) , suggesting that it may be difficult to determine these separately from observations .
i am currently investigating this point and will report on it at a later time .
l. s. finn , in _ proceedings of the cornelius lanczos international centenary conference _ , _ siam proceedings series _ , edited by j. d. brown , m. t. chu , d. c. ellison , and r. j. plemmons ( siam , philadelphia , 1994 ) , pp . | observations of binary inspiral in a single interferometric gravitational wave detector can be cataloged according to signal - to - noise ratio @xmath0 and chirp mass @xmath1 .
the distribution of events in a catalog composed of observations with @xmath0 greater than a threshold @xmath2 depends on the hubble expansion , deceleration parameter , and cosmological constant , as well as the distribution of component masses in binary systems and evolutionary effects . in this paper
i find general expressions , valid in any homogeneous and isotropic cosmological model , for the distribution with @xmath0 and @xmath1 of cataloged events ; i also evaluate these distributions explicitly for relevant matter - dominated friedmann - robertson - walker models and simple models of the neutron star mass distribution . in matter
dominated friedmann - robertson - walker cosmological models advanced ligo detectors will observe binary neutron star inspiral events with @xmath3 from distances not exceeding approximately @xmath4 , corresponding to redshifts of @xmath5 ( 0.26 ) for @xmath6 ( @xmath7 ) , at an estimated rate of 1 per week .
as the binary system mass increases so does the distance it can be seen , up to a limit : in a matter dominated einstein - desitter cosmological model with @xmath6 ( @xmath7 ) that limit is approximately @xmath8 ( 1.7 ) for binaries consisting of two @xmath9 black holes .
cosmological tests based on catalogs of the kind discussed here depend on the distribution of cataloged events with @xmath0 and @xmath1 .
the distributions found here will play a pivotal role in testing cosmological models against our own universe and in constructing templates for the detection of cosmological inspiraling binary neutron stars and black holes . |
mesoscopic analysis methods @xcite are among the most valuable tools available to applied network scientists and theorists alike .
their aim is to identify regularities in the structure of complex networks , thereby allowing for a better understanding of their function @xcite , their structure @xcite , their evolution @xcite , and of the dynamics they support @xcite .
community detection is perhaps the best - known method of all @xcite , but it is certainly not the only one of its kind @xcite .
it has been shown , for example , that the separation of nodes in a core and a periphery occurs in many empirical networks @xcite , and that this separation gives rise to more exotic mesoscopic patterns such as overlapping communities @xcite .
this is but an example there exist multitudes of decompositions in structures other than communities that explain the shape of networks both clearly and succinctly @xcite .
the stochastic block model ( sbm ) has proven to be versatile and principled in uncovering these patterns @xcite . according to this simple generative model ,
the nodes of a network are partitioned in blocks ( the _ planted partition _ ) , and an edge connects two nodes with a probability that depends on the partition .
the sbm can be used in any of two directions : either to generate random networks with a planted mesoscopic structure @xcite , or to infer the hidden mesoscopic organization of real complex networks , by fitting the model to network datasets @xcite perhaps its most useful application .
stochastic block models offer a number of advantages over other mesoscopic pattern detection methods @xcite .
one , there is no requirement that nodes in a block be densely connected , meaning that blocks are much more general objects than communities .
two , the sound statistical principles underlying the sbm naturally solve many hard problems that arise in network mesoscopic analysis ; this includes the notoriously challenging problem of determining the optimal number of communities in a network @xcite , or of selecting among the many possible descriptions of a network @xcite .
another consequence of the statistical formulation of the sbm is that one can rigorously investigate its limitations .
it is now known , for example , that the sbm admits a _ resolution limit _
@xcite akin to the limit that arises in modularity based detection method @xcite .
the limitations that have attracted the most attention , however , are the _ detectability limit _ and the closely related concept of _ consistency limit _ @xcite .
the sbm is said to be detectable for some parameters if an algorithm can construct a partition correlated with the planted partition , using no information other than the structure of a single infinitely large instance of the model .
it is said to be consistent if one can _ exactly _ recover the planted partition . therefore , consistency begets detectability , but not the other way around .
understanding when and why consistency ( or detectability ) can be expected is important , since one can not trust the partitions extracted by sbm if it operates in a regime where it is not consistent ( or detectable ) @xcite . due to rapid developments over the past few years
, the locations of the boundaries between the different levels of detectability are now known for multiple variants of the sbm , in the limit of infinite network sizes .
if the average degree scales at least logarithmically with the number of nodes , then the sbm is consistent @xcite , unless the constant multiplicative factor is too small , in which cas the sbm is then detectable , but not consistent .
if the average degree scales slower than logarithmically , then the sbm is at risk of entering an _ undetectable _ phase where no information on the planted partition can be recovered from the network structure @xcite .
this happens if the average degree is a sufficiently small constant independent of the number of nodes .
these asymptotic results are , without a doubt , extremely useful .
many efficient algorithms have been developed to extract information out of hardly consistent instances @xcite .
striking connections between the sbm and other stochastic processes have been established in the quest to bound the undetectable regime from below @xcite .
but real networks are not infinite objects .
thus , many of the findings of these asymptotic theories do not carry over to real networks
one can only _ assume _ that the asymptotic derivations are robust enough to inform us on finite cases .
the objective of our contribution is to investigate detectability in _
finite _ networks generated by the sbm .
the remainder of the paper is organized as follows .
we begin by formally introducing the sbm and the necessary background in sec .
[ section : stochastic_block_model ] .
we use this section to briefly review important notions , including inference ( sec .
[ subsection : stochastic_block_model - inference ] ) , as well as the consistency and detectability of the infinite sbm ( sec .
[ subsection : stochastic_block_model - related_work ] ) . in sec . [ section : finite_limit ] , we present a necessary condition for detectability , and show that it is always met , on average , by finite instances of the sbm .
we then establish the existence of a large equivalence class with respect to this notion of average detectability . in sec .
[ section : eta - detectability ] , we introduce the related concept of @xmath0detectability and investigate the complete detectability distribution , beyond its average . in sec . [
section : case_study ] , we apply the perfectly general framework of secs . [ section : finite_limit][section : eta - detectability ] to a constrained variant of the sbm : the general modular graph model of ref .
the results of this section hold for a broad range of models , since the general modular graphs encompass the symmetric sbm , the planted coloring model and many other models as special cases .
we gather concluding remarks and open problems in sec .
[ section : discussion ] .
two appendices follow .
the first investigates the interplay between noise and our notion of average detectability ( appendix [ appendix : noisy_sbm ] ) ; the second establishes a connection between our framework and random matrix theory ( appendix [ appendix : connection_with_rmt ] ) .
the stochastic block model is formally defined as follows : begin by partitioning a set of @xmath1 nodes in @xmath2 blocks of fixed sizes @xmath3 , with @xmath4 .
denote this partition by @xmath5 , where @xmath6 is the set of nodes in the @xmath7^th^ block . then , connect the nodes in block @xmath6 to the nodes in block @xmath8 with probability @xmath9 . in other words
, for each pair of nodes @xmath10 , set the element @xmath11 of the adjacency matrix @xmath12 to 1 with probability @xmath13 and to 0 otherwise , where @xmath14 is the block of @xmath15 .
note that for the sake of clarity , we will obtain all of our results for simple graphs , where edges are undirected and self - loops ( edges connecting a node to itself ) are forbidden .
this implies that @xmath16 and that @xmath17 .
we will think of this process as determining the outcome of a random variable , whose support is the set of all networks of @xmath1 nodes . due to the independence of edges , the probability ( likelihood ) of generating a particular network @xmath18
is simply given by the product of @xmath19 bernoulli random variables , i.e. , @xmath20^{1-a_{ij}}[p_{\sigma(v_i)\sigma(v_j)}]^{a_{ij}}\;,\ ] ] where @xmath21 is the @xmath22 matrix of connection probabilities of element @xmath9 ( sometimes called the affinity or density matrix ) , and @xmath23 is a shorthand for `` @xmath24 '' .
it is easy to check that the probability is properly normalized over the set of all networks of @xmath1 distinguishable nodes . a useful alternative to eq
. expresses the likelihood in terms of the number of edges between each pair of blocks @xmath25 rather than as a function of the adjacency matrix @xcite .
notice how the number of edges @xmath26 appearing between the sets of nodes @xmath6 and @xmath8 is at most equal to @xmath27 each of these @xmath28 edges exists with probability @xmath9 .
this implies that @xmath26 is determined by the sum of @xmath29 bernoulli trials of probability @xmath9 , i.e. , that @xmath26 is a binomial variable of parameter @xmath9 and maximum @xmath29 .
the probability of generating a particular instance @xmath18 can therefore be written equivalently as @xmath30 where @xmath31 and @xmath32 are jointly determined by the partition @xmath33 and the structure of @xmath18 , and @xmath34 denotes `` @xmath35 '' . having a distribution over all networks of @xmath1 nodes , one
can then compute average values over the ensemble .
for example , the average degree of node @xmath15 is given by @xmath36 where @xmath37 is the kronecker delta .
the expression correctly depends on the block @xmath38 of @xmath15 ; nodes in different blocks will in general have different average degree .
averaging over all nodes , one finds the average degree of the network @xmath39 this global quantity determines the density of the sbm when @xmath40 .
the sbm is said to be dense if @xmath41 , i.e. , if @xmath9 is a constant independent of @xmath1 .
it is said to be sparse if @xmath42 , i.e. , if @xmath43 goes to zero as @xmath44 . in the latter case , a node has a constant number of connections even in an infinitely large network a feature found in most large scale real networks @xcite . for finite instances
, it will often be more useful to consider the average density directly .
it is defined as the number of edges in @xmath18 , normalized by the number of possible edges , i.e. , @xmath45 where @xmath46 , and @xmath47 the dense versus sparse terminology is then clearer : the density of sparse networks goes to zero as @xmath48 , while dense networks have a nonvanishing density @xmath49 . depending on the elements of @xmath21 , the sbm can generate instances reminiscent of real networks with , e.g. , a community structure @xcite ( @xmath50 ) or a core - periphery organization @xcite ( @xmath51 and @xmath52 ) .
however , the sbm really shines when it is used to infer the organization in blocks of the nodes of real complex networks this was , after all , its original purpose @xcite . to have inferred the mesoscopic structure of a network ( with the sbm )
essentially means that one has found the partition @xmath53 and density matrix @xmath54 that best describes it . in principle , it is a straightforward task , since one merely needs to ( a ) assign a likelihood @xmath55 to each pair of partition and parameters [ see eqs .
] , then ( b ) search for the most likely pair ( @xmath53 , @xmath54 ) . since there are exponentially many possible partitions , this sort of enumerative approach is of little practical use .
fortunately , multiple approximate and efficient inference tools have been proposed to circumvent this fundamental problem .
they draw on ideas from various fields such as statistical physics @xcite , bayesian statistics @xcite , spectral theory @xcite and graph theory @xcite , to name a few , and they all produce accurate results in general
. one could expect perfect recovery of the parameters and partition from most of these sophisticated algorithms .
this is called the consistency property .
it turns out , however , that all known inference algorithms for the sbm , as diverse as they might be , fail on this account .
and their designs are not at fault , for there exists an explanation of this generalized failure .
consider the density matrix of elements @xmath56 @xmath57 , it is clear that the block partition is irrelevant the generated network can not and will not encode the planted partition .
thus , no algorithm will be abe to differentiate the planted partition from other partitions .
it is then natural to assume that inference will be hard or impossible if @xmath58 , where @xmath59 is a very small perturbation for networks of @xmath1 nodes ; there is little difference between the uniform case and this perturbed case .
in contrast , if the elements of @xmath21 are widely different from one another , e.g. , if @xmath60 and @xmath61 for @xmath62 , then easy recovery should be expected .
understanding where lies the transition between these qualitatively different regimes has been the subject of much recent research ( see ref .
@xcite for a thorough survey ) . as a result
, the regimes have been clearly separated as follows : ( i ) the undetectable regime , ( ii ) the detectable ( but not consistent ) regime and ( iii ) the consistent regime ( and detectable ) .
it has further been established that the scaling of @xmath63 with respect to @xmath1 determines which regime is reached , in the limit @xmath40 .
the sbm is said to be _ strongly consistent _ if its planted partition can be inferred perfectly , with a probability that goes to 1 as @xmath40 ( it is also said to be in the _ exact recovery _
another close but weaker definition of consistency asks that the probability of misclassifying a node goes to zero with @xmath40 ( the _ weakly consistent _ or _ almost exact recovery _
these regimes prevail when @xmath21 scales at least as fast as @xmath64 , where @xmath65 is a @xmath22 matrix of constants @xcite . predictably , most algorithms ( e.g. , those of refs .
@xcite ) work well in the exact recovery phase regime , since it is the easiest of all . in the _ detectable _ ( but not consistent ) regime , exact recovery is no longer possible ( the _ partial recovery _
the reason is simple : through random fluctuations , some nodes that belong to , say , block @xmath66 , end up connecting to other nodes as if they belonged to block @xmath67 .
they are thus systematically misclassified , no matter the choice of algorithms .
this occurs whenever @xmath68 , or @xmath69 , with @xmath70 a function of @xmath1 that scales slower than @xmath71 .
the discovery of the third regime the _
undetectable regime_arguably rekindled the study of the fundamental limits of the sbm . in this regime , which occurs when @xmath68 and @xmath65 is more or less uniform , it is impossible to detect a partition that is even correlated with the planted one .
that is , one can not classify nodes better than at random , and no information on the planted partition can be extracted .
thus , some parametrizations of the sbm are said to lie below the _ detectability limit_. this limit was first shown to exist using informal arguments from statistical physics and random matrix theory @xcite , and has since been rigorously investigated in refs .
@xcite , among others .
there exist many efficient algorithms that are reliable above the detectability limit , for almost all parameterizations of the sbm ; noteworthy examples include belief - propagation @xcite , and spectral algorithms based on the ordinary @xcite and weighted @xcite non - backtracking matrix , as well as matrices of self - avoiding walks @xcite .
when the number of blocks is too large , most of these algorithms are known to fail well above the information theoretic threshold , i.e. , the point where it can be proven that the partition is detectable .
it has been therefore conjectured in ref .
@xcite , that the undetectable regime is further separated in two phases : a truly undetectable regime , and a regime where detection is not achievable _
efficiently_. in the latter , it is thought that one _ can _ find a good partition , but only by enumerating all partitions a task of exponential complexity .
detectability and consistency are well separated phases of the infinite stochastic block model .
a minute perturbation to the parameters may potentially translate into widely different qualitative behaviors .
the picture changes completely when one turns to finite instances of the model .
random fluctuations are not smoothed out by limits , and transitions are much less abrupt .
we argue that , as a result , one has to account for the complete distribution of networks to properly quantify detectability , i.e. , define detectability for _ network instances _ rather than parameters .
this , in turn , commands a different approach that we now introduce .
consider a single network @xmath18 , generated by the sbm with some planted partition @xmath33 and matrix @xmath72 , where @xmath73 is a matrix of ones , @xmath7 a constant , and @xmath74 a matrix of ( small ) fluctuations .
suppose that the average density equals @xmath63 , and consider a second density matrix @xmath75 for which the block structure has no effect on the generative process .
if an observer with _ complete knowledge _ of the generative process and its parameters can not tell which density matrix , @xmath21 or @xmath75 , is the most likely to have generated @xmath18 , then it is clear that _ this particular instance _ does not encode the planted partition . as a result
, it will be impossible to detect a partition correlated with the planted partition .
this idea can be translated into a mathematical statement by way of a likelihood test . for a sbm of average density @xmath63 ,
call the ensemble of erds - rnyi graphs of density @xmath63 the ensemble of _ equivalent random networks_. much like the sbm ( see sec .
[ section : stochastic_block_model ] ) , its likelihood @xmath76 is given by the product of the density of @xmath19 independent and identically distributed bernoulli variables , i.e. , @xmath77 where @xmath78 is the total number of edges in @xmath18 .
the condition is then the following : given a network @xmath18 generated by the sbm of average density @xmath63 and density matrix @xmath21 , one can detect the planted partition @xmath33 if the sbm is more likely than its equivalent random ensemble of density @xmath63 , i.e. , @xmath79 a similar condition has been used in ref .
@xcite and @xcite to pinpoint the location of the detectability limit in infinite and sparse instances of the sbm .
but nothing forbids its application to the finite size problem ; one will see shortly that it serves us well in the context of finite size detectability
. the ( equivalent ) normalized log - likelihood ratio @xmath80 will be more practical for our purpose .
this simple transformation brings the line of separation between models from @xmath81 to @xmath82 , and prevents the resulting quantity from becoming too large .
more importantly , it changes products into sums , and allows for a simpler expression @xmath83 + \alpha_{rs}\!\log\!\!\left[\frac{1-p_{rs}}{1-\rho}\right]\!\!\bigg\}. \label{eq : normalized_log_likelihood}\ ] ] $ ] .
each color denotes a block @xcite .
( e - f ) empirical distribution of the normalized log - likelihood obtained from @xmath84 samples of @xmath85 . the bins in which the instances ( c - d ) fall are colored in red .
notice that a negative log - likelihood ratio is associated with some instances in ( f ) .
, title="fig : " ] $ ] .
each color denotes a block @xcite .
( e - f ) empirical distribution of the normalized log - likelihood obtained from @xmath84 samples of @xmath85 . the bins in which the instances ( c - d ) fall are colored in red .
notice that a negative log - likelihood ratio is associated with some instances in ( f ) .
, title="fig : " ] + $ ] .
each color denotes a block @xcite .
( e - f ) empirical distribution of the normalized log - likelihood obtained from @xmath84 samples of @xmath85 . the bins in which the instances ( c - d ) fall are colored in red .
notice that a negative log - likelihood ratio is associated with some instances in ( f ) .
, title="fig : " ] we will focus , for the remainder of this contribution , on the case where network instances @xmath18 of @xmath1 nodes are drawn from the sbm of parameters @xmath86 . from this point of view , @xmath85 can be seen as a random variable whose support is the networks of @xmath1 nodes with labeled nodes ( see fig . [
fig : log_likelihood_full_distribution ] ) . since @xmath21 , @xmath63 , @xmath87 and @xmath88 are all parameters
, @xmath85 is simply a weighted sum of binomial distributed random variables @xmath89 , with a constant offset .
its average will be a prediction of the detectability for the ensemble ( sec .
[ section : average_detectability ] ) , and the probability @xmath90 will give the fraction of instances that are undetectable for the selected parameters ( sec . [
section : eta - detectability ] ) .
because likelihood ratio tests can be understood as quantifying the amount of evidence for a hypothesis ( compared to a null hypothesis ) , there will be two interpretations of @xmath85 .
on the one hand , the condition @xmath91 will provide a lower bound on detectability ; if @xmath92 , then we can safely say that the instance @xmath18 is information theoretically undetectable .
however , @xmath93 does not necessarily mean that the instance is information theoretically detectable .
this is due to the fact that the condition @xmath91 is necessary but not sufficient , since we assume a complete knowledge of the generative process in calculating @xmath85 . on the other hand
, we will interpret @xmath85 operationally as a measure of the difficulty of the inference problem .
a large ratio of a hypothesis @xmath94 to its null model @xmath95 implies that the hypothesis is a much better explanation of the data than @xmath95 .
we may therefore say that @xmath85 is , in some sense , a measure of how easy the inference of @xmath33 is .
after all , since the @xmath96 and @xmath97 only differ by the mesoscopic organization of edges , @xmath98 should be taken to mean that the inferred partition is not an artifact of the procedure .
many empirical results will validate this interpretation ( sec .
[ section : case_study ] ) .
the average of a log - likelihood ratio is also known as the kullback - leibler ( kl ) divergence @xmath99 of two hypotheses @xcite , i.e. @xmath100 since the kl divergence is always greater or equal to zero , with equality if and only if @xmath101 , and since @xmath91 is only a necessary condition for detectability , the average @xmath102 will not be enough to conclude on detectability of the sbm , except for the case @xmath101 .
also goes to 0 at @xmath103 , and a more careful scaling analysis is necessary to conclude on the detectability of sparse instances . ] results pertaining to @xmath102 will therefore be best interpreted in terms of inference difficulty .
even if the average log - likelihood ratio is always positive ( assuming @xmath104 ) , it can be extremely close to zero for density matrix @xmath21 `` close '' to @xmath105 [ fig .
[ fig : log_likelihood_full_distribution ] ( f ) ] .
in fact , as we will see in sec . [ section : eta - detectability ] , @xmath106 implies that there are instances for which @xmath107 , i.e. , truly undetectable instances .
therefore , whenever the average is small , we may also interpret @xmath102 as a measure of detectability in the rigorous sense .
while eq . has a precise information theoretic interpretation , there exists an equivalent form , both more compact and easier to handle analytically .
it is given by @xmath108 where @xmath109 is the binary entropy of @xmath110 $ ] .
this expression can be derived in two different ways .
the simplest and most obvious of the two derivations is by direct computation of the average eq . over the joint distribution of the random variables
@xmath31 ; the second derivation follows from the extensivity of kl divergence for independent random variables.see supplemental material for more details on the mathematical development . ]
the bounds of @xmath102 will give us an intuition for what the easiest and hardest detectability problems might look like .
the kl divergence is never negative , and eq .
shows that the maximum of @xmath102 is @xmath111 ; the average of the normalized log - likelihood is thus confined to the interval @xmath112 an example of parameters that achieves the upper bound would be the sbm of density matrix @xmath113 , @xmath114 , with @xmath115 $ ] , i.e. , the `` ensemble '' of disconnected @xmath116cliques ( which contains a single instance ) . an example of parameters that achieves the lower bound would be @xmath101 , but also @xmath117 [ see eq . ] .
this goes to show that detectability is not guaranteed in sparse networks . equation also reveals connections between the qualitative regimes of the sbm , since it induces equivalence classes in the parameter space of the model , with respect to @xmath102 .
that is , it induces subset of parameters which all satisfy @xmath118 where @xmath119 is a constant that defines the equivalence class . in the next paragraphs
, we will characterize these equivalence classes in two complementary ways .
first , we will look for global transformations that preserve @xmath119 and map parameters @xmath120 to some other not necessarily close pair of parameters @xmath121 . provided that they satisfy a number of standard constraints , these transformations will be shown to correspond to the symmetry group of the set of _ hypersurfaces _ @xmath122 .
second , we will consider eq .
explicitly and use it to obtain an approximate hypersurface equation . this equation will be used in later sections to determine the location of the hypersurfaces that separate the parameter space of the sbm in different detectability phases .
we first look for the set of @xmath119preserving global transformations , i.e. , all transformations @xmath123 of the form @xmath124 valid at every point of the parameter space .
this is a broad definition and it must be restricted if we are to get anything useful out of it . intuitively , we do not want these transformations to change the space on which they operate , so it is natural to ask that they be space - preserving . under the ( reasonable ) constraint that these transformations are invertible as well , we can limit our search for @xmath119preserving transformations to the symmetry group of the parameter space
. we will be able to harness known results of geometry and algebra once the parameter space of the sbm is properly defined .
this space is in fact the cartesian product of two parameter spaces : the parameter space of @xmath87 and that of @xmath21 .
since there is @xmath125 free parameters in both @xmath87 and @xmath21 , the complete space is of dimension @xmath126 .
it is the product of the @xmath127dimensional hypercube in which every point corresponds to a choice of @xmath21- , and the @xmath128dimensional simplex in which every point corresponds to a choice of @xmath87 .
the latter is a simplex due to the normalization @xmath129 .
now , the symmetry group of the @xmath127dimensional hypercube and that of the @xmath128dimensional regular simplex are well - known @xcite : they are respectively the hyperoctahedral group @xmath130 and the symmetric group @xmath131 . their action on @xmath87 and @xmath21 can be described as @xmath132 where @xmath133 , and where both @xmath134 and @xmath135 are permutations of the indexes @xmath136 . while the symmetries of @xmath137 are automatically symmetries of the parameters , the converse is not true .
we therefore look for transformations @xmath138 that satisfy @xmath139 it turns out that this constraint is satisfied if and only if @xmath140 and @xmath141 @xmath57 , i.e. , for transformations of the form [ eq : symmetry ] @xmath142 with @xmath143 @xcite .
the permutation component of the symmetry is not to be confused with the symmetries generated by relabeling blocks : the latter only leads to @xmath144 different symmetries , whereas the former correctly generates @xmath145 symmetries , or a total of @xmath146 symmetries once they are compounded with @xmath147 .
the symmetries come about because the ordering of summation of the terms @xmath148 in eq . does not matter , and both @xmath149 and @xmath150 are preserved when @xmath151 . as an example of symmetry , let us focus on the special transformation @xmath152 @xmath57 with @xmath153 , i.e. , the only transformation that does not change the block structure of the model .
since networks generated by these parameters can be seen as complement of one another ( i.e. , an edge present in @xmath18 is absent from @xmath154 , and vice - versa ) , we may call this transformation the _ graph complement _ transformation .
the fact that it preserves detectability can be understood on a more intuitive level with the following argument .
suppose that we are given an unlabeled network @xmath18 generated by the sbm and that we are asked to confirm or infirm the hypothesis that it was , in fact , generated by the sbm .
even if nothing is known about the generative process , we can take the complement of the network a trivial ( and reversible ) transformation .
but this should not help our cause .
after all , this transformation can not enhance or worsen detectability since no information is added to or removed from @xmath18 in the process .
so we expect that @xmath119 be preserved , and it is . because all other symmetries affect the block structure through a change of @xmath87 ,
what the above result shows is that there is no other `` information - preserving '' transformation that can be applied to @xmath18 without a prior knowledge of its planted partition .
we now turn to the problem of finding compact and explicit formulas that describe the hypersurfaces of constant @xmath102 in the parameter space [ see eq . ] . in so doing
we will have to be mindful of the fact that the scale @xmath88 intervenes in the calculation , even though it is absent from our expression for @xmath102 .
this can be made explicit by rewriting eq .
as @xmath155 ; it is easy to see that any given hypersurface will be comparatively closer to the region @xmath156 in larger networks .
we focus on the universal behavior of the hypersurfaces and remove all references to the scale of the problem by defining @xmath157predictions for real systems can be recovered by reverting to the correct scale .
while eq .
is easily stated , it is not easily solved for , say , @xmath158 .
the average normalized log - likelihood ratio involves a sum of logarithmic terms ; the hypersurface equation is thus transcendental . to further complicate matters , there are @xmath159 degrees of freedom and the number of free parameters grow quadratically with @xmath2 . as a result
, little can be said of truly general instances of the sbm at least analytically .
all is not hopeless , however , because there are approximation methods that work well when the number of free parameters is not too large .
we sketch the idea here , and apply it to a simpler variant of the sbm in the case study of sec .
[ section : case_study ] .
expanding the binary entropy functions @xmath150 around @xmath56 @xmath160 drastically simplifies the hypersurface equation . leaving the term @xmath149 untouched
, we find from eq .
@xmath161\;.\ ] ] due to the normalization of @xmath162 , all terms in @xmath149 cancel out , and the definition @xmath163 allows us to eliminate the first order terms as well .
we are therefore left with @xmath164\;,\ ] ] where @xmath63 is fixed and @xmath120 take on values constrained by both eqs . and .
we then resort to a change of parameters and choose @xmath165 as one of the parameters . selecting the @xmath166 other parameters @xmath167 such that @xmath168 , we obtain the form @xmath169 hypersurfaces are therefore ellipsoids when @xmath170 . besides the simplicity of eq . , there are two additional arguments for dropping the higher order terms in eq . .
one , the series is invariant under the symmetry @xmath171 @xmath57 only if we limit ourselves to the second order expression : it easily verified that @xmath172(p_{rs}-\rho)^k\end{gathered}\ ] ] is off by a sign for odd powers of @xmath173 under the mapping @xmath174 , which also implies @xmath175 .
two , the true hypersurfaces enclose sets of parameters which are convex with respect to @xmath21 , and so does the hypersurface implicitly defined in eq . .
the convexity of the hypersurface follows from the fact that the sublevel set of a convex function encloses a convex set @xcite , and from the observation that @xmath102 is convex with respect to @xmath21 [ this is easy to show with eq . and the log - sum inequality @xcite ] .
the convexity of the approximate level set is trivial to the second order , since it is an ellipsoid [ eq . ] .
however , the approximate level set need not be convex when higher order terms are included . together , these two observations tell us that while not exact , eq . captures the important _ qualitative _ features of the problem , and that it is not necessarily true of approximate solutions with only a few extra terms .
in the previous section , we have computed the average @xmath102 and used it to obtain equivalence among the parameters , with respect to detectability .
we have also shown that @xmath176 for most parameters , i.e. , that we could not use the necessary condition @xmath91 to conclude on the _ undetectability _ of the finite sbm , on average . as we will now argue , this conclusion must be further refined ; the full distribution of @xmath85 leads to a more accurate picture of detectability . consider a parametrization @xmath177 of the sbm which yields @xmath178 . turning to the distribution of @xmath85 for this parametrization , one expect to find @xmath107 with non - zero probability ( unless the distribution of @xmath85 concentrates on @xmath179 ) .
therefore , @xmath102 could be indicative of detectability for some _ fraction _ of the networks generated by the sbm .
let us formalize this notion and introduce the concept of @xmath0detectability .
we will say that the ensemble of networks generated with the sbm of parameters @xmath86 is @xmath0detectable if @xmath180 that is , @xmath0 gives the fraction of networks in the ensemble which evades the necessary condition for undetectability . if @xmath181 , then detection is impossible , in the sense that most instances are best described by the null hypothesis @xmath96 . if @xmath182 , then most instances contain statistical evidence for @xmath33 ; detection can not be ruled out on the basis of the log - likelihood test .
we must compute the complete distribution or the cumulative distribution function of @xmath85 to determine @xmath0 .
an exact result is out of reach since the outcome of @xmath85 is determined by a weighted sum of independent binomial variables with non - identical distributions . in the following paragraphs , we give an approximate derivation based on the central limit theorem
it agrees extremely well with empirical results for all but the smallest networks .
equation gives the normalized log - likelihood ratio as a sum of independent binomial random variables ; it can be written as @xmath183 where the constants @xmath184 and @xmath185 are given by @xmath186\ ; , \label{eq : normalized_log_likelihood_x_rs}\\ c = \sum_{r\leq s}\alpha_{rs}\log\left[\frac{1-p_{rs}}{1-\rho}\right]\ ; , \label{eq : c_clt } \end{aligned}\ ] ] and where @xmath187 .
the central limit theorem ( clt ) predicts that the distribution of an appropriately rescaled and centered transformation of @xmath85 will converge to the normal distribution @xmath188 if the number of summed random variables @xmath125 goes to infinity . in the finite case
, @xmath127 obviously violates the conditions of the clt , but it nonetheless offers a good approximation of the distribution of @xmath85 ( see fig . [
fig : clt_quality ] ) . .
both histograms aggregate @xmath84 samples of @xmath85 .
the prediction of the clt is shown in red [ see eqs . ] .
we plot for comparison the gaussian kernel density estimate ( kde ) of @xmath189 ( dashed black line , hidden by the clt curve ) .
equation predicts @xmath190 and @xmath191 ; for the sample shown , numerical estimates yield @xmath192 and @xmath193 .
, title="fig : " ] + .
both histograms aggregate @xmath84 samples of @xmath85 .
the prediction of the clt is shown in red [ see eqs . ] .
we plot for comparison the gaussian kernel density estimate ( kde ) of @xmath189 ( dashed black line , hidden by the clt curve ) .
equation predicts @xmath190 and @xmath191 ; for the sample shown , numerical estimates yield @xmath192 and @xmath193 .
, title="fig : " ] to apply the clt , we first define the centered and normalized variable @xmath194 where @xmath195\notag\\ & = \sum_{r\leq s } \frac{\alpha_{rs}}{m^{\max}}\ p_{rs}(1-p_{rs})x_{rs}^2\end{aligned}\ ] ] is the sum of the variances of the @xmath127 scaled binomial variables @xmath196 , and where @xmath197 is the sum of their means [ we have used eq . in the last step ] .
the clt then tells us that @xmath198 , approximately .
recall that the cumulative distribution function of a normal random variable can be expressed in terms of the error function as @xmath199\;.\end{aligned}\ ] ] now , assuming that @xmath200 is indeed normally distributed we can use the fact that @xmath201 is equivalent to @xmath202 $ ] to compute @xmath0 . writing @xmath203 as @xmath102 [ see eq . ] , we find @xmath204\;,\ ] ] i.e. , a ( approximate ) equation in closed form for @xmath0 .
crucially , eq . predicts that @xmath0 can never be smaller than @xmath205 .
this comes about because ( i ) @xmath176 and ( ii ) @xmath131 is a sum of variances , i.e. , a positive quantity .
there are therefore two possible limits which will yield @xmath206 and @xmath207 : either @xmath156 or @xmath208 .
some care must be exerted in analyzing the case @xmath156 ; equations tell us that the distribution of @xmath85 is concentrated on @xmath209 when its average is exactly equal to 0 .
we conclude that @xmath207 is never reached but only approached asymptotically , for parameters that yield @xmath210 , with @xmath211 small but different from zero .
the consequence of @xmath212 is that at most half of the instances of the sbm can be declared undetectable on the account of the condition @xmath107 .
we can immediately reach a few conclusions on the interplay between the notions of average and @xmath0detectability .
first , the symmetries of @xmath102 , ( see sec .
[ subsubsection : symmetries_avg_detectability ] ) translates into symmetries for @xmath0 . to see this ,
first notice that @xmath131 is conserved under the mapping @xmath213
@xmath214 ^ 2 & \mapsto [ -x_{rs}(1-p_{rs},1-\rho)]^2\;,\\ p_{rs}(1-p_{rs } ) & \mapsto ( 1-p_{rs})p_{rs}\;.\end{aligned}\ ] ] and that a permutation of the indexes @xmath134 only changes the order of summation of the terms of @xmath131 .
second , hypersurfaces of constant average detectability need not be hypersurfaces of constant @xmath0detectability . to investigate this second important aspect of the connection between average detectability and @xmath0detectability ,
let us further approximate eq . .
the maclaurin series of the error function is , to the first order , @xmath215\right\}\notag\;,\\ & \approx \frac{1}{\sqrt{2\pi } } \frac{{\langle \mathcal{l } \rangle}}{s_{q^ * } } + \frac{1}{2}\;. \label{eq : eta_maclaurin}\end{aligned}\ ] ] this is a reasonably accurate calculation of @xmath0 when @xmath102 is small , i.e. , close to the _ average _ undetectable regime .
( recall that we do not allow diverging @xmath131 for the reasons stated in sec .
[ subsec : eta_clt ] ) .
it then becomes clear that on the hypersurfaces where @xmath216 is constant ( and close to 0 ) , @xmath217 is conserved rather than @xmath0 itself .
equation embodies a trade - off between accuracy ( @xmath0 ) and variance ( @xmath131 ) : in the regions of the hypersurface of constant @xmath102 where the variance is large , @xmath0 must be comparatively small , and vice - versa .
now , turning to the complementary case where @xmath102and consequently @xmath0is close to its maximum , we obtain a simple criterion for 1detectability based the asymptotic behavior of @xmath218 .
it is reasonable to define a ( small ) threshold @xmath138 beyond which @xmath219 for all practical purposes .
the error function goes asymptotically to @xmath220 with large values of its argument , but reaches its maximum of @xmath221 very quickly , so quickly , in fact , that @xmath222 is numerically equal to 1 to the 10^th^ decimal place .
asking that the argument of @xmath218 in eq .
be greater than this practical threshold , we obtain the inequality @xmath223 for @xmath220detectability .
whenever the inequality holds , the associated ensemble is 1detectable with a tolerance threshold @xmath138 , i.e. , we can say that for all practical purposes , there are no instances of the sbm which are necessarily is not sufficient for detectability , some instances could still be undetectable . ] undetectable .
the stochastic block model encompasses quite a few well - known models as special cases ; noteworthy examples include the _ planted partition model _
@xcite , the closely related _ symmetric sbm _ ( ssbm )
@xcite , the _ core - periphery model _ @xcite , and many more .
these simplified models are important for two reasons .
one , they are good abstractions of structural patterns found in real networks , and a complete understanding of their behavior with respect to detectability is therefore crucial .
two , they are simple enough to lend themselves to a thorough analysis ; this contrasts with the general case , where simple analytical expressions are hard to come by . in the paragraphs that follow ,
we investigate the _ general modular graph model _ ( gmgm ) @xcite , a mathematically simple , yet phenomenologically rich simplified model .
thanks to its simpler parametrization , we obtain easily interpretable versions of the expressions derived in secs .
[ section : finite_limit][section : eta - detectability ] .
the gmgm can be seen as constrained version of the sbm , in which _ pairs _ of blocks assume one of two roles : inner or outer pairs .
if a pair of blocks @xmath25 is of the `` inner type '' , then one sets @xmath224 .
if a pair of blocks @xmath25 is of the `` outer type '' , then one sets @xmath225 .
the resulting density matrices can therefore be expressed as @xmath226 where @xmath227 is a @xmath22 indicator matrix [ @xmath228 if @xmath229 is an inner pair ] , and where @xmath230 is a length @xmath2 vector of ones .
a non - trivial example of a density matrix of this form is shown in fig .
[ fig : general_modular_graph ] ( a ) .
the figure is meant to illustrate just how diverse the networks generated by the gmgm may be , but it is also important to note that the results of this section apply to _ any _ ensemble whose density matrix can be written as in eq . .
this includes , for example , the @xmath2block ssbm , obtained by setting @xmath231 and @xmath232 . whilst the parametrization in terms of @xmath233 and @xmath234 is simple
, we will prefer an arguably more convoluted parameterization which is also more revealing of the natural symmetries of the gmgm ( in line with the transformation proposed in sec .
[ subsubsection : hypersurfaces ] ) .
the first natural parameter is the average density , which can be computed from eqs . and and which equals [ eq : natural_gmg_params ] @xmath235\;,\notag\\ & = \beta p_{{\mathrm{in } } } + ( 1-\beta)p_{{\mathrm{out}}}\;,\end{aligned}\ ] ] where @xmath236 is the fraction of _ potential _ edges that falls between block pairs of the inner type .
the second natural parameter is simply the difference @xmath237 the absolute value of @xmath238 quantifies the distance between the parameters of the gmgm and that of the equivalent random ensemble ; its sign tells us which type of pairs is more densely connected . in this natural parametrization
the density matrix takes on the form @xmath239 , i.e. , a uniform matrix of @xmath63 with perturbation proportional to @xmath240 for the inner pairs .
it might appear that we have increased the complexity of the model description , since the additional parameter @xmath241 now appears in the definition of the density matrix .
it is , however , not the case , because we could consider the combined parameter @xmath242 .
therefore , eqs .
, together with @xmath227 and @xmath243 , suffice to unambiguously parametrize the model .
the average normalized log - likelihood ratio @xmath102 is tremendously simplified in the natural parametrization of the gmgm ; it is straightforward to show that the ratio takes on the compact ( and symmetric ) form @xmath244\big\ } \\- ( 1-\beta)\big\{h(\rho ) - h\bigl[\rho-\beta\delta\bigr]\big\ } \;,\end{gathered}\ ] ] by using @xmath245 together with the inverse of eqs .
: [ eq : natural_gmg_params_inverse ] @xmath246 in fig .
[ fig : general_modular_graph ] ( b ) , we plot @xmath247 in the @xmath248 space hereafter the density space for the indicator matrix @xmath227 shown in fig .
[ fig : general_modular_graph ] ( a ) ( and unequal block sizes , see caption ) .
unsurprisingly , @xmath102 is largest when the block types are clearly separated from one another , i.e. , when @xmath249 is the largest .
notice , however , how large separations are _ not _
achievable for dense or sparse networks .
this is due to the fact that not all @xmath248 pairs map to probabilities @xmath250 in @xmath251 $ ] .
the region of the density space which _ does _ yield probabilities is the interior of the quadrilateral whose vertices are , in @xmath248 coordinates : @xmath252 .
changing the value of @xmath241 skews this accessible region and , presumably , the functions that are defined on it , such as @xmath247 .
we also show on fig .
[ fig : general_modular_graph ] ( b ) two members of the level set defined by @xmath253 .
as mentioned previously , the exact functional form of this family of hypersurfaces ( here simply curves ) seems elusive , but an approximate solution is available . using the method highlighted in sec .
[ section : average_detectability ] , we find , to the second order , @xmath254 ^ 2 + ( 1-\beta)(\beta\delta)^2\;. \label{eq : hypersurface_gmg}\end{aligned}\ ] ] equation fixes the relative value of all parameters on the line where @xmath255 . solving for @xmath238 , we find @xmath256 also shown on fig . [
fig : general_modular_graph ] ( b ) for comparison .
figure [ fig : general_modular_graph ] highlights the accuracy of our approximation when @xmath119 is small .
but it also highlights its inaccuracy when @xmath119 is large ; @xmath257 forces @xmath258 to pass through a region where @xmath259 , i.e. , a region where the omitted terms on the rhs of eq .
contribute heavily .
fortunately , this is not so problematic , since most detectability related phenomena phase transitions , undetectable instances , etc.arise near @xmath260 , i.e. , where the approximation works .
while @xmath247 takes on a particularly compact form once we substitute @xmath158 by the natural parameters of the gmgm , the same can not be said of @xmath261 .
some analytical progress can be made by , e.g. , noticing that only two types of terms are involved in the calculation of @xmath131 , but , ultimately , the resulting expression is no more useful than the simple eq .
and eq . .
we will therefore omit the calculation of @xmath0 . in fig .
[ fig : general_modular_graph ] ( c ) we plot @xmath261 in the density space [ using eq . ] .
we also display the numerical solutions of @xmath262 for two values of @xmath263 .
the figure highlights just how quickly @xmath0 goes to 1 as a function of @xmath238 , even for the fairly small system sizes considered : we find that @xmath264 for _ any _ value of @xmath63 , as soon as @xmath265 .
the condition is therefore a _ weak _ one .
it allows us to determine that some parameters are overwhelmingly undetectable , but only when @xmath238 is very close to 0 .
figure [ fig : general_modular_graph ] also shows how increases in variance translate into decreases in accuracy [ see eq . ] : following a line of constant ( and relatively small ) @xmath238 , one can see that @xmath0 is minimized close to @xmath266 , i.e. , near the maximum of variance .
this is characteristic of many parameterizations of the sbm and gmgm ; it turns out that , for fixed @xmath1 , the hardest detection problems are not confined to vanishing densities .
in fact , values of @xmath63 closer to @xmath205 are associated with a comparatively larger interval of @xmath238 for which detection is hard . in secs .
[ section : average_detectability][section : eta - detectability ] , we have proved that there are @xmath146 transformations that preserve @xmath247 and @xmath261 .
we could therefore go about computing the symmetries of the gmgm by listing all of these transformations in terms of @xmath267 .
but since there are only 3 free parameters in the gmgm , we can also choose an alternative route and directly solve @xmath268 by , e.g. , obtaining a linear system from the taylor series of @xmath247 .
this simpler approach yields the following set of @xmath119preserving transformations for the model : [ eq : gmg_syms ] @xmath269 it is straightforward to check that these transformations form a group , whose product is the composition of two transformations .
a cayley table reveals that the group is isomorphic to the klein four - group @xmath270 . at a glance
, one immediately notices a large gap between the predicted number of symmetries ( @xmath146 ) and the actual number of symmetries ( @xmath271 , independent of @xmath2 ) .
it is explained by the fact that every symmetry of the general sbm maps onto one of the four transformations listed in eq .
pairs of blocks in the eyes of our formalism : a single inner pair and a single outer pair , with , respectively , a fraction @xmath241 and @xmath272 of all possible edges . ]
a sizable fraction of the symmetries reduce to eq , since permutations @xmath134 can not modify the natural parameters of the gmgm : the type of block pair @xmath25characterized by @xmath9is permuted alongside its share of potential edges @xmath273 .
another important fraction of the symmetries is accounted for by the `` graph complement transformation '' : any transformation @xmath274 plus a permutation reduces to eq . .
this leaves two symmetries , which happen to be artifacts of our choice of nomenclature . to see this ,
_ rename _ pair types , i.e. , call inner pairs `` outer pairs '' and vice - versa .
neither the density @xmath63 nor @xmath249 will change . but both the sign of @xmath238 and the value of @xmath241 will be altered . with this in mind , it becomes clear that eq
. corresponds to the permutation symmetry , and that eq
. corresponds to the graph complement symmetry , both up to a renaming of the types . and @xmath275 $ ] .
all empirical results are averaged over @xmath276 independent instances of the sbm .
we infer the partition of each of these instances with an optimal metropolis - hasting algorithm , seeded with the planted partition and the correct ensemble parameters @xcite .
( a ) average rnmi of the planted and the inferred partition of the ssbm ( of @xmath277 nodes ) in the density space .
solid red lines mark the border of the 1detectability region , with tolerance threshold @xmath278 , see eq . .
dotted black lines show the two solutions of @xmath279 , see eq . .
( b)(c ) phase transition at constant @xmath280 for networks of @xmath277 nodes ( b ) and @xmath281 nodes ( c ) .
circles indicate the fraction of instances for which a correlated partition could be identified , while diamonds show the average of the rnmi ( lines are added to guide the eye ) .
blue solid curves show @xmath282 , see eq . .
the shaded region lies below the kesten - stigum bound @xmath283 ( here with @xmath284 ) .
the dotted lines show the two solutions of @xmath285 .
, title="fig : " ] and @xmath275 $ ] .
all empirical results are averaged over @xmath276 independent instances of the sbm .
we infer the partition of each of these instances with an optimal metropolis - hasting algorithm , seeded with the planted partition and the correct ensemble parameters @xcite .
( a ) average rnmi of the planted and the inferred partition of the ssbm ( of @xmath277 nodes ) in the density space .
solid red lines mark the border of the 1detectability region , with tolerance threshold @xmath278 , see eq . .
dotted black lines show the two solutions of @xmath279 , see eq . .
( b)(c ) phase transition at constant @xmath280 for networks of @xmath277 nodes ( b ) and @xmath281 nodes ( c ) .
circles indicate the fraction of instances for which a correlated partition could be identified , while diamonds show the average of the rnmi ( lines are added to guide the eye ) .
blue solid curves show @xmath282 , see eq . .
the shaded region lies below the kesten - stigum bound @xmath283 ( here with @xmath284 ) .
the dotted lines show the two solutions of @xmath285 .
, title="fig : " ] and @xmath275 $ ] .
all empirical results are averaged over @xmath276 independent instances of the sbm .
we infer the partition of each of these instances with an optimal metropolis - hasting algorithm , seeded with the planted partition and the correct ensemble parameters @xcite .
( a ) average rnmi of the planted and the inferred partition of the ssbm ( of @xmath277 nodes ) in the density space .
solid red lines mark the border of the 1detectability region , with tolerance threshold @xmath278 , see eq . .
dotted black lines show the two solutions of @xmath279 , see eq . .
( b)(c ) phase transition at constant @xmath280 for networks of @xmath277 nodes ( b ) and @xmath281 nodes ( c ) .
circles indicate the fraction of instances for which a correlated partition could be identified , while diamonds show the average of the rnmi ( lines are added to guide the eye ) .
blue solid curves show @xmath282 , see eq . .
the shaded region lies below the kesten - stigum bound @xmath283 ( here with @xmath284 ) .
the dotted lines show the two solutions of @xmath285 .
, title="fig : " ] it will be instructive to put our framework to the test and compare its predictions with numerical experiments that involve inference , i.e. , the detection of the planted partition of actual instances of the gmgm . the procedure will be the following : ( i ) generate an instance of the model , ( ii ) run an inference algorithm on the instance , and ( iii ) compute the correlation of the inferred and planted partition .
the average detectability @xmath102 should predict the point where the average correlation becomes significant ( see below for a precise definition ) , and @xmath0detectability should give an upper bound on the fraction of correlated instances . before we proceed
, let us emphasize that the outcome of these experiments is influenced by a number of factors .
since it is conjectured that there exists a gap between information - theoretically feasible inference and efficiently achievable inference @xcite , we will have to be careful in choosing an inference algorithm otherwise we would run at risk of confounding the sources of error .
even for the size considered , enumeration is impossible ; we must therefore resort to an approximate algorithm .
we use an efficient algorithm based on the metropolis - hasting algorithm of ref . @xcite .
unlike belief propagation @xcite , it works well on dense networks with many short loops . in the spirit of refs .
@xcite we initialize the algorithm with the planted partition itself , to achieve our information theoretic threshold , even if efficient inference is not possible @xcite .
we must also define precisely what is meant by correlated inference if we are to quantify our experiment .
crucially , we have to account for finite size effects that could introduce spurious correlations .
the so - called renormalized normalized mutual information ( rnmi ) of ref .
@xcite appears a good choice .
much like the well - known nmi @xcite , the rnmi is bounded to the [ 0 , 1 ] interval , and @xmath286 means that the planted partition @xmath287 and the inferred partition @xmath288 are identical . unlike the nmi , @xmath289 signals the absence of correlation between the two partitions , even in finite networks . in fig .
[ fig : general_modular_graph_transition ] ( a ) , we plot @xmath290 in the density space of the gmgm .
we use the parameters @xmath291 , and @xmath115 $ ] ( i.e. , the ssbm ) , since the resulting ensemble is conjectured to be the hardest of all , with respect to detectability @xcite .
two important parallels can be drawn between the results shown in fig .
[ fig : general_modular_graph_transition ] ( a ) and the functional form of @xmath247 and @xmath261 [ shown in fig . [ fig : general_modular_graph ] ( b)(c ) for a different gmgm ] .
first , notice how the boundary that marks the onset of the ( theoretically ) 1detectable region partitions the density space in two qualitative regimes : a regime where perfect detection is possible _ for all instances _ , and a region where it is not .
there is , of course , some level of arbitrariness involved in selecting the threshold @xmath138 [ see eq . ] .
but the fact that a line of constant @xmath0 partitions the space is a hint that while @xmath107 is not sufficient for undetectability , there exists a level of significant @xmath292 for which @xmath85 properly separates detectable and undetectable instances . the second important parallel concerns hypersurfaces of constant @xmath102 and their connection with @xmath293 .
we have argued in sec .
[ section : average_detectability ] that @xmath102 can be seen as a measure of how easy the inference problem is : large positive values , in particular , indicate a high level of significance of the hypothesis @xmath97 .
it is not surprising , therefore , that there exists a hypersurface of constant @xmath102 which _ also _ partitions the density space in two qualitative regions : within the information - theoretical framework itself .
however , random matrix theory offers some insights as to why one should have @xmath294 , see appendix [ appendix : connection_with_rmt ] for details .
] one where @xmath295 and one where @xmath293 is clearly greater than zero . on this hypersurface ,
the average level of significance is the same for all parameterizations of the gmgm ; our results show that the inference problem is also equally hard .
one could argue that these parallels are not so obvious in fig .
[ fig : general_modular_graph_transition ] ( a ) ; we therefore focus on a subset of the density space in fig .
[ fig : general_modular_graph_transition ] ( b)-(c ) to make our case clearer . in these figures , we plot the same information , but only for networks of constant density @xmath280 and size @xmath277 ( b ) and @xmath281 ( c ) .
we also show the probability @xmath296 that the inferred partition is correlated with the planted partition . this a direct measurement of the fraction of detectable instances , which we compare against @xmath282 .
it never reaches 0 , because random fluctuations produce correlated partitions even when @xmath101 ( the rnmi corrects for the _ average _ correlation ) . if @xmath91 were a necessary and sufficient condition for detectability , then @xmath282 and @xmath297 would correspond perfectly .
but since @xmath91 is only a _ necessary _ condition , @xmath298 acts as an upper bound rather than an exact expression , i.e. , @xmath299 can never be greater than @xmath298 .
two further observations must be made .
first , it is known that in the sparse two - blocks ssbm , the transition between undetectable and detectable regions @xcite occurs on the so - called kesten - stigum ( ks ) bound located at @xmath283 for finite size instances . despite the fact that this bound was derived for infinite ensembles , it holds very well in the finite case , as shown in fig .
[ fig : general_modular_graph_transition ] ( b)(c ) .
but the finite size approach has the potential to be more precise .
building upon the interpretation of @xmath102 as a measure of the average difficulty of the inference problem , we set a threshold @xmath300 on the average detectability . for this choice of threshold ,
the approximate hypersurface equation predicts a transition at @xmath301 very close to the ks bound , but with a correction for nonvanishing densities .
interestingly , one can also motivate this choice of threshold with random matrix theory ( see appendix [ appendix : connection_with_rmt ] for details ) .
second , in asymptotic theories , the sbm is either said to be undetectable with overwhelming probability , or the converse .
the finite size approach is more nuanced in the sense that it accounts for random fluctuations , which are also manifest in empirical results [ see the curves @xmath296 ] .
while @xmath0detectability is not perfect , as is argued above , it nonetheless accounts for a sizable portion of the uncorrelated instances ( since it dips well below @xmath303 near @xmath260 ) .
building upon ideas from statistical theory , we have developed a framework to study the information - theoretic detectability threshold of the finite size sbm .
our analysis relies on two different interpretations of the log - likelihood ratio @xmath85 of the sbm and its equivalent random ensemble .
we have used the rigorous interpretation of @xmath85 to put a necessary condition on detectability .
we have then computed the distribution of @xmath85 , and proved that up to half of the instances of the finite size sbm could be declared undetectable on the basis of this simple test alone .
we have further argued that the average of @xmath85 could be interpreted as a measure of inference difficulty .
this interpretation has proved to be fruitful ; starting with a compact form for @xmath102 , we have established the existence of a large equivalence class with respect to average detectability . in appendix
[ appendix : noisy_sbm ] , we have shown that @xmath85 can also be used to prove that , quite naturally , detectability decreases when the datasets are noisy . using a correspondence with the finite size kesten - stigum bound ( as well as with random matrix theory , see appendix [ appendix : connection_with_rmt ] ) , we have presented numerical evidence that the hypersurface @xmath300 separates detectable from undetectable instances in a special case of the sbm .
the unifying theme of this contribution has been the idea that @xmath102 quantifies both detectability and consistency in the finite size sbm .
this interpretation leaves many questions open for future works .
perhaps the most important of all : can one justify the threshold @xmath304 within the framework of the theory itself , with no recourse to the ks bound or random matrix theory ?
a second important question pertains to sufficiency : can one modify the condition to make it necessary _ and _ sufficient ? or is a completely different approach needed ? in asymptotic analyses of the limit , one can use different conditions to bound the limit from above and below , as is done in ref .
can a similar approach be fruitfully applied to finite instances ? in closing , let us mention a few of the many possible generalizations of the methods introduced .
first , it will be important to verify how our approach behaves in the limit @xmath40 .
how this limit is taken will matter .
in particular , we believe that our framework has much to say about the limit where @xmath305 , since it does not assume poisson distributed degree , unlike other asymptotic theories of the limit .
second , we see no major obstacle to a generalization of our methods to other generative models of networks with a mesoscopic structure .
this includes , for example , the consistency of graphons , a subject whose study has been recently undertaken @xcite .
changing the null model from the equivalent random network ensemble to the configuration model @xcite could even allow an extension to degree - corrected sbm @xcite .
we thank guillaume st - onge and charles murphy for useful discussions and comments . this work has been supported by the fonds de recherche du qubec - nature et technologies ( j .- g.y .
, p.d ) , the conseil de recherches en sciences naturelles et en gnie du canada ( l.j.d . ) and the james s. mcdonnell foundation postdoctoral fellowship ( l.h .- d . ) , and grant dms-1622390 from the national science foundation ( l.h .- d . ) .
p.d . and e.l .
are grateful to d. ct ( p.d . , e.l ) and p. mathieu ( e.l . ) for financial support .
one almost never has a perfect knowledge of the structure of real networks .
the culprit can lie at the level of data collection , storage , transmission or a combination of the above , but the outcome is the same : some edges are spurious and others are omitted @xcite . to model imperfect knowledge
, we will suppose that instances of the sbm first go through a noisy channel where @xmath138 modifications
random edge removals or additions are applied to the structure . only then are we asked to tell which of hypotheses @xmath97 and @xmath96 is the most likely .
it should be clear that it will be harder to separate the two hypotheses , since noise is not necessarily aligned with the planted partition .
we will approach the problem with the following _ universal perturbation process _ ( upp ) : at each step @xmath306 of this process , a new random edge is added with probability @xmath307 ; otherwise , a random edge is removed .
if a new edge must be added , then it is selected uniformly from the set of non - edges .
if an edge must be removed , then it is selected uniformly from the set of edges already present in the network .
this randomization step is then repeated @xmath138 times .
we call this process universal because one can map arbitrary perturbation patterns onto one or successive upp with different parameters @xmath307 . to prove that @xmath102 decreases as a result of any sufficiently long upp
, we will show that the total derivative @xmath308 is negative everywhere . in so doing , we assume that the process can be approximated as a continuous one ( both with regards to `` time '' @xmath306 and discrete quantities such as @xmath26 ) .
admittedly , a more rigorous approach would be needed to settle the matter unequivocally , but we argue that the method presented in this appendix give a good intuition for the problem . without specifying the dynamics , and using eq .
, one can compute @xmath309 = \alpha_{rs } x_{rs}\;,\ ] ] where @xmath184 is identical to eq . .
this leaves the @xmath310 terms , whose expressions are determined by the perturbation dynamics . for the upp , the evolution of @xmath311 is determined by the set of differential equations @xmath312}{\sum_{r\leq s}m_{rs}(t)}\ ! + \ ! \frac{c\ [ m^{\max}_{rs } - m_{rs}(t)]}{m^{\max } - \sum_{r\leq s}m_{rs}(t)}.\ ] ] the first term accounts for edge removal events , which occur with probability @xmath313 and involve edges that connect nodes in blocks ( @xmath6 , @xmath8 ) with probability @xmath314 .
a similar argument leads to the second term , which accounts for edge creation events .
equation can be transformed into an equation for @xmath310 by dividing through by @xmath28 , and then using the definitions @xmath315 and @xmath316 .
we find @xmath317\ ; , \label{eq : appendix_p_rs_t}\ ] ] which , upon substitution in eq .
, yields @xmath318 \left [ \frac{f(c ) f(\rho)}{f(p_{rs } ) } - 1\right]\;,\ ] ] where @xmath319/[\rho n(n-1)]$ ] is a non - negative factor , and where we have defined @xmath320 .
it turns out that the sum is not only globally negative , but that each term is also individually negative , i.e. , @xmath321 \left [ \frac{f(c ) f(\rho)}{f(p_{rs } ) } - 1\right ] \leq 0\qquad \forall r\leq s.\ ] ] this comes about because the sign of the logarithm always matches that of the bracket . to prove this statement
, we treat 5 different cases and use the following identities repeatedly : @xmath322 the cases are : 1 . if @xmath323 : the logarithm equals 0 and the upper bound of eq
. holds .
2 . if @xmath324 and @xmath325 : the logarithm is positive [ see eq . ] .
the bracket is also positive , since inequality can be rewritten as @xmath326 using the fact that @xmath327 .
this simplifies to @xmath328 , in line with our premise .
if @xmath324 and @xmath329 : the logarithm is positive . using our premise , we conclude that @xmath330 and @xmath331
. therefore , @xmath332 , i.e. , the bracket is positive .
if @xmath333 and @xmath334 : the logarithm is negative . using our premise , we conclude that @xmath335 and @xmath336 .
therefore , @xmath337 , i.e. , the bracket is negative .
5 . if @xmath333 and @xmath338 : the logarithm is negative .
the bracket is also negative , since the converse of inequality can be rewritten as @xmath339 using the fact that @xmath340 .
this simplifies to @xmath341 , in line with our premise .
this list covers all cases , and therefore completes the proof that @xmath342 , i.e. , that average detectability decreases as a result of the application of a upp .
in refs . @xcite it is argued that sbm is undetectable when the extremal eigenvalues of the modularity matrix of its instances merge with the so - called `` continuous eigenvalue band '' .
it is proved in ref .
@xcite that this occurs when @xmath343 for the 2 block ssbm with poisson distributed degrees . since we are concerned with the finite case , let us first modify this result to account for binomial distributed degrees instead .
it turns out that the corrected condition is found by substituting the expectations of poisson variables [ in the rhs of eq .
] by that of binomial variables .
this leads to @xmath344}\;,\ ] ] or , in terms of the natural parameters of the gmgm , @xmath345 this equation bears striking similarity with eq . , our approximate equation for curves of constant @xmath102 .
in fact , for the 2 block ssbm ( @xmath346 ) , the latter reads @xmath347 one obtains an exact equivalence between the two expressions by setting @xmath348 . the fact that modularity based spectral methods can not infer a correlated partition if @xmath349 [ eq . ] can thus be understood as stemming from a lack of statistical evidence for the sbm .
66ifxundefined [ 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 * * , ( ) link:\doibase 10.1016/j.physrep.2009.11.002 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( )
@noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop * * , ( ) link:\doibase 10.1073/pnas.0605965104 [ * * , ( ) ] @noop _ _ ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) in @noop _ _ ( , ) pp . @noop * * , ( ) @noop ( ) @noop ( ) @noop ( ) @noop _ _ ( , ) @noop ( ) @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.6084/m9.figshare.1164194 [ `` , '' ] ( ) @noop _ _ ( , ) @noop _ _ ( , ) @noop _ _ ( , ) @noop * * , ( ) @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) * finite size analysis of the detectability limit of the stochastic block model + supplemental material *
[ thm : avg_l ] let @xmath97 be the distribution over all graphs of @xmath1 nodes prescribed by the stochastic block model of parameters @xmath158 and block size vector @xmath243 .
let @xmath96 be the distribution over all graphs of @xmath1 nodes , as prescribed by the erd - rnyi model @xmath350 , where @xmath351 is the average density and where @xmath352 is a function of @xmath243 .
define @xmath85 as the random variable @xmath353 whose support is all graphs of @xmath1 nodes , drawn from distribution @xmath97 .
then , @xmath354 where @xmath355 is the kullback - leibler divergence and @xmath356 is the binary entropy function .
the first equality follows directly from the definition of the average of @xmath85 . in the main text
, we argued that it was straightforward to show by direct calculation that @xmath102 led to @xmath357 ; we now carry out this calculation explicitly ( [ subsec : direct_proof_avg ] ) .
we also give an alternative proof of theorem [ thm : avg_l ] , where we obtain the expression in terms of difference of entropies directly from the kl divergence ( [ subsec : info_proof_avg ] ) .
note that to lighten the text , we will use `` @xmath34 '' as an abbreviation of the more precise `` @xmath358 '' . as we show in the main text
, one can think of the support of @xmath137 as all vertex labeled graphs of @xmath1 nodes , but also as all symmetric matrices @xmath359 with entries in @xmath360 and upper bounds given by @xmath361 .
in other words , all graphs drawn from the sbm of parameters @xmath362 can be mapped to a particular realization of @xmath363 , drawn with probability @xmath364 } = \prod_{r\leq s } { \mathrm{pr}\left[m_{rs}\right]}$ ] , where @xmath365}=\binom{m_{rs}^{\max}}{m_{rs } } p_{rs}^{m_{rs}}(1-p_{rs})^{m_{rs}^{\max}-m_{rs}}$]is the probability mass function of the binomial distribution . using this simpler definition of the random variable @xmath137
, we write its expectation as @xmath366 } \left\ { \sum_{r\leq s } \frac{m_{rs}}{m^{\max } } \log\left[\frac{p_{rs}}{\rho}\frac{1-\rho}{1-p_{rs}}\right ] + \sum_{r\leq s } \alpha_{rs } \log\left[\frac{1-p_{rs}}{1-\rho}\right]\right\}\;.\ ] ] it is straightforward to check that by the linearity of expectations @xmath367 } \sum_{r\leq s } \frac{m_{rs}}{m^{\max } } \log\left[\frac{p_{rs}}{\rho}\frac{1-\rho}{1-p_{rs}}\right ] & = \sum_{r\leq s } \log\left[\frac{p_{rs}}{\rho}\frac{1-\rho}{1-p_{rs}}\right ] \frac{1}{m^{\max}}\sum_{\bm{m } } { \mathrm{pr}\left[\bm{m}\right ] } m_{rs } \notag\;,\\ & = \sum_{r\leq s } \log\left[\frac{p_{rs}}{\rho}\frac{1-\rho}{1-p_{rs}}\right ] \frac{1}{m^{\max } } { \langle m_{rs } \rangle}\notag\;,\\ & = \sum_{r\leq s } \log\left[\frac{p_{rs}}{\rho}\frac{1-\rho}{1-p_{rs}}\right]\alpha_{rs}p_{rs}\;,\end{aligned}\ ] ] where we have used @xmath368 to get the last equality . since the second summation in eq .
is independent of @xmath363 , one obtains directly @xmath369 p_{rs } + \log\left[\frac{1-p_{rs}}{1-\rho}\right]\right\}\;,\notag\\ & = - \sum_{r\leq s } \alpha_{rs } \bigl[(1-p_{rs})\log(1-\rho ) + p_{rs}\log \rho\bigr ] + \sum_{r\leq s } \alpha_{rs } [ ( 1-p_{rs})\log(1-p_{rs } ) + p_{rs}\log p_{rs}]\;,\notag\\ & = h(\rho ) -\sum_{r\leq s } \alpha_{rs}\ , h(p_{rs } ) \;.\label{eq : average_likelihood}\end{aligned}\ ] ] the final result follows from ( i ) the definition @xmath370 , ( ii ) the normalization of @xmath371 and ( iii ) the fact that the average density can be written as @xmath372 .
we now provide an alternative proof of theorem [ thm : avg_l ] for the equation @xmath373 $ ] .
the following chain rule holds for the kullback - leibler divergence @xcite : @xmath374 where @xmath375 are joint distributions for @xmath376 on the same supports @xmath377 .
it is easy to see that if @xmath378 and @xmath379 are independent according to both distributions @xmath380 and @xmath2 , then the divergence is additive , i.e , @xmath381 this property trivially generalizes to an arbitrary number of random variables @xcite . because the distribution over all graphs can be seen as a product of independent distributions over all edges , we may write the kullback - leibler divergence as @xmath382 where @xmath383 and @xmath384 are the bernoulli random variables that govern the existence of the edge linking nodes @xmath15 and @xmath385 , according to the distribution @xmath97 and @xmath96 .
these variables have two outcomes : either the edge exists with probability @xmath386 ( resp .
@xmath387 ) , or does not with probability @xmath388 ( resp .
@xmath389 ) . we write
@xmath390 as an abbreviation of @xmath14 , the index of the block to which node @xmath15 belongs .
the divergence associated with an edge is therefore @xmath391 where @xmath356 is , again , the binary entropy . now , taking @xmath392 for all pairs and summing over all edges [ see eq . ] , we find , after grouping terms by types , @xmath393\;.\ ] ] the final result follows from the definition of the average density , after one substitutes @xmath28 for @xmath394 .
[ thm : symmetries ] all transformations @xmath395 of the parameter space of the sbm that are ( i ) reversible , ( ii ) space - preserving , and ( iii ) valid at every point of the parameter space can be written as [ eq_group : transformation_unconstrained_algebraic ] @xmath396 where @xmath397 and where @xmath398 and @xmath399 are permutations that acts on the set @xmath400 . under the additional constraint that @xmath401 be preserved by @xmath402 and equal to @xmath119 , one must have @xmath403 the ( constrained ) transformation that act on @xmath9 can be seen as an element of group @xmath404 . if @xmath405 , where @xmath2 is the dimension of @xmath87 and @xmath21 , then @xmath134 and @xmath135 can be seen as elements of the symmetric group @xmath406 .
the constrained symmetry group is therefore @xmath407 , and it is of order @xmath408 .
let us first introduce new notation to clarify the proof of theorem [ thm : symmetries ] .
first , we will define vectors @xmath409 and @xmath410 whose entries are the @xmath127 entries of the upper triangle ( and diagonal ) of @xmath21 and @xmath87 .
we use @xmath411 to denote the standard scalar product between vectors @xmath412 and @xmath413 in @xmath414 .
thus , in particular , we will write the average density as @xmath415 and the average detectability as @xmath416 \ ! p_{rs
} + \log\!\!\left[\frac{1-p_{rs}}{1-\rho}\right]\!\right\}\;,\notag \\ & = \braket{\alpha|u(\bm{\alpha},\bm{p})}\ ; , \label{eq : braket_def_l}\end{aligned}\ ] ] where @xmath417 is @xmath127dimensional vector parametrized by @xmath120 , whose entries are given by @xmath418 we also introduce @xmath419 and @xmath420 , two @xmath421 permutation matrices such that @xmath422 and @xmath423 , where @xmath424 is the element @xmath425 of vector @xmath412 ( we use double indexes in the vectors to emphasis the matrix to vector change in notation ) . in this notation , eqs .
are given by [ eq_group : transformation_unconstrained_matricial ] @xmath426 where @xmath427 is a diagonal matrix with element @xmath428 on the diagonal , where @xmath429 is the identity matrix , and where @xmath430 is also a diagonal matrix .
the first part of theorem [ thm : symmetries ] follows trivially from the definition of the parameter space , and from the fact that transformations which meet requirements ( i ) , ( ii ) and ( iii ) are the symmetries of this space . since @xmath87 takes its value in the standard simplex of dimension @xmath128 and
since @xmath21 is confined to the @xmath127dimensional unit cube , the complete parameter space is given by the cartesian product of both space we can simply compose the symmetry group of both spaces to obtain the complete symmetry group .
the symmetry group of the standard simplex and that of the unit cube are well - known results in geometry and algebra @xcite : the symmetry group of the @xmath127-dimensional unit cube is isomorphic to the hyperoctahedral group @xmath130 , and the symmetry group of the @xmath128-dimensional standard simplex is isomorphic to the symmetric group @xmath131 .
their action can be written as in eq .
, which proves the first part of the theorem .
to prove the second part of the theorem , we look for the subset of all transformations of the form that also preserve @xmath102 , i.e. , transformations @xmath138 in @xmath431 that map @xmath120 to @xmath121 and that satisfy @xmath432 it is easy to check that if @xmath433 and @xmath434 with @xmath435 , then the average density equals @xmath436 with the same constraint on the permutations and @xmath437 , the average of the normalized log likelihood is given by , in the transformed parameters , @xmath438 - \sum_{r\leq s}\alpha_{\pi(r , s)}h[\gamma + ( 1 - 2\gamma)p_{\pi(r , s)}]\;,\\ & = h(\rho ) -
\sum_{r\leq s}\alpha_{rs } h(p_{rs } ) = { \langle \mathcal{l}(\bm{\alpha},\bm{p } ) \rangle}\;,\end{aligned}\ ] ] since @xmath439 and permutations can be ignored ( they induce an irrelevant reordering of the sum ) .
the constrained transformations therefore preserve @xmath102 .
+ the above calculation is the sufficient part of the proof . to complete the proof
we must show that @xmath102 is conserved _ only if _ @xmath434 and @xmath433 .
first , we note that by the properties of the scalar product and permutation matrices , we have the following obvious symmetry @xmath440 which is valid for all permutation matrices @xmath419 .
we use this symmetry to `` shift '' all permutation matrices to the second part of the scalar product representation of @xmath102 , i.e. we write @xmath441 ( we use the fact that @xmath442 is a permutation matrix by definition ) .
now , from eq . , it is clear that we will have @xmath443 if and only if @xmath444 where @xmath445 .
since @xmath446 is analytic in @xmath87 , we can expand it by using taylor series ; this creates an infinite series of constraints that must all be satisfied .
in particular , condition will be satisfied only if @xmath447 this is true if and only if , for all @xmath136 , one has @xmath448 where @xmath449 . here
, @xmath450 is the transformed vector @xmath451 , on which the inverse of permutation @xmath134 is also applied .
note that in writing condition we have used @xmath452 in the denominators .
let us now suppose that @xmath87 tends to the point @xmath453 , which is such that @xmath454 for all @xmath136 except for @xmath455 ( i.e. , @xmath456 ) . in this limit , eq
. reads @xmath457 which is trivially satisfied when @xmath455 but not otherwise .
let us suppose @xmath458 and expand the equation around @xmath459
. from this second series expansion one concludes that the equality is satisfied if either @xmath460 or @xmath461 . in both cases
, the indices must match , which implies that @xmath462 . by repeating the same argument for all @xmath463 ,
we conclude that @xmath464 .
thus , the map @xmath465 is a symmetry only if @xmath466 .
this leaves the last part open , i.e. , the proof that @xmath434 .
let us , by contradiction , assume that @xmath428 differs from one set of indices to the other and define the sets @xmath467 and @xmath468 by @xmath469 then one can write @xmath470 where @xmath471 returning to eq . for @xmath472 and using the newfound fact that @xmath473 which implies @xmath474 ( no more permutations ) , we find @xmath475 this can only be true if @xmath476 , i.e. , if @xmath477 or @xmath478 .
therefore , @xmath141 @xmath479 , with @xmath435 .
[ thm : convexity ] @xmath401 is convex with respect to @xmath21 .
this property of @xmath102 is perhaps surprisingly not a consequence of the convexity of the kl divergence .
instead , it follows from the log - sum inequality .
we prove that @xmath401 is convex with respect to @xmath21 by showing that it satisfies the convexity condition @xmath480 explicitly for all @xmath481 $ ] .
again , for the sake of clarity , we will use the notation developed in the previous section , and , in particular , write the density as @xmath482 . focusing on one of the terms on the lhs of eq .
, we have @xmath483\log\frac{(1-t)p_{rs } + tq_{rs } } { ( 1-t)\braket{\alpha|p}+t\braket{\alpha|q}}+ [ 1-(1-t)p_{rs } - tq_{rs}]\log\frac{1-(1-t)p_{rs } - tq_{rs}}{1-(1-t)\braket{\alpha|p}-t\braket{\alpha|q } } \bigg\}\\ = \alpha_{rs}\bigg\{[(1-t)p_{rs } + tq_{rs } ] \log\frac{(1-t)p_{rs } + tq_{rs } } { ( 1-t)\braket{\alpha|p}+t\braket{\alpha|q}}\\+ [ ( 1-t)(1-p_{rs})+t(1-q_{rs})]\log\frac{(1-t)(1-p_{rs})+t(1-q_{rs})}{(1-t)(1-\braket{\alpha|p})+t(1-\braket{\alpha|q } ) } \bigg\ } \label{eq : term_explicit } \end{gathered}\ ] ] it is easy to see that the log - sum inequality @xmath484 can be applied to both parts of eq . to separate terms by their coefficients @xmath485 and @xmath306 . repeating the same operation on all terms yields inequality .
[ cor : detect_easy ] define @xmath486 as the average of the random variable @xmath487\;,\ ] ] where @xmath18 is drawn from the sbm of parameters @xmath120 , and where @xmath488 are the maximum likelihood estimator of @xmath489 computed from the planted partition @xmath33 , i.e. , @xmath490 where @xmath491 and @xmath492 .
then , the following inequality holds @xmath493 from the calculations of sec .
[ section : proof_avg_l ] , we know that @xmath102 can be written as @xmath494 or , equivalently , as @xmath495 where @xmath496 and @xmath497 are the expected values of the estimators .
jensen s inequality then prescribes @xmath498 corollary [ cor : detect_easy ] says that estimating the parameters simplifies the detection problem _ under certain particular conditions_a result whose consequence are not immediately obvious .
it can be understood on an intuitive level as follows : suppose that we use the density matrix @xmath499 ( with @xmath500 ) and that , for the sake of simplicity , @xmath115 $ ] . then , consider the instance with edge count matrix @xmath501,[n^2/4,0]]$ ]
. the likelihood ratio will be much smaller if one compares hypotheses with the true parameters instead of with the parameters estimated from @xmath33 .
one could argue that a large ratio would be appropriate here ; the partition is , after all , very clear .
but the instance actually contains little information on the planted partition , since the observed structure is _ not _ due to the planted partition but instead to an extreme random fluctuation .
corollary [ cor : detect_easy ] can therefore be taken as a cautionary tale : parameter estimation _ creates _ statistical evidence , and it must be handled carefully .
in the main text , we have analyzed the general modular graph model of ref .
@xcite and have shown how our framework yielded nicely interpretable expressions for such a low - dimensional model .
we now investigate another simple model the general two blocks sbm which also lends itself to exhaustive analysis . the parametrization of the two blocks sbm is specified with the 3 probabilities @xmath502 as well as the two blocks sizes , @xmath503 and @xmath504 .
it contains the core - periphery model ( with @xmath505 and @xmath506 ) and the ssbm ( @xmath507 ) as special cases .
we follow the path laid out in our analysis of the gmgm and obtain a natural parametrization in terms of average density and displacement from the equivalent random model .
the simplest parameterization of this kind is perhaps [ eq : two_blocks_params ] @xmath508 figure [ fig : two_block_sbm ] ( a ) illustrates the change of variables
. it amounts to placing ourselves in a plane of constant density in the @xmath509 space , then measuring the distance from the point @xmath510 in the @xmath511 and @xmath512 direction .
again , much like in the case of the gmgm , not all @xmath513 tuples correspond to probabilities it is only true for the region of the constant density plane contained inside @xmath251 ^ 3 $ ] .
when @xmath63 is close to 0 or 1 , this region is a triangle . at intermediate values of @xmath63 , parts of the triangle lie outside the cube of probabilities the region is then a polygon of more than 3 edges , see fig .
[ fig : two_block_sbm ] ( b ) for an example .
we do not investigate the general 2 blocks sbm as thoroughly as the gmgm , since much of the results and observations are identical to those of the previous case study .
it will , however , be instructive to consider the average detectability and symmetries of this simple model .
the average log - likelihood @xmath102 is , in the @xmath514 space , @xmath515\right\}\notag\\ + ( 1-\beta)^2&\left\{h(\rho ) - h\left [ \rho + \frac{\beta^2}{1 - 2\beta(1-\beta)}\delta_x - 2\beta(1-\beta)\delta_y \right]\right\}\notag\\ + 2\beta(1-\beta)&\left\{h(\rho ) - h\left [ \rho + [ 1 - 2\beta(1-\beta)]\delta_y \right]\right\}\;,\end{aligned}\ ] ] our main objective in studying the two - blocks sbm is to showcase anew the technique introduced in the main text to compute the hypersurfaces of constant @xmath102 .
the first step is to invert the parametrization @xmath516 upon substitution of these parameters in @xmath517\;,\ ] ] one finds @xmath518 + \delta_y^2[\alpha_{12}(1-\alpha_{12})]\;.\end{aligned}\ ] ] equation predicts that these hypersurfaces are ellipses in the plane of constant density @xmath63 , centered at @xmath519 , with major axis @xmath520 and minor axis @xmath521 .
this result is put to the test in fig .
[ fig : two_block_sbm ] ( b ) , where we also show numerical solutions for comparison
. agreement is , once again , excellent when @xmath119 is not too large , while @xmath522 leads to significant errors in our prediction . following the ellipses in fig .
[ fig : two_block_sbm]-(c ) shows that that the easiest inference problems those where @xmath102 is the largest are the ones in the corner on the edges of the accessible region .
these regions are those where block pairs are well segregated , i.e. , the bipartite ensemble with @xmath523 [ top - corner ] , the perfectly assortative cases with @xmath524 where @xmath525 is a constant [ bottom edge ] .
the purpose of the parametrization @xmath513 is to facilitate the calculation of hypersurfaces of constant @xmath102 . for the gmgm
, the natural parametrization also had the added benefit of highlighting the symmetries naturally .
this needs not be the case in all variants of the sbm , as we now show .
let us first simplify the notation ; we define @xmath241 as the fraction of nodes that belong to block @xmath66 , i.e. , @xmath526 and @xmath527 ( not to be confused with the @xmath241 of the gmgm ) . in the limit where @xmath528 , we then have @xmath529 since the parametrization assigns a special significance to the @xmath530 direction , the symmetries that involve these two blocks [ and _ not _ the @xmath531 pair ] , are the simplest . by direct enumeration
, one finds @xmath532 i.e. , the identity , the pure graph complement , the permutation of block pairs ( 0,0 ) and ( 1,1 ) , and the same permutation accompanied by the graph complement . notice how this subset of transformation forms , once again , a group isomorphic to the klein four - group .
the transformation equations would be , however , much less compact if we were to list transformations involving , say , cyclic permutations of the blocks .
this situation is common : choices of parameters that favor a particular pair of blocks will yield compact symmetry equations for this pair of blocks , but not the others .
therefore symmetries are , in general , best expressed directly in terms of @xmath87 and @xmath21 , unless the model has a special and all encompassing parametrization like the gmgm .
in this short section , we detail the inference algorithm used in the case studies of the main text .
optimal inference can be achieved ( quite inefficiently ) by evaluating the complete posterior distribution of the model @xmath533 for all possible partitions .
the idea is to construct the marginal distribution @xcite @xmath534 which gives the probability that node @xmath15 is in block @xmath6 , given a network @xmath18 and some generating parameter @xmath21 and @xmath243 [ @xmath14 denotes the index of the block of node @xmath15 in partition @xmath535 .
it is easy to see that we can then maximize the probability of guessing the partition correctly by assigning nodes according to @xmath536 while this method is attracting , it runs into an important problem very quickly , for it is impossible to compute the marginal distribution exactly for networks of even moderate size an exponential number of terms is involved .
there are multiple ways to circumvent this problem .
a popular method is to construct the marginal distribution directly , using a belief propagation ( bp ) algorithm @xcite ( see ref .
@xcite for an overview of related method ) .
bp based methods are extremely efficient , since their complexity is of the order of the number of edges in the network @xcite .
moreover , they work virtually perfectly for all practical purposes @xcite .
the only problem is that bp algorithms rely on the hypothesis that the network is locally tree - like ; this is never entirely true , but still a reasonable approximation for sparse instances of the sbm .
unfortunately , since our analysis calls for accurate results in all regimes ( i.e. , not only on sparse networks ) , we are forced to turn an inefficient alternative of bp , the markov chain monte carlo ( mcmc ) method @xcite .
_ sample _ from the posterior distribution and can therefore be used to estimate the marginal distribution of eq . .
we will use one of the simplest implementation of the mcmc method , the metropolis - hastings algorithm .
one only needs to specify a _
proposal distribution _
( given a partition @xmath537 , what is the probability that the next partition in the chain is @xmath538 ? ) and an _ acceptance probability _
@xmath539 proportional to @xmath540 for all proposed transitions .
the proposal distribution usually intervenes in the calculation of @xmath541 but will not be needed here , since we opt for a symmetric distribution ( see next paragraph ) .
the chain formed by accepting transitions @xmath542 with probability @xmath543 is then ergodic if a number of conditions are met ( e.g. , the proposal distribution must respect the detailed balance )
. our proposal distribution will be simple : take a node at random , and assign it to a random block ( see ref .
@xcite for a more efficient alternative ) .
one can easily check that this distribution meets all the necessary conditions for the ergodicity of the markov chain .
following an argument analogous to that of ref .
@xcite , we find that the acceptance probability associated to this proposal distribution is @xmath544^{k_r^{(i)}}\!\!\left[\frac{p_{ss}(1-p_{rs})}{p_{rs}(1-p_{ss})}\right]^{k_s^{(i)}}\!\ !
\left[\frac{1-p_{rs}}{1-p_{rr}}\right]^{n_r -1}\!\!\left[\frac{1-p_{ss}}{1-p_{rs}}\right]^{n_s } \prod_{l\neq r , s } \left[\frac{p_{ls}(1-p_{rl})}{p_{rl}(1-p_{ls})}\right]^{k_l^{(i)}}\!\!\!\!\left[\frac{1-p_{ls}}{1-p_{rl}}\right]^{n_l}\ ; , \label{eq : a_sbm_single}\end{gathered}\ ] ] if node @xmath15 is switched from block @xmath6 to block @xmath8 , with @xmath545 denoting the number of neighbors of node @xmath15 in block @xmath6 . since there are @xmath546 terms in eq .
, transition probabilities can be calculated in @xmath547 time , assuming that products and exponentiations are @xmath548 , and that one keeps @xmath549 items in memory to track the neighborhoods and the block assignments .
updating the memberships and neighborhoods is then a @xmath550 operation .
the memory cost can be reduced to @xmath551 if the neighborhoods @xmath552 are computed on the fly .
this comes at no significant costs since keeping the memory up - to - date is as expensive as computing the memberships every time .
we give a reference implementation of the algorithm at www.github.com/jg-you/sbm_canonical_mcmc .
also , we note that if one samples from the general modular graph model ( gmgm ) , the algorithmic complexity is dramatically reduced , since the acceptance probability then simplifies to @xmath553^{k_s^{(i)}-k_r^{(i)}}\left[\frac{1-p_{in}}{1-p_{out}}\right]^{(n_s - k_s^{(i)})-(n_r - k_r^{(i ) } ) + 1}\ ; , \label{eq : a_ppm_single}\ ] ] i.e. , the transition probability only depends on the blocks of the nodes involved in the mcmc move . in a normal
setting where the planted partition is not known
one would initialize the metropolis - hastings at random or using the partition generated by a simpler algorithm .
but since we are interested in _ detectability _ , we might as well initialize the chain with the planted partition itself .
if the partition is undetectable , then the algorithm will show no particular preference for this initial configuration and `` wander away '' towards partitions uncorrelated with the planted partition .
however , if the partition is detectable , then the chain will concentrate around the initial configuration , and the marginal distribution will yield a distribution correlated with the planted partition .
this choice of initial condition is not merely a trick to avoid transients : it also allows correlated recovery even in the _ hard phase _ , where the planted partition is detectable , but exponentially hard to find .
in this phase , the planted partition is identifiable but not globally stable , in the sense that many other locally optimal partitions exist @xcite .
when this is the case , an algorithm that relies on local information such as the metropolis - hastings algorithm cannot be expected to identify the correct partition . by seeding the algorithm with the planted partition ,
we ensure that it will be found ( if it is detectable ) . even though we seed the inference algorithm with the planted partition
, the chain can still go through a transient before it settles in its steady state , since the planted partition need not be typical of the stationary distribution .
this can cause problems , since one must ensure that the chain has settled before the sampling begins otherwise biased results are to be expected . in fig .
[ fig : mcmc_calibration ] , we show the distance between the partition currently consider by the algorithm and the initial condition for different parametrization of the gmgm .
these results show that the transient can be extremely long when there is little difference between the parameters of the ensemble and that of the equivalent random ensemble .
for instance , the bottom panel of fig .
[ fig : mcmc_calibration ] show that the algorithm is still in the transient after @xmath554 steps , when it is applied to instances of the gmg with @xmath291 , @xmath555 $ ] , @xmath280 and @xmath556 . to avoid biases
we only start sampling after an initial burn - in period of @xmath557 steps ( when @xmath277 ) and of @xmath558 steps ( when @xmath281 ) . the numerical results of fig [ fig : mcmc_calibration ] indicate that the chain will have settled by then , on average . nodes and ( bottom panel ) @xmath281 nodes .
the vi is a metric that quantifies the distance between two partitions @xcite .
its maximum equals @xmath559 when the two compared partition have both @xmath2 clusters and @xmath560 ( this is the case here ) . each line shows the average of ( top panel ) @xmath561 ( bottom panel ) @xmath562 independent mh run , using a random network instance and the planted partition as an initial condition each time .
we only show positive values of @xmath238 for the sake of clarity ; the results are similar for negative values of @xmath238 .
, title="fig : " ] nodes and ( bottom panel ) @xmath281 nodes .
the vi is a metric that quantifies the distance between two partitions @xcite .
its maximum equals @xmath559 when the two compared partition have both @xmath2 clusters and @xmath560 ( this is the case here ) .
each line shows the average of ( top panel ) @xmath561 ( bottom panel ) @xmath562 independent mh run , using a random network instance and the planted partition as an initial condition each time .
we only show positive values of @xmath238 for the sake of clarity ; the results are similar for negative values of @xmath238 .
, title="fig : " ] 11ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop _
( , ) @noop _ _ ( , ) @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop * * , ( ) | it has been shown in recent years that the stochastic block model is undetectable in the sparse limit , i.e. , that no algorithm can identify a partition correlated with the partition used to generate an instance , if the instance is sparse and infinitely large .
real networks are however finite objects , and one can not expect all results derived in the infinite limit to hold for finite instances . in this contribution , we treat the finite case explicitly .
we give a necessary condition for finite size detectability in the general sbm , using arguments drawn from information theory and statistics .
we then distinguish the concept of average detectability from the concept of instance - by - instance detectability , and give explicit formulas for both definitions . using these formulas
, we prove that there exist large equivalence classes of parameters , where widely different network ensembles are equally detectable with respect to our definitions of detectability . in an extensive case study
, we investigate the finite size detectability of a simplified variant of the sbm , which encompasses a number of important models as special cases .
these models include the symmetric sbm , the planted coloring model , and more exotic sbms not previously studied .
we obtain a number of explicit expressions for this variant , and also show that the well - known kesten - stigum bound does not capture the phenomenon of finite size detectability even at the qualitative level .
we conclude with two appendices , where we study the interplay of noise and detectability , and establish a connection between our information - theoretic approach and random matrix theory . |
from the statistical study of financial time series have arisen a set of properties or empirical laws sometimes called `` stylized facts '' or seasonalities .
these properties have the characteristic of being common and persistent across different markets , time periods and assets @xcite . as it has been suggested @xcite
, the reason why these `` patterns '' appear could be because markets operate in synchronization with human activities which leave a trace in the financial time series .
however using the `` right clock '' might be of primary importance when dealing with statistical properties and the patterns could vary depending if we use daily data or intra - day data and event time , trade time or arbitrary intervals of time ( e.g. @xmath2 , @xmath3 , @xmath4 minutes , etc . ) .
for example , it is a well - known fact that empirical distributions of financial returns and log - returns are fat tailed @xcite , however as one increases the time scale the fat - tail property becomes less pronounced and the distribution approach the gaussian form @xcite . as was stated in @xcite , the fact that the shape of the distribution changes with time makes it clear that the random process underlying prices must have a non - trivial temporal structure . in a previous work allez et al .
@xcite established several new stylized facts concerning the intra - day seasonalities of single and cross - sectional stock dynamics .
this dynamics is characterized by the evolution of the moments of its returns during a typical day .
following the same approach , we show the bin size dependence of these patterns for the case of returns and , motivated by the work of kaisoji @xcite , we extend the analysis to relative prices and show how in this case , these patterns are independent of the size of the bin , also independent of the index we consider but characteristic for each index .
these facts could be used in order to detect an anomalous behaviour during the day , like market crashes or intra - day bubbles @xcite .
the present work is completely empirical but it could offer signs of the underlying stochastic process that governs the financial time series .
the data consists in two sets of intra - day high frequency time series , the cac @xmath0 and the s&p @xmath1 . for each of the @xmath5 days of our period of analysis ( march @xmath6 )
, we dispose with the evolution of the prices of each of the stocks that composes our indexes during a specific day from @xmath7 a.m. to @xmath8 p.m. the main reasons why we chose to work with these two indexes are : the number of stocks that compose them ( @xmath9 and @xmath10 ) , the time gap between their respective markets and the different range of stock prices ( between @xmath3 and @xmath11 usd for the s&p @xmath1 and between @xmath3 and @xmath12 eu for the cac @xmath0 ) .
as the changes in prices are not synchronous between different stocks ( figure [ fig : fig1 ] ) , we manipulated our original data in order to construct a new homogeneous matrix @xmath13 of `` bin prices '' . in order to do this , we divided our daily time interval @xmath14 $ ] in @xmath15 bins of size @xmath16 ( minutes ) , i.e. @xmath17 $ ] , @xmath18 , ... , b_{k } = [ 16:00 - t,16:00]$ ] , where the right endpoints of these intervals are called `` bin limits '' . for a particular day
@xmath19 , the prices that conform the matrix @xmath13 are given by the last prices that stock @xmath20 reaches just before a specific bin limit .
are the stocks and @xmath21 are bins . the asynchronous prices are show in red and the bin limit in blue . ]
each row in the matrix below represents the evolution of the prices of a particular stock as function of the bins .
for example , the element @xmath22 , represents the price for a particular day @xmath19 of the stock @xmath20 and just before the bin limit of the bin @xmath21 .
@xmath23 in a similar way , we can construct the matrix @xmath24 for each of the @xmath25 stocks .
@xmath26 is the price of the stock @xmath27 in the day @xmath19 and just before the bin limit of the bin @xmath21.@xmath28 in the following and for simplicity , we will refer to the price @xmath29 of a particular stock @xmath30 during a particular day @xmath31 and just before the bin limit of the bin @xmath21 as @xmath32 where @xmath33 .
we will perform our statistical analysis over the variable @xmath34 that can be computed from the matrices above . for our interests
we will be working with returns @xmath35 and relative prices @xcite@xmath36 the single or collective stock dynamics is characterized by the evolution of the moments of the returns ( or relative prices ) .
below , we show how we computed these moments @xcite .
the distribution of the stock @xmath37 in bin @xmath38 is characterized by its four first moments : mean @xmath39 , standard deviation ( volatility ) @xmath40 skewness @xmath41 and kurtosis @xmath42 defined as@xmath43where @xmath44 is the median of all values of @xmath34 and time averages for a given stock in a given bin are expressed with angled brackets @xmath45 .
the cross - sectional distributions ( i.e. the dispersion of the values of the variable @xmath46 of the @xmath47 stocks for a given bin @xmath38 in a given day @xmath48 ) are also characterized by the four first moments@xmath49 \label{7 } \\
\sigma _ { d}^{2}(k , t ) & = & \left [ x_{\alpha } ^{2}(k , t)\right ] -\mu _ { d}^{2}(k , t ) \label{8 } \\ \zeta _ { d}(k , t ) & = & \frac{6}{\sigma _ { d}(k , t)}(\mu _ { d}(k , t)-m_{d}(k , t ) ) \label{9 } \\
\kappa _ { d}(k ) & = & 24\left ( 1-\sqrt{\frac{\pi } { 2}}\frac{\left [ |x_{\alpha } ( k , t)-\mu _ { \alpha } ( k)|\right ] } { \sigma _ { d}(k)}\right )
\label{10}\end{aligned}\]]where @xmath50 is the median of all the @xmath47 values of the variable @xmath46 for a given @xmath51 and the square brackets @xmath52 $ ] represent averages over the ensemble of stocks in a given bin and day . if @xmath34 are the returns , @xmath53 can be seen as the return of an index equi - weighted on all stocks .
the following results are in complete agreement with the results previously reported by @xcite .
figure [ fig : fig2 ] shows the stock average of the single stock mean @xmath54 $ ] , volatility @xmath55 $ ] , skewness @xmath56 $ ] and kurtosis @xmath57 $ ] for the cac @xmath0 ( blue ) and the s&p @xmath1 ( green ) , and @xmath58 minute bin .
as can be seen in figure [ fig : fig2](a ) , the mean tends to be small ( in the order of @xmath59 ) and noisy around zero .
the average volatility reveals the well known u - shaped pattern ( figure [ fig : fig2](b ) ) , high at the opening of the day , decreases during the day and increases again at the end of the day .
the average skewness ( figure [ fig : fig2](c ) ) is also noisy around zero .
the average kurtosis exhibits an inverted u - pattern ( figure [ fig : fig2](d ) ) , it increases from around @xmath60 at the beginning of the day to around @xmath61 at mid day , and decreases again during the rest of the day . + as the time average of the cross sectional mean is equal to the stock average of the single stock mean , the result we show in figure [ fig : fig3](a ) is exactly the same as the one shown in figure [ fig : fig2](a ) .
the time average of the cross sectional volatility @xmath62 ( figure [ fig : fig3](b ) ) reveals a u - shaped pattern very similar to the stock average volatility , but less noisy ( less pronounced peaks ) .
the dispersion of stocks is stronger at the beginning of the day and decreases as the day proceeds .
the average skewness @xmath63 is noisy around zero without any particular pattern ( figure [ fig : fig3](c ) ) .
the cross sectional kurtosis @xmath64 ( figure [ fig : fig3](d ) ) also exhibits an inverted u - pattern as in the case of the single stock kurtosis .
it increases from around @xmath65 at the beginning of the day to around @xmath66 at mid day , and decreases again during the rest of the day .
this means that at the beginning of the day the cross - sectional distribution of returns is on average closer to gaussian .
+ in figure [ fig : fig4 ] , we compare the stock average of single stock volatility @xmath55 $ ] ( black ) , the time average of the cross - sectional volatility @xmath67 ( red ) and the average absolute value of the equi - weighted index return @xmath68 ( blue ) for the cac @xmath0 , and for @xmath58 ( left ) and @xmath69 minute bin ( right ) .
similar results were obtained for the s&p @xmath1 . as can be seen
, the average absolute value of the equi - weighted index return also exhibits a u - shaped pattern and it is a proxy for the index volatility .
one thing that results interesting to observe is that the values of these volatilities actually depends of the size of the bin that we consider . for @xmath69 minute bin ,
the volatilities double the values found for @xmath58 minute bin ( we will discuss this result in the next sections ) . in order to compute the correlation between stocks ,
we first normalize the returns by the dispersion of the corresponding bin @xcite i.e. , @xmath70the @xmath71 correlation matrix for a given bin @xmath38 would be given by@xmath72 in figure [ fig : fig5](a ) we show the average correlation between stocks ( blue ) and top eigenvalue @xmath73 ( green ) for the cac @xmath0 .
as can be seen the largest eigenvalue is a measure of the average correlation between stocks @xcite .
this average correlation increases during the day from a value around @xmath74 to a value around @xmath75 when the market closes . for the case of smaller eigenvalues
, what we can see is that the amplitude of risk factors decreases during the day ( figure [ fig : fig5](b ) ) , as more and more risk is carried by the market factor ( figure [ fig : fig5](a ) ) @xcite . in order to simplify the computation of the @xmath76 correlation matrices for each bin
@xmath38 in the case of the s&p @xmath1 , we computed the correlation matrix @xmath77 for @xmath61 different sets of stocks : @xmath78 : composed by the @xmath79 first stocks of the s&p @xmath1 ; @xmath80 : composed by @xmath79 stocks randomly picked ; and @xmath81 : composed by @xmath82 stocks randomly picked .
figure [ fig : fig6](a ) shows @xmath83 as function of the bins .
although the values of the eigenvalues seem to be out of scale , it can be seen clearly that the average correlation increases during the day .
this scale conflict is solved by normalizing the value of the top eigenvalue not by @xmath47 but by the sample size @xmath84 ( i.e. @xmath79 or @xmath82 ) ( figure [ fig : fig6](b ) ) . as can be seen the average correlation of the index
can be computed by taking a subset of it which means that actually just the more capitalized stocks in the index drive the rest of stocks .
in this section we will report the results we found for the s&p @xmath1 .
similar results were found also for the cac @xmath0 .
we will see how in the case of the relative prices these intra - day seasonalities are independent of the size of the bin , also independent of the index we consider ( but characteristic for each index ) however this is not the case for the returns .
each path in figure [ fig : fig7 ] represents the evolution of a particular moment of one of the stocks that compose the s&p @xmath1 ( i.e. one path , one stock moment ) .
the stock average of the single stock mean @xmath85 $ ] , volatility @xmath86 $ ] , skewness @xmath56 $ ] and kurtosis @xmath57 $ ] of the s&p @xmath1 are shown in black .
the stock average of the single stock mean varies around zero .
the average volatility increases logarithmically with time .
the skewness varies between @xmath87 $ ] with an average value of zero .
the single stock kurtosis takes values between @xmath88 $ ] with an average value of one and its stock average starts from a value around @xmath60 in the very beginning of the day and decreases quickly to the mean value @xmath89 in the first minutes of the day .
+ each path in figure [ fig : fig8 ] represents the evolution of a particular index moment during a particular day ( i.e. one path , one day moment ) . as in the case of the single stock volatility ,
the cross - sectional dispersion @xmath90 increases logarithmically with respect to the time ( figure [ fig : fig8](b ) ) .
the cross - sectional skewness @xmath91 takes values in the interval @xmath92 $ ] with an average value of zero ( figure [ fig : fig8](c ) ) .
the average kurtosis @xmath64 starts from a value around @xmath65 in the very beginning of the day and decreases quickly to the mean value @xmath60 in the first minutes of the day ( figure [ fig : fig8](d ) ) .
+ similarly as we did in section 3.3 for returns , in figure [ fig : fig9 ] we show a comparative plot between the stock average of the single stock volatility @xmath55 $ ] , the time average of the cross - sectional volatility @xmath62 and the average absolute value of the cross - sectional mean @xmath68 for the relative prices of the s&p @xmath1 , and for @xmath58 and @xmath69 minute bin .
as can be seen , these three measures exhibit the same kind of intra - day pattern ( as it did in the case of the returns ) .
but the most important fact is to notice that this intra - day seasonality is independent of the size of the bin , also independent of the index we consider , but characteristic for each index ( see inset figure [ fig : fig9 ] ) .
as we saw in the last section , the volatilities for the relative prices exhibit the same kind of intra - day pattern ( figure [ fig : fig9 ] ) .
this intra - day seasonality is independent of the size of the bin , and the index we consider , but characteristic for each index . actually , this is not true in the case of the returns as we already suggested in section 3.3 from figure [ fig : fig4 ] . if we consider the odd moments ( mean and skewness ) of the returns , the behaviour is basically the same ( noisy around zero ) and without any particular pattern , independently of the bin size ( as can be seen in figures [ fig : fig2 ] , [ fig : fig3 ] and [ fig : fig10 ] ) .
but for the case of the even moments of returns , although they exhibit the well known u and inverted u - patterns , these patterns depend on the bin size .
this fact is well illustrated through figures [ fig : fig11 ] and [ fig : fig12 ] in where we have chosen @xmath3 different values of bin size from @xmath93 to @xmath94 minutes . in these figures
we show the time average of the cross - sectional volatility and kurtosis for the s&p @xmath1 but a similar bin size dependence can be shown for the cac @xmath0 or any other index and also for the time average of the single stock volatility and kurtosis . for @xmath3 different values of bin size @xmath16 .
] by other hand the kurtosis is a decreasing function of the size of the bin and the inverted u - pattern is evident just when we consider `` small '' bin sizes , in our case this occurs for @xmath58 and @xmath93 minute bin ( figure [ fig : fig12 ] ) .
this represents a confirmation that on small scales the returns have heavier tails , and on long time scales they are more gaussian @xcite . for @xmath3 different values of bin size @xmath16 . ]
one of the motivations we had in order to explore into the intra - day seasonalities for relative prices was due to kaisoji s previous work @xcite . in his work
he found that the upper tail of the complementary cumulative distribution function of the ensemble of the relative prices in the high value of the price is well described by a power - law distribution which when its exponent approached two , the japan s internet bubble burst .
taking into consideration our recent findings we suggest the use of the bin size independence for intra - day patterns in relative prices in order to characterize `` atypical days '' for indexes and `` anomalous behaviours '' for stocks .
the time average of the cross - sectional moments represents the average behaviour of a particular index moment during an average day . in figure
[ fig : fig8 ] we showed how each path represents the evolution of a particular index moment for one of the days of the period under analysis ( i.e. one path , one day moment ) .
if we look directly into the prices of the cac @xmath0 and s&p @xmath1 , we can observe during day @xmath95 a fall of the prices of the stocks that compose both indexes . during the days before and following day @xmath95 , the ( index ) moments move along our intra - day pattern . moreover ,
if we pick randomly one day from our period of analysis , in most of the cases our index during that day will behave as our intra - day seasonality ( as in figure [ fig : fig13 ] ) , but the one for day @xmath95 will not . in figure
[ fig : fig13 ] we show our ( cross - sectional ) intra - day seasonalities for the mean ( a ) and volatility ( b ) in blue and in clear blue the respective cross - sectional stock moments for 3 days randomly picked .
the average behaviour ( of the moments ) of our index during these days moves along with our intra - day pattern .
this is not the case of the curve corresponding to the day @xmath95 shown in red which clearly diverges from the expected behaviour .
this is what could be called as an `` atypical day '' for the s&p @xmath1 .
we could used the same reasoning as before in order to characterize `` anomalous behaviours '' in stocks . as we said in section 4.1 ,
each path in figure [ fig : fig7 ] represents the average evolution of a particular moment of one of the stocks that compose the s&p @xmath1 .
the stock average of those single stock moments represents the average behaviour of that moment for an average stock during an average day of our period of analysis .
that means that if from our set of stocks we pick randomly one stock , in most of the cases ( its moments ) will behave as our intra - day seasonality .
this is clearly illustrated in figure [ fig : fig14 ] where we show our intra - day seasonalities for the mean ( a ) and volatility ( b ) in blue and the respective single stock moments for 3 stocks randomly picked in clear blue .
as can be seen , the average behaviour of the moments of these stocks move along with our intra - day patterns .
however this is not the case for the curves shown in red which have been chosen on purpose to illustrate how in this case the stock @xmath96 behaves in an anomalous way with respect to what is expected .
in the present report , we have analysed the intra - day seasonalities of the single and cross - sectional ( or collective ) stock dynamics . in order to do this
, we characterized the dynamics of a stock ( or a set of stocks ) by the evolution of the moments of its returns ( and relative prices ) during a typical day .
what we have called `` single stock intra - day seasonalities '' is the average behaviour of the moments of the returns ( and relative prices ) of an average stock in an average day . in the same way ,
the cross - sectional intra - day seasonality is not more than the average day behaviour of an index moment .
we presented these intra - day seasonalities for returns ( figures [ fig : fig2 ] and [ fig : fig3 ] ) and relative prices ( figures [ fig : fig7 ] and [ fig : fig8 ] ) and compared the stock average of single stock volatility @xmath55 $ ] , the time average of the cross - sectional volatility @xmath62 and the average absolute value of the equi - weighted index @xmath68 ( figures [ fig : fig4 ] and [ fig : fig9 ] ) .
one thing that results interesting to observe , in the case of the returns , is that these `` patterns '' actually depend on the size of the bin .
this fact was well illustrated with @xmath3 different values of bin size through figure [ fig : fig11 ] for volatilities and figure [ fig : fig12 ] for kurtosis in which its inverted u - pattern is evident just when we consider `` small '' bin sizes . in the case of relative prices , the volatilities also exhibit the same kind of intra - day pattern ( figure [ fig : fig9 ] ) , but contrary with the returns , it is independent of the size of the bin , and the index we consider , but characteristic for each index .
we suggested in section @xmath97 how this bin size independence of intra - day patterns in relative prices could be used in order to characterize `` atypical days '' for indexes and `` anomalous behaviours '' in stocks .
this was showed in figures [ fig : fig13 ] and [ fig : fig14 ] where we presented our intra - day seasonalities for the mean ( a ) and volatility ( b ) in blue and the respective the cross - sectional moments for 3 days ( and the single stock moments for 3 stocks ) randomly picked in clear blue and we saw how the average behaviour of their moments move along with our intra - day patterns which was not the case for the day @xmath95 and the stock @xmath96 .
esteban guevara thanks anirban chakraborti , frederic abergel , remy chicheportiche and khashayar pakdaman for their support and discussions .
special thanks to the european commission , the ecuadorian government and the secretara nacional de educacin superior , ciencia , tecnologa e innovacin , senescyt . | in this paper we perform a statistical analysis over the returns and relative prices of the cac @xmath0 and the s&p @xmath1 with the purpose of analyzing the intra - day seasonalities of single and cross - sectional stock dynamics . in order to do that
, we characterized the dynamics of a stock ( or a set of stocks ) by the evolution of the moments of its returns ( and relative prices ) during a typical day .
we show that these intra - day seasonalities are independent of the size of the bin , and the index we consider , ( but characteristic for each index ) for the case of the relative prices but not for the case of the returns .
finally , we suggest how this bin size independence could be used to characterize `` atypical days '' for indexes and `` anomalous behaviours '' in stocks . |
the afterglow light curve of most grbs shows a smooth power law decay with the time after the burst .
this is consistent with the prediction of the simplest model of a spherical blast wave propagating into a uniform medium , where the spectrum consists of several power law segments in which @xmath9 ( sari , piran & narayan 1998 ) , and remains valid for an external density that drops as @xmath10 as expected for the stellar wind of a massive star progenitor ( chevalier & li 2000 ) and for a jetted outflow ( rhoads 1999 ; sari , piran & halpern 1999 ) .
however , a few grbs do show deviations from a smooth power law light curve .
grb 970508 showed a brightening at @xmath11days , which was observed in optical ( galama et al .
1998 ) and possibly also in x - rays ( piro et al . 1998 ) .
an achromatic bump was observed in the afterglow of grb 000301c at @xmath12 days ( berger et al . 2000 ; sagar et al . 2000 ) and was attributed , among other interpretations , to gravitational microlensing ( garnavich , loeb & stanek 2000 ) . in these two cases a single epoch of re - brightening was observed .
in contrast to this , grb 021004 displayed a highly variable light curve , with several wiggles and deviations from a simple power law behavior .
this variable light curve was interpreted as arising either due to a variable external density ( lazzati et al .
2002 ; nakar , piran & granot 2003 ; heyl & perna 2003 ) or due to a variable energy in the observed portion of the afterglow shock .
the latter can arise due to an irregular initial distrubotion of the energy per unit solid angle in the outflow ( the patchy shell model , e.g. kumar & piran 2000b ) .
the early slow decay of grb 021004 was interpreted by fox et al .
( 2003 ) as arising from delayed shocks and continuous energy ejection from the source ( rees & mszros 1998 ; kumar & piran 2000a ; sari & mszros 2000 ) .
grb 030329 was detected by hete - ii on march 29 , 2003 ( 11:37:14.67 ut ) , and its location was distributed ( vanderspek et al . , 2003 )
@xmath13hr later .
optical observations from ground based telescopes soon followed , and a bright optical afterglow with r=12.6 was detected at @xmath14hr ( peterson & price 2003 , torii 2003 ) .
observations at many different wavelengths followed , and a very bright afterglow emission was detected in the radio , sub - millimeter , millimeter and x - ray ( marshall & swank 2003 , berger , soderberg & frail 2003 , hoge et al .
2003 , kuno , sato & nakanishi 2003 ) .
its redshift , @xmath0 ( greiner et al . 2003 ) , places it as the closest grb whose redshift was firmly established . ] .
it is therefore not surprising that the afterglow of grb 030329 is by far the brightest afterglow detected so far , and is also the most well studied grb afterglow to date .
a spectral signature very similar to sn 1998bw began to emerge in the spectrum of grb 030329 after @xmath15days ( stanek et al . 2003 ) , and became more prominent in the following days .
this provides the best evidence so far for the association of grbs with core collapse supernovae of the type of sn 1998bw , which are also called hypernovae .
the optical light curve showed a jet break at @xmath16days , and several re - brightenings between @xmath17 and @xmath18days , each with a duration of @xmath19days .
we explore different possible explanations for the variability in the light curve using different physical mechanisms and show that the most likely explanation for the observed variability is refreshed shocks .
the optical light curve of grb 030329 steepened from @xmath20 to @xmath21 at @xmath22 days ( http://space.mit.edu/hete/bursts/grb030329/ ) .
the most likely explanation for this smooth change in the temporal decay index @xmath23 is a jet break , and in the following we shall assume that this is indeed the case . since the jet break is quite sharp , a roughly constant ( underlying ) external density is required .
a stellar wind profile would not produce a sufficiently sharp jet break ( kumar & panaitescu 2000 ) .
the fluence grb 030329 in the @xmath24kev range was measured by konus - wins to be @xmath25 ( golenetskii et al . 2003 ) , which for its redshift of @xmath0 and using the wmap cosmological parameters , implies an isotropic equivalent @xmath4-ray energy output of @xmath26erg .
the x - ray flux in the @xmath27kev range was measured by rxte at @xmath28hr to be @xmath29 , which implies an isotropic equivalent x - ray luminosity at @xmath30hr of @xmath31 ( marshall & swank 2003 ) . using the same formula and the fiducial parameter values of frail et al .
( 2001 ) we find @xmath32 which corresponds to a beaming factor of @xmath33 .
this implies a true @xmath4-ray output of @xmath34erg , and a true x - ray luminosity at @xmath30hr of @xmath35 .
both values are at the very low end of the standard distribution for grbs : frail et al .
( 2001 ) find @xmath36)=50.7\pm 0.5 $ ] , while berger , kulkarni & frail find @xmath37 .
the optical light curve shows a well monitored re - brightening between @xmath38 and @xmath39days .
this re - brightening begins to effect the light curve after about one day , which is only a factor of @xmath40 in time after @xmath41 .
therefore , we expect that the asymptotic decay slope @xmath42 has not yet been reached , and the value of @xmath43 inferred from observations may be regarded as a lower limit .
at least 3 additional and similar features appear in the light curve at @xmath44days , @xmath45days and t@xmath46days , though the latter are not as densely monitored ( see figure 1 ) .
another possible flattening of the light curve is observed at @xmath47days , however this may be at least in part due to the contribution of an the underlying sn 2003dh , and is therefore not included in our analysis . if the amplitude of each bump is measured as the offset between the decay epochs before and after the rise in the flux ( which is roughly constant , as the temporal decay indices @xmath23 after the different bumps are rather similar ) , we find that the amplitude of the first bump is a factor of @xmath40 , while the amplitude of the following bumps is @xmath48 .
the exact determination of the amplitude depends on the assumption for the underlying average " flux , and may vary somewhat .
variations in the external density was the first mechanism proposed to account for variability in the light curve of grb afterglows .
the variability due to small amplitude density fluctuations caused by interstellar turbulence was considered by wang & loeb ( 2000 ) . a density bump in the external medium
was suggested as a possible explanation for the bump in the light curve of grb 000301c ( berger et al .
2000 ) , and a highly variable external medium was suggested as an explanation for the large variability in the afterglow of grb 021004 ( lazzati et al .
2002 ; nakar , piran & granot 2003 ; heyl & perna 2003 ) .
a variable external density produces a much larger variability below the cooling break frequency @xmath49 ( typically optical or ir ) , compared to above @xmath49 ( typically x - ray ) . for a spherically symmetric external density distribution ,
it can not produce sharp variations in the flux on time scales @xmath50 ( nakar & piran 2003 ) , due to angular spreading ( i.e. the spreading in the arrival time of photons to the observer due to the curvature of the shock ) . turning to grb 030329
, we find that the rise in the normalization of the decaying part of the light curve after each bump implies that the external density would have to increase with radius , and to do so in discrete steps , rather than in a smooth manner .
however , such a density profile seems very unlikely .
if such a density profile would still somehow occur , then the observed optical light curve could be reproduced , if the jet would hardly expand sideways at @xmath51 . in this case , we expect much smaller fluctuations in the x - ray light curve .
as of yet , there is no good x - ray light curve available that would enable us to compare the variability in the light curve below and above @xmath49 . the patchy shell model ( kumar & piran 2000b ) suggests a variability in the energy per unit solid angle of the grb outflow , on some typical angular scale .
according to this model , the amplitude of the fluctuations in the afterglow light curve is expected to decrease with time @xmath52 ( nakar , piran & granot 2003 ) .
however , in grb 030329 the bumps occur after the jet break , when all the jet is visible , so that fluctuations in the energy per unit solid angle across the jet should no longer induce significant fluctuations in the light curve . therefore , assuming that the smooth steepening in the light curve at @xmath16 days is indeed a jet break , the patchy shell model can be ruled out as the source of the variability at @xmath53day .
according to the refreshed shocks scenario , slow moving shells that were ejected from the source with a modest lorentz factor catch up with the afterglow shock at late times well after the internal shocks that are responsible for the prompt @xmath4-ray emission have ceased ( rees & mszros 1998 ; kumar & piran 2000a ; sari & mszros 2000 ) .
each such shell injects energy into the afterglow shock as it collides with it from behind , and causes a re - brightening in the afterglow light curve . after the re - brightening the same power law decay as before the bump should be resumed , as long as the observed frequency remains within the same power law segment of the spectrum ( i.e. if no break frequency passes through the observed band ) .
similarly , under these conditions the color indices should not change .
the normalization of the decaying light curve after the bump is higher than before the bump by a factor @xmath54 , due to the increase in the energy of the afterglow shock .
the energy increases by a factor @xmath55 for @xmath56 and @xmath57 for @xmath58
. for grb 030329 @xmath59 and @xmath60 is around the optical , so that the increase in the energy is approximately linear in @xmath54 .
we find that across all the bumps in the first several days , this normalization increased by a factor of @xmath61 , and deduce that this implies a similar increase in the energy of the afterglow shocks due to energy injection from the shells that caught up with it . the conservative estimate of @xmath62 using the value of @xmath42 from just after @xmath41 is @xmath63 , while if we take @xmath42 from the decay after the second or third bumps we obtain @xmath64 ( see figure 1 ) .
therefore @xmath61 with an uncertainty of a factor @xmath40 seems like a reasonable estimate . in the original refreshed shocks scenario it was expected that these shocks occur before the jet spreading , and therefore @xmath65 ( kumar & piran 2000b )
. however , in grb 030329 the jet break is at @xmath66days , and the refreshed shocks took place at a later time , after the whole jet becomes visible as its lorentz factor is smaller than the inverse of its opening angle .
we expect that all the shells ejected from the central source will have roughly the same opening angle , @xmath67 ( see illustration in figure 2 ) .
the most forward shell sweeps up the ambient medium and produces the afterglow emission . since it is very hot , with an internal energy much higher than its rest energy , it could potentially expand sideways at close to the speed of light @xmath68 ( or @xmath69 ) in its local rest frame , though the lateral expansion could still be much smaller than this .
however , the slower shells that follow in the wake of the forward shell encounter very little material before they catch up with the forward shell , and therefore are expected to be cold , and should not expand sideways significantly before colliding with the afterglow shock . in this case
the duration of the bump in the light curve is given by the angular time @xmath70 , where @xmath71 if the lateral spreading of the jet is negligible , while @xmath72 if the lateral spreading is at the local sound speed .
this allows @xmath73 .
numerical simulations suggest that the lateral spreading is quite modest ( granot et al .
2001 ) , implying @xmath74 .
this is consistent with the almost constant duration of the first 3 bumps , and the slightly larger duration of the fourth bump , that are seen in grb 030329 . the temporal decay slope after the first bump is very close to @xmath42 , while after the second and third bumps @xmath23 is slightly larger than @xmath42 . unfortunately , it is hard to accurately determine the value of @xmath23 after the forth bump due to the contribution from the underlying sn , though it seems similar to the values for the preceding bumps .
the steeper temporal decay slopes , @xmath75 , may be due to the fact that the asymptotic power law decay after the jet break was not reached , and is underestimated by @xmath42 .
the small differences between the slopes may be attributed to a slight inhomogeneous distribution of the lorentz factor across the slow shell that catches up ( i.e. its lorentz factor decreases smoothly from its front edge to its back edge , so that the energy injection is not step - like but more gradual ) . we conclude that the refreshed shocks model provides the best explanation for the variability in the light curve of grb 030329 .
the timing of the bumps suggest that the lorentz factors of the slower shells ranged from @xmath76 to @xmath77 .
the lower limit is obtained for the extreme case of when after the jet break the jet expands sideways at the the local speed of sound and its radius is almost constant ( rhoads 1997 ) .
the higher values are obtained if the jet does not spread sideways significantly , as suggested by the results of granot et al .
in this case the radius of the front shell increases with time and the inner shells need a larger lorentz factor to catch up with the forward one .
we have compared the variable light curve of grb 030329 to the predictions of the different models for variability in grb afterglows and find that the only satisfactory explanation for the observed variability in grb 030329 is refreshed shocks that take place after the jet break .
this interpretation is supported by the almost constant durations of the different bumps , @xmath78days , which is also of the order of the jet break time , @xmath79days , and by the step - wise shape of the light curve . according to the refreshed shocks interpretation
, we find that the energy injected into the afterglow shock by all the slow shells that catch up with it , increases its initial energy by a factor of @xmath8 .
this implies that the total energy in this grb outflow is an order of magnitude larger than the energy inferred from the prompt @xmath4-ray emission , or from the early afterglow emission before @xmath80day , and thus explains the anomolously low values inferred for @xmath81 and @xmath82 .
a direct prediction of this interpretation is that we should expect very significant radio flares corresponding to the observed optical bumps .
these should arise from emission by the reverse shocks that form in the refreshed shocks .
this research was supported in part by funds for natural sciences at the institute for advanced study ( jg ) , by a grant from the israel space agency ( en and tp ) and by a grant from the horowitz foundation ( en ) . | grb 030329 is unique in many aspects .
it has a very low redshift for a grb , @xmath0 , and is therefore very bright and easy to monitor , making it the most well studied afterglow to date .
it shows a supernova bump in the light curve , with a spectrum very similar to sn 1998bw , thus establishing with much better confidence the connection between grbs and core collapse sne .
there are also two important physical characteristics that make this burst especially interesting , aside from its remarkably low redshift .
first , unlike most grb afterglows , the light curve of grb 030329 shows a very large variability a few days after the burst .
these fluctuations show a roughly constact amplitude , and a constant duration @xmath1 , while @xmath2 decreases with time @xmath3 .
second , its @xmath4-ray energy output and x - ray luminosity at @xmath5hr are a factor of @xmath6 and @xmath7 , respectively , below the average value around which most grbs are narrowly clustered .
we consider several interpretations for the variability in the light curve , in the context of different physical mechanisms , and find that the most likely cause is refreshed shocks , i.e. slow shells that are ejected from the source and catch up with the afterglow shock at late times . in grb 030329
this happens after the jet break , which implies an approximately constant duration @xmath1 of the bumps , in agreement with the observations .
this interpretation also explains the anomalously low initial energy of this burst , as the total energy of the afterglow shock is increased by a factor of @xmath8 due to the refreshed shocks , thus bringing the total energy output close to the average value for all grbs . |
discrete - time quantum walks have been introduced by aharanov , davidovich , and zagury@xcite as the quantum version of classical random walks .
quantum walks on lattices can spread out ballistically , in contrast with the diffusive behavior of classical random walks .
this characteristic has motivated many studies pursuing quantum algorithms that are faster than their classical counterparts@xcite .
discrete - time quantum walks have been investigated previously on many graphs .
the most studied graph is the one - dimensional line@xcite .
quantum walks have been analyzed on two - dimensional square lattices@xcite , and on the hypercube@xcite .
a spatial search using the discrete - time quantum walk model has been undertaken on the sierpinski gasket@xcite , and on the hanoi network of degree 3.@xcite a quantum walk on the dual sierpinski gasket using the continuous - time quantum walk model has been analyzed by agliari _ _ et .
al.__@xcite in this paper we focus our attention on discrete - time quantum walks on the sierpinski gasket .
we analyze the dynamics based on the standard evolution operator @xmath0 , where @xmath1 is the flip - flop shift operator , @xmath2 is the coin , and @xmath3 is the identity operator . throughout , we are using the grover coin .
the main physical quantities that we analyze are the mean - square displacement in form of the standard deviation in position , the limiting probability distribution , and the mixing time .
the results are compared with classical random walks on the sierpinski gasket and with quantum walks on other graphs , such as the square lattice .
the paper is organized as follows : in section [ sec : st ] we derive the evolution equation for quantum walks on sierpinski gaskets . in section [ sec : physq ] we present the numerical results for the standard deviation , the limiting probability distribution , and the mixing time . in the last section , we present our conclusions .
the sierpinski gasket of generation @xmath4 is a degree-4 regular graph , an example of which is depicted in fig .
[ fig : sierpinki ] for @xmath5 .
it has @xmath6 nodes and a fractal or hausdorff dimension of @xmath7 , which is larger than for the line and smaller than for the plane .
( 5.35686,4.6 ) ( 0,0)(8,4 ) ( 1.414214,0)(0.000000,0 ) ( 1.414214,0)(0.707107,1 ) ( 1.414214,0)(2.121320,1 ) ( 1.414214,0)(2.828427,0 ) ( 2.828427,0)(1.414214,0 ) ( 2.828427,0)(2.121320,1 ) ( 2.828427,0)(3.535534,1 ) ( 2.828427,0)(4.242641,0 ) ( 4.242641,0)(2.828427,0 ) ( 4.242641,0)(3.535534,1 ) ( 4.242641,0)(4.949747,1 ) ( 4.242641,0)(5.656854,0 ) ( 0.707107,1)(0.000000,0 ) ( 0.707107,1)(1.414214,0 ) ( 0.707107,1)(2.121320,1 ) ( 0.707107,1)(1.414214,2 ) ( 2.121320,1)(2.828427,0 ) ( 2.121320,1)(1.414214,0 ) ( 2.121320,1)(0.707107,1 ) ( 2.121320,1)(1.414214,2 ) ( 3.535534,1)(2.828427,0 ) ( 3.535534,1)(4.242641,0 ) ( 3.535534,1)(4.949747,1 ) ( 3.535534,1)(4.242641,2 ) ( 4.949747,1)(5.656854,0 ) ( 4.949747,1)(4.242641,0 ) ( 4.949747,1)(3.535534,1 ) ( 4.949747,1)(4.242641,2 ) ( 1.414214,2)(0.707107,1 ) ( 1.414214,2)(2.121320,1 ) ( 1.414214,2)(2.828427,2 ) ( 1.414214,2)(2.121320,3 ) ( 2.828427,2)(1.414214,2 ) ( 2.828427,2)(2.121320,3 ) ( 2.828427,2)(3.535534,3 ) ( 2.828427,2)(4.242641,2 ) ( 4.242641,2)(4.949747,1 ) ( 4.242641,2)(3.535534,1 ) ( 4.242641,2)(2.828427,2 ) ( 4.242641,2)(3.535534,3 ) ( 2.121320,3)(1.414214,2 ) ( 2.121320,3)(2.828427,2 ) ( 2.121320,3)(3.535534,3 ) ( 2.121320,3)(2.828427,4 ) ( 3.535534,3)(4.242641,2 ) ( 3.535534,3)(2.828427,2 ) ( 3.535534,3)(2.121320,3 ) ( 3.535534,3)(2.828427,4 ) ( -0.5,0)(-0.5,2 ) ( 0,-0.5)(2,-0.5 ) ( -0.7,1)@xmath8 ( 1,-0.7)@xmath9 a coined quantum walk on the sierpinski gasket embedded in the two - dimensional plane has a hilbert space @xmath10 , where @xmath11 is the @xmath12-dimensional coin subspace and @xmath13 is the @xmath14-dimensional position subspace .
@xmath13 is spanned by vectors of type @xmath15 with integers @xmath16 and @xmath17 restricted to be on the gasket , as shown in fig .
[ fig : sierpinki ] .
due to the embedding , we use the computational basis @xmath18 for the coin space @xmath11 , but only four of these basis vectors are utilized for each vertex . the shift operator for the internal vertices is @xmath19 where @xmath20 is the inverse of @xmath21 modulo 6 .
the coin value is inverted after the shift ( flip - flop shift ) .
functions @xmath22 and @xmath4 are defined in the table [ tab : f_g ] .
.auxiliary functions for the shift operator . [ cols="^,^,^,^,^,^,^",options="header " , ] [ tab : f_g ] for the external vertices @xmath23 , @xmath24 , @xmath25 , where @xmath4 is the generation level , we consider two cases of boundary conditions : ( 1 ) periodic and ( 2 ) reflective . in case ( 1 ) , the action of the shift operator is given by @xmath26 those special cases can be implemented through functions @xmath22 and @xmath4 and , in this case , @xmath22 and @xmath4 will depend on the location @xmath27 . in case
( 2 ) , the action of the shift operator is given by @xmath28 the grover coin is defined as @xmath29 where @xmath30 . its matrix representation is @xmath31 the generic state of the walker at time @xmath32 is described by @xmath33 where the coefficients @xmath34 are complex functions that obey the normalization condition @xmath35 for all time @xmath32 .
applying the evolution operator @xmath36 to the generic state , we obtain @xmath37 renaming the dummy indices , we obtain @xmath38 expanding the left hand side of the above equation in the computational basis and equating like coefficients , we obtain the evolution equation for the quantum walk , @xmath39 the matrix @xmath40 depends on @xmath41 , since there are six types of vertices that are distinct in their orientation . for each one
, we have to use the correct labels for their edges .
we use eq .
( [ pq_eq_evol_x_y ] ) to numerically simulate the evolution of the quantum walk using initial conditions of the form @xmath42 , where @xmath30 is the uniform vector in the coin space .
note that @xmath43 is not biased .
the same is true for the grover coin @xmath44 .
this coin and the flip - flop shift operator play an important role in spatial search algorithms@xcite .
in this section , we analyze the behavior of quantum walks on the sierpinski gasket with the focus on the diffusion processes .
the main physical quantities that we analyze are the mean - square displacement in form of the standard deviation in position , and the mixing time .
the physical quantities that we will analyze are defined using the probability distribution over the vertices of the graph .
it is one of the main physical quantities that is available in the analysis of the behavior of quantum walks .
the probability distribution is given by @xmath45 the position standard deviation @xmath46 is defined as @xmath47 where @xmath48 and @xmath49 for a walker that starts located on a specific vertex , the standard deviation at intermediate times @xmath50 increases as a power - law@xmath51 which defines the diffusion exponent @xmath52 , in analogy to a classical walk . on a finite system , the walk eventually reaches the farthest vertex which cuts off the growth in @xmath53 at some time @xmath54 , beyond that it oscillates around an average value . in order to analyze the diffusion process , we are interested in the asymptotic behavior of the power law regime for the infinite system , @xmath55 . therefore , on any finite sierpinski gasket , we bound the evolution time to be smaller than the time that the walker takes to reach the farthest vertex . to measure displacement ( standard deviation )
, we use reflective boundary conditions and the flip - flip shift operator .
the first result we have obtained from the simulations , which is strikingly different from the behavior of quantum walks on lattices , is that the scale of the standard deviation depends on the initial vertex .
for example , for generation @xmath56 , the fastest growth in the displacement is obtained when the walker starts on vertex @xmath57 fig .
[ fig : standev1 ] shows the standard deviation as function of the number of steps of a quantum walk with the initial state @xmath58 .
our fit yields @xmath59 , i.e. , @xmath60 , significantly faster than classical diffusion ( @xmath61 ) but still considerably slower than the standard deviation for quantum walks on a square lattice , for which @xmath62 .
the slowest - growing standard deviation of a flip - flop quantum walk on the sierpinski gasket of generation @xmath56 is obtained for the initial state @xmath63 . the best fit in this case using the data in fig .
[ fig : standev1 ] is @xmath64 ( or @xmath65 ) , which shows a very small spreading rate characterizing a sub - diffusive process .
since the displacement apparently depends on the initial vertex , it is interesting to define a mean standard deviation @xmath66 as function of time in the following way @xmath67 where sub - indices @xmath41 of @xmath68 indicate the initial location used to obtain the standard deviation . from each location
the walker evolves from an initially uniform state in coin space .
[ fig : standev1 ] depicts the behavior of @xmath66 .
the numerical results suggest a best fit of @xmath69 , or @xmath70 .
average over all initial locations makes this scaling exponent @xmath71 a characteristic of the sierpinski gasket of generation @xmath56 , which happens to be remarkably close to the result for classical diffusion , @xmath61 .
the histogram in fig .
[ fig : histogram ] shows the number of such initial conditions that have the same fitted exponent , @xmath71 , in eq .
( [ msd ] ) .
for example , for @xmath72 there are around 450 vertices that can be used as initial condition to obtain the same scaling .
the range in scaling ( for @xmath56 ) extends from @xmath60 to 3.45 , although the bulk of the distribution is centered very close to the result for classical diffusion , marked by a vertical line .
the width of this distribution is only about 17% of the mean , and it would be interesting to see whether the width narrows further for increasing system sizes @xmath55 . with the initial state @xmath42 , @xmath73 .
a vertical line indicates the corresponding result for classical diffusion .
, width=226 ] the average probability distribution is given by @xmath74 note that @xmath75 is a probability distribution for all @xmath76 , because @xmath77 the interpretation of @xmath75 uses projective measurements , therefore @xmath78 evolves stochastically , and converges to a limiting distribution when @xmath76 goes to infinity .
the definition of the limiting probability distribution is @xmath79 this limit exists and can be calculated explicitly if the expressions for the eigenvalues of the evolution operator are known .
the limiting distribution depends on the initial condition in general .
[ fig : limitingdist1 ] shows the limiting distribution as a function of position @xmath9 of a flip - flop quantum walk that departs from the central bottom vertex @xmath80 with periodic boundary conditions .
the probabilities in @xmath8-direction have been added up to generate a one - dimensional plot . in general ,
the numerical simulations show that the limiting distribution for the sierpinski gasket depends on the initial condition and is highly concentrated around the initial vertex .
note that a walker encounters frequent bottlenecks that inhibit spreading .
if the walker starts in vertex @xmath81 , there are only two passage points , @xmath82 and @xmath83 , towards the top of the sierpinski gasket ( in fig .
[ fig : sierpinki ] these points correspond to @xmath84 and @xmath85 ) .
a three - dimensional plot shows that the limiting probability is very close to zero when @xmath86 .
the average distribution @xmath75 converges to the limiting distribution @xmath87 .
this can be confirmed by the graph of the distance between these distributions as function of time .
the total variation distance between two probability distributions @xmath88 and @xmath89 is defined as @xmath90 fig .
[ fig : dtv ] shows the plot of @xmath91 as function of the number of steps represented by @xmath76 .
the best fit suggests that this distance scales as @xmath92 .
we note that it decays approximately as @xmath92 not only for the sierpinski gasket but also for two - dimensional lattices@xcite and hypercubes@xcite .
this numerical result can be explained analytically .
expanding the quantum walk state in the eigenbasis that diagonalizes the evolution operator , one can obtain an explicit expression for @xmath93 for any initial condition .
the expression has the form @xmath94 where @xmath95 is a difference between two non - equal eigenvalues , @xmath96 is a constant , and the inner sum is over all pairs of non - equal eigenvalues . some results of aharonov @xcite help to obtain that analytical expression .
the modulus of the term @xmath97 in the numerator of @xmath93 is a bounded oscillatory function .
so , the distance between the average and the limiting distribution scales as @xmath92 in general and oscillates around the curve @xmath92 , confirming the data shown in fig .
[ fig : dtv ] .
the mixing time @xmath98 is defined as @xmath99 which can be interpreted as the smallest number of steps such that the distance between the average distribution and the limiting distribution becomes permanently smaller than @xmath100 . if @xmath91 obeys an inverse power law as a function of time , then @xmath101 obeys an inverse power law as a function of @xmath100 .
the mixing time depends on the initial condition in general and on the size @xmath14 of the graph .
we have generated the same kind of data of fig .
[ fig : dtv ] for sierpinski gasket of generation 6 up to 10 . using the best - fitting curves we can estimate @xmath98 as a function of @xmath14 . fig .
[ fig : mixingtime ] shows that @xmath98 has a power law in terms of the number of vertices .
the data allows us to estimate that @xmath102 when we take @xmath103 as initial state . .
the best fit is @xmath104 .
the first point corresponds to @xmath105.,width=264 ]
we have analyzed the flip - flop discrete - time quantum walk on the sierpinski gasket of finite generation embedded in the two - dimensional plane using reflective and periodic boundary conditions .
our investigations focus on the following physical quantities : the position standard deviation ( with reflective boundary conditions ) and the mixing time ( with periodic boundary conditions ) . performing numerical simulations on sierpinski gaskets up to generation @xmath106
, we have obtained the scaling exponent of the standard deviation as function of the number of steps and the scale of the mixing time as function of the number of vertices .
for the system sizes studied , the results depend significantly on the initial condition .
as fractal lattices lack translational invariance , quantum interference effects likely vary strongly with the initial location . a characteristic way to assign a distinct diffusion exponent @xmath71 to the sierpinski gasket is provided by performing an average over all initial locations . in this case
, we obtain an average exponent @xmath107 that is remarkably close to the result for classical diffusion , @xmath61 , on this system .
therefore , a quantum walk on the sierpinski gasket spreads slower than on a square lattices .
the limiting distribution for the sierpinski gasket depends on the initial condition and is concentrated around the initial vertex .
this happens with other graphs such as the two - dimensional lattice@xcite .
the scaling of the mixing time for the sierpinski gasket @xmath108 is close to the scaling of mixing time for the two - dimensional lattice , which is believed@xcite to be @xmath109 .
our data is not precise enough to determine the presence of a term that depends on @xmath110 .
the result differs from the scaling on the cycle which is believed@xcite to be @xmath111 .
we acknowledge financial support from cnpq .
sb is grateful for the support and hospitality of lncc during this project . | we analyze discrete - time quantum walks on sierpinski gaskets using a flip - flop shift operator with the grover coin .
we obtain the scaling of two important physical quantities : the mean - square displacement and the mixing time as function of the number of points .
the sierpinski gasket is a fractal that lacks translational invariance and the results differ from those described in the literature for ordinary lattices .
we find that the displacement varies with the initial location .
averaged over all initial locations , our simulation obtain an exponent very similar to classical diffusion . |
the economic system is no doubt a many - particle system - it can be viewed as a collection of numerous interaction agents .
so it is possible that methods and concepts developed in the study of strongly fluctuation systems might yield new results in this area .
in fact , in the past decades the approaches from statistical physics have been applied in economics and a lot of interesting results , including empirical laws and theoretical models , have been achieved ( see @xcite and @xcite as a review of econophysics ) . among all these studies , a great deal of researchers is on the agent - based modelling and related non - trivial self - organizing phenomena . in the economic system ,
the agents learn from each other , and their activities may be influenced by others actions .
these interactions between agents may be simple and local , but they may have important consequence related with the emergence of global structure . to understand the mechanism behind these innovation phenomena , the methods and concepts in phase transitions and critical phenomena are helpful .
for instance , in the study of majority and minority game@xcite , opinion formation@xcite , and computational ecosystems@xcite . in this paper , we focus on the formation of labor division . roughly speaking
, an economic organizational pattern is said to involve the division of labor if it allocates the labor of different individuals to different activities .
hence the specialization of individuals and the number of professional activities are two sides of division of labor@xcite .
it is a common functional organization observed in many complex systems , and it is a fundamental way to improve efficiency and utilization so as to get global optimization for the system . in order to investigate the mechanism behind the formation of labor division
, we have constructed a simple model with many interacting agents .
every agent has only two kind of tasks , namely @xmath0 and @xmath1 we describe the level of specialization of agent by his working - time share spent on producing @xmath0 or @xmath2 . each agent make their decisions for working - time in different tasks , and receive payoffs according to their and other agents choices .
the agents can adapt by evaluating the performance of their strategies from past experience so as to get maximum returns .
the returns for any agent is determined by its production with endogenous technical progress - through the mechanism of learning by doing , and its cooperation with other agents . just like the hamiltonian in statistical physics
, the payoff function determines the behavior of the agent in economic system . because of the bounded rationality and incompleteness information in the system , we have introduced a parameter , named social temperature @xmath3 ( in the model , we have absorbed the @xmath4 into the other parameters .
such an approach is traditionally used in statistical physics , for example , let @xmath5 , so the new @xmath6 in hamiltonian of ising model means @xmath7 actually . ) , to describe the degree of randomness in decision - making .
then we assume that the system should obey the canonical ensemble distribution , that is the probability @xmath8 of a microstate @xmath9 is proportional to its ` boltzmann factor ' determined by the total returns of the microstate @xmath9 : @xmath10 then using the metropolis simulation method@xcite , we can get a master equation to investigate the evolution of the system . with the continual change of system parameters , we have found a so called ` social phase transitions ' phenomenon related to the emergence of labor division .
the model is defined in section 2 .
the economies of specialization is introduced by increasing returns from learning by doing . and the economies of complementarity is described by an additional payoffs from the combination of two products .
section 3 gives the numerical results .
the effects of parameters on critical point are discussed in detail .
it is revealed that although the technical progress is a key factor that determine whether or not the labor division will happen , the competitive cooperation among agents has very important effects on the critical point .
our results are summarized in section 4 .
let s consider a system consists of @xmath11 homogeneous agents . in a given time period @xmath3 , any agent has two different kind of tasks , namely @xmath0 and @xmath2 . in any time unit , the fraction of working time on task @xmath0 for @xmath12 agent is denoted by @xmath13 $ ] .
we are interested in the long - term ( @xmath14 ) evolution of the system , especially the emerging of labor division .
@xmath15 is a real number in @xmath16 $ ] , which describes the working pattern for the @xmath12 agent .
if @xmath15 equals @xmath17 or @xmath18 , the agent is full specialized in @xmath0 or @xmath2 and we call the system is in complete division of labor .
if @xmath15 equals @xmath19 , it means that the agent spends same time interval on tasks @xmath0 and @xmath2 , and we call it as the time - dividing working mode in the following discussion . usually @xmath15 could be any real number between @xmath18 and @xmath17 , in this case the agent does not have any preference on job @xmath0 or @xmath1we focus on the global behavior of the system . based on the above description of the agent , we introduce the following two order parameters to describe the behavior of the system on macroscopic level : @xmath20@xmath21where @xmath22 describes the intensity of labor division and cooperation in the system .
it has three special values , @xmath23 , represent time - dividing working mode , every @xmath15 equals @xmath19 , no - preference working mode , @xmath15 distributed in ( 0,1 ) randomly , and full specialization , every @xmath15 equals @xmath18 or @xmath17 , respectively .
@xmath24 gives the agents allocation on tasks @xmath0 and @xmath2 on global ( in average on macroscopic level ) . in the following discussion
, we try to specify a suitable evolution rule for @xmath25 based on the similar approach in statistical physics .
so we can then determine the dynamical equations for @xmath26 , and get the final steady states for @xmath27 and @xmath28 .
analogous with ising model in statistical physics , @xmath15 could be treated as the spin in @xmath12 point of the lattice , and then @xmath24 is the magnetization .
so if we can give the hamiltonian of the system , the evolution of the system could be determined . in the economic system ,
the agent switches his behavior according to his evaluations on the returns from production .
every agent try to get maximum return just as any practical tends to stay in the state with lowest energy .
so the payoff function from the agent s production should have the same effect as the hamiltonian . the working mode for each agent
should be determined by its returns .
according to the previous literatures on specialization and economic organization , we know that there are several factors that related to the division of labor , such as the increasing returns to specialization and transaction costs
. in fact , limited ability on learning and incomplete information on the technology will lead to the labor division directly .
but we do nt take them into account in this model . in our model ,
every agent know the technology for producing @xmath0 and @xmath2 . and for any agent
there is not any comparative advantages for producing @xmath0 or @xmath2 in the initial .
the technical progress is achieved by the mechanism of learning by doing without any cost .
the payoff function for agent @xmath29 is given by the following formulas : @xmath30where @xmath31@xmath32@xmath33 in the functions above , @xmath34 and @xmath35 are the technology for producing @xmath0 and @xmath2 respectively .
for simplicity and without losing any generality , we assume that unit production @xmath0 will get unit return .
and @xmath36 is the return of unit production @xmath2 .
@xmath37 gives the additional benefits for composite @xmath0 and @xmath2 into a final product .
so it is related to the less one between @xmath0 and @xmath2 .
if agent @xmath29 has some more single product after the composition himself , he can also get another return , named @xmath38 , from the cooperation with other agents .
but the return got from the composition with the product from other agent is factored by @xmath39 . because of the incomplete information in the system , any agent could nt know the situation of production on global . so in order to get the corresponding returns @xmath40 for every agent , we used the average - field approach similar with that in statistical physics .
that means , for every agent , he will match his product with all the other agents and then the result is the average .
this assumption indicate that there are already a public market in economics . in the function for @xmath38 , @xmath41 gives the surplus product @xmath2 for agent @xmath29 , and the sum of @xmath42 gives the surplus product @xmath0 of the all other agents .
when these two terms have different signs , their product is negative and @xmath38 , the corresponding payoff from the cooperation , will be also negative . in this case , the surplus product of the agent is the same kind as the final surplus of all the other agents . because of the transaction cost and the diseconomies of incomplementarity , it is rationale that the agent would get negative payoff . the way for technical progress is learning by doing .
this mechanism is related to the accumulation of agent s historical production behavior .
that means the development of comparative advantages is determined by related working time .
let @xmath43 denote the integral working time on @xmath0 or @xmath44 for agent @xmath29 : @xmath45 we have introduced two mechanisms for technical progress , named @xmath46-mechanism and @xmath47-mechanism .
they are given by following functions : @xmath48 @xmath49 these two mechanisms will give different results .
equation([gamma ] ) is an exponential function which gives unlimited growth on technology , while there is an upper limit in equation([mu ] ) .
figure([g_m ] ) shows three functions for @xmath50 in our computer simulations .
used in the simulations .
parameters are labelled in the graph . ] in the production function discussed above , term @xmath40 is the return from the agents cooperation .
it introduces interactions among all agents ( the same as the interactions among spins ) .
so if we discuss the dynamics in @xmath47-space , we should construct @xmath11 evolutionary equations for @xmath15 .
but they are difficult for both theoretical analysis and computer simulation . in the following discussion
, we try to describe the evolution of the system in @xmath51-space .
that is to discuss the evolutionary behavior of the joint density function @xmath52 in the space sponsored by @xmath53 .
then the total production for the system is : @xmath54where @xmath55 is the production of agent @xmath29 given by eq(single ) .
the same as the approach of critical dynamics in ising model , the master equation for the joint density function @xmath56 in @xmath51 space is @xmath57% \end{array } \label{master}\]]where @xmath58 is the transition probability which is determined by the boltzman factors as the following : @xmath59 in the above transition probabilities , @xmath60 is the change of returns related to the variation of working mode for @xmath12 agent . from eq.([total ] ) we can get @xmath61% \end{array}%\]]with the master equation([master ] ) and certain initial conditions , we can get evolving behaviors of @xmath22 and @xmath24 . then we can discuss the phenomena on labor division .
it is difficult to have some theoretical results .
so in the next section we will show some numerical simulation results by monte carlo simulation method .
based on the master equation([master ] ) and corresponding transition probability([jump ] ) , the simulation is proceed on the following metropolis algorithm : 1 .
for any given state with @xmath15 for agent @xmath29 , a new state @xmath62 is randomly selected ; 2 . if @xmath63 , then the transition from @xmath15 to @xmath62 is proceeded ; 3 . if @xmath64 , then a random number @xmath65 is selected .
if @xmath66 , the transition from @xmath15 to @xmath62 is proceeded or else the agent @xmath29 keeps the original state .
4 . for another agent @xmath67 ,
goto step 1 . for any given initial state ,
the system will achieve a certain steady state after some transient process .
figure([lambda - t ] ) gives a typical evolution behavior under @xmath68 mechanism .
@xmath69 , @xmath70 . ]
there are several parameters related to the mechanism of labor division in the model
. we will show the simple or maybe trivial results of @xmath46-mechanism first and then emphasize our discussion on the results of @xmath47-mechanism .
and we let @xmath71 in all the simulations before we discuss the effects of parameters @xmath72 and @xmath11 on the system evolution in the end . as shown in([ggamma ] ) because of @xmath46-mechanism is described by an exponential function , the system will achieve the state of labor division for any given initial state and @xmath74 no matter how small is @xmath75 .
that means if the intensity of technical progress is big enough , the agent would definitely specialized in a kind of task .
other factors such as competitive cooperation described by parameter @xmath76 have no effects on the long term evolution . in the final stationary state as a function of parameter @xmath77 under @xmath77-mechanism .
the final steady state of the system is determined by @xmath78 or @xmath79 . ]
but for the model with @xmath47-mechanism , @xmath81 has some important effects on the system evolution . as shown in figure([betamu ] ) , there are phase transitions results in labor division and @xmath81 has strong effects on the critical point @xmath82 for parameter @xmath68 .
@xmath83 decreases as the @xmath84 increasing .
the @xmath68-mechanism for learning by doing gives a logistic growth for technology .
that is much more realistic than the @xmath85-mechanism .
the results reveal that the competitive cooperation among agents is very important for the emergence of labor division . however , as discussed in the end of this section , when @xmath86 , that is there is no mechanism of increasing returns , the system has no phase transition to labor division .
this result is rationale for the model and it is consistent with the theory on specialization .
the comparative advantages in production can only be introduced by a positive feedback caused by increasing returns .
so the mechanism of learning by doing is a dominant factor for labor division .
, parameter @xmath87 in the final stationary state as a function of parameter @xmath80 under @xmath80-mechanism .
there is phase transition in the system .
@xmath88 has strongly effects on the critical point . ] in order to investigate the effects of all terms in production function in detail , we have simulated some special cases .
the first case is @xmath89 .
then @xmath37 and @xmath40 in equation ( single ) are all zero and technical progress is the only one factor that affects the final results . as shown in figure([no_beta ] ) , there is a phase transition in the system . under the given conditions , the critical point for @xmath68 ( @xmath82 ) is around 50 .
-mechanism and without the term @xmath90 that is the payoff from the cooperation among agents , parameter @xmath91 in the final stationary state as a function of parameter @xmath80 for different parameter @xmath92 .
the critical point @xmath93 is much larger than that in fig .
4 and even in fig . 6 . ]
then we have simulated the model with @xmath37 but no @xmath94 .
the results are shown in ( [ no_eco ] ) .
when @xmath95 is greater than a critical value @xmath96 , the system will stabilized in the state with @xmath97 .
that is @xmath98 for any agent @xmath29 .
every agent does the tasks @xmath0 and @xmath2 himself and spends the same time on both tasks .
the system is in the time - dividing working mode . only when @xmath99 is smaller than @xmath96 , does the system have phase transition to labor division . but the critical value @xmath83 is much larger than that of in figure ( [ betamu ] ) .
it is even larger than the @xmath83 in figure ( [ no_beta ] ) .
so the term @xmath37 has the effect of anti - specialization . from the simulation results in figure ( [ betamu ] ) and
just as we have indicated in the beginning of this section , @xmath38 is helpful to labor division .
actually the term @xmath38 is the benefit for agent @xmath29 get from the cooperation with all the other agents .
it should has the same effect as the decreasing of transaction costs .
so it is not surprising that this term will enforce specialization .
-mechanism , parameter @xmath91 in the final stationary state as a function of parameter @xmath80 when there is only the term for technical progress in the production function .
the critical point @xmath93 is much larger than that in fig .
another interesting case is the system without technical progress , that is @xmath100 and @xmath35 are all equal 1 . in this case , the system is exactly the same as ising model .
let @xmath101 and @xmath102 , then the total returns for the system is @xmath103 where @xmath104 .
the distribution of the state for the system is determined by @xmath105 . comprising with ising model ,
let the corresponding hamiltonian is @xmath106and let @xmath107 , then the system could be described by canonical ensemble with @xmath108 .
that is an anti - ferromagnetic ising model with global interaction .
the similar approach shows that this system either doest have phase transition related to labor division .
the result is shown in figure([no_gamma ] ) . and @xmath109 as a function of parameter @xmath92 when there is no progress of productivity .
the system has no phase transition . ]
parameter @xmath111 reflects the preference for tasks @xmath0 and @xmath2 .
so it is expected that parameter @xmath72 , the ratio of returns for @xmath0 and @xmath2 , will affect final results on @xmath111 .
the diagram in phase space for @xmath112 when @xmath113 is shown in figure ( [ lambda_2 ] ) , in which we can see @xmath114 fluctuates around zero , and no phase transition with symmetry breaking emerges . and as shown in figure([lambda_a ] ) , @xmath115 indeed has some effects on @xmath111 .
but because of the combination mechanism described by parameter @xmath81 , @xmath115 has only a little effect on @xmath111 .
it also has some detailed effects on labor division described by @xmath116 , as shown in figure ( [ alpha5_1 ] ) .
but @xmath115 could not change the qualitative behavior of phase transition . in the final stationary state as a function of parameter @xmath80 for different @xmath92
when @xmath117 .
there is no symmetry breaking with the phase transition . ] in the final stationary state as a function of parameter @xmath110 .
because of the cooperation mechanism described by parameter @xmath92 , @xmath110 has only a little effect on the final distribution . ] here are other two phenomena that should be pay attention to , first , the model has no scale effect for limited @xmath11 , as shown in figure ( [ 500 - 1000 ] ) .
the phase curves for different @xmath11 are almost consistent .
it has only some effects on the accurate of simulation .
second , when @xmath118 in @xmath119-mechanism for technical progress , the productivity declines with the time evolution .
then the system could not reach the state of specialization .
this result is also meaningful for real economic system . in the final stationary state as a function of parameter @xmath80 for different @xmath110 . ]
@xmath120 , @xmath121 curves for different population size @xmath11 .
the curves are almost consistent . ]
a distinctive feature of the organization of a human society is the division of labor . from the classical economic theory[11 ] , the division of labor comes from the development of endogenous comparative advantages .
so there is the intrinsic relationship between technical progress and the evolution of the division of labor . but
how and why it emerges from the system consists of identical individuals ?
we have studied the formation of labor division by the approach of statistical physics .
the results reveal that there is a phase transition with this pattern formation .
although the progress of productivity dominated the phase transition occurs or not , the competitive cooperation among the agents has important effects on the critical point .
so the market formation and labor division are usually reinforced each other .
all the above results give us deep understanding to the evolution of labor division .
studying the economy as an evolving complex system , we can avoid the standard economic assumptions of equilibrium based on rational behavior by agents . and
the concepts and methods developed in statistical physics is helpful to uncover fundamental principles governing the evolution of complex adaptive systems .
so the approaches presented here have potential applications in a variety of economical , biological and financial problems .
this work was supported by the national science foundation under grant no.10175008 and national 973 program under grant no.g2000077307 .
we also thanks to all members of professor yang s group and faculties from department of system science for their warm discussion . | the emergence of labor division in multi - agent system is analyzed by the method of statistical physics . considering a system consists of n homogeneous agents .
their behaviors are determined by the returns from their production . using the metropolis method in statistical physics , which in this model can been regarded as a kind of uncertainty in decision making
, we constructed a master equation model to describe the evolution of the agents distribution .
when we introduce the mechanism of learning by doing to describe the effect of technical progress and a formula for the competitive cooperation , the model gives us the following interesting results : ( 1 ) as the results of long term evolution , the system can reach a steady state .
( 2 ) when the parameters exceed a critical point , the labor division emerges as the result of phase transition .
( 3 ) although the technical progress decides whether or not phase transition occurs , the critical point is strongly effected by the competitive cooperation . from the above physical model and the corresponding results , we can get a more deeply understanding about the labor division .
key words : division of labor , statistical physics , phase transition , metropolis method pacs:05.45.-a ; 87.23-n |
since the recent discovery of topological insulators@xcite , the search for topological transitions in condensed matter has become a priority task . in particular , the prediction and the observation of a new topological order , namely the @xmath0 topological order@xcite , has stimulated a great interest in the scientific community .
graphene , a two - dimensional carbon - based crystal , is a wonderful platform for investigating topological transitions .
it is a semi - metal , whose band structure of the @xmath1 orbitals consists in two bands that touch linearly at the fermi level at two inequivalent points of the brillouin zone@xcite .
this peculiar structure gives rise to massless dirac - like low energy excitations .
on the other hand , several ideas to induce topological states of matter by means of time periodic external potentials have been proposed@xcite , and recently observed in temporal modulated photonic crystals @xcite .
this approach renews considerably the possibilities of inducing new topological phases .
remarkably in graphene , a different topological transition between a dirac semi - metallic phase and an insulating phase , can also be achieved by merging the pair of dirac points ( pdp)@xcite .
the resulting insulating phase is not a @xmath0 topological phase , but may nevertheless host zero - energy edge states whose topological origin is well understood in terms of a one - dimensional bulk winding number , namely the zak phase@xcite .
the search for this transition has stimulated experimental efforts beyond the solid state community : the merging of the dirac points , as well as the emergence of zero - energy edge states have been recently observed in anisotropic traps of cold atoms@xcite and in microwave tight - binding analogue experiments of a honeycomb lattice@xcite .
however , the merging transition has not been observed yet in electronic systems . in particular
its achievement in graphene itself@xcite , by means of mechanical manipulations like stretching , is unfortunately out of reach , because the graphene sheet would be destroyed far before the expected transition@xcite .
honeycomb lattice irradiated by an ac electric field with arbitrary polarization .
the inset shows the merging of the dirac points induced by the external field . ] in the present work , we propose an alternative to mechanical distortions to achieve the merging transition in the honeycomb lattice .
in particular , we show that an ac electric field can drive the topological merging transition , or inversely induce the creation of new dirac points .
this mechanism is different to the ones usually studied in previous works dealing with floquet topological insulators and summarized in @xcite .
more precisely , we demonstrate that in the high frequency regime , the ac field acts similarly to a mechanical strain , and therefore allows for the manipulation of the dirac points , their annihilation and their creation in a controllable way .
furthermore , the ac field is able to induce the localization of the electrons in specific directions by aligning the dirac points , which marks another topological transition between two insulating or two semi - metallic phases .
our analysis , when restricted to linear polarization and high frequency regime , agrees with the results obtained in shaken optical lattices@xcite .
in addition , we also consider different field polarizations and finite frequency effects .
importantly , at lower frequencies , we find that the coupling between the floquet bands , that characterize the properties of the driven system , gives rise to out - of - equilibrium phases with no analogue in the static system in which multiple pdp emerge . our approach does not rely neither on low energy models nor on approximations valid just close to resonance ( as the rotating wave approximation)@xcite , but takes into account the full lattice model together with arbitrary field amplitudes and phase polarizations .
this allows us to keep track of the relative position between the dirac points and thus properly describe their annihilation or creation .
the hamiltonian of a @xmath2-periodic driven system fulfills @xmath3 , and its eigenvectors can then be written in the floquet form : @xmath4 ( @xmath5 ) , being @xmath6 the quasi - energy , and @xmath7 the floquet state@xcite .
this ansatz , consequence of the time translation invariance , maps the time - dependent schrdinger equation to the eigenvalue equation @xmath8 , in which the quasi - energies and the floquet states are the eigenvalues and eigenvectors of the floquet operator @xmath9 , respectively . in order to deal with the time dependence of the floquet operator @xmath10 , it is useful to introduce the sambe space , which consists in a composed hilbert space , where the basis states are time - independent@xcite . in this space ,
the quasi - energies are given by ( see @xcite and appendix a ) : @xmath11 where @xmath12 with @xmath13 , and where @xmath14 is the @xmath15 fourier component of the floquet state .
the coupling between the floquet side - bands is encoded in the fourier components @xmath16 of the hamiltonian . for a honeycomb lattice embedded in
a periodic time - dependent in - plane electric field @xmath17 ( see fig.[fig : schematic ] ) , the peierls substitution leads to a time dependent hamiltonian : @xmath18 where @xmath19 , @xmath20 being @xmath21 the nearest neighbor hopping parameter , @xmath22 are the basis vectors of the bravais lattice : @xmath23 , @xmath24 , and @xmath25 .
the vectors joining the nearest neighbors @xmath26 are given by @xmath27 , @xmath28 , @xmath29 .
we consider electric fields with arbitrary amplitude and polarization , and write the vector potential as @xmath30 .
the calculation of the fourier components @xmath31 can be performed analytically ( see appendix b ) and yields : @xmath32 with @xmath33 , and where we have defined the time independent floquet hoppings @xmath34 , being @xmath35 the @xmath36 order bessel function of the first kind .
the dimensionless functions @xmath37 and @xmath38 encode all the information of the field configuration : @xmath39,\\ { \cal a}_{3}= & a_{x}a,\quad\psi_{3}=0\ .
\end{split } \label{eq : functions}\ ] ] as a consequence , a spatial anisotropy can be tuned by varying the polarization or the amplitude of the field .
this is the key result of the present work . for the sake of clarity
, we shall first focus on the high frequency regime @xmath40 . in that limit ,
the floquet bands are uncoupled and the zeroth fourier component @xmath41 of the floquet operator is the dominant one .
thus , eq.[eq : epsilon ] is block - diagonal in the fourier space , and simply consists in a collection of identical @xmath42 time - independent hamiltonians separated in energy by @xmath43 .
the quasi - energy of a side - band @xmath44 , is then simply given by @xmath45 , which shows that the quasi - energy of each floquet band is , in the high - frequency regime , identical to the energy bands of the undriven honeycomb lattice with renormalized hopping parameters @xmath46 .
in this section , we analyze the fate of the dirac points of the quasi - energy spectrum in the high frequency regime as a function of the parameters of the electric field .
note that in this regime , the renormalized hoppings @xmath47 are real parameters , so that the system is still time - reversal invariant for all field polarizations , and the dirac nodes are thus well defined@xcite .
a direct consequence of the ac field induced anisotropy of the hopping parameters , is to change the location of the dirac points .
the two dirac points , related by time - reversal and inversion symmetry , move as the anisotropy is modified and merge at one of the four time - reversal ( inversion ) symmetric points of the brillouin zone m@xmath48 , whenever a specific relation between the hopping parameters is fulfilled @xcite : @xmath49 where the index @xmath50 has been dropped out for clarity .
the merging transition corresponds to the creation / annihilation of a pdp . at the transition ,
the so - called semi - dirac dispersion relation is quadratic in one direction , but remains linear in the other one , leading to the prediction of striking properties such as an unusual temperature dependence of the specific heat and magnetic field dependence of the landau levels @xcite .
such transitions can now be achieved for specific values of amplitude @xmath51 and polarization @xmath52 of the electric field .
the system exhibits several distinct semi - metallic and insulating phases that we now describe .
fig.[fig : ellips ] shows a typical example of a semi - metallic / insulating phase diagram obtained in the high frequency regime when varying the amplitude and the polarization of the field ( other examples are shown in the appendix c ) .
the creation / annihilation of a pdp at a m@xmath48 point is represented by a continuous _ merging line _ that separates a semi - metallic phase from an insulating phase .
thus , there are four different merging lines , one for each point m@xmath48 .
it follows that distinct semi - metallic or insulating phases ( coloured differently ) may emerge when different merging lines are crossed .
this distinction can be made in terms of both the points m@xmath48 and topological ( winding ) numbers : each semi - metallic phase can be labeled by a point m@xmath48 where a pdp _ can not _ be annihilated .
for instance , a pdp can not be annihilated at m@xmath53 ( green line ) from the semi - metallic phase @xmath54 ( which is the one connected to the undriven graphene phase ) .
the reason is that all the hopping parameters in the @xmath54 phase have the same sign , so that the merging at @xmath55 would require first a change in sign of any of them , and according to eq.[eq : condition ] , it would fulfill the merging conditions at other m@xmath56 point . therefore , one can distinguish four semi - metallic phases , denoted as @xmath57 , one for each @xmath58 point where the merging transition can not occur .
in addition , two winding numbers can be introduced to characterize the topology of these four semi - metallic phases .
indeed , the topological properties of out - of - equilibrium floquet systems can be characterized in a similar way to those of static systems ( see for instance ref.@xcite ) .
the first one is the quantized berry phase around a dirac point , giving rise to a topological charge @xmath59 assigned to each dirac point , where @xmath60 is the polar angle parameterizing the bloch sphere .
the two dirac points of graphene carry opposite topological charges that annihilate at the merging transition @xcite . besides
, the topological transition is accompanied with a change of a second winding number , _ the berry phase evaluated across a reduced one - dimensional brillouin zone _ , namely the zak phase @xmath61 , where @xmath62 is a vector of the reciprocal lattice .
chiral symmetry guaranties integer values of the zak phase @xcite which depend on the direction in the reciprocal space .
this reflects the edge orientations dependence for the density of zero - energy edge modes @xcite .
thus , the four semi - metallic phases differ by the set of values the zak phase takes for all directions in the reciprocal space , and therefore by the range of existence in @xmath63-space of edge zero - energy modes for a given edge orientation .
similarly , different insulating phases can be distinguished as well by the set of values the zak phase takes in all possible directions , even though a pdp can , in principle , be created at any of the four points m@xmath48 .
we find one normal insulating ( ni ) phase for which the zak phase is zero in every directions ( meaning the absence of zero - energy edge states ) , and , for polarizations different than @xmath64 , two _ zak insulating _
phases for which the zak phase is non zero in different directions .
the zak insulating phases reflect , in 2d , the underlying non - trivial topology expected for a bdi symmetry class in 1d@xcite .
interestingly , the phase diagrams show multiple crossings between the merging lines , meaning that a pdp _ can be created and annihilated simultaneously at two different @xmath58 points_. following eq.[eq : condition ] , such crossings actually imply the vanishing of one of the three hopping parameters @xmath65 , what directly induces localization of the electrons in the direction @xmath26 .
this striking feature corresponds to a transition between two insulating or two semi - metallic phases . at the transition ,
the gap closes along lines parallel to the @xmath26 direction that passes through the two distinct merging points . unlike a single merging transition
, it follows that the dispersion relation is flat along the @xmath26 direction but remains linear in the other one , as shown in fig.[fig : circular ] ( c ) . at the transition ,
the system is then reduced to an array of uncoupled one - dimensional chains , hosting one - dimensional massless dirac fermions .
remarkably , for phase polarization @xmath64 , the merging at @xmath66 and @xmath67 is always coincident .
this gives rise to critical _ localization lines _ that separate two sm phases as shown in fig.[fig : circular ] .
at the transition between two semi - metallic phases , the topological charges assigned to the dirac points change sign ( see fig.[fig : circular ] ( b ) and ( d ) ) . in that sense ,
the localization ( or double merging ) transition is also a topological transition . finally , we report that the phase diagram for polarization @xmath64 also exhibits crossings between the four merging lines simultaneously , the crossing of three merging lines only is not possible . ) ] . at such critical points ,
all the hopping parameters @xmath65 vanish and the quasi - energy of each side - band is perfectly flat which leads to full charge localization in all directions . ( a )
semi - metallic ( sm ) / insulating ( i ) phase diagram in the high frequency regime for @xmath64 , where only three phases appear , due to the overlap between the m@xmath68 and m@xmath69 merging lines .
( c ) this overlap gives rise to localization lines separating two semi - metallic phases . at the transition between these two semi - metallic phases , the topological charges assigned to each dirac point change sign ( ( b ) and ( d ) )
this topological transition goes together with a change of the value of the zak phase , which is @xmath70 ( @xmath50 ) in the sm@xmath71 ( sm@xmath53 ) phase along the path represented by a full ( dashed ) arrow.,scaledwidth=43.0% ]
although the merging transition in the high frequency regime is simple to describe , it might be difficult to achieve experimentally in graphene , since it requires high field amplitudes which also would involve heating effects .
for that reason , we now investigate the effect of a frequency decrease and show that the merging transition may also be achieved by varying the frequency , and thus requires smaller intensities of the ac field .
importantly , besides being more relevant experimentally , it turns out that such a regime is particularly interesting since it allows us to engineer additional dirac points .
sketch of the first uncoupled side - bands @xmath72 as a function of a parameter ( for instance the field amplitude ) , for different driving frequencies .
the colored area highlight the range where the side - bands are inverted . ]
the high frequency regime holds as long as the frequency is larger than the bandwidth of the undriven system , namely , @xmath73 . for lower frequencies ( @xmath74 )
the coupling between the floquet side - bands becomes relevant , and the high frequency effective theory is not accurate . besides , unless the polarization of the field is linear , the couplings break time reversal symmetry and will therefore open a gap at the dirac points , non linear field polarizations induce complex hoppings , and thus break time reversal symmetry opening gaps
. however , similarly to the boron nitride model , the topological charges assigned to the non - trivial berry phases in each valley remain , so that their annihilation / creation can still be achieved by varying the parameters of the field . ] . in the following ,
we therefore restrict our analysis to linearly polarized fields only , although our formalism remains valid for any phase polarization .
quasi - energy spectra for @xmath75 , @xmath76 and @xmath77 with ( top , left ) @xmath78 , ( top , right ) @xmath79 , ( bottom , left ) @xmath80 and ( bottom , right ) @xmath81 .
, title="fig : " ] quasi - energy spectra for @xmath75 , @xmath76 and @xmath77 with ( top , left ) @xmath78 , ( top , right ) @xmath79 , ( bottom , left ) @xmath80 and ( bottom , right ) @xmath81 .
, title="fig : " ] quasi - energy spectra for @xmath75 , @xmath76 and @xmath77 with ( top , left ) @xmath78 , ( top , right ) @xmath79 , ( bottom , left ) @xmath80 and ( bottom , right ) @xmath81 .
, title="fig : " ] quasi - energy spectra for @xmath75 , @xmath76 and @xmath77 with ( top , left ) @xmath78 , ( top , right ) @xmath79 , ( bottom , left ) @xmath80 and ( bottom , right ) @xmath81 .
, title="fig : " ] as illustrated in fig.[fig : sketch ] , the floquet side - bands overlap as the frequency is decreased .
this yields new couplings which can now include the absorption or the emission of photons . at around about @xmath82 , the ( uncoupled ) side - bands @xmath83 and @xmath84 are expected to touch at @xmath85 for vanishing amplitude of the field , giving rise to a new crossing at the @xmath55 point . this new crossing corresponds to a band inversion , in which the symmetries of the conduction band and the valence band are exchanged .
as the high frequency approximation fails to describe this situation , the floquet operator is now diagonalized numerically . for more concreteness
, we fix the system in a normal insulating phase in the high frequency regime by choosing suitable field amplitudes ( fig.[fig : low ] , top , left ) .
then , we decrease the frequency such that the side - bands @xmath83 and @xmath84 cross .
this leads to the appearance of a pdp at the center of the brillouin zone @xmath55 for @xmath79 with the typical semi - dirac dispersion relation ( quadratic in one direction and linear in the other one , as it is shown in fig.[fig : low ] , top , right ) . when the frequency is further decreased , the two dirac points move away to each other while the rest of the bands bend ( fig.[fig : low ] , bottom , left )
. a larger decrease of frequency makes the bands close the gap once again at the @xmath86 point of the fbz .
this creates a _ second _ pdp with the same typical semi - dirac shape , which move away as the frequency is lowered , as it is shown in fig.[fig : low ] , bottom , right .
then , for lower frequencies the band structure now owns _ four _ dirac points at the same quasi - energy .
therefore , the driving field offers not only the opportunity to manipulate the dirac points , but also to create new ones and observe the merging transition by just tuning the frequency of the ac field .
we have shown that the merging of the dirac points in the honeycomb lattice can be driven in a controllable way by an ac electric field .
we have obtained rich phase diagrams in which distinct topological semi - metallic and insulating phases emerge . at high frequency ,
@xmath40 , the behavior is easily described by an effective floquet operator , equivalent to the hamiltonian of undriven graphene , but with renormalized hoppings tuned by the field amplitude and the phase polarization .
the resulting phases reflect interesting topological features , and the critical lines in which pairs of dirac points are created and annihilated simultaneously show unexpected localization properties for which electrons behave relativistically in one direction , while they are localized in the other one .
importantly , we also proved that out of the high frequency regime ( @xmath87 ) , and for linearly polarized fields , a change in frequency can drive a novel transition with the emergence of new pairs of dirac points , although it does not necessarily correspond to a transition between an insulating and a semi - metallic phase .
this new transition lies on multi - side - bands couplings , or in other words , multi - photon assisted transitions .
the resulting new semi - metallic phase with four dirac points can clearly be distinguished from the one with two dirac points by various transport measurements .
for instance , a hall conductance measurement would provide a direct signature of the two additional dirac points , since each valley brings a contribution @xmath88 to the total transverse conductivity , where @xmath89 is the number of landau levels .
the conductivity measurements would be accomplished by considering the floquet sum rule obtained in ref.@xcite , due to the far from equilibrium situation of the setup at intermediate frequencies @xmath90 .
we also notice that for samples of the size of or smaller than the wave length ( @xmath91 micron ) , the effect of the magnetic field of the irradiation can be ignored . for the realization in graphene , the high frequency regime requires field frequencies in the near ultraviolet and field amplitudes at least of the order @xmath92 @xcite .
the energy corresponding to this electric field is about half the ionization energy of carbon atoms .
however , real graphene also possesses @xmath93-orbitals which can be affected by such a high frequency field .
thus , in order to achieve the high frequency regime , a frequency larger than the actual band width of the graphene bands must be considered . in addition , for fields of both high frequency and high amplitude , the heating of the sample will certainly be an issue because of the dissipative processes due to phonons and electron - electron scattering .
one way to partially avoid these issues can be the use of alternative platforms with similar properties , such as artificial graphene@xcite .
this would help in two different ways : the increase of the lattice parameter allows one to consider lower frequencies and thus lower the field intensities , while in addition , the contribution of the @xmath93-orbitals would vanish .
finally , we expect that the results obtained in the lower frequency regime ( @xmath87 ) to be more easily achievable in real graphene , if only because the undesirable heating of the sample will be reduced . furthermore , we stress that this regime allows one not only to manipulate the dirac points but also to create new ones just by tuning the frequency .
+ * acknowledgments : * the authors would like to thank janos asboth , jian li , markus bttiker , hector ochoa , alexey kuzmenko and f. koppens for inspiring discussions .
p. d. was supported by the european marie curie itn nanoctm .
gmez - len acknowledges jae program .
g. l. and g. p. acknowledge mat 2011 - 24331 and itn , grant 234970 ( eu ) for financial support .
references m. hasan and c. kane , rev .
mod . phys . *
82 * , 3045 ( 2010 ) .
qi and s .- c .
zhang , rev .
* 83 * , 1057 ( 2011 ) . c. l. kane and e. j. mele , phys .
lett . * 95 * , 146802 ( 2005 ) .
b. a. bernevig , t. a. hughes , and s .- c .
zhang , science * 314 * , 1757 ( 2006 ) .
markus knig , steffen wiedmann , christoph brne , andreas roth , hartmut buhmann , laurens w. molenkamp , xiao - liang qi , and shou - cheng zhang et al .
, science * 318 * , 766 - 770 ( 2007 ) .
a. h. c. neto , f. guinea , n. m. r. peres , k. s. novoselov , and a. k. geim , rev .
* 81 * , 109 ( 2009 ) .
inoue and a. tanaka , phys .
rev . lett . *
105 * , 017401 ( 2010 ) .
n. h. lindner , g. refael , and v. galitski , nat .
phys * 7 * , 490 ( 2011 ) .
z. gu , h. a. fertig , d. p. arovas , and a. auerbach , phys .
lett . * 107 * , 216601 ( 2011 ) .
t. kitagawa , t. oka , a. brataas , l. fu , and e. demler , phys .
b * 84 * , 235108 ( 2011 ) .
a. gmez - len and g. platero , phys .
* 110 * , 200403 ( 2013 ) .
j. cayssol , b. dora , f. simon and r. moessner , phys .
status solidi rrl * 7 * , 101 ( 2013 ) .
m. c. rechtsman et al .
, nature * 496 * , 196 ( 2013 ) .
p. dietl , f. pichon , and g. montambaux , phys .
* 100 * , 236405 ( 2008 ) .
b. wunsch , f. guinea , and f. sols , new j. phys . *
10 * , 103027 ( 2008 ) .
v. m. pereira , a. h. castro neto , and n. m. r. peres , phys .
b * 80 * , 045401 ( 2009 ) .
j. zak , phys .
* 62 * , 2747 ( 1988 ) .
s. ryu and y. hatsugai , phys .
* 89 * , 077002 ( 2002 ) .
p. delplace , d. ullmo , and g. montambaux , phys .
b * 84 * , 195452 ( 2011 ) .
l. tarruell , d. greif , t. uehlinger , g. jotzu , and t. esslinger , nature * 483 * , 302 ( 2012 ) .
n . fuchs , and g. montambaux , phys .
* 108 * , 175303 ( 2012 ) .
m. c. rechtsman , y. plotnik , j. m. zeuner , and m. s. a. szameit , arxiv:1211.5683 ( 2012 ) .
m. bellec , u. kuhl , g. montambaux , and f. mortessagne , phys .
lett . * 110 * , 033902 ( 2013 ) .
g. montambaux , f. pichon , j .-
fuchs , and m. o. goerbig , phys .
b * 80 * , 153412 ( 2009 ) .
s. koghee , l .- k .
lim , m. o. goerbig , and c. m. smith , phys .
a * 85 * , 023637 ( 2012 ) .
m. busl , g. platero , and a .-
jauho , phys .
b * 85 * , 155449 ( 2012 ) .
g. platero and r. aguado , physics reports * 395 * , 1 ( 2004 ) .
h. sambe , phys .
rev . a * 7 * , 2203 ( 1973 ) .
gmez - len , p. delplace , and g. platero , arxiv : 1309.5402 ( 2013 ) .
g. montambaux , f. pichon , j .-
fuchs , and m .- o .
goerbig , eur .
j. b. * 72 * , 509 ( 2009 ) .
a. p. schnyder , s. ryu , a. furusaki , and a. w. w. ludwig , phys .
b * 78 * , 195125 ( 2009 ) . a. kundu and b. seradjeh , phys .
lett . * 111 * , 136402 ( 2013 ) .
e. j. g. santos and e. kaxiras , nano letters * 13 * , 898 ( 2013 ) .
j. karch et al .
lett . * 107 * , 276601 ( 2011 ) .
s. tani , f. m. c. blanchard , and k. tanaka , phys .
lett . * 109 * , 166603 ( 2012 ) .
m. polini , f. guinea , m. lewenstein , h. c. manoharan , and v. pellegrini , nature nanotechnology * 8 * , 62533 ( 2013 ) .
crystals coupled to in - plane ac electric fields have lattice and time translation invariance : @xmath94 , being @xmath95 the lattice vectors and @xmath96 a period of the driving field .
thus , one can assume solutions in floquet - bloch form @xmath97 , where @xmath98 is the quasi - energy for the @xmath44 floquet state , and @xmath99 the wave - vector .
the floquet - bloch states @xmath100 are periodic in both @xmath101 and @xmath102 , and fulfill the floquet eigenvalue equation : @xmath103 where @xmath104 is the floquet operator .
the time dependent hamiltonian is obtained through the peierls substitution in the tight binding model , leading to the time dependent hoppings @xmath105 , and to the time dependent hamiltonian : @xmath106 where we have included the sub - lattice indexes @xmath44 , @xmath107 , and the time dependent annihilation and creation operators of floquet - bloch states . due to its time periodicity , we expand in fourier series @xmath108 and @xmath109 : @xmath110 eq.[eq : floquet - tight - fourier ] gives a description of the time dependent hamiltonian in terms of the time independent operators @xmath111 .
the calculation of the quasi - energies is easily performed in sambe space by considering the composed scalar product : @xmath112 where @xmath113 represents the usual scalar product in the hilbert space .
this leads to a set of fourier components for the time dependent hamiltonian given by : @xmath114 where @xmath115 .
the quasi - energies are then obtained by direct diagonalization of the infinite dimensional floquet operator in sambe space : @xmath116 where @xmath117 is the identity matrix for the sub - lattice subspace .
note that this representation of the tight binding hamiltonian simply describes a set of hoppings : @xmath118 which include the emision or absorption of @xmath119 photons .
the last term of eq.[eq : floquet - tight - binding ] @xmath120 is simply the fourier space representation of the time derivative operator @xmath121 .
the honeycomb lattice is made of a primitive unit cell with two inequivalent atoms ( a , b ) and unit cell translation vectors @xmath23 , and @xmath24 , where @xmath122 is the distance between the atom a and and the atom b. each a atom has three nearest neighbors ( b type ) at positions @xmath29 , @xmath28 and @xmath27 . in @xmath99-space
the energy spectrum shows two inequivalent gapless points at @xmath123 and @xmath124 called dirac points .
we also define the time reversal invariant momentum ( trim ) points as : @xmath125 in absence of driving , the tight binding hamiltonian up to nearest neighbors ( nn ) for the honeycomb lattice is : @xmath126 where @xmath127 is referred to the hopping along the @xmath26 direction , and for simplicity we have included @xmath25 to take into account the hopping within the same unit cell .
explicitly @xmath128 consists in the the sum of the next three terms : @xmath129 the spectrum of @xmath130 has two bands with dispersion relation given by : @xmath131 when the system is coupled to an ac electric field in the dipolar approximation , one can follow appendix a to obtain the time dependent hamiltonian and the expression for the floquet operator in fourier space . in general , we assume without loss of generality that the vector potential for an homogeneous in - plane electric field is given by : @xmath133 which represents the case of an elliptically polarized field in the x - y plane . note that @xmath77 reduces eq.[eq : graphene - ac - field ] to the case linear polarization , and @xmath134 with @xmath64 reduces eq.[eq : graphene - ac - field ] to the case of circular polarization . to obtain the quasi - energies
, one needs to calculate the matrix elements of the floquet operator by means of eq.[eq : floquet - tight - binding ] .
the integrals are of the form : @xmath135 which has been solved using the identity : @xmath136 where @xmath137 , and @xmath138 is obtained from the relation : @xmath139 i.e. , @xmath140 the fourier space representation of the time dependent tight binding is thus given by blocks ( @xmath115 ) : @xmath141 with @xmath33 and the renormalized hoppings : @xmath142 the functions @xmath38 and @xmath143 encode all the information of the ac field configuration : @xmath144 in the high frequency regime ( @xmath73 ) , the floquet bands are decoupled ( @xmath145 is block diagonal ) , and the system can be described by an effective time independent floquet operator , which is given by the 2 by 2 matrix : @xmath146 [ [ appendix - c - different - field - configurations - in - the - high - frequency - regime ] ] appendix c : different field configurations in the high frequency regime ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1 .
we fix @xmath147 and vary @xmath148 and @xmath52 independently . in this configuration
the relation @xmath147 for the ac field amplitudes compensates the asymmetry in x - y of the honeycomb lattice .
it gives rise to very simple expressions for the renormalized hoppings .
we fix @xmath64 and vary @xmath149 and @xmath148 independently .
note that this is reduced to circular polarization if @xmath134 .
we fix @xmath77 , and vary @xmath149 and @xmath148 independently .
this case contains all possible configurations of linearly polarized fields .
the functions @xmath143 and @xmath38 for the condition @xmath150 are given by : @xmath151}\right),\ \psi_{3}=0\ , \end{aligned}\ ] ] with the renormalized hoppings @xmath152 , @xmath153 , and @xmath154 . the figure [ fig : merging - phases ] shows the corresponding phase diagram . for the case of phase difference @xmath64 ,
the functions are given by : @xmath155 with a corresponding phase diagram plotted in fig.[fig : merging - phases2 ] .
note that the case of circular polarization is obtained by tracing a line @xmath134 .
interestingly , the case @xmath64 only presents a trivial insulating phase ( description in the main text ) , and the two merging lines @xmath66 and @xmath67 overlap .
thus it is possible to find the merging of one or four pdp , being the latter related with full localization of the electrons .
phase diagram as a function of the external ac electric field parameters @xmath156 and @xmath157 for the condition @xmath64 .
the different colors label the four inequivalent @xmath158 points for the merging . note that in this case the merging lines @xmath159 and @xmath160 overlap , giving rise to the existence of two and four simultaneous crossings . ] for linear polarization ( @xmath77 ) the functions are given by : @xmath161 the corresponding phase diagram is plotted in fig.[fig : merging - phases3 ] .
as we can see , this phase diagram has analogies with the one in fig.[fig : merging - phases ] . merging lines as a function of the external ac electric field parameters @xmath156 and @xmath157 for the condition @xmath77 .
the different colors label the four inequivalent @xmath158 points for the merging . ] in addition , in this plot it is clearly seen the imprint of the lattice symmetry ( hexagonal pattern ) around the @xmath54 phase ( the one at @xmath162 which is connected to undriven graphene ) .
this can be expected from the fact that linearly polarized fields do not modify the phase adquired by the electron during the hopping ( @xmath163 for all @xmath164 ) .
thus , the effective lattice with renormalized hoppings must conserve the hexagonal symmetry .
this effect can be understood as well in terms of the hoppings : tuning the electric field orientation , we favor one of the three different hoppings .
it leads to three different merging lines ( blue , red and brown ) and the six - fold symmetry occurs because of the positive / negative value of the field amplitude . | we investigate the effect of an in - plane ac electric field coupled to electrons in the honeycomb lattice and show that it can be used to manipulate the dirac points of the electronic structure .
we find that the position of the dirac points can be controlled by the amplitude and the polarization of the field for high frequency drivings , providing a new platform to achieve their merging , a topological transition which has not been observed yet in electronic systems .
importantly , for lower frequencies we find that the multi - photon absorptions and emissions processes yield the creation of additional pairs of dirac points .
this provides an additional method to achieve the merging transition by just tuning the frequency of the driving .
our approach , based on floquet formalism , is neither restricted to specific choice of amplitude or polarization of the field , nor to a low energy approximation for the hamiltonian . |
the recent successful preparation of monolayer graphene has generated a lot of interest in the physics community and efforts are underway to study the electronic properties of graphene @xcite .
the nature of quasiparticles called dirac electrons in these two - dimensional systems is very different from those of the conventional two - dimensional electron gas ( 2deg ) realized in semiconductor heterostructures .
graphene has a honeycomb lattice of carbon atoms .
the quasiparticles in graphene have a band structure in which electron and hole bands touch at two points in the brillouin zone . at
these dirac points the quasiparticles obey the massless dirac equation . in other words , they behave as massless dirac particles leading to a linear dispersion relation @xmath1 ( with the characteristic velocity @xmath2 .
this difference in the nature of the quasiparticles in graphene from conventional 2deg has given rise to a host of new and unusual phenomena such as anomalous quantum hall effects and a @xmath3 berry phase@xcite@xcite . among the electronic properties of interest
is the interaction of electrons with artificially created periodic potentials .
it has been observed that if conventional 2deg is subjected to artificially created periodic potentials in the submicrometer range it leads to the appearance of weiss oscillations in the magnetoresistance .
this type of electrical modulation of the 2d system can be carried out by depositing an array of parallel metallic strips on the surface or through two interfering laser beams @xcite .
besides the fundamental interest in understanding the electronic properties of graphene there is also serious suggestions that it can serve as the building block for nanoelectronic devices @xcite . in conventional 2deg systems ,
electron transport in the presence of electric and magnetic modulation has continued to be an active area of research @xcite .
recently , electrical transport in graphene monolayer in the presence of electrical modulation was considered and theoretical predictions made @xcite .
we have also carried out a study of magnetoconductivity when graphene monolayer is subjected to magnetic modulation alone@xcite . along the same lines , in this work we investigate low temperature magnetotransport of dirac electrons in a graphene monolayer subjected to both electric and magnetic modulations with the same period . from a practical point of view , this is important to consider as magnetic modulation of graphene can be realized by magnetic or superconducting stripes placed on top of graphene which in turn act as electrical gates that induce electric modulation . the relative phase of the two modulations can have important consequences for magnetotransport in the system as was seen in conventional 2deg . therefore in this work
we investigate weiss oscillations in magnetoconductivity @xmath0 in a graphene monolayer for both the cases when the electric and magnetic modulations are in - phase and when they are out of phase .
we consider two - dimensional dirac electrons in graphene moving in the @xmath4plane .
the magnetic filed ( @xmath5 ) is applied along the @xmath6direction perpendicular to the graphene plane .
we consider the perpendicular magnetic field @xmath5 modulated weakly and periodically along one direction such that @xmath7 . here
@xmath8 is the strength of the magnetic modulation and @xmath9 with @xmath10 being the modulation period .
we consider the modulation to be weak such that @xmath11 .
we use the landau gauge for vector potential @xmath12 . in effective mass approximation
the one electron hamiltonian is @xmath13 the low energy excitations are described by the two - dimensional ( 2d ) dirac like hamiltonian @xcite@xmath14here @xmath15 are the pauli matrices and @xmath16 characterizes the electron velocity .
we employ the landau gauge and write the vector potential as @xmath17 .
the hamiltonian can be written as @xmath18where @xmath19 is the unmodulated hamiltonian given as @xmath20 and @xmath21the landau level energy eigenvalues without modulation are given by@xmath22 where @xmath23 is an integer and @xmath24 as has been pointed out @xcite the landau level spectrum for dirac electrons is significantly different from the spectrum for electrons in conventional 2deg which is given as @xmath25 , where @xmath26 is the cyclotron frequency .
the eigenfunctions without modulation are given by @xmath27{c}% -i\varphi_{n-1}(x , x_{0})\\ \varphi_{n}(x , x_{0 } ) \end{array } \right ) \label{2}%\ ] ] where @xmath28 are the normalized harmonic oscillator eigenfunctions@xmath29 is the magnetic length , @xmath30 @xmath31 is the @xmath32-dimension of the graphene layer and @xmath33 are the hermite polynomials .
since we are considering weak modulation @xmath34 @xmath5 , we can apply standard perturbation theory to determine the first order corrections to the unmodulated energy eigenvalues in the presence of modulation with the result @xmath35\right ) $ ] where @xmath36 , @xmath37 and @xmath38 are the laguerre polynomials .
hence the energy eigenvalues in the presence of the periodic magnetic modulation are @xmath39 with @xmath40 $ ] .
we observe that the degeneracy of the landau level spectrum of the unmodulated system with respect to @xmath41 is lifted in the presence of modulation with the explicit presence of @xmath41 in @xmath42 the @xmath43 landau level is different from the rest as the energy of this level is zero and electrons in this level do not contribute to diffusive conductivity calculated in the next section .
the rest of the landau levels broaden into bands .
the landau bandwidths @xmath44 oscillates as a function of @xmath23 since @xmath38 are oscillatory functions of the index @xmath23 .
since we are interested in electron transport in the presence of both electric and magnetic modulations , we consider an additional weak electric modulation potential given as @xmath45 on the system . here
@xmath46 is the amplitude of modulation .
we can determine the energy eigenvalues in the presence of weak electric modulation where we take @xmath47 to be an order of magnitude smaller than the fermi energy @xmath48with @xmath49 is the magnitude of fermi wave vector with @xmath50 being the electron concentration .
hence we can apply standard first order perturbation theory to determine the energy eigenvalues in the presence of modulation .
the first order correction in the energy eigenvalues when electric modulation is present is given as@xmath51 here , @xmath52 $ ] , @xmath53 and , @xmath38 and @xmath54are laguerre polynomials .
the energy eigenvalues of dirac electrons in the presence of both modulations can be expressed as@xmath55 to better appreciate the modulation effects on the landau levels we determine the asymptotic expression for the bandwidth ( @xmath56 ) next .
the width of the @xmath23th landau level in the presence of periodic electric and magnetic modulation is given as @xmath57 where @xmath58 is width of the electric modulation and @xmath59 is the width of the magnetic modulation : @xmath60 \label{6}%\ ] ] the asymptotic expression of bandwidth can be obtained by using the following asymptotic expression for the laguerre polynomials by taking the large @xmath23 ( @xmath61 ) limit as@xmath62 with the result that the asymptotic expression for @xmath59 is@xmath63 similarly , for electric modulation , the bandwidth @xmath58 is given as@xmath64 and the asymptotic expression for @xmath58 is@xmath65 therefore the bandwidth in the presence of both electric and magnetic modulations can be expressed as@xmath66 in fig.(1 ) we present the bandwidths as a function of the magnetic field for temperature @xmath67 , electron density @xmath68 the period of modulation @xmath69the strength of the electric modulation @xmath70 where as @xmath71 which corresponds to @xmath72 with the result that both the modulations have equal strengths . in the same figure
we have also shown the bandwidths when either the magnetic or electric modulation alone is present .
we observe that electric and magnetic bandwidths are out of phase while the positions of the extrema of combined bandwidth are shifted with respect to electric and magnetic bandwidths .
the combined bandwidth when both modulations are present will affect the conductivity and that is considered in the next section .
to calculate the electrical conductivity in the presence of weak electric and magnetic modulations we use kubo formula to calculate the linear response to applied external fields . in a magnetic field ,
the main contribution to weiss oscillations comes from the scattering induced migration of the larmor circle center .
this is diffusive conductivity and we shall determine it following the approach in @xcite where it was shown that the diagonal component of conductivity @xmath0 can be calculated by the following expression in the case of quasielastic scattering of electrons@xmath73\tau(e_{\zeta})(\upsilon_{y}^{\zeta})^{2 } \label{10}%\ ] ] @xmath74 , @xmath31 , are the dimensions of the layer , @xmath75 is the inverse temperature with @xmath76 the boltzmann constant , @xmath77 is the fermi dirac distribution function , @xmath78 is the electron relaxation time and @xmath79 denotes the quantum numbers of the electron eigenstate .
the diagonal component of the conductivity @xmath0 is due to modulation induced broadening of landau bands and hence it carries the effects of modulation in which we are primarily interested in this work . @xmath80 does not contribute as the component of velocity in the @xmath81-direction is zero here . the collisional contribution due to impurities is not taken into account in this work .
the summation in eq.(10 ) over the quantum numbers @xmath79 can be written as@xmath82 the component of velocity required in eq.(10 ) can be calculated from the following expression@xmath83 substituting the expression for @xmath84 obtained in eq.(5 ) into eq.(12 ) yields @xmath85 \label{13}%\ ] ] as a result @xmath86 the corresponding velocity given by eq.(13 ) contains the contribution from both the modulations ( electric and magnetic ) obtained in eq.(5 ) compared to one term in velocity component for each@xcite .
this term ( @xmath87 ) has important consequences for the quantum transport phenomena in modulated systems .
with the results obtained in eqs.(11 ) , ( 12 ) and ( 13 ) we can express the diffusive contribution to the conductivity given by eq.(10 ) as@xmath88 where@xmath89 and @xmath90 is given as @xmath91^{2}}\left [ \frac{\sqrt { u}e^{-u/2}v_{0}}{2}(l_{n}(u)+l_{n-1}(u))+\omega_{0}e^{-u/2}\sqrt{2n}% ( l_{n-1}(u)-l_{n}(u))\right ] ^{2 } \label{16}%\ ] ] where @xmath92 $ ] and @xmath93 is the fermi energy . in fig.(2 )
we show the in - phase conductivity ( the magnetic and the electric modulations are in - phase ) @xmath0 given by eqs(14,15,16 ) as a function of the inverse magnetic field for temperature @xmath67 , electron density @xmath68 the period of modulation @xmath69 the dimensionless magnetic field @xmath94 is introduced where @xmath95 for @xmath69 the strength of the electric modulation @xmath70 where as @xmath71 which corresponds to @xmath72 with the result that both the modulations have equal strengths . in the same figure
we have also shown the conductivity when either the magnetic or electric modulation alone is present .
the @xmath96 phase difference in the bandwidths results in the same phase difference appearing in the conductivity for electric and magnetic modulations as can be seen in the figure .
to better understand the effects of in - phase modulations on the conductivity we consider the asymptotic expression of the quantity @xmath90 given by eq.(16 ) that appears in the magnetoconductivity @xmath0 .
the asymptotic results are valid when applied magnetic field is weak such that many landau levels are filled .
the asymptotic expression is obtained in the next section .
to get a better understanding of the results of the previous section we will consider the asymptotic expression of conductivity where analytic results in terms of elementary functions can be obtained following @xcite .
the asymptotic expression of @xmath90 can be obtained by employing the following asymptotic expression for the laguerre polynomials which is valid in the limit of large @xmath23 when many landau levels are filled @xcite@xmath97 with the result that the in - phase bandwidth can be written as@xmath98 where the ratio between the two modulation strengths @xmath99 the flat band condition from the above equation is @xmath100 where @xmath101 is an integer .
this condition can also be expressed as @xmath102 , where @xmath103 is the highest fermi integer .
we see that the flat band condition in this case depends on the relative strength of the two modulations .
we now take the continuum limit:@xmath104 to express @xmath90 in eq.(16 ) as the following integral@xmath105^{2}}\sin^{2}(1/2\sqrt{u / n})\sin^{2}(2\sqrt{nu}-\frac{\pi}{4}+\phi ) \label{19}%\end{aligned}\ ] ] where @xmath106 with @xmath107 and @xmath108 . now assuming that the temperature is low such that @xmath109 and replacing @xmath110 , we rewrite the above integral as@xmath111 where @xmath112 is the dimensionless fermi momentum of the electron . to obtain an analytic solution
we have also replaced @xmath113 by @xmath93 in the above integral except in the sine term in the integrand .
the above expression can be written as @xmath114 the above integration can be performed by using the following identity @xcite:@xmath115 with the result@xmath116 \right ] \label{23}%\end{aligned}\ ] ] where @xmath117 @xmath118and @xmath119 from the asymptotic expression of @xmath90 given by eq.(23 ) , we observe that the effect of the in - phase electric and magnetic modulations is the appearance of a phase factor @xmath120 in the conductivity . the shift in the weiss oscillations when in - phase electric and magnetic modulations are present can be seen in fig.(3 ) .
the phase factor @xmath120 depends on the relative strength of the two modulations .
how the weiss oscillations are affected as @xmath120 as well as the magnetic field is varied can be seen in fig.(3 ) .
the results shown are for a fixed magnetic modulation of strength @xmath72 and the electric modulation is varied .
the change in @xmath46 results in a corresponding change in both @xmath121 and @xmath120 . from fig.(3 ) , we observe that the position of the extrema in magnetoconductivity as a function of the inverse magnetic field depends on the relative strength of the modulations .
the effects of electric and magnetic modulations that are out - of - phase on the conductivity can be better appreciated if we consider the asymptotic expression for @xmath90 in this case .
this is taken up in the next section .
in this section , we calculate @xmath90 when electric and magnetic modulations are out of phase by @xmath122 we consider magnetic modulation out of phase with the electric one : we take the electric modulation to have the same phase as given in the previous section with the @xmath123 phase difference incorporated in the magnetic field .
the energy eigenvalues are@xmath124 and the bandwidth is@xmath125 the term responsible for weiss oscillations is the @xmath126 term under the square root which can be readily seen by considering the large @xmath23 limit of the bandwidth
. therefore for @xmath127 weiss oscillations are no longer present in the bandwidth .
the velocity component @xmath128 is given as@xmath129 \label{26}%\ ] ] and @xmath90 is given as @xmath91^{2}}\left [ \frac { ue^{-u}v_{0}^{2}}{4}(l_{n}(u)+l_{n-1}(u))^{2}+\omega_{0}^{2}e^{-u}% 2n(l_{n-1}(u)-l_{n}(u))^{2}\right ] .
\label{27}%\ ] ] the asymptotic expression for @xmath90 in the presence of both electric and magnetic modulations that are out of phase is obtained by substituting the asymptotic expressions for the laguerre polynomials and converting the sum into integration with the result @xmath130 \right ) \right ] .
\label{28}%\end{aligned}\ ] ] from the expression of the out - of - phase bandwidth given by eq.(25 ) we find that weiss oscillations in the bandwidth are absent for relative modulation strength @xmath127 , the same is reflected in @xmath90 as the term responsible for weiss oscillations ( @xmath131 $ ] ) vanishes for @xmath127 as can be seen from the above equation .
therefore the magnetoconductivity @xmath0 does not exhibit weiss oscillations when the relative modulation strength @xmath132the conductivity as a function of magnetic field when the electric and magnetic modulations are out - of - phase is shown in fig.(4 ) .
the results shown are for a fixed magnetic modulation of strength @xmath133 and the electric modulation @xmath46 is allowed to vary between positive and negative values .
the other parameters are the same as in figs.(1,2,3 ) .
as @xmath46 is varied there is a corresponding change in @xmath134 we find that the positions of the extrema of @xmath0 as a function of the inverse magnetic field do not change as @xmath121 is varied since the phase factor @xmath120 does not appear in the expression of conductivity when the two modulations are out of phase .
it is also observed in fig.(4 ) that there is a @xmath96 phase difference between the curves for @xmath135 and @xmath136 the same behavior is observed in the bandwidth which is reflected in the magnetoconductivity . in this work
we have considered weiss oscillations and have not taken pure shubnikov de hass ( sdh ) oscillations into account but sdh oscillations do appear superimposed on weiss oscillations in the region of strong magnetic field as can been seen in all of our figures .
in addition , since this work was motivated by@xcite , it is important to compare the results obtained here for graphene monolayer with those presented for standard 2deg .
in contrast to the work presented here , in @xcite sdh are explicitly taken into account but we find that our principal expressions eq.(23 ) , ( 28 ) reduce to the expressions there when the terms contributing to sdh oscillations are ignored . in the end , we would also like to mention relevance of this work in experimental studies of transport in graphene .
experiments on graphene monolayer in the presence of modulated electric and magnetic fields have not been realized yet , this theoretical work is in anticipation of experimental work .
we expect that modulation effects predicted here can be observed in graphene employing established techniques that were used for the two - dimensional electron gas systems found in semiconductor heterostructures @xcite . in conventional semiconductor systems ,
modulation of the potential seen by electrons can be produced by molecular beam epitaxy , chemical vapor deposition as well as sputtering techniques @xcite . in graphene
, we expect that modulation effects can be introduced by adsorbing adatoms on graphene surface using similar techniques , by positioning and aligning impurities with scanning tunneling microscopy or by applying top gates to graphene .
epitaxial growth of graphene on a patterned substrate is also possible @xcite . in conclusion
, we have determined the effects of both the electric and magnetic modulations on the magnetoconductivity of a graphene monolayer .
appearance of weiss oscillations in the magnetoconductivity @xmath0 is the main focus of this work .
these oscillations are affected by the relative phase of the two modulations and position of the oscillations depends on the relative strength of the two modulations . we find complete suppression of weiss oscillations for particular relative strength of the modulations when the modulations are out - of - phase .
we gratefully acknowledge f. m. peeters for suggesting the problem and for comments on the manuscript .
one of us ( k.s . ) would like to acknowledge the support of the pakistan science foundation ( psf ) through project no .
c - qu / phys ( 129 ) .
m. t. would like to acknowledge the support of the pakistan higher education commission ( hec ) .
k. s. novoselov _
_ , _ _ a. k. geim , s. v. morozov , d. jiang , m. i. katsnelson , i. v. grigorieva , s. v. dubonos , and a. a. firsov , nature * 438 , * 197 ( 2005 ) ; y. zhang , y. w. tan , h. l. stormer , and p. kim , * 438 * , 201 ( 2005 ) .
y. zheng and t. ando , phys . rev .
b * 65 * , 245420 ( 2002 ) ; v. p. gusynin and s. g. sharapov , phys . rev
. lett . * 95 * , 146801 ( 2005 ) ; n. m. r. perez f. guinea , and a. h. castro neto , phys .
b * 73 * , 125411 ( 2006 ) ; m. i. katsnelson , k. s. novoselov , and a. k. geim , nat . phys .
* 2 * , 620 ( 2006 ) ; k. s. novoselov , e. mccann , s. v. morozov , v. i. falko , m. i. katsnelson , u. zeitler , d. jiang , f. schedin , and a. k. geim , nat .
* 2 * , 177 ( 2006 ) . f. m. peeters and p. vasilopoulos , phys . rev .
b * 46 * , 4667 ( 1992 ) ; a. manolescu , r. r. gerhardts , phys . rev .
b * 56 * , 9707 ( 1997 ) ; r. r. gerhardts , phys .
b * 53 * , 11064 ( 1996 ) ; r. menne , r. r. gerhardts , phys . rev .
b * 57 * , 1707 ( 1998 ) ; u. j. gossmann , a. manolescu , r. r. gerhardts , phys . rev .
b * 57 * , 1680 ( 1998 ) ; j. shi and f. m. peeters , k. w. edmonds and b. l. gallagher , phys . rev .
b * 66 * , 035328 ( 2002 ) ; p. de ye , d. weiss , r. r. gerhardts , m. seeger , k. v. klitzing , _ _ k. eberl and h. nickel , phys . rev
. lett . * 74 * , 3013 ( 1995 ) ; j. h. ho , y. h. lai , y. h. chui and m. f. lin , nanotechnology * 19 * , 035712 ( 2008 ) ; p. vasilopoulos , f. m. peeters , superlattices and microstructures * 7 * , 393 ( 1990 ) ; f. m. peeters and a. matulis , phys . rev .
b * 48 * , 15166 ( 1993 ) ; f. m. peeters , p. vasilopoulos and jirong shi , j. phys . :
matter * 14 * , 8803 ( 2002 ) ; x. f. wang and p. vasilopoulos , f. m. peeters , phys .
b * 71 * , 125301 ( 2005 ) ; x. f. wang , p. vasilopoulos and f. m. peeters , phys . rev .
b * 69 * , 035331 ( 2004 ) ; d. p. xue and g. xiao , phys . rev .
b * 45 * , 5986 ( 1992 ) .
h. a. carmona,_et.al _ , phys .
lett . * 74 * , 3009 ( 1995 ) ; k. w. edmonds,_et.al _ , phys .
b * 64 * , 041303 ( 2001 ) ; p. de ye , d. weiss , r. r. gerhardts , m. seeger , k. v. klitzing , _ _ k. eberl and h. nickel , phys . rev . lett .
* 74 * , 3013 ( 1995 ) . | we have investigated the electrical transport properties of dirac electrons in a monolayer graphene sheet in the presence of both electric and magnetic modulations .
the effects of the modulations on quantum transport when they are in phase and out of phase are considered .
we present the energy spectrum and the bandwidth of the dirac electrons in the presence of both the modulations .
we determine the @xmath0 component of the magnetoconductivity tensor for this system which is shown to exhibit weiss oscillations.asymptotic expressions for @xmath0 are also calculated to better illustrate the effects of in - phase and out - of - phase modulations.we find that the position of the oscillations in magnetoconductivity depends on the relative strength of the two modulations .
when the two modulations are out - of - phase there is complete suppression of weiss oscillations for particular relative strength of the modulations . |
the jets observed emanating from the nuclei of quasars and other active galactic nuclei ( agn ) represent the most energetic long - lived phenomenon in the universe .
although jet imaging is no longer the sole prerogative of radio astronomy , the exquisite resolution of very long baseline interferometry ( vlbi ) at radio wavelengths remains an unreached goal for sub - millimeter ( sub - mm ) and shorter wavelengths .
it is thought that accretion onto a black hole drives the jet outward via magnetic forces ( e.g. * ? ? ?
currently , the most direct way to provide observational evidence for such a model is to study the linear polarization of jets at different frequencies ( from radio to optical ) along with changes in the innermost jet structure .
the most promising objects in such investigations are blazars , flat - spectrum radio - loud quasars and bl lac objects characterized by high optical polarization up to 46% @xcite , pronounced and rapid variability of flux , and one - sided jets with knots that move at superluminal apparent velocities , as fast as @xmath430@xmath5 @xcite . according to @xcite ,
there is a highly significant correlation between optical polarization and dominance of a compact radio core - jet structure , which suggests beaming of the optical flux along with the radio emission .
comparison of radio - loud quasars with low and high optical polarization shows a higher fractional polarization of the radio core for the latter @xcite , which implies a co - spatial origin of the emission at these wavelengths . in support of this , the optical polarization in blazars is affected by the formation and emergence of new vlbi knots @xcite .
these results indicate that simultaneous multifrequency polarization monitoring , together with high resolution polarimetric imaging of the radio jets , provides a unique tool for identifying the location of the regions responsible for variability at different wavelengths and for relating the magnetic field geometry to the structure of the jet .
we have obtained total and polarized intensity images of 15 agns with the very long baseline array ( vlba ) at 7 mm ( 43 ghz ) at 17 epochs over three years .
the vlba observations are accompanied at many epochs by nearly simultaneous ( within two weeks ) measurements of polarization at 1.35/0.85 mm ( 230/350 ghz ) and at optical wavelengths . in the second half of the program simultaneous polarization observations at 3 mm
were performed at several epochs .
the main goals of the project are to relate emission regions at high frequencies to the parsec - scale jet structure and to investigate the strength , direction , and variability of the magnetic field close to the central engine .
these can be achieved only after detailed study of the jet kinematics .
this paper is devoted to an analysis of the jet structure and its variability associated with ejection and propagation of disturbances down the jet , which appear on radio maps as knots of enhanced brightness .
the paper is the first in a series based on the entire data set collected during the project .
some results on individual sources have been published already by ( * ? ? ?
* 120 ) , ( * ? ? ?
* bl lac ) , and ( * ? ? ? * 279 ) .
other papers will include ( 1 ) comparison of the polarization parameters at different frequencies with the jet structure and disturbances in the jet ; ( 2 ) analysis of the available radio , sub - mm , optical , and x - ray light curves to relate the flux variability to activity in the jet ; ( 3 ) results regarding stability of the vlbi core position based on phase - referencing observations obtained for five sources in the sample ; and ( 4 ) structure of the core and intraday total and polarized intensity variability in the parsec - scale jets .
although much progress has been made in studying the kinematics of jets since the vlba started to operate fully in 1995 @xcite , our program is unique regarding the number of sources observed in such detail over a rather long period of time .
this allows us to separate the fast and slow , ballistic and curved motion in the jet flow , define physical parameters for both individual jet features and entire parsec - scale jets , and compare the results across different classes of agns .
our program is designed for comparison of the linear polarization at high frequencies ( mm , sub - mm , and optical wavelengths ) with the parsec scale jet structure of agns .
this has defined the main criteria for selecting the sample and frequency of our vlba observations : 1 .
the sources should be bright , @xmath60.5 jy , and polarized , @xmath63% , at sub - mm wavelengths .
the size of the sample and brightness of sources should allow us to perform vlba observations at a single epoch during 24 hours with sufficient uv - coverage to produce total and polarized intensity images at 43 ghz with high dynamic range .
3 . the sample should contain sources with resolved radio structure from different sub - classes of agns for which variability in the jet flow can be expected on timescales of months .
the sources should be convenient for monitoring in the northern hemisphere and their coordinates should cover the whole range of right ascensions .
following these constraints , we have formed a sample of 15 agns that consists of 8 quasars , 5 bl lac objects , and 2 radio galaxies ( 3c 120 of fanaroff - riley type 1 and 3c 111 of type 2 ) after obtaining information on the linear polarization at sub - mm wavelengths from @xcite .
the sources are listed in table [ sample ] .
we have observed the objects in our sample in four different regions of the electromagnetic spectrum : at 43 ghz ( 7 mm ) with the vlba , at 350/230 ghz ( 0.85/1.3 mm ) with the _ james clerk maxwell telescope _
( jcmt , mauna kea , hawaii ) using scuba @xcite and its polarimeter @xcite , at the steward observatory 1.5 m telescope ( mt .
lemmon , arizona ) with the two - holer polarimeter / photometer @xcite over an effective wavelength range of @xmath760007000@xmath8 , and at 86 ghz ( 3 mm ) with the _ berkeley - illinois - maryland array _
( bima , hat creek , california ) .
the observations were performed from 1998 march to 2001 april .
the vlba monitoring was carried out roughly bimonthly ( 17 epochs ) .
for the jcmt and optical observations the number of epochs depends on the source , with the maximum being 11 and seven epochs and the minimum five and three epochs , respectively , except for the bl lac object 1803 + 784 , which was not observed in the optical region owing an inaccessibly high declination .
the majority of the jcmt polarization observations were accompanied by total flux measurements .
differential @xmath9-band photometry ( @xmath10-band photometry in the case of 3c 273 ) was carried out when conditions were photometric .
the bima polarization observations started in 2000 april and for many sources in the sample were performed simultaneously with the vlba observations at 3 - 4 epochs .
however , each of these epochs usually was accompanied by a series of bima observations separated by several days so that for some sources polarization measurements at 3 mm were obtained at 20 epochs .
the polarization data at 3 mm include all four stokes parameters . in this paper
the optical , jcmt , and bima polarization results are shown only graphically to illustrate the data collected .
detailed descriptions of the data reduction and tables of measurements will be presented in a later paper .
the 43 ghz observations were carried out with the vlba recording system using eight 8 mhz wide channels , each in right and left circular polarization , with 15 - 20 scans of 3 - 5 minute duration for each object .
all 10 antennas were used for each source except at epochs affected by weather or receiver failure .
table [ antenna ] lists the antennas operated at each epoch and typical parameters of the synthesized beam ( those of the oj 287 observations ) for uniform weighting ( used for imaging all sources ) to illustrate the consistency of the @xmath11-coverage over epochs .
initial correlation was carried out at the national radio astronomy observatory ( nrao ) array operations center in socorro , nm .
subsequent calibration was performed with the astronomical image processing system ( aips ) software supplied by nrao , while images were made with the caltech software difmap @xcite .
the calibration included application of the nominal antenna - based gain curves and system temperatures , and correction for sky opacity , followed by iterative imaging plus phase and amplitude self - calibration .
for each epoch we calculated the total flux density in the images of sources that are known to have very weak emission outside the angular size range of the vlba images ( 0420 - 014 , 0528 + 134 , oj 287 , and bl lac ) .
these values were compared with total flux densities obtained by interpolating in time the measurements of the monitoring program at 37 ghz at the metshovi research station , finland @xcite .
the comparison produced the flux density correction factors , @xmath12 , given in table [ antenna ] .
the factors were applied for the final adjustment of the flux density scale in the images . we have constructed light curves of the vlbi core for each source to confirm the absence of correlation between their behavior , indicating no evidence for residual amplitude calibration errors .
the light curves of the 15 sources peak at 9 different epochs and the correlation coefficients are spread between @xmath130.4 and + 0.3 , consistent with no systematic calibration errors in the flux density scaling .
we performed a cross - hand fringe fit using a scan of 3c 279 averaged over all baselines ( the data were corrected for parallactic angle rotation of phases in advance ) .
the resulting right - left phase and rate delay corrections were processed with the aips task polsn and applied to the full data set .
after preliminary images were produced , the images were used in aips task calib to self - calibrate the phases of all sources ; this includes the removal of residual r - l phase differences owing to calibration errors .
the instrumental polarization `` d - terms '' were determined via the method of @xcite and @xcite .
a final set of d - terms at each epoch was obtained by averaging of the solutions for those sources in the sample for which there is the best agreement between the d - terms ( usually , 3c 111 , oj 287 , 3c 345 , and bl lac ) . the electric vector position angle ( evpa ) calibration was obtained by different methods : comparison between the very large array ( vla ) and vlba integrated evpas at quasi - simultaneous epochs , d - terms method ( see below ) , and using evpa - stable features in the images of the jets of 3c 279 , oj 287 , and cta102 . over the period 1998 march 1999 february we used results of the vla observations performed for 0420@xmath13014 , 3c 120 , oj 287 , 3c 279 , bl lac , and 3c 454.3 by @xcite .
later we obtained polarization measurements for 0420@xmath13014 , 0528 + 134 , oj 287 , and 1803 + 784 at two epochs , 2000 april and 2000 july , with the vla @xmath14 and @xmath15 configuration , respectively . at epochs where vla data were not available we used the d - terms method , which is based on the assumption that the instrumental polarization parameters change slowly with time @xcite . a detailed description of the method is given in @xcite .
in addition , later we checked our calibration for these epochs using the nrao data base at * http://www.vla.nrao.edu/astro/calib/polar/ * for sources 0420@xmath13014 , 0528 + 134 , oj287 , and bl lac when the nrao data were available at epochs close to the vlba dates .
the results are completely consistent with our calibration .
the evpa - stable feature in 3c 279 is the well - known superluminal component @xmath16 ( e.g. , * ? ? ?
* ) , which underwent a change in evpa in the beginning of 1998 @xcite but maintained this evpa until the end of our program .
the evpa - stable features in oj 287 and cta 102 are quasi - stationary components at @xmath71 mas and @xmath72 mas from the core , respectively .
however , these both became very weak in the second half of 2000 in the 43 ghz images .
the final evpa calibration is a result of the best agreement among the different methods .
the accuracy of the calibrated evpa measurements , @xmath17(evpa ) , is indicated in table [ antenna ] .
we have combined the images obtained for every source in a sequence of total intensity maps convolved with the same beam , corresponding to the average beam during the epochs when all 10 antennas were in operation .
the contours are in terms of the global peak of the maximum total intensity observed over all epochs .
the sequences are presented in figures [ alla]-[allo ] , which also show the jcmt , bima , and optical polarization measurements .
all images are oriented with north toward the top and east toward the left .
the scale in mas is indicated along one of the axes , while the epoch of each image is given along the other axis .
the peaks of the total , @xmath18 , and polarized , @xmath19 , intensity , parameters of the average beam , the total intensity of the lowest contour , @xmath20 , and the lowest level of the polarized intensity , @xmath21 , of the combined maps are indicated in table [ sample ] .
the polarization at 7 mm is shown inside the total intensity images by line segments , which are plotted if the local polarized intensity exceeds @xmath21 .
the segments are oriented in the local direction of the polarization and their length is proportional to the local polarized intensity , with the maximum length corresponding to the global peak of the polarized intensity over all epochs , @xmath19 .
table [ antenna ] marks epochs when the bima , jcmt , and optical observations were performed within two weeks of the corresponding vlba epoch .
we employed the task modelfit in difmap to represent each total intensity image ( stokes parameter @xmath22 ) as a sequence of circular gaussian components that are characterized by flux density , size , and position relative to the map center .
initially , point - like components are used to obtain an image with 100:1 dynamic range .
when a set of point - like components was found , a final group of 100 iterations was executed with all parameters of all components allowed to vary to define their properties .
in addition , because we have roughly bimonthly observations , we repeated the model fitting with the components from the previous epoch used as the initial model ( except for the first epoch , 1998 march ) .
this improves identification of components over epochs and provides an estimate of the accuracy of the parameters by comparison of the outcomes obtained with different initial models . for several sources ( 3c 66a ,
oj 287 , 3c 279 , pks 1510@xmath13089 , and 1803 + 784 ) at four epochs ( 1998 march 1998 , 1999 october , 2000 july 2000 , and 2001 january ) we derived estimates of the 1@xmath17 uncertainties of the best - fit models using the method described by @xcite and realized in the package difwrap @xcite .
the uncertainties depend significantly on the brightness and size of components : 1 ) for _ bright _ ( flux of knot @xmath23 noise level ) and _ compact _ ( size @xmath240.1 mas ) features , uncertainties in flux density @xmath71% , in position @xmath70.01 mas , and in size @xmath25 1% ; 2 ) for the majority of components with size 0.1 - 0.3 mas and flux @xmath650 mjy , uncertainties in the flux density @xmath73% , in position @xmath71/5 of the beam size , and in size @xmath75% ; 3 ) for diffuse components , uncertainties in flux density @xmath710% , in position are comparable with half the size of the knot , and in size @xmath710% ; 4 ) for components weaker than 50 mjy , uncertainties in flux density @xmath750% and positional uncertainties correspond to the size of the average beam ( see table [ sample ] ) ; however , for the majority of weak components these uncertainties were later modified as discussed in 4 ( errors of polynomial parameters ) .
we determined the parameters of polarized components by applying the task modelfit to the @xmath11-data in stokes parameters @xmath26 and @xmath27 separately .
for this purpose only point - like components were used in the models . because the q and u components can be either positive or negative , the procedure needs to be carried out carefully by checking with the polarized intensity image to avoid the inclusion of false features generated by the imaging procedure .
the derived positional parameters of the @xmath26 and @xmath27 components were compared with the parameters of @xmath22 components .
we consider a @xmath26 and/or @xmath27 component to be associated with an @xmath22 component if they are co - spatial to within the uncertainties .
if both polarized components are present , the position of the polarized component is defined by the average location , weighted by the absolute values of the @xmath26 and @xmath27 fluxes .
there are few cases when a significant polarized component does not have an @xmath22 counterpart .
uncertainties in the values of the parameters of polarized components are difficult to define .
estimates of the accuracy of polarization parameters in jet features at 15 and 22 ghz are discussed by @xcite . for components that are bright in the total intensity and highly polarized ( percent polarization @xmath285% )
, we find that the uncertainties in fractional polarization @xmath29% and in polarization angle @xmath30 .
a table of parameters of jet features for each source over all epochs can be found at the web - site * www.bu.edu / blazars / multi.html*. the columns of these tables are as follows : 1 - epoch , 2 - flux density in jy , 3 - relative right ascension in mas , 4 - relative declination in mas , 5 - distance from the core in mas , 6 - position angle relative to the core in degrees , 7 - angular size in mas , 8 - polarized flux density of a polarized component associated with a total intensity component , in jy , 9 - distance of polarized component from the core in mas , 10 - position angle of polarized component relative to the core in degrees , 11 - evpa in degrees .
the data allow us to follow the evolution of major features of the jets that are observed in the total and polarized intensity vlba images of the 15 agns at 17 epochs over 3 years . for each image
we identify a component , @xmath31 , as the vlbi core .
for all sources @xmath31 is located at one end of the jet , although it is not always the brightest feature .
the position of the core in right ascension @xmath32 and declination @xmath33 is defined as @xmath34 and @xmath35 . in our analysis
we assume that the core is stationary over all epochs .
locations of other features are determined relative to the core .
components are categorized as follows : knots @xmath36 ( other than @xmath31 ) are features that either are stationary ( the proper motion is less than or equal to its uncertainty ) , undergo reverse motion ( toward the core ) , or move at subluminal apparent speeds .
knots @xmath37 , @xmath14 , and @xmath38 are superluminal features , where @xmath37 components are the fastest knots in the jet , @xmath38 components are the farthest knots from the core , ejected before our monitoring period started , and the remainder are labeled as @xmath14 .
the designation of components is different for 3c 120 , 3c 279 , 3c 345 , and bl lac , where the naming corresponds to @xcite , @xcite , @xcite , and @xcite , respectively .
each component is characterized by the following parameters : total @xmath39 and polarized @xmath40 flux density ; positions @xmath32 ( ra ) , @xmath33 ( dec ) , and @xmath10 , where @xmath41 ; position angle @xmath42 ; size ( fwhm ) of component @xmath43 ; and electric vector position angle , evpa . to define the temporal evolution of the jet and determine the apparent velocities of the jet flow we perform the following steps .
_ identification of components at different epochs . _ the identification of a component is based on comparison of the parameters during the epochs when it is visible on the images .
we assume that the same component has similar total , @xmath39 , and polarized , @xmath44 , fluxes , position angles @xmath45 and evpa , and size , @xmath43 , at successive epochs . however , some components evolve dramatically even over a two - month interval between epochs , splitting into one or more subcomponents or merging with other features of the jet . in this paper
we analyze those jet features whose identification is supported by the similarity of a number of parameters at different epochs .
_ fitting various polynomials . _
the thorough sequences of images allow us to search for acceleration / deceleration of the jet flow and non - ballistic projected trajectories .
we fit the @xmath32,@xmath33 positions of a component over @xmath46 epochs by different polynomials of order @xmath47 : @xmath48 @xmath49 where @xmath50 is the epoch of observation , @xmath51=1, ... ,n , and @xmath52=@xmath53 .
we use the program lsqpol of the fortran version of the package datan @xcite to find the optimal polynomials of order @xmath47 , where @xmath47 runs from @xmath54 to @xmath55 . the upper limit ,
@xmath56 , is a consequence of the maximum number of epochs of observation of a knot , equal to 17 .
the program provides the value @xmath57 , which is the goodness - of - fit by a polynomial of order @xmath47 . for each order
, we perform a @xmath58 test to determine the polynomial that best fits the data .
we choose a polynomial of the lowest order for which @xmath59 , where @xmath60 is the value of the @xmath58 distribution corresponding to significance level @xmath61=0.05 for @xmath62 degrees of freedom @xcite .
examples of selection of polynomials that fit the data are given in tables [ poly1 ] and [ poly2 ] for component @xmath63 identified in the jet of 3c 66a at 16 epochs and component @xmath64 , also seen at 16 epochs , in 3c 273 .
these tables show parameters of the best - fit polynomials of order from 0 to 4 , the corresponding values of the goodness of fit to the data @xmath65 , and the goodness required by the @xmath58-test .
table [ poly1 ] shows that @xmath47=1 satisfies the @xmath58 test for coordinates @xmath32 and @xmath33 in the case of @xmath63 for 3c 66a . for @xmath64
, @xmath47=4 is required to match motion in ra , while a second - order polynomial adequately describes the data for declination . in a few cases the @xmath58 test
is not satisfied even by polynomials of order 4 .
an increase in the order of the polynomial does not improve the situation since , given the limited number of observations , the number of degrees of freedom becomes too small .
such cases are approximated by a straight line with special consideration for uncertainties in the @xmath32,@xmath33 values ( see below ) .
these cases are marked in table [ speed ] by `` 1 * '' under the polynomial order .
_ uncertainties in polynomial parameters .
_ after the best - fit polynomial is found , we run the subroutine lsqasn of the package datan to estimate errors in the derived parameters .
this method is valid for a normal distribution of the unknowns and when the true value lies with probability @xmath66 within the confidence region around the estimated value ( @xmath66=0.95 is used ) .
however , the errors depend significantly on the uncertainties in the individual observations that are obtained from the model fitting ( see 3 ) .
this allows us to lessen the weight of data at epochs with bad weather , a failure of one or more antennas , or when a given component was very weak or diffuse . for component motion that can not be represented well by polynomials @xmath674 ( see above ) , we approximate the motion by a first - order polynomial and determine uncertainties in the @xmath32,@xmath33 position using the program lsqpol and the method suggested by @xcite .
we set the uncertainty for each data point equal to the average beam size and calculate the optimal polynomial of first order corresponding to a preliminary @xmath58 value . taking this preliminary @xmath58 value , we uniformly re - scale the uncertainties in the data points such that @xmath58 would correspond to the @xmath58 value for @xmath61=0.05 and @xmath68 .
_ calculation of proper motion , acceleration , and ejection time .
_ we define the average proper motion as a vector ( @xmath69,@xmath70 ) , where @xmath69 represents the mean angular speed of motion and @xmath70 gives the average direction of motion .
the average values are derived as follows : @xmath71 and @xmath72 , where @xmath73 and @xmath74 . in the case of first or second - order polynomials , @xmath75=@xmath76 and @xmath77=@xmath78 . for a polynomial of order higher than unity
we compute an average vector of acceleration , ( @xmath79,@xmath80 ) , where @xmath79 is along the direction of the average velocity , @xmath70 . in the case of a second - order polynomial @xmath81 and @xmath82 . for a higher order polynomial , @xmath83 and @xmath84 .
we find superluminal apparent speeds for 19 of 22 components in the radio galaxies , 19 of 31 knots in the bl lac objects , and 46 of 53 knots in the quasars .
table [ speed ] lists the average apparent speed , @xmath85 , calculated in units of the speed of light , @xmath5 , using the average proper motion , @xmath69 .
an inhomogeneous friedmann - lema@xmath86tre - robertson - walker cosmology , with @xmath87 , @xmath88 , and hubble constant @xmath89=70 km s@xmath90 mpc@xmath90 @xcite , is adopted for the calculations .
we derive the uncertainties in the average proper motion , acceleration , and apparent speed from the uncertainties in the polynomial coefficients . table [ speed ]
contains also the number of epochs , @xmath46 , at which the component has been observed ; the average total flux , @xmath91 , in jy ; the average distance from the core , @xmath92 , in mas ; the average position angle of the component , @xmath93 , in degrees , with uncertainty corresponding to the scatter across the epochs ; the average direction of the velocity vector , @xmath70 , in degrees ; and the ejection time ( epoch of zero separation ) , @xmath94 .
the ejection time is the extrapolated time of coincidence of the position of a moving knot with the core in the vlba images .
@xmath94 is the average of @xmath95 and @xmath96 weighted by their uncertainties , where @xmath95 and @xmath96 are roots of the best - fit polynomials . in the case of a polynomial of order 3 and higher ,
the roots are computed by successive iterations with an accuracy of @xmath97 .
table [ accel ] gives the parameters of acceleration for components in table [ speed ] having best - fit polynomials of order @xmath282 ( some columns repeat information from table [ speed ] for convenience ) .
figure [ ident ] shows the positions of the jet components relative to the core for both the total ( open circles ) and polarized ( filled circles ) intensity images of each source .
the solid lines / curves indicate the best polynomial fit to the data .
components presented in figure [ ident ] are marked at the first epoch in figures [ alla]-[allo ] if they are detected then or at the later epoch at which they appear in the jet .
figure [ map ] displays an image from one particular epoch for each source when the most prominent jet features are seen .
table [ tmap ] lists the parameters of the maps shown in figure [ map ] ( rms(i ) and rms(ip ) are the root mean squares of the residual total and polarized intensity on the images , respectively ) .
the trajectories of all components with non - ballistic motion are plotted in figure [ traj ] .
appendix a contains a description of our results for each individual object in the sample .
our intensive , prolonged high - frequency vlbi monitoring reveals complex , changing structure of parsec scale jets , including significant variations in the apparent speed of knots in the jet , @xmath98 , within individual objects .
the apparent speed is determined by the intrinsic velocity , @xmath99 , and the angle between the trajectory and the line of sight , @xmath100 : @xmath101 the intrinsic velocity defines the lorentz factor of the knot , @xmath102 where @xmath99 is in units of the speed of light .
large scatter in the apparent speed within a source can be caused by ( 1 ) different patterns of jet components , such as `` blobs '' of energetic plasma or forward , reverse , or stationary shocks ; ( 2 ) variable lorentz factor of the jet flow ; or ( 3 ) different trajectories of components in the jet .
the latter can result from a change in the direction of the jet or in the paths of components if each given knot does not fill the entire cross - section of the jet .
the observed flux density of a superluminal component is boosted in the direction of the observer by a factor @xmath103 , where @xmath104^{-1}\label{e4}\ ] ] is the doppler factor and @xmath105 is the spectral index ( @xmath106 ) of the knot .
the flux of a component decreases with distance from the core as the result of radiative energy losses and expansion ( `` adiabatic '' cooling ) .
the emergence of a new component is associated with one or more flares in the radio light curves ( e.g. , * ? ? ?
the flare usually has a sharp peak and nearly symmetric exponential rise and decay at high frequencies @xcite . in the case of the shock - in - jet model for the appearance of superluminal knots , variations in flux , spectral energy distribution , and polarization @xcite
, the symmetric light curves of flares suggest that the flux variability is controlled by light travel delays across the shocked region ( see * ? ? ?
this assumption allows us to calculate the doppler factor for each superluminal component using its flux density variability and size measured from the vlba data .
we can then use the derived doppler factors to study the physical parameters of the jets in our sample .
figure [ flux ] shows light curves of all superluminal components listed in table [ speed ] ( except knots classified as trailing features , see 6.6 ) .
each light curve is normalized by the corresponding average flux indicated in table [ speed ] .
we define the timescale of the variability for each superluminal component as @xmath107 @xcite , where @xmath108 and @xmath109 are the measured maximum and minimum flux densities , respectively , and @xmath110 is the time in years between @xmath108 and @xmath109 .
the variability doppler factors are derived as @xmath111 where @xmath38 is the luminosity distance , @xmath112 is the angular size of the component , equal to @xmath113 for a gaussian with fwhm=@xmath43 measured at the epoch of maximum flux if the true geometry is similar to a uniform face - on disk . in the case of a point - like knot
, we adopt @xmath43=0.1 mas , which yields an upper limit to @xmath114 .
this definition of the doppler factor assumes that the variability timescale corresponds to the light - travel time across the knot
. this will be true if the radiative cooling time is shorter than the light crossing time , which in turn is shorter than the timescale for cooling by adiabatic expansion .
we can roughly verify this assumption by using the relation between flux density and size of a shocked region derived in @xcite during the `` adiabatic '' stage .
we calculate the timescale , @xmath115 , of the variability in size for each non - point - like component as @xmath116 ( we use @xmath117 and @xmath118 if @xmath119 , which is valid for the majority of components ) .
figure [ tau_tau ] plots the size variability timescale versus the flux variability timescale .
the straight line indicates the expected relation between @xmath115 and @xmath120 for adiabatic losses for optically thin shocked gas with @xmath121 @xcite .
figure [ tau_tau ] shows that the majority of components have shorter flux variability timescales than those predicted for adiabatic expansion .
this implies that at high radio frequencies the decay in flux is driven by radiative losses and , therefore , equation [ e5 ] should apply to our data .
a widely used method to estimate the doppler factor from vlbi data assumes that the highest apparent speed detected in a source defines the lowest possible lorentz factor of the jet , @xmath122 .
the viewing angle is then taken to be @xmath123 ; this leads to @xmath124 . in figure [ delta_delta ]
doppler factors thus derived are plotted versus the doppler factor computed from the flux variability and size of components having the highest apparent speed in the jets .
figure [ delta_delta ] demonstrates that there is reasonable agreement between the doppler factors estimated by these two different methods .
the best least - square linear fit to the dependence is @xmath125 , consistent with @xmath114 being the true value and @xmath126 being a lower limit taking in account that @xmath126 is estimated for the lowest possible lorentz factor .
we conclude that the combination of variability of flux and measurement of angular sizes provides a new , robust method for deriving doppler factors from well - sampled sequences of vlbi data
. tables [ q_param ] , [ b_param ] , and [ g_param ] give estimates of the lorentz factor , viewing angle , doppler factor , observed brightness temperature , @xmath127 , and intrinsic brightness temperature , @xmath128 , for 43 knots in the quasars , 19 knots in the bl lacs , and 15 knots in the radio galaxies .
the lorentz factor and viewing angle are solutions of the system that combines equations ( [ e3 ] ) and ( [ e4 ] ) for a knot having measured apparent speed with the doppler factor derived via equation ( [ e5 ] ) .
the observed brightness temperature is computed based on the vlbi measurements as @xmath129 k , where @xmath108 is the maximum flux of the component in jy and @xmath112 is its measured angular size in mas at the epoch of maximum flux ( see above ) .
we apply the doppler factors derived with equation ( [ e5 ] ) to estimate the intrinsic brightness temperature , @xmath130 , where we adopt @xmath121 .
we assume that jets have conical structure and constant angle , @xmath131 , between the jet axis and surface of the cone that contains the entire region of emission .
therefore , @xmath131 is the actual half opening angle of the jet .
this assumption is likely correct for the section of the jet within 1 - 2 mas from the core , while at larger distances ( @xmath2810 mas ) the jet could trace out a helical structure covering a wide range of position angles even if the body of the jet remains narrow ( e.g. , * ? ? ?
we estimate the projected half opening angle , @xmath132 , for each source using the ratio between apparent transverse size , @xmath133 , of the jet and apparent longitudinal distance , @xmath134 , of components : @xmath135 , where @xmath136 is the slope of the best linear fit to the relation between @xmath133 and @xmath134 as defined at the position of each component that is brighter than 1% of the peak intensity .
the values of @xmath133 and @xmath134 are calculated as follows : @xmath137 , where @xmath10 is the observed separation of the component from the core ; and @xmath138 , where @xmath139 is the projected direction of the jet as defined by the mean position angle of all sufficiently bright components over all epochs , @xmath45 is the position angle of the component , and @xmath43 is the size of the component .
we first plot the values of @xmath133 against @xmath134 to estimate a preliminary linear dependence .
the preliminary relationship is used to remove points that deviate from the dependence by more than 3@xmath17 of the linear model .
then the value of @xmath136 is obtained by minimizing @xmath58 of the adjusted data . in figure [ opan ]
all pairs of @xmath140 are plotted and the best linear fit of the adjusted data is presented for each source .
figure [ opan ] shows that a jet model with constant opening angle provides a good approximation to the inner jet , although the plots of two bl lac objects ( 3c 66a and 1803 + 784 ) display an increase in @xmath132 beyond 2 and 1 mas , respectively , which might be a general feature of many bl lac jets . in the radio galaxy 3c 111 and quasar 3c 345 a decrease in @xmath132
is observed , and is most likely caused by the weakness and diffuse nature of components at large distance from the core , such that our images only contain portions of these features .
table [ source_par ] gives the parameters @xmath139 , @xmath132 , @xmath141 , @xmath131 , @xmath142 , and @xmath143 for each source .
parameters @xmath142 , @xmath143 , and @xmath141 are weighted averages of the values for individual components listed in tables [ q_param ] , [ b_param ] , and [ g_param ] , with weights inversely proportional to the uncertainty in apparent speed .
the intrinsic half opening angle , @xmath131 , is estimated as @xmath144 , where @xmath141 is the angle between the jet axis and the line of sight .
we have constructed the dependence between the doppler beaming factor , @xmath114 , and apparent speed ( fig .
[ v_doppler ] ) to compare derived parameters with results from the 2-cm vlba survey @xcite .
these authors determined doppler factors using the variability method of @xcite , with an intrinsic brightness temperature of @xmath145 k that they assume corresponds to the region of the vlbi core where the main variability at radio frequencies occurs @xcite . in figure [ v_doppler ]
the majority of points follow the expectation that @xmath146 and lie inside the `` @xmath147 cone '' . moreover , despite our sample consisting of blazars or blazar - like sources for which a higher lorentz factor is expected , the upper limit to @xmath98 corresponding to @xmath148 @xcite applies to our sample as well ( dotted curve in fig .
[ v_doppler ] ) .
figure [ h_gamma ] shows the distribution of lorentz factors of superluminal components . there is a significant scatter in @xmath1 for the quasars and bl lacs , and the lorentz factors of the radio galaxies are lower than for the other classes .
the distributions , in general , agree with the distribution of lorentz factors found in the 2-cm survey .
monte - carlo simulations for flux limited samples of radio sources predict a connection - although not direct correlation - between the lorentz factor and viewing angle of the jets : the higher the lorentz factor , the smaller is the viewing angle @xcite .
our sample of the blazars and blazar - like sources shows a significant correlation between 1/@xmath1 and @xmath100 ( coefficient of correlation 0.83 , fig .
[ gt ] ) ; this supports the above prediction .
small intrinsic bends in jets oriented very close to the line of sight should be greatly amplified in projection on the sky @xcite .
the expected morphology has not been confirmed by vlbi surveys .
for example , @xmath149-ray blazars , which exhibit the highest apparent speeds @xcite , do not possess more pronounced bends than sources not yet detected in @xmath149-rays @xcite .
estimates of the viewing angle obtained for our sample allow us to test the relation between morphology and viewing angle of the jets more directly .
we have determined the average projected position angle of the jet , @xmath139 , and its standard deviation , @xmath150 , for each source ( see table [ source_par ] ) . in sources with jet direction close to the line of sight ,
a small intrinsic change of the component trajectory should result in a significant bend in the projected path , leading to a large scatter in @xmath139 and yielding a high value of @xmath150 .
a similar test is to compare the projected opening angle with the viewing angle of the jets : jets viewed very close to the line of sight should have a wide projected opening angle on vlbi maps .
figure [ sigmat ] presents both plots .
the left panel shows the relationship between @xmath100 and @xmath150 .
although there is a general decrease in @xmath150 with increasing angle between the jet axis and line of sight ( coefficient of correlation @xmath130.58 ) , the relationship is very weak or absent for viewing angles @xmath151 ( coefficient of correlation @xmath130.18 ) .
the right panel shows that there is no connection between the projected opening angle and viewing angle of the jet ( coefficient of correlation @xmath130.16 ) .
this result implies that amplification of the projected size by a small angle between the jet axis and line of sight is partly ( or sometimes completely ) canceled by the proportionality between opening angle and @xmath152 ( see below ) .
possible reasons behind the lack of correlation between degree of apparent bending and small viewing angle include : ( 1 ) an inverse relationship between bulk lorentz factor and degree of intrinsic bending , given that the momentum of the jet is proportional to @xmath1 , ( 2 ) broader opening angles in slower jets , with filamentary structure appearing similar to bending of the axis , and ( 3 ) the resolution available transverse to the jet not being sufficient to detect bending in many high-@xmath1 objects .
the lorentz factor plays a significant role in determining the jet geometry . according to standard models of relativistic jets ,
the opening angle of the jet should be inversely proportional to the lorentz factor ( e.g. , * ? ? ?
* ) . in figure [ open_g ] ,
the estimated half opening angles are plotted versus the derived lorentz factors .
there is an obvious decrease of @xmath131 toward higher values of @xmath1 .
according to a @xmath58 test the _ observed _ dependence can be described by the relation @xmath153 rad , where @xmath154 ( solid line in fig .
[ open_g ] ) at a 2.5% level of significance ( @xmath58=6.875 , f=2 ) .
the current models for the formation of relativistic jets that employ confinement of a jet by magnetic forces ( e.g. * ? ? ?
* ; * ? ? ?
* ) do not yet specify the opening angle of the jet as a function of the model parameters .
such an expression , however , has been derived based on gas dynamics for a relativistic jet confined by pressure equilibrium with its surroundings @xcite . in this model
the opening angle depends on the lorentz factor of the flow and the ratio of the external pressure , @xmath155 , to the initial pressure , @xmath156 , of the plasma in the core region , @xmath157 . using equations ( 18 ) , ( 19a ) , and ( 19b ) in @xcite
, we calculate dependences of the opening angle on lorentz factor for different values of @xmath158 ( dotted curves in fig .
[ open_g ] ) .
the best fit of the _ observed _ dependence ( 5% level of significance ) coincides with the model in which @xmath159 ( @xmath58=4.586 , f=2 ) , corresponding to a pressure ratio in the core region @xmath160 .
figure [ ti ] presents the distribution of intrinsic brightness temperature of the jet components .
note that @xmath128 is measured at distances ranging from several parsecs to a few kiloparsecs from the core ( see fig .
[ ti ] , _ right panel _ ) , where the knots are generally optically thin .
the distribution peaks at @xmath161 k for the quasar knots and @xmath162 k for the bl lac knots , while for the radio galaxies the temperatures are evenly distributed between these values .
the maxima should be adjusted to slightly higher temperatures since for 30% of the quasar and bl lac components we have obtained only upper limits to the doppler factors .
comparison of these brightness temperatures with the intrinsic equipartion brightness temperature of the optically thick part of the jet ( @xmath163 k ) , thought to depend weakly on source parameters @xcite , implies a faster drop of the intrinsic temperature from the compact core region to the more extended structure in the bl lac jets .
one possible explanation for a lower intrinsic brightness temperature is the presence of a stronger magnetic field @xcite .
this would lead to more severe radiative losses and weaker jet components relative to the core in the bl lac objects , and perhaps a lower intrinsic brightness temperature in the cores as well . in figure [ dt ]
we plot the average doppler factor vs. average viewing angle ( _ left panel _ ) , and the average viewing angle vs. the intrinsic half opening angle of the jet for each source ( _ right panel _ ) .
the bold crosses indicate the average position of the quasars , bl lac objects , and radio galaxies , with the size of each cross equal to the 1@xmath17 uncertainty of the parameters .
although each subclass has a large scatter around the average values and statistically the difference in the parameters of the quasars and bl lac objects is negligible , the points form a continuous sequence on a 3-d plot of @xmath164 , @xmath100 , @xmath131 with : quasars ( @xmath143=23@xmath16511 , @[email protected]@xmath167 , @[email protected]@xmath167 ) @xmath169 bl lacs ( @[email protected] , @[email protected]@xmath167 , @[email protected]@xmath167 ) @xmath169 radio galaxies ( @[email protected] , @[email protected]@xmath167 , @[email protected]@xmath167 ) . according to the @xmath58 test ,
one component in the radio galaxy 3c 120 , five components in two bl lac objects , and 13 components in six quasars exhibit a statistically significant change in the proper motion with time corresponding to acceleration / deceleration with distance from the core .
therefore , in 9 out of 15 sources the apparent speed of individual components varies .
moreover , although 86 knots are classified as moving ballistically , @xmath170 have trajectories that we are unable to fit well by a polynomial of any order , causing us to suspect non - ballistic motion for these as well .
figure [ v_change ] shows the evolution of apparent velocity , including direction along the jet , derived from the best polynomial approximation of component positions across epochs for knots with a detected change in apparent speed .
we have analyzed the results to search for a general trend in the variation of apparent speed with distance from the core .
for each component , we compare the instantaneous apparent speed at different distances from the core with the average ( indicated in table [ speed ] ) , and assign a value representing the apparent speed at each angular distance of @xmath171 if the velocity is higher than the average , @xmath172 if it is lower than the average , and @xmath54 if it equals the average .
then we construct the distribution of such changes along the jet using the derived viewing angle for each component ( tables [ q_param]-[g_param ] ) , which allows deprojection of observed distances from the core .
the result is shown in figure [ h_vchange ] where the distribution above the @xmath32-axis represents accelerating components and the distribution under the @xmath32-axis indicates decelerating components at corresponding deprojected distances from the core .
figure [ h_vchange ] reveals that at distances larger than @xmath75 pc from the core an increase of apparent speed is more common . however , there are some jet components that undergo alternating periods of both acceleration and deceleration
. the latter could be a signature of helical motion ( e.g. , * ? ? ?
* ; * ? ? ?
* ) or pinch instabilities that cause the jet cross - section , pressure , and lorentz factor to oscillate with distance from the core ( e.g. , * ? ? ?
the cause of the accelerations could be bending combined with selection of objects with high doppler factors such that the mean angle to the line of sight of the region near the core is less than optimal for superluminal motion .
statistically , such jets are more likely to bend away from the line of sight , thus increasing their apparent speeds with distance from the core .
alternatively , the acceleration could be physical , caused by considerably higher energy density in relativistic particles than in rest mass @xcite or magnetic acceleration @xcite .
our data reveal the coexistence of both very fast and much slower ( but moving ) knots in the jets of the majority of the sources in the sample ( see fig .
[ ident ] ) .
diversity in the apparent speeds of jet features might reflect intrinsic variations in the pattern speed of disturbances in the jet flow . in the shock wave model for interpretation of the radio light curves and features propagating down a relativistic jet , a disturbance such as an increase in velocity or energy flux in a jet can create both forward and
reverse shocks ( e.g , * ? ? ?
both move away from the central engine , but the reverse shock has a lower velocity than does the forward shock . in this context
fast jet features can be associated with the forward shock and slow moving knots with the reverse shock .
analysis of the brightness of these features might reveal a prevalence of different types of shock waves in the jets for different classes of agns . for each knot in table [ speed ] ( except knots classified as trailing features ; see 6.6 ) we have computed parameter @xmath173 , which characterizes the flux of a knot relative to the core , where @xmath108 is the maximum observed flux density of the knot and @xmath174 is the average flux density of the core over epochs .
figure [ h_fr ] shows the distributions of the derived values of this parameter for fast ( @xmath175 ) and slow ( @xmath176 ) moving knots in the quasars , bl lac objects , and radio galaxies .
the separation into fast and slow features is not strictly defined , and the distributions do not change significantly if the dividing velocity is in the range from 2@xmath5 to @xmath55c .
the distributions of @xmath177 of fast knots in the quasars and bl lacs are different at 99.5% level of confidence ( @xmath178=5 ) according to the @xmath58 test .
the brightness of @xmath750% of the fast knots in the quasars is comparable to the brightness of the core ( @xmath1790.5 ) , while the brightness of fast knots in the bl lac objects never exceeds half the brightness of the core , and the distribution peaks at @xmath1801/4 .
the distributions of @xmath177 in the radio galaxies can not be classified due to the small number of objects .
the distributions for slow moving knots indicate the presence of two populations : bright ( @xmath1811 ) and faint ( @xmath1800.5 for the quasars and radio galaxies and @xmath1821 for the bl lacs ) .
the population of bright slow knots consists of jet features with subluminal apparent speeds that most likely represent stationary shocks in the jets .
this population is most prominent in the quasars .
the population of faint slow knots is prominent in the bl lac objects , including 39% of all identified jet features , while in the quasars such features comprise only 7% of detected knots .
about 70% of faint slow knots in the bl lac objects have flux within 25% to 100% of the flux of the core , significantly brighter than fast knots in these objects .
if fast knots represent forward shocks and slow knots correspond to reverse shocks , then the forward shock is stronger in the quasars and , perhaps , the radio galaxies , while the reverse shock dominates in the jets of bl lac objects
. a reverse shock will be strong relative to its corresponding forward shock when the disturbance is prolonged such that the faster flow enters the rear of the shock structure over an extended period of time .
otherwise , only the forward shock will be prominent and the knot will have a lorentz factor only slightly ( @xmath1836% ) less than that of the forward shock front ( e.g. , * ? ? ?
bl lac objects might therefore have more prolonged disturbances of lower amplitude than those in quasars . if jets possess spine - sheath structure @xcite , then an alternative explanation for the above effect could be that the power of high-@xmath1 spines in bl lacs is lower than in quasars .
the proposal that the nature of shock waves in quasars is different from that of bl lac objects has been suggested by @xcite , who interpreted the evolution of the flux and polarization of the quasar 3c 345 as forward shocks that have greater speed than the underlying jet .
this is in contrast to the bl lac object oj 287 , studied by @xcite , who modeled the polarization , kinematics , and x - ray variability as reverse shocks , which are slower than the underlying jet .
our data support the idea that this could be a key difference between the two classes of blazars .
numerical hydrodynamical simulations @xcite show that the interaction of disturbances in the jet flow with the underlying jet and/or the external medium can play a significant role in the variability of the jet emission .
in particular , multiple conical shocks can form behind a strong shock wave propagating down the jet .
these _ trailing _ shocks appear to be released in the wake of the primary superluminal component rather than ejected from the core .
the simulations show that the ratio between the apparent velocity of the main and trailing components is a function of distance along the jet : the closer to the core a trailing component first appears the slower it moves , although for a given trailing component the simulations predict deceleration with distance from the core .
trailing components are oblique shocks in three - dimensional models and should possess different polarization properties than the leading shock .
@xcite identified several knots in the jet of the radio galaxy 3c 120 as having the characteristics of trailing shocks .
our sequences of high resolution vlba images reveal a number of jet features in different sources that have properties matching those expected for this phenomenon .
these are ( see fig .
[ ident ] ) : @xmath184 behind knot @xmath63 in the radio galaxy 3c 111 ; @xmath185 behind @xmath186 found previously by @xcite in 3c 120 ; @xmath187 following @xmath188 , respectively , in the quasar 3c 273 ; @xmath189 behind @xmath190 in 3c 345 ; @xmath191 following @xmath192 , respectively , in cta 102 ; and @xmath193 behind @xmath194 in 3c 454.3
. components @xmath195 ( 3c 111 ) , @xmath196 ( 3c 273 ) , and @xmath197 and @xmath198 ( cta 102 ) emerge from bright knots at some distance from the core ( see fig .
[ ident ] ) , while the remainder are already trailing bright features from the first epoch of our observations .
all of them have apparent speeds less than the corresponding main component and exhibit different evpas than the fast feature ( see fig . [ map ] ) .
although our observations do not allow us to check every trailing component for deceleration , a decrease in apparent speed is pronounced for @xmath199 and @xmath185 in 3c 111 and 3c 120 , respectively , and can be inferred for @xmath197 in cta 102 ( see fig .
[ v_trail ] ) . for
@xmath197 in cta 102 a second - order polynomial does not satisfy the @xmath58 test but improves goodness - of - fit to the data significantly .
we characterize each trailing component by three parameters : @xmath200 is the average apparent speed listed in table [ speed ] , @xmath201 is the ratio of apparent velocities of trailing component to corresponding main component , and @xmath10 is the average deprojected distance from the core at which a trailing component is detected in our data ( @xmath10 is calculated using @xmath92 indicated in table [ speed ] and jet parameters given in table [ source_par ] ) .
figure [ trail ] ( _ left panel _ ) demonstrates that trailing components form in the wake of bright knots having values of @xmath1 extending from 5 to @xmath20220 and different values of @xmath201 ( all of them @xmath251 ) are observed for similar lorentz factors .
figure [ trail ] ( _ right panel _ ) shows that there is an increase of the apparent speed of trailing components with distance from the core ( coefficient of correlation 0.71 ) . according to @xcite ,
trailing components represent pinch waves excited by the main disturbance , and an increase of their speed at larger distance reflects acceleration of the expanding jet .
our data shown in figure [ trail ] imply such an acceleration of the underlying jet that appears not to depend strongly on the lorentz factor of the main disturbance , consistent with the conversion of internal energy into bulk kinetic energy that accompanies expansion .
figure [ trail ] ( _ left panel _ ) shows a possible inverse correlation between @xmath201 and lorentz factor of the main disturbance ( coefficient of correlation @xmath130.40 ) .
the correlation is most likely the result of selection effects .
detection of a trailing component close to a leading shock with a high lorentz factor is complicated by stretching of the longitudual size of the main component in the observer s frame by a factor @xmath203 owing to light - travel time delays .
detection of trailing components that lag greatly behind a main disturbance with a low lorentz factor is hampered by difficulties in associating with confidence trailing components to the leading knot .
for this reason , many of the subluminal or quasi - stationary features detected near the core ( @xmath36-components ) in the majority of sources in the sample could represent a superposition of many trailing components formed behind a number of superluminal features .
this possibility is valid only if the direction of the jet `` nozzle '' changes between ejections , since a second major disturbance passing through a trailing shock would destroy it .
@xcite have found that there is a systematic decrease in @xmath200 with increasing wavelength , which they suggest results from sampling different parts of the jet structure at different frequencies .
although this might be the case for average apparent speeds derived from surveys at different wavelengths , for individual sources with superluminal components detected at similar distances at different frequencies , the effect can be caused by the structure containing the leading perturbation plus slower trailing shocks being unresolved at the longer wavelengths .
perhaps the difference in the proper motions of the polarized and total intensity components @xmath204 and @xmath205 noted in the quasar 0528 + 134 ( see table [ speed ] ) is an example of this effect , with the polarized intensity image ( which reveals the leading compact component ) playing the role of the finer resolution of shorter wavelength observations .
our monitoring is unique in terms of the number of blazars observed in a regular manner at high angular resolution over 3 years .
the excellent time coverage allows us to determine the rate of superluminal ejections , thought to be controlled by activity in the cental engine .
we list in table [ source_par ] the rate , @xmath206 ( multiplied by the time dilation factor 1+z ) , of superluminal ejections over three successive years based on the results given in table [ speed ] . in figure [ eject ] we plot the ejection rate versus the average jet lorentz factor for blazars .
there is a trend suggesting a positive correlation between @xmath206 and @xmath1 ( coefficient of correlation 0.5 ) the formal @xmath207 test rejects the hypothesis that there is no relation between @xmath206 and @xmath1 at a confidence level @xmath61=0.05 .
however , the trend could be an artifact of the bias toward highly doppler boosted objects in our sample .
nevertheless , blazars form the largest class of identified @xmath149-ray sources @xcite , for which a connection between @xmath149-ray events and radio jet activity has been found @xcite .
if nonthermal flares are associated with ejections of new superluminal knots , as seems to be the case at mm wavelengths @xcite , then we would expect @xmath149-ray light curves to reflect the rate of ejections . in this case
, the possible correlation between the lorentz factor and ejection rate would imply that the most variable @xmath149-ray sources ( in terms of major flares per year ) should possess the highest lorentz factors .
this can be tested by @xmath149-ray light curves of blazars obtained by the gamma ray large area space telescope ( glast , expected to begin operation in 2007 ) , along with vlbi monitoring of the radio jets .
in this paper we have presented the entire data set collected during a three - year program of monitoring agns using the vlba , and have discussed the main results relating to the kinematics of jets .
the sequences of images reveal short timescales ( for some sources shorter than 2 months ) of variability of the jet structure and even more rapid variability in polarization .
this program illustrates the importance of intensive monitoring for understanding jet physics in superluminal radio sources .
we have measured the apparent speed of 106 features in the inner jets ( within 4 mas of the core ) of two radio galaxies , five bl lac objects , and eight quasars from 1998 march to 2001 april .
superluminal apparent speeds occur in 80% of the knots , 26% of which show statistically significant deviations from ballistic motion .
the majority of non - ballistic components undergo an increase of apparent speed with distance from the core , although local decelerations are observed in some cases alongside the general acceleration
. this could be the result of physical accelerations or from selection of sources whose angles to the line of sight @xmath208 near the core and closer to this value farther out .
many of the jets contain both very fast and much slower moving features , which might be explained as forward and reverse shocks , respectively .
it appears that fast features are pronounced in the jets of the quasars , while slower moving knots dominate in the bl lac jets .
this suggests a different nature of the main disturbances seen in the jets of the two classes of blazars .
the properties of 11% of the superluminal components are consistent with the characteristics of trailing shocks expected to form in the wake of strong disturbances in the flow . trailing components emerging from bright knots farther down the jet have a faster apparent speed than those generated near the core while a given trailing feature most likely decelerates along the jet . according to the numerical simulations by @xcite ,
the trailing components are caused by the triggering of pinch modes by the main disturbance .
the jet accelerates close to the core as internal energy is converted into bulk kinetic energy , leading to trailing shocks being faster farther downstream . in four sources ( 3c 120 , oj 287 , 3c 345 , and cta 102 ) the behavior of jet components appears to be affected by the interaction with the external medium (
see a ) . for each source
a number of components decelerate , change trajectory , brighten in total and polarized flux , and undergo a rotation of the evpa at the same distance from the core along an edge of the jet .
this implies the existence of gas clouds a few parsecs to @xmath2021 kpc ( deprojected ) from the central engine , intermediate between the locations of the very dense clouds of the broad line regions and more rarefied clouds of the narrow line regions .
using measurements from the vlbi images of parameters such as flux density , apparent speed , and size of components , we have estimated lorentz and doppler factors as well as viewing angles for superluminal knots and the opening angle of each jet .
this is a new method to define jet parameters , based on the assumption that the decay in flux of the superluminal components is caused by radiative losses rather than by cooling from expansion , and is subject to light - travel delays .
we demonstrate that at high radio frequencies these assumptions are most likely correct .
the derived parameters of the jets indicate that in our sample the quasars have the highest doppler factors ( @xmath164 ) and smallest viewing ( @xmath100 ) and opening ( @xmath131 ) angles , while the two radio galaxies possess significantly lower doppler factors , larger viewing angles , and wider opening angles despite their `` blazar - like '' radio properties .
this implies that in the 3-dimensional parameter space ( @xmath164,@xmath100,@xmath131 ) the radio galaxies , bl lacs , and quasars in our sample occupy different regions .
the regions of the quasars and bl lac objects partly overlap , while the radio galaxies are significantly distinguished from the blazars in all three parameters . since our sample is not a complete one , the major differences suggested by these results need to be investigated further with larger , complete samples .
the inferred relationship between the half opening angle and the lorentz factor agrees with the expectation of gas - dynamic models that predict smaller values of @xmath131 for higher lorentz factors and a dependence of @xmath131 on the ratio of the external and internal jet pressure .
the best approximation to the relation is very close to @xmath209 .
this does not , however , exclude collimation by magnetic pinching , which might produce a similar @xmath210 relation .
we have estimated the intrinsic brightness temperatures of jet components in the quasars , bl lacs , and radio galaxies on parsec scales , obtaining averages of @xmath211 k , @xmath212 k , and @xmath213 k , respectively .
comparison of these values with the equipartition brightness temperature of the optically thick part of the jets , @xmath214 k , suggests a stronger magnetic field in the bl lac objects . there is a possible positive correlation between the lorentz factor of the jet and ejection rate of superluminal components for the blazars
this can be tested with the @xmath149-ray light curves that will be measured by the glast mission .
* 3c 66a : * the radio jet consists of a bright , polarized core with weak , diffuse extended structure ( fig . [ alla ] ) .
the set of images allows us to trace components @xmath63 , @xmath215 , @xmath216 , and @xmath16 at many epochs ( fig .
[ ident ] ) .
they are characterized by moderate velocity , 2 - 5c , typical of those measured in bl lac objects at lower frequencies @xcite .
component @xmath63 is probably feature @xmath14 from @xcite , where it was thought to be stationary .
the centroids of @xmath217 and @xmath218 appear to move toward the core at mildly superluminal speeds ; however , these are extended features , hence the shifts in position could be due to changes in their brightness distributions .
this is supported in the case of @xmath218 , whose proper motion makes an angle of @xmath219 to the line between the core and the feature s centroid .
figure [ ident ] shows possible identifications of very weak components @xmath220 , and @xmath221 .
the designation of knots follows that of @xcite ; figure [ ident ] includes their data , marked by triangles .
@xmath37 knots move at a high apparent speed , @xmath222 .
the reality of this faster motion is supported by a region of enhanced polarized intensity that progresses rapidly downstream at the same epochs as when the high apparent speed is detected in the total intensity images ( see fig . [ 3c66a_pol ] ) .
it appears that the more slowly moving @xmath14 components and highly polarized , fast @xmath37 components coexist in the jet .
one of possible physical explanations for this dichotomy is a double shock structure , with the faster forward shock being weaker than slower reverse shock ( see 6.5 ) .
* 3c 111 : * the innermost jet of this galaxy exhibits strong activity .
at least 7 moving components ( @xmath223 , see fig . [ ident ] ) can be identified inside 1 mas of the core .
these knots move ballistically at a typical apparent speed @xmath224c .
there are two bright features near the core , @xmath217 and @xmath218 , that are stationary according to the @xmath58 test .
the most prominent feature of the jet is an extended , polarized blob northeast of the core .
the feature appears to be associated with a major outburst observed at 90 ghz in 1996 january @xcite , at 37 ghz in 1996 march , and at 22 ghz in 1996 august @xcite .
the first ejection around this time appears to have occurred earlier than the derived epoch of ejection @xmath225 ( see table [ speed ] ) , since complex structure of the core region was detected already in a 43 ghz vlbi map obtained in 1996 july @xcite .
this implies slower motion of the disturbance near features @xmath217 and @xmath218 , which agrees with the description of the structure by @xcite as `` a mix of dimming and flaring of stationary and moving emission components . '' in our observations we identify the brightest part of the outburst remnant as @xmath63 , which moves ballistically at a higher apparent speed than that of the innermost components .
knots @xmath199 and @xmath195 detected behind @xmath63 seem to belong to the same disturbance . at the last 3
epochs the parameters of @xmath63 , @xmath199 , and @xmath195 are obtained after tapering the images by reducing the weighting of the data from the longest baselines , and are not used to compute the jet velocities . *
0420@xmath13014 : * the parsec - scale jet of this quasar is strongly dominated by the vlbi core , which has a complex polarization pattern ( see figs .
[ allc ] and [ map ] ) , implying multi - component substructure in the core region .
we detect two moving components in the innermost part of the jet .
knot @xmath64 has slower motion near the core , @xmath77@xmath5 ( fig .
[ v_change ] ) , and accelerates significantly beyond 0.2 mas as its trajectory turns from @xmath226 to @xmath227 ( fig .
[ traj ] ) .
the bright component @xmath228 traced inside 0.15 mas from the core moves ballistically along @xmath229 at a similar apparent speed as that derived for @xmath64 near the core .
the behavior of the knots is consistent with the common curved trajectory for jet components found by @xcite . *
3c 120 : * our observations continue to trace components @xmath230 , @xmath231 , @xmath47 , @xmath186 , and @xmath232 detected in @xcite , and we identify new components @xmath233 , @xmath234 , @xmath235 , and @xmath236 ( fig .
[ ident ] ) .
components @xmath230 and @xmath231 move ballistically with the same apparent speed found previously , while @xmath47 has decelerated since earlier epochs
. components @xmath186 and @xmath185 represent the front section of the major disturbance designated in @xcite as _ complex o_. component @xmath186 brightens in total and polarized flux at @xmath73.4 mas , similar to the behavior seen for @xmath231 and @xmath47 @xcite . after flaring , @xmath186 and @xmath185 diverge in their behavior : @xmath186 accelerates and @xmath185 decelerates ( fig .
[ ident ] ) .
such a development is expected if @xmath185 is a trailing component that forms in the wake of the accelerating knot @xmath186 , which represents the major disturbance in the flow @xcite .
the most prominent feature at later epochs is @xmath233 , which evolves in the same manner as @xmath47 .
the components have nearly equal proper motions , [email protected] ( @xmath47 , @xcite ) and [email protected] mas yr@xmath90 ( @xmath233 ) , and similar trajectories ( fig .
[ traj ] ) .
they both undergo a flare in total and polarized flux at a location between 2.5 and 3.5 mas from the core and the evpas of both rotate during the flare ( fig .
[ 3c120_cloud ] ) .
@xcite explained these properties as the results of interaction between the jet and an interstellar cloud .
the behavior of component @xmath233 is consistent with this , which suggests that it interacts with the same cloud .
however , there is a significant shift ( @xmath237 mas ) between the projected positions of the peaks of the flares ( see fig . [ 3c120_cloud ] ) .
most likely , the shift is a consequence of the slightly more southern trajectory of @xmath233 ( fig .
[ traj ] ) . for a viewing angle @xmath238 and opening angle @xmath239
( see table [ source_par ] ) , the shift by 0.6 mas gives a lower limit to the size of the cloud of 1 pc at a deprojected distance from the vlbi core of @xmath77 - 8 pc .
* 0528 + 134 : * components @xmath194 and @xmath240 detected in @xcite are seen during the early epochs of our monitoring and four new components are ejected later ( fig .
[ ident ] ) .
all components have similar trajectories , with a sharp bend to the north beyond 0.5 mas from the core , as previously noted by other authors ( e.g. , * ? ? ?
the projected trajectories of the knots are shifted relative to each other ( see fig .
[ traj ] ) , which suggests different angles of ejection .
this is supported by a difference in the apparent speed near the core , from @xmath241 for the fastest and most eastern components to @xmath242 for the most northern component .
we find progressive acceleration in at least three components ( fig .
[ v_change ] ) .
the most northern knot , @xmath205 , which is strongly polarized and the brightest feature in the jet ( fig . [ map ] ) , shows significantly different motion of the total and polarized intensity peaks .
the polarized region , @xmath204 , moves ballistically ahead of the main concentration of total intensity , @xmath205 .
table [ speed ] lists the parameters of @xmath205 and @xmath204 separately .
we suggest that @xmath204 represents a leading shock of a strong disturbance with complex structure ( see 6.6 ) .
* oj 287 : * the jet appearance changes considerably during the observations ( fig .
[ allf ] ) , largely due to brightness variability of three prominent jet features , @xmath243 , and @xmath244 ( fig .
[ map ] ) . on average , @xmath217 and @xmath218 separate from the core at subluminal speeds ,
although significant fluctuations relative to linear motion are observed ( see fig .
[ oj 287_a ] ) .
component @xmath218 might be associated with @xmath63 , classified by @xcite as a stationary feature .
table [ speed ] gives the parameters of @xmath218 based on our data alone , and also based on these data combined with those of @xcite under the assumption that @xmath218 and @xmath63 are the same feature .
the latter is listed as @xmath245 in table [ speed ] .
the addition of the previous data decreases the proper motion by a factor of @xmath72 and in both cases the proper motion does not exceed zero at the 2@xmath17 level .
this suggests that the feature may fluctuate about a stationary position rather than translate downstream .
we identify several fast moving components in the jet .
knots @xmath246 ( the designation continues the scheme of * ? ? ?
* ) have very short lifetimes of @xmath71 yr at 43 ghz , and can be resolved only after they have passed @xmath218 ( see fig
. [ map ] ) .
however , a notable brightening of @xmath218 occurs at the epochs when a @xmath37 component likely reached @xmath218 ( see fig . [ ident ] ) .
the @xmath37 components show high superluminal speeds that are slightly slower than those of knot @xmath247 in @xcite .
knot @xmath247 is observed to accelerate and its motion at 22/15 ghz is measured @xmath72 times farther from the core than the locations of the @xmath37 components in our images .
a distinguishing feature of the jet is component @xmath244 , the formal solution for the motion of which is upstream , toward the core .
this motion is especially pronounced in the likely event that @xmath244 is associated with @xmath215 identified by @xcite , in which case @xmath244 also undergoes a significant change in position angle relative to the core ( fig .
[ traj ] ) .
the latter can be explained by swinging / precession of the jet nozzle or by a translation of the core , either of which agrees with changes in trajectory of components near the core from @xmath248 for @xmath64 @xcite to @xmath249 for @xmath250 .
the alignment between the innermost jet direction and the position of component @xmath244 suggests that the feature is the result of interaction between the jet flow and the external medium .
the upstream apparent motion of @xmath244 might then trace the location of an interstellar cloud ( projected on the sky ) when the intrinsic position angle of the jet changes . in this case , the dramatic fading of the feature toward the end of our observations ( see fig . [ allf ] ) could be connected with an intrinsic change of the inner jet direction ( component @xmath250 has the most southern projected position angle during 1995 - 2000 ) so that the jet flow misses the cloud .
this interpretation agrees with the findings of @xcite .
these authors used the radio reference frame image database data and determined that the projected position angle of the inner jet of oj 287 rotated clockwise by @xmath251 in 8 yr , in agreement with a ballistic precession model if the velocities of the components decrease beyond @xmath71 mas from the core .
we suggest that the interaction with an extragalactic cloud might cause the deceleration of components . from the derived parameters of the jet ( see table [ source_par ] ) , we place the cloud at a deprojected distance of @xmath252 pc from the core .
acceleration of @xmath247 detected by @xcite takes place between 0.2 and 1 mas from the core and reflects an increase in the apparent speed beyond @xmath217 and @xmath218 and before @xmath244 located at @xmath71 mas . *
3c 273 : * the most prominent features in the images ( see fig .
[ allg ] ) are an unpolarized core and strongly polarized moving knots @xmath64 and @xmath228 that both brighten at @xmath70.8 mas ( fig . [ 3c273_fluxb ] ) .
each component becomes strongly polarized just as it leaves the core region ( @xmath70.3 mas from the core , fig .
[ map ] ) while the core has a low polarization ( comparable with the noise level ) .
the similarity of the evpa and fractional polarization at 7 ( brighest polarized feature ) , 3 , and 1.3/0.85 mm suggests a low fractional polarization in the core at the high frequencies , where faraday effects are not important .
this implies that the core has intrinsically low polarization , which might be caused by a strongly turbulent magnetic field on scales smaller than the 7 mm synthesized beam .
after they flare , @xmath64 and @xmath228 fade dramatically and expand both along and transverse to the jet direction .
this results in the formation of new features @xmath253 and @xmath254 related to @xmath64 , and @xmath196 related to @xmath228 ( fig . [ ident ] ) .
the @xmath255 components might form in the interaction of relativistic shocks connected with features @xmath64 and @xmath228 and the underlying jet flow .
these components should have lower apparent speed than the main disturbances @xcite , but this is not the case for @xmath254 .
alternatively , the @xmath37 components are fairly extended features in a broad jet and the @xmath255 knots might represent sections of the jet where the velocity vectors lie at different angles to the line of sight .
the latter interpretation is supported by a difference in the projected trajectories of the @xmath37 and @xmath255 components ( fig .
[ traj ] ) and the strong dependence between the projected position angle and apparent speed of components , with the more northern knots being slower ( fig .
[ 3c273_cross ] ) . a substantial velocity gradient across the jet
is also expected according to a model where the underlying jet flow has a double - helix structure and a slower lorentz factor than that of disturbances @xcite .
the brightening of @xmath64 and @xmath228 at @xmath70.8 mas from the core might be connected with either ( 1 ) a change of the jet direction ( a slight curvature of the projected trajectories of both components is seen in fig .
[ traj ] at this distance ) , ( 2 ) an intensity peak of a threadlike pattern of the underlying flow where the local velocity vector bents toward the line of sight @xcite , or ( 3 ) interaction with the external medium . however , a stable polarization direction in @xmath228 before , during , and after the flare ( see epochs 1999.76 , 1999.93 , and 2000.07 in fig . [ allg ] ) does not support ( 3 ) .
at many epochs the images contain a diffuse , highly polarized ( up to 50% ) feature , @xmath63 , farther down the jet ( see fig .
[ 3c273_c1 ] ) , moving at an apparent speed of @xmath256 . according to the time of ejection , @xmath63 corresponds to knot @xmath257 found by @xcite .
if @xmath63 and @xmath257 are the same feature , it has maintained a nearly constant proper motion over @xmath75 yr .
* 3c 279 : * our data reveal motion of at least 8 components , @xmath258 ( our designations continue those of @xcite ) , with apparent speeds up to 17@xmath5 , three times higher than previously reported velocities measured since 1991 @xcite .
the identification of components in figure [ ident ] includes results at 43 ghz from @xcite ( triangles ) and @xcite ( squares ) .
components @xmath258 are shown in figure [ map ] at the last epoch of monitoring .
@xcite suggest that the remarkable change in apparent speed is caused by a shift in the direction of the jet `` nozzle '' while the lorentz factor remains constant .
this conclusion is supported by a gradual decrease of the projected position angle @xmath45 of each successively ejected component ( starting with @xmath190 ) . in this interpretation ,
the jet flow is characterized by @xmath259 while @xmath100 changes from 0.5@xmath167 for components with the lowest speed ( 5.5c for @xmath260 found by @xcite ) to 1.6@xmath167 for components with the highest speed .
it is also possible that component @xmath260 accelerates ( see fig .
[ ident ] ) due to an increase in the lorentz factor of the flow .
we designate this possible component as @xmath261 ; it corresponds to @xmath260 before the acceleration and to @xmath190 after the acceleration . as indicated in table [ accel ] , the change in the velocity component parallel to jet , @xmath79 , is significant .
another method , which we apply in this paper to estimate doppler and lorentz factors and viewing angle for each superluminal knot ( see 6 ) , suggests that both the lorentz factor and the angle between the jet axis and line of sight for @xmath258 are different from those of @xmath260 . in any event , these results are generally consistent with viewing angle playing the major role in the change of apparent speed .
contemporaneous to these changes in the inner jet , an increase in the apparent speed is observed for @xmath16 at @xmath73 mas from the core at epoch @xmath71998.2 @xcite .
our data , combined with previously published results at 43 ghz @xcite , show that variations in the trajectory and apparent speed of @xmath16 occurred at least twice between 1995 and 2001 .
this identifies three segments ( @xmath43 , @xmath255 , and @xmath5 ) of the trajectory ( denoted by dotted lines in fig .
[ traj ] ) where the motion is ballistic according to a @xmath58 test .
the ballistic sections of the trajectory define two epochs , [email protected] and [email protected] , when the apparent speed changes .
the values of @xmath200 corresponding to these paths are given in table [ speed ] using designations @xmath262 , and @xmath263 , respectively .
table [ speed ] shows that @xmath16 had a high apparent speed ( @xmath264 ) in 1995 and decelerated dramatically in 1996 , then accelerated again from 1998 until at least 2002 according to @xcite .
both periods of acceleration correspond to fading of the component , while during times of lower apparent speed @xmath16 brightened .
such behavior may be the result of helical motion , with the radius of the helix increasing outward .
* 1510@xmath13089 : * the radio emission of this quasar is strongly core dominated with usually weak extended jet structure .
our data indicate that the quasar possesses an ultra - relativistic jet with apparent speed up to @xmath746@xmath5
. however , even bimonthly observations are not sufficient to state that the identification of components shown in figures [ alli ] , [ ident ] , and [ 1510_bw4 ] is unique . on the other hand , this identification is supported by the total and polarized intensity images along with the total intensity light curves at 37 ghz @xcite and 14.5 ghz ( umrao data base ) .
we detect three moving components , @xmath64 ( fig .
[ alli ] ) , @xmath228 , and @xmath194 ( fig .
[ 1510_bw4 ] ) . in all three cases
we observe a similar polarization behavior in the core region when the component is emerging from the core - the core region is strongly polarized with evpa along the jet direction ( see fig .
[ alli ] ) . in the cases of components
@xmath228 and @xmath194 polarized features with similar evpa are subsequently observed propogating down the jet ( fig .
[ 1510_bw4 ] ) .
for all three components the time of the ejection falls on the rising branch of an outburst in the total intensity light curves ( an especially prominent flare is connected with the ejection of component @xmath228 ) .
the modeling of the jet structure shows two features , @xmath217 and @xmath218 , that fluctuate near positions 0.14 and 0.54 mas , respectively ( fig .
[ ident ] ) .
component @xmath217 is usually canot be resolved from a moving component at epochs when the latter is close to @xmath217 , while @xmath218 shifts downstream when a moving component is approching and returns back when the moving knot has passed by ( see fig . [ ident ] ) .
such a behavior of jet features formed by stationary shocks is seen in numerical simulations of jets with variable input flow velocities @xcite . at many
epochs , a feature , @xmath39 , with flux significantly above the noise is seen southeast of the core ( fig .
[ map ] ) , close to the position angle of the arcsecond scale jet @xcite . on the other hand ,
this component is located at a similar distance from the core as @xmath217 but in the opposite direction .
this suggests possible artificial generation of the knot owing to modeling of an actually slightly extended core using a point - like source . however , there is no correlation between the positions and fluxes of @xmath39 and @xmath217 .
the position angle of @xmath39 varies within the range 71@xmath167 to 168@xmath167 and the flux varies from 0.03 jy to 0.5 jy , while the values of @xmath45 and the flux of @xmath217 are fairly stable ( parameters of @xmath217 and @xmath218 in table [ speed ] are computed for epochs when the features are not confused with moving knots ) .
differences in the properties of @xmath39 and @xmath217 suggest that these features along with @xmath31 constitute a complex core region inside the 43 ghz synthesized beam . *
3c 345 : * at many epochs the core of this quasar is not the brightest feature of the jet ( fig .
[ allj ] ) .
the core is usually unpolarized at 43 ghz ( fig .
[ map ] ) , although the fractional polarization of the component nearest to the core , @xmath217 , can be as high as 20% .
the latter fact suggests the low polarization of the core is intrinsic rather than caused by faraday effects .
we detect seven superluminal knots with average apparent speeds in the range 9 - 24@xmath5 ( fig .
[ ident ] , where the component designations follow @xcite ) , and at least four of these exhibit acceleration or deceleration ( figs .
[ traj ] and [ v_change ] ) . for knots
@xmath265 statistically significant acceleration is found only for @xmath266 .
similar acceleration was reported for components @xmath16 and @xmath267 by @xcite and @xcite , although beyond 2 mas from the core @xmath267 decelerates as indicated by both @xmath79 and interpolation of its position relative to 1996.81 as observed by @xcite ( asterisks in fig .
[ ident ] ) .
it appears that @xmath260 repeats the stages of @xmath267 , although it experiences an earlier and greater deceleration , possibly signifying a more complex trajectory ( see fig .
[ traj ] ) .
the change in apparent speed and the twisted trajectory are also pronounced for @xmath190 .
although the historically extrapolated position of @xmath190 ( using the average apparent speed of @xmath268 ) agrees well with the measured position at epoch 1996.81 @xcite , it decelerates significantly from @xmath269 near the core to @xmath270 at 1 mas , and then accelerates beyond 1.5 mas .
the deceleration is accompanied by brightening of the knot and rotation of the evpa by @xmath271 , making it almost perpendicular to the jet direction ( see epochs 1998.76 and 1998.94 in fig .
[ allj ] ) .
in addition @xmath190 has the most southern position angle in the jet during the deceleration period ( @xmath272 ) , similar to that observed for @xmath260 by ( * ? ? ?
* @xmath273 ) .
@xcite interpreted a change they observed in apparent speed in terms of a flow with constant lorentz factor along a bent trajectory .
component @xmath190 expands as it propagates down the jet and leaves behind a complex structure that consists of knot @xmath189 moving at apparent speed @xmath274 ( fig .
[ ident ] ) plus two extended features , @xmath275 and @xmath276 , that form later at the northern and southern edges of the jet , respectively ( fig .
[ map ] ) .
the majority of moving features have a trajectory with zig - zags in the north - south direction ( although the net motion is to the west ) , an increase in flux , and rotation of the evpa at a distance @xmath70.8 - 1 mas from the core ( fig . [ 3c345_cflux ] ) .
these variations could be the result of interaction with the external medium , as suggested by @xcite .
we estimate the deprojected distance of this possible interaction as @xmath7300 pc . *
1803 + 784 : * the curved jet of this bl lac object has been interpreted as the result of helical structure @xcite .
our high resolution 43 ghz images show at least two superluminal components near the core , @xmath64 and @xmath228 .
these follow a similar curved path ( fig .
[ traj ] ) and have average apparent speeds of @xmath277 , with acceleration and deceleration along the trajectory ( fig . [ v_change ] ) . a passage from deceleration to acceleration and vice versa
is detected three times for @xmath64 , and locations of these transitions along the jet coincide with bends in the trajectory ( compare figs .
[ traj ] and [ v_change ] ) .
this behavior supports the hypothesis of helical structure of the jet .
@xcite classify the jet feature @xmath71.5 mas from the core as a stationary knot , @xmath39 . in our data
this feature appears to be associated with @xmath218 , and has an upper limit to its apparent speed of @xmath278 ( fig .
[ ident ] ) .
no motion is detected for bright knot @xmath217 near the core .
@xcite found that a superluminal component at 8.5 ghz only becomes visible at a distance @xmath61 mas from the core . moving component @xmath63
is detected beyond @xmath218 and has an apparent speed of @xmath270 , similar to that found by @xcite .
the fractional polarization in the jet is high .
for example , @xmath218 is polarized up to 30% with the evpa aligned with the jet direction .
in contrast , the evpa in the core region is essentiallyradial ( see fig .
[ map ] ) , supporting the inference of a toroidal magnetic field geometry @xcite , but only if the viewing angle is almost @xmath279 in the immediate vicinity of the core .
it is interesting to note that the viewing angle of the jet derived in 6 is indeed quite small . *
1823 + 568 : * the 43 ghz images of this bl lac object contain bright stable features @xmath243 , and @xmath244 within 0.5 mas of the core ( fig .
[ ident ] ) .
all of these components are strongly polarized , especially @xmath244 , where the fractional polarization reaches 25% ( fig .
[ map ] ) .
four components , @xmath280 and @xmath63 , detected in our images can be identified with components @xmath281 , and @xmath282 , respectively , from @xcite .
the locations of @xmath218 and @xmath244 are slightly farther from the core than @xmath283 and @xmath284 ( epoch 1995.58 ) .
this can be attributed to a shift in the core position at 43 ghz relative to 22 ghz due to opacity effects , in which case the positions of these features would be stable over a period of 5 years . despite excellent time coverage ( fig .
[ alll ] ) , we do not detect motion inside 1 mas of the core in our observations .
this has several possible explanations : 1 ) ejections of superluminal components are rare in this object , with mean time between ejections @xmath6 20 years in the observer s frame ( see table [ speed ] ) ; 2 ) moving components are much fainter than stationary features ; or 3 ) the innermost part of the jet is very close to the line of sight .
however , we do detect two moving components , @xmath63 and @xmath215 , beyond 1 mas from the core where the jet bends slightly to the west ( see fig .
[ map ] ) .
knot @xmath63 has a low apparent speed of 3.8c that appears to be essentially constant , since the linear extrapolation of its motion back to epoch 1995.58 agrees well ( within 0.03 mas ) with the position of knot @xmath282 in the image of @xcite . for the fastest moving feature in the jet , @xmath215 ( @xmath285 ) , extrapolation from our sequence produces a position of the component at epoch 1995.58 that is farther from the core by @xmath70.1 mas than was observed for @xmath286 .
this implies a slower apparent speed farther from the core .
motion of @xmath215 is not detected in our images after it passes @xmath63 ( fig .
[ ident ] ) .
* bl lac : * our data for bl lac have been analyzed in @xcite .
at least three new components are ejected during the monitoring program ( fig .
[ ident ] ) .
these , together with previously ejected knots observed by @xcite , form a sequence of superluminal features that appear to emerge from the core ballistically along different position angles , indicating a change in direction of the jet `` nozzle . ''
the trajectories of the knots exhibit curvature that is especially pronounced for components @xmath287 , and @xmath288 ( fig .
[ traj ] ) . despite different average apparent speeds , the components accelerate in a similar manner suggestive of a small change in the lorentz factor and a substantial change in direction ( see table [ accel ] and fig .
[ v_change ] ) .
the core region can be represented by two point - like features , @xmath63 and @xmath215 ( * ? ? ? * and fig .
[ map ] ) . in the inner jet two bright stationary knots , @xmath217 and @xmath218 ,
are observed at all epochs ( see figs .
[ ident ] and [ map ] ) .
* cta 102 : * the apparent velocity of jet components in this quasar is fairly high , @xmath289 , in agreement with previous findings @xcite .
we label newly detected moving components as @xmath290 and @xmath250 ( fig .
[ ident ] ) following the scheme of @xcite and rename @xmath291 in @xcite as @xmath240 based on its estimated ejection time .
components @xmath290 and @xmath250 have ballistic motion .
however , at @xmath7 0.2 mas from the core both knots split , forming more slowly moving components @xmath197 and @xmath198 behind the main features . at some epochs ( e.g. , fig .
[ map ] ) @xmath290 is significantly brighter than the core both in total and polarized intensity .
there are two features , @xmath217 ( near the core ) and @xmath14 ( @xmath72 mas from the core ) , that do not move according to a @xmath58 test .
component @xmath14 was detected by @xcite and classified as probably stationary , although @xcite associated the feature with different moving components . as noted by @xcite
, @xmath14 is strongly variable in brightness and located where the jet bends sharply to the south . in 1998 - 1999
, @xmath14 was bright , very diffuse , and showed a double structure that includes knot @xmath5 closer to the core ( see fig . [ map ] ) .
we detect no motion of @xmath5 ( fig .
[ ident ] ) .
both components fade dramatically by the end of the monitoring ( fig .
[ alln ] ) despite the approach of the growing inner part of the jet toward @xmath14 .
this behavior appears similar to that observed for component @xmath244 in oj 287 ( see above ) , where we associate the fading of @xmath244 with a change in the jet direction .
figure [ traj ] shows the trajectories of @xmath290 and @xmath250 , as well as @xmath64 and @xmath194 from @xcite , plus the position of @xmath14 during both sets of observations .
the location of @xmath14 fluctuates over the range 1 - 1.5 mas from the core in right ascension and within 1.4 - 1.6 mas ( to the south ) in declination .
the innermost jet shows a significant ( @xmath292 ) difference in projected position angle of @xmath290 and @xmath250 relative to @xmath293 , suggestive of wobbling of the jet `` nozzle '' and/or change of the vlbi core position .
these properties imply interaction between the moving flow and external medium at a deprojected distance @xmath71.4 kpc from the core that results in the formation of a diffuse feature . *
3c 454.3 : * this quasar has very complex structure at 43 ghz .
figure [ ident ] plots only bright components ( flux density @xmath2020.1 jy ) .
one of the most interesting features of the jet is highly polarized component @xmath14 .
at all epochs ( see figs .
[ allo ] and [ map ] ) the evpa of the brightness peak of @xmath14 aligns with the jet axis .
no change in separation of the component from the core has been seen during either previous observations @xcite or our current sequence of images . however , combining the data from the two studies reveals a drift of @xmath14 to the southwest ( see fig .
[ traj ] ) at an apparent speed of @xmath294 between 1995 and 2001 ( the component defined in this way is referred to as @xmath295 in table [ speed ] ) , while the size of the knot remains fairly stable at @xmath296 mas .
the absence of proper motion within each data set suggests that the observed difference in position of the feature ( @xmath70.2 mas in declination and @xmath70.1 mas in right ascension ) could be caused by a displacement to the northeast of the vlbi core between 1997.6 and 1998.2 . beyond @xmath14 ,
trajectories of moving components separate into two groups : northwestern ( @xmath297 ) and southwestern ( @xmath240 ) , with a difference between the projected trajectories that can reach @xmath298 ( fig .
[ traj ] ) .
this is consistent with the direction of the jet near the location of @xmath14 lying very close to the line of sight , thereby producing large differences in projected trajectories of moving components beyond @xmath14 even if the intrinsic trajectories differ only slightly .
the proper motions that we measure cover the entire range from 0.12 to 0.53 mas yr@xmath90 found previously in the jet of 3c 454.3 ( * ? ? ?
* ; * ? ? ?
* and references therein ) .
component @xmath250 accelerates markedly ( see table [ accel ] ) .
a similar acceleration is derived for a bright feature detected in 1998 beyond @xmath14 if we associate it with @xmath194 found by @xcite based on its brightness , position , and time of ejection .
this gives an increase in apparent speed of @xmath194 from 6@xmath5 near the core to 13@xmath5 beyond @xmath14 .
both of these components belong to the northwestern group ( fig .
[ traj ] ) .
northwestern components appear to accelerate significantly without a change in direction of motion while southwestern knots ( e.g. , @xmath240 ) show mostly a change in direction ( fig [ v_change ] ) . a diffuse feature ( designated in fig .
[ ident ] by the symbol @xmath299 ) is located in the northern part of the jet and appears to move after @xmath240 passes @xmath14 .
figure [ map ] displays component @xmath39 to the south of @xmath14 when @xmath240 approaches @xmath14 . as observed in other objects discussed above , more slowly moving knots @xmath193 and @xmath198 trail some of the bright fast components ( @xmath194 and @xmath250 , respectively ) .
figure [ 3c454_d ] shows knot @xmath38 @xmath76 mas from the core .
this can be associated with a feature ( @xmath300 ) imaged at longer wavelengths by @xcite .
according to @xcite @xmath300 accelerated in 1984 - 1989 from 16@xmath5 to 30@xmath5 and then decelerated to 20@xmath5 in 1992 .
our observations suggest a possible continuation of the deceleration such that the apparent speed of the knot has decreased to @xmath301 . this material is based on work supported by the national science foundation under grants no .
ast-0098579 and ast-0406865 .
gmez acknowledges support from the spanish ministerio de educacin y ciencia and the european fund for regional development through grant aya2004 - 08067-c03 - 03 .
the vlba is a facility of the national radio astronomy observatory , operated by associated universities inc . under cooperative agreement with the national science foundation
this research has made use of data from the university of michigan radio astronomy observatory , which is supported by the national science foundation and by funds from the university of michigan .
we are grateful to the referee for the thorough , thoughtful , and useful criticism .
agudo , i. , et al .
2001 , , 549 , 183 alef , w. , preuss , e. , kellermann , k. i. , & gabuzda , d. c. 1998 , in iau colloq .
164 , radio emission from galactic and extragalactic compact sources , ed .
zensus , j.a . ,
taylor , g.b . , & wrobel j.m .
, asp conf .
ser . , 144 , 129 aloy , m .- . , et al . 2003 , , 585 , l109 biretta , j. a. , cohen , m. h. , unwin , s. c. , & pauliny - toth , i. i. k. 1983 , nature , 306 , 42 blandford , r. d. & knigl , a. 1979 , , 232 , 34 bowker , a. h. & lieberman , g. j. 1972 , engineering statistics ( englewood cliffs , new jersey : prentice - hall , inc . ) brandt , s. 1999 , data analysis : statistical and computational methods for scientists and engineers ( springer : springler - verlag new york inc . ) britzen , s. & krichbaum , t. p. 1995 , in proc .
10th working meeting on european vlbi for geodesy and astronomy , ed .
r. lanotte & g. bianco ( matera : asi ) , 172 britzen , s. , witzel , a. , krichbaum , t. p. , campbell , r. m. , wagner , s. j. , & qian , s. j. 2000 , , 360 , 65 britzen , s. , witzel , a. , krichbaum , t. p. , macslow , t. , & matveyenko , l. i. 2001 , astronomy letters , 27 , 1 britzen , s. , witzel , a. , krichbaum , t. p. , qian , s. j. , & campbell , r. m. 1999 , , 341 , 418 burbidge , g. r. , jones , t. w. , & odell , s. l. 1974 , , 193 , 43 daly , r. a. & marscher , a. p. 1988 , , 334 , 539 denn , g. r. , mutel , r. l. , & marscher , a. p. 2000 , , 129 , 61 cawthorne , t. v. , and wardle , j. f. c. 1988 , , 332 , 696 gabuzda , d. c. , 2000 , in astrophysical phenomena revealed by space vlbi , eds
h. hirabayashi , p.g .
edwards , & d.w .
murphy , isas , tokyo , 121 gabuzda , d. c. , & cawthorne , t. v. 2000 , , 319 , 1056 gabuzda , d. c. , & chernetskii , v. a. 2003 , , 339 , 669 gabuzda , d. c. , sitko , m. l. , & smith , p. s. 1996 , , 112 , 1877 gabuzda , d. c. and sitko , m. l. 1994 , , 107 , 884 gabuzda , d. c. , mullan c. m. , cawthorne , t. v. , wardle , j. f. c. , & roberts , d. h. 1994 , , 435 , 140 gmez , j. l. , et al .
2002 , vlba scientific memo 30 ( nrao ) gmez , j. l. , marscher , a. p. , & alberdi , a. 1999 , , 522 , 74 gmez , j. l. , marscher , a. p. , alberdi , a. , jorstad , s. g. , & agudo , i. 2001 , , 561 , l161 gmez , j. l. , marscher , a. p. , alberdi , a. , jorstad , s. g. , & garca - mir , c. 2000 , science , 289 , 2317 gmez , j. l. , mart , j. m. a. , marscher , a. p. , ibnez , j. m. a. , & alberdi , a. 1997 , , 482 , 74 greaves , j. s. , et al .
2003 , , 340 , 353 hartman , r. c. , et al . 1999 , , 123 , 79 holland , w. s. , et al . 1999 , , 303 , 659 homan , d. c. , lister , m. l. , kellermann , k. i. , cohen , m. h. , ros , e. , zensus , j. a. , kadler , m. , & vermeulen , r. c. 2003 , , 589 , l9 homan , d. c. , ojha , r. , wardle , j. f. c. , roberts , d. h. , aller , m. f. , aller , h. d. , & hughes , p. a. 2001 , , 549 , 840 homan , d. c. , ojha , r. , wardle , j. f. c. , roberts , d. h. , aller , m. f. , aller , h. d. , & hughes , p. a. 2002a , , 568 , 99 homan , d. c. , wardle , j. f. c. , cheung , c. c. , roberts , d. h. , & attridge , j. m. 2002b , , 580 , 742 hughes , p. a. , aller , h. d. , & aller , m. f. 1991 , , 374 , 57 hughes , p. a. , aller , h. d. , & aller , m. f. 1985 , , 298 , 301 impey , c. d. & tapia , s. 1990 , 354 , 124 jorstad , s. g. , marscher , a. p. , lister , m. l. , stirling , a. m. , cawthorne , t. v. , gmez , j. l. , & gear , w. k. 2004 , , 127 , 3115 jorstad , s. g. marscher , a. p. , mattox , j. r. , aller , m. f. , aller , h. d. , wehrle , a. e. , & bloom , s. d 2001 , , 556 , 738 jorstad , s. g. , marscher , a. p. , mattox , j. r. , wehrle , a. e. , bloom , s. d. , & yurchenko , a. v. 2001 , , 134 , 181 kantowski , r. , kao , j. k. , & thomas , r. c. 2000 , , 545 , 549 kellermann , k. i. , lister , m. l. , homan , d. c. , vermeulen , r. c. , cohen , m. h. , ros , e. , kadler , m. , zensus , j. a. , & kovalev , y. y. 2004 , , 609 , 539 kellermann , k. i. , vermeulen , r. c. , zensus , j. a. , & cohen , m. h. 1998 , , 115 , 1295 lhteenmki , a. & valtaoja , e. 1999 , , 521 , 493 laing , r. a. , parma , p. , de ruiter , h. r. , & fanti , r. 1999 , , 306 , 513 leppnen , k. j. , zensus , j. a. , & diamond , p. j. 1995 , , 110 , 2479 lister , m. l. , kellermann , k. i. , vermeulen , r. c. , cohen , m. h , zensus , j. a. , & ros , e. 2003 , , 584 , 135 lister , m. l. 2001 , , 561 , 676 lister , m. l. & smith , p. s. 2000 , , 541 , 66 lister , m. l. & marscher , a. p. 1997
, , 476 , 572 lobanov , a. p. & zensus , j. a. 2001 , science , 294 , 128 lovell , j. 2000 , in astrophysical phenomena revealed by space vlbi , eds .
h. hirabayashi , p.g .
edwards , & d.w .
murphy , isas , tokyo , 301 marscher , a. p. & gear , w. k. 1985 , , 298 , 114 marscher , a. p. , jorstad , s. g. , gmez , j. l. , aller , m.f . ,
tersranta , h. , lister , m.l . , & stirling , a.m. 2002 , nature , 417 , 625 mead , a. r. g. , ballard , k. r. , brand , p. w. j. l. , hough , j. h. , brindle , c. , & bailey , j. a. 1990 , , 83 , 183 meier , d. l. , koide , s. , & uchida , y. 2000 , science , 291 , 84 nartallo , r. , et al .
1998 , , 297 , 557 pauliny - toth , i. i. k. 1998 , in iau colloq .
164 , radio emission from galactic and extragalactic compact sources , eds .
zensus , j.a . ,
taylor , g.b . , & wrobel j.m .
, asp conf .
ser . , 144 , 75 pearson , t. j. , et al . 1998 , in iau colloq . 164 , radio emission from galactic and extragalactic compact sources , eds .
zensus , j.a . ,
taylor , g.b . , & wrobel j.m .
, asp conf .
ser . , 144 , 17 rantakyr , f. t. , wiik , k. , tornikoski , m. , valtaoja , e. , & bth , l. b. 2003 , , 405 , 473 readhead , a. c. s. 1994 , , 426 , 51 readhead , a. c. s. , cohen , m. h. , pearson , t. j. , & wilkinson , p. n. 1978 , nature , 276 , 768 roberts , d. h. , wardle , j. f. c. , & brown , l. f. 1994 , , 427 , 718 ros , e. , zensus , j. a. , & lobanov , a. p. 2000 , , 354 , 55 savolainen , t. , wiik , k. , valtaoja , e. , jorstad , s. g. , & marscher , a. p. 2002
, , 394 , 851 shepherd , m. c. 1997 , in asp conf .
proc . , astronomical data analysis software and systems vi , ed .
g. hunt & h. e. payne ( san francisco : asp ) , 125 , 77 sitko , m. l. , schmidt , g. d. , & stein , w. a. 1985 , , 59 , 323 sokolov , a. s. , marscher , a. p. , & mchardy , i. 2004 , , 613 , 725 stirling , a. m. , et al .
2003 , , 341 , 405 tateyama , c. e. & kingham , k. a. 2004 , , 608 , 149 tateyama , c. e. , kingham , k. a. , kaufmann , p. , & de lucena , a. m. p. 2002
, , 573 , 496 tersranta , h. et al .
2004 , , 427 , 769 tersranta , h. & valtaoja , e. 1994 , , 283 , 51 tzioumis , a. k. et al .
2000 , , 98 , 36 vlahakis , n. & knigl , a. 2004 , , 605 , 656 wardle , j. f. c. , cawthorne , t. v. , roberts , d. h. , & brown , l. f. 1994 , , 437 , 122 wehrle , a. e. , et al .
2001 , , 133 , 297 wills , b. j. , wills , d. , breger , m. , antonucci , r. r. j. , & barvainis , r. 1992 , , 398 , 454 ccrcccccc 3c 66a&bl&0.444&[email protected]&@xmath1315&745&60&1.9&5 + 3c 111&rg&0.048&[email protected]&@xmath1310&2960&14&6.0&3 + 0420@xmath13014&q&0.915&[email protected]&@xmath138&6265&211&12.5&11 + 3c 120&rg&0.033&[email protected]&@xmath136&1520&36&7.6&3 + 0528 + 134&q&2.06&[email protected]&@xmath137&3930&131&18.6&6 + oj 287&bl&0.306&[email protected]&@xmath137&2330&96&8.1&5 + 3c 273&q&0.158&[email protected]&@xmath136&12800&943&44.8&45 + 3c 279&q&0.538&[email protected]&@xmath136&17240&1240&26.0&62 + 1510@xmath13089&q&0.361&[email protected]&@xmath136&2580&147&9.0&7 + 3c 345&q&0.595&[email protected]&@xmath139&5055&358&17.7&21 + 1803 + 784&bl&0.68&[email protected]&25&1490&94&5.2&5 + 1823 + 568&bl&0.664&[email protected]&@xmath1310&1580&131&7.9&5 + bl lac&bl&0.069&[email protected]&@xmath133&3810&184&9.5&9 + cta 102&q&1.037&[email protected]&@xmath133&4180&124&10.4&6 + 3c 454.3&q&0.859&[email protected]&@xmath135&5070&358&17.7&18 + ccrrrcccc 03/25/98(1998.23)&all&[email protected]&@xmath135&[email protected]&7&no&yes&no + 05/30/98(1998.41)&all&[email protected]&@xmath137&[email protected]&5&no&yes&no + 08/01/98(1998.58)&all&[email protected]&@xmath1311&[email protected]&5&no&yes&no + 10/05/98(1998.76)&all&[email protected]&@xmath134&[email protected]&8&no&yes&no + 12/10/98(1998.94)&all&[email protected]&@xmath139&[email protected]&8&no&yes&no + 02/11/99(1999.12)&all&[email protected]&@xmath1312&[email protected]&8&no&yes&yes + 04/29/99(1999.33)&all , bad weather&[email protected]&@xmath136&[email protected]&8&no&yes&yes + 07/18/99(1999.55)&all&[email protected]&@xmath1312&[email protected]&10&no&no&no + 10/06/99(1999.76)&all&[email protected]&@xmath133&[email protected]&6&no&yes&no + 12/05/99(1999.94)&all besides sc&[email protected]&@xmath1320&[email protected]&8&no&yes&yes + 01/24/00(2000.07)&all&[email protected]&@xmath135&[email protected]&6&no&no&no + 04/05/00(2000.26)&all&[email protected]&@xmath138&[email protected]&6&yes&yes&yes + 07/17/00(2000.54)&all&[email protected]&@xmath137&[email protected]&8&no&no&no + 10/01/00(2000.75)&all besides br&[email protected]&@xmath138&[email protected]&8&no&yes&no + 12/11/00(2000.95)&all besides mk&[email protected]&@xmath132&[email protected]&12&yes&yes&yes + 01/28/01(2001.08)&all , bad weather&[email protected]&@xmath131&[email protected]&15&yes&yes&no + 04/14/01(2001.28)&all&[email protected]&@xmath138&[email protected]&5&yes&no&no + lllllllll 0&@xmath130.1566&&&&&15&8.687&7.261 + 1&@xmath130.1575&@xmath130.0236&&&&14&6.344&6.571 + 2&@xmath130.1458&@xmath130.0233&@xmath130.0123&&&13&5.933&5.892 + 3&@xmath130.1448&@xmath130.0060&@xmath130.0129&@xmath130.0109&&12&5.731&5.226 + 4&@xmath130.1327&@xmath130.0065&@xmath130.0577&@xmath130.0106&0.0201&11&5.261&4.575 + @xmath47&@xmath303&@xmath78&@xmath304&@xmath305&@xmath306&@xmath178&@xmath65&@xmath60 + 0&@xmath130.6114&&&&&15&36.30&7.261 + 1&@xmath130.6164&@xmath130.1288&&&&14&5.194&6.571 + 2&@xmath130.6410&@xmath130.1293&@xmath130.0258&&&13&4.395&5.892 + 3&@xmath130.6421&@xmath130.1486 & 0.0264&@xmath130.0121&&12&4.284&5.226 + 4&@xmath130.6424&@xmath130.1486 & 0.0277&@xmath130.0121&@xmath130.0006&11&4.284&4.575 + lllllllll
0&@xmath130.9252&&&&&15&1009.0&7.261 + 1&@xmath131.384 & @xmath130.6840&&&&14&11.87&6.571 + 2&@xmath131.385 & @xmath130.6361&0.0415&&&13&10.13&5.892 + 3&@xmath131.367 & @xmath130.5981&@xmath130.0114&@xmath130.0545&&12&7.188&5.226 + 4&@xmath131.364 & @xmath130.7077&@xmath130.1515&0.0150&0.0791&11&4.242&4.575 + @xmath47&@xmath303&@xmath78&@xmath304&@xmath305&@xmath306&@xmath178&@xmath65&@xmath60 + 0&@xmath130.7363&&&&&15&215.4&7.261 + 1&@xmath131.018&@xmath130.4385&&&&14&7.262&6.571 + 2&@xmath131.029&@xmath130.3713&0.0683&&&13&4.002&5.892 + 3&@xmath131.027&@xmath130.3657&0.0649&@xmath130.0051&&12&3.988&5.226 + 4&@xmath131.026&@xmath130.3740&0.0527&@xmath130.0001&0.0065&11&3.975&4.575 + lrrrrrlrrrl 3c
66a&@xmath16&6&0.20&0.12&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath216&12&0.10&0.23&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath215&16&0.06&0.37&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath63&16&0.03&0.63&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath221&5&0.03&0.48&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath250&11&0.02&1.55&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath240&7&0.03&2.04&@[email protected]&1@xmath307&[email protected]&[email protected]&[email protected]&[email protected] + & @xmath217&16&0.02&1.19&@[email protected]&1@xmath307&[email protected]&[email protected]&[email protected] & + & @xmath218&17&0.05&2.50&@[email protected]&1@xmath307&@[email protected]&[email protected]&[email protected] & + 3c 111&@xmath221&5&0.23&0.43&[email protected]&1&[email protected]&[email protected]&[email protected]&[email protected] + & @xmath250&6&0.09&0.75&[email protected]&1&[email protected]&[email protected]&[email protected]&[email protected] + & @xmath290&8&0.16&0.94&[email protected]&1&[email protected]&[email protected]&[email protected]&[email protected] + & @xmath240&7&0.25&0.66&[email protected]&1&[email protected]&[email protected]&[email protected]&[email protected] + & @xmath194&8&0.20&0.93&[email protected]&1&[email protected]&[email protected]&[email protected]&[email protected] + & @xmath228&6&0.03&0.91&[email protected]&1&[email protected]&[email protected]&[email protected]&[email protected] + & @xmath64&4&0.05&1.08&[email protected]&1&[email protected]&[email protected]&[email protected]&[email protected] + & @xmath63&13&0.20&4.38&[email protected]&1&[email protected]&[email protected]&[email protected]&[email protected] + & @xmath199&16&0.09&3.73&[email protected]&2&[email protected]&[email protected]&[email protected]&[email protected] + & @xmath195&9&0.07&5.29&[email protected]&1&[email protected]&[email protected]&[email protected]&[email protected] + & @xmath217&17&0.58&0.12&[email protected]&0&[email protected]&[email protected]&[email protected] & + & @xmath218&12&0.37&0.27&[email protected]&0&[email protected]&[email protected]&[email protected] & + 0420@xmath13014&@xmath228&4&1.30&0.13&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath64&14&0.27&0.32&@xmath13147@xmath16525&3&[email protected]&[email protected]&[email protected]&[email protected] + 3c 120&@xmath236&4&0.09&1.03&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath234&8&0.27&1.28&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath235&5&0.13&1.10&@[email protected]&1@xmath307&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath233&10&0.17&1.95&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath186&16&0.27&3.28&@[email protected]&2&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath185&14&0.42&1.85&@[email protected]&2&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath47&7&0.16&2.85&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath231&16&0.16&5.49&@[email protected]&1@xmath307&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath230&7&0.06&6.95&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath217&15&0.22&0.19&@[email protected]&1&[email protected]&[email protected]&[email protected] & + 0528@xmath299134&@xmath205&11&1.33&0.13&77@xmath16516&2&[email protected]&[email protected]&[email protected]&[email protected] + & @xmath204&9&0.04&0.19&60@xmath16522&1&[email protected]&[email protected]&[email protected]&[email protected] + & @xmath221&6&0.43&0.16&108@xmath16522&1@xmath307&[email protected]&[email protected]&[email protected]&[email protected] + & @xmath250&13&0.32&0.34&86@xmath16520&3&[email protected]&[email protected]&[email protected]&[email protected] + & @xmath290&10&0.08&0.43&75@xmath16512&1@xmath307&[email protected]&[email protected]&[email protected]&[email protected] + & @xmath240&9&0.71&0.39&69@xmath16514&3&[email protected]&[email protected]&[email protected]&[email protected] + & @xmath194&6&0.65&0.53&68@xmath16517&1@xmath307&[email protected]&[email protected]&[email protected]&[email protected] + oj 287&@xmath250&4&0.28&0.39&@[email protected]&1@xmath307&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath290&6&0.28&0.39&@[email protected]&1@xmath307&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath240&5&0.27&0.33&@[email protected]&1@xmath307&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath217&17&0.24&0.10&@xmath1390@xmath16527&1@xmath307&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath218&17&0.41&0.30&@xmath13111@xmath1654&1@xmath307&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath245&24&0.36&0.28&@xmath13109@xmath1659&1@xmath307&@[email protected]&[email protected]&[email protected]&@xmath71970.5 + & @xmath244&16&0.12&1.00&@[email protected]&1&[email protected]&[email protected]&[email protected] & + 3c 273&@xmath228&10&3.29&1.16&@[email protected]&2&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath196&6&1.70&1.19&@[email protected]&1@xmath307&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath64&16&7.58&1.61&@[email protected]&4&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath253&17&4.04&1.16&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath254&8&0.55&2.67&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath217&17&1.14&0.16&@[email protected]&0&@[email protected]&[email protected]&[email protected] & + & @xmath218&11&5.74&0.78&@[email protected]&1@xmath307&@[email protected]&[email protected]&[email protected]&1984@xmath16512 + & @xmath63&10&0.32&8.27&@[email protected]&1@xmath307&@[email protected]&[email protected]&[email protected]&[email protected] + 3c 279&@xmath309&6&3.33&0.30&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath310&8&2.51&0.42&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath311&10&1.34&0.53&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath312&13&2.03&0.59&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath313&11&2.02&0.54&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath266&7&1.93&0.38&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath314&14&1.75&0.70&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath190&21&2.23&0.98&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath261&27&3.13&0.81&@[email protected]&3&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath16&46&1.28&3.54&@[email protected]&3&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath315&5&1.33&2.66&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath316&11&1.56&3.19&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath263&30&1.26&3.81&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath317&4&1.12&0.28&@[email protected]&1@xmath307&@[email protected]&[email protected]&[email protected]&[email protected] + 1510@xmath13089&@xmath194&5&0.24&0.55&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath228&4&0.10&0.64&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath64&4&0.04&0.78&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath217&15&0.28&0.14&@[email protected]&1@xmath307&@[email protected]&[email protected]&[email protected] & + & @xmath218&16&0.05&0.54&@[email protected]&1@xmath307&[email protected]&[email protected]&[email protected] & + & @xmath39&9&0.18&0.13&[email protected]&1@xmath307&17@xmath1651&[email protected]&[email protected] & + 3c 345&@xmath312&5&0.96&0.40&@[email protected]&1@xmath307&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath313&7&0.74&0.59&@[email protected]&1@xmath307&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath266&10&0.78&0.69&@[email protected]&2&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath314&14&0.35&0.83&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath190&17&0.40&1.23&@[email protected]&4&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath189&8&0.78&1.23&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath260&10&0.33&1.54&@[email protected]&3&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath267&16&0.14&2.89&@[email protected]&2&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath217&14&2.29&0.10&@[email protected]&0&@[email protected]&[email protected]&[email protected] & + 1803@xmath299784&@xmath228&13&0.12&0.38&@[email protected]&4&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath64&14&0.14&0.70&@[email protected]&4&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath217&17&0.21&0.28&@[email protected]&1@xmath307&[email protected]&[email protected]&[email protected] & + & @xmath218&17&0.11&1.44&@[email protected]&1@xmath307&@[email protected]&[email protected]&[email protected] & + & @xmath63&7&0.12&1.66&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + 1823@xmath299568&@xmath217&17&0.54&0.06&@[email protected]&0&@[email protected]&[email protected]&[email protected] & + & @xmath218&17&0.07&0.21&@[email protected]&0&@[email protected]&[email protected]&[email protected] & + & @xmath244&17&0.11&0.52&@[email protected]&0&@[email protected]&[email protected]&[email protected] & + & @xmath215&11&0.04&1.61&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath63&16&0.04&1.91&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + bl lac&@xmath288&10&0.10&2.27&@[email protected]&3&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath318&9&0.10&1.83&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath319&10&0.11&1.58&@[email protected]&3&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath320&10&0.33&2.20&@[email protected]&2&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath217&17&0.73&0.10&@[email protected]&0&@[email protected]&[email protected]&[email protected] & + & @xmath218&17&0.34&0.29&@[email protected]&1@xmath307&[email protected]&[email protected]&[email protected] & + cta 102&@xmath250&10&0.92&0.30&[email protected]&1&[email protected]&[email protected]&[email protected]&[email protected] + & @xmath198&5&0.24&0.32&[email protected]&1&[email protected]&[email protected]&[email protected]&[email protected] + & @xmath290&16&0.99&0.50&[email protected]&1&[email protected]&[email protected]&[email protected]&[email protected] + & @xmath197&12&0.29&0.52&[email protected]&1@xmath307&[email protected]&[email protected]&[email protected]&[email protected] + & @xmath217&16&0.59&0.12&[email protected]&0&@[email protected]&[email protected]&[email protected] & + & @xmath14&15&0.13&1.99&[email protected]&0&[email protected]&[email protected]&[email protected] & + 3c 454.3&@xmath250&7&1.64&0.31&@[email protected]&2&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath198&4&0.65&0.26&@[email protected]&0&@[email protected]&[email protected]&[email protected] & + & @xmath290&5&0.31&0.19&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath240&15&0.68&0.61&@[email protected]&2&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath194&7&0.85&1.37&@[email protected]&1&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath193&6&0.79&0.73&@[email protected]&1@xmath307&@[email protected]&[email protected]&[email protected]&[email protected] + & @xmath14&17&1.27&0.63&@[email protected]&1@xmath307&[email protected]&[email protected]&[email protected] & + & @xmath321&29&1.44&0.63&@[email protected]&1@xmath307&@[email protected]&[email protected]&[email protected] & + & @xmath46&14&0.37&0.86&@[email protected]&1@xmath307&@[email protected]&[email protected]&[email protected]&1992@xmath1655 + & @xmath38&16&0.11&5.80&@[email protected]&1&@[email protected]&[email protected]&[email protected] & + rrrrrrrrr 0420@xmath13014&@xmath64&14&[email protected]&[email protected]&[email protected]&[email protected] + 3c 111&@xmath199&16&[email protected]&[email protected]&@[email protected]&@[email protected] + 3c 120&@xmath186&16&[email protected]&@[email protected]&[email protected]&[email protected] + & @xmath185&14&[email protected]&@[email protected]&@[email protected]&@[email protected] + 0528@xmath299134&@xmath205&11&[email protected]&[email protected]&[email protected]&@[email protected] + & @xmath250&13&[email protected]&[email protected]&[email protected]&@[email protected] + & @xmath240&9&[email protected]&[email protected]&[email protected]&@[email protected] + 3c 273&@xmath64&16&[email protected]&@[email protected]&@[email protected]&[email protected] + & @xmath228&10&[email protected]&@[email protected]&[email protected]&[email protected] + 3c 279&@xmath261&27&[email protected]&@[email protected]&[email protected]&[email protected] + & @xmath16&46&[email protected]&@[email protected]&[email protected]&@[email protected] + 3c
345&@xmath266&10&[email protected]&@[email protected]&[email protected]&[email protected] + & @xmath190&17&[email protected]&@[email protected]&[email protected]&@[email protected] + & @xmath260&10&[email protected]&@[email protected]&@[email protected]&@[email protected] + & @xmath267&16&[email protected]&@[email protected]&@[email protected]&@[email protected] + 1803@xmath299784&@xmath228&13&[email protected]&@[email protected]&[email protected]&@[email protected] + & @xmath64&14&[email protected]&@[email protected]&[email protected]&@[email protected] + bl lac&@xmath288&10&[email protected]&@[email protected]&@[email protected]&@[email protected] + & @xmath319&10&[email protected]&@[email protected]&[email protected]&@[email protected] + & @xmath320&10&[email protected]&@[email protected]&@[email protected]&@[email protected] + 3c 454.3&@xmath250&7&[email protected]&@[email protected]&[email protected]&[email protected] + & @xmath240&15&[email protected]&@[email protected]&@[email protected]&@[email protected] + crrccccc 3c 66a&[email protected]&@xmath1313&500&30&1.5&0.7&0.5 + 3c 111&[email protected]&@xmath1310&803&6&2.0&0.9&0.9 + 0420@xmath13014&[email protected]&@xmath138&3504&112&10.5&4.0&2.7 + 3c 120&[email protected]&@xmath136&650&11&2.9&0.7&1.4 + 0528 + 134&[email protected]&@xmath137&1426&53&3.8&1.8&1.2 + oj 287&[email protected]&@xmath138&1115&32&4.4&2.5&2.6 + 3c 273&[email protected]&@xmath136&7178&655&35.9&7.6&14.1 + 3c 279&[email protected]&@xmath137&9695&432&24.2&3.5&3.1 + 1510@xmath13089&[email protected]&@xmath135&1517&36&7.6&1.5&3.4 + 3c 345&[email protected]&@xmath1310&1470&144&7.4&1.2&1.4 + 1803 + 784&[email protected]&30&1211&61&4.2&1.3&2.2 + 1823 + 568&[email protected]&@xmath1311&781&55&3.9&0.6&1.4 + bl lac&[email protected]&@xmath134&2709&128&6.8&2.2&2.7 + cta 102&[email protected]&@xmath139&2456&109&6.1&3.1&4.6 + 3c 454.3&[email protected]&@xmath136&3729&206&9.3&1.8&2.9 + lrrcrcc 0420@xmath13014&@xmath228&10.4&2.4&@xmath2517.5&4.69&1.09 + & @xmath64&11.5&3.6&@xmath2515.0&1.64&0.49 + 0528@xmath299134&@xmath205&27.4&0.4&@xmath2553.2&7.94&0.62 + & @xmath221&35.2&0.5&@xmath2563.6&2.17&0.125 + & @xmath250&25.5&1.2&39.2&2.58&0.34 + & @xmath290&21.1&3.3&17.3&0.38&0.20 + & @xmath240&27.5&0.8&48.5&3.72&0.34 + & @xmath194&39.5&0.6&69.2&1.07&0.05 + 3c 273&@xmath228&13.8&6.9&7.4&15.4&6.62 + & @xmath64&8.3&6.1&9.4&16.4&4.66 + & @xmath63&8.2&3.8&12.6&0.23&0.04 + 3c 279&@xmath309&17.1&2.9&19.7&20.2&2.65 + & @xmath310&18.2&2.3&23.9&25.3&2.38 + & @xmath311&17.4&2.8&20.0&9.37&1.19 + & @xmath312&16.3&3.3&17.6&16.1&2.57 + & @xmath313&16.6&2.5&22.0&23.3&2.52 + & @xmath266&12.9&2.2&@xmath2520.9&12.6&1.49 + & @xmath314&14.9&1.7&24.7&4.96&0.44 + & @xmath190&16.7&1.6&27.3&3.74&0.28 + & @xmath260&17.2&0.6&33.4&6.47&0.35 + & @xmath16&10.5&3.9&13.8&1.05&0.25 + 1510@xmath13089&@xmath194&30.1&1.3&41.3&1.10&0.03 + & @xmath228&35.2&1.5&39.0&0.73&0.02 + & @xmath64&47.9&1.6&34.2&0.21&0.01 + 3c 345&@xmath312&16.2&2.3&23.1&7.41&0.79 + & @xmath313&24.6&1.6&32.9&3.25&0.19 + & @xmath266&29.8&3.8&12.0&7.37&2.39 + & @xmath314&19.3&3.3&17.6&12.1&2.05 + & @xmath190&16.2&4.6&11.9&1.51&0.49 + & @xmath260&19.9&0.9&36.5&0.50&0.025 + & @xmath267&13.3&2.0&21.9&0.02&0.003 + cta 102&@xmath250&20.5&1.7&@xmath2529.8&3.54&0.37 + & @xmath290&15.7&3.0&18.8&14.4&3.29 + 3c 454.3&@xmath250&25.2&1.9&29.7&17.6&1.58 + & @xmath290&14.8&1.5&@xmath2525.8&1.58&0.18 + & @xmath240&20.0&3.9&13.9&4.54&1.49 + & @xmath194&13.4&3.7&15.4&24.7&6.76 + & @xmath14&14.2&0.2&28.2&6.76&0.66 + & @xmath38&10.2&1.7&18.7&0.03&0.007 + lrrcrcc 3c 66a&@xmath16&5.0&1.8&@xmath259.7&0.73&0.285 + & @xmath216&5.9&2.8&@xmath2510.9&0.44&0.141 + & @xmath215&7.7&2.8&@xmath2513.5&0.35&0.079 + & @xmath63&11.4&0.8&22.2&0.15&0.014 + & @xmath221&29.7&3.3&@xmath2515.4&0.21&0.037 + & @xmath250&26.1&4.0&12.1&0.12&0.032 + & @xmath240&27.6&2.5&22.1&0.02&0.002 + oj 287&@xmath250&20.7&4.7&10.6&0.41&0.117 + & @xmath290&12.3&3.4&16.1&1.31&0.183 + & @xmath240&19.9&1.8&@xmath2528.8&1.70&0.088 + 1803@xmath299784&@xmath228&17.2&2.2&23.7&0.34&0.038 + & @xmath64&15.7&3.6&16.1&0.25&0.054 + & @xmath63&20.8&0.8&38.7&0.02&0.001 + 1823@xmath299568&@xmath215&9.8&2.9&15.7&0.20&0.005 + & @xmath63&13.8&0.6&27.0&0.14&0.012 + bl lac&@xmath288&9.1&7.0&8.1&0.03&0.009 + & @xmath318&8.6&7.9&7.2&0.16&0.061 + & @xmath319&7.2&13.2&3.9&0.07&0.082 + & @xmath320&5.1&6.0&8.0&0.07&0.025 + lrrcrcc 3c 111&@xmath221&2.8&24.8&2.3&1.49&3.82 + & @xmath250&3.7&22.9&2.4&8.00&0.20 + & @xmath290&3.5&19.2&@xmath253.0&1.44&2.40 + & @xmath240&4.0&24.4&2.1&1.61&4.94 + & @xmath194&4.8&10.4&@xmath255.5&0.06&0.04 + & @xmath228&4.2&21.6&2.5&0.29&0.69 + & @xmath64&3.2&12.5&@xmath254.3&0.23&0.21 + & @xmath63&6.6&13.0&4.1&0.15&0.15 + 3c 120&@xmath236&6.6&23.0&1.7&0.19&0.85 + & @xmath234&4.3&22.3&2.3&0.75&1.89 + & @xmath233&4.9&16.9&3.2&0.11&0.16 + & @xmath186&7.0&22.0&1.7&3.69&15.2 + & @xmath47&4.0&21.4&2.5&0.88&1.91 + & @xmath231&4.5&19.8&2.7&0.31&0.61 + & @xmath230&3.9&11.3&4.9&0.03&0.02 + lrrrrrrrl 3c 66a&2.46&@xmath13162.0@xmath16525&[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&1.9 + 3c 111&0.222&[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&2.4 + 0420@xmath13014&5.925&@[email protected]&21@xmath16510&[email protected]&[email protected]&[email protected]&[email protected]&1.1 + 3c 120&0.145&@[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&2.4 + 0528 + 134&16.12&[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&7.0 + oj 287&1.59&@[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&1.3 + 3c 273&0.76&@[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&0.7 + 3c 279&3.10&@[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&3.5 + 1510@xmath13089&1.93&@[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&2.6 + 3c 345&3.50&@[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&3.7 + 1803 + 784&4.11&@[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&1.0 + 1823 + 568&4.00&@[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&0.5 + bl lac&0.31&@[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&1.1 + cta 102&6.92&[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&1.2 + 3c 454.3&5.48&@[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&[email protected]&1.9 + | we present total and polarized intensity images of 15 active galactic nuclei obtained with the very long baseline array at 7 mm wavelength at 17 epochs from 1998 march to 2001 april . at some epochs
the images are accompanied by nearly simultaneous polarization measurements at 3 mm , 1.35/0.85 mm , and optical wavelengths . here
we analyze the 7 mm images to define the properties of the jets of two radio galaxies , five bl lac objects , and eight quasars on angular scales @xmath0 milliarcseconds .
we determine the apparent velocities of 106 features in the jets . for many of the features we derive doppler factors using a new method based on comparison of timescale of decline in flux density with the light - travel time across the emitting region .
this allows us to estimate the lorentz factors , @xmath1 , intrinsic brightness temperatures , and viewing angles of 73 superluminal knots , as well as the opening angle of the jet for each source .
the lorentz factors of the jet flows in the different blazars range from @xmath2 to 40 with the majority of the quasar components having @xmath3 while the values in the bl lac objects are more uniformly distributed .
the brightest knots in the quasars have the highest apparent speeds , while the more slowly moving components are pronounced in the bl lac objects .
the quasars in our sample have similar opening angles and marginally smaller viewing angles than the bl lacs .
the two radio galaxies have lower lorentz factors and wider viewing angles than the blazars .
opening angle and lorentz factor are inversely proportional , as predicted by gas - dynamic models .
the brightness temperature drops more abruptly with distance from the core in the bl lac objects than in the quasars and radio galaxies , perhaps owing to stronger magnetic fields in the former resulting in more severe synchrotron losses of the highest energy electrons . in nine sources
we detect statistically meaningful deviations from ballistic motion , with the majority of components accelerating with distance from the core . in six sources we identify jet features with characteristics of trailing shocks that form behind the primary strong perturbations in jet simulations .
the apparent speeds of these components increase with distance from the core , suggestive of acceleration of the underlying jet . |
compact steep spectrum ( css ) sources are a population of powerful radio sources with projected linear size less than 20 kpc and steep high radio frequency spectrum @xmath4 . ]
( peacock & wall 1982 , fanti et al . 1990 , and review by odea 1998 * and fanti 2009 * ) .
kinematical studies of the hot spots and analysis of the high - frequency turnover in the radio spectrum due to radiative cooling imply ages for css sources in the range 10@xmath510@xmath6 yr ( e.g. , owsianik , conway & polatidis 1998 ; murgia et al .
the sub - galactic size of css sources has been used to argue that css sources are probably young radio sources , ( the ` youth ' model : fanti et al .
1995 ; readhead et al .
however , another interpretation attributes the apparent compactness of the css sources to being strongly confined by the dense ism in the host galaxy ( the ` frustration ' model : van breugel , miley & heckman 1984 ) .
spectroscopic observations of css sources provide evidence for abundant gas reservoirs in the host galaxies and strong interaction between the radio sources and the emission - line clouds @xcite .
some css sources have been observed to have high - velocity clouds ( as high as @xmath7 kms@xmath8 ) in the narrow - line region ( nlr ) , presumably driven by radio jets or outflows ; an example is 3c 48 @xcite .
in addition , many css sources show distorted radio structures , suggestive of violent interaction between the jet and the ambient interstellar medium @xcite .
the ample supply of cold gas in their host galaxies and their strong radio activity , which results in a detection rate as high as @xmath9 per cent in flux - density limited radio source surveys @xcite , make css sources good laboratories for the study of agn triggering and feedback .
3c 48 ( @xmath10 ) is associated with the first quasar to be discovered @xcite in the optical band .
its host galaxy is brighter than that of most other low redshift quasars .
the radio source 3c 48 is classified as a css source due to its small size and steep radio spectrum @xcite .
optical and nir spectroscopic observations suggest that the active nucleus is located in a gas - rich environment and that the line - emitting gas clouds are interacting with the jet material @xcite .
vlbi images @xcite have revealed a disrupted jet in 3c 48 , indicative of strong interactions between the jet flow and the dense clouds in the host galaxy . although some authors @xcite have suggested that the vigorous radio jet is powerful enough to drive massive clouds in the nlr at speeds up to 1000 km s@xmath8 , the dynamics of the 3c 48 radio jet have yet to be well constrained . due to the complex structure of the source , kinematical analysis of 3c 48 through tracing proper motions of compact jet components
can only be done with vlbi observations at 4.8 ghz and higher frequencies , but until now the required multi - epoch high - frequency vlbi observations had not been carried out . in order to study the kinematics of the radio jet for comparison with the physical properties of the host galaxy , we observed 3c 48 in full polarization mode with the vlba at 1.5 , 4.8 and 8.3 ghz in 2004 , and with the evn and merlin at 1.65 ghz in 2005 . combined with earlier vlba and evn observations ,
these data allow us to constrain the dynamics of the jet on various scales .
our new observations and our interpretation of the data are presented in this paper .
the remainder of the paper is laid out as follows .
section 2 describes the observations and data reduction ; section 3 presents the total intensity images of 3c 48 ; and section 4 discusses the spectral properties and the linear polarization of the components of the radio jet . in section 5 , we discuss the implications of our observations for the kinematics and dynamics of the radio jet .
section 6 summarizes our results . throughout this paper
we adopt a cosmological model with hubble constant @xmath11=70 km s@xmath8 mpc@xmath8 , @xmath12 , and @xmath13 . under this cosmological model
, a 1-arcsec angular separation corresponds to a projected linear size of 5.1 kpc in the source frame at the distance of 3c 48 ( @xmath10 ) .
the vlba observations ( which included a single vla antenna ) of 3c 48 were carried out at 1.5 , 4.8 , and 8.3 ghz on 2004 june 25 .
the evn and merlin observations at 1.65 ghz were simultaneously made on 2005 june 7 .
table [ tab : obs ] lists the parameters of the vlba , evn and merlin observations .
in addition to our new observations , we made use of the vlba observations described by worrall et al .
( 2004 ) taken in 1996 at 1.5 , 5.0 . 8.4 and 15.4
the total 12 hours of vlba observing time were evenly allocated among the three frequencies . at each frequency
the effective observing time on 3c 48 is about 2.6 hours .
the data were recorded at four observing frequencies ( ifs ) at 1.5 ghz and at two frequencies at the other two bands , initially split into 16 channels each , in full polarization mode .
the total bandwidth in each case was 32 mhz .
the detailed data reduction procedure was as described by worrall et al .
( 2004 ) and was carried out in aips .
we used models derived from our 1996 observations to facilitate fringe fitting of the 3c 48 data . because the source structure of 3c 48 is heavily resolved at 4.8 and 8.3 ghz , and missing short baselines adds noise to the image , the initial data were not perfectly calibrated .
we carried out self - calibration to further correct the antenna - based phase and amplitude errors .
this progress improves the dynamic range in the final images .
polarization calibration was also carried out in the standard manner .
observations of our bandpass calibrator , 3c 345 , were used to determine the r - l phase and delay offsets .
the bright calibrator source da 193 was observed at a range of parallactic angles and we used a model image of this , made from the stokes @xmath14 data , to solve for instrumental polarization .
our observing run included a snapshot observation of the strongly polarized source 3c 138 . assuming that the polarization position angle ( or the e - vector position angle in polarization images , ` evpa ' ) of 3c 138 on vlbi scales at 1.5 ghz is the same as the value measured by the vla , and we used the measured polarization position angle of this source to make a rotation of @xmath15 of the position angles in our 3c 48 data .
we will show later that the corrected evpas of 3c 48 at 1.5 ghz are well consistent with those derived from the 1.65-ghz evn data that are calibrated independently . at 4.8 ghz ,
3c 138 shows multiple polarized components ; we estimated the polarization angle for the brightest polarized component in 3c 138 from figure 1 in cotton et al .
2003 , and determined a correction of @xmath16 for the 3c 48 data .
after the rotation of the evpas , the polarized structures at 4.8 ghz are basically in agreement with those at 1.5 ghz . at 8.3
ghz the polarized emission of 3c 138 is too weak to be used to correct the absolute evpa ; we therefore did not calibrate the absolute evpas at 8.3 ghz .
the effective observing time on 3c 48 was about 8 hours .
apart from occasional rfi ( radio frequency interference ) , the whole observation ran successfully .
the data were recorded in four ifs .
each if was split into 16 channels , each of 0.5-mhz channel width .
in addition to 3c 48 we observed the quasars da 193 and 3c 138 for phase calibration .
3c 138 was used as a fringe finder due to its high flux density of @xmath09 jy at 1.65 ghz .
the amplitude of the visibility data was calibrated using the system temperatures , monitored during the observations , and gain curves of each antenna that were measured within 2 weeks of the observations .
the parallactic angles were determined on each telescope and the data were corrected appropriately before phase and polarization calibration .
we corrected the ionospheric faraday rotation using archival ionospheric model data from the cddis .
da 193 and oq 208 were used to calibrate the complex bandpass response of each antenna .
we first ran fringe fitting on da 193 over a 10-minute time span to align the multi - band delays .
then a full fringe fitting using all calibrators over the whole observing time was carried out to solve for the residual delays and phase rates .
the derived gain solutions were interpolated to calibrate the 3c 48 visibility data .
the single - source data were split for hybrid imaging .
we first ran phase - only self - calibration of the 3c 48 data to remove the antenna - based , residual phase errors .
next we ran three iterations of both amplitude and phase self - calibration to improve the dynamic range of the image . da 193
is weakly polarized at centimetre wavelengths ( its fractional polarization is no more than 1 per cent at 5 ghz , xiang et al .
2006 ) , and was observed over a wide range of parallactic angles to calibrate the feed response to polarized signals .
the instrumental polarization parameters of the antenna feeds ( the so - called ` d - terms ' ) were calculated from the da 193 data and then used to correct the phase of the 3c 48 data . the absolute evpa was then calibrated from observations of 3c 138 @xcite . a comparison between the apparent polarization angle of 3c 138 and the value from the vla calibrator monitoring program ( i.e. , @xmath17 at 20 cm wavelength ) leads to a differential angle @xmath18 , which was applied to correct the apparent orientation of the e - vector for the 3c 48 data .
after correction of instrumental polarization and absolute polarization angle , the cross - correlated 3c 48 data were used to produce stokes @xmath19 and @xmath20 images , from which maps of linear polarization intensity and position angle were produced .
the merlin observations of 3c 48 were performed in the fake - continuum mode : the total bandwidth of 15 mhz was split into 15 contiguous channels , 1 mhz for each channel .
a number of strong , compact extragalactic sources were interspersed into the observations of 3c 48 to calibrate the complex antenna gains .
the merlin data were reduced in aips following the standard procedure described in the merlin cookbook .
the flux - density scale was determined using 3c 286 which has a flux density of 13.7 jy at 1.65 ghz .
the phases of the data were corrected for the varying parallactic angles on each antenna .
magnetized plasma in the ionosphere results in an additional phase difference between the right- and left - handed signals , owing to faraday rotation .
this time - variable faraday rotation tends to defocus the polarized image and to give rise to erroneous estimates of the instrumental polarization parameters .
we estimated the ionospheric faraday rotation on each antenna based on the model suggested in the aips cookbook , and corrected the phases of the visibilities accordingly .
da 193 , oq 208 , pks 2134 + 004 and 3c 138 were used to calibrate the time- and elevation - dependent complex gains .
these gain solutions from the calibrators were interpolated to the 3c 48 data .
the calibrated data were averaged in 30-second bins for further imaging analysis .
self - calibration in both amplitude and phase was performed to remove residual errors .
the observations of oq 208 were used to calculate the instrumental polarization parameters of each antenna assuming a point - source model .
the derived parameters were then applied to the multi - source data .
we compared the right- and left - hand phase difference of the 3c 286 visibility data with the phase difference value derived from the vla monitoring program ( i.e. , @xmath21 at 20 cm , cotton et al .
1997b ; taylor & myers 2000 ) , and obtained a differential angle of @xmath22 .
this angle was used to rotate the evpa of the polarized data for 3c 48 .
after self - calibration , the evn and merlin data of 3c 48 were combined to make an image with intermediate resolution and high sensitivity .
the pointing centre of the merlin observation was offset by 0.034 arcsec to the west and 0.378 arcsec to the north with respect to the evn pointing centre ( table [ tab : obs ] ) . before combination , we first shifted the pointing centre of the merlin data to align with that of the evn data .
the lovell and cambridge telescopes took part in both the evn and merlin observations .
we compared the amplitude of 3c 48 on the common lovell cambridge baseline in the evn and merlin data , and re - scaled the evn visibilities by multiplying them by a factor of 1.4 to match the merlin flux .
after combination of evn and merlin visibility data , we performed a few iterations of amplitude and phase self - calibration to eliminate the residual errors resulting from minor offsets in registering the two coordinate frames and flux scales .
figures [ fig : mercont ] and [ fig : vlbimap ] exhibit the total intensity images derived from the merlin , vlba and evn data .
the final images were created using the aips and miriad software packages as well as the mapplot program in the caltech vlbi software package .
figure [ fig : mercont ] shows the total intensity image of 3c 48 from the merlin observations .
we used the multi - frequency synthesis technique to minimize the effects of bandwidth smearing , and assumed an optically thin synchrotron spectral index ( @xmath23 ) to scale the amplitude of the visibilities with respect to the central frequency when averaging the data across multiple channels .
the final image was produced using a hybrid of the clark ( bgc clean ) and steer ( sdi clean ) deconvolution algorithms .
the image shows that the source structure is characterized by two major features : a compact component contributing about half of the total flux density ( hereafter referred to as the ` compact jet ' ) , and an extended component surrounding the compact jet like a cocoon ( hereafter called the ` extended envelope ' ) .
the compact jet is elongated in roughly the north - south direction , in alignment with the vlbi jet .
the galactic nucleus corresponding to the central engine of 3c 48 is associated with vlbi component a @xcite .
it is embedded in the southern end of the compact jet .
the emission peaks at a location close to the vlbi jet component d ; the second brightest component in the compact jet is located in the vicinity of the vlbi jet component b2 ( figure [ fig : vlbimap ] : see section [ section : vlbimap ] ) .
the extended envelope extends out to @xmath01 arcsec north from the nucleus . at @xmath00.25 arcsec north of the nucleus , the extended component bends and diffuses toward the northeast .
the absence of short baselines ( @xmath24 ) results in some negative features ( the so - called ` negative bowl ' in synthesis images ) just outside the outer boundary of the envelope . the integrated flux density over the whole source is [email protected] jy ( very close to the single - dish measurement ) , suggesting that there is not much missing flux on short spacings .
the uncertainty we assign includes both the systematic errors and the _ r.m.s .
_ fluctuations in the image .
since the calibrator of the flux density scale , 3c 286 , is resolved on baselines longer than 600 k@xmath26 @xcite , a model with a set of clean components was used in flux density calibration instead of a point - source model .
we further compared the derived flux density of the phase calibrator da 193 from our observations with published results @xcite .
the comparison suggests that the flux density of da 193 from our merlin observation was consistent with that from the vlbi measurements to within 7 per cent .
we note that this systematic error includes both the amplitude calibration error of 3c 286 and the error induced by the intrinsic long - term variability of da 193 ; the latter is likely to be dominant .
the optical and nir observations @xcite detect a secondary continuum peak , denoted 3c 48a , at @xmath01 arcsec northeast of the optical peak of 3c 48 .
although merlin would be sensitive to any compact structure with this offset from the pointing centre , we did not find any significant radio emission associated with 3c 48a .
there is no strong feature at the position of 3c 48a even in high - dynamic - range vla images @xcite .
it is possibly that the radio emission from 3c 48a is intrinsically weak if 3c 48a is a disrupted nucleus of the companion galaxy without an active agn @xcite or 3c 48a is an active star forming region @xcite . in either case
, the emission power of 3c 48a would be dominated by thermal sources and any radio radiation would be highly obscured by the surrounding interstellar medium .
figure [ fig : vlbimap ] shows the compact radio jet of 3c 48 on various scales derived from the vlba and evn observations . table [ tab : figpar ] gives the parameters of the images .
the vlbi data have been averaged on all frequency channels in individual ifs to export a single - channel dataset .
the visibility amplitudes on each if have been corrected on the assumption of a spectral index of @xmath27 .
the total - intensity images derived from the 1.5-ghz vlba and 1.65-ghz evn data are shown in figures [ fig : vlbimap]-a to [ fig : vlbimap]-c .
the jet morphology we see is consistent with other published high - resolution images @xcite .
the jet extends @xmath00.5 arcsec in the north - south direction , and consists of a diffuse plume in which a number of bright compact knots are embedded .
we label these knots in the image using nomenclature consistent with the previous vlbi observations @xcite ( we introduce the labels b3 and d2 for faint features in the b and d regions revealed by our new observations ) . the active nucleus
is thought to be located at the southern end of the jet , i.e. , close to the position of component a @xcite .
the bright knots , other than the nuclear component a , are thought to be associated with shocks that are created when the jet flow passes through the dense interstellar medium in the host galaxy @xcite .
figure [ fig : vlbimap]-b enlarges the inner jet region of the 3c 48 , showing the structure between a and b2 . at @xmath00.05 arcsec north away from the core a , the jet brightens at the hot spot b. b is in fact the brightest jet knot in the vlbi images .
earlier 1.5-ghz images ( figure 1 : wilkinson et al . 1991 ; figure 5 : worrall et al .
2004 ) show only weak flux ( @xmath28 ) between a and b , but in our high - dynamic - range image in figure [ fig : vlbimap]-b , a continuous jet is distinctly seen to connect a and b. from component b , the jet curves to the northwest . at @xmath00.1 arcsec north of the nucleus
, there is a bright component b2 . after b2 , the jet position angle seems to have a significant increase , and the jet bends into a second curve with a larger radius . at 0.25 arcsec north of the nucleus
, the jet runs into a bright knot c which is elongated in the east - west direction . here
a plume of emission turns toward the northeast .
the outer boundary of the plume feature is ill - defined in this image since its surface brightness is dependent on the _ r.m.s .
_ noise in the image .
the compact jet still keeps its northward motion from component c , but bends into an even larger curvature . beyond component d2 ,
the compact vlbi jet is too weak to be detected . at 4.8 and 8.3
ghz , most of the extended emission is resolved out ( figures [ fig : vlbimap]-d to [ fig : vlbimap]-g ) and only a few compact knots remain visible .
figure [ fig : vlbimap]-e at 4.8 ghz highlights the core - jet structure within 150 pc ( @xmath030 mas ) ; the ridge line appears to oscillate from side to side . at the resolution of this image
the core a is resolved into two sub - components , which we denote a1 and a2 .
figure [ fig : vlbimap]-g at 8.3 ghz focuses on the nuclear region within 50 pc ( @xmath010 mas ) and clearly shows two well - separated components . beyond this distance the brightness of the inner jet is below the detection threshold .
this is consistent with what was seen in the 8.4- and 15.4-ghz images from the 1996 vlba observations @xcite .
figure [ fig : core ] focuses on the core a and inner jet out to the hot spot b. figure [ fig : core]-a shows the 1.5-ghz image from 2004 . unlike the image already shown in figure [ fig : vlbimap]-b , this image was produced with a super - uniform weighting of the @xmath29 plane ( see the caption of figure [ fig : core ] for details ) .
the high - resolution 1.5-ghz image reveals a quasi - oscillatory jet extending to a distance of @xmath040 mas ( @xmath0200 pc ) to the north of the core a. interestingly , figure [ fig : core]-b shows similar oscillatory jet structure at 4.8-ghz on both epoch 2004 ( contours ) and epoch 1996 ( grey - scale , worrall et al .
the consistency of the jet morphology seen in both 1.5- and 4.8-ghz images and in both epochs may suggest that the oscillatory pattern of the jet seen on kpc scales ( figure [ fig : vlbimap ] ) may be traced back to the innermost jet on parsec scales .
figure [ fig : core]-c shows the 8.3-ghz images in 2004 ( contours ) and 1996 ( grey scale , worrall et al .
2004 ) . in 1996 ( the image denoted ` 1996x ' ) the core is only slightly resolved into the two components a1 and a2 , while these are well separated by 3.5 mas ( 2 times the synthesized beam size ) in the 2004 observations ( ` 2004x ' ) .
direct comparison of 1996x and 2004x images thus provides evidence for a northward position shift of a2 between 1996 and 2004 .
figure [ fig : core]-d overlays the 2004x contour map on the 1996u ( 15.4 ghz , worrall et al .
2004 ) grey - scale map . neglecting the minor positional offset of a1 between 1996u and 2004x , possibly due to opacity effects
, this comparison of 1996u and 2004x maps is also consistent with the idea that a2 has moved north between 1996 and 2004 .
we will discuss the jet kinematics in detail in section [ section : pm ] .
in order to measure the spectral properties of the 3c 48 jet , we re - mapped the 4.99-ghz merlin data acquired on 1992 june 15 @xcite and compared it with the 1.65-ghz evn+merlin data described in the present paper .
the individual data sets were first mapped with the same _ uv _ range , and convolved with the same 40@xmath3040 ( mas ) restoring beam .
then we compared the intensities of the two images pixel by pixel to calculate the spectral index @xmath31 .
the results are shown in figure [ fig : spix ] .
component a shows a rather flat spectrum with a spectral index @xmath32 .
all other bright knots show steep spectral indices , ranging from @xmath33 to @xmath34 .
the extended envelope in general has an even steeper spectrum with @xmath35 .
spectral steepening in radio sources is a signature of a less efficient acceleration mechanism and/or the depletion of high - energy electrons through synchrotron / compton radiation losses and adiabatic losses as a result of the expansion of the plasma as it flows away from active acceleration region .
the different spectral index distribution seen in the compact jet and extended envelope may indicate that there are different electron populations in these two components , with the extended component arising from an aged electron population .
figure [ fig : merpol ] displays the polarization image made from the merlin data . the majority of the polarized emission is detected in the inner region of the source , in alignment with the compact jet .
the polarized intensity peaks in two locations .
the brightest one is near the vlbi jet component c , with an integrated polarized intensity of 0.31 jy and a mean percentage of polarization ( defined as @xmath36 , where @xmath37 represents the @xmath37th polarized sub - components ) of @xmath38 per cent .
the secondary one is located between vlbi jet components b and b2 , with an integrated polarized intensity of 0.23 jy and a mean degree of polarization @xmath39 per cent .
both of the two peaks show clear deviations from the total intensity peaks in figure [ fig : mercont ] .
these measurements of polarization structure and fractional polarization are in good agreement with those observed with the vla at 2-cm wavelength with a similar angular resolution @xcite .
the integrated polarized flux density in the whole source is [email protected] jy and the integrated fractional polarization is ( [email protected] ) per cent .
since the integrated polarized intensity is in fact a vector sum of different polarized sub - components , the percentage polarization calculated in this way represents a lower limit .
we can see from the image ( figure [ fig : merpol ] ) that the percentage of the polarization at individual pixels is higher than 5 per cent , and increases toward the south of the nucleus .
a maximum value of @xmath40 per cent is detected at @xmath00.045 arcsec south of the nucleus .
the fractional polarization ( @xmath41 per cent ) measured from our merlin observation at 18 cm is at least an order of magnitude higher than the vla measurement at 20 cm , although it is consistent with the values measured by the vla at 6 cm and shorter wavelengths .
this difference in the fractional polarization at these very similar wavelengths is most likely to be an observational effect due to beam depolarization , rather than being due to intrinsic variations in the faraday depth ( r. perley , private communication ) . the averaged polarization angle ( or evpa ) is @xmath42 in the polarization structure . on the basis of the new measurements of the rotation measure ( rm ) towards 3c 48 by mantovani et al .
( 2009 ) , i.e. , rm=@xmath43 rad m@xmath44 and intrinsic position angle @xmath45 @xcite , we get a polarization angle of @xmath46 at 1.65 ghz .
this result suggests that the absolute evpa calibration of 3c 48 agrees with the rm - corrected evpa within 3@xmath47 .
we show in figure [ fig : merpol ] the rm - corrected evpas . the evpas are well aligned in the north - south direction , indicating an ordered magnetic field in the faraday screen . at the resolution of the evn ,
most of the polarized emission from extended structures is resolved out . in order to map the polarized emission with modest sensitivity and resolution , we created stokes @xmath19 and @xmath20 maps using only the european baselines .
figure [ fig : vlbipol]-a shows the linear polarization of 3c 48 from the 1.65 ghz evn data .
the polarized emission peaks at two components to the east ( hereafter , ` c - east ' ) and west ( hereafter , ` c - west ' ) of component c. the integrated polarized flux density is 24.8 mjy in ` c - west ' and of 22.9 mjy in ` c - east ' , and the mean percentage polarization in the two regions is 6.3 per cent and 10.7 per cent respectively .
the real fractional polarization at individual pixels is much higher , for the reasons discussed above ( section 4.2.1 ) .
there is clear evidence for the existence of sub - components in ` c - west ' and ` c - east ' ; these polarized sub - components show a variety of evpas , and have much higher fractional polarization than the ` mean ' value .
the polarization is as high as 40 per cent at the inner edge of the knot c , which would be consistent with the existence of a shear layer produced by the jet - ism interaction and/or a helical magnetic field ( 3c 43 : cotton et al . 2003 ; 3c 120 : gmez et al .
component b , the brightest vlbi component , however , is weakly polarized with an intensity @xmath484.0 mjy beam@xmath8 ( percentage polarization less than 1 per cent ) .
the nucleus a shows no obvious polarization .
the 20-cm vlba observations were carried out in four 8-mhz bands , centred at 1404.5 , 1412.5 , 1604.5 and 1612.5 mhz . in order to compare with the 1.65-ghz evn polarization image
, we made a vlba polarization image ( figure [ fig : vlbipol]-b ) using data in the latter two bands .
this image displays a polarization structure in excellent agreement with that detected at 1.65 ghz with the evn , although the angular resolution is 3 times higher than the latter : the polarized emission mostly comes from the vicinity of component c and the fractional polarization increases where the jet bends ; the hot spot b and the core a are weakly polarized or not detected in polarization . the 1.65- and
1.61-ghz images show detailed polarized structure in the component - c region on a spatial scale of tens of parsecs : the polarization angle ( evpa ) shows a gradual increase across component c , with a total range of @xmath49 , and the percentage of polarization gradually increases from 5 per cent to @xmath5030 per cent from the western edge to the eastern edge at both ` c - west ' and ` c - east ' .
figure [ fig : vlbipol]-c and [ fig : vlbipol]-d show the 4.8- and 8.3-ghz polarization images made with the vlba data .
both images were made by tapering the visibility data using a gaussian function in order to increase the signal - to - noise of the low - surface brightness emission .
similar to what is seen in the 1.65 and 1.61-ghz images , component ` c - west ' shows a polarization angle that increases by @xmath51 across the component , but these images show the opposite sense of change of fractional polarization fractional polarization decreases from 60 per cent down to 20 per cent from the northwest to the southeast .
another distinct difference is that hot spot b shows increasing fractional polarization toward the higher frequencies , @xmath52 per cent at 4.8 ghz and @xmath53 per cent at 8.3 ghz in contrast with @xmath54 per cent at 1.6 ghz .
the difference in the fractional polarizations of b at 1.6/4.8 ghz and 8.3 ghz imply that a component of the faraday screen is unresolved at 1.6 and 4.8 ghz and/or that some internal depolarization is at work .
the non - detection of polarization from the core a at all four frequencies may suggest a tangled magnetic field at the base of the jet .
we found at all four frequencies that the polarization angles undergo a rotation by @xmath55 across the jet ridge line at both the ` c - east ' and ` c - west ' components .
there are four possible factors that may affect the observed polarization angle : ( 1 ) the calibration of the absolute evpas ; ( 2 ) faraday rotation caused by galactic ionized gas ; ( 3 ) faraday rotation due to gas within the 3c 48 system and ( 4 ) intrinsic polarization structure changes .
the correction of absolute evpas applies to all polarization structure , so it can not explain the position - dependent polarization angle changes at component c ; in any case , the fact that we see similar patterns at four different frequencies , calibrated following independent procedures , rules out the possibility of calibration error .
galactic faraday rotation is non - negligible ( section [ section : rm - reference ] ; @xmath43 rad m@xmath44 implies rotations from the true position angle of @xmath56 at 1.4 ghz , @xmath57 at 1.6 ghz , @xmath58 at 4.8 ghz and @xmath59 at 8.3 ghz ) , and means that we expect significant differences between the evpa measured at our different frequencies ; however , the galactic faraday screen should vary on much larger angular scales than we observe . only factors ( 3 ) and
( 4 ) , which reflect the situation internal to the 3c 48 system itself , will give rise to a position - dependent rotation of the evpas .
the evpa gradient is related to the gradient of the rm and the intrinsic polarization angle by : @xmath60 , where the first term represents the rm gradient and the latter term represents the intrinsic polarization angle gradient . if the systematic gradient of evpas , @xmath61 , were solely attributed to an rm gradient , then @xmath62 would show a strong frequency dependence ; on the other hand , if @xmath61 is associated with the change of the intrinsic polarization angle , there is no frequency - dependence .
we compared the @xmath63 at 1.6 and 4.8 ghz and found a ratio @xmath64 .
this number falls between 1.0 ( the value expected if there were no rm gradient ) and 8.8 ( the ratio of @xmath65 ) , suggesting that a combination of rm and intrinsic polarization angle gradients are responsible for the systematic gradient of evpas at c. accordingly , it is worthwhile to attempt to measure the rm in the vlbi components of 3c 48 .
the first two bands of the 20-cm vlba data ( centre frequency 1.408 ghz ) are separated from the last two bands ( centre frequency 1.608 ghz ) by 200 mhz , indicating a differential polarization angle of @xmath66 across the passband .
the low integrated rotation measure means that the effects of faraday rotation are not significant ( @xmath67 ) between 4.8 and 8.3 ghz , while the absolute evpa calibration at 8.3 ghz is uncertain ; moreover , the _ uv _ sampling at 8.3 ghz is too sparse to allow us to image identical source structure at 1.5 and 4.8 ghz .
therefore we used the 1.408 , 1.608 and 4.78 ghz data to map the rm distribution in 3c 48 .
we first re - imaged the stokes @xmath19 and @xmath20 data at the three frequencies with a common _ uv _ cutoff at @xmath68400 k@xmath26 and restored with the same convolving beam .
we tapered the @xmath29 plane weights when imaging the 4.78-ghz data in order to achieve a similar intrinsic resolution to that of the images at the two lower frequencies .
we then made polarization angle images from the stokes @xmath19 and @xmath20 maps .
the three polarization angle images were assembled to calculate the rm ( using aips task rm ) .
the resulting rm image is shown in figure [ fig : rm ] .
the image shows a smooth distribution of rm in the component - c region except for a region northeast of ` c - west ' . the superposed plots present the fits to the rm and intrinsic polarization angle ( @xmath69 , the orientation of polarization extrapolated at @xmath70 ) at four selected locations .
the polarization position angles at individual frequencies have multiples of @xmath71 added or subtracted to remove the @xmath72 amibiguity .
the errors in the calculated rms and @xmath69 are derived from the linear fits .
we note that the systematic error due to the absolute evpa calibration feeds into the error on the observed polarization angle .
all four fits show a good match with a @xmath65 law .
the fitted parameters at ` p4 ' in the ` c - east ' region are consistent with those derived from the single - dish measurements for the overall source @xcite .
the western component ( ` c - west ' ) shows a gradient of rm from @xmath73 rad m@xmath44 at ` p1 ' to @xmath74 rad m@xmath44 at ` p3 ' , and the intrinsic polarization angle varies from @xmath75 ( or @xmath76 ) at ` p1 ' , through @xmath77 ( or @xmath78 ) at ` p2 ' to @xmath79 at ` p3 ' .
this result is in good agreement with the qualitative analysis of the evpa gradients above .
a straightforward interpretation of the gradients of the rms and the intrinsic polarization angles is that the magnetic field orientation gradually varies across the jet ridge line ; for example , a helical magnetic field surrounding the jet might have this effect . an alternative interpretation for the enhancement of the rotation measurement at the edge of the jet is that it is associated with thermal electrons in a milliarcsec - scale faraday screens surrounding or inside the jet due to jet - ism interactions @xcite .
more observations are needed to investigate the origins of the varying rm and @xmath69 .
the hot spot b shows a much larger difference of evpas between 4.8 and 8.3 ghz than is seen in component c. this might be a signature of different rms at b and c. a rough calculation suggests a rm of @xmath80 rad m@xmath44 at b. the high rotation measure and high fractional polarization ( section 4.2 ) is indicative of a strong , ordered magnetic field in the vicinity of b. this might be expected in a region containing a shock in which the line - of - sight component of the magnetic field and/or the density of thermal electrons are enhanced ; in fact , the proper motion of b ( section 4.5 ) does provide some evidence for a stationary shock in this region . in order to make a quantitative study of the radiation properties of the compact vlbi components in 3c 48 , we fitted the images of compact components in the vlbi images from our new observations and from the vlba data taken in 1996 @xcite with gaussian models .
measurements from the 1996 data used mapping parameters consistent with those for the 2004 images .
table [ tab : model ] lists the fitted parameters of bright vlbi components in ascending frequency order .
the discrete compact components in the 4.8- and 8.3-ghz vlba images are well fitted with gaussian models along with a zero - level base and slope accounting for the extended background structure .
the fit to extended emission structure is sensitive to the _ uv _ sampling and the sensitivity of the image .
we have re - imaged the 1.5-ghz vlba image using the same parameters as for the 1.65-ghz evn image , i.e. , the same _ uv _ range and restoring beam . at 1.5 and 1.65
ghz , gaussian models are good approximations to the emission structure of compact sources with high signal - to - noise ratio , such as components a and b. for extended sources ( i.e. , components b2 to d2 ) whose emission structures are either not well modelled by gaussian distribution , or blended with many sub - components , model fitting with a single gaussian model gives a larger uncertainty for the fitted parameters .
in particular , the determination of the integrated flux density is very sensitive to the apparent source size .
the uncertainties for the fitted parameters in table [ tab : model ] are derived from the output of the aips task jmfit .
these fitting errors are sensitive to the intensity fluctuations in the images and source shapes . in most cases ,
the fitting errors for the peak intensities of gaussian components are roughly equal to the _ r.m.s . _ noise .
we note that the uncertainty on the integrated flux density should also contain systematic calibration errors propagated from the amplitude calibration of the visibility data , in addition to the fitting errors .
the calibration error normally dominates over the fitting error .
the amplitude calibration for the vlbi antennas was made from the measurements of system temperature ( @xmath81 ) at two - minute intervals during the observations combined with the antenna gain curves measured at each vlbi station .
for the vlba data , this calibration has an accuracy @xmath825 per cent of the amplitude scale . because of the diversity of the antenna performance of the evn elements , we adopted an averaged amplitude calibration uncertainty of 5 per cent for the evn data .
the positions of the vlbi core a1 at 4.8 , 8.3 and 15.4 ghz show good alignment within 0.4 mas at different frequencies and epochs .
the positions of the unresolved core a at 1.5 and 1.65 ghz show a systematic northward offset by 24 mas relative to the position of a1 at higher frequencies .
due to the low resolution and high opacity at 1.5 ghz , the position of a at this frequency reflects the centroid of the blended emission structure of the active galactic nucleus and inner 40-pc jet . the parameters that we have derived for the compact components a
, b and b2 in epoch 1996 are in good agreement with those determined by worrall et al .
( 2004 ) at the same frequency band .
the results for fitting to extended knots at 1.5 and 1.65 ghz are in less good agreement .
this is probably because of the different _ uv _ sampling on short spacings , meaning that the vlba and evn data sample different extended structures in the emission .
the integrated flux densities of the vlbi components a1 and a2 in 1996x ( 8.3 ghz ) are higher than those in 2004x ( 8.3 ghz ) by @xmath0100 per cent ( a1 ) and @xmath060 per cent ( a2 ) , respectively . the large discrepancy in the flux densities of a1 and a2 between epochs 1996x and 2004x
can not easily be interpreted as an amplitude calibration error of larger than 60 per cent since we do not see a variation at a comparable level in the flux densities of components b , b2 and d. although the _ total _ flux densities of css sources in general exhibit no violent variability at radio wavelengths , the possibility of small - amplitude ( @xmath82100 per cent ) variability in the vlbi core and inner jet components is not ruled out .
component a1 has a flat spectrum with @xmath83 between 4.8 and 8.3 ghz in epoch 2004 ; component a2 has a rather steeper spectrum with @xmath84 ( epoch 2004 ) .
the spectral properties of these two components support the idea that a1 is associated with the active nucleus and suffers from synchrotron self - absorption at centimetre radio wavelengths ; in this picture , a2 is the innermost jet .
the spectral indices of components b and b2 in epoch 2004 are @xmath85 ( b ) and @xmath86 ( b2 ) , respectively .
this is consistent with the measurements from the 1.65 and 4.99 ghz images ( figure [ fig : spix ] ) .
component d shows a relatively flatter spectrum in epoch 2004 with @xmath87 , in contrast to the other jet knots . while this spectral index is consistent with those of the shock - accelerated hot spots in radio galaxies , the flattening of the spectrum in d might also arise from a local compression of particles and magnetic field .
table [ tab : tb ] lists the brightness temperatures ( @xmath88 ) of the compact vlbi components a1 , a2 and b. all these vlbi components have a brightness temperatures ( @xmath88 ) higher than @xmath89k , confirming their non - thermal origin .
these brightness temperatures are well below the @xmath90 k upper limit constrained by the inverse compton catastrophic @xcite , suggesting that the relativistic jet plasma is only mildly beamed toward the line of sight .
the @xmath88 of a1 is about 3 times higher than that of a2 at 4.8 and 8.3 ghz , and the @xmath88 of a1 decreases toward higher frequencies .
together with the flat spectrum and variability of a1 , the observed results are consistent with a1 being the self - absorbed core harbouring the agn .
@xmath88 is much higher in 1996x than 2004x for both a1 and a2 , a consequence of the measured flux density variation between the two epochs .
the gaussian fitting results presented in table [ tab : model ] may be used to calculate the proper motions of vlbi components . in order to search for proper motions in 3c 48 ,
maps at different epochs should be aligned at a compact component such as the core @xcite .
however , thanks to our new vlbi observations we know that aligning the cores at 1.5-ghz is not likely to be practical , since the core structure appears to be changing on the relevant timescales . even at 4.8 ghz
, the core still blends with the inner jet a2 in epoch 1996c ( figure [ fig : core ] ) .
in contrast to these two lower frequencies , the 8.3-ghz images have higher resolution , better separation of a1 and a2 , and less contamination from extended emission . these make 8.3-ghz images the best choice for the proper motion analysis . in the following discussion of proper motion measurements we rely on the 8.3 ghz images .
we have already commented on the shift of the peak of a2 to the north from epochs 1996x to 2004x in figure [ fig : core ] .
a quantitative calculation based on the model fitting results gives a positional variation of 1.38 mas to the north and 0.15 mas to the west during a time span 8.43 yr , assuming that the core a1 is stationary .
that corresponds to a proper motion of @xmath91 mas yr@xmath8 ( minus mean moving to the west ) and @xmath92 mas yr@xmath8 , corresponding to an apparent transverse velocity of @xmath93 and @xmath94 .
the error quoted here includes both the positional uncertainty derived from gaussian fitting and the relative offset of the reference point ( _ i.e. _ , a1 ) .
that means that we detect a significant ( @xmath95 ) proper motion for a2 moving to the north .
the apparent transverse velocity for a2 is similar to velocities derived from other css and gps sources in which apparent superluminal motions in the pc - scale jet have been detected , e.g. , 3.39.7@xmath96 in 3c 138 @xcite .
we also searched for evidence for proper motions of the other jet knots .
the proper motion measurement is limited by the accuracy of the reference point alignment , our ability to make a high - precision position determination at each epoch , and the contamination from extended structure .
we found only a @xmath97 proper motion from b2 , which shows a position change of @xmath98 mas and @xmath99 mas in 8.43 yr , corresponding to an apparent velocity of @xmath100 to the northeast .
the measurements of the position variation of the hot spot b between 1996x and 2004x show no evidence for proper motion with @xmath101 mas yr@xmath8 and @xmath102 mas yr@xmath8 .
worrall et al . ( 2004 ) earlier reported a @xmath97 proper motion for b by comparing the the 1.5-ghz vlba image taken in 1996 with wilkinson et al s 1.6-ghz image from 11.8 years previously .
however , as mentioned above , the 1.5-ghz measurements are subject to the problems of lower angular resolution , poor reference point alignment and contamination from structural variation .
in particular , if we extrapolate the observed angular motion of a2 back , the creation of jet component a2 took place in 1984 , therefore in 1996 a2 would still have been blended with a1 in the 1.5-ghz image within @xmath103 beam .
the fitting of a gaussian to the combination of a1 and a2 at 1.5 ghz on epoch 1996 would then have suffered from the effects of the structural changes in the core due to the expansion of a2 . for these reasons
we conclude that the hot spot b is stationary to the limit of our ability to measure motions . for the other jet components ,
the complex source structure does not permit any determination of proper motions .
most css sources show double or triple structures on kpc scales , analogous to classical fr i or fr ii galaxies . however , some css sources show strongly asymmetric structures . at small viewing angles ,
the advancing jet looks much brighter than the receding one , due to doppler boosting .
the sidedness of radio jets can be characterized by the jet - to - counterjet intensity ratio @xmath104 . in vla images
@xcite , 3c 48 shows two - sided structure in the north - south direction . the southern ( presumably receding )
component is much weaker than the north ( advancing ) one . in vlbi images
( wilkinson et al . 1991 ; worrall et al . 2004 ; the present paper ) 3c 48 shows a one - sided jet to the north of the nucleus .
if the non - detection of the counterjet is solely attributed to doppler deboosting , the sideness parameter @xmath104 can be estimated from the intensity ratio of jet knots to the detection limit ( derived from the @xmath97 off - source noise ) . assuming the source is intrinsically symmetric out to a projected separation of 600 pc ( the distance of b2 away from a1 )
, the sideness parameter would be @xmath105 for b2 and b in the 1.5-ghz image ( figure [ fig : vlbimap]-a ) . in the highest - sensitivity image on epoch 2004c ( figure [ fig : vlbimap]-d ) ,
the off - source noise in the image is 40 @xmath106jy beam@xmath8 , so that the derived @xmath104 at component b could be as high as @xmath50900 . for a smooth jet which consists of a number of unresolved components , the jet - to - counterjet brightness ratio @xmath104 is related to the jet velocity ( @xmath107 ) and viewing angle ( @xmath108 ) by @xmath109 assuming an optically thin spectral index @xmath110 for the 3c 48 jet ( figure [ fig : spix ] ) , the sideness parameter @xmath111 estimated above
gives a limit of @xmath112 for the projected jet velocity in the line of sight .
using only the combination of parameters @xmath113 it is not possible to determine the kinematics ( jet speed @xmath107 ) and the geometry ( viewing angle @xmath108 ) of the jet flow .
additional constraints may come from the apparent transverse velocity , which is related to the jet velocity by @xmath114 . in section
[ section : pm ] we determined the apparent velocities for components b and b2 , @xmath115 , @xmath116 , and so we can combine @xmath113 and @xmath117 to place a constraint on the kinematics and orientation of the outer jet . the constraints to the jet velocity and source orientation are shown in figure [ fig : viewangle ] .
the results imply that the 3c 48 jet moves at @xmath118 along a viewing angle less than @xmath119 . as discussed in section [ section : vlbimap ] the bright jet knots define a sinusoidal ridge line .
this is the expected appearance of a helically twisted jet projected on to the plane of the sky .
helical radio jets , or jet structure with multiple bends , can be triggered by periodic variations in the direction of ejection ( e.g. , precession of the jet nozzle ) , and/or random perturbations at the start of the jet ( e.g. , jet - cloud collisions ) .
for example , the wiggles in the ballistic jets in ss 433 are interpreted in terms of periodic variation in the direction of ejection @xcite .
alternatively , small perturbations at the start of a coherent , smooth jet stream might be amplified by the kelvin - helmholtz ( k - h ) instability and grow downstream in the jet . in this case , the triggering of the helical mode and its actual evolution in the jet are dependent on the fluctuation properties of the initial perturbations , the dynamics of the jet flow , and the physical properties of the surrounding interstellar medium @xcite . in the following subsections we consider these two models in more detail .
we use a simple precession model @xcite , taking into account only kinematics , to model the apparently oscillatory structure of the 3c 48 radio jet .
figure [ fig : sketch ] shows a sketch map of a 3-d jet projected on the plane of the sky .
the x- and y - axis are defined so that they point to the right ascension and declination directions , respectively . in the right - handed coordinate system ,
the z - axis is perpendicular to the xoy plane and the minus - z direction points to the observer .
the jet axis is tilting toward the observer by an inclination angle of ( @xmath120 ) .
the observed jet axis lies at a position angle @xmath121 . in the jet rest frame ,
the kinematic equation of a precessing jet can be parameterized by jet velocity ( @xmath122 ) , half - opening angle of the helix cone ( @xmath123 ) and angular velocity ( or , equivalently , precession period @xmath124 ) . to simplify the calculations , we assume a constant jet flow velocity @xmath122 , a constant opening angle @xmath123 of the helix , and a constant angular velocity .
we ignore the width of the jet itself , so we are actually fitting to the ridge line of the jet .
the jet thickness does not significantly affect the fitting unless it is far wider than the opening angle of the helix cone .
( we note that , although we have measured lower proper motion velocities in b and b2 than the velocity in the inner jet a2 , this does not necessarily imply deceleration in the outer jet flow , since the brightening at b , and to some extent at b2 , may arise mostly from stationary shocks ; the proper motions of b and b2 thus represent a lower limit on the actual bulk motions of the jet . )
we further assume the origin of the precession arises from the central black hole and accretion disk system , so that ( @xmath125,@xmath126,@xmath127 ) can be taken as zero .
in the observer s frame the jet trajectory shown in the clean image can be acquired by projecting the 3-d jet on the plane of the sky and then performing a rotation by an angle @xmath121 in the plane of the sky so that the y - axis aligns to the north ( declination ) and x - axis points to the east ( right ascension ) .
in addition to the above parameters , we need to define a rotation sign parameter @xmath128 ( @xmath129 means counterclockwise rotation ) and jet side parameter ( @xmath130 means the jet moves toward the observer ) . since we are dealing with the advancing jet , the jet side parameter
is set to 1 .
based on our calculations , we found that a clockwise rotation pattern ( @xmath131 ) fits the 3c 48 jet . to estimate the kinematical properties of the precessing jet flow , we use the proper motion measurements of component a2 as an estimate of the jet velocity and orientation ( figure [ fig : viewangle ] ) .
we have chosen a set of parameters consistent with the curve for @xmath132 and a viewing angle of @xmath133 .
other combinations of angles to the line of sight and velocities give qualitatively similar curves .
for example , if we use a lower flow speed instead , a similar model structure can be produced by adjusting other parameters accordingly , e.g. by increasing the precessing period by the same factor .
the high - resolution vlbi images ( figure [ fig : core ] ) show that the innermost jet aligns to the north .
so an initial position angle @xmath134 should be a reasonable estimate .
the vlbi images ( figure [ fig : vlbimap ] ) suggest that the position angle of the jet ridge line shows an increasing trend starting from the hot spot b. moreover , we found that a model with a constant position angle does not fit simultaneously to both the inner and outer jet . to simplify the calculation
, we introduced a parameter @xmath135 to account for the increasing position angle in the outer jet .
the fitted jet ridge line is shown ( thick green line ) in the upper panel of figure [ fig : helicalfit ] overlaid on the total intensity image .
the assumed and fitted parameters are listed in table [ tab : helicalfit ] .
the modelled helix fits the general wiggling jet structure with at least two complete periods of oscillation .
the fitted opening angle of @xmath136 suggests that the line of sight falls outside the helix cone .
the initial phase angle @xmath69 is loosely constrained ; it is related to the reference time of the ejection of the jet knot , @xmath137 .
the fits suggest that the reference time is @xmath138 yr . in the presence of the gradual tilting of the jet axis as well as the helical coiling around the jet axis ,
the fits most likely represents a superposition of the precession of the jet knots and the nutation of the jet axis , analogous to ss 433 ( e.g. katz et al .
1982 ; begelman , king & pringle 2006 ) . the fitted period of 3500 yr is then a nutation period , about 0.4 times the dynamical time scale of the jet , assuming a flow speed of @xmath139 , while the precession period is much longer . from the rate of the jet axis tilting , we estimate a precession period of @xmath140 yr . the ratio of the estimated precession period to the nutation period is 57:1 , 2.2 times the ratio in ss 433 ( which has a 162-day periodic precession and 6.3-day nodding motion : see begelman , king & pringle 2006 and references therein ) .
the precessing jet model predicts a smooth structure on small scales , and a constant evolution of the wavelength so long as the jet kinetic energy is conserved and the helix cone is not disrupted ( the opening angle of the helix cone is constant ) .
however , the real 3c 48 jet probably does not conserve kinetic energy , as it is characterized by a disrupted jet and violent jet - ism interactions .
in particular , the inner - kpc jet is seen to be physically interacting with a massive gas system , and the observed blue - shifted nir clouds could be driven by the radio jet to move at velocities up to 1000 km s@xmath8 @xcite .
the 3c 48 radio jet thus might lose a fraction of its kinetic energy , resulting in a slowing down of the jet flow and the shrinking of the wavelength in the outer jet , assuming that the precessing periodicity is not destroyed .
we next investigate the interpretation of a hydrodynamic or magnetized jet instability for a helical structure @xcite .
we used the simple analytic model described in steffen et al .
( 1995 ) to fit to the helical jet trajectory in 3c 48 .
the kinematic equations of this toy model are solved on the basis of the conservation of kinetic energy @xmath141 and the specific momentum in the jet motion direction ( case 2 : steffen et al .
it is in fact identical to the isothermal hydrodynamic model @xcite under the condition of a small helix opening angle .
model fitting with an adiabatically expanding jet can basically obtain similar helical twisting jet as well , but the initial amplitude growth is much faster @xcite than that of the isothermal jet . in this analysis
we confine our discussion to the isothermal case .
to make the calculations simple but not to lose generality , we used similar assumptions to those of model 1 on the jet kinematics and geometry .
( we should note that although we used an apparent velocity @xmath142 with same value in model 1 , the jet speed @xmath122 in the k - h model is the pattern speed , and therefore the real flow speed and the viewing angle in the k - h model are more uncertain than for the ballistic case . ) in addition , we assume that the initial perturbations originate from a region very close to the central engine .
the calculations thus start from an initial distance of zero along the jet axis and a small displacement @xmath143 in the rotation plane away from the jet axis .
moreover , we assumed an initial position angle @xmath144 , and again introduced a rate @xmath145 to explain the eastward tilting of the jet axis .
the half opening angle , which is a parameter to be fitted , is assumed constant .
this assumption is plausible since the jet width seems not to change much within 0.5 arcsec , indicating that the trajectory of the jet is not disrupted even given the occurrence of a number of jet - ism interactions .
in addition to the above morphological assumptions , the model also assumes the conservation of specific momentum and kinetic energy @xmath146 along the jet axis .
the conservation of specific momentum is equivalent to a constant velocity along the jet axis if mass loss or entrainment are negligible .
the combination of the conservation of specific momentum and kinetic energy along the jet axis results in a constant pitch angle along the helical jet .
furthermore , the constant jet opening angle and pitch angle lead to a helical geometry in which the oscillatory wavelength linearly increases with time .
the parameter @xmath143 controls how fast the wavelength varies ( equation 12 : steffen et al .
the model describes a self - similar helical trajectory with a number of revolutions as long as the helical amplitude is not dampened too rapidly .
the modelled curve is exhibited in the lower panel of figure [ fig : helicalfit ] .
the assumed and fitted parameters are listed in table [ tab : helicalfit ] . as mentioned above
, this k - h instability model predicts that , when the helical amplitude is not dampened and the opening angle @xmath123 is small ( @xmath147 ) , the oscillating wavelength ( or period ) along the jet axis increases linearly with time .
the fits give an initial wavelength of 60 mas and initial period of 370 yr . the period increases to @xmath148 yr at the end of the plot window of 9000 yr . the fitted curve displays more oscillations in the inner part of the jet and smoother structure in the outer part , due to the decreasing angular velocity downstream . the initial transverse distance @xmath143 represents the location where the k - h instability starts to grow in the surface of the jet .
it is associated with the varying rate of the wavelength .
a value of @xmath149 mas corresponds to a projected linear distance of 9.2 pc off the jet axis .
as discussed above , the major discrepancy between the helical model and the real 3c 48 jet could be the assumption of the conservation of kinetic energy @xmath141 .
we have tried to fit the helical model without the conservation of kinetic energy but with conserved angular momentum , which is in principle similar to case 4 in steffen et al .
however , in this case , the modelled helix rapidly evolves into a straight line , and thus fails to reproduce the observed 3c 48 jet on kpc scales .
both two models give fits to the overall jet structure of 3c 48 within 0.45 arcsec with 23 complete revolutions , but they have some differences in detail .
the helical shape of the precessing jet is a superposition of ballistic jet knots modulated by a nodding motion ( nutation ) . in this case
, the whole jet envelope wiggles out and shows a restricted periodicity .
the observed jet structure displays a smooth shape on rather smaller scales .
if , alternatively , the coherent , smooth jet stream is initially disturbed at the jet base , and is amplified by the kelvin - helmholtz instability downstream in the jet , the jet stream itself is bent .
the resulting helical jet flow rotates faster at the start and gradually slows down as it moves further away .
if the twisted inner jet morphology detected at 1.5 and 4.8 ghz ( figure [ fig : core ] ) is real , this would support the k - h instability model .
further high - dynamic - range vlbi maps of the inner jet region could test this scenario .
in addition to the morphological discrepancy , the two models require different physical origins .
in the precessing - jet model , ballistic knots are ejected in different directions which are associated with an ordered rotation in the jet flow direction in the vicinity of the central engine .
if the precession results from a rotating injector at the jet base ( see discussion in worrall et al . 2007 ) , the precession period of 0.2 million yr requires a radius of @xmath150 pc , assuming the injector is in a keplerian motion around the black hole .
this size scale is much larger than the accretion disk , and so we may simply rule out the possibility of an injection from the rotating accretion disk . instead the long - term precession can plausibly take place in a binary smbh system or a tilting accretion disk ( e.g. @xcite ) .
for example , the precessing period caused by a tilting disk is @xmath151 yr , assuming a @xmath152 smbh for 3c 48 , a dimensionless viscosity parameter @xmath153 and the dimentionless specific angular momentum of the black hole @xmath154 @xcite . in this scenario , the short - term nodding motion can then be triggered by the tidally - induced torque on the outer brim of the wobbling accretion disk , analogous to ss 433 @xcite .
on the other hand , the helical k - h instabilities modes can be triggered by ordered or random perturbations to the jet flow . the fits with model 2
give an initial perturbation period @xmath155 yr , which leads to a radius of @xmath156 pc where perturbations take place .
this radius is still larger than the size of the accretion disk , but at this size scale it is still plausible for the perturbations to be due to interactions between the jet flow and the broad - line - region clouds ( e.g. 3c120 : @xcite ) .
however , the high faraday depth and/or the possible internal depolarization structure in the radio core a makes it difficult to investigate this scenario through vlbi polarimetric measurements .
in addition , k - h instabilities would not only produce simple helical modes , but also many other instability modes mixed together ; the k - h interpretation of the oscillatory 3c 48 jet on both pc and kpc scales requires a selection of modes or a simple mix of low - order modes .
however , it is difficult to see how these required modes are excited while others with higher growth rates are suppressed ( see the discussion of the wiggling filament in ngc 315 by worrall et al .
moreover , the k - h model does not have a ready explanation for the observed large - scale gradual bend of the jet axis .
simple kinematical models , such as a reflection by an oblique shock or a pressure gradient in the narrow - line - region ism , may not be adequate to explain the bends of the robust ( @xmath157 ) jet flow .
we have observed 3c 48 at multiple frequencies with the vlba , evn and merlin with spatial resolutions between tens and hundreds of parsec .
our principal results may be summarized as follows : \(1 ) the total - intensity merlin image of 3c 48 is characterized by two components with comparable integrated flux density .
a compact component aligns with the vlbi jet , while an extended envelope surrounds it .
the extended emission structure becomes diffuse and extends toward the northeast at @xmath00.25 arcsec from the nucleus .
the extended component shows a steeper spectrum than the compact jet .
\(2 ) in the vlba and evn images , the compact jet seen in the merlin image is resolved into a series of bright knots .
knot a is further resolved into two smaller features a1 and a2 in 4.8- and 8.3-ghz vlba images .
a1 shows a flat spectrum with spectral index @xmath158 .
a2 shows a steep spectrum with @xmath159 , and may be identified with the inner jet .
the brightness temperature of a1 is @xmath160 k and much higher than the @xmath88 of a2 .
the flux densities of a1 and a2 in epoch 2004 show a 100 and 60 per cent decrease compared with those in 1996 .
the high brightness temperature , flat spectrum and variability imply that a1 is the synchrotron self - absorbed core found close to the active nucleus .
\(3 ) comparison of the present vlba data with those of 1996 january 20 strongly suggests that a2 is moving , with an apparent velocity @xmath161 to the north .
combining the apparent superluminal motion and the jet - to - counterjet intensity ratio yields a constraint on the jet kinematics and geometry : the jet is relativistic ( @xmath162 ) and closely aligned to the line of sight ( @xmath163 ) .
\(4 ) we present for the first time vlbi polarization images of 3c 48 , which reveal polarized structures with multiple sub - components in component c. the fractional polarization peaks at the interface between the compact jet and the surrounding medium , perhaps consistent with a local jet - induced shock .
the systematic gradient of the evpas across the jet width at c can be attributed to the combination of a gradient in the emission - weighted intrinsic polarization angle across the jet and possibly a systematic gradient in the rm .
changing magnetic field directions are a possible interpretation of the rm gradient , but other alternatives can not be ruled out .
the fractional polarization of the hot spot b increases towards higher frequencies , from @xmath164 per cent ( 1.6 ghz ) , @xmath165 per cent ( 4.8 ghz ) to @xmath166 per cent ( 8.3 ghz ) . the relatively low degree of polarization at lower frequencies probably results from a unresolved faraday screen associated with the nlr clouds and/or the internal depolarization in the jet itself .
hot spot b has a higher rm than c , which can perhaps be attributed to a stationary shock in the vicinity of b. the core a at all frequencies is unpolarized , which may be the result of a tangled magnetic field in the inner part of the jet .
\(5 ) the combined evn+merlin 1.65-ghz image and 1.5-ghz vlba images show that the bright knots trace out a wave - like shape within the jet .
we fitted the jet structure with a simple precession model and a k - h instability model .
both models in general reproduce the observed oscillatory jet trajectory , but neither of them is able to explain all the observations .
more observations are required to investigate the physical origin of the helical pattern .
further monitoring of the proper motion of the inner jet a2 should be able to constrain the ballistic motion in the framework of the precessing jet .
high - resolution vlbi images of the inner jet region will be required to check whether or not the jet flow is oscillating on scales of tens of mas , which might give a morphological means of discriminating between the two models .
sophisticated simulations of the jet would be needed to take into account the deceleration of the jet flow due to kinetic energy loss via jet - cloud interaction and radiation loss , but these are beyond the scope of the present paper .
ta and xyh are grateful for partial support for this work from the national natural science foundation of pr china ( nsfc 10503008 , 10473018 ) and shanghai natural science foundation ( 09zr1437400 ) .
mjh thanks the royal society ( uk ) for support .
we thank mark birkinshaw for helpful discussions on the jet kinematics .
the vlba is an instrument of the national radio astronomy observatory , a facility of the us national science foundation operated under cooperative agreement by associated universities , inc .
the european vlbi network ( evn ) is a joint facility of european , chinese , south african and other radio astronomy institutes funded by their national research councils .
merlin is a national facility operated by the university of manchester at jodrell bank observatory on behalf of the uk science and technology facilities council ( stfc ) .
99 akujor c.e . ,
spencer r.e . , zhang f.j . ,
davis r.j . ,
browne i.w.a .
, fanti c. , 1991 , mnras , 250 , 215 an t. , hong x.y . ,
wang w.h . , 2004 ,
chjaa , 4 , 28 bate m.r . , bonnell i.a . ,
clarke c.j . ,
lubow s.h . ,
ogilvie g.i .
, pringle j.e . ,
, 2000 , mnras , 317 , 773 begelman m.c .
, blandford r.d .
, rees m.j . , 1980 ,
nature , 287 , 307 begelman m.c . , king a.r . , pringle j.e . , 2006 , mnras , 370 , 399 briggs d.s . , 1995 ,
thesis , new mexico institute of mines and technology chatzichristou e.t .
, vanderriest c. , jaffe w. , 1999 , a&a , 343 , 407 canalizo g. , & stockton a. , 2000 , apj , 528 , 201 camenzind m. , 1986 , a&a , 156 , 137 condon j.j .
, cotton w.d . ,
greisen e.w . ,
yin q.f . , perley r.a . ,
taylor g.b . ,
broderick j.j . , 1998 , aj , 115 , 1693 cotton w.d . , fanti c. , fanti r. , dallacasa d. , foley a.r .
, schilizzi r.t . ,
spencer r.e . , 1997a , a&a , 325 , 479 cotton w.d . , dallacasa d. , fanti c. , fanti r. , foley a.r .
, schilizzi r.t . ,
spencer r.e .
, 1997b , a&a , 325 , 493 cotton w.d . , dallacasa d. , fanti c. , fanti r. , foley a.r .
, schilizzi r.t . ,
spencer r.e . , 2003 ,
a&a , 406 , 43 fanti c. , fanti r. , parma p. , schilizzi r.t . , van breugel w.j.m .
, 1985 , a&a , 143 , 292 fanti r. , fanti c. , schilizzi r.t .
, spencer r.e .
, nan r.d . ,
parma p. , van breugel w.j.m .
, venturi t. , 1990 , a&a , 231 , 333 fanti c. , fanti r. , dallacasa d. , schilizzi r.t . , spencer r.e . , stanghellini c. , 1995 , a&a , 302 , 317 * fanti c. , 2009 , an , 330 , 120 * feng w.x . ,
an t. , hong x.y . , zhao j .- h .
, venturi t. , shen z .- q . , wang w.h . , 2005 , a&a , 434 , 101 gmez j .-
, marscher a.p .
, alberdi a. , jorstad s.g . , garca - mir c. , 2000 , science , 289 , 2317 gmez j .-
, marscher a.p .
, jorstad s.g . ,
agudo i. , roca - sogorb m. , 2008 , apj , 681 , l69 greenstein j.l . , 1963 , nature , 197,1041 gupta n. , srianand r. , & saikia d.j . , 2005 , mnras , 361 , 451 hardee p.e .
, 1987 , apj , 318 , 78 hardee p.e . , 2003 ,
apj , 597 , 798 hjellming r.m . , &
johnston k.j . , 1981 , apj , 246 , l141 katz j.i . ,
anderson s.f . , grandi s.a . , margon b. , 1982 , apj , 260 , 780 kellermann k.i . , & pauliny - toth i.i.k
. , 1969 , apj , 155 , l71 krips m. , eckart a. , neri r. , zuther j. , downes d. , scharwachter j. , 2005 , a&a , 439 , 75 lu j .- f . & zhou b .- y . , 2005 ,
apj , 635 , l17 mantovani f. , mack k .- h . ,
montenegro - montes f.m .
, rossetti a. , kraus a. , 2009 , a&a , 502 , 61 matthews t. a. , bolton j. g. , greenstein j. l. , munch g. & sandage a. r. , 1961 , sky and telescope , 21 , 148 murgia m. , fanti c. , fanti r. , gregorini l. , klein u. , mack k .- h . , vigotti m. , 1999 , a&a , 345 , 769 nan r.d . , schilizzi r.t . ,
fanti c. , fanti r. , 1991 , a&a , 252 , 513 nan r.d . , cai z.d . , inoue m. , kameno s. , schilizzi r.t . , fanti c. , fanti r. , 1992 , pasj , 44 , 273 odea c.p . , 1998 , pasp , 110 , 493 odea c.p .
, de vries w.h . , koekemoer a.m. , baum s.a . , morganti r. , fanti r. , capetti a. , tadhunter c.n . , barthel p.d . ,
axon d.j . , gelderman r. , 2002 , apj , 123 , 2333 peacock j.a . ,
& wall j.v . , 1982 , mnras , 198 , 843 shen z .- q . , jiang d.r .
, kameno s. , chen y.j .
, 2001 , a&a , 370 , 65 simard - normandin m. , kronberg p.p . , & button s. , 1981 , apjs , 45 , 97 simon r.s . ,
readhead a.c.s .
, moffet a.t . ,
wilkinson p.n . ,
booth r. , allen b. , burke b.f . , 1990 , apj , 354 , 140 spencer r.e .
, schilizzi r.t .
, fanti c. , fanti r. , parma p. , van breugel w.j.m . , venturi t. , muxlow t.w.b .
, nan r.d . , 1991 , mnras , 250 , 225 stanghellini c. , odea c.p . ,
dallacasa d. , baum s.a .
, fanti r. , fanti c. , 1998 , a&as , 131 , 303 steffen w. , zensus j.a .
, krichbaum t.p . ,
witzel a. , qian s.j . , 1995 , a&a , 302 , 335 stockton a. , & ridgway s.e . , 1991 ,
aj , 102 , 488 stockton a. , canalizo g. , fu h. , keel w. , 2007 , apj , 659 , 195 taylor g.b . , &
myers s.t .
, 2000 , vlba scientific memorandum 26 van breugel w. , miley g. , & heckman t. , 1984 , aj , 89 , 5 xiang l. , reynolds c. , strom r.g . , dallacasa d. , 2006 , a&a , 454 , 729 wilkinson p.n . , spencer r.e .
, readhead a.c.s .
, pearson t.j . ,
simon r.s . , 1984 , iau symposium 110 , ` vlbi and compact radio sources ' , eds .
r. fanti , k.i .
kellermann , g. setti , p.25 wilkinson p.n . ,
tzioumis a.k . ,
akujor c.e . , benson j.m . ,
walker r.c . ,
simon r.s . , 1990 , in zensus j.a . , pearson t.j .
, eds . parsec - scale radio jets , cambridge univ .
press , cambridge , p. 152
wilkinson p.n . ,
tzioumis a.k . , benson j.m . ,
walker r.c . ,
simon r.s . ,
kahn f.d . , 1991 , nature , 352 , 313 worrall d.m . , hardcastle m.j . , pearson t.j . ,
readhead a.c.s . , 2004 , mnras , 347 , 632 worrall d.m . , birkinshaw m. , laing r.a . , cotton w.d .
, & bridle a.h . , 2007 , mnras , 380 , 2 zuther j. , eckart a. , scharwachter j. , krips m. , straubmeier c. , 2004 , a&a , 414 , 919 ) image of 3c 48 from the merlin observation at 1.65 ghz .
the image was made with uniform weighting .
the restoring beam is 138@xmath30115 ( mas ) , pa=65.1 .
the phase centre is at ra=01@xmath16737@[email protected] , [email protected] .
the _ r.m.s .
_ noise in the image measured in an off - source region is @xmath01.3 mjy b@xmath8 , corresponding to a dynamic range of @xmath02800:1 in the image .
the contours are 6 mjy b@xmath171(-2 , 1 , 2 , 4 , ... , 512 ) .
the cross denotes the location of the hidden agn .
the square marks the region in which compact jet dominates the emission structure . ] ) image of 3c 48 from vlba and evn observations .
the phase centres of all images have been shifted to ra=01@xmath16737@[email protected] , [email protected] .
table [ tab : figpar ] presents the image parameters .
the horizon bars in the subpanels illuminates the length scale in projection .
a number of bright components are labeled in the images . ] ) image of 3c 48 from vlba observations in 2004 and 1996 . *
top left : * 1.5 ghz image on epoch 2004 .
the image was made with a super - uniform weighting ( robust=@xmath172 and uvbox=3 in aips task imagr ) and restored with a beam of [email protected] ( mas ) , pa=@xmath173 contours : 0.4 mjy b@xmath171 ( 1,2, ... ,128 ) ; * top right : * 4.8 ghz contours ( epoch 1996 ) superposited on the grey - scale image ( epoch 2004 ) ; * bottom left : * 8.3 ghz contours ( epoch 2004 ) and grey - scale ( epoch 1996 ) , * bottom right : * 8.3 ghz contours ( epoch 2004 ) and 15.4 ghz grey - scale ( epoch 1996 ) . ] 40 ( mas ) .
the lowest contour is 0.7 mjy b@xmath8 , increasing in a step of 4 .
the 1.65 ghz data are obtained from the combined evn and merlin data observed on 2005 june 7 ( the present paper ) .
the two images are re - produced using visibility data on the common _ uv _ range , and restored with the same 40@xmath3040 ( mas ) beam .
compact vlbi components are labeled in the image .
] image ( figure [ fig : mercont ] ) .
the polarization image is derived from stokes @xmath19 and @xmath20 images above a 4@xmath47 cutoff ( 1@xmath47=6 mjy b@xmath8 ) .
the wedge at the top indicates the percentage of the polarization .
the length of the bars represents the strength of polarized emission , 1 arcsec represents 0.5 jy b@xmath8 .
the orientation of the bar indicates the rm - corrected evpa . ]
): quantitative calculations ( figure [ fig : rm ] ) show that the rms in the component c region is about 1.4 times the value measured from the overall source ; moreover , the intrinsic polarization angles rotate by @xmath174 from the northwest edge of component c to the southeast edge
. therefore the correction based on the overall - source rm might not be sufficiently accurate to all sub - components , while this uncertainty tends to small toward the higher frequencies .
* ( a ) * : the evn image at 1.65 ghz .
contours are 4 mjy b@xmath171 ( 1,4,16,64,256 ) .
the stokes @xmath19 and @xmath20 maps were convolved with a 20-mas circular beam , and we used intensities above 4@xmath47 to calculate the polarized intensity and polarization angle ; * ( b ) * : contours : 1 mjy b@xmath171 ( 1,4,16,64,256 ) ; * ( c ) * : contours : 1 mjy b@xmath171 ( 1,4,16,64,256 ) ; * ( d ) * : contours : 1.2 mjy b@xmath171 ( 4,16,64,256 ) ; the contours in the inset panel are 1.2 mjy b@xmath171 ( 1,4,8,16,32,64,128,256,512 ) .
, title="fig:",scaledwidth=45.0% ] ): quantitative calculations ( figure [ fig : rm ] ) show that the rms in the component c region is about 1.4 times the value measured from the overall source ; moreover , the intrinsic polarization angles rotate by @xmath174 from the northwest edge of component c to the southeast edge
. therefore the correction based on the overall - source rm might not be sufficiently accurate to all sub - components , while this uncertainty tends to small toward the higher frequencies . *
( a ) * : the evn image at 1.65 ghz .
contours are 4 mjy b@xmath171 ( 1,4,16,64,256 ) .
the stokes @xmath19 and @xmath20 maps were convolved with a 20-mas circular beam , and we used intensities above 4@xmath47 to calculate the polarized intensity and polarization angle ; * ( b ) * : contours : 1 mjy b@xmath171 ( 1,4,16,64,256 ) ; * ( c ) * : contours : 1 mjy b@xmath171 ( 1,4,16,64,256 ) ; * ( d ) * : contours : 1.2 mjy b@xmath171 ( 4,16,64,256 ) ; the contours in the inset panel are 1.2 mjy b@xmath171 ( 1,4,8,16,32,64,128,256,512 ) .
, title="fig:",scaledwidth=45.0% ] + ): quantitative calculations ( figure [ fig : rm ] ) show that the rms in the component c region is about 1.4 times the value measured from the overall source ; moreover , the intrinsic polarization angles rotate by @xmath174 from the northwest edge of component c to the southeast edge . therefore the correction based on the overall - source rm might not be sufficiently accurate to all sub - components , while this uncertainty tends to small toward the higher frequencies . *
( a ) * : the evn image at 1.65 ghz .
contours are 4 mjy b@xmath171 ( 1,4,16,64,256 ) .
the stokes @xmath19 and @xmath20 maps were convolved with a 20-mas circular beam , and we used intensities above 4@xmath47 to calculate the polarized intensity and polarization angle ; * ( b ) * : contours : 1 mjy b@xmath171 ( 1,4,16,64,256 ) ; * ( c ) * : contours : 1 mjy b@xmath171 ( 1,4,16,64,256 ) ; * ( d ) * : contours : 1.2 mjy b@xmath171 ( 4,16,64,256 ) ; the contours in the inset panel are 1.2 mjy b@xmath171 ( 1,4,8,16,32,64,128,256,512 ) .
, title="fig:",scaledwidth=45.0% ] ): quantitative calculations ( figure [ fig : rm ] ) show that the rms in the component c region is about 1.4 times the value measured from the overall source ; moreover , the intrinsic polarization angles rotate by @xmath174 from the northwest edge of component c to the southeast edge . therefore the correction based on the overall - source rm might not be sufficiently accurate to all sub - components , while this uncertainty tends to small toward the higher frequencies . *
( a ) * : the evn image at 1.65 ghz .
contours are 4 mjy b@xmath171 ( 1,4,16,64,256 ) .
the stokes @xmath19 and @xmath20 maps were convolved with a 20-mas circular beam , and we used intensities above 4@xmath47 to calculate the polarized intensity and polarization angle ; * ( b ) * : contours : 1 mjy b@xmath171 ( 1,4,16,64,256 ) ; * ( c ) * : contours : 1 mjy b@xmath171 ( 1,4,16,64,256 ) ; * ( d ) * : contours : 1.2 mjy b@xmath171 ( 4,16,64,256 ) ; the contours in the inset panel are 1.2 mjy b@xmath171 ( 1,4,8,16,32,64,128,256,512 ) . ,
title="fig:",scaledwidth=45.0% ] ( 1,4,8,16 ) .
the wedge at the top indicates the rm in the observer s frame , in unit of rad m@xmath44 .
the insets show the measured values of the observed polarization angle for four selective locations as a function of @xmath65 along with a linear fitting of rotation measure . ] .
the jet axis op is inclining by an angle of ( @xmath120 ) with respect to the line of sight .
the angle @xmath121 between the op and y - axis is defined as the position angle in the 2-dimension clean image . ]
+ @xmath175 : a precessing jet model ; + @xmath176 : a helical - mode k - h instability model ; + @xmath122 : in the precessing model , @xmath122 represents the flow speed in the observer s frame , taking into account relativistic aberration effects ; in the k - h model , @xmath122 denotes the pattern velocity .
the velocity is expressed in terms of proper motion in order to agree with the coordinates used in clean images ; + @xmath177 : the angle between the jet axis and the sky plane ; + @xmath178 : the initial position angle of the jet axis , measured from north to east ; + @xmath179 : the rate of change of position angle with time . in the precessing model
, it gives an estimate of the angular velocity of the precession ; + @xmath123 : half of the opening angle of the helix cone ; + @xmath180 : initial phase angle of the helical jet flow ; + @xmath124 : in the model 1 , the fitted @xmath124 is actually the nutation period , see discussion in section 5.2.1 ; in the k - h model , @xmath124 represents an initial period for the triggered perturbations ; + @xmath143 : the initial radius where the k - h instabilities starts to grow ; + [ tab : helicalfit ] | we present results on the compact steep - spectrum quasar 3c 48 from observations with the very long baseline array ( vlba ) , the multi - element radio linked interferometer network ( merlin ) and the european vlbi network ( evn ) at multiple radio frequencies . in the 1.5-ghz vlbi images ,
the radio jet is characterized by a series of bright knots .
the active nucleus is embedded in the southernmost vlbi component a , which is further resolved into two sub - components a1 and a2 at 4.8 and 8.3 ghz .
a1 shows a flat spectrum and a2 shows a steep spectrum .
the most strongly polarized vlbi components are located at component c @xmath00.25 arcsec north of the core , where the jet starts to bend to the northeast .
the polarization angles at c show gradual changes across the jet width at all observed frequencies , indicative of a gradient in the emission - weighted intrinsic polarization angle across the jet and possibly a systematic gradient in the rotation measure ; moreover , the percentage of polarization increases near the curvature at c , likely consistent with the presence of a local jet - ism interaction and/or changing magnetic - field directions .
the hot spot b shows a higher rotation measure , and has no detected proper motion .
these facts provide some evidence for a stationary shock in the vicinity of b. comparison of the present vlbi observations with those made 8.43 years ago suggests a significant northward motion for a2 with an apparent transverse velocity @xmath1 .
the apparent superluminal motion suggests that the relativistic jet plasma moves at a velocity of @xmath2 if the jet is viewed at an inclination angle less than @xmath3 . a simple precessing jet model and a hydrodynamical isothermal jet model with helical - mode kelvin - helmholtz instabilities are used to fit the oscillatory jet trajectory of 3c 48 defined by the bright knots .
[ firstpage ] galaxies : active , galaxies : kinematics , galaxies : jets , quasars : individual : 3c 48 |
if electrons did not interact , only the integral quantum hall effect ( iqhe ) @xcite would occur in nature .
the discovery of the fractional quantum hall effect ( fqhe ) @xcite signaled the existence of a new correlated state of matter , the essence of which lies in the formation of quantum particles called composite fermions , @xcite which are bound states of electrons and an even number of quantized vortices .
electrons avoid each other most efficiently by turning into composite fermions , which , in turn , interact much more weakly than electrons . for a large number of phenomena , it is a valid first approximation to neglect the interaction between composite fermions altogether , indicating that the interaction between composite fermions can often be treated perturbatively and does not cause any phase transitions .
the fundamental property of composite fermions , which is responsible for the dramatic phenomenology of two - dimensional systems in high magnetic fields , is that as they move about , the aharonov bohm phase is partly canceled by the phase produced by the vortices tied to other composite fermions and as a result they experience an _ effective _ magnetic field @xmath3 given by @xmath4 where @xmath5 is the external magnetic field , @xmath6 is the electron density ( in two dimensions ) , the even integer @xmath7 is the vorticity of composite fermions ( cf s ) , and @xmath8 .
( the composite fermion with vorticity @xmath7 is denoted by @xmath9cf or cf-@xmath7 . )
they form landau - like levels in this reduced magnetic field , with their filling factor ( @xmath10 ) related to the electron filling factor ( @xmath11 ) by @xmath12 the wave functions for composite fermions are given by @xcite @xmath13 where @xmath14 denotes the position of the @xmath15th particle , @xmath16 is an antisymmetric wave function for fermions at the effective filling @xmath10 , and @xmath17 is the lowest landau level ( lll ) projection operator .
the wave function @xmath18 is of course the wave function of correlated electrons at @xmath11 , but can also be interpreted as the wave function of composite fermions at filling factor @xmath10 , because on the right hand side , each electron has @xmath7 vortices bound to it through the multiplicative factor @xmath19 , which converts each electron in into a composite fermion .
let us begin by neglecting the interaction between composite fermions . in this case
, a gap opens up when the composite fermions fill an integral number of levels , i.e. , when @xmath20 , which corresponds to electron filling factors given by @xmath21 where @xmath22 is a positive or negative integer .
[ the negative values of @xmath22 give @xmath23 , and correspond to the situation in which the effective field @xmath3 is opposite to the external field @xmath5 . ] the model of non - interacting composite fermions thus predicts quantum hall effect ( qhe ) of electrons at these fractions , and also at @xmath24 due to particle hole symmetry in the lowest landau level .
( the fractions in eq . [ f2 ] can be obtained by formulating the problem in terms of _ holes _ rather than electrons in the lowest landau level , and then making composite fermions out of them . ) the qhe at these filling factors will be termed
cf-@xmath7 qhe . "
cf-2 qhe and cf-4 qhe are routine , and cf-6 qhe and cf-8 qhe have also been observed @xcite .
these sequences exhaust most of the more than 50 observed fractions to date in the lowest landau level .
the wave functions have been tested in detail for finite systems , and are very accurate , without involving any adjustable parameter@xcite . in short ,
most of the qhe can be understood as the integral qhe of appropriate fermions .
it ought to be noted that the qhe of higher order composite fermions can be viewed as a consequence of interactions between lower order composite fermions .
for example , the cf-4 qhe originates from interactions between cf-2 s . nonetheless , the physics of all states at filling factors of eqs .
( [ f1 ] ) and ( [ f2 ] ) is correctly described in terms of _ non_-interacting composite fermions of the appropriate kind .
while the above fractions are obtained most immediately in the cf theory , it has been known since the very beginning of the composite fermion theory that other fractions are not ruled out.@xcite new qhe _ between _ two successive integral hall states of composite fermions will be called nqhe , where n stands for nested . " a simple scenario for nqhe
is as follows .
consider electrons in the filling factor range @xmath25 which map into @xmath9cf s in the range @xmath26 containing @xmath22 levels fully occupied and the @xmath27 level partially occupied .
it is a natural conjecture that inter - cf interaction can possibly cause gaps at @xmath28 which will produce new fractions between the familiar fractions @xmath29 and @xmath30 . here , the @xmath9cf s in the topmost partially filled level capture @xmath31 vortices ( as a result of the residual @xmath9cf-@xmath9cf interaction ) to turn into higher order composite fermions ( @xmath32cf s ) and condense into @xmath33 levels , thereby opening a gap and producing quantum hall effect .
this qhe state is a nested state with higher order composite fermions forming in the background of the original composite fermions .
certain qualitative consequences of the above scenario are as follows .
the nqhe of composite fermions is analogous to qhe of electrons between two successive iqhe states of electrons , that is , to fqhe in _ higher _ electronic landau levels .
it is known that fqhe is rare in second electronic landau levels ( i.e. , in the range @xmath34 , including spin degeneracy ) and non - existent in third or higher landau levels ; by analogy , nqhe is expected to be rare . extending the analogy further , the strongest nqhe states for fully spin polarized composite fermions are expected to be at @xmath35 and @xmath36 , which correspond to electron filling factors @xmath37 and @xmath38 ; these are expected to be the strongest nested
cf states @xcite .
recently , pan _
_ @xcite have reported observation of 4/11 and 5/13 , heralding a new generation of quantum hall effect that can not be understood as the iqhe of composite fermions .
this discovery has given fresh impetus to the issue of nqhe , especially because it is possible that more nqhe states will be observed in the future as the sample quality improves and the temperature drops , reminiscent of how the fqhe appeared on top of the iqhe .
what other nested states are possible ? what is the true nature of these states ?
such questions have motivated us to investigate this topic further . even though the theoretical scenario described above
is plausible , and indeed natural , there is no _ a priori _ guarantee that it actually occurs , and it is important to carry out quantitative tests to ascertain its applicability to the real world . it is worth recalling here that not all fractions that can occur _ in principle _ do really occur in nature .
for example , fqhe can occur for some hypothetical interaction at very small filling factors , but is preempted by wigner crystal for the coulomb interaction , and in higher landau levels , the fqhe is often believed to be unstable to charge - density - wave type state .
therefore , it would require further theoretical work before one could claim with any degree of confidence that nqhe based on the physics described above is actually possible .
the quantitative theoretical investigations so far suggest that there should be _ no _ fqhe at fractions other than those in eqs .
( [ f1 ] ) and ( [ f2 ] ) for fully spin - polarized electrons .
an exact diagonalization study @xcite of @xmath39 electrons at 4/11 finds a non - uniform ground state , which is an evidence against fqhe liquid . in another approach @xcite
, a model has been developed for the effective interaction between composite fermions in the second level . in ref .
it was argued that no new fqhe is obtained @xcite .
ref . found that the form of this interaction favors a bubble crystal " of composite fermions at filling factors like 4/11 rather than the quantum hall effect .
these studies indicate that the effective interaction between composite fermions is not sufficiently strongly repulsive to stabilize fqhe at 4/11 and several other fractions .
both approaches have their problems . in exact diagonalization ,
the system is effectively small , with only @xmath40 composite fermions in the second level , and one may question if the conclusion based on this system will hold up as @xmath41 increases .
the effective interaction model involves several assumptions , because of which the results are plausible but not conclusive . in ref
. showed that fqhe at 4/11 is very likely possible for a _ partially _ spin polarized system or a spin unpolarized system .
it is not yet known for sure if the fqhe states at 4/11 and 5/13 are partially or fully polarized in the experiments of pan _
et al . _ , @xcite but given the relatively high magnetic fields at which these states have been observed , it is quite possible that they are fully spin polarized , and therefore it is worth revisiting the issue of whether fully polarized incompressible states can be found theoretically at these filling fractions .
that is the prime motivation behind our present study . in this work
, we will look for fqhe at several filling fractions in the range between 1/3 and 2/5 .
we will study finite systems of @xmath41 particles at total flux @xmath42 ( in units of the flux quantum @xmath8 ) , where the relation between @xmath41 and @xmath43 for a given filling factor will be derived from the physics described above . the most reliable theoretical method would be exact diagonalization . however , because of the exponentially growing hilbert space , exact diagonalization is possible only for very small systems , and very few filling factors .
for example , for @xmath37 , exact diagonalization has been possible for @xmath39 electrons.@xcite the next system contains @xmath44 ( in the spherical geometry , discussed below ) , for which the lowest landau level hilbert space contains a total of @xmath45 slater determinant basis states ; the diagonalization of matrices of such sizes is beyond the reach of present day computers .
we have developed a powerful new method , namely diagonalization in low - energy composite - fermion basis ( lecfb ) , which we believe gives accurate results .
it exploits the fact that the composite fermion theory allows us to directly identify low - energy states of the full hilbert space , which lets us work within a sub - space much smaller than the full hilbert space .
for example , we have been able to obtain the low - energy spectrum for the 4/11 state with as many as 24 electrons , where the dimension of the full hilbert space is @xmath2 .
a brief outline of the method is as follows .
consider the filling factor range given in eq .
( [ eq2 ] ) .
given that the two ends of this range are well described as @xmath46 and @xmath22 filled levels of composite fermions , it is natural to expect that the low energy states at @xmath11 contain @xmath22 composite fermion levels completely occupied , and the remaining composite fermions in the @xmath27 level .
the configurations that involve promotion of composite fermions to higher kinetic energy levels are expected to cost substantially higher energy and are neglected . in other words , we include all configurations of type shown in fig .
( [ fig_1 ] ) but neglect configurations of the type shown in fig .
( [ fig_2 ] ) .
this neglect of composite - fermion landau level mixing is the only assumption in our approach , which can be tested for consistency at the end of the calculation .
other than that the method is exact .
the most important point here is that we do not make any assumption with regard to the interaction between composite fermions , or their structure in the partially filled level . in our calculation
, we first construct all states of composite fermions at @xmath47 , with the lowest @xmath22 levels fully occupied and the next one with filling factor @xmath48 , and then diagonalize the coulomb hamiltonian within this sector to obtain the spectrum of low energy states .
the calculation is rather involved , requiring extensive monte carlo , but still numerically stable , and yields reliable results .
the present method is a significant advance over the variational method of ref . .
it is much more reliable because it eliminates several approximations made in that work . in ref .
@xcite , the problem of composite fermions at @xmath47 was mapped into that of fermions at @xmath48 .
further , it was assumed that the interaction energy of the system of composite fermions in the partially filled level can be approximated by a sum of two - body interactions .
the interaction between composite fermions was obtained by keeping only two composite fermions in the @xmath27 level and integrating out the composite fermions of the lower filled levels . in general , when there are many composite fermions in the @xmath27 level , the integrating out of the lower - level composite fermions will produce a quite complex interaction , with two- , three- , and @xmath22-body terms , because the full system is strongly correlated .
the hope was that the @xmath22-body terms do not cause any phase transitions , in which case the two - body terms will produce the correct state . however , this approximation was untested . finally , the method was a variational study , in which the energies of certain wave functions were compared to determine which had the lowest energy ; there was of course no guarantee that this state would describe the true ground state .
in contrast , in the present study , we work with the full composite fermion system and do not assume anything about the form of the interaction between composite fermions or the nature of their ground state . the following section contains a discussion of how we construct the low - energy composite fermion basis and how we carry out the gram - schmid orthogonalization and diagonalization .
the reader not interested in the technical details can skip this section and directly proceed to the subsequent section , wherein the results are given and their implications are discussed .
we first summarize certain relevant facts from the introduction .
the electron filling factor @xmath11 given by @xmath49 corresponds to composite fermion filling factor @xmath10 .
we will be interested below in the specific values of @xmath10 given by @xmath50 here , composite fermions fill @xmath22 levels completely and occupy @xmath48 fraction of the @xmath27 level .
if the @xmath9cf s in the partially occupied level capture @xmath31 additional vortices and condense into @xmath33 filled landau level state , then a gap would open up and nqhe would be obtained .
the question is whether this mechanism really occurs in nature . in the composite fermion theory ,
there are several possible approaches for constructing a low energy basis . in this work ,
we construct a lecfb of _ all _ states at @xmath47 in which @xmath22 @xmath9cf levels are fully occupied and the next level has filling @xmath48 .
we do not make any approximation regarding the structure of @xmath9cf s in the partially filled level .
this basis is in principle straightforward : we simply need to take all _ electronic _ states at @xmath47 and attach @xmath7 vortices to obtain the composite fermion states .
the method , however , is technically rather demanding , and the present section is devoted to it .
we will employ the spherical geometry in which @xmath41 electrons move on the surface of a sphere under the influence of a radial magnetic field @xmath5 created by a magnetic monopole at the center .
the magnitude of the @xmath5 field is given by @xmath51 where @xmath52 is known as the flux quantum , @xmath53 is the radius of the sphere , and @xmath43 is called the monopole strength which should be either an integer or a half - integer because of dirac s quantization condition .
the interacting electron system at monopole strength @xmath43 maps into a system of weakly interacting composite fermions at an effective monopole strength @xmath54 .
the wave functions for interacting electrons at @xmath43 are given by @xmath55 where @xmath56 is the wave function for the fully occupied lowest landau level , and @xmath57 are antisymmetric wave functions at @xmath58 .
@xmath59 is in general a linear superposition of slater - determinant basis states made up of the monopole harmonics , @xmath60 , given by @xcite : @xmath61 where @xmath62 @xmath63=0,1,2 , ... is the landau level ( ll ) index , to be differentiated from @xmath22 , the number of filled landau levels .
@xmath64 represents the angular coordinates @xmath65 and @xmath66 of the @xmath15th electron , and @xmath67 @xmath68 it was shown in ref . that @xmath69 , the wave function of interacting electrons at @xmath70 , is obtained from @xmath57 by replacing @xmath60 by @xmath71 , defined as : @xmath72\;\end{aligned}\ ] ] where @xmath73 here the prime denotes the condition @xmath74 . for @xmath75 ,
that is at the special filling factors @xmath76 the wave function for the ground state has the simple form : @xmath77 where @xmath78 is the slater determinant wave function for @xmath22 filled landau levels . since @xmath78 is unique , the wave function @xmath79 also contains no adjustable parameters .
it has been tested in the past and was found to be remarkably accurate @xcite . of interest in this work
is the situation when @xmath80 . here
, we obtain the low energy spectrum through the following steps . 1 .
we perform exact diagonalization for @xmath81 particles at filling factor @xmath48 in the lowest landau level to get all eigenstate .
the form of the interaction used for the diagonalization is of no significance , because its only role is to produce basis wave functions , but we work with the coulomb interaction .
let us denote these states by @xmath82 , where @xmath83 labels different states .
2 . we then promote each eigenstate to the @xmath27 landau level , and fill the lower @xmath22 landau levels completely with additional particles to get the @xmath41 particle state at @xmath0 .
this gives _ all _ wave functions @xmath84 for which the lowest @xmath22 landau levels are fully occupied and the next landau level has filling @xmath48 .
these states are denoted by @xmath85 , where @xmath86 denotes addition of @xmath22 landau levels .
it ought to be remembered that the operator @xmath86 changes the number of particles .
3 . we then multiply each wave function by @xmath87 and carry out the lowest landau level projection by the method discussed above .
this gives us correlated basis functions @xmath88 at @xmath11 .
these steps are summarized as + @xmath89 + @xmath90 + the advantage of constructing basis states in this way is that it directly gives us basis functions with well defined orbital angular momentum @xmath91 , which is a good quantum number in the spherical geometry .
we can work within each @xmath91 sector independently , because they are not coupled by the interaction .
( note that in the above equation , @xmath82 is a wave function for @xmath81 electrons only . )
the basis thus obtained is not orthogonal .
we obtain an orthogonal basis following the gram - schmid procedure .
some relevant details are outlined below .
we construct the hamiltonian matrix .
diagonalization of the hamiltonian gives the low energy eigenspectrum at @xmath11 .
if an incompressible state is obtained , the ground state wave function is likely to be well approximated by @xmath92 \label{cfgs}\ ] ] where @xmath93 is the wave function for the ground state for @xmath81 particles at @xmath94 .
since our goal is to look the feasibility of nqhe described above , we do our calculations at those values of @xmath43 where it is possible for @xmath9cf s in the partially filled level to condense into a standard type of fqhe state . for that
, we first need to determine the relation between @xmath43 and @xmath41 for a given filling factor . clearly , in the limit @xmath95 we must have @xmath96 where @xmath42 is the number of flux quantum . for finite systems
, however , the ratio @xmath97 is not exactly equal to @xmath11 , and the identification of what filling factor a finite system corresponds to requires additional theoretical input . let us first consider the incompressible states at filling factor @xmath98 , the wave function for which is given by @xmath99 . for these states ,
the relation between @xmath43 and @xmath41 can be obtained by noting that the product of two monopole harmonics at @xmath100 and @xmath101 gives a monopole harmonic at @xmath102 , i.e. the monopole strengths add . for reasons that will become clear below ,
let us denote the number of particles by @xmath103 and the monopole strength by @xmath104 .
the lowest ll projection is unimportant for the question of the relationship between @xmath103 and @xmath104 , as it does not alter @xmath104 . in the spherical geometry ,
for monopole strength @xmath105 the degeneracy of the lowest landau level is @xmath106 , for the next ll it is @xmath107 , and so on . from that , it is clear that the @xmath33 filled ll state is obtained for @xmath108 given by @xmath109 in particular , one filled ll is obtained at @xmath110 from the addition rule , the monopole strength of @xmath99 is given by @xmath111 it can be verified that @xmath112 gives the filling factor @xmath113 in the thermodynamic limit .
now we ask what is the monopole strength @xmath58 for the state at @xmath114 for @xmath41 electrons .
let us denote by @xmath103 the number of fermions in the @xmath27 partially filled landau level . in order for @xmath103 fermions to form the @xmath115 state , the degeneracy in this level must satisfy : @xmath116 which implies that @xmath117 the total number of fermions , including the fermions in the lowest @xmath22 fully occupied landau levels is @xmath118 eliminating @xmath103
, one gets the following relation between @xmath41 and @xmath58 : @xmath119 the state at @xmath120 is obtained for all vaules of ( @xmath41,@xmath58 ) related by the above equation , provided @xmath41 is an integer and @xmath58 is an integer or a half integer .
the state at @xmath121 in turn is obtained for all values of ( @xmath41,@xmath43 ) with @xmath122 .
we have considered in this work the following filling factors .
* @xmath123 : here , @xmath124 .
the parameters are : @xmath125 , which gives + @xmath126 * @xmath127 : here , @xmath128 .
the parameters are : @xmath129 and @xmath130 , which gives @xmath131 * @xmath132 : here , @xmath133 .
the parameters are : @xmath129 and @xmath134 , which gives @xmath135 * @xmath136 : here , @xmath137 .
the parameters are : @xmath138 and @xmath139 , which gives @xmath140 of course , only integer values are allowed for @xmath41 , @xmath141 , and @xmath42 .
the table [ tab1 ] explicitly lists the systems that we have studied below .
the basis obtained in the way described above is not orthogonal for a given angular momentum sector @xmath91 .
we orthogonalize these using the gram - schmid procedure and than diagonalize the coulomb hamiltonian in the orthogonal basis .
these require the knowledge of matrix elements of the type @xmath142 and @xmath143 , where @xmath144 and @xmath145 ( @xmath146 ) are unorthonormalized wave functions , which we obtain by monte carlo .
since the monte carlo evaluation is most efficient when the integrand is positive definite , we determine these using the equations @xcite @xmath147 and @xmath148 where @xmath149 denotes the coulomb interaction hamiltonian .
we denote @xmath150 as the unorthogonal but normalized basis set obtained from the set @xmath151 , and define @xmath152 and @xmath153 these are real numbers .
we then follow the gram - schmid procedure to construct an orthogonal basis set @xmath154 in terms of the unorthogonal basis set @xmath155 .
the former can be expressed as @xmath156 with @xmath157 and @xmath158 .
while the above relation allows an iterative evaluation of @xmath154 , we find it more convenient to work with the following expressions : @xmath159 and @xmath160 \nonumber \\ & = & \frac{1}{n_i^2 } \left [ u_{ij } - \sum_{k=1}^{i-1 } f_{ki}f_{kj}n_k^2 \right ] \ , , \end{aligned}\ ] ] eq .
( [ gs ] ) was used to obtain these expressions , as well as the fact that @xmath161 and @xmath162 are real .
the hamiltonian matrix elements @xmath163 in the basis set @xmath154 are thus obtained as @xmath164 \ , \end{aligned}\ ] ] which are again obtained iteratively .
we finally diagonalize the matrix @xmath165 to obtain the eigenvalue spectrum for a given @xmath91 sector . the major computation time is spent in calculating the matrix elements @xmath166 and @xmath167 in the monte carlo method , because at each step we must evaluate @xmath168 slater determinants , @xmath168 being the size of the lecfb .
to give an example , at @xmath169 we have @xmath170 . to minimize the computational time , we sample all the states relative to the @xmath171 state , which is obtained from the lowest energy state at @xmath10 .
we typically perform six to ten monte carlo runs for each @xmath172 with ( 0.6 - 1.0)@xmath173 monte carlo iterations in each run . for larger systems , we use parallel computers .
( we divide all of the monte carlo steps into several configurations with each configuration placed on a single node of a beowolf class pc cluster .
one node consists of a dual 1 ghz pentium iii processor . to obtain one data point , at a particular @xmath41
, we use as many as 30 nodes repeatedly until the standard deviation in energies is sufficiently low to produce the desired accuracy . )
approximately 80 cpu days were needed for the calculation of the spectrum for @xmath169 .
the computation time increases very quickly with @xmath41 , approximately as @xmath174 .
it is difficult to ascertain how the statistical error in each element of the matrix would propagate into the eigenvalues .
we instead obtain the spectrum in several different runs , and use the eigenvalues from those different runs to obtain the average and the standard deviation for each eigenvalue .
we have obtained the low energy spectra at several values of @xmath175 which were identified earlier with systems at special filling factors 4/11 , 5/13 , 7/19 , and 6/17 .
( [ fig1 ] ) , ( [ fig2 ] ) , ( [ fig3 ] ) , and ( [ fig4 ] ) show results for 4/11 ; figs .
( [ fig5 ] ) , and ( [ fig6 ] ) for 5/13 ; figs .
( [ fig7 ] ) , and ( [ fig8 ] ) for 7/19 ; and figs .
( [ fig9 ] ) , ( [ fig10 ] ) , and ( [ fig11 ] ) for 6/17 . the energy spectrum for @xmath176
have been given in the past both from exact diagonalization and the composite fermion theory @xcite .
the accuracy of the energies is good enough to be able to state whether the ground state has @xmath171 or not .
the error increases with increasing @xmath91 , which is a consequence of our use of an @xmath171 state for sampling ; in principle , the error at large @xmath91 can be reduced by sampling with another reference state , but we do not see any need for it for the question of interest in this work . the orthogonalization makes a qualitative difference in the spectrum . figs .
( [ fig12 ] ) and ( [ fig13 ] ) show the energies of the basis states _ prior _ to orthogonalization , which ought to be compared to the spectra in figs .
( [ fig3 ] ) and ( [ fig4 ] ) , respectively
. the results explicitly confirm that our method indeed gives very low energy states .
for example , consider 4/11 .
the ground state energy is approximately @xmath177 per particle , which compares very favorably with @xmath178 , which is what one would get from a linear extrapolation between the energies @xcite at 1/3 ( @xmath179 ) and 2/5 ( @xmath180 ) .
further , a large fraction of the states obtained here are within an energy band less than the effective cyclotron gap at 1/3 , which is on the order of 0.1 @xmath181 . in other words ,
most of the states considered here lie within the 1/3 gap .
this justifies our approximation of neglecting the states that involve transitions between composite fermion landau levels , which are expected to cost energy of order 0.1 @xmath181 .
now we ask whether the spectra imply incompressibility .
a characteristic feature of _ all _ incompressible states known to date is that they have an @xmath171 ground state , which is separated by a gap from other states ; in addition , there is also a well defined branch of neutral excitation .
the ground state is extremely well described as a filled landau level state of composite fermions , and the excitation as a particle hole pair excitation of composite fermion @xcite .
this is the case , with no exception , for for systems with @xmath182 related by @xmath183 , which are identified with @xmath184 .
none of the spectra have these standard features of an incompressible state . even in cases
when the ground states has @xmath171 , there is no well defined cf - exciton branch .
the final spectrum at @xmath11 thus bears little resemblance to the spectrum at @xmath113 that we started out with , implying that the interaction between composite fermions in the partially filled level is quite different from that of electrons at @xmath113 in the lowest landau level .
the scenario described in the introduction for nqhe is thus not borne out .
the ground states at the flux values considered above are in general not incompressible uniform states with @xmath171 .
consider 4/11 for example .
it has been known @xcite that for @xmath39 particle system at @xmath185 , which is identified above with 4/11 , the ground state does not have @xmath171 . however , the system is effectively very small here , with only @xmath186 composite fermions in the second level , and one may question if the conclusion will hold up as @xmath41 increases .
we find that for @xmath44 and 20 the ground state has @xmath171 , but for @xmath187 and 24 it does not .
this provides an illustration for why a study of several values of @xmath41 is crucial ; if one had only studied @xmath44 , that would have led one to exactly the opposite conclusion .
another example is the 12 particle system at 6/17 , which has an @xmath171 ground state and might suggest fqhe at 6/17 ; however , the spectra for @xmath188 and 24 at 6/17 explicitly clarify that this is only a finite size effect that does not survive in the thermodynamic limit .
our study implies that even though the ground state sometimes has @xmath171 , it does _ not _ describe an incompressible state at any of the filling factors considered here .
the following observations support this statement : 1 . when the ground state has @xmath189 , the @xmath171 state is a rather highly excited state .
2 . in the plot of the ground state energy with @xmath190 , the energies of the @xmath171 ground states are not particularly low . see fig .
( [ fig14 ] ) for 4/11 . 3 .
the ground state at @xmath113 generally produces an excited state at @xmath11 , even in the @xmath171 sector ( prior to orthogonalization ) , which indicates a breakdown of analogy between @xmath113 and @xmath11 .
the energy per particle is not a smooth function of @xmath190 , but instead shows fluctuations characteristic of a compressible state .
see fig .
( [ fig14 ] ) . 5 .
it is not inconceivable for a compressible state to have @xmath171 , because a uniform state avoids hartree energy cost , but it would be difficult to understand how an incompressible state might have @xmath189 . as a result , the physical mechanism , in which some of the composite fermions turn into higher order composite fermions and condense into landau levels to exhibit qhe , does not appear to be relevant for fully spin polarized electrons .
the interaction between composite fermions in the second level is thus not of a form to produce nqhe .
it must be stressed , however , that the results do _ not _ rule out nqhe through some hitherto unknown mechanism , which would imply a different relation between @xmath42 and @xmath41 for a given filling factor than the one assumed above . in ref .
, which also investigated some of the above fractions , it was predicted that the ground state is a bubble crystal of composite fermions , similar to what was found in hartree fock studies of corresponding electronic states in higher landau levels@xcite . aside from several approximations , the results of that study were also predicated upon the quantitative validity of certain trial wave functions , making an independent confirmation worthwhile .
unfortunately , the spherical geometry used here is not convenient for the study of charge density wave states .
we do note in passing that a crucial step in ref . was to model the states at @xmath191 in terms of fermions at @xmath113 with an effective two - body interaction . a close similarity between our fig .
( [ fig9 ] ) and fig .
( 5 ) of ref . , obtained from the two - body interaction model [ both for @xmath192 provides support to the validity of this model . in the end
, we note that the above calculations assume a strictly two dimensional system ( with zero width ) , no disorder , and no mixing with higher electron landau levels .
these approximations represent an idealization of the actual experimental system .
it is expected that the idealized model , while not accurate _ quantitatively _
, should give the correct qualitative behavior .
it is a pleasure to acknowledge partial support by the national science foundation under grant no .
we are grateful to horst stormer , wei pan , and chia - chen chang for valuable discussions , to david weiss for suggesting the name `` nqhe , '' to the high performance computing ( hpc ) group led by v. agarwala , j. holmes , and j. nucciarone , at the penn state university aset ( academic services and emerging technologies ) for assistance and computing time with the lion - xe cluster , and acknowledge nsf dge-9987589 for computer support .
we warn the reader that the assignment of filling factor to a finite system involves physical interpretation , and therefore must be treated with caution .
the relation between @xmath41 and @xmath42 at filling factor @xmath11 must have the form @xmath193 , where @xmath83 is of oder unity , so as to produce the correct filling factor in the limit @xmath95 ( @xmath194 ) , but the value of @xmath83 , which is crucial for finite systems , depends on the physics of the incompressible state . in this paper ,
the filling factor is attributed to a finite system assuming the familiar physics that has been successful in explaining the fractions @xmath195 . a different origin of incompressibility
could possibly imply a different relation between @xmath41 and @xmath42 .
.the monopole strength @xmath43 as a function of @xmath41 for the nqhe states studied in this work . the quantities @xmath11 and @xmath10 are the the filling factors of electrons and composite fermions , with @xmath113 being the cf filling factor in the partially filled level .
@xmath41 is the total number of composite fermions , and @xmath81 is the number of composite fermions in the partially filled level , at filling @xmath113 .
@xmath43 and @xmath58 are the monopole strengths for the electrons and composite fermions .
[ tab1 ] [ cols="^,^,^,^,^,^",options="header " , ] | almost all quantum hall effect to date can be understood as _ integral _ quantum hall effect of appropriate particles , namely electrons or composite fermions .
this paper investigates theoretically the feasibility of nested states of composite fermions which would lead to a quantum hall effect that can not be understood as integral quantum hall effect of composite fermions .
the weak residual interaction between composite fermions will play a crucial role in the establishment of such quantum hall states by opening a gap in a partially filled composite - fermion level . to treat the problem of interacting composite fermions
, we develop a powerful method that allows us to obtain the low energy spectra at composite fermion fillings of @xmath0 without making any assumption regarding the structure of composite fermions in the topmost partially filled level .
the method is exact aside from neglecting the composite - fermion landau level mixing , and enables us to study rather large systems , for example , 24 particles at a total flux of 62 @xmath1 , for which the dimension of the lowest landau level hilbert space is @xmath2 .
we have investigated , for fully spin polarized composite fermions , several filling factors between 1/3 and 2/5 using this approach .
the results indicate that any possible incompressibility at these fractions is likely to have a fundamentally different origin than that considered earlier . |
in this paper we prove a product - to - sum formula that holds for the quantization of the moduli space of flat @xmath0-connections on the torus .
a quantization of this moduli space was given by witten in @xcite in conjunction with the jones polynomial , using a path integral on the space of connections .
witten s path integrals were put into rigorous framework by reshetikhin and turaev in @xcite using quantum groups .
this paper is based on their construction .
the algebra of observables is densely generated by the functions of the form @xmath2 , @xmath3 integers .
the quantization of such a function consists of the coloring of the curve of slope @xmath4 on the torus by the virtual representation @xmath5 , where @xmath6 is the greatest common divisor of @xmath7 and @xmath8 , @xmath9 , @xmath10 , and we denote by @xmath11 the @xmath12-dimensional irreducible representation of the quantum group of @xmath1 . in the work reshetikhin and turaev
@xmath13 is only defined for @xmath14 , but we will see below how this definition is extended to all integers .
let us denote by @xmath15 the operator that is associated by the quantization to @xmath2 .
a product - to - sum formula for the version of these operators defined using the kauffman bracket was discovered in @xcite and that is what inspired the following result .
[ mainresult ] in any level @xmath16 and for any integers @xmath17 the following product - to - sum formula holds @xmath18 so we see that the quantization of the algebra of the observables on the moduli space of flat @xmath0-connections on the torus is isomorphic to a subalgebra of the noncommutative torus defined by rieffel @xcite .
to be more rigorous , the above product - to - sum formula induces a @xmath19-product on the algebra of functions on the character variety on the torus .
this @xmath19-algebra is a subalgebra of the noncommutative torus .
the noncommutative torus is a @xmath20-algebra generated by @xmath21 , @xmath22 , @xmath23 , @xmath24 where @xmath21 and @xmath22 are two unitary operators satisfying the exponential form of the heisenberg noncommutation relations @xmath25 .
the inclusion is given by @xmath26 the elements @xmath27 are the noncommutative cosines , so theorem [ mainresult ] is nothing but the product - to - sum formula for noncommutative cosines .
of course , @xmath28 root of unity is the situation rieffel did not consider , for here the rotation angle in the definition of the noncommutative torus is rational .
what we consider to be a nice feature of this paper is that the proof of this result about the quantization of moduli spaces uses topological cut - and - paste techniques .
let us briefly discuss the case of the torus and then state the main result .
the moduli space of flat @xmath0-connections on the torus is the `` pillow case '' , the quotient of the complex plane by the lattice @xmath29 $ ] and by the symmetry with respect to the origin .
this space is the same as the character variety of @xmath0-representations of the fundamental group of the torus .
the algebra of observables , i.e. the algebra of functions on this space , is generated for example by the functions @xmath30 , where @xmath31 are integer numbers with @xmath32 and @xmath3 coprime .
a function of this form has a geometric interpretation , namely it associates to a conjugacy class of flat connections the trace of the holonomy of the connection along the curve of slope @xmath33 , where the trace is computed in the @xmath6-dimensional irreducible representation of @xmath0 .
witten defined the path integral only for functions of this form , but the definition can be extended to any ( smooth ) function on the character variety to yield an operator @xmath34 where @xmath35 is defined using approximations of @xmath36 by sums of traces of holonomies of the type describe above , @xmath37 is the chern - simmons functional for connections on the cylinder over the torus , @xmath12 being the level of quantization , and the path integral is taken over all connections @xmath38 that interpolate between two flat connections on the boundary components of the cylinder over the torus .
witten s construction was made rigorous by reshetikhin and turaev using the quantum group of @xmath1 at roots of unity . here
is how this is done for the torus .
fix a level @xmath39 , and let @xmath40 . for
an integer @xmath6 define @xmath41=(t^{2n}-t^{-2n})/(t^2-t^{-2})$ ] .
the quantum analogue of the universal enveloping algebra of @xmath42 is the algebra @xmath43 with generators @xmath44 satisfying @xmath45 this is a hopf algebra , thus its representations form a ring under direct sum and tensor product .
modulo a small technicality where one factors by the part of quantum trace zero , this ring contains a subring generated by finitely many irreducible representations @xmath46 . the representation @xmath13 has the basis @xmath47 and @xmath43 acts on this basis by @xmath48e_{j+1}\\ & & ye_j=[m - j+1]e_{j-1}\\ & & ke_j = t^{2j}e_j.\end{aligned}\ ] ] using the quantum version of the clebsch - gordon theorem , these irreducible representations can be defined recursively by @xmath49 .
this definition works only for @xmath50 . with these representations we color any knot @xmath51 in a 3-dimensional manifold , and the coloring
is denoted by @xmath52 .
let @xmath53 be the core @xmath54 of the solid torus @xmath55 , @xmath56 .
the hilbert space of the torus @xmath57 has an orthonormal basis given by @xmath58 , @xmath59 .
now the convention is made that @xmath60 and with this in mind we can extend the definition of @xmath61 to all @xmath6 via the recursion @xmath62 here @xmath63 is just @xmath53 colored by the representation @xmath64 . to make sense of these notations recall the pairing @xmath65 on @xmath57 obtained by gluing two solid tori such that the meridian of the first is identified with the longitude of the second and vice versa , as to obtain a 3-sphere .
this pairing is not an inner product but is nondegenerate .
pairing two elements of @xmath57 yields a colored link in the 3-sphere , and reshetikhin and turaev described in @xcite a way of associating a number to this link using the ribbon algebra structure of @xmath66 ( see also @xcite , @xcite ) .
this number is the value of the pairing .
for example it is well known that @xmath67.\end{aligned}\ ] ] let @xmath7 and @xmath8 be two integers , @xmath6 their greatest common divisor and @xmath9 , @xmath10 .
we describe now the operator @xmath68 that corresponds to the function @xmath69 ( @xmath70 stands for sine ) .
since the pairing is nondegenerate , it suffices to describe @xmath71 .
this number is computed by placing the curve of slope @xmath4 on the boundary of the solid torus , then gluing another torus to it to obtain the 3-sphere , coloring the @xmath4-curve by @xmath72 , the core of the first torus by @xmath13 and the core of the second torus by @xmath73 and then evaluating the colored link diagram as in @xcite .
we neglected the discussion about the framing , and the invariants of reshetikhin and turaev are defined for framed knots and links .
the curve @xmath74 will always be framed by the annulus @xmath75 $ ] while the curve of slope @xmath4 has the vertical framing ( i.e. the framing defined by the vector field orthogonal to the torus ) .
the quantization of the function @xmath76 is the operator @xmath77 here @xmath78 stands for cosine and the definition is motivated by the trigonometric formula @xmath79 these are the operators for which we prove the product - to - sum formula .
unlike the case of @xcite , here an easy inductive proof will not work .
the proof we give below was inspired by the computation of the colored jones polynomials of torus knots from @xcite .
it uses an apparatus called _ topological quantum field theory with corners _ , which enables the computation of quantum invariants of 3-manifolds through a successive application of axioms .
this system of axioms is in spirit analogous to the eilenberg - maclane system of axioms for homology .
the fundamental principles of a tqft with corners were summarized by walker in @xcite .
the construction of the basic data for the case of the quantum group @xmath43 was initiated in @xcite and completed in @xcite .
let us discuss just the facts about the @xmath1 tqft with corners needed in the proof of theorem [ mainresult ] . as explained in @xcite , a tqft in dimension three consists of a modular functor @xmath22 that associates vector spaces to surfaces and isomorphisms to homeomorphisms , and a partition function that associates to each three dimensional manifold a vector in the vector space of its boundary . walker refined this point of view in @xcite .
he considered decompositions of surfaces into disks , annuli , and pairs of pants , called dap - decompositions , which correspond through the functor to basis of the vector space .
his dap - decompositions ( which should probably be called rigid structures ) involve more structure than that .
this structure consists of some curves , called seams and drawn like dotted lines in diagrams , which keep track of the twistings that occur and are not detected by the change in dap - decompositions .
transformations of dap - decompositions are called moves and have the vector space correspondent of the change of basis . they are useful in preparing the boundary of a 3-manifold for a gluing with corners .
let us point out that for @xmath43 one needs also an orientation of the decomposition curves , as explained in @xcite where it was encoded using the klein four group , but this fact is irrelevant for all computations below , so we do not refer to it again .
also , walker considered framed , or extended manifolds , which are pairs @xmath80 with @xmath81 a manifold and @xmath6 an integer , necessary to make the partition function well defined . in this paper we begin the computation with manifolds with framing zero .
the framing only changes when we perform moves on boundary tori , and there the change is computed using the shale - weil cocycle .
however all relevant changes are performed back and forth , so the framing cancels .
thus we simply ignore it .
following @xcite we denote @xmath82 ^ 2}$ ] ( not to be confused with the generator of @xmath43 ) .
the vector space of an annulus has basis @xmath83 , @xmath84 , with pairing given by @xmath85.\end{aligned}\ ] ] gluing the boundaries of an annulus we obtain a torus , with the same basis for the vector space .
the vector space of the torus @xmath57 is the same as the hilbert space of the quantization of the moduli space of flat @xmath0-connections on the torus .
the pairing on the torus makes @xmath83 an orthonormal basis , but we do nt need this fact now .
the moves @xmath70 and @xmath86 on the torus are described in figure 1 .
the @xmath87-entry of the matrix of @xmath70 is @xmath88 $ ] .
the move @xmath86 is diagonal and its @xmath89th entry is @xmath90 .
the quantum invariant ( i.e. partition function ) of the cylinder over a surface is the identity matrix .
so the quantum invariant of the cylinder over an annulus is @xmath91}{x}\beta_n^n\otimes \beta_n^n\otimes \beta_n^n\otimes\beta_n^n$ ] ( when taking the cylinder over the annulus the boundary of the solid torus will be canonically decomposed into four annuli ) .
if the dap - decomposition of a 3-manifold involves two disjoint annuli , then its invariant can be written in the form @xmath92 gluing the two annuli produces a 3-manifold that in the newly obtained dap - decomposition has the invariant equal to @xmath93}v_{k , k}$ ] .
[ rootsofunity ] let @xmath94 , @xmath95 , @xmath96 , @xmath97 and @xmath98 be integers .
then @xmath99t^{bx^2}[cy]([x(y+d)]t^{2ey}+[x(y - d)]t^{-2ey})=\\ x^2t^{bc^2+be^2 - 2de}([a(c+e)]t^{2(be - d)c}+[a(c - e)]t^{-2(be - d)c}).\end{aligned}\ ] ] the proof of the main result is based on a theorem that is of interest in itself .
it describes how the observable @xmath15 acts on the hilbert space of the quantization . with our convention for defining @xmath100 for
all @xmath12 we have let @xmath97 to be the greatest common divisor of @xmath7 and @xmath8 , @xmath104 and @xmath105 we concentrate first on the computation of @xmath106 and @xmath107 here is the place where we use the topological quantum field theory with corners from @xcite .
the expression in ( [ invariant1 ] ) is the invariant of the link that has one component equal to the @xmath108-curve on a torus colored by @xmath109 , and the other two components the cores of the two solid tori that lie on one side and the other of the torus knot , colored by @xmath13 ( the one inside ) and @xmath73 ( the one outside ) . the expression in ( [ invariant2 ] ) is the invariant of the same link but with the @xmath108-curve colored by @xmath110 .
it was shown in @xcite that this number is equal to @xmath111 times the coordinate of @xmath112 of the vector that is the quantum invariant of the link complement .
let us produce the complement of this link by gluing together two simple 3-manifolds , whose quantum invariants are easy to compute .
consider first the cylinder over an annulus @xmath38 and glue its ends to obtain the manifold @xmath113 .
in the basis of the vector space of @xmath114 determined by the dap - decomposition @xmath115 the invariant of this manifold is @xmath116 .
take another copy of the same manifold . change the decomposition curves of the exterior torus of the first manifold to the @xmath4-curve and of the exterior torus of the second manifold to the longitude .
of course , to do this on the second manifold we apply the @xmath70-move and so the invariant of the second manifold changes to @xmath117\beta_\delta^\delta \otimes \beta _ { j_{n+1}}^{j_{n+1}}.\end{aligned}\ ] ] with the first manifold the story is more complicated .
consider the continued fraction expansion @xmath118 the required move on the boundary is then @xmath119 .
so the invariant of the first manifold in the new dap - decomposition is @xmath120t^{-a_n(j_n^2 - 1)}[j_nj_{n-1 } ] \cdots [ j_2j_1]t^{-a_1(j_1 ^ 2 - 1)}[j_1k]\beta_k^k\otimes \beta _ { j_{n+1}}^{j_{n+1}}.\end{aligned}\ ] ] now expand one annulus in the exterior tori of each of the two manifolds .
then glue just one annulus from the the first manifold to one annulus from the second . this way we obtain the complement of the link in discussion .
one of its boundary tori is decomposed into two annuli .
contract one of them .
since the gluing introduces a factor of @xmath121 $ ] , the invariant of the manifold is @xmath122}{[j_{n+1 } ] } [ j_{n+1}j_n]t^{-a_n(j_n^2 - 1)}[j_nj_{n-1 } ] \cdots [ j_2j_1]t^{-a_1(j_1 ^ 2 - 1)}\\ \times [ j_1k]\beta _ { j_{n+1}}^{j_{n+1}}\otimes \beta_\delta^\delta\otimes \beta_k^k.\end{aligned}\ ] ] at this moment we have the right 3-manifold but with the wrong dap - decomposition .
we need to fix the dap - decomposition of the torus that corresponds to the basis element @xmath123 ( the boundary of the regular neighborhood of the @xmath108-curve ) such as to transform the decomposition curve into the meridian of the link component . for this
we apply the move @xmath124 .
we obtain the following expression for the invariant of the extended manifold @xmath125[j_{2n+2}j_{2n+1 } ] t^{a_1j_{2n+1}^2}\cdots [ j_{n+2}j_{n+1}]\frac{[\delta j_{n+1}]}{[j_{n+1 } ] } \\ \times [ j_{n+1}j_n ] t^{-a_nj_n^2}[j_nj_{n-1 } ] \cdots [ j_2j_1]t^{-a_1j_1 ^ 2}[j_1k]\beta_\delta^\delta \otimes\beta_k^k\otimes \beta_m^m\end{aligned}\ ] ] ( in this formula we already reduced @xmath126 and @xmath127 , @xmath128 ) . now fix @xmath12 and @xmath129 , let @xmath130 and focus on the coefficients of @xmath131
. multiplied by @xmath132 these are the colored jones polynomial of the link @xcite with the @xmath108-curve colored by the @xmath133- , respectively @xmath134-dimensional irreducible representation of @xmath43 .
since @xmath135 and also @xmath136}{[j_{n+1}]}-\frac{[(d-1)j_{n+1}]}{[j_{n+1}]}= t^{2j_{n+1}}+t^{-2j_{n+1}},\end{aligned}\ ] ] we deduce that the value of @xmath137 is equal to @xmath138[j_{2n+2}j_{2n+1 } ] t^{a_1j_{2n+1}^2}\cdots [ j_{n+2}j_{n+1}](t^{2dj_{n+1}}+t^{-2dj_{n+1 } } ) \\ \times [ j_{n+1}j_n]t^{-a_nj_n^2}[j_nj_{n-1 } ] \cdots [ j_2j_1]t^{-a_1j_1 ^ 2}[j_1k].\end{aligned}\ ] ] we want to compute these iterated gauss sums .
we apply successively lemma [ rootsofunity ] starting with @xmath139 , then @xmath140 and so on to obtain @xmath141[j_{2n+2}j_{2n+1 } ] t^{a_1j_{2n+1}^2}\cdots [ j_{n+3}j_{n+2}]\\ & & \times ( [ j_{n-1}(j_{n+2}+d)]t^{-2a_ndj_{n+2 } } + [ j_{n-1}(j_{n+2}-d)]t^{2a_ndj_n+2 } ) \\ & & \times t^{-a_{n-1}j_{n-1}^2}[j_{n-1}j_{n-2 } ] \cdots [ j_2j_1]t^{-a_1j_1 ^ 2}[j_1k]\end{aligned}\ ] ] @xmath142[j_{2n+2}j_{2n+1 } ] t^{a_1j_{2n+1}^2}\cdots [ j_{n+4}j_{n+3}]\\ & & \times ( [ j_{n-2}(j_{n+3}+a_nd)]t^{-2(a_na_{n-1}-1)dj_n+3}+[j_{n-2}(j_{n+3}-a_nd ) ] t^{2(a_na_{n-1}-1)dj_n+3 } ) \\ & & \times t^{-a_{n-2}j_{n-2}^2 } [ j_{n-2}j_{n-3 } ] \cdots [ j_2j_1]t^{-a_1j_1 ^ 2}[j_1k]=\cdots \end{aligned}\ ] ] @xmath143[j_{2n+2}j_{2n+1}]\\ & & \times ( [ kj_{2n+1}+kdq']t^{-2dp'j_{2n+1 } } + [ kj_{2n+1}-kdq ' ] t^{2dp'j_{2n+1}})\\ & & = x^{-1 } t^{-pq}\sum_{j_{2n+1},j_{2n+2}=1}^{r-1 } [ mj_{2n+2}][j_{2n+2}j_{2n+1}]\\ & & \times ( [ kj_{2n+1}+kq]t^{-2pj_{2n+1 } } + [ kj_{2n+1}-kq ] t^{2pj_{2n+1}})\end{aligned}\ ] ] let us see how theorem [ mainresult ] follows from this result .
we compute : @xmath145 also @xmath146 on the one hand , and @xmath147 on the other .
an easy check of coefficients shows that the product - to - sum formula holds on the basis of the hilbert space of the quantization , and we are done .
the same method can be applied to prove the product - to - sum formula for the kauffman bracket skein algebra of the torus ( see @xcite ) using the topological quantum field theory with corners constructed in @xcite .
we conclude with a question .
is the product - to - sum formula true when the deformation variable @xmath28 is not a root of unity ?
the quantization for arbitrary @xmath28 is described in @xcite , and a proof of the formula should involve the analysis of the quasitriangular @xmath148 matrix of @xmath43 .
kirillov , n.y .
reshetikhin , _ representation of the algebra @xmath151 , @xmath8-orthogonal polynomials and invariants of links _ , in : infinite dimensional lie algebras and groups , ( ed . by v.g .
kac ) , adv . ser . in math
7 , world scientific , singapore 1988 , 285339 . | this paper exhibits a product - to - sum formula for the observables of a certain quantization of the moduli space of flat @xmath0-connections on the torus .
this quantization was defined using the topological quantum field theory that was developed by reshetikhin and turaev from the quantum group of @xmath1 at roots of unity . as a corollary
it is shown that the algebra of quantum observables is a subalgebra of the noncommutative torus with rational rotation angle .
the proof uses topological quantum field theory with corners , and is based on the description of the matrices of the observables in a canonical basis of the hilbert space of the quantization . |
powder samples of fesc@xmath0s@xmath1 were synthesized by solid state reaction of elemental fe , sc and s in an evacuated quartz tube .
all materials were handled in a glove box using standard air free techniques .
the sample was initially characterized by powder x - ray diffraction at 300 k using a cu-@xmath13 x - ray diffractometer .
additional high resolution synchrotron x - ray diffraction measurements were acquired @xmath14100 k using the powder diffractometer 11-bm located at the advanced photon source , these are described in appendix [ sec : s_xrd ] .
specific heat measurements were carried out on a cold pressed pellet of fesc@xmath0s@xmath1 using a quantum designs ppms .
the macs spectrometer at the nist center for neutron research was used for neutron scattering measurements . for measurements at ambient pressures ,
the 0.8 g sample of fesc@xmath0s@xmath1 was mounted in a standard al sample can .
measurements under hydrostatic pressure were conducted with the sample loaded into a stainless steel pressure cell and pressurized to 1 gpa with he gas .
the lattice parameter of a pyrolytic graphite standard loaded into the pressure cell just above the sample was used to monitor the quasi - hydrostatic pressure at low temperature .
macs was operated with a fixed final neutron energy of either 3.7 mev or 2.4 mev with appropriate combinations of be and beo filters located before and after the sample to suppress higher order neutron contamination .
no incident beam filter was used for measurements with incident energies above 5 mev , and data for these energies were corrected to account for contributions to the monitor count rate from higher order neutrons @xcite . to compare absolute intensities between the 0 kbar and
10 kbar measurements data collected in the neutron attenuating pressure cell were corrected for the measured energy dependent neutron transmission of the cell .
neutron diffraction measurements on the same sample were performed on bt-1 at nist utilizing a cu(311 ) monochrometer , neutron wavelength 1.54 , and 60 collimation before the monochrometer .
a set of high resolution neutron scattering measurements were conducted on the cncs spectrometer located at the spallation neutron source at oak ridge national lab @xcite . for these measurements a separate 0.45 g powder sample was prepared and cncs was operated with a fixed incident neutron energy of 1.55 mev and a resulting elastic energy resolution of 0.034 mev ( fwhm ) . for all scattering measurements background signal contributions from the sample environment were subtracted and signal count rates were converted to absolute values of the scattering cross - section using the integrated intensity from a nuclear bragg reflection @xcite .
we explicitly define the normalized magnetic neutron scattering intensity as used in this work to aid accurate discussion of sum - rules presented later . the magnetic neutron scattering cross section for a powder sample as a function of momentum transfer @xmath15 , and energy transfer @xmath16
is expressed as @xmath17 where n is the number of unit cells , @xmath18 and @xmath19 are the incident and final neutron wavevectors respectively , @xmath20 cm is the magnetic scattering length , and @xmath21 is the debye waller factor which we set to unity for low t and low q. the spherically averaged dynamic structure factor is given by @xmath22 where @xmath23 is the land g - factor and @xmath24 is the magnetic form factor . in this work
we report the normalized inelastic neutron scattering intensity per fe @xmath25 rietveld refinement of neutron and synchrotron diffraction data was carried out using the general structure analysis system ( gsas ) complimented with the sara_h _ program @xcite .
our main experimental finding is presented in fig . [
fig : tdepbragg ] ( a ) which displays a false color map of the temperature dependent neutron diffraction intensity as measured on macs .
the most prominent features are two structural bragg peaks at @xmath26 @xmath12 and @xmath27 @xmath12 that saturate the color scale and are independent of temperature below 20 k. more importantly , there are five temperature dependent bragg reflections which emerge in unison below 12 k. located at : 0.6 @xmath12 , 0.84 @xmath12 , 1.33 @xmath12 , 1.46 @xmath12 , and 1.77 @xmath12 , these peaks were not reported for fesc@xmath0s@xmath1 , although published diffraction data are not inconsistent with these findings as their sensitivity is insufficient to access this weak signal . in the following
we will argue that these bragg peaks arise from a magnetic ordering transition which is enabled by a higher temperature orbital ordering transition .
details of the lowest q magnetic bragg reflection are shown in fig .
[ fig : tdepbragg ] ( b ) and ( c ) .
we fit the high resolution data using a lorentzian profile convolved with a gaussian instrumental resolution of 0.045 @xmath12 full width at half maximum ( fwhm ) .
this reveals an intrinsic lorentzian half width at half maximum of @xmath28 @xmath12 and corresponding correlation length of @xmath29 .
similarly the energy dependence of the q - integrated peak was fit to a quasielastic lorentzian profile convolved with the gaussian instrumental resolution to yield a lower bound on the correlation time of @xmath30 ns .
the finite magnetic correlation length might result from exchange disorder built into the system at the higher temperature orbital occupational ordering transition as discussed below .
the specific heat of our powder sample shown in fig [ fig : specificheat ] has qualitative similarities to previous reports , but there are many quantitative differences @xcite .
no sharp anomaly signifying a phase transition is visible around 12 k and instead there is a very broad maximum in c / t at 8 k. this slightly higher than the 6 k maximum reported in earlier studies @xcite .
the turnover at 8 k is also considerably shaper in temperature ; both of these observations are consistent with a reduced disorder in the samples studied here .
phonon contributions to the specific heat were estimated by appropriately scaling the measured specific heat of the non - magnetic isostructural compound cdin@xmath0s@xmath1 as described in appendix [ sec : phonon_cp ] . the phonon contribution , shown as a dashed line in fig .
[ fig : specificheat ] ( a ) , was subtracted from the total specific heat to yield the magnetic component . from this
we calculate the magnetic entropy , @xmath31 shown in fig .
[ fig : specificheat ] ( b ) . rather than exceeding the @xmath32 entropy available for a @xmath33 system , the magnetic entropy plateaus at approximately 70 % of the available spin entropy ,
approaching @xmath32 at 80 k. we find no evidence for orbital fluctuations in the magnetic entropy at low temperatures , only above 80 k does the entropy begin to recover the additional @xmath34 available for a two - fold orbital degeneracy and our analysis indicates that the orbital degeneracy in fesc@xmath0s@xmath1 is entirely quenched below @xmath980 k. specific heat for powder samples of fesc@xmath0s@xmath1 used in this work .
( a ) total specific heat ( grey circles ) and estimated magnetic contribution ( blue triangles ) . dashed yellow line is the estimated phonon contribution .
( b ) estimated magnetic entropy of fesc@xmath0s@xmath1 . ]
an orbitally non - degenerate state at low temperatures is at odds with the previously accepted crystal structure for fesc@xmath0s@xmath1 we shall thus first re - examine the low temperature crystal structure .
the t=20 k neutron diffraction pattern of fesc@xmath0s@xmath1 is shown in fig .
[ fig : structure ] ( a ) .
all peaks can be indexed by the cubic @xmath35 space group and no peak splitting characteristic of a symmetry lowering structural distortion is apparent .
nonetheless , rietveld refinement of our diffraction data using the @xmath35 space group is not satisfactory as it yields a systematic discrepancy between measured and calculated peak intensities . amongst the maximal subgroups of @xmath35 we find a significantly improved description of the data using the tetragonal @xmath36 space group .
specifically , and as described more fully in appendix [ sec : struct_models ] , the goodness of fit @xmath37 is reduced from @xmath38 to @xmath39 in going from @xmath35 to @xmath36 .
[ fig : structure ] ( b ) shows the new unit cell and the resulting rietveld refinement is shown in fig .
[ fig : structure ] ( a ) , details of the refinement are presented in table [ tab : tetrag_struct ] .
( a ) neutron powder diffraction pattern for fesc@xmath0s@xmath1 measured on bt-1 at 20 k , open symbols are observed intensities and solid red line is a rietveld refinement of the nuclear structure to the @xmath36 space group .
( b ) proposed @xmath36 crystal structure of fesc@xmath0s@xmath1 , the refined lattice parameters are @xmath40 and @xmath41 .
there are two crystallographically distinct sulfur sites , labelled s(1 ) and s(2 ) , resulting in two different fe@xmath4 coordinating tetrahedra alternately expanded and contracted along the tetragonal ( 001 ) direction .
fe@xmath4 orbital occupancies for small distortions of the s tetrahedra are shown on the right .
the parent cubic cell is shown as a light grey dashed outline . ] within our measurement resolution the fesc@xmath0s@xmath1 unit cell maintains its cubic metric and the tetragonal distortion is limited to @xmath42 , where the parent cubic cell lattice parameter @xmath43 and @xmath44 is the tetragonal cell lattice parameter .
the @xmath36 unit cell contains _ two _ crystallographically distinct fe sites , each site comprising a _ centered tetragonal _ ( ct ) sublattice .
importantly , the fe - coordinating sulfur tetrahedra for each respective sublattice are alternately compressed or elongated along the ( 001 ) direction , as shown in fig .
[ fig : structure ] ( b ) .
the tetragonal distortion thus relaxes the tetrahedral symmetry around each iron site , lifting the orbital degeneracy of the fe @xmath6 orbitals and results in a hole occupying the @xmath45 orbital for one sublattice and the @xmath46 orbitals for the other . from an energetic perspective
, this particular distortion minimizes macroscopic strain while fully quenching the orbital moment .
high resolution x - ray diffraction measurements show that the average structure is cubic at least to 100 k , but reveal a large anisotropic microstrain broadening of the diffraction peaks possibly indicating an incipient structural transition ( see appendix [ sec : s_xrd ] .
further neutron and x - ray diffraction experiments that span the relevant temperature regime where the phase transition may occur are under way .
while no deviations from cubic symmetry have been reported thus far , the particular low temperature structure we report resolves a number of inconsistencies in the literature .
first , mssbauer spectroscopy consistently indicates two distinct fe sites @xcite .
this has been ascribed to result from fe - sc site mixing at the level of 30 % @xcite , but seems an extremely unlikely scenario given the disparate ionic radii and charge of fe@xmath4 and sc@xmath47 , and is furthermore inconsistent with neutron diffraction , which offers excellent contrast between fe and sc .
even at room temperature the mssbauer spectra necessitate a second fe site in small fractions and below @xmath950 k the quadrupolar splitting from each respective site diverges .
these observations seem consistent with a thermally fluctuating , or dynamic , distortion which dramatically slows down below 50 k. second , optical measurements in the far infrared show signatures of symmetry breaking @xcite .
two bands around 467 cm@xmath12 are observed in the 300 k absorption spectra which are forbidden in the @xmath35 space group and have lacked a satisfactory explanation @xcite .
the extra absorption bands indicate a local distortion is present , at least dynamically , even at room temperature .
.atomic parameters for the proposed low temperature tetragonal structure of fesc@xmath0s@xmath1 at 20 k. the space group is @xmath36 ( 119 ) with lattice parameters @xmath40 and @xmath41 .
reitveld refinements resulted in a @xmath37 of 1.33 and @xmath49% . [ cols="<,^,^,^,^,^,^",options="header " , ] guided by the relative strength of spin correlators and magnetic bond energies extracted from the static structure factor and first moment sum rule analysis , the magnetic exchange interactions where chosen to provide the best qualitative match of the data with the simplest combination of exchange terms .
exchange constants used for each respective spin wave calculation are detailed in table [ tab : spinw_params ] .
the ground state energy of each magnetic structure are as follows : @xmath50 ground state energies of models 3 and 4 are identical . for these models heisenberg nn ( @xmath51 ) and out - of - plane nnn ( @xmath52 ) exchange terms cancel at the mean field level and the systems is effectively a _ two - dimensional _ heisenberg square lattice . for models 1 and 2
the nearest neighbor spins are orthogonal and heisenberg terms have no effect .
46ifxundefined [ 1 ] ifx#1 ifnum [ 1 ]
# 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty link:\doibase 10.1126/science.1163196 [ * * , ( ) ] link:\doibase 10.1038/nature08917 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.78.2799 [ * * , ( ) ] link:\doibase 10.1143/jpsj.67.3703 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.85.3946 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.85.3950 [ * * , ( ) ] link:\doibase 10.1126/science.1212154 [ * * , ( ) ] link:\doibase 10.1103/physrevx.2.041013 [ * * , ( ) ] link:\doibase 10.1103/physrevb.73.014413 [ * * , ( ) ] link:\doibase 10.1038/nphys622 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.101.047201 [ * * , ( ) ] link:\doibase 10.1103/physrevb.79.134406 [ * * , ( ) ] link:\doibase 10.1073/pnas.1107861108 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.92.116401 [ * * , ( ) ] http://stacks.iop.org/1367-2630/6/i=1/a=191 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.94.237402 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.102.096406 [ * * , ( ) ] link:\doibase 10.1103/physrevb.80.224409 [ * * , ( ) ] link:\doibase 10.1103/physrevb.91.125112 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.114.207201 [ * * , ( ) ] link:\doibase 10.1103/physrevb.92.094413 [ * * , ( ) ] link:\doibase 10.1103/physrevb.69.014502 [ * * , ( ) ] \doibase http://dx.doi.org/10.1063/1.3626935 [ * * , ( ) ] link:\doibase 10.1103/physrevb.56.8091 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/s0921-4526(99)01722-6 [ * * , ( ) ] , @noop * * , ( ) \doibase http://dx.doi.org/10.1016/j.jmmm.2008.04.151 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/s0925-8388(01)01772-8 [ * * , ( ) ] link:\doibase 10.1103/physrevb.81.184426 [ * * , ( ) ] link:\doibase 10.1103/physrevb.88.174415 [ * * , ( ) ] link:\doibase 10.1103/physrevb.89.174431 [ * * , ( ) ] link:\doibase 10.1103/physrevb.78.144417 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.84.3718 [ * * , ( ) ] link:\doibase 10.1103/physrevb.10.128 [ * * , ( ) ] link:\doibase 10.1103/physrevb.64.144405 [ * * , ( ) ] http://stacks.iop.org/0953-8984/27/i=16/a=166002 [ * * , ( ) ] link:\doibase 10.1103/physrevb.82.041105 [ * * , ( ) ] link:\doibase 10.1103/physrevb.84.064438 [ * * , ( ) ] link:\doibase 10.1103/physrevb.73.132409 [ * * , ( ) ] link:\doibase 10.1134/1.1921323 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.62.2056 [ * * , ( ) ] link:\doibase 10.1103/physrevb.19.3580 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.100.015703 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.62.2507 [ * * , ( ) ] link:\doibase 10.1107/s0021889898006001 [ * * , ( ) ] link:\doibase 10.7566/jpsj.82.024702 [ * * , ( ) ] | we present neutron scattering measurements on powder samples of the spinel fesc@xmath0s@xmath1 that reveal a previously unobserved magnetic ordering transition occurring at 11.8(2 ) k. magnetic ordering occurs subsequent to a subtle cubic - to - tetragonal structural transition which distorts fe coordinating sulfur tetrahedra lifting the orbital degeneracy .
the application of 1 gpa hydrostatic pressure appears to destabilize this nel state , reducing the transition temperature to 8.6(8 ) k and redistributing magnetic spectral weight to higher energies .
the relative magnitudes of ordered @xmath2 and fluctuating moments @xmath3 show that the magnetically ordered ground state of fesc@xmath0s@xmath1 is drastically renormalized and in proximity to criticality .
= 4 in a quantum spin liquid quantum fluctuations compete with , and ultimately overwhelm , any tendency towards the formation of a classical long range magnetic order , effectively melting the staggered magnetization at zero temperature .
the search for this precarious state of matter in a real material is now a decades old preeminent theme of condensed matter physics @xcite . in insulating antiferromagnets
, research has largely been concentrated on geometrically frustrated materials , where the underlying lattice structure results in a competition amongst magnetic exchange interactions .
the result is an extensive degeneracy that promotes quantum fluctuations and precludes the development of a staggered magnetization @xcite .
but frustration is not the only game in town .
other degrees of freedom may enhance magnetic fluctuations .
for example , orbital degeneracy which can in turn enhance quantum fluctuations and suppress magnetic order @xcite .
the orbital degrees of freedom may even remain disordered in the presence of a staggered magnetization forming an orbital liquid state @xcite , or both spins and orbitals may fail to develop temperature independent correlations in a so - called spin - orbital liquid state .
spinel compounds , ab@xmath0x@xmath1 , with magnetic ions occupying the a - site diamond sublattice form a simple , and therefore important , three dimensional frustrated lattice where competing exchange interactions promote unconventional magnetic ground states stabilized by fluctuations . amongst the a - site spinels
, fesc@xmath0s@xmath1 holds a special place as the fe@xmath4 ( 3d@xmath5 ) ions are both magnetically and orbitally active .
the tetrahedral crystal field of the a - site environment splits the fe 3d manifold into a lower @xmath6 doublet and higher energy @xmath7 triplet , while hund s coupling yields a high spin ( s=2 ) configuration , with a single hole occupying the @xmath6 doublet .
thus , the fe@xmath4 ion in fesc@xmath0s@xmath1 is jahn - teller active yet , surprisingly , there have been no reported observations of any structural distortion or magnetic ordering down to 50 mk . in particular , while the magnetic susceptibility exhibits curie - weiss like behaviour with @xmath8 k @xcite ; there is no clear indication of a phase transition .
moreover the specific heat only exhibits a broad peak with maximum centered at 8 k. neutron scattering measurements on polycrystalline samples conducted so far have not provided evidence for magnetic ordering .
the data show intense dispersive spin excitations emerging from momentum transfer corresponding to the cubic ( 100 ) position with an apparent excitation gap of @xmath90.2 mev @xcite . based on these experimental observations
it was proposed that fesc@xmath0s@xmath1 is a spin - orbital liquid : an exotic state arising from competition between on site spin - orbit coupling and kugel - khomskii exchange in the fe diamond sub - lattice @xcite . while spin and orbital ordering is favoured by the exchange interactions , atomic spin orbit coupling favours local spin - orbital singlets and fesc@xmath0s@xmath1
is predicted to lie near the quantum critical point separating the spin - orbital singlet phase from a magnetically and orbitally ordered state .
this hypothesis appears supported by terahertz and far infrared optical spectroscopy measurements which have revealed low energy ( 4.5 mev ) excitations consistent with a so - called spin - orbiton @xcite , an excitation of entangled spin and orbital degrees of freedom . in this scenario , varying the magnetic exchange interactions relative to the spin orbit coupling through the application of hydrostatic pressure might drive fesc@xmath0s@xmath1 through the quantum critical point , stabilizing antiferromagnetism @xcite .
( a ) overview of the temperature dependent neutron diffraction intensity in fesc@xmath0s@xmath1 as measured on macs .
( b ) magnetic bragg reflection measured on cncs integrated between @xmath10 mev , an identical constant background has been subtracted from low and high temperature data sets .
( c ) energy dependence of magnetic bragg reflection integrated over @xmath11 @xmath12 . horizontal bars in ( b ) and ( c ) indicate the instrumental resolution as determined from corresponding cuts through the nuclear ( 111 ) bragg peak at 1.04 @xmath12 .
error bars represent one standard deviation . ]
here we show that contrary to the spin - orbital liquid scenario , fesc@xmath0s@xmath1 undergoes a magnetic ordering transition concomitant with the maximum observed in specific heat .
our data also indicate a subtle tetragonal , jahn - teller - like , distortion occurring at higher temperatures which lifts the orbital degeneracy so that the low temperature dynamics can be described in terms of a spin - only hamiltonian . additionally , we find that the application of hydrostatic pressure of 1 gpa reduces the magnetic ordering temperature and shifts inelastic spectral weight to higher energies , apparently destabilizing the ordered state .
the nel state in fesc@xmath0s@xmath1 is strongly renormalized with a sublattice magnetization that is reduced to 44(1)% of the full saturation moment and 40(2)% of the inelastic spectral weight associated with non - spin wave fluctuations .
these results show that fesc@xmath0s@xmath1 is indeed near a quantum critical point , but surprisingly , on the magnetically ordered side of this critical point . |
the moutard transformation plays a fundamental role in projective differential geometry of surfaces and was extensively investigated by bianchi , darboux , demoulin , guichard and others . in recent publications
@xcite we gave an application of the moutard transformation to the explicit construction of two - dimensional schrdinger operators @xmath1 with fast decaying smooth rational potentials such that their @xmath2-kernels contain at least two - dimensional subspaces spanned by rational eigenfunctions as well as a @xmath0-dimensional extension of the moutard transformation which was able to produce explicit rational blowing - up solutions to the novikov
veselov ( nv ) equation with fast decaying smooth rational cauchy data .
purely algebraic constructions of the aforementioned papers imply that one should look for an algebraic interpretation of the moutard transformation as a simple transformation in the ring of partial differential operators .
the one - dimensional case of darboux transformation was studied in @xcite .
here we give an analogue of their results ; as it turned out , for the moutard transformation one needs to recourse to a more complicated ring of _ rational _ pseudodifferential operators ( ore localization of the ring of partial differential operators ) . as a by - product
, we obtain a more natural version of the dubrovin - krichever - novikov formalism @xcite of algebro - geometric two - dimensional schrdinger operators . in the last section we briefly sketch our previous results @xcite .
first we give an account of the algebraic theory of the darboux transformations .
let @xmath3 be a one - dimensional schrdinger operator and let @xmath4 satisfy the equation @xmath5 the function @xmath4 determines a factorization of @xmath6 : @xmath7 indeed we have @xmath8 and the equation @xmath9 is equivalent to @xmath10 . if @xmath11 is real - valued we have @xmath12 _ the darboux transformation _ is the swapping of @xmath13 and @xmath14 : @xmath15 or in terms of @xmath16 : @xmath17 it is easy to check by simple computations that [ prop1 ] if @xmath18 satisfies the equation @xmath19 with @xmath20 then @xmath21 satisfies the equation @xmath22 remark . in general the darboux transformation
is defined for any solution to the equation @xmath23 with @xmath24 ( see , for instance , @xcite ) . in this case
it reduces to the transformation of @xmath25 for which @xmath26 .
this transformation was invented by th .
moutard for the hyperbolic equation of the form @xmath27 in the context of local differential geometry of surfaces and was extensively studied and used in many areas of the local surface theory @xcite . here
we give an elliptic version @xcite of this transformation suitable for our purposes .
let @xmath6 be a two - dimensional potential schrdinger operator and let @xmath4 be a solution to the equation @xmath28 then the moutard transformation of @xmath6 is defined as @xmath29 [ prop3 ] if @xmath18 satisfies the equation @xmath30 , then the function @xmath31 defined from the system @xmath32 satisfies @xmath33 . by this definition , if @xmath31 satisfies ( [ eigen - moutard ] ) then @xmath34 satisfies ( [ eigen - moutard ] ) for any constant @xmath35 .
we shall use the following notation for the moutard transformation : @xmath36 for one - dimensional potentials the moutard transformation reduces to the darboux transformation . indeed ,
let @xmath37 depend on @xmath38 only and @xmath39 .
then @xmath40 satisfies the one - dimensional schrdinger equation @xmath41 and the moutard transformation reduces to the darboux transformation of @xmath42 defined by @xmath40 : @xmath43 if @xmath44 satisfies @xmath45 , then @xmath30 with @xmath46 . we derive from ( [ eigen - moutard ] ) that @xmath47 satisfies @xmath48 if @xmath49 , i.e. @xmath50 is a multiple of the darboux transform of @xmath51 : @xmath52 where @xmath53 is the factorization of @xmath54 defined by @xmath40 .
the inverse darboux transformation is given by @xmath55 .
. we can rewrite ( [ eigen - moutard ] ) as @xmath56 which implies @xmath57 where , as usual , @xmath58 , @xmath59 , @xmath60 .
this representation will be used later .
there is another two - dimensional generalization of the darboux transformation called the laplace transformation .
its relation to integrable systems was recently studied in @xcite .
since weakly algebraic two - dimensional schrdinger operators are defined in terms of their spectral curves on the zero energy level we recall the definitions of the spectral curves of one- and two - dimensional schrdinger operators .
let a one - dimensional schrdinger operator @xmath61 commute with an ordinary differential operator @xmath62 of order @xmath63 .
then the burchnall chaundy theorem @xcite guarantees that these two operators satisfy a polynomial equation @xmath64 where @xmath65 is a polynomial of degree @xmath63 .
this equation defines _ the spectral curve _
@xmath66 of the operator @xmath6 .
its points parameterize the joint `` eigenfunctions '' , i.e. solutions of the equations @xmath67 moreover these functions are glued ( after some normalization ) into a function @xmath68 which is meromorphic in @xmath69 and after completing @xmath66 by adding an infinity point @xmath70 the eigenfunction @xmath71 gets a singularity at @xmath72 : @xmath73 where @xmath74 is a local parameter on @xmath66 near the infinity point @xmath72 @xcite . therewith it is said that the operator @xmath6 is _ algebraic _ ( or _ algebro - geometric _ ) .
_ the spectral curve of a constant potential .
_ let @xmath75 be a constant potential , then @xmath66 is given by @xmath76 , @xmath77 and the covering @xmath78 ramifies at @xmath79 where @xmath80 . given a partial differential operator @xmath6
, one says that operators @xmath81 generate a commutative @xmath82 algebra @xmath83 if they satisfy the following commutation relations : @xmath84=d_{ij}h , \ \ \ [ l_i , h ] = d_i h\ ] ] where @xmath85 , are partial differential operators .
this definition was introduced in @xcite in which they considered in detail the particular case when @xmath6 is the two - dimensional schrdinger operator ( probably with an electromagnetic field ) in two variables and introduced the following definition : * a two - dimensional operator @xmath86 is called _ ( weakly ) algebraic _ if it is included in a nontrivial commutative @xmath82 algebra @xmath83 generated by operators @xmath87 and @xmath2 in two variables , the operators @xmath87 and @xmath2 satisfy the polynomial relation @xmath88 and to a generic point of the algebraic curve @xmath89 there corresponds a @xmath90-dimensional space of functions @xmath71 which satisfy the equations @xmath91 the dimension @xmath90 is called the rank of @xmath83 .
the darboux transformation admits the well - known representation as a conjugation in the ring of pseudodifferential operators which implies interesting corollaries concerning algebraic operators .
let @xmath92 and @xmath93 .
then in the ring of pseudodifferential operators in @xmath38 we have @xmath94 where the functions @xmath4 and @xmath95 are identified with the operator of multiplication by them . from
that we conclude @xmath96 fixing @xmath4 let us denote by @xmath97 the conjugation of a pseudodifferential operator @xmath98 .
it is clear that if @xmath62 commutes with @xmath6 , then @xmath99 commutes with @xmath100 : @xmath101= 0 \ \ \ \rightarrow \ \ \ [ \widetilde{h},\widetilde{l}]=0.\ ] ] however , given an arbitrary differential operator @xmath98 , we have @xmath102 where @xmath103 is the function obtained by applying @xmath98 to @xmath4 and @xmath104 is a differential operator . hence @xmath105 @xmath106 and
we conclude that _
@xmath107 is a differential operator if and only if @xmath108 , which means @xmath109 _ let @xmath110 and let @xmath62 be a differential operator of odd order which commutes with @xmath6 .
we assume that its order is minimal with respect to this property .
let , by the burchnall chaundy theorem , @xmath111 then we have ( see @xcite ) : 1 .
if @xmath112 and @xmath113 has no a multiple root at @xmath114 , then @xmath100 and @xmath99 generate a commutative ring of differential operators and @xmath115 2 .
if @xmath116 , then @xmath100 and @xmath117 generate a commutative ring of differential operators and @xmath118 3 . if @xmath112 and @xmath113 has a multiple root at @xmath114 , then the action of the darboux transformation is inverse to one described in the previous statement : @xmath100 and some operator @xmath119 generate commutative ring of differential operators with @xmath120 in particular , this implies that * the darboux transformation preserves the class of algebro - geometric one - dimensional schrdinger operators .
moreover it always preserves the normalization of the spectral curve .
in @xcite it is noted that some facts mentioned above were discovered by burchnall , chaundy and drach .
the first two cases are demonstrated by the following example . _
the darboux transformation of the constant potential .
_ let @xmath121 and @xmath122 . 1 .
let @xmath123 .
then @xmath124 , @xmath125 , and the darboux transformation is even trivial . 2 .
let @xmath126 . then @xmath127 @xmath128 if @xmath129 , then @xmath130 is not periodic .
an example of the third case is easily derived from the following explicitly computable examples which are interesting in themselves .
_ rational solitons via the darboux transformation .
_ let @xmath131 and @xmath132 .
then @xmath133 the spectral curve @xmath134 is transformed to @xmath135 and @xmath136 ( we normalize it by condition @xmath137 as @xmath138 ) .
the iterations of the darboux transformation initially applied to the trivial potential @xmath139 give all rational solitons discovered in @xcite .
the spectral curve of the potential @xmath140 obtained after @xmath141 iterations , is given by the equation @xmath142 .
the spectral curves of these are singular : topologically they are spheres but at @xmath114 they have the following singularities : _ any rational function @xmath40 on @xmath143 , the spectral curve of @xmath144 , which is holomorphic near @xmath114 satisfies the condition _
@xmath145 this is easily explained by the normalization mapping @xmath146 which has the form @xmath147 indeed any rational function @xmath40 which is holomorphic at @xmath114 is a ratio of polynomials @xmath148 such that @xmath149 and any such function written in terms of @xmath35 satisfies the conditions ( [ singular-1 ] ) . in order to give an analogous interpretation of the moutard transformation as a conjugation we first recall some basic facts from the ore theory of localization of noncommutative rings and
introduce the rings @xmath150 and @xmath151 $ ] which we shall use .
we recall that a ring @xmath152 is an algebra with two operations , the addition and the multiplication , such that @xmath152 is a commutative group with respect to the addition and the distribute laws @xmath153 hold .
it is said that a ring @xmath152 is regular if it satisfies _ the ore conditions _ : * for @xmath154 there exist @xmath155 such that @xmath156 ; * if @xmath157 or @xmath158 for @xmath159 then @xmath160 ( so it has no zero divisors ) .
ore showed in @xcite that _ any regular ring can be embedded as a subring into a non - commutative field ( skew field of fractions ) _ as follows .
let us consider the set @xmath161 of all formal fractions @xmath162 _ note _ that due to non - commutativity of @xmath152 one shall always keep in mind that in our construction the denominators @xmath163 are always on the _ right _ of the numerators @xmath164 !
it is easy to propose an analogous construction with denominators standing on the left , but the resulting skew field will be isomorphic to the field we are to construct .
we say that two such fractions are equal : @xmath165 if and only if @xmath166 where @xmath167 and @xmath168 satisfy the equality ( see the first ore condition ) @xmath169 it is easy to show that the equality is independent on the choice of @xmath167 and @xmath168 satisfying the last equality
. then the addition is defined by @xmath170 where @xmath167 and @xmath168 satisfy ( [ ore - pair ] ) .
the multiplication is given by @xmath171 where @xmath172 the unit element is defined as @xmath173 this definition is independent on the choice of @xmath164 and with such a law of multiplication we have @xmath174 .
we say that @xmath161 ( with these operations ) is _ the ore localization _ of @xmath152 .
one can routinely check all the usual properties of the addition and multiplication operations defined on @xmath161 , which becomes a _
skew field_. this means that we can use expressions like @xmath175 for arbitrary elements of this skew field .
all operations in @xmath161 given above are constructive , unlike the case of the ring of formal pseudodifferential operators defined as infinite series customary in solitonics . for a commutative ring @xmath152 without divisors of zero the ore localization
coincides with the standard localization .
let us introduce the main example with non - commutative ring @xmath152 which we shall use . example .
let @xmath176 be a field formed by all @xmath177-valued functions of the form @xmath178 where @xmath179 and @xmath180 are analytical functions of @xmath181 or even formal power series in these variables .
we consider two cases which are \1 ) @xmath182 and @xmath183 ; \2 ) @xmath184 , @xmath185 and @xmath186 , and , for simplicity and our needs , we even assume that @xmath187 , @xmath188 , and @xmath189 .
let us consider @xmath190 as an operator algebra @xmath152 which acts on itself by multiplications @xmath191 take the partial derivative operators @xmath192 and consider the operator algebra @xmath193 $ ] which is generated by elements of @xmath152 and by the operator @xmath194 that acts on @xmath190 .
the ring @xmath193 $ ] is noncommutative because @xmath195 = \partial_y \cdot f - f \cdot \partial_y = \frac{\partial f}{\partial y } = f_y \ \
\mbox{for $ f \in f$}.\ ] ] it is straightforward to check @xmath193 $ ] and @xmath196 $ ] satisfy the ore conditions .
let us take the ore localization @xmath150 of @xmath193 $ ] .
we have [ commutator ] for any @xmath197 we have @xmath198 = - \partial_y^{-1}\cdot \frac{\partial f}{\partial y}\cdot \partial^{-1}_y.\ ] ] proof .
let us take the equality ( [ comm ] ) , multiply every its term by @xmath199 from both sides and obtain @xmath200 = \partial_y^{-1 } \cdot f_y \cdot \partial_y^{-1}$ ] .
proposition is proved .
let us consider @xmath151 $ ] , the ring of formal differential operators in @xmath38 with coefficients from @xmath150 .
this ring is embedded into its ore localization @xmath201 . we need to define the commutation of @xmath202 with elements of @xmath151 $ ] . since @xmath203 = - \partial_x^{-1}[\partial_x , f]\partial_x^{-1},\ ] ] it is enough to define the commutators of @xmath204 $ ] with @xmath205 . here
we do that even for the ore localization ring @xmath201 .
if @xmath206 $ ] is a differential operator in @xmath38 and @xmath207 with coefficients from @xmath190 , then its derivative in @xmath38 ( i.e. the operator with coefficients being the derivatives of the respective coefficients of @xmath62 ) may be defined by the commutation formula @xmath208.\ ] ] let us take this formula as a definition of the derivative of any element of @xmath209 w.r.t .
@xmath38 . using the leibnitz identity
@xmath210 we derive the following proposition by straightforward computations .
given @xmath211 with @xmath212 $ ] , we have @xmath213 = m(l^{-1})_x + m_x l^{-1 } = -m\,l^{-1}l_xl^{-1 } + m_x l^{-1}.\ ] ] [ cor1 ] if @xmath214 , then @xmath215 \in f(\partial_y)$ ] , i.e. @xmath216 contains no derivations in @xmath38 .
let us consider the general moutard transformation which is applied to an operator : @xmath217 where @xmath218 or @xmath219 via the formulas @xmath220 it relates solutions @xmath18 and @xmath31 to the equations @xmath221 where @xmath222 let us consider the ore localization @xmath223 of the ring of differential operators in @xmath168 and the ring @xmath224 $ ] of differential operators in @xmath167 with coefficients from ( noncommutative ) ring @xmath223 . given @xmath4 , a solution to the equation @xmath110
, we consider the differential operators @xmath225$]:@xmath226 and their `` ratio '' , the operator @xmath227 of the form @xmath228 we denote by @xmath98 and @xmath107 the following formal operators of the first order from @xmath224 $ ] : @xmath229 m = _ s^-1 = _ r - _ s^-1 . @xmath230 [ th1 ] given @xmath4 satisfying @xmath231 and the moutard
transform @xmath130 of @xmath16 ( we assume that the transformation is generated by @xmath4 ) , the operators @xmath98 and @xmath107 are conjugated in @xmath224 $ ] by @xmath227 : @xmath232 remark .
it is easy to check that @xmath6 and @xmath100 are not conjugated by @xmath227 : @xmath233 .
first we expose some auxiliary facts .
[ l3 ] @xmath234 proof of lemma .
it is enough to show that @xmath235 .
let us write down the right - hand side of this equality as @xmath236 @xmath237 this proves the lemma .
[ sled4 ] @xmath238 now we are ready to prove theorem [ th1 ] .
it to enough to establish the equivalent equality @xmath239 i.e. @xmath240 by lemma [ l3 ] , the left - hand side of ( [ sopr3 ] ) is @xmath241 _ s-(-u+)(1 - _
s^-1_s)= @xmath242 = + @xmath242 + u(1 - _ s^-1_s ) - ( 1 - _ s^-1_s ) = @xmath242 = ( ( ) _ s _
s^-1_s + _ s _ s^-1_s . .
- ( ) _ s _ s^-1_rs - _ s_s^-1_rs ) + @xmath242 + _
sa^-1 b_r + u - _ s^-1_s - + _ s^-1_s = @xmath242 = ( _ s^-1_s - _ s^-1_s + + ( ) _ s^-1_rs - _ rs)+ @xmath242 + _
sa^-1 b_r + u - _ s^-1_s - + _ s^-1_s = @xmath242 = ( ) _ s^-1(u ) - 2u + _
sa^-1 b_r + u = @xmath242 = ( ) _ s^-1(u ) + _
sa^-1 b_r - u .
sa^-1 b_s^-1(_r_s - u)= _
sa^-1 b_r - _ s(1 - _ s^-1_s)_s^-1 u= @xmath242 = _
sa^-1 b_r - u + ( ) _ s_s^-1 _ s_s^-1 u + ( ) _ s_s^-1_s_s^-1 u= @xmath242 = _
sa^-1 b_r - u - _ s^-1_s_s^-1 u + _
s^-1 u. @xmath230 let us apply proposition [ commutator ] to the third term @xmath244 in the last formula and finally derive @xmath245 @xmath246 which coincides with the left - hand side of ( [ sopr3 ] ) .
theorem is proved .
let us assume that @xmath6 is a ( weakly ) algebraic operator , i.e. , there are differential operators @xmath87 and @xmath2 such that @xmath247 = d_0 h , \ \ \ [ l_1,h ] = d_1 h , \ \ \ [ l_2,h ] = d_2 h\ ] ] and there is a polynomial @xmath180 in two variables with constant coefficients such that @xmath248 by applying the euclid algorithm ( division with remainder in the ring of formal linear _ ordinary _ differential operators @xmath224 $ ] ) we obtain @xmath249 such that @xmath250 with @xmath251 $ ] .
[ th2 ] 1 .
@xmath252 = 0 $ ] ; 2 .
@xmath253 ; 3 .
@xmath254 = [ r_2,m]=0 $ ] . proof . 1 . using ( [ kri1 ] ) we compute that @xmath255=[l_1-q_1 h , l_2-q_2 h]=s_1\cdot h , \ \
\ s_1 \in f(\partial_s)[\partial_r].\ ] ] however in the left - hand side we have an element of @xmath224 $ ] which contains no derivations in @xmath167 which implies that @xmath256 .
the equality @xmath257 means that @xmath258 in @xmath259 $ ] which , by ( [ kri1 ] ) , implies @xmath260.\ ] ] however in the left - hand side we have an element of @xmath224 $ ] with no derivations in @xmath167 which implies @xmath261 .
we have @xmath262=[l_1-q_1 h , h]=s_3\cdot h,\ ] ] @xmath263=[r_1,(\partial_s)^{-1}h]=[r_1,(\partial_s)^{-1}]h+ \ ] ] ( _ s)^-1[r_1,h]=s_4h = s_5 m , @xmath230 where @xmath264 $ ] for @xmath265 . by corollary [ cor1 ] , this implies @xmath263=[r_1,\partial_r - ( \partial_s)^{-1}\cdot u(r , s)]= \ ] ] -(r_1)_r + [ r_1,(_s)^-1u(r , s ) ] f(_s ) , @xmath230 and again as above we see that @xmath266 .
the proof of the equality @xmath267=0 $ ] is completely the same .
theorem is proved . in view of theorems [ th1 ] and [ th2 ]
we guess that the moutard transformation should preserve the class of weakly algebraic two - dimensional schrdinger operators .
in @xcite we gave a simple examples of fast decaying rational potentials for the two - dimensional schrdinger operator which has a degenerated @xmath2-kernel .
these examples are constructed by using the moutard transformation as follows . main construction .
let @xmath268 be an operator with a potential @xmath269 and let @xmath270 and @xmath271 satisfy the equations @xmath272 we take the moutard transformations @xmath273 and @xmath274 defined by @xmath275 and @xmath276 and obtain the operators @xmath277 where @xmath278 . by the construction , we have @xmath279 let us choose some function @xmath280 and put @xmath281 these functions define the moutard transformations of @xmath282 and @xmath283 and we obtain the operators @xmath284 and @xmath285 with the potentials @xmath286 the following key lemma is checked by straightforward computations which we omit .
we note that in this construction we have a free scalar parameter @xmath35 ( see ( [ theta - family ] ) ) for the choice of @xmath292 .
this parameter can be used in some cases to build a non - singular potential @xmath16 and functions @xmath293 and @xmath294 .
for example if we apply this construction to the situation when @xmath295 and @xmath275 and @xmath276 are real - valued harmonic polynomials @xmath296 then for some appropriate constant of integration in @xmath297 we obtain @xmath298 @xmath299 and @xmath300 ( here we simplify @xmath293 and @xmath294 by multiplying by some constant ) .
if one takes appropriate two harmonic polynomials of the third order @xmath275 and @xmath276 ( cf .
@xcite ) then one can construct the potential @xmath16 and the functions @xmath293 and @xmath294 which are smooth , rational , the potential decays like @xmath306 for @xmath302 , the functions @xmath293 and @xmath294 decay like @xmath307 for @xmath302 and span a two - dimensional space in the kernel of the operator @xmath308 . remark .
we guess that for every @xmath309 by applying this construction to other harmonic polynomials one can construct smooth rational potentials @xmath16 and the eigenfunctions @xmath293 and @xmath294 decaying faster than @xmath310 . in @xcite
we used a time - dependent extension of the moutard transformation constructing explicit solutions of the novikov
veselov equation @xcite @xmath311 some of our solutions show a very special behavior : the initial data for @xmath312 are smooth decaying rational functions of @xmath38 and @xmath207 ; nevertheless for @xmath313 the solutions to the nv equation ( [ nv ] ) blow up ( become singular ) .
in particular the following solution @xmath314 of the novikov
veselov equation can be obtained using this technology : @xmath315 with @xmath316 @xmath317 this solution decays as @xmath318 at infinity , it is nonsingular for @xmath319 and have singularities for @xmath320 . for @xmath321 , as we can see on figure 1 , the solution @xmath322 oscillates with growing amplitudes in the neighborhoods of the points @xmath323 and @xmath324 since the denominator @xmath283 vanishes at these points for @xmath325 .
the numerator @xmath282 has zeros or order 3 at the points @xmath179 and @xmath180 so their respective neighborhoods are subdivided by smooth lines into 6 sectors with different signs of the numerator
. the complicated behavior of the potential in the neighborhood of one of the singular points for @xmath326 is shown on figure 2 .
( 64.67,130.00 ) ( -20.67,17.00 ) ( 5.67,5.00)(0,0)[lc]figure 1 : the potential @xmath327 for @xmath328 .
( 120.67,17.00 ) ( 145.67,5.00)(0,0)[lc]figure 2 : the potential @xmath327 for @xmath329 ( 133.67,-5.00)(0,0)[lc ] in the neighborhood of the point @xmath330 . | in this paper we consider the moutard transformation @xcite which is a two - dimensional version of the well - known darboux transformation .
we give an algebraic interpretation of the moutard transformation as a conjugation in an appropriate ring and the corresponding version of the algebro - geometric formalism for two - dimensional schrdinger operators .
an application to some problems of the spectral theory of two - dimensional schrdinger operators and to the @xmath0-dimensional novikov
veselov equation is sketched . |
the buildup of stellar mass in dark matter halos has been measured in the local universe with a variety of methods , indirectly with abundance matching techniques @xcite and more directly with measurements of stellar and total masses of individual halos @xcite .
these stellar mass halo mass relations provide strong constraints on models of galaxy formation ( e.g. , * ? ? ?
while much of the attention has been focussed on the stellar content of @xmath9 halos , recent work has highlighted the relevance of the higher mass scales represented by groups and clusters , where various feedback processes are predicted to operate .
@xcite for instance show discrepancies between stellar - to - halo mass ratios computed for clusters from abundance matching and from direct observations .
meanwhile , at @xmath10 , @xcite measure a dramatic increase in the number density of galaxy groups over an epoch when star formation activity is greatly diminished , both in relative intensity ( i.e. , ssfr , * ? ? ?
* ) and in the number of galaxies that display active levels of star formation at a fixed mass
. given that these massive halos at high redshift represent sites where galaxies are quenched at levels above that in the field @xcite , an analysis of environmental trends provides insight into the mechanisms that contribute toward the growth of the red sequence . directly measuring the mass scales of these halos in which galaxies undergo quenching will provide stronger constraints on the relevant mechanisms than indirect methods for inferring halo membership and halo mass scales . the two relevant measurements for directly probing group mass scales are the halo mass and the total group stellar mass .
at high redshift , reliable detection of galaxy groups as well as halo mass estimates require deep x - ray imaging .
we therefore utilize the ultra deep chandra+xmm newton group catalog in the cdfs of @xcite . to find galaxies within these groups , we require a survey with sufficient redshift precision , which is uniform , and which has well - characterized completeness corrections to relatively low stellar masses .
the carnegie - spitzer - imacs ( csi ) survey @xcite meets these criteria and is used in this work .
we assume a cosmology with @xmath11 km s@xmath12 mpc@xmath12 , @xmath13 , and @xmath14 .
stellar masses are based on a @xcite imf .
we utilize combined chandra and xmm - newton x - ray observations in the cdfs @xcite to select groups at high redshift .
these observations represent the deepest x - ray data to date . for the analysis presented here
, we cull the list of @xmath15 x - ray groups from @xcite to those with redshifts in the range @xmath2 and flags of @xmath16 or @xmath17 , indicating well defined centroids and redshift determinations from followup spectroscopy .
in addition , we include only those groups with csi coverage in the ecdfs , out of @xmath18 sq degs in the swire cdfs field .
only @xmath19 groups remain in the catalog .
total masses , @xmath20 , given in @xcite are estimated from the @xmath21 relation in @xcite , which was calibrated with weak lensing based mass estimates in the cosmos field .
though the lowest halo masses were derived by extrapolating the latter relation , @xcite confirmed the reliability of the halo mass estimates from stacked weak lensing and clustering analyses . @xmath22
values follow from these mass estimates .
we utilize a redshift catalog in the cdfs constructed from new csi observations .
the handling of the data is similar to that in the xmm - lss field @xcite , though our deep decam photometry provides for more uniform completeness as a function of magnitude .
briefly , csi targets were selected in the _ spitzer _ irac 3.6 band with magnitudes brighter than @xmath23 ab mag and observed with the uniform dispersion prism ( udp ) on imacs @xcite at magellan .
redshifts were measured by fitting the available spectrophotometry ( spectrum @xmath24 ) with seds generated from stellar population synthesis models . in comparisons with high - resolution redshifts from @xcite we find a biweight scatter of @xmath25 for both red and blue galaxies . at @xmath26 ,
the stellar mass completeness limit is @xmath27 .
we note that group numbers @xmath28 and @xmath29 do not contain csi galaxies above the mass cut of @xmath30 ( some galaxies may have been missed due to slit collisions and other sources of incompleteness ) , though both contain galaxies below this limit .
these are both the lowest mass groups in the sample with halo mass estimates of @xmath31 @xmath32 .
also , number 63 had a large uncertainty on the stellar mass estimate ( see below ) .
we exclude these groups from our analysis leaving a total sample of 20 x - ray groups in the cdfs .
the spectroscopic completeness for csi in the cdfs is @xmath33 over @xmath34 ab mag and a broad range of color . to accurately account for galaxies that were not observed with the udp or which did not have well constrained redshifts ,
galaxies are assigned weights according to their inverse completeness in bins of observed magnitude ( @xmath35 $ ] ) , observed color ( @xmath36 $ ] ) , and local source density ( see , e.g. , * ? ? ?
* ) . even though the cdfs has been extensively studied by other surveys , with numerous high - resolution redshifts available in the literature ( e.g. , *
* ) , we refrain from combining the multiple available datasets .
the different selection functions in other work would make it difficult to produce a homogeneous dataset from the aggregate of all the available surveys , despite potentially boosting the level of completeness .
the depth and homogeneity of the csi data simplify the analysis , and the relatively uniform @xmath33 completeness is high enough to robustly characterize the stellar mass content of these x - ray groups .
lastly , while other works have used photometric catalogs to assemble or calibrate galaxy samples in groups in the cdfs ( e.g. , * ? ?
? * ; * ? ? ?
* ) , our sample is exclusively spectroscopic . for each x - ray group
, we assign csi galaxies above the mass limit within the projected virial radii , @xmath22 , and within @xmath37 .
this large redshift window encompasses both members and contaminants ( see below for correction ) .
we sum the stellar masses of the galaxies down to @xmath38 , multiplying by the completeness weights . on spatial scales comparable to those of the groups at @xmath2 ( e.g. , @xmath39 )
, the completeness correction accurately reproduces galaxy counts to within @xmath40 , an uncertainty that is factored into the total error budget on the aggregate stellar mass measurements . to account for galaxies below our stellar mass completeness limit
, we correct these group stellar masses assuming a schechter function with @xmath41 and @xmath42 . integrating down to @xmath43 @xmath32 we find a correction factor of @xmath44 .
changing @xmath45 and @xmath46 within reasonable ranges yields an uncertainty of @xmath47 in this correction , which is added to the error budget .
we derive a contamination correction for every group by subtracting off the stellar mass contribution from foreground and background galaxies .
we measured the stellar mass within circular apertures of radius @xmath22 at spatial locations that did not overlap with other groups near @xmath48 .
we compute the aggregate stellar mass of galaxies at @xmath49 in these apertures and take the mean and standard deviation to represent the background and its uncertainty ( typically @xmath50 ) . after measuring the stellar mass within @xmath22 , we correct the value to that expected within @xmath51 assuming a spherically symmetric nfw profile @xcite seen in projection ( e.g. , * ? ? ?
* see their equation 7 ) .
we assume an nfw profile with concentration of @xmath52 resulting in a multiplicative correction factor of @xmath53 .
the range of typical concentration values associated with high mass halos ( @xmath54 , * ? ? ?
* ) impacts the correction factor by only @xmath55 and therefore does not significantly affect our results or conclusions .
lastly , there is an additional source of uncertainty due to the statistical uncertainties in the individual mass measurements .
the typical statistical uncertainty , based on repeat udp observations and spectrophotometric sed fitting , is @xmath56 dex per galaxy .
this does not include contributions due to systematic effects such as different stellar population synthesis models and imf variations .
the total uncertainty in the aggregate stellar mass measurement for a typical group in our sample is @xmath57 dex .
figure [ fig_map ] shows two example low - mass groups in the cdfs .
color images were constructed using hst candels wfc3 @xmath58 and @xmath59 imaging @xcite . within the virial radii ( dashed green circles ) ,
figure [ fig_map ] shows clear overdensities of group members ( blue circles ) relative to field galaxies within the same redshift slice ( red and orange circles ) , indicating that we are able to reliably place galaxies into groups with our data and compute robust aggregate stellar mass totals .
we compare our results for low - mass groups at @xmath2 to low - mass groups at @xmath60 and to high - mass groups at @xmath2 , both derived from the wider field , but shallower x - ray data from cosmos ( for a detailed study , see * ? ? ?
we standardize the group stellar masses in cosmos with ours in the cdfs by measuring them in the same manner as described in the previous section and using the photometric redshifts and stellar masses from @xcite .
while this approach differs from summing redshift likelihood functions within the redshift windows of the groups , as carried out in @xcite , there is very good agreement between the two methods with a scatter in the aggregate stellar masses of only @xmath61 dex and a negligible offset .
we convert the cosmos stellar masses from chabrier to kroupa imf by adding @xmath62 dex . in total , the cosmos sample consists of @xmath63 groups , @xmath64 at @xmath2 and @xmath65 at @xmath60 with halo masses spanning @xmath66 and with x - ray flag=1 or flag=2 and no other flags .
figure [ fig_selection]@xmath67 shows the x - ray luminosity of x - ray groups as a function of redshift .
redshifts were determined from followup spectroscopy of red sequence candidate galaxies as described in @xcite and ( * ? ? ? * submitted ) .
the filled symbols represent x - ray groups with centroids determined from the x - ray image ( flag=1 ) , while the open symbols represent centroids determined from the positions of the red sequence galaxies ( flag=2 ) .
the median redshift for the cdfs sample is @xmath68 .
figure [ fig_selection]@xmath69 shows the cosmos and cdfs group masses , @xmath20 , as a function of redshift .
the error bars represent the statistical uncertainty and do not include the contribution due to the scatter in the @xmath21 relation ( @xmath57 dex ) . as a result of the lower flux limits
, cdfs reaches significantly lower group masses at high redshift compared to cosmos . in particular , halos at @xmath70 with mass below @xmath71 @xmath32 are almost exclusively detected in the cdfs .
the aggregate stellar mass within x - ray groups at @xmath2 is compared to their total mass in figure [ fig_stellar_halo_mass ] . in figure
[ fig_stellar_halo_mass]@xmath67 , the cdfs groups are shown using the black points .
the cosmos groups are shown in light gray . owing to its depth
, the cdfs sample provides the only leverage on @xmath72 for low mass groups ( @xmath73 @xmath32 ) .
the blue line is a biweight fit to the cdfs and cosmos data of the form @xmath74 where @xmath75 and @xmath76 ( see table [ table_fits ] ) .
the slope is consistent with the value of @xmath77 found in @xcite for their higher mass cosmos group sample spanning @xmath78 .
the implied stellar fraction for the cosmos groups ( @xmath79 at @xmath80 @xmath32 ) is also consistent with the results of @xcite .
the green data points in figure [ fig_stellar_halo_mass ] represent measurements for clusters from the gclass survey @xcite .
the green line is a fit to all three datasets and results in @xmath81 and @xmath82 .
the inclusion of the cluster data results in a shallow slope which is consistent with the general trend found at high halo masses .
the observed biweight scatter about the relation in equation [ eq_cc ] is @xmath83 dex , and is comparable for both the cdfs and cosmos samples . the similar scatter is reassuring and implies that our completeness corrections are robust and unbiased .
furthermore , this observational scatter is remarkably small .
when factoring in the uncertainties on the aggregate stellar mass measurements , the intrinsic scatter must be quite low .
determination of the intrinsic scatter requires accurate estimates of the observational errors .
because they total @xmath57 dex , we are unable to make a precise determination of the intrinsic scatter . in any case , an upper limit for the intrinsic scatter in the group stellar mass versus halo mass relation of @xmath84 dex reflects a relatively tight correlation suggesting that the former serves as a rough proxy for the latter .
this has also been noted by other authors ( e.g. , * ? ? ?
* ) . figure [ fig_stellar_halo_mass]@xmath69 shows the median stellar to halo mass for different bins in halo mass .
the bulk of the low mass cdfs group sample lies in the two lowest mass bins spanning @xmath6 . in these halos , which are 10 - 20 times more massive than the milky way s halo , the stellar mass constitutes @xmath85 of the total mass budget .
lcccc cdfs , cosmos & @xmath87 & @xmath88 & @xmath89 & 89 + cdfs , cosmos , gclass & @xmath90 & @xmath91 & @xmath89 & 93 here we examine the redshift evolution , if any , of the stellar mass halo mass relation for groups .
the orange curve in figure [ fig_stellar_halo_mass]@xmath69 indicates the stellar - to - halo mass ratio from abundance matching , empirically derived by ( * ? ? ?
* figure 12 ) at @xmath92 . given the error bars and any additional uncertainty due to systematic differences between the two measurement methods
, there does not appear to be any significant evolution at @xmath10 in the stellar - to - halo mass relation for @xmath93 @xmath32 .
we can further examine and confirm the lack of evolution by using additional measurements at lower redshifts ( @xmath94 ) that were made in the same manner as our own .
figure [ fig_evolution ] shows the redshift evolution for the median stellar - to - halo mass ratio for two halo mass bins .
uncertainties were computed by bootstrapping .
note that in the lowest mass bin , @xmath95 ( blue ) , the entire sample at @xmath70 is comprised of groups in the cdfs where the x - ray imaging is of sufficient depth to detect these systems and stellar masses have been measured with csi .
the low redshift ( @xmath96 ) sample is comprised exclusively of cosmos x - ray groups ( see section [ sec_cosmos ] ) .
the open squares indicate the results at @xmath92 from abundance matching @xcite and the associated error bars the range of median values spanned by halos with masses in the 25 - 75th percentile of the two sub - samples .
again , we do not find any significant evolution in the stellar - to - halo mass ratio in the group regime at @xmath10 .
we have measured the stellar mass halo mass relation for low mass groups at @xmath2 , complementing other work at these redshifts ( e.g. , * ? ? ?
* ; * ? ? ?
for the low mass groups in our cdfs sample , we find that about @xmath5 of the total mass is contained in stars , less than that of @xmath9 haloes but more than that of clusters , as is also the trend in the local universe .
the lack of ( or slow ) evolution in the stellar fraction of such halos is relevant given that the number density of these groups is observed to increase dramatically at @xmath97 @xcite indicating the growing significance of such environments .
late - time formation of massive halos is expected from hierarchical growth .
massive halos in the mass range @xmath98 @xmath32 are predicted to comprise @xmath99 of the total mass of all halos above @xmath100 @xmath32 ( approximately the halo mass scale that would host central galaxies at our stellar mass limit ) at @xmath86 ( based on hmfcalc , * ? ? ?
* ; * ? ? ?
by @xmath92 , this percentage rises substantially to @xmath101 . for groups in the mass range @xmath102 , the percentage increases from @xmath103 at @xmath86 to @xmath104 at @xmath92 .
clusters at @xmath105 @xmath32 clearly contribute toward the mass growth at @xmath97 but even these structures are built , in part , from groups @xcite .
the increase in number densities of galaxy groups , and the associated increase in the stellar mass density within such halos , has implications for the buildup of quiescent galaxies ( qgs ) since @xmath86 . from @xmath86
to @xmath92 the quiescent fraction ( qf ) increases from @xmath106 to @xmath107 for galaxies with @xmath30 @xcite . down to this mass limit ,
qfs are higher in rich environments compared to the field @xcite with qfs of @xmath108 and @xmath109 in groups / clusters and the field , respectively ( e.g. , * ? ? ?
as a consequence , the rise of groups over the past @xmath1 gyr contributes substantially to the evolving qf over that time . by assessing the star formation activity and quiescence of galaxies in groups over the full 15 degs@xmath110 survey field of csi , we will accurately constrain the impact of large scale structure growth on the buildup of qgs and the decline in global sfrd since @xmath0 . | since @xmath0 , the stellar mass density locked in low mass groups and clusters has grown by a factor of @xmath1 .
here we make the first statistical measurements of the stellar mass content of low mass x - ray groups at @xmath2 , enabling the calibration of stellar - to - halo mass scales for wide - field optical and infrared surveys .
groups are selected from combined @xmath3 and @xmath4 x - ray observations in the chandra deep field south ( cdfs ) .
these ultra - deep observations allow us to identify bona fide low mass groups at high redshift and enable measurements of their total halo masses .
we compute aggregate stellar masses for these halos using galaxies from the carnegie - spitzer - imacs ( csi ) spectroscopic redshift survey .
stars comprise @xmath5 of the total mass of group halos with masses @xmath6 ( about the mass of fornax and 1/50th the mass of virgo ) . complementing our sample with higher mass halos at these redshifts , we find that the stellar - to - halo mass ratio decreases toward higher halo masses , consistent with other work in the local and high redshift universe .
the observed scatter about the stellar - halo mass relation is @xmath7 dex , which is relatively small and suggests that total group stellar mass can serve as a rough proxy for halo mass . we find no evidence for any significant evolution in the stellar - halo mass relation since @xmath8 . quantifying the stellar content in groups
since this epoch is critical given that hierarchical assembly leads to such halos growing in number density and hosting increasing shares of quiescent galaxies . |
a fundamental problem in cosmology is to understand how the spatial distribution of galaxies ( and of galaxy clusters ) is related to that of the underlying mass . in standard models of galaxy formation
, it is assumed that galaxies form by the cooling and condensation of gas within dark matter halos ( white & rees 1978 ; white & frenk 1991 ) .
nongravitational processes which can not be modeled reliably are likely to be critical in determining the properties of individual galaxies , yet they should have little effect on the formation and clustering of dark halos . as a result , the problem of galaxy biasing can be approached by first understanding how dark halos are distributed relative to the mass .
dark halos are highly nonlinear objects and their formation and clustering has usually been studied using n - body simulations ( e.g. frenk 1991 ; gelb & bertschinger 1994 and references therein ) .
such simulations are limited both in resolution and in dynamical range and can be difficult to interpret .
our understanding of their results could be substantially enhanced by simple physical models and the analytic approximations they provide .
mo & white ( 1996 ) have developed a model for the second order correlation functions of dark halos based on the press - schechter formalism ( press & schechter 1974 ; hereafter ps ) .
here we extend their work to derive a model for the higher order correlations .
we also work out a similar model for peaks in the initial density field . in the ps formalism
, one defines dark matter halos at any given time as regions of the initial density field which just collapse at that time according to a spherical infall model .
this formalism can be extended so that it predicts not only the mass function of dark halos , but also a wide range of other statistical properties of the hierarchical clustering process ; comparisons with n - body simulation data show detailed agreement ( e.g. bond et al . 1991 ; bower 1991 ; kauffmann & white 1993 ; and particularly , lacey & cole , 1993 , 1994 ) .
mo & white ( 1996 ) showed that the ps formalism and its extensions can be used to construct a model for the spatial correlation of dark halos in hierarchical models .
they used the standard ps formalism both to define dark halos from the initial density field , and to specify how their mean abundance within a large spherical region is modulated by the linear mass overdensity in that region .
the gravitationally induced evolution of clustering was treated by assuming that each region evolves as if spherically symmetric .
the model was found to give an accurate description of the bias function , @xmath13 , where @xmath14 is the mean overdensity of halos within spheres which have radius @xmath15 and _ mass _ overdensity @xmath16 .
a simple extension provides a surprisingly accurate prediction for the halo - halo two - point correlation measured in n - body simulations ( mo & white 1996 ; mo , jing & white 1996 ) .
it is clearly interesting to see if such a model can also be constructed for the higher order correlations of dark halos . in the peak formalism
, one assumes that galaxies form at those peaks of a suitably smoothed version of the initial density field which rise above some density threshold .
kaiser ( 1984 ) introduced this idea to show how the strong clustering of abell clusters could result from the statistics of high peaks in a gaussian initial field .
this formalism was later developed extensively by bardeen et al .
( 1986 , hereafter bbks ) .
these authors showed that if galaxies can be associated with high peaks of the initial density field then they should be more strongly clustered than the mass , an effect usually called `` galaxy biasing '' .
unfortunately it is not known how _
well _ galaxies correspond to high peaks of the initial field , and it is unclear how to deal with the problem that the present - day clustering of peaks differs substantially from that in the initial ( lagrangian ) space because of gravitationally induced motions . as a result ,
the predictions of the theory have only been checked quantitatively by direct n - body simulation of the motion of the material associated with density peaks ( e.g. davis et al .
1985 ; frenk et al .
1988 ; katz , quinn & gelb 1993 ) . in this paper
we will show that the idea underlying the model of mo & white can also be used to construct a model for the correlation functions of peaks in _ physical _ space .
we use the theory of bardeen et al . to define peaks from the initial density field , and to specify how their mean abundance within a large spherical region is modulated by the linear mass overdensity in that region .
the gravitationally induced evolution of clustering is then treated , as in the model of mo & white , by assuming that each region evolves according to a spherical collapse model .
we describe our model in 2 , and present detailed tests of its predictions against n - body simulations in 3 .
then in 4 we demonstrate how it can be used to interpret the observed high order correlation functions of galaxies and of clusters of galaxies , and to determine whether or not galaxies are biased relative to mass .
finally , in 5 we summarize our main results .
to calculate the high order moments of peaks and halos ( together called `` galaxies '' ) , we use the general formalism developed by fry & gaztanaga ( 1993 ) .
consider the present - day mass overdensity field @xmath17/\barrho$ ] , where @xmath18 is the local mass density and @xmath19 the mean density .
when smoothed in spherical window @xmath20 with characteristic radius @xmath15 , @xmath21 gives rise to a smoothed density field : @xmath22 for a top - hat window , @xmath23 is just the volume average of @xmath24 within a sphere of radius @xmath15 .
the statistical properties of @xmath25 are described by the one - point distribution function of the density field .
now let us denote the overabundance of `` galaxies '' in the same window by @xmath26 .
if @xmath27 is completely determined by @xmath28 , then we can write @xmath29 as a function of @xmath16 , @xmath30 , which should not depend on @xmath31 .
( note that this can hold at best approximately ; see below . ) to simplify notation , we will omit writing explicitly the smoothing radius @xmath15 associated with @xmath16 and @xmath29 . in general , we can expand @xmath32 in taylor series : @xmath33 where @xmath34 are constant . fry & gaztanaga ( 1993 )
have shown that if the volume averaged j - point mass correlation functions @xmath35 have the hierarchical form : @xmath36 then the transformation given by equation ( 2 ) preserves the hierarchical structure in the limit @xmath37 . in this case
one can write @xmath38 and for @xmath1 , 4 and 5 , which are relevant for our later discussion , one has @xmath39
@xmath40 @xmath41 where @xmath42 and @xmath43 .
thus to obtain the high order moments @xmath44 for halos and peaks in the quasilinear regime , we need to work out the coefficients @xmath34 in the bias relation ( equation 2 ) .
we will do this in 2.2 and 2.3 .
it is important to emphasize that the bias relation given by equation ( 2 ) is at best approximate .
this is because @xmath29 in a spherical region with mean mass overdensity @xmath16 must depend not only on @xmath16 but also on the internal structure of the region and on the external tides acting on it . as a result
, there must be scatter in the `` galaxy '' counts among spheres with the same @xmath16 and @xmath15 .
such scatter should also contribute to the high order moments . as discussed in mo & white ( 1996 ) , this contribution must be important on scales which are not much larger than the linear _ lagrangian _
sizes of the `` galaxies '' . on larger scales it may or may not be important , and since it can not be estimated within our formalism , in the present paper we will assume it to be negligible and test the resulting predictions against n - body data . the initial overdensity field @xmath45/{\bar { \rho}}$ ] is assumed to be gaussian and so to be described by a power spectrum @xmath46 , where @xmath47 is the fourier transform of @xmath21 and @xmath48 is the dirac delta function .
we smooth the field @xmath21 by convolving it with a spherically symmetric window function @xmath20 having _ comoving _ characteristic radius @xmath15 ( measured in current units ) .
the smoothed field can be written as @xmath49 where @xmath50 is the fourier transform of the window function @xmath20 .
following bbks , we define the order @xmath51 moment of the smoothed field by @xmath52 the order zero moment , which we denote by @xmath53 , is just the _ rms _ fluctuation of mass in the smoothing window . for a given window function
the smoothed field @xmath23 is gaussian and so has the following one - point distribution function @xmath54 since both @xmath16 and @xmath55 grow with time in the same manner in linear perturbation theory , it is convenient to use their values linearly extrapolated to the present time .
these extrapolated quantities will still obey equation ( 8) . in the following , we write our formulae in terms of these extrapolated quantities .
we will also omit writing explicitly the smoothing radius @xmath15 , but often use subscripts to distinguish @xmath56 , and other quantities , at different smoothing lengths [ e.g. @xmath57 , @xmath58 . as another convention
, we will always label the properties of `` galaxies '' using a subscript `` 1 '' .
the subscript `` 0 '' is reserved for the properties of larger uncollapsed spherical regions .
mo & white ( 1996 ) have described in considerable detail how to derive the bias relation , @xmath59 , for dark halos ( so subscript `` h '' denotes `` halos '' ) in the ps formalism .
here we adopt their notation and work out the coefficients @xmath34 in equation ( 2 ) that are needed in the calculations of @xmath60 .
we assume that dark halos are spherically symmetric , virialized clumps of dark matter .
the mass @xmath61 of a halo is then related to the comoving lagrangian radius @xmath62 of the region from which it formed by @xmath63 where @xmath19 is the mean density of the universe . in an einstein - de sitter universe ( which we assume throughout this paper ) , a spherical perturbation of linear overdensity @xmath64 collapses at redshift @xmath65 , where the critical overdensity for collapse @xmath66 . according to the ps formalism
, the comoving number density of halos , expressed in current units , as a function of @xmath61 and @xmath67 is : @xmath68 notice that in this formalism a class of halos must be defined by specifying _ both _ their mass @xmath61 ( or equivalently @xmath62 or @xmath69 ) and their redshift of identification @xmath67 ( or equivalently @xmath64 ) . to derive the bias relation ,
we need formulae which relate halo abundances to the density field on larger scales .
bower ( 1991 ) and bond et al .
( 1991 ) extend the original ps formalism to show that the fraction of the mass in a region of lagrangian radius @xmath70 and linear overdensity @xmath71 which at redshift @xmath67 is contained in dark halos of mass @xmath61 ( where by definition @xmath72 ) is given by @xmath73 thus the average number of @xmath61 halos identified at redshift @xmath67 in a spherical region with comoving radius @xmath70 and overdensity @xmath71 is @xmath74 following mo & white ( 1996 ) , we obtain the _ physical _ space overabundance of halos in spheres which at the desired redshift @xmath75 have radius @xmath15 and overdensity @xmath16 , using a spherical model .
in such a model , each spherical shell moves as a unit and different shells do not cross until they collapse through the zero radius .
thus the mass interior to each mass shell is a constant , which gives @xmath76 furthermore , since dark halos in our ps model are defined to be objects identified at some specific redshift , the mean abundance of equation ( 12 ) can be taken as referring to halos of mass @xmath61 identified at redshift @xmath67 within spheres of radius @xmath77 and overdensity @xmath78 . under these assumptions the average overdensity of dark halos in spheres with current radius @xmath15 and current mass overdensity @xmath79 can be obtained from equations ( 10 ) and ( 12 ) : @xmath80 where @xmath81 , @xmath82 , and @xmath83 is determined from @xmath16 by the spherical collapse model , as described in the appendix . when considered as a function of @xmath79 , @xmath84 in equation ( 14 )
just gives the bias relation .
we assume that @xmath85 so that @xmath86 in equation ( 11 ) can be replaced by @xmath87 .
assuming also @xmath88 and using equation ( a4 ) in the appendix we can expand @xmath84 in the form of equation ( 2 ) .
it turns out that the first five coefficients ( which are relevant in our discussion ) are @xmath89 @xmath90 @xmath91 @xmath92 @xmath93\left({\nu_1\over \delta_1}\right)^2(\nu_1 ^ 2 - 3)\ ] ] @xmath94 where @xmath95 ; @xmath96 , @xmath97 and @xmath98 are the coefficients in the expansion of @xmath99 ( see equation a4 ) . inserting @xmath34 into equation ( 5 ) we can obtain @xmath100 ( skewness ) , @xmath101 ( kurtosis ) and @xmath102 for dark halos . as we can see from equation ( 15 ) , for a given @xmath67 ,
the high order moments depend on the mass @xmath61 ( or @xmath69 ) of the halos .
when halos ( identified at a given redshift @xmath67 ) with a range of masses are considered , @xmath34 in equation ( 5 ) should be replaced by the values obtained by averaging them over @xmath61 with a weighting of @xmath103 .
the high order moments depend also on the dynamical evolution of the underlying mass density field , as manifested by the dependence on @xmath96 , @xmath97 and @xmath98
. however , as shown by bernardeau ( 1992 ) , the @xmath71-@xmath16 relation ( which determines @xmath104 ) given by the spherical collapse model , and the high order moments ( @xmath105 ) of the mass distribution , depend only weakly on cosmological model in the quasilinear regime . therefore the results for @xmath8 given by equations ( 5 ) and ( 15 ) should not depend significantly on a particular choice of cosmological parameters . before moving on to the next subsection ,
let us consider some asymptotic properties of the @xmath34 in equation ( 15 ) and of the high order moments @xmath106 derived from them .
when @xmath107 and @xmath108 is not large , i.e. for small halos identified at early times , equation ( 15 ) gives @xmath109 and @xmath110 for @xmath111 .
it then follows from equation ( 5 ) that @xmath112 , meaning that such halos are not biased relative to mass .
in contrast , when @xmath113 and @xmath64 is not large , i.e. for big halos identified at low redshift , we have @xmath114 for @xmath111 , and @xmath115 are determined completely by the statistical properties of the _ initial _ density field , independent of both @xmath105 and @xmath104 .
the numerical values of the first several moments are : @xmath116 if @xmath117 and @xmath64 is not large , i.e. for small halos identified at low redshift , then @xmath118 and @xmath119 for @xmath120 . in this case @xmath115 may depend significantly on the dynamical evolution of the underlying mass density field .
the skewness of such halos will be @xmath121s_3 -6\delta_1(1+a_2)/(\delta_1 - 1)^2 $ ] , which can be larger than @xmath122 . for halos with @xmath123 ,
the skewness is @xmath124 , which is substantially smaller than @xmath122 unless @xmath64 ( and so @xmath67 ) is high .
the argument in 2.2 can also be used to construct a model for the high order moments of density peaks . in the peak theory
, we consider peaks in the initial density field after smoothing with a ( spherical ) window function with a given radius @xmath62 ( corresponding to a _
rms _ mass fluctuation of @xmath69 ) , and examine the distribution of peaks with respect to the peak height @xmath95 .
according to bbks , the comoving differential peak density is @xmath125 where @xmath126 with @xmath127 , @xmath128 , @xmath129 defined in equation ( 7 ) , and @xmath130\over [ 2\pi(1-\gamma^2)]^{1/2 } } , \eqno(19)\ ] ] with @xmath131 + \erf\left[\left({5\over 2}\right)^{1/2}{x\over 2}\right]\right\rbrace\ ] ] @xmath132 \eqno(20)\ ] ] ( see equation a19 in bbks ) . to derive the bias relation for peaks , we also need formulae which relate peak number density to the mass density field on larger scales .
let us consider a spherical top - hat region with lagrangian radius @xmath70 and linear mass overdensity @xmath71 .
the number density of peaks ( with characteristic radius @xmath133 ) in such a region is modulated by the background field , and assuming @xmath134 , it can be written as @xmath135 where @xmath136 with @xmath137 and @xmath138 .
the last relation for @xmath139 follows from the argument of bower ( 1992 ) that @xmath140 when @xmath85 . under the same assumptions made for equation ( 14 ) ,
the average overabundance of peaks @xmath141 ( subscript `` p '' for `` peaks '' ) in spheres with current radius @xmath15 and current mass overdensity @xmath16 can be obtained from equations ( 17 ) and ( 21 ) : @xmath142 where @xmath143 , and @xmath83 is determined from @xmath16 by the spherical collapse model .
as we have done for dark halos , we now assume @xmath85 ( so that @xmath144 ) and @xmath88 , and expand @xmath145 in the form of equation ( 2 ) .
it follows that the first five coefficients are @xmath146 @xmath147 @xmath148 @xmath149 @xmath150 , \eqno(24d)\ ] ] @xmath151\left({\nu_1\over \delta_1}\right)^2 \left(\nu_1 ^ 2 - 1 + 2g_1+{2g_2\over \nu_1 ^ 2}\right)\ ] ] @xmath152\ ] ] @xmath153 , \eqno(24e)\ ] ] where @xmath96 , @xmath97 , @xmath98 are , as before , the coefficients in the expansion of @xmath99 ( see equation a4 ) , and @xmath154 since @xmath155 and its derivatives involve only single integrations ( see equation 19 ) , it is straightforward to calculate @xmath34 in equation ( 24 ) numerically . as one can see from equation
( 24 ) , for a given peak scale @xmath62 the high order moments of peaks given by equation ( 5 ) depend on the peak height @xmath108 ( or @xmath64 ) .
when peaks with a range of heights are considered , @xmath34 in equation ( 5 ) should be replaced by the values obtained by averaging them over @xmath108 with a weighting of @xmath156 .
it is interesting to note that @xmath34 in equation ( 24 ) would have the same forms as those in equation ( 15 ) , if @xmath157 and @xmath158 for @xmath111 .
since @xmath159 have finite values for any realistic power spectra , it is clear from equation ( 24 ) that in general the high order moments of peaks have different asymptotic values from those of halos discussed in 2.2 .
we now test our analytic theory by comparison with the results from a series of large cosmological n - body simulations of einstein - de sitter universes .
we use results for four different spectra . the first two have cdm - like forms , where the transfer functions are given by equation ( g3 ) in bbks , with the shape parameter @xmath160 equal to 0.5 and 0.2 , respectively .
we normalize the initial power spectra by specifying @xmath161 , the linear _
rms _ mass fluctuation in a spherical top - hat window of radius @xmath162 .
these simulations were performed using a particle - particle / particle - mesh ( @xmath163 ) code with @xmath164 particles and a force resolution of about @xmath165 .
the simulation box size is 256 @xmath166 for @xmath167 and @xmath168 for @xmath169 .
the two - point correlation functions and mass functions of dark halos in these simulations have been analysed by mo , jing & white ( 1996 ) .
the other two are power - law spectra with @xmath170 and @xmath171 .
these simulations were performed using the @xmath163 code described by efstathiou et al .
( 1988 ) and are very similar to the simulations of that paper . however , they are substantially larger ( @xmath172 ) and have higher resolution ( gravitational softening length equal to @xmath173 , where @xmath174 is the side of the fundamental cube of the periodic simulation region ) .
the initial power spectrum was normalized as described by efstathiou et al .
( 1988 ) and `` time '' is measured by expansion factor @xmath175 since the start of the simulation ( @xmath176 for the initial conditions ) .
jain , mo & white ( 1996 ) have used these ( and some other ) simulations to test the similarity scalings in the mass correlation functions and power spectra .
these simulations were also used by mo & white ( 1996 ) to study the two - point statistics of dark halos .
the peaks considered in this paper are defined as those above a certain threshold @xmath177 in the primordial density field smoothed with a gaussian window @xmath178 .
the window width is taken to be @xmath179 so that it is relevant for galactic - sized objects .
we follow the prescription of white et al .
( 1987 ) to select peaks in the numerical simulations ( see jing et al .
1994 for a detailed description of our algorithm ) .
the algorithm gives an expectation number of peaks for each simulation particle .
since this number is always less than 1 ( i.e. each particle carries less than one peak ) , we select peaks by randomly culling simulation particles with a selection probability for each particle equal to its expectation number .
the correlation functions are calculated by peak counts in spheres regularly placed on a @xmath180 grid . since the simulations are periodic , there are no difficulties with spheres overlapping the boundary of the simulated region .
the circles in figure 1 show our simulation results for the skewness of peak count @xmath181 as a function of the radius @xmath15 of the spherical counting cell .
results are shown for peaks with heights above the values indicated in the panels .
the solid curves show the predictions of equation ( 5a ) with @xmath182 and @xmath183 given by equations ( 24b ) and ( 24c ) , respectively . in our model calculations we have used the values of @xmath184 ( the skewness of the mass distribution ) estimated directly from the simulations . in practice @xmath184 in the quasilinear regime can also be obtained from the initial density spectrum ( fry 1984 , juszkiewicz et al .
1993 ; lucchin et al .
1994 ; bernardeau 1994 ; baugh et al .
1995 ) . for comparison
, we plot the skewness of the mass distribution in the simulations as the dashed curves .
it is clear from fig.1 that the theoretical model works well over a wide range of scales .
the thick ticks on the horizontal axis mark the value of @xmath15 where the second moment of the mass distribution @xmath185 .
our model can work well even for @xmath186 ; this is particularly the case when the peak - height threshold is not very high .
since very high peaks are preferentially located in high density regions where nonlinear evolution of the density field can be strong , our model fails at small @xmath15 for such peaks . in figure 2
we compare the kurtosis @xmath187 of the density peaks in the simulations ( circles ) with our model predictions ( solid curves ) . here again the skewness @xmath184 and kurtosis @xmath188 of the mass distribution used in the model ( equation 5b ) are estimated directly from the simulations . for comparison , we show @xmath188 in fig.2 as the dashed curves .
figure 2 shows that our model also works reasonably well for @xmath189 over a wide range of scales .
to define dark halos in the simulations , we will use the standard ` friends - of - friends ' ( fof ) group finder with a linkage length equal to 20% of the mean interparticle distance ( e.g. davis et al .
this algorithm is easy to implement and has been extensively tested against the ps mass function ( see lacey & cole 1994 for a careful discussion ) .
the two - point correlation functions of the halos found by this algorithm have been tested against analytical theories of the kind we analyse here by mo & white ( 1996 ) and mo et al .
the statistics in this section are based on counts of halos or of individual particles within spheres . when evaluating such statistics for the simulations we count objects within spheres centered on each grid point of a regular @xmath190 cubic mesh .
figure 3 shows the comparison of our model prediction for the skewness of halos @xmath100 ( solid curves ) against results obtained from simulations ( circles ) .
results are shown for halos in different mass ranges . in figs .
3a-3c , halos are selected at an earlier epoch ( when the expansion factor is @xmath191 ) than the one when the skewness is calculated ( at @xmath192 ) . in general , halos identified at @xmath193 will , by @xmath96 , have increased their mass by accretion or lost their identity by merging .
however , galaxies which were forming at their centers at @xmath193 may still remain distinct at @xmath96 . thus the results shown in figs .
3a-3c may be relevant to galaxies . in the simulations
the position of each halo at the later epoch is assumed to be that of the particle which was closest to its center at @xmath193 .
the model predictions are obtained from equations ( 5 ) and ( 15 ) with @xmath64 taking the value @xmath194 , and with the skewness of mass distribution @xmath122 estimated directly from the simulations . in this case , the agreement between the analytic model and the simulation results is reasonably good for large @xmath15 where the second moment of mass distribution @xmath195 .
( the values of @xmath15 where @xmath196 are marked by the thick ticks on the horizontal axis . )
the skewness of halos shown in the figure does not change significantly with halo mass , because the dynamical range covered by the simulation is still too small to allow us to select halos with masses small enough to see such a dependence . for comparison , the crosses in fig.3a
show the simulation result for halos which contain only about 7 particles .
we see that the skewness of such small halos can indeed be as high as that for the mass , as predicted by our model .
unfortunately such small halos are very poorly sampled by our current simulation . for small @xmath15
, two effects may cause our model to fail .
both the nonlinear evolution of the mass density field , and the fact that halos are spatially exclusive at the time of their identification in the simulations , may change halo clustering properties on small scales .
halo exclusion effects reduce the variance in the halo count to significantly below the poisson value and also affect higher moments of the counts ( see mo & white 1996 for more detailed discussion ) .
for the same reason , our model is less successful for halos which are identified at the epoch when the skewness is calculated ( fig.3d ) ; halo exclusion effects are clearly more important in this case . in figure 4
we compare the kurtosis @xmath197 of dark halos in the simulations ( circles ) with our model predictions ( solid curves ) .
results are shown for the same models as in fig.3 .
as before the skewness @xmath184 and kurtosis @xmath188 of the mass distribution , which are required by the model , are estimated directly from the simulations .
the values of @xmath188 are shown in fig.4 as the dashed curves .
this figure shows that our model predictions for @xmath101 agree reasonably well with simulation results on large scales when halos are selected earlier than the epoch when @xmath101 is analysed . for the reasons discussed above ,
our model is less successful on small scales and for halos identified at the epoch when the kurtosis is analysed .
the last section shows that our model for the skewness and kurtosis of density peaks and dark halos ( `` galaxies '' ) works reasonably well . in this section
we demonstrate how this model can be used to interpret the observed high order correlation functions of real galaxies and clusters of galaxies .
the assumption we make is , of course , that these objects are associated with initial density peaks or with the dark halos present at some given redshift . in figure 5
we show predictions for the high order moments @xmath8 ( @xmath1 , 4 and 5 ) of `` galaxies '' as a function of the linear bias parameter @xmath198 [ defined by equations ( 15b ) and ( 24b ) for halos and peaks , respectively ] .
each curve corresponds to a particular choice of @xmath64 , as parameterized by the value of @xmath65 given in the figure caption .
as an example , we show results for a cdm - like spectrum with @xmath167 and @xmath199 and for a radius @xmath200 .
the spectrum chosen here is consistent with that given by the angular correlation functions of galaxies in the apm survey ( efstathiou , sutherland & maddox 1990 ; see also maddox et al .
1996 for a recent discussion ) .
the main features in @xmath8 do not change significantly if we change the value of @xmath201 from 0.2 to 0.5 .
the choice of @xmath15 is based on the fact that the mass density field in the universe is still in the quasilinear regime on this scales and that high order moments of galaxies are difficult to measure on much larger scales . for halos with fixed @xmath67 and @xmath10 ,
@xmath202 depend only on @xmath203 which , in turn , depend only on the effective power index at @xmath15 : @xmath204 . for peaks , however , @xmath205 depend , in addition , also on the shape of the spectrum on the peak scales @xmath62 [ through the dependence on @xmath206 defined in equation 18 ] . for a given shape of power spectrum , the dependence of @xmath207 on @xmath161 is weak ,
once @xmath10 and @xmath64 are fixed .
the values of @xmath8 in the quasilinear regime depend only weakly on cosmology , as pointed out in 2.2 . from fig.5 we see that the @xmath8 for `` galaxies '' are systematically smaller than those for the mass , unless the linear bias parameter @xmath10 is comparable to or less than unity .
the values of @xmath8 are more or less constant for large @xmath10 , but decrease rapidly with increasing @xmath10 for @xmath208 . for a given @xmath10 ,
@xmath8 are higher for dark halos that are identified at an earlier epoch and for density peaks with higher @xmath64 .
the values of @xmath8 are the lowest for `` galaxies '' with @xmath209 and @xmath12 .
these results have interesting implications for the observed high order moments of galaxies and clusters of galaxies .
let us start with clusters of galaxies .
clusters of galaxies are found to have much larger two - point correlation amplitude than galaxies ( see e.g. bahcall & west 1992 ; peacock & west 1992 ; dalton et al .
1992 ; nichol et al .
if the two - point correlation function of galaxies is neither significantly _ smaller _ than that of the mass nor larger than that of the mass by a factor exceeding three , the value of the linear bias factor for clusters should lie somewhere between 2 and 5 .
if we take clusters to be virialized halos identified at present time , figure 5 shows that the skewness ( @xmath210 ) and kurtosis ( @xmath211 ) of clusters should be about 2 and 7 , respectively .
this is consistent with current observations .
based on the three - point correlation functions of abell clusters , jing & zhang ( 1989 ) found @xmath212 which corresponds to @xmath213 ( see also plionis & valdarnini 1994 and cappi & maurogordato 1995 who did count - in - cell analysis for abell clusters and confirmed that @xmath213 ) .
recently , gaztanaga , croft & dalton ( 1995 ) obtained @xmath213 and @xmath214 for clusters in the apm survey .
these values of skewness and kurtosis appear to be much smaller than the corresponding values for the mass distribution obtained from a power spectrum which has the shape expected given the angular correlation functions of apm galaxies ( see e.g. gaztanaga et al .
1995 ) . from our model
we see that such low values are a result of clusters being high mass halos identified at low redshift . for optical galaxies ,
the current best estimates of the high order moments are those of gaztanaga ( 1994 ) based on apm survey .
the values he got are @xmath215 , @xmath216 and @xmath217 for @xmath9 .
these results are plotted in fig.5 as circles with errorbars . as noticed by gaztanaga & frieman ( 1994 )
, the observed values for apm galaxies appear to be close to those for the mass ( the values predicted by quasilinear theory are indicated by the horizontal solid lines in fig.5 ) . comparing our predictions with the observational results we infer that galaxies in the apm survey should not be strongly biased relative to the mass .
namely , the linear bias parameter for apm galaxies should not be significantly larger than unity .
mild antibias ( i.e. @xmath208 ) is possible if most apm galaxies are associated with galactic - sized halos which form rather late .
we note , however , that gaztanaga ( 1992 ) found significantly smaller values @xmath218 and @xmath219 from the fully 3-dimensional cfa and ssrs surveys .
gaztanaga ( 1994 ) argues that these small values reflect the fact that the volumes of the local surveys are too small to be fair .
redshift distortion present in these surveys may also complicate the determination of the high - order correlations in real space .
nevertheless , if these lower values turn out to be correct ( for example if there is some problem in deriving @xmath220 and @xmath221 from the 2-dimensional apm data ) then our analysis would suggest that the apm galaxies could be substantially biased relative to the mass .
future redshift samples from large digital sky surveys will certainly help to resolve the problem .
the skewness and kurtosis of spiral galaxies and iras galaxies appear to be much lower than those for apm galaxies , with @xmath222 and @xmath223 ( e.g. jing et al . 1991 ; meiksin et al .
1992 ; bouchet et al .
since these galaxies have weaker two - point correlations than apm galaxies ( and therefore lower @xmath10 values ) , the observational results seem to require these galaxies to be associated with halos identified at late time , or with peaks at low overdensity @xmath64 .
in this paper , we have developed an analytic model for the high - order moments of the distribution of density peaks and dark halos in the quasi - linear regime .
such a model allows the high - order correlation functions of density peaks and dark halos to be calculated analytically for any given initial ( gaussian ) density spectrum .
tests against results from a variety of n - body simulations have shown that our model works successfully for density peaks and for halos identified at an earlier epoch than the time when the moments are calculated .
our model is only qualitatively correct for halos identified at the same time as the moments are calculated , because halos are spatially exclusive at the time of their identification .
we have found that the skewness ( @xmath220 ) , kurtosis ( @xmath221 ) and @xmath224 for both halos and peaks decrease rapidly with the linear bias parameter @xmath10 ( of these objects ) for @xmath12 .
thus if galaxies are associated with peaks in the initial density field , or with dark halos formed at different redshifts , a measurement of @xmath8 ( @xmath1 , 4 , 5 ) of the galaxy distribution in the quasilinear regime should allow us to determine whether or not galaxies are significantly biased relative to the mass .
we have used our model to interpret the observed high order correlation functions of galaxies and clusters .
we have found that if the values of @xmath8 for galaxies are indeed as high as those given by the apm survey , then apm galaxies should not be significantly biased .
there is , however , a significant uncertainty in the comparison between our model predictions and observed galaxy distribution . at
any given time massive dark halos may contain more than one galaxy and galactic - sized peaks in the initial density field may merge with each other to form a single galaxy .
thus the observed galaxies may not correspond uniquely to the centers of the halos present at any single epoch or to galactic - sized peaks in the initial density field . as a result ,
it is not straightforward to apply our results directly to galaxies . however , if more detailed modeling allows a prediction of how galaxies form in dark halos ( e.g. white & frenk 1991 ; kauffmann , nusser & steinmetz 1996 ) , our results can readily be extended to study the high - order moments of galaxy distribution .
such a study will also help us to assess the importance of nongravitational effects in the measurements we are suggesting here .
ypj acknowledges the receipt of an alexander - von - humboldt research fellowship .
hjm is supported by the ambrose monell foundation in the institute for advanced study .
bahcall n. a. , west m. , 1992 , apj , 392 , 419 bardeen j. , bond j.r . , kaiser n. , szalay a.s . , 1986 ,
apj , 304 , 15 ( bbks ) baugh c.m . , gaztanaga e. , efstathiou g. , 1995 , mnras , 274 , 1049 bernardeau f. , 1992 , apj , 392 , 1 bernardeau f. , 1994 , a&a , 291 , 697 bond j.r .
, cole s. , efstathiou g. , kaiser n. , 1991 , apj , 379 , 440 bower r.j . , 1991 , mnras , 248 , 332 bouchet f.r . , strauss m. , davis m. , fisher k. , yahil a. , huchra j. , 1993 , apj , 417 , 36 cappi a. , maurogordato s. , 1995 , apj , 438 , 507 dalton g.b . , croft r.a.c .
, efstathiou , g.,sutherland w.j . , maddox s.j . , davis m. , 1994 , mnras , 271 , l47 davis m. , efstathiou g. , frenk c. , white s.d.m . , 1985 , apj , 292 , 371 efstathiou g. , frenk c.s .
, white s.d.m . , davis m. , 1988 , mnras , 235 , 715 efstathiou g. , sutherland w.j . , maddox s.j . , 1990 , nat , 348 , 705 frenk c.s . , 1991 , physica scripta , t36 , 70 frenk c.s . , white s.d.m .
, davis m. , efstathiou g. , 1988 , apj , 327 , 507 fry j.n . , 1984 , apj , 279 , 499 fry j.n . , gaztanaga e. , 1993 , apj , 413 , 447 gaztanaga e. , 1994 , mnras , 268 , 913 gaztanaga e. , croft r.a.c . , dalton g.b .
, 1995 , mnras , 276 , 336 gaztanaga e. , frieman j.a . , 1994 , apj , 437 , l13 gelb j.m . , bertschinger e. , 1994 , apj , 436 , 491 lucchin f. , matarrese s. , melott a.l . , moscardini l. , 1994 , apj , 422 , 430 jain b. , mo h.j .
, white s.d.m . , 1995 , mnras , 276 , l25 jing y.p . , mo h.j .
, brner g. , 1991 , a&a , 252 , 449 jing y.p . , mo h.j .
, brner g. , fang l.z . , 1994 , a&a , 284 , 703 jing y.p .
, zhang j.l . , 1989 , apj , 342 , 639 juszkiewicz r. , bouchet f. , colombi s. , 1993 , apj , 412 , l9 kaiser n. , 1984 , apj , 284 , l9 katz n. , quinn t. , gelb j.m . , 1993 , mnras , 265 ,
689 kauffmann g. , nusser a. , steinmetz m. , mnras , submitted kauffmann g. , white s.d.m .
, 1993 , mnras , 261 , 921 lacey c. , cole s. , 1993 , mnras , 262 , 627 lacey c. , cole s. , 1994 , mnras , 271 , 676 maddox s. , et al . , 1996 , preprint meiksin a. , szapudi i. , szalay a.s . , 1992 ,
apj , 394 , 87 mo h.j . , jing y.p . , white s.d.m . ,
1996 , mnras , submitted mo h.j .
, white s.d.m . , 1996 , mnras , in press nichol r.c .
, collins c.a . , guzzo l. , lumsden s.l . , 1992 , mnras , 255 , 21p peacock j.a .
, west m.j . , 1992 , mnras , 259 , 494 peebles p.j.e . , 1980 , the large - scale structure of the universe , princeton university press , princeton plionis m. , valdarnini r. , 1995 , mnras , 272 , 869 press w.h . , schechter p. , 1974 ,
apj , 187 , 425 ( ps ) white s.d.m . ,
frenk c.s . , 1991 , apj , 379 , 52 white s.d.m . , frenk c.s .
, davis m. , efstathiou g. , 1987 , apj , 313 , 505 white s.d.m . , rees m.j . , 1978 , mnras , 183 , 341
for a spherical perturbation in an einstein - de sitter universe , the physical radius @xmath15 of a mass shell which had initial lagrangian radius @xmath70 and mean linear overdensity @xmath71 is given for @xmath225 by ( see peebles 1980 ) @xmath226 @xmath227 for @xmath228 , we just replace @xmath229 in equation ( a1 ) by @xmath230 and @xmath231 in equation ( a2 ) by @xmath232 . without loss of generality , we assume @xmath233 at the time when the moments of halos and peaks are examined
. then @xmath71 depends only on the present mass overdensity @xmath234 . for @xmath235 , we can expand @xmath99 in power series : @xmath236 where the first five coefficients ( which are used in our model ) are @xmath237 ( see bernardeau 1992 ) . as shown by bernardeau
, the @xmath71-@xmath16 relation depends only very weakly on cosmological model in the quasilinear regime . | we develop an analytic model for the hierarchical correlation amplitudes @xmath0 [ where @xmath1,4,5 , and @xmath2 is the @xmath3th order connected moment of counts in spheres of radius @xmath4 of density peaks and dark matter halos in the quasi - linear regime .
the statistical distributions of density peaks and dark matter halos within the initial density field ( assumed gaussian ) are determined by the peak formalism of bardeen et al .
( 1986 ) and by an extension of the press - schechter formalism , respectively .
modifications of these distributions caused by gravitationally induced motions are treated using a spherical collapse model .
we test our model against results for @xmath5 and @xmath6 from a variety of n - body simulations .
the model works well for peaks even on scales where the second moment of mass ( @xmath7 ) is significantly greater than unity .
the model also works successfully for halos that are identified earlier than the time when the moments are calculated .
because halos are spatially exclusive at the time of their identification , our model is only qualitatively correct for halos identified at the same time as the moments are calculated . for currently popular initial density spectra ,
the values of @xmath8 at @xmath9 are significantly smaller for both halos and peaks than those for the mass , unless the linear bias parameter @xmath10 [ defined by @xmath11 for large @xmath4 is comparable to or less than unity .
the @xmath8 depend only weakly on @xmath10 for large @xmath10 but increase rapidly with decreasing @xmath10 at @xmath12 .
thus if galaxies are associated with peaks in the initial density field , or with dark halos formed at high redshifts , a measurement of @xmath8 in the quasi - linear regime should determine whether galaxies are significantly biased relative to the mass .
we use our model to interpret the observed high order correlation functions of galaxies and clusters .
we find that if the values of @xmath8 for galaxies are as high as those given by the apm survey , then apm galaxies should not be significantly biased . |
a combined quantum mechanical / molecular mechanical ( qm / mm ) method is a powerful computational tool for studying chemical reactions in solution and in biological systems.@xcite it treats a chemically active part of the entire system with accurate qm methods while the rest of the system with mm force fields .
the quality of a given qm / mm calculation depends primarily on the electronic structure method used . in the calculation of statistical properties like free energy ,
it is also important to adequately sample the relevant phase space.@xcite however , this phase space sampling is very demanding computationally , because one needs to calculate qm electronic energy for a large number of statistical samples .
one can ensure sufficient statistics by using fast semiempirical methods , but the resulting energetics may be less satisfactory than obtained with ab initio methods .
on the other hand , highly correlated qm methods require too much computational time and thus it becomes difficult to explore the phase space .
a variety of approaches have been proposed in order to address the above trade - off between accuracy and efficiency .
one approach is a family of dual - level methods , in which a classical or semiempirical potential is used for statistical sampling and an accurate qm method for energetic corrections.@xcite another approach is to introduce some approximation to the qm
mm electrostatic interactions in order to reduce the number of qm calculations .
our main interest in this paper is in the second approach above . in particular
, we are concerned with the following three _ embedding _ schemes that prescribe how to calculate the qm wavefunction in the mm environment : \(1 ) _ gas - phase embedding scheme .
_ this scheme totally neglects electrostatic perturbations of the mm environment on the qm subsystem .
the qm wavefunction is calculated in the gas phase , and the resulting charge density or partial charges are embedded into the mm environment .
the reaction path is also determined by the gas - phase calculation .
the free energy profile in solution is obtained via free energy perturbation ( fep ) calculations along the pre - determined reaction path .
this approach was first utilized by jorgensen and co - workers@xcite to study organic reactions in solution , and later by kollman and co - workers@xcite to study enzyme reactions .
it should be noted however that this approach may not be appropriate for a certain class of enzyme reactions.@xcite \(2 ) _ mean - field embedding scheme .
_ this method calculates the qm wavefunction in the presence of the mean field of the environment .
the averaged polarization ( or distortion ) of the qm wavefunction is thus correctly taken into account , while statistical fluctuations of the qm wavefunction are totally neglected .
indeed , this mean - field approximation has been the basis of many conventional solvation models like the pcm@xcite and rism - scf@xcite methods .
the mean - field idea has also been applied to the qm / mm framework by several authors .
for example , aguilar and co - workers @xcite performed geometry optimization on an approximate qm / mm free energy surface using the averaged solvent electrostatic potential ( asep)/md method .
more recently , the mean - field idea was exploited by warshel and co - workers@xcite in order to accelerate qm / mm calculation of solvation free energy .
\(3 ) _ polarizable embedding scheme . _
this method first develops a polarizable model of the qm subsystem and then embeds the resulting model into the mm environment .
the polarizable qm model can be developed , for example , by taylor expanding the qm electronic energy up to second order .
@xcite the qm / mm minimum free energy path ( mfep ) method by yang and co - workers @xcite is based on this perturbative expansion idea , which has been applied to chemical reactions in solution and enzymes . among the three embedding schemes above ,
the polarizable one is most accurate by allowing statistical fluctuations of the qm wavefunction .
the first goal of this paper is to formulate the mean - field embedding scheme above by starting from a variational principle for the qm / mm free energy ( ) .
as mentioned above , conventional solvation models are based on the mean - field embedding approximation .
they often start with a variational principle for the following free energy @xcite [ a_conv ] a ( ) = | | + .
minimization of @xmath2 in terms of @xmath3 gives a nonlinear equation for @xmath3 that is subject to the mean field of the environment .
very often , analytical gradient of free energy , @xmath4 , is obtained by utilizing the variational nature of @xmath2 . since those solvation models are quite successful in studying solution - phase chemistry , it is natural to try to extend them to the qm / mm framework .
the main benefits of this extension are as follows .
first , qm / mm models can describe inhomogeneous as well as homogeneous environments on an equal theoretical footing .
this makes it more straightforward to compare the chemical reactivity of a system in different environments ( e.g. , in solution and enzymes ) .
second , since the mean - field qm wavefunction is calculated only for a `` batch '' of mm configurations , the number of qm calculations can be made significantly smaller than a direct qm / mm statistical calculation .
as mentioned above , such a mean - field qm / mm approach has been explored by several authors in the literature .
for example , @xcite discussed such a method quite a few years ago based on a variational principle and linear - response approximation ( lra ) .
more recently , kato and co - workers@xcite developed the qm / mm lrfe method using a different type of variational / lra idea and applied it to chemical reactions in solution and enzymes .
on the other hand , aguilar and co - workers@xcite took a different approach in the asep / md method , where they did not invoke a variational principle nor lra but rather approximated the free energy gradient as follows : [ aguilar_grad ] = e ( , ) .
here , @xmath5 is the total energy of the qm / mm system and @xmath6 denotes the statistical average over mm degrees of freedom . note that in `` the average of the energy gradient '' ( the exact expression ) is replaced by `` the gradient of the averaged energy '' in the spirit of the mean - field approximation .
while aguilar demonstrated its accuracy via comparison with direct qm / mm calculations,@xcite the detailed derivation of was not provided and it was used as an ansatz . therefore , our first aim in this paper is ( i ) to formulate a mean - field qm / mm framework by starting from a variational principle ( but not invoking the lra ) , ( ii ) obtain analytical gradient of the associated free energy , and ( iii ) discuss a possible rationale for the approximate gradient in .
the second goal of this paper is to understand the relation of the above variational / mean - field procedure with the underlying exact qm / mm free energy as well as existing qm / mm theories ( secs .
[ sec : mf_pert ] and [ sec : fluc ] ) .
first , it is shown that the above variational procedure is equivalent with the first - order expansion of effective qm energy ( in the exact free energy expression ) about the self - consistent reference field .
as mentioned above , the qm / mm - mfep method @xcite is based on this type of perturbative expansions . therefore , it is interesting to compare the present approach with the qm / mm - mfep method in detail ( ) . from this comparison
it follows that the variational procedure is essentially equivalent with model 3 of the qm / mm - mfep method with charge response kernel @xmath7 neglected .
note however that the full version of model 3 includes that response kernel @xmath7 and thus it is more accurate by describing statistical fluctuations of the qm wavefunction .
therefore , in we also discuss several possible ways for evaluating such non - mean - field effects on top of the variational / mean - field calculation . as an illustration ,
the present method is applied to an reaction in water ( ) .
free energy profiles are obtained by integrating the free energy gradient and they are compared with free energy perturbation ( fep ) results . non - mean - field effects are also evaluated using a gaussian fluctuation model for the environment .
the obtained results suggest that the non - mean - field effects are rather small for the present reaction in water .
we consider the following qm / mm free energy ( or the potential of mean forces acting on qm atoms ) [ aqmmm ] ( ) = - d \ { - e ( , ) } , where @xmath8 and @xmath9 are cartesian coordinates of the qm and mm atoms , respectively , @xmath10 is the reciprocal temperature , and @xmath11 is the total energy given by [ total_energy ] e ( , ) = ( , ( , ) ) + ( , ) . here , @xmath12 is the electronic energy of the qm subsystem in the presence of an external electrostatic field ( called the effective qm energy ) . in the standard electronic embedding scheme ,
it is defined via the following equation [ se_cont ] [ + d ( ) ( ) ] | = | ( note that the prime symbol will be attached on variables and functions of `` dummy '' nature ) .
@xmath13 is the qm hamiltonian in the gas phase , @xmath14 is the qm charge density operator , [ rho ] ( ) = ^ _ z _ ( - _ ) - ^_i ( - _ i ) , and @xmath15 is an ( arbitrary ) external electrostatic field . in this paper
we will utilize a discretized approximation to given by [ se ] [ _ + _ _ v _ ] | ( , ) = _ ( , ) | ( , ) , where @xmath16 are a set of `` partial charge '' operators associated with qm atoms @xmath17 , and @xmath18 .
the motivation for using is that the external field can be parametrized by an @xmath19-dimensional vector @xmath20 , where @xmath19 is the number of qm atoms .
this fact makes the following discussion somewhat simpler .
nevertheless , we stress that there is no fundamental difficulty in using the original equation in ; see for such a formulation . in
, we summarize the present definition of the partial charge operator @xmath21 based on the electrostatic potential ( esp ) fitting procedure.@xcite now going back to , @xmath22 are mm electrostatic potentials acting on qm atoms , [ vmm_alpha ] v _ , = ( _ , ) , where [ vmm ] ( , ) = _ l with @xmath23 being partial charges of the mm atoms .
@xmath24 is the sum of the van der waals interactions between qm
mm subsystems and the internal energy of the mm subsystem , ( , ) = e_^ ( , ) + ( ) . in the following we will sometimes drop the arguments of @xmath25 and @xmath26 for notational simplicity , i.e. @xmath27 and @xmath28 . the free energy in may be rewritten as [ aqmmm_explicit ] ( ) = - d \ { - [ ( , ) | + | ( , ) + ] } , by explicitly writing @xmath29 in terms of the qm wavefunction .
note that @xmath25 always stands for @xmath30 as mentioned above . a direct evaluation of @xmath31 is computationally demanding because @xmath32 depends on @xmath9 through @xmath27 .
to avoid repeated qm calculations , let us replace the true wavefunction @xmath32 by some trial one @xmath33 that best approximates the true wavefunction in a statistically averaged sense .
to do so , we consider a free energy functional of the form [ a_vp ] = \ { - [ | + | + ] } .
since the following inequality holds by definition for arbitrary @xmath27 [ we assume that @xmath34 is the ground state of @xmath35 , ( , ) | + | ( , ) ( ) | + namely , @xmath36
$ ] is a strict upper bound on @xmath31 , and the best approximation to @xmath31 is obtained by minimizing @xmath36 $ ] with respect to @xmath37 .
this variational principle is indeed a direct qm / mm analog of the standard ones used in conventional solvation theories.@xcite by minimizing the following lagrangian to account for the normalization of @xmath37 , l [ , , ] = - \ { @xmath37 : [ nlse ] | = | . here
@xmath38 represents the statistical average over mm degrees of freedom , [ mm_ave ] _ = , and @xmath39 = \langle \psi ' | \qvecop | \psi ' \rangle$ ] .
[ in this paper the double bracket @xmath38 indicates that the average is of `` classical '' nature , i.e. , it does not require repeated qm calculations . ]
since is nonlinear with respect to @xmath37 , it is usually solved via iteration .
it follows from comparison between that @xmath37 and @xmath40 may be written as & = & ( , ) , + & = & ( , ) , where @xmath41 is the self - consistent response field determined by [ sc_cond ] ( ) & = & ( , ) _ , + ( ) & = & | | , with @xmath42 .
in this paper , will be called the self - consistent ( embedding ) condition .
the minimum value of the free energy functional is obtained by inserting @xmath43 into : _ & = & \ { - [ | + | + ] } + & = & | | + ( , ) + & & ( ) , [ amf_var ] where @xmath44 is defined by [ amm ] ( , ) = e ^ - [ + ] .
note that @xmath45 has been extracted from the integral over @xmath9 since it is independent of @xmath9 . by the last line of
, we define the qm / mm free energy with mean - field embedding approximation , @xmath46 .
the analytical gradient of @xmath46 can be obtained using a standard procedure as follows .
first , we rewrite the @xmath46 in terms of the lagrangian , ( ) = = l [ , , ] with @xmath47 . recalling that @xmath48 $ ] is stationary with respect to @xmath37 and @xmath40 , we obtain ( ) = . | _ = ,
= , where the @xmath8 derivative in the right - hand side does not act on @xmath37 nor @xmath40 .
we then obtain the analytical gradient in , which will be discussed in the next section .
@xmath46 in can also be obtained from the exact @xmath49 in by taylor expanding the effective qm energy @xmath29 up to first order : @xmath50 here , @xmath51 is an arbitrary reference potential that is assumed to be independent of @xmath9 . using the following hellman - feynman theorem for @xmath12 , [ partial_charge ] & = & ( , ) | | ( , ) + & & ( , ) , and introducing the internal qm energy as [ eqm ] ( , ) = ( , ) | | ( , ) , or alternately via [ internal_qm_energy ] ( , ) = ( , ) + ( , ) , may be rewritten as ( , ) ( , ) + ( , ) .
( see appendix [ sec : non - var ] for cases where the hellmann - feynman theorem does not hold ) . inserting the above expansion of @xmath12 into the exact @xmath31 and extracting @xmath52 from the integral over @xmath9 , we obtain [ aqmm_ref ] ( ) ( , ) + ( , ) with @xmath53 .
the above equation is very similar to @xmath46 in , and indeed , the latter can be recovered simply by setting @xmath54 to the self - consistent potential @xmath41 : [ amf ] ( ) = ( , ) + ( , ) .
therefore , the variational principle in is essentially equivalent with the first - order expansion of the effective qm energy about the self - consistent reference field .
the analytical gradient of @xmath46 can be obtained by first writing @xmath46 in terms of the effective qm energy as ( ) & = & ( , ( ) ) - ( ) ( ) + & & + ( , ( ) ) , taking the @xmath8 derivative of the right - hand side , and using the following symmetric relations : .
= = ( ) , .
it then follows that the derivatives of @xmath55 and @xmath56 cancel with each other and we are left with the following : [ grad ] ( ) = .
this is our working equation for the gradient of qm / mm free energy with mean - field embedding [ see also for related equations ] .
the first term is the partial @xmath8 derivative of the effective qm energy ( _ not _ of internal qm energy @xmath57 ) , which may be written using the hellman - feynman theorem as [ eeffqm_deriv ] .
= = | + _ _ | .
the second term in is the partial @xmath8 derivative of the ( classical ) solvation free energy , which may be written using as [ amm_deriv ] .
= = q^ _ + _ .
an alternative way to obtain the free energy gradient in is as follows ( see also ) .
first we define the mean - field approximation to the total energy @xmath11 and the qm / mm free energy @xmath49 as [ emf ] ( , ) = ( , ) + ( - ) + , and ( ) = - d \ { - ( , ) } . the gradient of @xmath46 then becomes ( ) & = & _ + & = & _ .
inserting into the above equation and using the self - consistency condition in gives the free energy gradient in .
we note that if the reference potential @xmath54 is not set at the self - consistent one , i.e. @xmath58 , the following term appears due to incomplete cancellation among terms , , which requires the derivative of qm charges calculated in the reference potential , @xmath59 .
we now compare the above perturbative approach with the qm / mm - mfep@xcite and asep / md methods.@xcite the qm / mm - mfep method develops a series of polarizable qm models by taylor expanding its energy and esp charges up to first or second order .
their comparison with the present approach is made in . from this comparison
it follows that @xmath46 is essentially equivalent with model 3 of the qm / mm - mfep method with charge response kernel @xmath7 neglected .
the free energy gradient of model 3 ( with @xmath7 neglected ) looks somewhat different from the present result at first sight .
however , the former can be rewritten using the self - consistency condition as follows ( see for the notation ) + _ , [ yang_grad ] which is essentially equivalent with the present gradient expressions [ eqs .
( [ grad ] ) , ( [ grad_cont ] ) , and ( [ grad_mixed ] ) ] .
the above equation provides some rationale for the approximate gradient used by the asep / md method [ ] . to see this , let us rewrite as [ aguilar_grad_detail ] | + | , with @xmath60 , \end{gathered}\ ] ] which suggests that the gradient of @xmath2 may be viewed as the gradient of effective qm energy calculated in the averaged mm potential . here it should be noted that the @xmath61-derivative above does not act on @xmath62 nor @xmath63 , no matter whether the self - consistency condition is assumed or not ( see for details ) . as seen above , the variational / mean - field approach totally neglects statistical fluctuations of the qm wavefunction about the mean - field state .
the aim of this section is thus to discuss several ways for evaluating such non - mean - field effects on qm / mm free energy .
first , let us separate the total energy into the mean - field and non - mean - field contributions as follows : [ e_separate ] e ( , ) = ( , ) + ( , ) , where @xmath64 is defined by and @xmath65 is the remaining part of the total energy . using the definition of @xmath11 in
, the non - mean - field term can be written more explicitly as @xmath66 inserting into the exact @xmath49 in gives [ a_separate ] ( ) = ( ) + ( ) , where [ afluc ] ( ) = - ( - ) _ .
we note that up to this point are still exact .
the statistical average in can be evaluated rather rigorously as follows .
first , one calculates a long trajectory of the mm subsystem using the sampling function @xmath67 , selects a relatively small subset of mm configurations from the long trajectory ( say , 500 samples ) , and calculates @xmath68 for those selected configurations in order to take the average of @xmath69 .
indeed , this is a type of dual - level qm / mm sampling method , where @xmath67 is used as a low - cost sampling function while @xmath68 gives energetic corrections .
although the above dual - level method is rigorous , it requires hundreds of qm calculations and thus may be rather expensive .
one approach for reducing the computational cost is to truncate the expansion of effective qm energy at the second order,@xcite [ efluc_approx ] ^(2 ) = ( - ) ( - ) , where @xmath70 is defined by [ chiqm ] ( , ) = = with @xmath71 .
@xmath70 is also called the charge response kernel due to the second equality in .@xcite once @xmath70 is obtained , the statistical average of @xmath72 can be evaluated with no extra qm calculations , thus significantly reducing the computational cost .
the above second - order expansion is also utilized by models 2 and 3 of the qm / mm - mfep method [ see in the present paper ] in order to describe statistical fluctuations of the qm wavefunction .
@xcite a further simplification can be made by introducing a gaussian fluctuation model for the mm environment .
specifically , we assume that the mm electrostatic potential acting on qm atoms , @xmath27 , takes a ( multi - dimensional ) gaussian distribution : @xcite @xmath73 .\end{gathered}\ ] ] here , @xmath74 is the covariance matrix of @xmath25 , = ( - ) ( - ) ^t _ . by combining @xmath75 in and the gaussian fluctuation model above
, we obtain an approximate analytical expression for @xmath76 : [ afluc_gf ] ( ) [ 1 + ] .
note that since @xmath70 is negative definite , @xcite @xmath75 and @xmath76 are always negative .
the basic appeal of is that once the charge response kernel @xmath70 is obtained , @xmath76 can also be obtained simultaneously by combining with @xmath74 that is available from the mean - field calculation .
@xmath70 can be evaluated most efficiently by solving a coupled - perturbed hartree - fock or kohn - sham equation , @xcite or more primitively , by numerically differentiating the esp charges @xmath77 with respect to @xmath78 based on the second equality in .
we now apply the above method to a type - ii reaction in water ( the menshutkin reaction ) [ menshutkin ] .
this reaction is known to exhibit greatly enhanced rates in polar solvents than in the gas phase due to strong electrostatic stabilization of the products.@xcite this is in contrast to type - i reactions like @xmath79 which are decelerated by greater electrostatic stabilization of the reactant than of the transition state . due to the great acceleration in rate ,
the menshutkin reaction became the subject of many theoretical studies.@xcite gao and xia performed the first extensive qm / mm study using the am1 model,@xcite and demonstrated that the transition state in water is shifted remarkably toward the reactant region .
continuum solvent models were also applied to the same reaction at various levels of qm methods .
@xcite while those studies observed that continuum models can provide free energetics similar to the qm / mm results,@xcite it was also argued that those models may not be appropriate for reaction ( [ menshutkin ] ) due to the presence of hydrogen bonds.@xcite since then , several qm / mm(-type ) calculations were performed , @xcite including the rism - scf method,@xcite a mean - field qm / mm approach,@xcite and a dual - level method.@xcite overall , those calculations are in reasonable agreement with each other , predicting the free energy of activation @xmath80 to be 20 @xmath81 30 kcal / mol and the free energy of reaction @xmath82 to be @xmath83 @xmath81 @xmath84 kcal / mol ( both including solute entropic contributions ) . among those studies ,
the present one is most similar in spirit to the mean - field qm / mm calculation by aguilar and co - workers .
@xcite following previous studies , we define the reaction coordinate as [ rc ] ( ) = r ( ) - r
the mean - field free energy @xmath46 is minimized with respect to @xmath8 under the constraint @xmath85 .
the resulting optimized geometry will be denoted as @xmath86 .
our goal here is to obtain the free energy profile @xmath87 as a function of @xmath88 . in this paper
we constructed such a profile by integrating @xmath89 along the optimized reaction path @xmath86 [ i.e. , via thermodynamic integration ( ti ) ] : @xmath90 where @xmath91 .
we calculated @xmath92 and @xmath93 for equally spaced grid points , @xmath94 @xmath95 , and evaluated the above integral via cubic spline interpolation . in practice , the following trapezoid rule was also sufficient for this small size of grid spacing : @xmath96 \qquad\qquad\qquad \\
\qquad \times \frac{1}{2 } \left [ \nabla \amf ( \ropt ( \rc_{k } ) ) + \nabla \amf ( \ropt ( \rc_{k-1 } ) ) \right ] .\end{gathered}\ ] ] the geometry optimization on @xmath46 was performed by adapting the sequential sampling / optimization method by yang and co - workersfor the present purpose . since the free energy gradient in assumes that the self - consistency ( sc ) condition in is satisfied
, one might think that the latter must be solved for each step of the optimization .
then , the optimization procedure may appear the following : 1 . given a geometry @xmath97 at an intermediate step @xmath98 ,
solve the sc condition in for obtaining @xmath41 and @xmath99 , and evaluate the analytical gradient @xmath100 via ; 2 . advance @xmath97 one step , e.g. , as @xmath101 ; and 3 .
repeat steps 1 and 2 until a given convergence criterion is met , e.g. , @xmath102 . however , this scheme is somewhat too restrictive because the gradient does not have to be exact nor very accurate at the early stages of the optimization .
what is needed is that the gradient becomes increasingly more accurate as the optimization proceeds .
this observation leads to the following variant of the sequential sampling / optimization procedure , which performs the statistical sampling of the mm environment and the optimization of the qm geometry in an iterative manner : 1 .
cycle 0 : _ qm optimization .
_ + the qm subsystem is optimized in the gas phase to prepare the initial state . the resulting qm geometry and partial charges are denoted as @xmath103 and @xmath104 .
no mm / md simulation is performed at this cycle .
cycle @xmath98 : ( i ) _ mm sampling . _ + the qm geometry @xmath105 and charges @xmath106 of the previous cycle are embedded into the mm environment .
an mm / md simulation is then performed to evaluate the averaged electrostatic potential and the gradient of @xmath44 in : ^(n ) = _ ^(n-1 ) , ^(n-1 ) , ^(n ) _ & = & .
^(n-1 ) , ^(n-1 ) + & = & q^(n-1 ) _ + _ ^(n-1 ) , ^(n-1 ) .
since @xmath107 and @xmath108 are simple @xmath19- and @xmath109-dimensional vectors ( @xmath19 is the number of qm atoms ) , they can be accumulated directly in the md simulation .
cycle @xmath98 : ( ii ) _ qm optimization . _
+ the qm geometry is optimized in the presence of @xmath107 and @xmath110 . to this end , we employ the following target function for optimizing the qm geometry @xmath8 : [ target_e ] a^(n ) ( ) & = & ( , ^(n ) ) + & & + _ ^(n ) _ ( _ - ^(n-1 ) _ ) .
the gradient of this target function is [ grad_approx ] a^(n ) ( ) & = & + ^(n ) _ + & = & .
= ^(n ) + & & + . | _
^(n-1 ) , ^(n-1 ) .
we note that @xmath111 gives a local approximation to @xmath46 in that its gradient @xmath112 approximates the analytical gradient @xmath113 in . by minimizing @xmath111 under the constraint @xmath114 [ i.e. ,
by eliminating the orthogonal component of @xmath112 to @xmath115 , we obtain a new qm geometry @xmath97 for the next cycle . in this paper the optimization of @xmath111
was performed by adding linear external potentials @xmath116 and forces @xmath117 for individual qm atoms in the gamess quantum chemistry package.@xcite 4 . by iterating over the above cycles , the qm geometry , esp charges , and mm
mean potentials converge to their asymptotic values , namely @xmath118 , @xmath119 , and @xmath120 .
this means that the sc condition is satisfied to a good accuracy . since the qm geometry does not move any further
, we may regard @xmath121 as providing a good approximation to @xmath122 at the optimized geometry @xmath123 .
the above iterative optimization was used to obtain the free energy gradient in .
the other computational details are as follows .
the qm and mm calculations were performed using modified versions of gamess@xcite and dl_poly packages.@xcite following truong , @xcite we used the bhhlyp/6 - 31+g(d , p ) method in most calculations .
previous study shows that this method gives results similar to the mp4/aug - cc - pvdz level for the present reaction.@xcite the esp charge operator and associated fitting grid were defined following ten - no @xcite and spackman.@xcite the charge response kernel @xmath70 was calculated via finite difference of the esp charges @xmath77 with respect to @xmath78 .
@xmath124 symmetry was enforced on the qm subsystem .
the optimization tolerance was set to @xmath125 hartree / bohr , which is five times larger than the default setting in gamess . although @xmath107 and @xmath110 above should satisfy @xmath124 symmetry in principle , they do not in practice due to statistical errors .
these errors generate small artificial components of overall translation and rotation .
we thus removed those components manually such that the optimization could be completed to a given tolerance .
md calculations were performed by solvating one solute molecule into 253 water molecules ( with the tip3p potential)@xcite in a cubic box of side length 19.7 at @xmath126 = 300 k. periodic boundary condition was applied , and electrostatic potentials were calculated using the ewald method .
the lennard - jones parameters are listed in table [ tab : lj ] .
the timestep for integration was 2 fs .
one iterative optimization cycle consisted of 50000 md steps for equilibration and 300000 steps for production .
although 100000 production steps were sufficient for obtaining a similar result , we did not attempt to minimize the computational efforts . rather , we aimed at obtaining a highly converged result , such that the statistical errors become comparable to the width of the plotting line .
.[tab : lj]lennard - jones parameters for the solute molecule .
parameters of the cl atom are taken from gao and xia ( ) , and those of the other atoms are from the amber94 force field ( ) . [
cols="^,^,^",options="header " , ]
_ numerical stability of the esp charge operator .
_ esp charges and associated charge operator @xmath127 are sometimes numerically unstable , as often stressed in the literature.@xcite for example , we observed an oscillatory behavior of partial charges within the ch@xmath128 group during the optimization cycles .
this was typical for @xmath129 , where the ion pair products start to form .
since these oscillations are partly due to ambiguous assignment of partial charges for `` buried '' atoms , @xcite the resp method@xcite was of great help in suppressing those oscillations . however , the resp method was of little help in removing a divergent behavior of partial charges within the nh@xmath128 group observed in the reactant asymptotic region ( @xmath130 ) .
specifically , the partial charge on the n ( h@xmath131 ) atom kept on growing in the negative ( positive ) direction during the optimization cycles .
this might be due to inherent limitations of the present charge model , where partial charges are placed only on atomic nuclei and the lone pair on the n atom may be poorly described .
@xcite in this respect , it may be more straightforward to use the continuous or mixed representation in or [ sec : mixed_rep ] , where one embeds the mm point charges directly into the qm hamiltonian . see for this type of implementation .
_ fep that connects optimized geometries .
_ if one is interested only in the free energy difference between two stationary points ( e.g. , activation free energy ) , it is probably more efficient to use fep than ti . specifically , one first searches the free energy surface for stationary points by using the free energy gradient ( and possibly the hessian ) , and then connects these points via fep .
the qm geometries and charges of intermediate points could be generated by linear interpolation of two end points .
see for such a calculation . in this way
, one can reduce the number of costly free energy optimization .
if one also needs to know a rough free energy profile , one could perform additional optimization for a limited number of intermediate points and then connect them via fep .
_ solute thermal / entropic contributions . _ the method in this paper calculates the qm / mm free energy for a given fixed qm geometry .
the thermal / entropic contributions of the qm subsystem thus need to be taken into account separately , e.g. , via harmonic vibration approximations .
this is a well - known limitation of the present type of method , which is also shared by conventional solvation theories . to overcome this limitation ,
several methods have been proposed for including the solute flexibility into the qm / mm free energy calculation at a reasonable computational cost .
@xcite to conclude , we have presented a direct qm / mm analog of conventional solvation theories based on variational and perturbative frameworks .
the main approximation in this paper is that the true qm wavefunction is replaced by an averaged one that is calculated in the mm mean field .
we stress however that the electrostatic interactions between the averaged qm wavefunction and the mm environment are calculated correctly without further approximations .
the basic appeal of the mean - field qm / mm approach is that it can describe different environments ( e.g. , solutions and enzymes ) on an equal theoretical footing , while the number of qm calculations can be made significantly smaller than a direct qm / mm calculation .
this work was supported by grant - in - aid for the global coe program , `` international center for integrated research and advanced education in materials science , '' from the ministry of education , culture , sports , science and technology of japan .
the author also thanks prof .
weitao yang for a critical reading of the manuscript and suggesting detailed comparison with the qm / mm - mfep method .
the main text is based on the approximate equation in , where qm / mm electrostatic interactions are `` discretized '' in terms of the esp charge operator . in this section
we summarize an alternative formulation using the continuous equation in .
first , the total energy is given by e ( , ) = + ( , ) , where @xmath132 $ ] is defined via with @xmath133 .
we then expand @xmath132 $ ] in terms of @xmath134 up to first order , & & + d ( | ) \ { ( , ) - ( | ) } + & = & + d ( | ) ( , ) , where @xmath135 = \langle \psisc | \hqm | \psisc \rangle$ ] and we have used the following hellmann - feynman theorem : = | ( ) | . @xmath136 and @xmath137 are obtained from the following self - consistency condition , [ sc_cond_cont ] [ vsc_cont ] ( | ) = ( , ) _ , ( | ) = | ( ) | , with @xmath138 $ ] .
note that @xmath139 and @xmath140 depend parametrically on @xmath8 via .
@xmath38 is the statistical average defined by _
= , with [ eeffmm_cont ] = d ( ) ( , ) + ( , ) . inserting the above first - order expansion into @xmath31 in
gives the qm / mm free energy with mean - field embedding , [ amf_cont ] ( ) = + , with [ amm_cont ] = e^ - .
the gradient of @xmath46 can be obtained via similar arguments as [ grad_cont ] ( ) = .
the first term represents the energy gradient in a fixed external field .
the second term may be rewritten using as [ amm_deriv_cont ] .
note that the above equation lacks the electrostatic term like @xmath141 that is present in the the discretized case [ ] .
this discrepancy originates from the different physical meaning of @xmath142 / \del \rvec $ ] ( in the continuous representation ) and @xmath143 ( in the discretized representation ) . in the discretized case ,
the external potential values @xmath78 acting on qm atoms are kept constant while varying the nuclear coordinates @xmath8 . in the continuous case , the external potential _
@xmath15 is kept constant while varying @xmath8 .
this means that the potential values acting on qm atoms , @xmath144 , may vary as a function of @xmath8 .
the situation becomes clear by considering the following relation : .
| _ = & & .
( , \ { ( _ | ) } ) | _ = + & & . | _
= + & + & q^ _ ( ) _ , where we have used the following identity obtained from : [ vmm_deriv ] . |
_ = & = & . |
_ = _ + & = & .
_ | _ = _ .
therefore , it follows that the missing electrostatic term in is now accounted for by the energy gradient term in .
the esp charges @xmath145 for a given wavefunction @xmath3 are obtained by minimizing the following function@xcite [ esp_l ] l ( , ) & = & _ l^ \ { d - _ } ^2 + & & - , where @xmath146 are the esp fitting grid and @xmath147 is the total charge . by requiring that @xmath148 ,
inverting the resulting linear equations for @xmath149 , and determining @xmath40 via @xmath150 , we obtain q _
= | _ | , where @xmath151 is an explicit function of @xmath152 and @xmath147 , with @xmath153 being the electron coordinates .
see previous work for the explicit form of @xmath151 in the atomic orbital basis .
@xcite the above definition of @xmath127 suggests that one may make the following replacement d _ , as long as @xmath154 is located outside the core region of the qm charge density .
then , the continuous qm / mm electrostatic interaction may be discretized as d ( ) ( , ) _ _ ( _ , ) , by inserting the definition of @xmath155 in .
this is the present rationale for using @xmath127 in .
here we compare the perturbative treatment of @xmath46 in with the qm / mm - mfep method .
@xcite the starting point is the same as ( here expressed using the notation in ) , [ yang_aqmmm ] ( ) = - d \ { - e ( , ) } , where @xmath156 , @xmath157 , and the total energy is given by [ yang_e ] e ( , ) = | | + ( , ) .
@xmath158 is the effective qm hamiltonian in the presence of the mm electrostatic field , [ yang_heff ] = + d ( ) ( , ) , and other quantities like @xmath13 are defined in the main text .
we then introduce the mm mean field as [ yang_vmmref ] ( ) = ( , ( ) ) , where @xmath159 are a set of mm configurations obtained from the previous cycle of the sequential sampling / optimization method.@xcite the qm hamiltonian in the presence of the mm mean field is defined by [ yang_heffref ] = + d ( ) ( ) . the eigenfunction and eigenenergy of @xmath160 are denoted as @xmath161 and @xmath162 , and the esp charges derived from @xmath161 are written as @xmath163 . the internal qm energy associated with @xmath160
is defined as [ yang_einternal ] ( ) - _ i ( ) ( ) , i.e. , by subtracting the qm / mm electrostatic interaction energy expressed in terms of esp charges from the effective qm energy .
the qm / mm - mfep method then develops a series of polarizable qm models by taylor expanding its energy and esp charges up to first or second order . among others ,
model 3 ( `` qm point charges with polarization due to mm and qm atoms '' ) approximates the total energy as follows [ eqs . ( 36 ) and ( 40 ) of ] : ( , ) & = & ( ) + & + & _ i ( ) ( , ) + & + & [ ( , ) - ( ) ] + & & _ ij [ ( , ) - ( ) ] + & + & ( , ) , [ yang_e_model3 ] where @xmath164 is the charge response kernel in .
the above equation may be viewed as the second - order expansion of the effective qm energy in terms of mm electrostatic potential [ cf . ] .
the gradient of qm / mm free energy is obtained by inserting into the following , [ yang_grad0 ] _ , or alternately , into an fep - type expression [ eq .
( 6 ) of ] [ yang_fep ] , where @xmath165 is the reference sampling function that is obtained from the previous cycle of the sequential sampling / optimization method . the main difference of the present approach from the qm / mm - mfep method is that the present one utilizes the self - consistency condition in order to simplify the gradient expression . to see this ,
let us insert @xmath166 in into the statistical average in , & = & + & & + ( , ) _ + & & + _ i ( ) _ + & & + _ , [ yang_grad1 ] where terms depending on @xmath164 have been neglected ( they are treated separately in ) . using , we may rewrite the above equation as & = & + \ { ( , ) _ - ( ) } + & & + _ i ( ) \ { _ - } + _ .
[ yang_grad2 ] now let us assume that the reference mm coordinates @xmath167 satisfies the following self - consistency condition [ yang_sc_cond ] f ( , ) _ f ( , ( ) ) , f , which is expected to hold well for the last few cycles of the sequential sampling / optimization method.then , by setting @xmath168 or @xmath169 , we have [ yang_vmm_deriv ] ( , ) _ & & ( , ( ) ) + & = & ( ) , + _ & & + & = & , which suggest that the curly brackets in vanish , and as a result we obtain a simpler expression for the free energy gradient , + _ .
[ yang_grad_alt ] this form is found to be equivalent with the present gradient expressions , e.g. , .
as seen from , the qm / mm - mfep method is based on a `` mixed '' representation of the qm / mm electrostatic interactions .
that is , the qm wavefunction is calculated with the continuous equation in , while the internal qm energy etc are defined in terms of esp charges . in this mixed representation
, @xmath46 may be defined as [ amf_mixed ] ( ) & = & - _ q^ _ ( ) ( _ | ) + & & + ( , ) , and the mixed form of the self - consistency condition is [ sc_cond_mixed ] [ sc_cond_mixed_vmm ] ( | ) & = & ( , ) _ , + q^ _ ( ) & = & | _ | , where @xmath170 $ ] .
the gradient of @xmath46 then becomes ( ) & = & .
v = + _ \ { ( _ , ) _ - ( _ | ) } + & & + q^ _ ( ) \ { _ - . | _ = } + _ + & & + d ( | ) - _ q^ _ ( ) _ = _ , [ mixed_deriv1 ] where @xmath171
. the curly brackets in the above equation vanish by using and its derivative with respect to @xmath172 [ see also ] .
the third line also vanishes approximately since @xmath99 represent the esp charges that correspond to @xmath140 .
therefore , we obtain the following gradient : [ grad_mixed ] ( ) . | _ v = + _ .
the main text assumes that the underlying qm wavefunction is exact or calculated using qm methods with variational nature ( e.g. , hartree - fock and dft ) .
this means that the hellmann - feynman theorem holds and it can be used to define partial charges via .
however , this is not the case for non - variational qm methods like the mp2 theory . in the latter case
, one needs to generalize the definition of partial charges as follows , [ gen_partial_charge ] ( , ) , since the first derivative of effective qm energy plays the role of partial charges as described in .
accordingly , one needs to define the internal qm energy as ( , ) ( , ) - ( , ) . with these definitions
the discussion in remains valid .
however , the actual calculation of generalized partial charges in may be tedious unless some analytical algorithms are available .
fortunately , in the mp2 method one can avoid such a calculation by discarding higher - order terms in correlation energy.@xcite to see this , let us denote relevant quantities at the mp2 level as @xmath173 and @xmath174 etc , and the difference between the mp2 and hf levels as @xmath175 etc .
then , the mean - field free energy at the mp2 level may be written as a _ & = & e _ ( ) + ( ) + & = & _ ( ) - + ( ) . by inserting @xmath176 and @xmath177 , and making the first - order expansion in terms of @xmath178 and @xmath179
, we have a _ & = & _ ( ) - + ( ) + o ( ^2 ) + & = & a _ + _ + o ( ^2 ) , where _ = _ ( ) - _ ( ) . since @xmath180 is of higher order in correlation energy , @xcite it
may safely be neglected at the mp2 level . the mp2 correction for free energy
is thus given by @xmath181 , and we do not need to calculate @xmath182 nor @xmath183 explicitly .
the @xmath181 can be evaluated using the standard expression _
= _ abrs , where @xmath184 etc are obtained with @xmath185 . | conventional quantum chemical solvation theories are based on the mean - field embedding approximation .
that is , the electronic wavefunction is calculated in the presence of the mean field of the environment . in this paper a direct quantum mechanical / molecular mechanical ( qm / mm ) analog of such a mean - field theory
is formulated based on variational and perturbative frameworks . in the variational framework ,
an appropriate qm / mm free energy functional is defined and is minimized in terms of the trial wavefunction that best approximates the true qm wavefunction in a statistically averaged sense .
analytical free energy gradient is obtained , which takes the form of the gradient of effective qm energy calculated in the averaged mm potential . in the perturbative framework ,
the above variational procedure is shown to be equivalent with the first - order expansion of the qm energy ( in the exact free energy expression ) about the self - consistent reference field .
this helps understand the relation between the variational procedure and the exact qm / mm free energy as well as existing qm / mm theories .
based on this , several ways are discussed for evaluating non - mean - field effects ( i.e. , statistical fluctuations of the qm wavefunction ) that are neglected in the mean - field calculation . as an illustration , the method is applied to an menshutkin reaction in water , @xmath0 for which free energy profiles are obtained at the hf , mp2 , b3lyp , and bh&hlyp levels by integrating the free energy gradient .
non - mean - field effects are evaluated to be @xmath1 kcal / mol using a gaussian fluctuation model for the environment , which suggests that those effects are rather small for the present reaction in water . |
v2051 oph ( @xmath0 ) belongs to a small group of ultrashort period eclipsing dwarf nova together with z cha , oy car and ht cas . during 1996 ,
it was seen at a low brightness state ( @xmath1 ) , in which mass accretion was reduced to a minimum ( baptista et al .
1998 ) .
high - speed ultraviolet and optical spectroscopy was secured with the faint object spectrograph onboard the hubble space telescope on 1996 jan 29 ( uv : g400h grating , @xmath2 , @xmath3 pixel@xmath4 ; optical : g160l grating , @xmath5 , @xmath6 pixel@xmath4 ) at a time resolution of 3.4s .
the runs cover the eclipse cycles 109988 ( uv ) and 109989 ( optical ) , according to the ephemeris of baptista et al .
( 2003 ) .
the out - of - eclipse uv spectra show prominent ly@xmath7 @xmath8 , cii @xmath9 , siiv @xmath10 and civ @xmath11 emission lines as well as broad absorption bands possibly due to feii .
the optical spectra show the balmer continuum in emission as well as he emission lines and strong balmer emission lines .
the uv spectra were divided into 34 narrow passbands ( 19 - 60 wide in the continuum and velocity - resolved bins of 2000 km / s or 3000 km / s for the lines ) while the optical spectra were divided into 68 passbands ( 15 - 42 wide in the continuum and velocity - resolved bins of 1000 km / s for the lines ) and light curves were extracted for each one
. maximum - entropy eclipse mapping techniques ( baptista & steiner 1993 ) were used to solve for a map of the disc brightness distribution and for the flux of an additional uneclipsed component in each band .
the line centre maps show broad brightness distributions from an extended and optically thick region , possibly ionized by the boundary layer .
velocity - resolved h@xmath12 line light curves show the expected behavior for the eclipse of gas rotating in prograde sense ( rotational disturbance ) , with the blue wing ( @xmath13 ) being eclipsed earlier than the red wing ( @xmath14 ) .
net line emission maps show that the balmer lines are in absorption at the disc centre , and that there is also little contribution from the bright spot at the disc rim .
in contrast with the observed in the line maps , the continuum maps show pronounced emission from the compact white dwarf and bright spot sources as well as enhanced emission along the ballistic stream trajectory of gas stream , an evidence of gas stream overflow .
we sliced the disc into concentric annullar regions of width @xmath15 to compute disc spectra as a function of distance .
motivated by the asymmetric emission patterns observed in the eclipse maps , we extracted spatially resolved spectra for three distinct disc regions : `` front '' ( the disc side closest to the secondary star ) , `` back '' ( the side disc farther away from the secondary star ) and `` gas stream '' ( a thin slice around the stream trajectory in each annullar region ) .
the disc ( front and back ) spectra show that the balmer decrement becomes shallower and the lines are more prominent and narrower with increasing radii .
the `` front '' side is systematically brighter than the `` back '' side in agreement with the results of vrielmann et al .
the difference becomes more pronounced as the radius increases .
the comparison of the `` front '' and `` back '' spectra show that the latter contain broad absorption bands , possibly due to feii , that become more conspicuous in the outer parts of the disc suggesting that they arises from absorption along the line of sight in a vertically extended region above the disc ( e.g. horne et al . 1994 ) .
the spectrum of the central pixel shows a continuum filled with broad and shallow absorption lines and a balmer jump in absorption , and resembles that of a da white dwarf .
we computed the ratio between the spectrum of the `` gas stream '' and that of the `` front '' side at the same radius to investigate the emission along the gas stream trajectory .
the spectrum of the gas stream is clearly distinct from the disc spectrum for a large range of radii .
in contrast to the observed in the disc spectra , the lines appear in absorption .
the distinct emission along the gas stream trajectory underscores the evidence of gas stream overflow .
the fact that the lines appear in absorption in the stream spectra suggest the presence of matter outside of the orbital plane ( e.g. a chomosphere + wind ) .
the uneclipsed component is dominated by the balmer continuum and strong balmer emission lines in the optical and by resonant emission lines in the uv , suggesting that it arises in a vertically extended region .
the strong lines in the optical and uv spectra as well as the balmer continuum accounts for @xmath16 per cent of the total light at the respective wavelength .
the continuum in the uneclipsed spectrum yield negligible fractional contributions to the total light at the corresponding wavelengths . | we report a study of the spectra and structure of the accretion disc of the v2051 oph while the star was in an unusual , faint brightness state during 1996 . |
the issue of identifiers is crucial in distributed computing @xcite .
indeed , the correct operation of deterministic protocols often relies on the assumption that each processor @xmath7 comes with with a unique _ identity _ , @xmath8 @xcite .
informally , in network computing , such an identity assignment is crucial for tackling two of the fundamental difficulties that are inherent to distributed computing , namely : ( 1 ) _ symmetry breaking _ , and ( 2 ) _ topological information gathering_. the use of identities for tackling the above two difficulties is illustrated well in the context of _ local _ algorithms @xcite .
indeed , in the @xmath9 model @xcite , an algorithm that runs in @xmath3 communication rounds , assuming an identity assignment , can be viewed as composed of two parts : first , collecting at each node @xmath7 , the ball @xmath10 of radius @xmath3 around it ( together with the inputs of nodes ) , and second , deciding the output at @xmath7 based solely on the information in @xmath10 . to achieve these two tasks ,
one should first obtain the ball @xmath10 , which may not be possible if the underlying graph is anonymous ( i.e. , without identities ) . moreover , even if obtaining the ball is possible , e.g. , if the structure of the graph allows it , the absence of unique identities given to the nodes may prevent the algorithm from breaking symmetry .
for example , in the absence of unique identities , it is impossible to design a distributed deterministic coloring algorithm , even for the symmetric connected graph composed of two nodes only . in fact , to the best of our knowledge , all algorithms in the @xmath9 model are designed assuming the presence of pairwise distinct identities or some other type of node - labeling or edge - labeling , including , e.g. , sense of direction @xcite . the seminal paper of naor and stockmeyer @xcite provides an important insight regarding the role of identities in local computation .
informally , they show that , even though identities are necessary , in many cases the actual values of identities is not crucial , and only their relative order matters . specifically , @xcite shows that for a particular class of problems , called lcl ( for _ locally checkable languages _ ) , if there exists a local algorithm that , for any identity assignment , constructs an instance of a problem in lcl in constant number of rounds , then there exists an _ _ order invariant _
_ algorithm for that problem that runs in the same number of rounds .
lcl restricts its concern to graphs with constant maximum degree , and to problems with a constant number of inputs .
the assumption on the size of the inputs of problems in lcl was shown necessary in @xcite , by exhibiting a natural problem that is locally checkable , has unbounded input size , can be solved in 1 round with identities , but can not be solved in constant time by any order invariant algorithm .
the role of identities can also be gauged by comparing their impact to that of `` orientation mechanisms '' .
for instance , gs et al .
@xcite have shown that for a large class of optimization problems , called po - checkable problems , local algorithms do not benefit from any kind of identifiers : if a po - checkable optimization problem can be approximated with a local algorithm , the same approximation factor can be achieved in anonymous networks if the network is provided with a port - numbering and an orientation .
the discussion above involved distributed _ construction _ tasks , including , e.g. , graph coloring @xcite , maximal independent set @xcite , and maximal matching @xcite .
when it comes to distributed _ decision _ tasks @xcite , symmetry breaking issues do not however seem to play a role .
informally , a decision task requires the nodes to `` collectively decide '' whether the given instance ( i.e. , a graph with inputs to the nodes ) satisfies some specific properties . for instance , deciding coloring requires , given a colored graph , to check whether this graph is properly colored .
the meaning of `` collectively decide '' is as follows . on a legal instance
, all nodes should output `` yes '' , and on an illegal one , at least one node should output `` no '' .
note that it is not really important whether this node is unique or not ; hence , this specification does not inherently require any symmetry breaking .
therefore , assuming that each node @xmath7 can obtain the ball @xmath10 , it makes sense that the assumption of having an identity assignment may not be crucial for achieving correct decision .
we tackle the question of whether identities play a role in decision problems in the framework of the aforementioned @xmath0 model @xcite , which is a standard distributed computing model capturing the essence of locality .
recall that , in this model , processors are nodes of a connected network @xmath11 , have pairwise distinct identities , and have inputs . more formally ,
a _ configuration _ is a triplet @xmath12 where @xmath13 is a connected graph , every node @xmath14 is assigned as its _ local input _ a binary string @xmath15 , and @xmath16 denotes the identity of node @xmath17 .
( in some problems , the local input of every node is empty , i.e. , @xmath18 for every @xmath14 , where @xmath19 denotes the empty binary string ) .
processors are woken up simultaneously , and computation proceeds over the input configuration @xmath12 in fault - free synchronous _ rounds _ during which every processor exchanges messages of unlimited size with its neighbors in the underlying network @xmath13 , and performs arbitrary individual computations on its data . in many cases , the running time of an algorithm
is measured with respect to the size @xmath20 of @xmath13 : the running time of an algorithm is defined as the maximum number of rounds it takes to terminate at all nodes , over all possible @xmath20-node networks . similarly to @xcite , we consider algorithms whose running time is independent of the size of the network , that is they run in constant time .
let @xmath10 be the ball centered at @xmath7 , of radius @xmath3 , excluding the edges between two nodes at distance exactly @xmath3 from @xmath7 . as mentioned before , without loss of generality ,
any algorithm running in time @xmath21 in the @xmath0 model consists of : 1 . collecting ( in @xmath3 rounds ) at every node @xmath7 the structure of the ball @xmath10 together with all the inputs @xmath22 and identities @xmath16 of these nodes , and , 2 .
performing some individual computation at every node .
we define the _ anonymous _
@xmath0 model similarly to the @xmath0 model , except that nodes have no identities .
more precisely , an input configuration in the anonymous @xmath0 model is just a pair @xmath23 .
an algorithm running in time @xmath21 in the anonymous @xmath0 model consists of : 1 .
getting at every node @xmath7 a snapshot of the structure of the ball @xmath10 together with all the inputs of the nodes in this ball , and , 2 . performing some individual computation at every node . note that the _ anonymous
_ @xmath0 model does not explicitly involve communications between nodes .
instead , it implicitly assumes that the underlying network supports the snapshot operation .
clearly , this model is not stronger than the @xmath0 model , and possibly even strictly weaker , since a node @xmath7 can no longer base its individual computation on the identities of the nodes in the ball @xmath10 .
one can think of various other `` anonymous '' models , i.e. , which do not involve node identities .
in particular , there is a large literature on distributed computing in networks without node identities , where symmetry breaking is enabled thanks to locally disjoint port numbers ( see , e.g. , @xcite ) .
we consider the anonymous @xmath0 model to isolate the role of node identities from other symmetry breaking mechanisms .
model is the strongest model among all models without node identities .
indeed , there are network problems that can be solved in the anonymous @xmath0 model which can not be solved in the aforementioned model that is based on locally disjoint port numbers .
a simple example is to locally detect the absence of a 3-node cycle .
] our aim is to compare the power of the anonymous @xmath0 model with the standard @xmath0 model in order to capture the impact of identities on local distributed decision .
recall from @xcite that a _ distributed language _ is a decidable collection @xmath24 of configurations .
( since an undecidable collection of configurations remains undecidable in the distributed setting too , we consider only decidable collections of configurations ) .
a typical example of a language is @xmath25 where @xmath26 denotes the ( open ) neighborhood of @xmath17 , that is , all nodes at distance exactly 1 from @xmath17 . still following the terminology from @xcite
, we say that a distributed algorithm @xmath27 _ decides _ a distributed language @xmath24 if and only if for every configuration @xmath23 , every node of @xmath13 eventually terminates and outputs `` yes '' or `` no '' , satisfying the following decision rules : * if @xmath28 , then each node outputs `` yes '' ; * if @xmath29 , then at least one node outputs `` no '' . in the ( non - anonymous ) @xmath0 model , these two rules must be satisfied for every identity assignment .
that is , all processes must output `` yes '' on a legal instance , independent of their identities . and ,
on an illegal instance , at least one node must output `` no '' , for every identity assignment .
note that this node may potentially differ according to the identity assignment .
some languages can be decided in constant time ( e.g. , @xmath30 ) , while others can easily be shown not to be decidable in constant time ( e.g. , `` is the network planar ? '' ) .
in contrast to the above examples , there are some languages whose status is unclear . to elaborate on this , consider the particular case where it is required to decide whether the network belongs to some specified family @xmath31 of graphs .
if this question can be decided in a constant number of communication rounds , then this means , informally , that the family @xmath31 can somehow be characterized by relatively simple conditions .
for example , a family @xmath31 of graphs that can be characterized as consisting of all graphs having no subgraph from @xmath32 , where @xmath32 is some specified finite set of graphs , is obviously decidable in constant time . however , the question of whether a family of graphs can be characterized as above is often non - trivial .
for example , characterizing cographs as precisely the graphs with no induced @xmath33 , attributed to seinsche @xcite , is not easy , and requires nontrivial usage of modular decomposition .
we are now ready to define one of our main subjects of interest , the classes @xmath1 and @xmath2 .
specifically , @xmath1 ( for _ local decision _ ) is the class of all distributed languages that can be decided by a distributed algorithm that runs in a constant number of rounds in the @xmath0 model @xcite .
similarly , @xmath2 , the anonymous version of @xmath1 , is the class of all distributed languages that can be decided by a distributed algorithm that runs in a constant number of rounds in the anonymous @xmath0 model . by definition ,
we conjecture that @xmath34 in this paper , we provide several evidences supporting this conjecture . in addition , we investigate the _ non - deterministic _ version of these classes , and prove that they coincide .
more specifically , a distributed _ verification _ algorithm is a distributed algorithm @xmath27 that gets as input , in addition to a configuration @xmath23 , a global _ certificate vector _ @xmath35 , i.e. , every node @xmath17 of a graph @xmath13 gets as input two binary strings , an input @xmath36 and a certificate @xmath37 . a verification algorithm @xmath27 verifies @xmath24 if and only if for every input configuration @xmath23 , the following hold : * if @xmath28 , then there exists a certificate @xmath35 such that every node outputs `` yes '' ; * if @xmath29 , then for every certificate @xmath35 , at least one node outputs `` no '' . again , in the ( non - anonymous ) @xmath0 model , these two rules must be satisfied for every identity assignment , but the certificates must be the same regardless of the identities .
we now recall the class @xmath38 , for _ non - deterministic local decision _ , as defined in @xcite : it is the class of all distributed languages that can be verified in a constant number of rounds in the @xmath0 model .
similarly , we define @xmath39 , the anonymous version of @xmath38 , as the class of all distributed languages that can be verified in a constant number of rounds in the anonymous @xmath0 model . by definition ,
@xmath40 . in this paper
, we give several evidences supporting the conjecture @xmath5 .
in particular , we show that it holds for languages defined on paths , with a finite set of input values .
more generally , we show that the conjecture holds for _ hereditary _ languages , that is , languages closed under node deletion . regarding arbitrary languages , and arbitrary graphs , we prove that the conjecture holds assuming that every node knows an upper bound on the total number of nodes in the input graph .
( this upper bound can be arbitrary , and may not be the same for all nodes ) .
moreover , we prove that equality between non - anonymous decision and anonymous decision holds in the context of _ non - deterministic _ local decision , where nodes are given certificates ( independent of the identities , if they exist ) , and the decision consists in verifying these certificates .
more precisely , we prove that @xmath6 .
this latter result is obtained by characterizing both @xmath38 and @xmath39 .
the question of how to locally decide ( or verify ) languages has received quite a lot of attention recently . inspired by classical computation complexity theory , it was suggested in @xcite that the study of decision problems may lead to new structural insights also in the more complex distributed computing setting . indeed , following that paper , which focused on the @xmath9 model
, efforts were made to form a fundamental computational complexity theory for distributed decision problems in various other aspects of distributed computing @xcite .
the classes ld , nld and bpld defined in @xcite are the distributed analogues of the classes p , np and bpp , respectively .
the paper provides structural results , developing a notion of local reduction and establishing completeness results .
one of the main results is the existence of a sharp threshold for randomization , above which randomization does not help ( at least for hereditary languages ) .
more precisely , the bpld classes were classified into two : below and above the randomization threshold . in @xcite ,
the authors show that the hereditary assumption can be lifted if we restrict our attention to languages on path topologies .
these two results from @xcite are used in the current paper in a rather surprising manner .
the authors in @xcite then `` zoom '' into the spectrum of classes below the randomization threshold , and defines a hierarchy of an infinite set of bpld classes , each of which is separated from the class above it in the hierarchy .
the precise knowledge of the number of nodes @xmath20 was shown in @xcite to be of large impact on non - deterministic decision .
indeed , with such a knowledge every language can be decided non - deterministically in the model of nld .
we note , however , that the knowledge of an arbitrary upper bound on @xmath20 ( as assumed here in one of our results ) seems to be a much weaker assumption , and , in particular , will not suffice for non - deterministically deciding all languages . in the context of construction problems ,
it was shown in @xcite that in many case , the knowledge of @xmath20 ( or an upper bound on @xmath20 ) is not essential .
the original theoretical basis for non - determinism in local computation was laid by the theory of _ proof - labeling schemes _ ( pls ) @xcite originally defined in @xcite . as mentioned , this notion resembles the notion of nld , but differs in the role identities play .
specifically , in pls the designer of the algorithm may base the certificates ( called labels in the terminology of pls ) construction on the given identity assignment .
in contrast , in the model of nld , the certificates must be the same regardless of the identities of nodes .
indeed , this difference is significant : while every language can be verified by a proof labeling scheme , not every language belongs to nld @xcite .
these notions also bear some similarities to the notions of _ local computation with advice _
@xcite , _ local detection _
@xcite , _ local checking _
@xcite , or _ silent stabilization _ @xcite .
in addition , as shown later on , the notion of nld is related also to the theory of _ lifts _ or _ covers _
@xcite .
finally , the classification of decision problems in distributed computing has been studied in several other models .
for example , @xcite and @xcite study specific decision problems in the @xmath41 model . in @xcite , the authors study mst verification in the pls sense but under the @xmath41 model of communication .
in addition , decision problems have been studied in the asynchrony discipline too , specifically in the framework of _ wait - free computation _ @xcite and _ mobile agents computing _ @xcite . in the wait - free model ,
the main issues are not spatial constraints but timing constraints ( asynchronism and faults ) .
the main focus of @xcite is deterministic protocols aiming at studying the power of the `` decoder '' , i.e. , the interpretation of the results . while this paper essentially considers the and - checker ( since a global `` yes '' corresponds to all processes saying `` yes '' ) , @xcite deals with other interpretations , including more values ( not only `` yes '' and `` no '' ) , with the objective of designing checkers that use the smallest number of values .
we conjecture that @xmath42 .
a support to this conjecture is that it holds for a large class of languages , namely for all _ hereditary _ languages , that is languages closed under node deletion . for instance , @xmath30 and @xmath43 are hereditary , as well as all languages corresponding to hereditary graph families , such as planar graphs , interval graphs , forests , chordal graphs , cographs , perfect graphs , etc .
[ theo : ldhereditary ] @xmath5 for hereditary languages . to prove the theorem , it is sufficient to show that @xmath44 for hereditary languages .
this immediately follows from the statement and proof of theorem 3.3 in @xcite .
indeed , let @xmath27 be a non - anonymous local algorithm deciding @xmath24 .
this deterministic algorithm is in particular a randomized algorithm , with success probabilities @xmath45 for legal instances , and @xmath46 for illegal instance .
that is , algorithm @xmath27 is a @xmath47-decider for @xmath24 , according to the definition in @xcite . since @xmath24 is hereditary , and since @xmath48 , the existence of @xmath27 implies the existence of a specific deterministic anonymous local algorithm @xmath49 for @xmath24 .
indeed , the algorithm @xmath49 described in the proof of theorem 3.3 in @xcite is in fact anonymous : it simply collects the ball @xmath10 of radius @xmath3 around each node @xmath7 for some constant @xmath3 , and @xmath7 then decides `` yes '' or `` no '' according to whether @xmath50 or not , regardless of the identities .
a similar proof , based on theorem 4.1 in @xcite , enables to establish the following : [ theo : ldpath ] @xmath5 for languages defined on the set of paths , with a finite set of input values .
another evidence supporting the conjecture @xmath42 is that it holds assuming that nodes have access to a seemingly weak oracle .
specifically , this oracle , denoted @xmath51 , simply provides each node with an arbitrarily large upper bound on the total number of nodes in the actual instance .
( it is not assumed that all the upper bounds provided to nodes are the same ) .
we denote by @xmath52 the class of languages that can be decided by an anonymous local algorithm having access to oracle @xmath53 , and we prove the following : [ theo : ld = ld ] @xmath54 .
we just need to prove that @xmath55 .
let @xmath56 , and let @xmath27 be a local ( non - anonymous ) algorithm deciding @xmath24 .
assume that the running time of @xmath27 is @xmath3 .
we transform @xmath27 into an anonymous algorithm @xmath57 deciding @xmath24 in time @xmath3 , assuming each node @xmath7 in a given input @xmath13 has an access to the oracle @xmath53 , i.e. , it knows an arbitrary upper bound @xmath58 on the number of nodes in @xmath13 .
algorithm @xmath57 works as follows .
each node @xmath7 collects the ball @xmath10 of radius @xmath3 around it .
then , for every possible assignment of identities to the nodes of @xmath10 taken from the range @xmath59 $ ] , node @xmath7 simulates the behavior of the non - anonymous algorithm @xmath27 on the ball @xmath10 with the corresponding identities .
if , in one of these simulations , algorithm @xmath27 decides `` no '' , then @xmath57 decides `` no '' .
otherwise , @xmath57 decides `` yes '' .
we now prove the correctness of @xmath57 .
if the input @xmath60 , then @xmath27 accepts it for every identity assignment to the nodes of @xmath13 .
therefore , since , for every node @xmath7 , every possible identity assignment to the nodes of the ball @xmath10 can be extended to an identity assignment to all the nodes of @xmath13 , all the simulations of @xmath27 by @xmath7 return `` yes '' , and hence @xmath57 accepts @xmath24 as well . on the other hand , if @xmath61 then @xmath27 rejects it for every identity assignment to the nodes of @xmath13 .
that is , for every identity assignment to the nodes of @xmath13 , at least one node @xmath7 decides `` no '' .
( note that , this node @xmath7 may be different for two different identity assignments ) .
let us fix one identity assignment @xmath62 to the nodes of @xmath13 , in the range @xmath63 $ ] , and let @xmath7 be one node that decides `` no '' for @xmath62 .
let @xmath64 be the ball @xmath10 with the identities of the nodes given by @xmath62 . in @xmath57
, since @xmath7 tries all possible identity assignments of the ball @xmath10 in the range @xmath59 $ ] with @xmath65 , in one of its simulations of @xmath27 , node @xmath7 will simulate @xmath27 on @xmath64 . in this simulation
, node @xmath7 decides `` no '' , and hence algorithm @xmath57 rejects @xmath24 as well .
note that the inclusion @xmath55 holds when one imposes no restrictions on the individual sequential running time .
however , the transformation of a ( non - anonymous ) local algorithm into an anonymous local algorithm as described in the proof of theorem [ theo : ld = ld ] is very expensive in terms of individual computation .
indeed , the number of simulations of the original local algorithm @xmath27 by each node @xmath7 can be as large as @xmath66 where @xmath58 is the upper bound on @xmath20 given by the oracle @xmath53 , and @xmath67 is the number of nodes in the ball @xmath10 .
this bound can be exponential in @xmath20 even if the oracle provides a good approximation of @xmath20 ( even if it gives precisely @xmath20 ) .
it would be nice to establish @xmath55 by using a transformation not involving a huge increase in the individual sequential computation time .
in the previous section , we have seen several evidences supporting the conjecture that @xmath5 , but whether it holds or not remains to be proved . in this section
, we turn our attention to the non - deterministic variants of these two classes , and show that they coincide .
more formally , we have : [ theo : nld = nld ] @xmath6 . to prove @xmath6 , it is sufficient to prove @xmath68 . to establish this inclusion , we provide a sufficient condition for @xmath39-membership , and prove that it is a necessary condition for @xmath38-membership .
let @xmath69 and @xmath70 be two input instances .
a _ homomorphism _ from @xmath71 to @xmath72 is a function @xmath73 that preserves the edges of @xmath13 as well as the inputs to the nodes .
specifically , @xmath74 and @xmath75 maps every node @xmath76 to a node @xmath77 satisfying @xmath78 for instance , assuming the nodes have no inputs , and labeling the nodes of the @xmath20-node cycle @xmath79 by consecutive integers from 0 to @xmath80 , modulo @xmath20 , then the map @xmath81 defined by @xmath82 is a homomorphism . the trivial map @xmath83 defined by @xmath84 , where @xmath85 is the 2-node clique , is also a homomorphism . to establish conditions for @xmath38- and @xmath39-membership , we require the involved homomorphisms to preserve the local neighborhood of a node , and define the notion of _
@xmath3-local isomorphism_. let @xmath3 be a positive integer .
we say that @xmath71 is @xmath3-local isomorphic to @xmath72 if and only if there exists an homomorphism @xmath75 from @xmath71 to @xmath72 such that , for every node @xmath14 , @xmath75 restricted to @xmath86 is an isomorphism from @xmath86 to @xmath87 .
we call such a homomorphism @xmath75 a @xmath3-local isomorphism .
note that a homomorphism is not necessarily a 1-local isomorphism .
for instance , the aforementioned map @xmath81 defined by @xmath82 is a 1-local isomorphism , but the map @xmath83 defined by @xmath84 is not a 1-local isomorphism . to be a 1-local isomorphism
, a homomorphism should also insure isomorphism between the balls of radius 1 .
also observe that any @xmath3-local isomorphism @xmath88 is onto ( because if a node of @xmath89 has no pre - image , then neither do its neighbors have a pre - image , since homomorphisms preserve edges , and so forth ) . to avoid confusion , it is thus useful to keep in mind that , informally , a @xmath3-local isomorphism goes from a `` larger '' graph to a `` smaller '' graph .
[ def : lift ] for positive integer @xmath3 , we say that @xmath24 is _ @xmath3-closed under lift _ if , for every two instances @xmath90 such that @xmath71 is @xmath3-local isomorphic to @xmath72 , we have : @xmath91 so , informally , defintion [ def : lift ] states that , for a language @xmath24 to be @xmath3-closed under lift , if a `` smaller '' instance @xmath72 is in @xmath24 then any `` larger '' instance @xmath71 that is a lift of @xmath72 , i.e. , satisfying that @xmath71 is @xmath3-local isomorphic to @xmath72 , must also be in @xmath24 .
the following lemma gives a sufficient condition for @xmath39-membership .
[ claim : sufficient ] let @xmath24 be a language . if there exists @xmath92 such that @xmath24 is @xmath3-closed under lift , then @xmath93 .
let @xmath24 be a language , and assume that there exists @xmath92 such that @xmath24 is @xmath3-closed under lift .
we describe an anonymous non - deterministic local algorithm @xmath27 deciding @xmath24 , and performing in @xmath3 rounds .
the certificate of each node @xmath17 is a triple @xmath94 where @xmath89 is an @xmath20-node graph with nodes labeled by distinct integers in @xmath63=\{1,\dots , n\}$ ] , @xmath95 $ ] , and @xmath96 is an @xmath20-dimensional vector .
informally , the certificates are interpreted by @xmath27 as follows .
the graph @xmath89 is supposed to be a `` map '' of @xmath13 , that is , @xmath89 is interpreted as an isomorphic copy of @xmath13 .
the integer @xmath97 is the label of the node in @xmath89 corresponding to node @xmath17 in @xmath13 .
finally , @xmath96 is interpreted as the input of the nodes in @xmath89 .
the algorithm @xmath27 performs as follows .
every node @xmath17 gets @xmath86 , the ball of radius @xmath3 around it ; hence , in particular , it collects all the certificates of all the nodes at distance at most @xmath3 from it .
then , by comparing its own certificate with the ones of its neighbors , it checks that the graph @xmath89 , and the input @xmath96 in its certificate , are identical to the ones in the certificates of its neighbors .
it also verifies consistency between the labels and the nodes in its ball of radius @xmath3 .
that is , it checks whether the labels and inputs in the certificate of the nodes in @xmath86 are as described by its certificate . whenever a node fails to pass any of these tests , it outputs `` no '' .
otherwise it output `` yes '' or `` no '' according to whether @xmath98 or not , respectively .
( this is doable because we are considering languages that are decidable in the usual sense of sequential computation ) .
we show that @xmath27 performs correctly . if @xmath60 , then by labeling the nodes in @xmath13 by distinct integers from 1 to @xmath99 , and by providing the node @xmath17 labeled @xmath97 with @xmath100 , the algorithm @xmath27 output `` yes '' at all nodes , as desired .
consider now a instance @xmath101 .
assume , for the purpose of contradiction that there exists a certificate @xmath35 leading all nodes to output `` yes '' .
let @xmath102 be defined by @xmath103 where @xmath97 is the label of @xmath17 in its certificate .
since @xmath35 passes all tests of @xmath27 , it means that ( 1 ) @xmath94 where the instance @xmath70 is the same for all nodes , ( 2 ) @xmath75 restricted to @xmath86 is an isomorphism from @xmath86 to @xmath87 , for every node @xmath17 , and ( 3 ) @xmath98 .
in view of ( 2 ) , @xmath71 is @xmath3-local isomorphic to @xmath72 .
therefore , ( 3 ) implies that @xmath104 , because @xmath24 is @xmath3-closed under lift .
this is in contradiction with the actual hypothesis @xmath61 .
thus , for each certificate @xmath35 , there must exist at least one node that outputs `` no '' . as a consequence
, @xmath27 is a non - deterministic algorithm deciding @xmath24 , and thus @xmath93 .
the following lemma shows that the aforementioned sufficient condition for @xmath39-membership is a necessary condition for @xmath38-membership .
[ claim : necessary ] let @xmath24 be a language . if @xmath105 , then there exists @xmath92 such that @xmath24 is @xmath3-closed under lift .
let @xmath24 be a language in @xmath38 , and let @xmath27 be a non - deterministic ( non - anonymous ) local algorithm deciding @xmath24 .
assume , for the purpose of contradiction that , for any integer @xmath92 , @xmath24 is _ not _ @xmath3-closed under lift .
that is , for any @xmath3 , there exist two input instances @xmath90 such that @xmath71 is @xmath3-local isomorphic to @xmath72 , with @xmath106 and @xmath107 .
assume that @xmath27 runs in @xmath3 rounds .
without loss of generality , we can assume that @xmath108 .
let @xmath101 and @xmath109 satisfying @xmath71 is @xmath3-local isomorphic to @xmath72 .
since @xmath107 , there exists a certificate @xmath110 such that when @xmath27 is running on @xmath72 with certificate @xmath110 , every node output `` yes '' for every identity assignment .
since @xmath71 is @xmath3-local isomorphic to @xmath72 , there exists an homomorphism @xmath111 such that , for every node @xmath112 , @xmath75 restricted to @xmath86 is an isomorphism from @xmath86 to @xmath87 .
let @xmath35 be the certificate for @xmath71 defined by @xmath113 .
consider the execution of @xmath27 running on @xmath71 with certificate @xmath35 , and some arbitrary identity assignment @xmath62 .
since @xmath27 performs in @xmath3 rounds , the decision at each node @xmath17 is taken according to the inputs , certificates , and identities in the ball @xmath86 , as well as the structure of this ball . by the nature of the homomorphism @xmath75 , and by the definition of certificate @xmath35 , the structure , inputs and certificates of the ball @xmath86 , are identical to the corresponding structure , inputs and certificates of the ball @xmath87 .
balls may however differ in the identities of their nodes .
so , let @xmath114 be the node in @xmath13 deciding `` no '' for @xmath23 with certificate @xmath35 .
there exists such a node since @xmath115 .
let @xmath116 , and assign the same identities to the nodes in @xmath117 as their corresponding nodes in @xmath118 .
arbitrarily extend this identities to an identity assignment @xmath119 to the whole graph @xmath89 . by doing so , the two balls are not only isomorphic , but every node in @xmath118 has the same input , certificate and identity as its image in @xmath117 .
therefore , the decision taken by @xmath27 at @xmath120 under @xmath62 is the same as its decision at @xmath121 under @xmath119 .
this is in contradiction to the fact that @xmath114 decides `` no '' while @xmath122 decides `` yes '' .
lemmas [ claim : sufficient ] and [ claim : necessary ] together establish the theorem .
the proof of lemma [ claim : sufficient ] also provides an upper bound on the size of the certificates for _ graph languages _ in @xmath38 , that is , for languages in @xmath38 with no input .
( this includes , e.g. , recognition of interval graphs , and recognition of chordal graphs ) .
indeed , given @xmath105 , algorithm @xmath27 in the proof of lemma [ claim : sufficient ] verifies @xmath24 using a certificate at each node which is essentially an isomorphic copy of the input instance @xmath23 , with nodes labeled by consecutive integers in @xmath63 $ ] . if @xmath24 is a graph language , then there is no input @xmath123 , and thus the size of the certificates depends only on the size of the graph . more precisely , we have : [ cor : mlogn ] let @xmath105 be a graph language .
there exists an algorithm verifying @xmath24 using certificates of size @xmath124 bits at each node of every @xmath20-node graph in @xmath24 .
we now argue that the above bound is tight , that is , we prove the following .
there exists a graph language @xmath105 such that every algorithm verifying @xmath24 requires certificates of size @xmath125 bits .
recall that @xcite showed that there exists a graph language for which every proof labeling scheme ( pls ) requires labels of size @xmath125 bits ( the proof of this latter result appears in a detailed version @xcite ) . still in the context of pls
, @xcite showed that this lower bound holds for two _ natural _ graph families : specifically , @xcite showed that verifying symmetric graphs requires labels of size @xmath125 bits , and verifying non-3 colorable graphs requires almost the same size of labels , specifically , @xmath126 bits .
note that the certificate size required for verifying a language in @xmath38 is at least as large as the minimum label size required for verifying the language via a proof labeling scheme .
unfortunately , however , one can not obtain our claim directly from the aforementioned results since it turns out that neither of the two graph languages ( namely , symmetric graphs and non-3 colorable graphs ) belongs to @xmath38 .
we therefore employ an indirect approach . specifically , consider a graph @xmath13
we say that @xmath127 is a _ seed _ of @xmath13 if there exists a 1-local isomorphism from @xmath13 to @xmath127 .
suppose @xmath31 is a family of graphs .
let denote the family of graphs @xmath13 , for which there exists a seed of @xmath13 that belongs to @xmath31 .
then , by definition , is @xmath128-closed under lift . indeed
, assume that there is a 1-local isomorphism @xmath129 from @xmath89 to @xmath13 , and let @xmath130 be a seed of @xmath13 that belongs to @xmath31 .
then let @xmath75 be the 1-local isomorphism from @xmath13 to @xmath127 .
we have that @xmath131 is a 1-local isomorphism from @xmath89 to @xmath127 , because , for every @xmath132 , @xmath133 is isomorphic to @xmath134 , which in turn is isomorphic to @xmath135 .
thus @xmath127 is also a seed of @xmath89 .
is therefore in @xmath38 .
now , in the proof of corollary 2.2 in @xcite , the authors construct , for every integer @xmath20 , a family @xmath136 of @xmath20-node graphs that requires proof labels of size @xmath125 .
note that for every prime integer @xmath137 , a graph @xmath13 of size @xmath137 belongs to @xmath138 if and only if it belongs to .
therefore , there exists a graph language , namely , , that requires certificates of size @xmath125 bits ( at least for prime @xmath20 s ) .
again , in this paper , we provide some evidences supporting the conjecture @xmath5 . for instance , theorem [ theo : ld = ld ] shows that if every node knows any upper bound on the number of nodes @xmath20 , then all languages in @xmath1 can be decided in the anonymous @xmath0 model .
one interesting remark about the @xmath0 model is that it is guaranteed that at least one node has an upper bound on @xmath20 .
this is for instance the case of the node with the largest identity . in the anonymous @xmath0 model
, however , there is no such guarantee . finding a language
whose decision would be based on the fact that one node has an upper bound on @xmath20 would disprove the conjecture @xmath5 .
nevertheless , it is not clear whether such a problem exists . in this paper
, we also prove that @xmath6 , that is , our conjecture holds for the non - deterministic setting .
it is worth noticing that @xcite proved that there exists an @xmath38-complete problem under the local one - to - many reduction .
it is not clear whether such a problem exists for @xmath39 .
indeed , the reduction in the completeness proof of @xcite relies on the aforementioned guarantee that , in the @xmath0 model , at least one node has an upper bound on @xmath20 .
a. das sarma , s. holzer , l. kor , a. korman , d. nanongkai , g. pandurangan , d. peleg and r. wattenhofer .
distributed verification and hardness of distributed approximation . in _ proc .
43rd acm symp . on theory of computing ( stoc ) _ , 2011 .
p. fraigniaud , c. gavoille , d. ilcinkas and a. pelc .
distributed computing with advice : information sensitivity of graph coloring . in _ proc .
34th colloq . on automata , languages and programming ( icalp ) _ , 231242 , 2007 .
e. fusco and a. pelc .
communication complexity of consensus in anonymous message passing systems . in _ proc .
15th international conference on principles of distributed systems ( opodis 2011 ) _ , lncs 7109 , 191206 .
h. hasemann , j. hirvonen , j. rybicki , and j. suomela . deterministic local algorithms , unique identifiers , and fractional graph colouring . in _ proc .
19th international colloquium on structural information and communication complexity ( sirocco ) _ , 2012 .
a. korman , s. kutten , and t. masuzawa .
fast and compact self - stabilizing verification , computation , and fault detection of an mst . in _ proc .
30th acm symp . on principles of distributed computing ( podc ) _ , 2011 .
a. korman , j.s .
sereni , and l. viennot . toward more localized local algorithms : removing assumptions concerning global knowledge . in _ proc .
30th acm symp . on principles of distributed computing ( podc ) _ , 4958 , 2011 . | the issue of identifiers is crucial in distributed computing .
informally , identities are used for tackling two of the fundamental difficulties that are inherent to deterministic distributed computing , namely : ( 1 ) _ symmetry breaking _ , and ( 2 ) _ topological information gathering_. in the context of _ local computation _ , i.e. , when nodes can gather information only from nodes at bounded distances , some insight regarding the role of identities has been established .
for instance , it was shown that , for large classes of _ construction _ problems , the role of the identities can be rather small . however , for the identities to play no role , some other kinds of mechanisms for breaking symmetry must be employed , such as edge - labeling or sense of direction . when it comes to
local distributed _ decision _ problems , the specification of the decision task does not seem to involve symmetry breaking .
therefore , it is expected that , assuming nodes can gather sufficient information about their neighborhood , one could get rid of the identities , without employing extra mechanisms for breaking symmetry .
we tackle this question in the framework of the @xmath0 model .
let @xmath1 be the class of all problems that can be _ decided _ in a constant number of rounds in the @xmath0 model .
similarly , let @xmath2 be the class of all problems that can be decided at constant cost in the anonymous variant of the @xmath0 model , in which nodes have no identities , but each node can get access to the ( anonymous ) ball of radius @xmath3 around it , for any @xmath3 , at a cost of @xmath3 .
it is clear that @xmath4 .
we conjecture that @xmath5 . in this paper , we give several evidences supporting this conjecture .
in particular , we show that it holds for _ hereditary _ problems , as well as when the nodes know an arbitrary upper bound on the total number of nodes .
moreover , we prove that the conjecture holds in the context of _ non - deterministic _ local decision , where nodes are given certificates ( independent of the identities , if they exist ) , and the decision consists in verifying these certificates . in short , we prove that @xmath6 . *
keywords : * distributed complexity ; locality ; identities ; decision problems ; symmetry breaking ; non - determinism . |
the bethe - heitler ( bh ) reaction @xcite is a basic process for the production of real photons .
when the real photons are produced from the interaction of an incoming electron with the coulomb field of an atom or a nucleus , it corresponds to the bremsstrahlung reaction which is commonly used to generate linearly and circularly polarized photon beams of various energies @xcite . when the interaction occurs with the nuclear electromagnetic field , the bethe - heitler process carries information about the internal structure of the target nucleus represented by its electromagnetic form factors .
while the inclusive process is a radiative correction to the lepton scattering cross section @xcite , the exclusive process is an essential tool for the experimental determination of generalized nucleon polarizabilities @xcite and generalized parton distributions ( gpd ) @xcite . in the effort to improve our understanding of the nuclear structure @xcite and to unravel the parton structure of nucleons and nuclei @xcite via the ( deeply ) virtual compton scattering process ( ( d)vcs ) @xcite
, the bh reaction appears as a powerful contamination of the global process of the lepto - production of real photons .
( d)vcs corresponds to the absorption of a virtual photon by the nucleus ( parton ) followed by the emission of a real photon from the excited nucleus ( parton ) .
involving the same initial and final states the bh and ( d)vcs processes interfere coherently . in the multi - gev energy range relevant for these studies ,
the cross section for the lepto - production of real photons contains a strong , often dominant , contribution of the bh amplitude .
the nuclear structure information of interest is extracted @xcite from the unpolarized lepton cross section as a deviation from the bh contribution that should consequently be accurately subtracted . considering the cross section difference for incident polarized leptons of opposite helicities or opposite charges
, the bh process serves as a magnifier of the ( d)vcs signal , the precise knowledge of its magnitude being still required to access the nuclear structure information .
specifically , the bh process is the reference reaction of the lepto - production of real photons , and any attempt to access the nuclear structure via this reaction requires a very accurate knowledge of this reference process . in most of the available calculations of this process ( see @xcite for a review )
, the mass of the incoming and scattered leptons is neglected , at variance with the crossed process ( lepton pair production ) where a few finite lepton mass calculations exist .
indeed , the ultra - relativistic approximation seems well - justified in the energy range spanned by electron scattering experiments at mami @xcite , jlab and desy @xcite , and still justified for muon scattering experiments at cern @xcite .
however , a recent work on the bremsstrahlung and pair - creation processes may motivate a closer look at this approximation .
revisiting cross sections and polarization observables for these processes , this work @xcite showed that finite electron mass effects may result in strong differences as compared to previous massless calculations @xcite . in the bremsstrahlung process
, finite electron mass effects are shown to be significant in the end - point region of the cross section , and regularize polarization observables in that kinematical domain .
the differences are even more striking for the pair - creation process , the reciprocal of bremsstrahlung , where the finite electron mass calculation allows for consistent polarization observables at energies as low as the reaction threshold . if it is intuitive that finite electron mass calculations should be more accurate at low incoming energy , it is less obvious to expect persistent strong differences as the energy increases .
it is the merit of these calculations to show that additional terms involving the electron mass may significantly affect experimental observables even at relativistic energies .
the purpose of the present work is to investigate the effects of finite lepton mass on the bh process in the perspective of the on - going and future vcs and dvcs experimental programs .
the next section describes the formalism for the unpolarized bh process off the nucleon and discusses the comparison with a previous massless calculation .
the following sections focus on the specific features of the calculation here - below developed .
within the born approximation , the bethe - heitler elastic process @xmath1 is described by the two diagrams in fig .
[ fig : fd ] . in the laboratory frame
the four - momenta indicated in the figure are @xmath2 where @xmath3 and @xmath4 are the nucleon and lepton mass , respectively , @xmath5 is the equivalent virtual photon four - momentum corresponding to the true virtual photon of the ( d)vcs process , and @xmath6 .
the four - momentum transfer to the nucleon writes @xmath7 the corresponding elementary cross section can be expressed as and @xmath8 . ]
@xmath9 integrating over the final nucleon and the real photon energy , the differential cross section with respect to the final lepton three - momentum @xmath10 and to the outgoing photon solid angle @xmath11 reads @xmath12 the @xmath13-matrix for the bethe - heitler process is expressed in terms of the leptonic and hadronic tensors according to @xmath14 where @xmath15 the sum over spin states stands for unpolarized electron scattering off unpolarized nucleon target .
the electromagnetic nucleon current matrix element is @xmath16 u(p)~,\ ] ] where @xmath17 and @xmath18 are the dirac and pauli electromagnetic form factors of the nucleon .
@xmath19 is the real photon polarization vector and @xmath20 u(k)\ ] ] is the lepton current written in terms of the lepton propagator @xmath21 the sum over the photon polarization components reads @xmath22 where the gauge term does not contribute to the scattering amplitude because of current conservation ( @xmath23 ) .
introducing @xmath24 and performing the spin traces , the leptonic tensor can be recast in the following form @xmath25 where @xmath26 ^ 2}{p_1 p_2}~ , & b = 1+\mu^2 \ , \frac{p_1}{p_2}~ , \\ & c = 1+\mu^2 \ , \frac{p_2}{p_1}~,\ \ \ & d = -k^{\prime}_\delta + 2\mu^2 \ , \frac{k^{\prime}_\delta + k_\delta}{p_2}~ , \\ & e = -k_\delta+2\mu^2 \ , \frac{k^{\prime}_\delta + k_\delta}{p_1}~ , & f = -g = -2\mu^2~ , \end{array}\ ] ] being @xmath27 , @xmath28 , and @xmath29 . the electromagnetic hadronic tensor can be written as @xcite @xmath30 where @xmath31 and @xmath32 having introduced @xmath33 and the sachs electric and magnetic nucleon form factors @xcite . by contracting the reduced leptonic tensor
( eq . [ eq : lmunu ] ) and the hadronic tensor ( eq . [ eq : wmunu ] ) we get @xmath34 \nonumber\end{aligned}\ ] ] being @xmath35 and @xmath36 .
note that @xmath37 , @xmath38 and @xmath39 do not contribute . in the @xmath40 limit , @xmath41 , @xmath42 , @xmath43 , and the above expression simplifies to @xmath44 - w_2 \ , \left [ k^{\prime 2}_v + k^2_v \right]~.\ ] ]
different contributions can be distinguished in the tensor contraction of eq .
[ eq : lwm ] , following the relation @xmath45 expressing a polynomial dependence in the squared reduced lepton mass @xmath46 , in addition to the intrinsic mass dependence built into kinematics .
the polynomial coefficients write @xmath47 with @xmath48 @xmath49 is formally identical to the massless tensor contraction ( eq . [ eq : nomass ] ) and differs from it only by the finite lepton mass kinematics . @xmath50 and @xmath51 are additional contributions to the cross section attached to the second and fourth power of the reduced lepton mass , respectively , and originating from the finite mass leptonic tensor ( eq . [ eq : lmunu ] ) .
finally , the bethe - heitler differential cross section with respect to the scattered lepton momentum and the real photon solid angle , for unpolarized incoming lepton and initial target , writes @xmath52 where @xmath53 is the fine structure constant .
( color online ) bh differential cross section off proton at kinematics relevant of the jlab and cern experimental program ( @xmath54 is the incident lepton energy and @xmath55 the bjorken variable ) .
the full ( green ) line corresponds to the massless lepton calculation of ref .
the numerical results of the present work are shown for different lepton mass : dotted ( red ) line for massless leptons , dashed ( black ) line for electrons , and dash - dotted ( blue ) line for muons.,title="fig : " ] ( color online ) bh differential cross section off proton at kinematics relevant of the jlab and cern experimental program ( @xmath54 is the incident lepton energy and @xmath55 the bjorken variable ) .
the full ( green ) line corresponds to the massless lepton calculation of ref .
the numerical results of the present work are shown for different lepton mass : dotted ( red ) line for massless leptons , dashed ( black ) line for electrons , and dash - dotted ( blue ) line for muons.,title="fig : " ] fig .
[ fg : bh - mass - effect ] shows the results of the present calculations for typical jlab and compass kinematics , using the galster parametrization for the proton electromagnetic form factors @xcite .
the particle angles and momenta are referenced in the laboratory frame of the equivalent virtual photon ( @xmath56 ) with @xmath57-axis parallel to @xmath58 ; in that frame @xmath59 is the out - of plane angle between the leptonic plane @xmath60 and the hadronic plane @xmath61 .
the results are displayed as function of the polar angle @xmath62 of real photons for in - plane kinematics ( @xmath63 ) .
the massless lepton result of eq .
[ eq : nomass ] ( dotted line ) is first compared with the previous massless calculation of ref .
@xcite ( full line ) : both approaches are in this case strictly identical , yielding the same numerical values .
it is then compared with the finite mass calculations for electron- ( dashed line ) and muon- ( dash - dotted line ) induced bh .
a first region is identified and singled - out in the upper left corner of the graphs , corresponding to the quasi - singularities of the intermediate leptons propagators where strong differences between massless and finite mass results are observed , particularly for muons . on the right panel , a second region is identified at lower @xmath0 , corresponding to the kinematical threshold ( minimum absolute momentum transfer ) where significant finite mass effects are also observed .
note that the existence of these regions is in agreement with earlier observations of l. mo and y. tsai who used similar arguments to derive specific peaking approximations for the calculation of radiative corrections to elastic and inelastic lepton scattering @xcite .
the next sections detail the mechanisms at play in these two specific regions .
the intermediate lepton propagators entering the bh cross section are responsible for the local singular behaviour , observed on fig .
[ fg : bh - mass - effect ] , through the denominators ( @xmath64 , @xmath65 ) .
simple algebra shows that the condition @xmath66 is equivalent to @xmath67 ( similarly , @xmath68 is equivalent to @xmath69 ) , leading to @xmath70 this condition is obviously never satisfied for finite mass leptons and is verified only by massless leptons emitting a real photon in the forward direction . at fixed lepton kinematics , the @xmath71-location of these in - plane quasi - singularities are @xmath72 for @xmath64 , and @xmath73 for @xmath65 . at these momentum transfers ,
massless cross sections are infinite while finite mass cross sections yield finite values whose magnitudes depend on the lepton mass .
this singular behaviour of the cross section was already noticed in ref .
considering finite lepton mass in the kinematics regularizes the cross section but does not allow for a precise determination of the cross section in the quasi - singularity regions , since it ignores the extra terms associated to the full leptonic tensor .
evolution of the ratio @xmath74 for electron ( left ) and muon ( right ) scattering off proton calculated at the scattered lepton quasi - singualirity @xmath75 for different initial energy and momentum transfer of the equivalent virtual photon .
the additional full line on the left panel ( upper curve ) indicates the muon ratio at @xmath76 gev@xmath77/@xmath78 ; other @xmath79 can not be distinguished from the electron ratio.,title="fig : " ] evolution of the ratio @xmath74 for electron ( left ) and muon ( right ) scattering off proton calculated at the scattered lepton quasi - singualirity @xmath75 for different initial energy and momentum transfer of the equivalent virtual photon .
the additional full line on the left panel ( upper curve ) indicates the muon ratio at @xmath76 gev@xmath77/@xmath78 ; other @xmath79 can not be distinguished from the electron ratio.,title="fig : " ] the importance of extra contributions to the leptonic tensor specific of the finite lepton mass formalism can be expressed by the ratio @xmath80 identical to the corresponding differential cross section ratio .
[ bh - pole ] shows the @xmath0 evolution of this ratio calculated for typical jlab and compass lepton kinematics at the quasi - singularity @xmath75 corresponding to emission of real photons in the direction of scattered leptons .
globally , the extra terms @xmath50 and @xmath51 are significantly contributing to the cross section at the quasi - singularity .
notably , @xmath74 exhibits a saturation behaviour more or less rapidly reached depending on kinematics , the saturation value being roughly independent of the lepton mass .
[ eq : f0]-[eq : gam1 ] help understand these features . at the quasi - singularity @xmath75 , @xmath64 is quasi - null and extra contributions can be reduced to their dominant contribution in @xmath81 @xmath82 \\ { \cal f}_4 & \simeq & - \frac{p_2}{p_1 } \ , w_1 \ .\end{aligned}\ ] ] moreover , from eq .
[ eq : prop1 ] and eq .
[ eq : prop2 ] one obtains at @xmath75 @xmath83 such that the @xmath84 dependence of @xmath85 leads to a lepton mass independent ratio @xmath74 up to @xmath86 order . assuming further @xmath87 , leading to @xmath88 and @xmath89
, we obtain @xmath90 in the approximation of the dominance of the magnetic form factor contribution to the cross section , the electric contribution to @xmath91 in eq .
[ eq : w2 ] can be neglected and eq .
[ eq : ratmu ] becomes @xmath92 this expression shows that @xmath74 reaches a maximum value about @xmath93 at small @xmath94 and reproduces the @xmath95-dependence observed on the left panel of fig .
[ bh - pole ] at @xmath87 .
a minimum value about @xmath96 is also reproduced at the maximum kinematically allowed @xmath95 , consistently with exact calculations .
similar arguments can also be developed for the quasi - singularity at @xmath97 corresponding to emission of real photons in the lepton beam direction . to conclude this section , it is worth stressing that an accurate evaluation of the bh cross section in the quasi - singularity regions requires the higher order terms in @xmath84 specific of the finite lepton mass formalism , independently of the kinematics of the reaction and most notably even in the ultra - relativistic case .
significant finite lepton mass effects are also identified at minimum momentum transfer @xmath98 ( right panel of fig .
[ fg : bh - mass - effect ] ) , a region of particular interest for the experimental determination of the ji sum rule @xcite leading to the measurement of parton contribution to the total angular momentum of the nucleon . at this momentum transfer ,
real photons are emitted in the direction of the equivalent virtual photon with an energy @xmath99 where @xmath100 fig .
[ bh - ml ] shows that the bh differential cross section in the @xmath98 region is strongly sensitive to the lepton mass , several orders of magnitude distinguishing massless and finite mass calculations .
these effects originate again from the extra terms of the leptonic tensor .
region at fixed kinematics and for different incident lepton masses . ] the ratio @xmath74 ( eq . [ eq : ratmu ] ) characterizing the deviation of the finite mass with respect to the massless leptonic tensor is shown on fig .
[ bh - mass - tmi ] for different equivalent virtual photon squared four - momentum at the kinematical point @xmath98 , for an incident beam energy of 190 gev .
the muon ratio for a proton target ( upper curves of the left panel ) shows very large effects in the small @xmath0 region .
the electron ratio ( lower curves of the left panel ) exhibits the same behaviour and scales with the muon ratio as the squared lepton mass ratio .
most notably , @xmath74 appears almost independent of @xmath95 except in the region of the kinematical boundaries of the phase space .
additionally , the initial beam energy has little influence on this behaviour and essentially acts as a phase space magnifier : higher beam energies open access to smaller @xmath0 where larger ratios are obtained .
it is interesting to note that extra contributions effects appear negligible for a neutron target ( right panel of fig .
[ bh - mass - tmi ] ) .
evolution of the ratio @xmath74 calculated for different @xmath95 at @xmath98 and an incident beam energy of 190 gev of muons ( upper curves ) and electrons ( lower curves ) : proton ( left panel ) , neutron ( right panel).,title="fig : " ] evolution of the ratio @xmath74 calculated for different @xmath95 at @xmath98 and an incident beam energy of 190 gev of muons ( upper curves ) and electrons ( lower curves ) : proton ( left panel ) , neutron ( right panel).,title="fig : " ] these features can be understood by considering a simplified expression of the leptonic tensor . introducing the physics scale parameter @xmath101 , the dimensionless lepton energy @xmath102 , and @xmath103
, we have @xmath104 \\ p_1 & = & - \ , ( 1 - x_b ) \ , \left [ \frac{1}{\sqrt{1 + 4z } } + \frac{\zeta}{x_b } \left ( 1 + \frac{1}{\sqrt{1 + 4z } } \right ) \right ] \\
p_2 & = & - \ , ( 1 - x_b ) \ , \left [ \frac{1}{\sqrt{1 + 4z } } + \left ( \frac{1}{2z } - \frac{\zeta}{x_b } \right ) \left ( 1 + \frac{1}{\sqrt{1 + 4z } } \right ) \right ] \ , .\end{aligned}\ ] ] in the limit @xmath105 , equivalent to @xmath106 , the lepton propagators reduce to @xmath107 replacing these expressions in eqs .
[ eq : al1]-[eq : gam1 ] , one obtains at leading order in @xmath108 @xmath109 \ , \zeta^2 \\
\beta_1 & \simeq & -6 \ , { \left ( \frac{x_b-1}{x_b } \right)}^2 \ , \zeta^2 \\ \beta_2 & \simeq & 6 \ , \left [ \frac{x_b-1}{x_b } + \tau_{min } { \left ( \frac{x_b-1}{x_b } \right)}^2 + \frac{1}{4\tau_{min } } \right ] \ , \zeta^2 \\ \gamma_1 & \simeq & 6 \ , .\end{aligned}\ ] ] it is readily seen from these relations that the leptonic tensor ratio ( eq . [ eq : ratmu ] ) is approximately independent of the initial beam energy and of @xmath95-dependence .
the additional approximation @xmath110 , valid in the kinematical region @xmath111 , leads to @xmath112 the leptonic tensor ratio then writes @xmath113 } w_2}{w_1 } \simeq 6 \ , \frac{m^2}{m^2 } \ , \frac{(1-x_b)^2}{x_b^4 } \ , \frac{g_e^2}{g_m^2 } \ , .
\label{eq : ratfin}\ ] ] this approximate expression shows the strong @xmath0-dependence of the ratio responsible for the large effects observed at small @xmath0 , and provides an explanation for its sensivity to the nucleon isospin ( fig .
[ bh - mass - tmi ] ) .
it also shows that mass effects become sizeable ( @xmath114 ) for the proton when @xmath115 for muon scattering and @xmath116 for electron scattering , in agreement with the exact results of fig .
[ bh - mass - tmi ] . it may be somehow surprising that finite lepton mass effects appear sensitive to the nucleon target isospin .
it is indeed quite unconventional that finite mass effects are , as in the present case , linked to the electromagnetic structure of the nucleon .
the difference between the bh cross section for protons and neutrons expresses not only in the magnitude of the massless lepton cross section but also in the sensitivity to mass corrective terms , as seen from the ratio dependence on the electric form factor ( eq . [ eq : ratfin ] ) which vanishes at small momentum transfer for the neutron . summarizing , extra contributions to the leptonic tensor
have been shown to dominate the bh differential cross section off protons in the minimum momentum transfer region .
this effect is definitely important for muon scattering and becomes significant for electron scattering when the beam energy is high enough to allow exploring the region of @xmath0 below a few @xmath117 , that is for example , the kinematical range of the next generation of electron - ion colliders @xcite .
this work presents a full calculation of the bh process of unpolarized leptons off unpolarized nucleons where , at variance with previous calculations , the initial and final lepton masses are not neglected . the inclusion of the finite lepton mass has been shown to generate additional contributions to the cross section that become determinant in two specific domains : * lepton propagator quasi - singularities : here extra contributions to the leptonic tensor are required for an accurate determination of the bh cross section , independently of the lepton considered ; * minimum momentum transfer regions : in the @xmath98 region , of particular interest for the experimental determination of the ji sum rule , extra contributions dominate muon scattering off protons and must also be considered for electron scattering when @xmath0 becomes significantly small ( a few @xmath117 ) . moreover , they cancel for a neutron target as a consequence of their dependence on the electric nucleon form factor .
the bh cross section is generally sensitive to the parametrization of the nucleon electromagnetic form factors , however the magnitude of lepton mass effects as expressed by the leptonic tensor ratio ( eq . [ eq : ratmu ] ) is only weakly depending on the model .
this dependence increases with the four - momentum transfer in the quasi - singularity region , while it does not show up in the minimum momentum region because of the smallness of the four - momentum transfer . before concluding it is worth emphasizing that the effects investigated in this work are genuine finite mass effects in the sense that they do not rely on the extreme relativistic limit of the kinematics , usually employed in high energy calculations , but are related to extra terms appearing in the leptonic tensor .
moreover , their magnitude increases with the beam energy , in opposition to the ultra - relativistic limit .
these effects have been shown to be very sizeable in the above mentioned kinematic regions and may affect the quantitative analysis of high energy lepton scattering experimental dvcs data .
we would like to thank markus diehl for stimulating discussions .
this work was supported by the italian istituto nazionale di fisica nucleare under contract mb31 , by the french centre national de la recherche scientifique and by the gdr 3034 chromodynamique quantique et physique des hadrons .
( jefferson lab hall a collaboration ) c. muoz camacho et al .
* 97 * ( 2006 ) 262002 ; m. mazouz et al . ,
* 99 * ( 2007 ) 242501 ; h. fonvieille et al .
c * 86 * ( 2012 ) 015210 .
( jefferson lab clas collaboration ) s. stepanyan et al . , phys .
* 87 * ( 2001 ) 182002 ; s. chen at al .
* 97 * ( 2006 ) 072002 ; f.x .
girod et al . , phys . rev .
* 100 * ( 2008 ) 162002 ; g. gavalian et al . ,
c * 80 * ( 2009 ) 035206 .
( hermes collaboration ) a.airapetian et al .
* 87 * ( 2001 ) 182001 ; phys .
d * 75 * ( 2007 ) 011103(r ) ; jhep * 0806 * ( 2008 ) 066 ; jhep * 0911 * ( 2009 ) 083 ; nucl .
b * 829 * ( 2010 ) 1 ; jhep * 1006 * ( 2010 ) 019 ; phys .
b * 704 * ( 2011 ) 15 ; nucl .
b * 842 * ( 2011 ) 265 ; jhep * 1207 * ( 2012 ) 032 ; jhep * 1210 * ( 2012 ) 042 .
( color online ) bh differential cross section off proton in replacement of the right panel of fig . 2 , and @xmath74 evolution at @xmath75 in replacement of the left panel of fig . 3 , title="fig : " ] ( color online )
bh differential cross section off proton in replacement of the right panel of fig . 2 , and @xmath74 evolution at @xmath75 in replacement of the left panel of fig . 3 , title="fig : " ] the correct expressions for the coefficients @xmath118 and @xmath119 in the leptonic tensor defined on page 506 are @xmath120 as a consequence , eqs .
27 - 29 should be replaced with @xmath121 the main effects of these corrections are shown on fig .
[ fignew ] .
our conclusions remain unchanged in the quasi - singularity regions , while the cross section enhancement observed at minimum transfer cancels . for completeness , eqs .
35 - 36 should also be replaced with @xmath122 \nonumber \\
{ \cal f}_4 & \sim & 2\,\frac{p_2}{p_1 } \ , w_1 \nonumber\end{aligned}\ ] ] leading to the substitution of eq .
39 by @xmath123 and eq . 41 with @xmath124 | we develop the full finite lepton mass formalism for the production of real photons via the bethe - heitler reaction of unpolarized leptons off unpolarized nucleons .
genuine lepton mass effects are described , in particular their dependence upon the lepton mass and the initial beam energy , as well as their sensitivity to the nucleon isospin . in the minimum momentum transfer region , these effects dominate the muon induced proton cross section and become significant for electron scattering at small @xmath0 . |
recently , a model for a finite stack of tensionless membranes @xcite was studied with respect to the effects of higher - order terms of the curvature energy @xcite .
the approach was perturbative , using the renormalization group to sum infinitely many terms .
it was shown that thermal fluctuations induce the melting of the stack into a vertically disordered phase . by its nature , the perturbative expansion was able to give a satisfactory description only for the ordered phase . for a description of the disordered phase and a better understanding of the entire transition , we analyze in this paper the behavior of a stack of tensionless membranes exactly for very large dimension @xmath1 of the embedding space .
since the model is exactly solvable in this limit , we can calculate all its relevant properties explicitly , in particular its complete phase diagram as a function of the interlayer separation @xmath0 .
as in ref .
@xcite , we consider a model in which a multilayer system is made up of @xmath2 fluid membranes , parallel to the @xmath3-plane of a cartesian coordinate system , separated a distance @xmath0 . if the vertical displacement of the @xmath4th membrane with respect to this reference plane is described by a function @xmath5 , where @xmath6 , the energy of the stack reads : @xmath7.\ ] ] here , @xmath8 is the mean curvature , where @xmath9 is the unit normal to the @xmath4th membrane , and @xmath10 the induced metric , with @xmath11 , @xmath12 and @xmath13 $ ] .
the parameter @xmath14 is the bending rigidity of a single membrane , and @xmath15 the compressibility of the stack . in eq .
( [ mod0 ] ) , as in the following , the subscript @xmath16 denotes bare quantities , whereas renormalized parameters will carry no subscript . for slow spatial variations , the discrete variable @xmath17 may be replaced with a continuous one , and @xmath18 , where @xmath19 . in this limit
, the energy ( [ mod0 ] ) reduces to @xmath20.\ ] ] here we have introduced the bulk version of the bending rigidity @xmath21 , and defined @xmath22 .
the gradient energy @xmath23 has the unphysical feature that it gives @xmath24-dependent reparametrizations a kinetic energy .
it should therefore be replaced by the normal gradient energy @xmath25 .
we have seen in ref .
@xcite that this changes the critical exponent with which the renormalized @xmath26 vanishes as the bending rigidity becomes critical .
however , in the limit of large @xmath1 , which we are going to investigate , the difference between the two gradient energies will be negligible .
for arbitrary @xmath1 , the vertical displacement of the @xmath4th membrane in the stack becomes a @xmath28-vector field @xmath29 .
it is useful to consider @xmath30 as an independent field @xcite , and impose relation ( [ gmunu ] ) with help of a lagrange multiplier @xmath31 .
we write the partition function as a functional integral over all possible configurations @xmath32 of the individual membranes in the stack , as well as over all possible metrics @xmath33 . after taking again the continuum limit
, the partition function reads : @xmath34 with @xmath35 where @xmath36 is a @xmath28-dimensional vector - function of @xmath37 .
note that the functional integral over @xmath38 in ( [ partfunc3 ] ) has to be performed along the imaginary axis to result in a @xmath39-function .
we have also introduced a term proportional to @xmath40 .
this term is necessary to renormalize the theory , and its coefficient @xmath41 corresponds to the large-@xmath1 in - plane compressibility of the membranes . since we take the membranes to be incompressible , we shall set the renormalized @xmath42 equal to zero at the end of our calculations .
we have also included a surface tension @xmath43 , again to absorb infinities and to be set equal to zero after renormalization .
the functional integral over @xmath36 in eq .
( [ partfunc3 ] ) is gaussian and can be carried out to yield an effective energy @xmath44 with @xmath45 \label{s0tilde},\ ] ] and @xmath46 , \label{s1}\ ] ] where the functional trace tr is here an integral over space as well as the integral over wavevectors @xmath47 and @xmath48 , after replacing @xmath49 and @xmath50 .
note that the discrete nature of the stack restricts the integral over the wavevectors @xmath48 to the first brillouin zone @xmath51 . in the large-@xmath1 limit
, the partition function ( [ partfunc3 ] ) is dominated by the saddle point of the effective energy ( [ effectaction ] ) with respect to the metric @xmath30 and the lagrange multiplier @xmath52 .
for very large membranes , the saddle point can be assumed to be symmetric and homogeneous @xcite : @xmath53 with constant @xmath54 and @xmath55 . at the saddle point
the effective energy ( [ effectaction ] ) becomes the free energy of the system . in the following
we shall investigate both the case of an infinite and a finite stack of membranes .
as we will see , the large-@xmath1 approximation allows for the vertical melting even in an infinite stack , which is not found perturbatively @xcite .
let us first analyze the case of an infinite stack . to simplify our calculations , we assume the number @xmath56 of membranes in the stack to be very large , making the distance @xmath0 between them very small .
in this regime , we may extend the limits @xmath57 of the integral over @xmath48 to infinity .
the explicit @xmath0-dependence will be introduced later into our calculations .
after evaluating the functional trace in eq .
( [ s1 ] ) , we obtain @xmath58\right\},\ ] ] where ultraviolet divergences are regularized by introducing a sharp transverse wavevector cutoff @xmath59 and @xmath60 has been replaced by @xmath1 for large @xmath1 .
we may now absorb the first term in ( [ s1linf ] ) by renormalizing @xmath43 , so that @xmath61 is the physical surface tension , which is set equal to zero . the second , quadratically divergent term in ( [ s1linf ] ) is used to define the critical temperature as @xmath62 the next divergent term , proportional to @xmath63 , is regularized by introducing a renormalization scale @xmath64 and modifying the the in - plane compressibility to @xmath65 the physical in - plane compressibility @xmath42 is now set equal to zero , as explained in the previous section .
the effective energy thus becomes @xmath66 \right\},\ ] ] with the constants @xmath67 , @xmath68 . from the second derivative matrix of @xmath69 with respect to @xmath70 and @xmath38 we find that the stability of the saddle point is guaranteed only for @xmath71 .
extremizing the above expression with respect to @xmath70 , we find two solutions for @xmath38 , namely @xmath72 and @xmath73 , with @xmath74 = \frac{\sqrt{k}}{a } \left ( \frac{1}{t } - \frac{1}{t_{\rm c } } \right).\ ] ] for @xmath75 , this equation has no solution for @xmath76 . in this case , the only possible solution is @xmath72 , which corresponds to the ordered phase as we shall verify later . for @xmath77 ,
the saddle point lies at @xmath78 , which is now well - defined .
this is the vertically disordered phase . the free energy density at the extremum is given by @xmath79 and its behavior is similar to the one found perturbatively in ref.@xcite .
( see fig .
[ fig : lambda ] ) .
= 7.cm extremizing the effective energy ( [ eefflinf ] ) with respect to @xmath38 ,
we find @xmath70 as a function of temperature .
for @xmath80 it is given by @xmath81 this as @xmath82 approaches @xmath83 from below , indicating the vertical melting at @xmath83 .
in the disordered phase , @xmath70 is found to be @xmath84 as @xmath82 approaches @xmath83 from above , @xmath76 tends to zero , and @xmath70 goes again to infinity .
the positivity of @xmath70 and the stability of the saddle point imply that there is a maximum temperature , given by @xmath85 below which our assumption that the membranes in the stack are in - plane incompressible does not lead to a stable system . let us now analyze the case of a finite stack of size @xmath86 . now the functional trace in ( [ s1 ] ) involves a sum over the discrete wavevectors @xmath87 , given by @xmath88 for small @xmath55 , a series expansion leads to @xmath89 with @xmath90 \nonumber \\ & & + \sqrt{\pi } \sum_{m=3}^{\infty } \frac{(-1)^{m+1}\lambda_0^m}{m \ , 2^{2 m } \pi^m } l_\parallel^{m-2 } \left ( \frac{k_0}{b_0}\right)^{\frac{m-1}{2 } } \frac{\gamma(\frac{m-1}{2})}{\gamma(\frac{m}{2 } ) } \zeta(m-1).\end{aligned}\ ] ] as in the case of the infinite stack , we absorb the logarithmic divergence by renormalizing the in plane compressibility via eq.([compressibility ] ) , setting @xmath42 equal to zero for incompressible membranes .
the surface tension receives now an @xmath91-dependent renormalization @xmath92 and @xmath93 is again set equal to zero to describe a stack of tensionless membranes .
extremization of the renormalized combined effective action ( [ s0tilde ] ) and ( [ s1finl ] ) with respect to @xmath70 leads again to two possible solutions for the saddle point , namely @xmath94 or @xmath95 , with @xmath96 + \lambda_{l_\parallel } \left [ 1 - \gamma + \ln \left(\frac{l_{\perp}^2}{l_\parallel } \frac{1}{8 \pi } \sqrt{\frac{b}{k } } \right)\right ] \nonumber \\ & & + 32 \pi^{3/2 } \sum_{m=3}^{\infty } \frac { ( -1)^{m+1 } \lambda_{l_\parallel}^{m-1}}{m \ , 2^{2 m } \pi^m } l_{\parallel}^{m-2 } \left ( \frac{k}{b}\right)^{\frac{m-2}{2 } } \frac{\gamma(\frac{m-1}{2})}{\gamma(\frac{m}{2 } ) } \zeta(m-1 ) = \frac{\sqrt{k}}{a } \left ( \frac{1}{t } - \frac{1}{t_{l_\parallel } } \right)\end{aligned}\ ] ] where @xmath97 is the inverse critical temperature for a stack of size @xmath91 . for @xmath98 , eq .
( [ lambdafinl ] ) has no solution . in this case , the stack is in the ordered phase , the only available solution for the saddle point being @xmath72 . for @xmath99 , there exists a nonzero solution @xmath100 , where the system is in the vertically disordered phase .
let us now examine the saddle point solutions for @xmath70 . in the vertically disordered phase where @xmath95 is nonzero , we may expand the effective energy in a small-@xmath86 series .
extremization with respect to @xmath101 leads to @xmath102 the positivity of @xmath70 and the stability of the saddle point again define a maximal temperature , given by @xmath103 above which our assumption that the membranes in the stack are in - plane incompressible can not be maintained . in the ordered phase ,
the situation is more delicate . for @xmath72 ,
@xmath70 can be calculated exactly , and we obtain @xmath104,\ ] ] with an infrared regulator @xmath105 equal to the inverse lateral size of the membranes in the stack . if the size @xmath86 of the stack is large , ( [ rhoftl0 ] )
may be approximated by @xmath106 for smaller stacks , however , the positivity of @xmath70 is not guaranteed . for a fixed , but small stack size @xmath86 , and for fixed lateral size @xmath105 of the membranes in the stack ,
there is a characteristic temperature defined by @xmath107 above which @xmath70 changes sign , and ( [ rhoftl0 ] ) is no longer applicable .
interestingly , for all @xmath105 and for all finite sizes @xmath91 of the stack , the critical temperature @xmath108 is lower than @xmath109 , so that the vertical melting still occurs .
the behavior of @xmath70 is depicted in fig .
[ fig : rho ] .
note that eq .
( [ tcr ] ) reflects the existence of a characteristic horizontal length scale . at fixed temperature @xmath110 , and for membranes of lateral size
@xmath111 smaller than @xmath112 the height fluctuations of the individual membranes are not strong enough to destroy the ordered phase .
the characteristic length @xmath113 corresponds to the the de gennes - taupin persistence length @xmath114 @xcite of the individual membranes , below which crumpled membranes appear flat .
until now we have performed our calculations in the somewhat unphysical continuum approximation , by letting the interlayer separation @xmath0 be very small making the number of membranes in the stack very large .
let us now investigate the properties of the stack for a fixed number @xmath56 of membranes at a finite interlayer distance @xmath0 . for this purpose
, we replace the continuum derivative @xmath115 in the @xmath24-direction with the discrete gradient operator @xmath116 , whose eigenvalues are given by @xmath117 where @xmath118 is some test function .
the discrete wavevectors @xmath87 are now given by @xmath119 for small interlayer separation @xmath0 , the free energy is given by ( [ s1finl ] ) with @xmath120 \nonumber \\
& + & \frac{\lambda_0 ^ 2}{64 \pi } \sqrt{\frac{k_0}{b_0 } } \left [ 1 - 2 \ln \left ( \frac{4 \lambda^2}{\lambda_0 } \right)\right ] + \frac{1}{n \sqrt{\pi } } \sum_{m=2}^{\infty } \frac{(-1)^{m+1}\lambda_0^m}{m \ , 2^{\frac{3 m+1}{2 } } } l^{m-2 } \left ( \frac{k_0}{b_0}\right)^{\frac{m-1}{2 } } \frac{\gamma(\frac{m-1}{2})}{\gamma(\frac{m}{2 } ) } \bar{\zeta}_{n}(m-1 ) , \nonumber \\\end{aligned}\ ] ] where we have defined the modified zeta - function @xmath121^\frac{m}{2}}.\ ] ] we proceed by renormalizing the in - plane compressibility via eq.([compressibility ] ) , setting @xmath42 equal to zero for incompressible membranes as before .
the surface tension receives an @xmath0-dependent renormalization @xmath122 and @xmath93 is set equal to zero to describe a stack of tensionless membranes , as before .
but now the bulk bending rigidity is also modified to @xmath123 note that this renormalization agrees with the known result for a single membrane @xcite .
the saddle point for @xmath38 is now given by @xmath72 or @xmath124 , with @xmath125 + 32 \sqrt{\pi } \sum_{m=2}^{\infty } \frac { ( -1)^{m+1 } \lambda_{l}^{m-1}}{m \ , 2^{\frac{3 m + 1}{2 } } } l^{m-2 } \left ( \frac{k}{b}\right)^{\frac{m-2}{2}}\frac{\gamma(\frac{m-1}{2})}{\gamma(\frac{m}{2 } ) } \bar{\zeta}_n(m-1 ) \nonumber \\ & & = \frac{\sqrt{k}}{a } \left ( \frac{1}{t } - \frac{1}{t_l } \right)\end{aligned}\ ] ] where @xmath126 \right\}\ ] ] is the inverse @xmath0-dependent critical temperature , and @xmath127 is the bending rigidity of a single membrane in the stack .
the two solutions for @xmath38 again imply the existence of two different phases , with a phase transition at the critical temperature @xmath128 , which , as in the perturbative case , depends only weakly on the number of membranes in the stack .
the corresponding solutions for @xmath70 , obtained by extremizing the effective energy with respect to @xmath38 , are given by @xmath129 for @xmath130 , that is , in the ordered phase , and @xmath131 in the vertically molten phase , where @xmath132 .
the stability of the saddle point requires a minimum interlayer separation @xmath133 below which the stack becomes unstable .
@xmath134 is inversely proportional to de genne s penetration depth @xmath135 @xcite , which is of the order of the interlayer separation in smectic liquid crystals .
the phase diagram of the stack is depicted in fig .
[ fig : phase ] = 6.cm let us now characterize both phases in more detail .
for temperatures higher than @xmath128 , the solution of the saddle point is @xmath124.this corresponds to the disordered phase , where the stack melts .
the normals to the membranes are uncorrelated beyond a length scale @xmath136 , as can be derived from the expression for the orientational correlation function , which in the limit @xmath137 , with constant @xmath138 , reads : @xmath139 the length scale @xmath136 may thus be identified with the persistence length @xmath114 @xcite . in the low - temperature phase ,
the solution of the saddle point is @xmath72 .
this corresponds to the ordered lamellar phase as can be seen by examining the orientational correlation function in the planes of the membranes .
we find in the limit @xmath140 , with constant @xmath138 : @xmath141 this slow , algebraic fall - off of the correlation function implies that , at large distances , the normal vectors to the membranes are still parallel , so that the surfaces remain flat on the average .
the effect of thermal fluctuations is suppressed , and they do not disorder the stack .
let us now calculate the entropy loss in the ordered phase . by a simple scaling argument @xcite , this quantity should be inversely proportional to the quadratic interlayer spacing , @xmath142 where @xmath143 is a temperature dependent proportionality constant .
as we shall see , as the interlayer distance increases , logarithmic corrections must be added to ( [ deltas ] ) .
the entropy loss in the ordered phase can be computed by calculating the difference between the free energy density of a single , isolated membrane , an the free energy density of the stack . for small values of @xmath38
, it is given by @xmath144 strictly speaking , the ordered phase corresponds to @xmath145 . in that case ,
( [ deltaf ] ) agrees with ( [ deltas ] ) , and we see no correction to the entropy loss .
however , if the size of the membranes in the stack is smaller than the characteristic length @xmath113 , an ordered phase still exists for small values of @xmath38 [ see discussion after eq .
( [ lp ] ) ] , in which case the corrections to the first term in ( [ deltaf ] ) will appear .
in the limit of large embedding dimension @xmath1 , we have shown that a stack of tensionless and incompressible membranes melts vertically upon approaching a critical temperature , where the lamellar phase goes over into a disordered phase .
in contrast to the low - temperature ordered phase , where the decay of orientational correlations is power - like , the high - temperature disordered phase is characterized by an exponential decay of orientational correlations , with different length scales in the transversal and longitudinal directions .
99 w. helfrich , z. naturforsch . *
33a * , 305 ( 1978 ) .
m. e. s. borelli , h. kleinert and a.m.j .
schakel , _ vertical melting of a stack of membranes _ , berlin preprint ( 2000 ) , cond - mat/0004432 .
a. m. polyakov , nucl .
b * 268 * , 406 ( 1986 ) .
h. kleinert , phys .
b * 189 * , 187 ( 1987 ) ; phys .
lett . * 58 * , 1915 ( 1987 ) .
p. olesen and s. yang , nucl .
b * 283 * , 73 ( 1987 ) .
f. david and e. guitter , nucl .
b * 295 * , 332 ( 1988 ) .
p. g. de gennes and c. taupin , j. phys
. chem . * 86 * , 2294 ( 1982 ) .
l. peliti and s. leibler , phys .
54 * , 1690 ( 1985 ) , h. kleinert , phys .
lett . a * 114 * , 263 ( 1986 ) , d. frster , phys .
a * 114 * , 115 ( 1986 ) .
p. g. de gennes , _ the physics of liquid crystals _ ( oxford university press , london , 1974 )
. s. leibler in _ statistical mechanics of membranes and surfaces _ ,
proceedings of the fifth jerusalem winter school for theoretical physics , 1987/1988 , edited by d. nelson , t. piran , and s. weinberg ( world scientific , singapore , 1989 ) . | the phase diagram of a stack of tensionless membranes with nonlinear curvature energy and vertical harmonic interaction is calculated exactly in a large number of dimensions of configuration space . at low temperatures , the system forms a lamellar phase with spontaneously broken translational symmetry in the vertical direction . at a critical temperature , the stack disorders vertically in a melting - like transition .
the critical temperature is determined as a function of the interlayer separation @xmath0 . |
quantum field theory predicts that a relativistically accelerated observer views the minkowski vacuum as a thermal bath with a temperature proportional to the proper acceleration of the observer .
this phenomenon , known as the unruh effect @xcite , is extremely weak , and even the acceleration as high as @xmath0 should result in the temperature as low as @xmath1 .
although several experimental proposals have been put forth @xcite , verification of the unruh effect still remains an open question @xcite .
the theory of the unruh effect has attracted widespread interest in the physics community and it has been extensively examined from different perspectives .
these include the approaches involving bogolyubov transformation between the minkowski and rindler frames of reference @xcite , the application of the unruh - dewitt detector model which offers an operational definition of a particle as a trigger for the transitions between the atomic energy levels @xcite , the manifestation of thermalization of open quantum systems involving decoherence and dissipation @xcite , understanding as a quantum noise channel @xcite , and the thermal correction to the energy shift and spontaneous excitation of atoms @xcite . in recent years
, the unruh effect has been extensively considered in the field of quantum information .
it has been studied how the unruh effect affects quantum entanglement between different observers @xcite , as well as other types of non - classical correlations @xcite and consequently quantum nonlocality @xcite , quantum teleportation @xcite , entropic uncertainty relations @xcite , and parameter estimation @xcite .
the unruh effect has also been used to analyze how to improve information processing protocols @xcite and quantum measurement techniques @xcite . in order to detect the unruh effect
several metrological tools have been employed .
different approaches include using gaussian probes @xcite , applying the framework of open quantum systems to compute fisher information and quantum fisher information @xcite , and using the technique of quantum channel discrimination @xcite . in this work , we study an overlooked possibility of increasing the detection precision for the unruh effect and
show how entangled states can be used to enhance sensibility of the accelerated detectors .
we use quantum state discrimination technique @xcite to search for the optimal strategies that can facilitate the detection of the unruh effect .
we employ the model of a two - level atom coupled to the massless scalar fields in the minkowski vacuum state as a probe and study its time evolving state to discriminate between two possible scenarios : one , in which the atom has been inertial , and the other , in which the atom has been uniformly accelerated .
we discuss different strategies of detection and show how initial entanglement between the atom and an external degree of freedom can enhance the performance of the atom as a thermometer .
our paper is constructed as follows : in section [ section2 ] we introduce the model of our probe , we analyze the optimal detection of the unruh effect in section [ section3 ] , and section [ section4 ] concludes the paper .
let us note that the detector model we will consider was firstly introduced in ref .
@xcite by f. benatti and r. floreanin , where the unruh effect has been reexamined from the perspective of the thermalization phenomenon and entanglement generation of accelerated atoms .
recently this model has been fruitfully applied to understand the hawking and gibbons - hawking effects @xcite . here
we will use this model to investigate the detection of the unruh effect with the assistance of entanglement .
a combined system consists of a detector and an external fluctuating vacuum field with the total hamiltonian of the general form : @xmath2 where @xmath3 is the hamiltonian of the detector , @xmath4 is the free hamiltonian of the scalar field , and @xmath5 represents their mutual interaction .
the detector will be described with a two - level atom governed by the hamiltonian @xmath6 , where @xmath7 is its energy - level spacing and @xmath8 is the pauli matrix .
the interaction hamiltonian takes the form : @xmath9 where @xmath10 is the coupling constant , @xmath11 ( @xmath12 ) is the atomic raising ( lowering ) operator , and @xmath13 corresponds to the scalar field operator taken at @xmath14 representing the classical trajectory of the atom .
we assume that the initial state of the atom , @xmath15 , is separable from the initial state of the field , which is assumed to be in the minkowski vacuum @xmath16 .
therefore the initial state of the total system is given by @xmath17 .
in the reference frame of the atom , the equation of motion of the total system is given by : @xmath18,\end{aligned}\ ] ] where @xmath19 is the proper time along the trajectory of the atom , and @xmath20 denotes the time - dependent density matrix of the atom and the field .
generally , we will be interested in the evolution of the atom and this dynamics can be obtained by tracing out the field degrees of freedom , i.e. , @xmath21 $ ] .
the reduced state density of the atom , in the limit of weak coupling , is found to obey the kossakowski - lindblad equation of the form @xcite : @xmath22+\mathcal{l}[\rho(\tau)]\end{aligned}\ ] ] with the effective hamiltonian @xmath23 given below , and the lindblad superoperator : @xmath24=\frac{1}{2}\sum^3_{i , j=1}a_{ij}[2\sigma_j\rho\sigma_i-\sigma_i\sigma_j\rho-\rho\sigma_i\sigma_j].\end{aligned}\ ] ] both the matrix @xmath25 and the effective hamiltonian @xmath23 are associated with the field correlation functions , @xmath26 . by defining the fourier and hilbert transforms of the field correlation functions , namely : @xmath27 we have computed that the coefficients of the kossakowski matrix @xmath25 can be expressed as : @xmath28 with @xmath29 $ ] , @xmath30 $ ] .
the effective hamiltonian @xmath23 , which contains the so called _ lamb shift _
@xcite , then can be written as : @xmath31\right\}\sigma_z,\ ] ] where we have defined a renormalized energy level spacing @xmath32 , which contains the original atomic energy level spacing @xmath7 and the energy shift @xmath33 $ ] .
assuming the atom is initially prepared in the pure state @xmath34 , by solving the master equation ( [ lindblad equation ] ) we can obtain the time - dependent reduced density matrix of the atom , @xmath35 let us now consider the evolution of the uniformly accelerated two - level atom , whose trajectory can be described as : @xmath36 in order to obtain the parameters of the reduced density matrix of the accelerated atom , we need to evaluate the correlation function of the field , which can be found from the following two - point function of the massless scalar field : @xmath37 substituting the trajectory of the atom ( [ trajectory ] ) into ( [ two - point - function ] ) , one can obtain the field correlation function in the frame of the accelerated atom @xcite : @xmath38.\end{aligned}\ ] ] in this case , the fourier transform of the field correlation function is : @xmath39 consequently , the coefficients of the kossakowski matrix @xmath25 and the effective level spacing @xmath32 of the atom are given by : @xmath40 where @xmath41 is the spontaneous emission rate of the inertial atom in minkowski spacetime .
using the above results and assuming that the time of evolution is sufficiently long , i.e. @xmath42 , one can write the steady state of the accelerated atom as : @xmath43},\end{aligned}\ ] ] which is a thermal state with the temperature @xmath44 .
this result shows that the accelerated atom has been equilibrating with the thermal bath characterized by the mean number of unruh particles equal to @xmath45 . for further convenience ,
let us parameterize the number of unruh particles by @xmath46 .
then , in the interaction representation we can rewrite the time - dependent state of the accelerated atom as : @xmath47 where @xmath48 , @xmath49 and @xmath50 .
let us note that if @xmath51 , then eq . ( [ bloch - representation ] ) represents the the evolving quantum state of the inertial atom in the minkowski vacuum .
we will now use the detector model described in the previous section to act as a thermometer detecting the unruh effect . in order to find out how to optimally detect this effect
, we choose to measure the atom in the basis that has the lowest probability of misidentifying the two possible states : one characterizing a static , and the other - uniformly accelerated trajectory of the atom .
it is clear that this question has been reduced to the classical problem of quantum state discrimination @xcite . in this scheme ,
the minimum probability of making an error in the identification of the states @xmath52 or @xmath53 is given by : @xmath54 where @xmath55 denotes the trace norm of @xmath56 .
for hermitian operators @xmath56 their norms also equal the sum of the absolute values of their eigenvalues . in our problem , @xmath52 is the evolution state of the inertial atom , i.e. , the case of @xmath51 in eq . , and @xmath53 corresponds to the quantum state of the accelerated atom , i.e. , the case of @xmath57 in eq . .
from this , we obtain : @xmath58 with @xmath59 ^ 2,\end{aligned}\ ] ] where @xmath60 , @xmath61 , and @xmath62 . for sufficiently long interaction times
, the atom reaches equilibrium with the state of the external field , then the distinguishability between the two equilibrium states in eq .
( [ delta ] ) is determined by @xmath63 . moreover , in the long limit of interaction times the level of distinguishability is independent of the initial state of such a thermometer .
let us now study the distinguishability of the state for finite interaction times . to better quantify the performance of the atom thermometer , let us normalize the euclidean distance @xmath64 to @xmath65 . in fig .
[ fig1 ] , we plot @xmath66 as a function of the interaction time @xmath19 and the parameter @xmath67 characterizing the initial state of the atom thermometer .
we can see from fig .
[ fig1 ] that the two possible trajectories are always less distinguishable as compared to the equilibrium case .
the largest distance between the two alternative final states is obtained in the limit of long interaction times , and it is equal to @xmath68 .
interestingly the largest distance in this case is independent of initial states of the atom thermometer .
it means that the optimal condition to detect the unruh effect is letting the atom evolve for long enough times regardless of its initial state .
thus , we arrive at the conclusion that quantum coherence @xcite plays a trivial role in the optimal detection of the unruh effect .
let us note that our results obtained here are consistent with that of ref .
@xcite where the technique of quantum metrology has been used .
we also find that if the probe is prepared in the excited state , i.e. , @xmath69 , then there is a finite time , @xmath70 ( namely , when @xmath71 ) , at which the trace distance goes to zero .
this is because at this time , both the accelerated atom and the inertial atom are approximatively driven to the same thermal state with the parameterized temperature @xmath72 .
however , after that time , the state of the accelerated atom remains almost unchanged , while the state of the inertial atom will be eventually driven to the ground state .
thus , we can see that the euclidean distance between these two cases will increase again and finally reach @xmath68 .
we can also see from fig .
[ fig1 ] that the increase of @xmath67 leads to a quicker equilibration of the system .
, normalized to @xmath65 , plotted as a function of time @xmath19 ( in units of @xmath73 ) and the parameter @xmath67 characterizing the initial state of the atom .
@xmath72 is assumed.,title="fig:",scaledwidth=46.0% ] , normalized to @xmath65 , plotted as a function of time @xmath19 ( in units of @xmath73 ) and the parameter @xmath67 characterizing the initial state of the atom .
@xmath72 is assumed.,title="fig:",scaledwidth=42.0% ] we will now turn our attention to a more general scenario , in which the considered atom is initially entangled with another atom . in particular
we assume that the additional atom does not interact with the external field .
let us consider the initial state of the two atoms to be : @xmath74 note that this state is an @xmath75type state and the parameter @xmath76 satisfies @xmath77 ; the two particular cases , @xmath78 and @xmath79 correspond to one of the bell states and the werner state , respectively .
then the evolution of the system leads to the following final state of the atoms : @xmath80 with this in hand , we can compute the trace distance between the state @xmath81 corresponding to the resting detector case and the state @xmath82 corresponding to the accelerated detector case , which is given by : @xmath83,\end{aligned}\ ] ] where @xmath84 .
let us note that in the limit of infinite evolution time , the trace distance also approaches to @xmath68 regardless of the initial state of the two atoms .
consider the werner initial state , for which @xmath85 with @xmath86 .
such an input state is only entangled when @xmath87 . the resulting trace distance given by eq .
is plotted in fig .
we find that the trace distance between the accelerated and non - accelerated scenario increases as a function of the initial entanglement ( quantified by @xmath88 ) .
besides , for some entangled states there is a global maximum in the distinguishability before they achieve the equilibrium with the environment , and interestingly this value can even exceed the equilibrium value , @xmath68 .
this interesting result proves that the performance of the thermometer with the auxiliary component can be enhanced with the aid of the initial entanglement .
it also follows that the optimal bipartite state for the detection of unruh effect is the maximally entangled state ( @xmath89 ) .
( in units of @xmath73 ) .
@xmath72 is taken.,scaledwidth=45.0% ] let us now elucidate some of the phenomena observed in fig .
there is a sudden change point for the trace distance , which occurs when @xmath90 .
thus , the bigger @xmath88 is , the later the sudden change occurs . if @xmath91 ( it is equal to @xmath92 for @xmath72 in fig .
[ fig2 ] ) , the trace distance can become larger than the equilibrium one , @xmath68 . in this case
there is also the maximum trace distance corresponding to the the optimal detection of the acceleration of the atom . in the case of @xmath89 ( @xmath93 ) ,
the maximum trace distance is @xmath94 ( @xmath95 ) when @xmath96 ( @xmath97 ) . to find out which thermometer
, the one with or without the initial entanglement , performs better in the detection of the unruh effect , we compare the trace distance of bipartite system case with that of a single atom case in fig .
it is clear that the use of the initial entanglement provides a significant advantage .
this interesting result provides a hint that the use of entanglement can be useful in the detection of the absolute acceleration , and as a consequence in experimental verification of the unruh effect .
( in units of @xmath73 ) .
the green curve denotes that for the initial maximally entangled state case , the red curve denotes that for the single atom initially prepared in the ground state , and the blue curve denotes the equilibrium case .
@xmath72 is assumed here.,scaledwidth=45.0% ]
we have used a simple model of a two - level atom interacting relativistically with the fluctuating massless scalar quantum field in the minkowski vacuum to probe the unruh effect .
the purpose of the study was to determine the optimum methods of the detection , in particular to show a relevant role of the initial entanglement , which has been overlooked in the previous studies . within the framework of open quantum systems ,
we have obtained the analytical dynamical evolution of the thermometer .
we have then compared the evolution of the accelerated atom with that of the inertial one using the quantum state discrimination techniques .
we have found that regardless of the initial state of the atom , the largest trace distance between this two cases is @xmath68 , which is achieved in the limit of infinitely long interaction times .
this results corresponds to the situation when the probe equilibrates with the surrounding thermal bath of the field .
however , we have shown that this result can be improved by considering a more general situation in which the atom thermometer is initially entangled with another two - level system that does not interact with the considered field .
our results shows that the elusive unruh effect , which has been too difficult to be experimentally verified so far , perhaps can be better explored with the aid of entangled systems in accelerated motion .
let us note that our method here , based the equivalence principle , can also be applied to the detection of hawking effect @xcite and gibbons - hawking effect @xcite , and so on .
besides , with the development of simulation experiments of relativistic quantum systems @xcite , our results can be relevant to upcoming experimental tests of related phenomena in analogue systems .
z. tian and a. dragan thank for the financial support to the national science center , sonata bis grant no .
dec-2012/07/e / st2/01402 .
j. jing and j. wang are supported by the national natural science foundation of china under grant nos .
11475061 , and 11305058 .
n. friis , m. huber , i. fuentes , and d. e. bruschi , phys .
d * 86 * , 105003 ( 2012 ) ; d. e. bruschi , a. dragan , a. r. lee , i. fuentes , and j. louko , phys .
lett . * 111 * , 090504 ( 2013 ) ; d. e. bruschi , j. louko , d. faccio , and i. fuentes , new journal of physics * 15 * ( 2013 ) 073052 ( 12pp ) . c. w. helstrom , _ quantum detection and estimation theory _
( academic press , new york , 1976 ) ; a. chefles , contemp . phys .
* 41 * , 401 ( 2000 ) ; v. kargin , ann
. stat . * 33 * , 959 ( 2005 ) ; m. tsang , phys .
* 108 * , 170502 ( 2012 ) . | we show how the use of entanglement can enhance the precision of the detection of the unruh effect with an accelerated probe .
we use a two - level atom interacting relativistically with a quantum field as the probe , and treat it as an open quantum system to derive the master equation governing its evolution . by means of quantum state discrimination , we detect the accelerated motion of the atom by examining its time evolving state .
it turns out that the optimal strategy for the detection of the unruh effect , to which the accelerated atom is sensitive , involves letting the atom - thermometer equilibrate with the thermal bath .
however , introducing initial entanglement between the detector and an external degree of freedom leads to an enhancement of the sensitivity of the detector .
also , the maximum precision is attained within finite time , before equilibration takes place .
= 0.45 cm |
the na48 experiment at the cern sps aims at a precise measurement of the @xmath0 parameter , characterising direct cp violation in the neutral kaon decays .
this parameter can be accessed experimentally through the double ratio @xmath1 of the decay rates : @xmath2 to achieve the required accuracy of @xmath3 @xmath4 on @xmath0 , a large statistic is needed . to minimise systematic effects , na48 collects the four modes at the same time using simultaneous @xmath5 and @xmath6 beams and from the same decay region , such that most effects cancel at first order in the double ratio .
a description of the analysis of the data taken in 1997 can be found in @xcite .
the @xmath7 decay mode is identified by measuring four photons in the calorimeter , in the energy range 3 - 100 gev , with an average energy of @xmath3 25 gev .
the calorimeter is located @xmath3 100 m downstream of the decay region .
photon angles are not directly measured and therefore the distance @xmath8 between the kaon decay point and the calorimeter is computed assuming the kaon mass for the four photons , and using the measured photon energies and positions transverse to the beam axis : @xmath9 } / m_k \nonumber\end{aligned}\ ] ] the fiducial decay region is defined by applying cuts on the reconstructed value of @xmath8 .
this decay region should be defined identically for neutral and charged decays and this implies a good control of systematic effects in measuring photon energies and positions .
the invariant masses of the two photon pairs are then computed using @xmath8 and compared to the nominal @xmath10 mass to select @xmath11 candidates . to achieve a good rejection of the @xmath12 background ( whose decay rate is @xmath3 200 times higher than the cp violating @xmath13 mode ) , a @xmath10 mass resolution of @xmath3 1 mev is required .
in addition , na48 is using proton tagging to distinguish @xmath5 and @xmath6 decays : the time of each proton directed towards the @xmath6 production target is recorded and compared to the decay time measured in the detector .
if a coincidence is found the event is classified as @xmath6 , otherwise it is classified as @xmath5 .
this requires a very reliable measurement of the event time with a good ( better than 500 ps ) resolution .
the requirements on the performances of the calorimeter can be summarised as follows : + 1 ) energy resolution : should be better than 1% at 25 gev and above , with a good uniformity over the calorimeter ( long range variations not exceeding @xmath3 0.1% ) + 2 ) position resolution : of the order of 1 mm + 3 ) time resolution : better than 500 ps + 4 ) non linearity : residual non linearities should be understood at the level of 0.1% in the energy range 3 - 100 gev + 5 ) stability : the calorimeter should give a stable response over several years , in a high rate of kaon decays ( the @xmath5 decay rate seen by the detector is @xmath3 500 khz ) to fulfil these goals , na48 has decided to use a quasi homogeneous liquid krypton calorimeter , operated as ionization calorimeter .
this calorimeter is almost fully active , which ensures a very good intrinsic energy resolution .
the use of cold noble liquid allows a very good stability of the response .
initial current readout technique with a fast shaping allows to operate in a high rate environment and to achieve a good time resolution .
table [ table1 ] summarises the characteristics of noble liquids .
clearly , liquid argon is not dense enough for a homogeneous calorimeter .
xenon would offer a smaller radiation length than krypton , but its moliere radius , giving the transverse size of showers , which is in our case a more important parameter , is in fact close to the one of the krypton .
liquid krypton was therefore chosen as medium .
the calorimeter is briefly described in the next section .
the reconstruction of pulses and showers are then described with the performances achieved , with emphasis on the energy resolution and the linearity . .
_ summary of noble liquid characteristics . _ [ cols="<,^,^,^,^",options="header " , ] [ table1 ]
the calorimeter is made of a bath of @xmath3 10 m@xmath14 of liquid krypton at 120 k with a total thickness of 125 cm ( @xmath3 27 radiation lengths ) and a octagonal shaped active cross - section of @xmath3 5.5 m@xmath15 .
a 8 cm radius vacuum tube goes through the calorimeter to transport the undecayed neutral beam .
thin copper - beryllium ribbons ( of dimensions 40@xmath16 m @xmath17 18 mm @xmath17 127 cm ) stretched between the front and the back of the calorimeter form a tower structure readout .
the 13212 readout cells each have a cross - section of @xmath3 2@xmath18 cm@xmath15 and consist ( along the horizontal direction ) of a central anode ( at the high voltage ) in the middle of two cathodes ( kept at the ground ) .
the gap size @xmath20 is therefore @xmath3 1 cm . there is a 2 mm vertical separation between the electrodes .
the drift of the ionization electrons towards the anode produces a current @xmath21 which in the limit of uniform ionization across the gap and constant electric field can be written as : @xmath22 where @xmath23 is the charge one wants to measure , @xmath24 the drift velocitys@xmath25 for a high voltage of 3000 v ] and @xmath26 the total drift time across the gaps ] . to guarantee a good uniformity of the initial current ( @xmath27 ) , a good accuracy of the gap
is required .
this is enforced by the use of five precisely machined spacer plates ( fibreglass reinforced epoxy , 5 mm thick , @xmath3 0.025 x0 ) , normal to the detector axis and located every 21 cm .
the spacer plates guide the ribbons in a accordion geometry with a @xmath28 48 mrad zig - zag angle , in the horizontal direction .
configurations where the shower core is aligned with an electrode , resulting in a drop of the response , are thus avoided .
the accuracy achieved in the position of the holes in the spacer plates and therefore in the gap size is typically @xmath28 45 @xmath16 m .
the overall tower structure is projective towards the middle of the kaon decay region , which is located @xmath3 114 m upstream of the calorimeter .
the measurement of photon positions becomes thus independent of fluctuations in the longitudinal shower development .
because of the projectivity , the gap size is increasing by 1% between the front and the back of the calorimeter .
figure [ fig_lkr1 ] shows the electrode structure of one quarter of the calorimeter .
the calorimeter is housed in a vacuum insulated cryostat .
the warm windows are made of aluminium foil 4 mm thick followed on the beam entrance side by a cold double walled stainless window . the outer wall which holds the pressure is convex and 2.9 mm thick , the inner is flat and only 0.5 mm thick .
a light honeycomb aluminium structure connects the two foils .
the total amount of matter before the beginning of the liquid krypton volume is @xmath3 0.8 x0 including all the na48 detector elements upstream of the calorimeter , out of which @xmath3 0.63 x0 comes from the cryostat and the calorimeter itself .
krypton evaporates continuously at the top of the cryostat and flows through an external purifier to be subsequently recondensed and returned to the cryostat .
the krypton purity is such that the lifetime of secondary electrons exceeds 100 @xmath16s ( to be compared to a shaping time of @xmath3 70 ns ) .
long term temperature variations within the calorimeter are @xmath29 @xmath280.3 k , with negligible influences on the calorimeter performances . preamplifiers are located on the back plane of the electrode structure . a dual hybrid version of cold charge integration preamplifier , of bnl type using silicon jfet technology ,
was designed @xcite .
the restoring time chosen is 150 ns .
the output signal from the preamplifiers travels in 6 m long coaxial cables to the feed - throughs .
transceiver boards @xcite are mounted directly on the warm side of the feed - throughs .
they perform pole - zero compensation ( to restore the quasi ideal triangular shape of the signal , with a rise time of @xmath3 22 ns ) , amplification and drive differential output to the digitising electronics via 10 m long shielded twisted pairs cables . the last stage of the electronics @xcite performs the final shaping of the signal using a bessel filter .
the shape of the pulse after this filter is almost symmetrical with a width of 70 ns , risetime of 43 ns and a fall time of 52 ns , with an undershoot of @xmath3 - 3% of the maximum lasting during the total drift time ( @xmath3 3.16 @xmath16s ) .
this signal is fed to a gain switching amplifier , where the gain choice is based on the signal before shaping , and then to a 10 bit 40 mhz flash adc .
the event times are asynchronous with the clock frequency .
four gains are used with relative values of 1,2.86,6.89 and 16.80 , allowing to cover the energy range @xmath3 -1.5 gev to 55 gev per channel . in the highest gain ,
the sensitivity is @xmath3 3.5 mev / count .
the total electronic noise per channel ( dominated by the preamplifier noise ) is @xmath3 10 mev .
the total dispersion of the gains of the electronic chain from channel to channel is around 3% .
the shaped pulse is also used as input of the trigger system @xcite .
the response uniformity and stability of the full electronic chain is controlled by a calibration system .
calibration pulses are produced locally in the liquid krypton so as not to suffer from distortion or cross - talk dues to long transmission lines .
the generator produces a step of injected current followed by an exponential decay .
the amplitude of the signal is given by a reference voltage @xmath30 controlled by a 15 bit dac .
the injected current can be written as @xmath21 = @xmath31 , where the constant @xmath32 depends on the individual components of the calibration circuit .
the intrinsic dispersion of @xmath32 is several percent . to achieve a good calibration accuracy ,
@xmath32 was measured for each channel on a test bench , by equalising the calibration signal , produced for a given dac value , to the signal obtained by directly injecting a calibrated current .
the injected current is generated via an external high precision resistor .
the ratio of these two calibrations gives the calibration constant @xmath32 , expressed in ma / v .
these constants were measured at the level of @xmath3 1% accuracy , for each channel , at room temperature and at 120 k. the technique used simulates the behaviour of the real physics pulses , including the attenuation due to the hv decoupling capacitor , the capacity of the cell , and the capacitive coupling of adjacent cells . the calorimeter assembly has been completed in 1996 when test data with part of the readout electronic were taken .
the calorimeter was fully operational in 1997 for the first data taking period for the @xmath0 measurement . during these two years
, the calorimeter had to be operated at a high voltage of 1500 v , lower than expected , because of problems with a small fraction of the blocking capacitors .
this led to a small space charge effect from accumulation of positive ions in the gap , distorting the electric field @xcite . in the winter 1997/1998
, all the blocking capacitors were replaced .
this allowed to operate the calorimeter at a high voltage of 3000 v , reducing the space charge effect to a negligible level , and allowing a noise reduction of @xmath3 25% ( thanks to the higher drift velocity ) . in 1998,1999 and 2000 , data were taken with this configuration . typically , there are @xmath3 50 faulty channels ( out of @xmath3 13000 in the calorimeter ) .
30 of those are dead preamplifiers in the liquid krypton .
this small mortality happened when the calorimeter was cooled down in 1998 .
this number has been very stable since , the calorimeter being always kept cold and full of krypton .
for each channel readout , the information of 10 adc samples is available .
the timing is adjusted such that the first two samples are before the beginning of the triggered signal .
they can therefore be used to perform a baseline check to determine if the pulse sits on the undershoot of a out - of - time shower . if this is the case , the corresponding baseline is subtracted .
the gain choice is done typically 2 time samples before the maximum of the signal and the same gain is kept for five or more time samples .
thus usually , the same gain value is used for all the samples needed to reconstructed the pulse .
the pulses are converted to energy using a linear relation @xmath33 = @xmath34 @xmath17 ( adc - offset@xmath35 ) derived from calibration events . from this calibrated pulse , the energy ( @xmath36 ) and time ( @xmath37 ) of the signal are reconstructed using a digital filter method : @xmath38 where @xmath39 is the sampled calibrated pulse , and @xmath40 and @xmath41 are digital filter coefficients , derived from the measured pulse shape in calibration events .
these coefficients are binned as a function of the time of the signal , and one iteration is performed to measure both @xmath36 and @xmath37 . to improve the accuracy of the procedure , the channels are divided into 10 categories according to the observed pulse width , and coefficients
are computed for each category . in calibration events ,
the pulse reconstruction accuracy is @xmath3 0.1% on the energy and @xmath29150 ps on the time ( for high enough signals such that the noise becomes negligible ) .
three time samples centred around the maximum are used to perform this digital filter computation .
this is a compromise between noise reduction and sensitivity to accidental showers .
the sample immediately after a gain change is not well measured and in a small fraction of the cases only two samples can be used to reconstruct energy and time .
in this quasi - homogeneous calorimeter , with a moliere radius of @xmath3 6.1 cm , the dominant intrinsic fluctuations in the energy measurement come from fluctuations in the energy loss outside the finite calorimeter area used to measure the shower . to illustrate the importance of this effect , the expected resolution from the geant 3.21 @xcite monte - carlo program
can be computed for various dimensions of the area used to collect the energy . for an infinite size ,
the `` sampling '' term would be @xmath3 1.2%/@xmath42 . using a r=11 cm radius ,
this increases to @xmath3 2.8%/@xmath42 , and @xmath3 3.5%/@xmath42 for a 7 cm radius .
the size used is a compromise between the electronic noise ( which increases with r ) and the sampling term ( which decreases with r ) .
the value used is r=11 cm , which contains @xmath3 100 calorimeter cells .
this value is kept constant with energy in order not to bias the linearity .
only the most energetic central cells are used for the time measurement , and the time of all the other cells is fixed to this time in the digital filter computation , again to avoid bias on the linearity . in events with several showers , some care has to be taken to subtract the energy leakage from one shower to another , up to shower separation as large as @xmath3 60 cm , to eliminate small systematic biases in the energy measurement .
the measurement of the impact point of the electron or photon is based on the centre of gravity of the shower measured in 3@xmath173 cells , to benefit from the narrow central core of the shower .
the position resolution has been studied using data taken with a monochromatic electron beam in 1996 , by comparing the measured electron shower position to the extrapolated track impact point given by the na48 spectrometer .
the resolution of the later quantity is @xmath3 100@xmath16 m and is negligible compared to the calorimeter resolution .
the results achieved are shown in figure [ fig_sigmaxyvse ] , as a function of the electron energy , for both the @xmath43 ( horizontal ) and @xmath44 ( vertical ) coordinates .
the position resolution is better than 1 mm above 25 gev .
the photon time resolution can be studied using events of the type @xmath45 , by comparing the photon time to the time of the electron tracks ( given by a scintillator array in front of the calorimeter ) .
the time resolution per photon is better than 500 ps . in @xmath46 events , the final event time is computed combining the time informations from the four photons . the event time resolution is better than 250 ps .
the tails outside @xmath282 ns are smaller than @xmath47 level , which is a crucial point for the validity of the @xmath6/@xmath5 identification method used by na48 @xcite .
the main tool to study in situ the detailed performances of the calorimeter is given by a sample of @xmath48 decays ( called ke3 decays ) , which is very abundant .
the electron track can be measured in the spectrometer upstream of the calorimeter . this spectrometer @xcite consists of four large drift chambers with a magnet between the second and third chambers ( which gives a @xmath49 kick of 265 mev / c ) .
the position resolution of the drift chambers is better than 100@xmath16 m .
this allows to reconstruct the electron momentum ( @xmath50 ) with a resolution given by : @xmath51 where @xmath50 is expressed in gev .
this measurement of @xmath50 can be compared to the electron energy @xmath36 measured with the calorimeter . in an ideal world
, the ratio @xmath52 should be 1 . taking p as `` perfect ''
, this allows to study the variations in the energy response , the uniformity of the response over the full calorimeter , the energy resolution and the linearity .
a total of @xmath3 150@xmath1710@xmath53 ke3 decays were recorded in 1998 and 1999 for this purpose .
figures [ fig_eop_xcell ] show the variation of @xmath52 as a function of the impact point of the electron within one cell .
the @xmath3 0.5% variation with @xmath43 ( horizontal direction ) comes from the fact that when the electron shower develops near the anode ( @xmath43 @xmath3 0 ) , the response is slightly reduced because part of the deposited charge drifts during a time smaller than the @xmath3 70 ns integration time of the shaping .
thanks to the accordion angle , this effect is small and has a smooth variation with the impact point , it can therefore be easily corrected .
the drop of the response ( @xmath3 1% ) near the vertical separations between cells comes from the fact that in the small vertical gap between two anodes , the electric field is reduced .
this variation can also be corrected using the measured shower position .
both these variations can be reproduced by a monte - carlo which includes a detailed electric field map , as well as a simulation of the readout electronics .
after corrections , the residual variations with the impact point within the cell are @xmath3 0.1% .
the uniformity of the response over the full calorimeter is studied by computing the average value of @xmath52 per cell ( of the maximum energy deposition of the shower ) . using only the electronic calibration , and the values of @xmath32 measured as described above
, the rms dispersion of @xmath54 is 0.41% .
this is well consistent with the expected accuracy in the measurement of @xmath32 of @xmath3 1% , given the fact that @xmath55 40% of the shower energy is deposited in the impact cell . to improve the uniformity , ke3 events are used to equalise @xmath54 for all impact cells .
a simple iterative procedure is used to obtain cell per cell coefficients @xcite . for this procedure ,
only electrons in the energy range 25 to 40 gev are used , to avoid possible interplay between non - linearity and non - uniformity .
this gives intercalibration factors which are one static factor per cell , equivalent to a redefinition of the @xmath32 constants . applying to the 99 ke3 data the intercalibration factors derived from the 98 ke3 sample , the rms dispersion of @xmath54 as
a function of the impact cell becomes @xmath3 0.15% , showing the stability of the intercalibration procedure ( part of this dispersion is directly coming from the event statistic used in 98 to derive the intercalibration factors ) .
the validity of this procedure is also checked using photons produced in @xmath10 and @xmath56 decays , produced with @xmath57 beam striking thin targets during special runs .
the long range uniformity is found to be better than @xmath3 0.1% . from the observed @xmath52 resolution
, the energy resolution can be deduced unfolding the momentum resolution .
figure [ fig_resolution_ke3_99 ] shows the @xmath52 and @xmath36 resolutions in the ke3 sample collected in 99 , after the intercalibration procedure described above . because of the different zero suppression scheme used for ke3 events and @xmath58 events , the energy measurement in ke3 events
is based on 7@xmath177 calorimeter cells .
the corresponding energy resolution extrapolated to the full shower size used for photons is : @xmath59 where e is in gev .
the uncertainties quoted include uncertainties in the momentum resolution and in the extrapolation to the full shower size . above 20
gev , the energy resolution is better than 1% . at the average energy of 25 gev ,
the dominant term in the energy resolution is the sampling term .
the coherent noise contribution to the total noise of @xmath3 90 mev is almost negligible .
several effects are expected to contribute to the observed constant term : a geant based simulation gives a constant term of @xmath3 0.2% ( residual impact point variations , fluctuations in longitudinal shower development , ... ) ; the intercalibration accuracy is @xmath3 0.15% , the gap size variations ( @xmath2845 @xmath16 m ) should give a constant term contribution of @xmath3 0.1 to 0.2% ; the pulse reconstruction has a @xmath3 0.1% accuracy in calibration events , and because the physics signal shape is slightly different from calibration , one expects an additional @xmath3 0.15% constant term contribution .
these effects can account qualitatively for the constant term observed in the data , and this shows that there is not one single dominant contribution to the residual constant term .
if one did not use the ke3 intercalibration , the constant term would increase to @xmath3 0.6% .
the resolution quoted above is the result of a gaussian fit .
some non - gaussian tails are present in the energy response : @xmath3 1% of the showers have a measured energy lower by more than 3@xmath60 from the average value , with
@xmath3 0.1% having a response 20% or more lower .
this tail can be interpreted as coming from @xmath61 production in electromagnetic showers through hadronic photoproduction .
the average value of @xmath52 as a function of the electron momentum is shown in figure [ fig_linearity_ke3_99 ] , where 45 mev has been added to the electron energy to account for energy loss before the liquid krypton .
the energy response is linear in the energy range 5 - 100 gev within @xmath3 0.1% .
this figure also shows the expectation from monte - carlo , which has a small @xmath3 0.05% non linearity coming from the gap opening : as the energy increases , the shower develops later in the calorimeter at a place where the gap is larger and therefore the initial current smaller .
the small structure visible in the data at @xmath36 @xmath3 30 to 50 gev is probably coming from a small non linearity in the adc .
the resolution of the @xmath10 mass in @xmath46 events observed in the data is @xmath3 1 mev , consistent with the expectations based on the energy and position resolutions described above .
thanks to this very good resolution , the level of residual @xmath62 background is smaller than 0.1% and induces a negligible systematic uncertainty in the measurement of @xmath0 . as discussed in the introduction , the decay region definition relies on the measured photon energies and positions : the distance between the reconstructed decay vertex position and the calorimeter is for instance directly proportional to the calorimeter energy scale .
systematic effects from non linearity are small as shown from the ke3 analysis .
one is then only left with fixing the overall energy scale factor of the calorimeter . given the small difference in electron / photon response from the dead matter before the calorimeter , this is done using the @xmath46 events themselves . in the @xmath6 beam ,
an anticounter located at the beginning of the decay region ( 122 m upstream of the calorimeter ) is used to sharply define its beginning , by vetoing decays occurring upstream .
the calorimeter energy scale is adjusted such that the corresponding sharp rise in the reconstructed decay vertex position matches the known geometrical position of this counter with respect to the calorimeter .
this is done with an accuracy of few 10@xmath63 . as a cross - check of non linearity
, one can check that this factor is constant with the kaon energy ( in the range 70- 170 gev ) to better than @xmath64 .
the performances of the na48 liquid krypton calorimeter have been studied extensively in situ .
the resolutions achieved ( @xmath3 500 ps time resolution per photon , better than 1 mm position resolution above 25 gev , better than 1% energy resolution above 20 gev with a constant term smaller than 0.5% ) , together with the small non linearity in the energy response and the good stability of the calorimeter over several years , match the requirements for a precise measurement of direct cp violation in the neutral kaon system .
99 = 11.6pt v. fanti _ et al .
letters b * 465 * , 335 ( 1999 ) . the moliere radius is computed according to the pdg parameterisation : @xmath65 = x0@xmath17(21 mev)/ec with ec = 610 mev/(z+1.24 ) . c. cerri in : a. antonelli _
vi int . conf .
on calorimetry in hep , frascati physics serie 6 , p 841 ( 1996 ) .
g. martin - chassard _ et al . _ , in : a. antonelli _ et al _ , proc .
vi int . conf .
on calorimetry in hep , frascati physics serie 6 , p 799 ( 1996 ) .
b. hallgren _
et al . _ ,
instr . meth .
a * 419 * , 623 ( 1998 ) .
s. guidici , to appear in the same proceedings .
s. palestini _ et al .
instr . meth .
a * 421 * , 75 ( 1999 ) .
cern program library long writeup w5013 , 1993 .
s. crp , thse , lal 98 - 34 , orsay , may 1998 .
d. ddrde _ et al .
. meth . a * 367 * , 88 ( 1995 ) .
j. ocariz , thse , lal 99 - 24 , orsay , may 1999 . | the na48 experiments aims at a precise measurement of direct cp violation in the neutral kaon system .
this puts stringent requirements on the electromagnetic calorimeter used to detect photons of average energy 25 gev .
the choice of na48 is a quasi homogeneous liquid krypton calorimeter with fast readout .
the operation of this device and the performances achieved are described .
= 11.6pt |
one of the greatest success of the chandra x - ray observatory ( cxo ) is the discovery of equatorial tori and polar jet structures in many pulsar wind nebula ( pwn ) systems .
it is now believe that these features are common among young neutron stars . in the @xcite picture , when the highly relativistic pulsar wind decelerates in the external medium , a termination shock is formed at a characteristic scale @xmath2 where @xmath3 is the pulsar spin - down power and
@xmath4 is the filling factor . in general ,
if the pulsar is subsonic in the ambient medium , i.e. @xmath5 for a pulsar speed @xmath6 through a density of @xmath7 , a toroidal shock structure is expected ; faster objects produce bow shock nebulae . indeed
, many young pulsars still inside their high pressure supernova remnant birth sites do show such toroidal symmetry .
the best - known example is the pwn around the as observed by the _ cxo
recently , several relativistic mhd models , e.g. @xcite and @xcite , have shown how such toroidal structure can form if the pulsar wind has a latitudinal variation .
@xcite ( hereafter ) developed a fitting procedure to measure the 3d orientation of the pulsar wind torus and applied to a few x - ray observations . while this simple geometrical model does not capture the fine details of the mhd simulations , is does allow one to extract the torus ( and hence pulsar spin ) orientation from relatively low signal - to - noise ratio ( s / n ) data . also gave quantitative estimates for the statistical errors arising from poisson statistics .
however the systematic errors due to unmodeled components such as jets or background were neglected . for bright objects e.g.the
crab and vela pulsars , the s / n is high and such systematic errors dominate . in this study
, we attempt to quantify these systematic errors and apply the fitting to more _ cxo _ pwn observations , thus providing a more comprehensive study .
in , we developed a generic torus fitting procedure to capture the characteristic structure of pwne . here
we give a brief summary of the model : consider a torus in 3d with radius @xmath8 and gaussian cross section of thickness @xmath9 .
the emissivity is described by @xmath10\ ] ] where @xmath11 is the direction along the torus axis and the observer line of slight lies in the @xmath12-plane .
as the post - shock flow in the pwne is expected to be mildly relativistic , the torus intensity in the observer s frame is doppler boosted by @xmath13 here we assume a constant flow speed @xmath14=@xmath15@xmath16 radially from the pulsar in the equatorial plane .
of course this does not capture the complex latitudinal structure and speed variations seen in the mhd simulations ( e.g. * ? ? ? * ) , but it does give a characteristic speed and captures the gross doppler boosting . during the fitting
we leave @xmath14 as a free parameter , but assume a fixed photon spectral index @xmath17 in the rest frame , which is a typical value for pwne .
we have also tried other values of @xmath18 , e.g. @xmath19 and @xmath20 as reported for the vela and crab pulsars respectively @xcite .
the geometrical parameters are quite insensitive to @xmath18 .
the phenomenological speed @xmath14 is slightly affected ( with @xmath21 ) , so large spectral index variations could induce trends in the fit speed .
altogether this gives @xmath22 \ , \ ] ] where @xmath23 is the total number of counts in the torus and @xmath1 is the inclination of the torus . in order to compare with data ,
the 3-d torus is projected onto the sky plane .
we set up the coordinate system for the image plane with @xmath24 along celestial north , @xmath25 along the observer line of sight and the ccd in the @xmath26-plane .
then the transformation between the frames is given by @xmath27 where @xmath0 is the position angle ( n through e ) of the projected torus axis .
the counts in each ccd pixel ( x , y ) result from the integration of emissivity through the line of sight @xmath28 the basic torus model is characterized by the parameters @xmath0 , @xmath1 , @xmath8 , @xmath9 and @xmath14 . in addition , a point source and constant background are also included in the fit .
the counts in each component are determined by the fit , while the total counts in the model is constrained to match the image counts . in a few cases where
twin tori are observed , there is an extra fitting parameter @xmath29 for the separation . for simplicity
we assume a symmetrical offset along the torus axis , with the pulsar in the middle .
a few other cases have inner and outer tori , at different @xmath8 . to estimate the goodness of fit for particular model parameters , a likelihood function is defined using poisson statistics .
the probability of observing @xmath30 counts out of the expected @xmath31 counts in pixel @xmath32 is given by @xmath33 in the high - count limit , we switch to gaussian statistics for @xmath34 , hence the likelihood function passes to the usual @xmath35 statistics .
the best - fit parameters are then determined by the maximum likelihood method , i.e. minimization of the figure - of - merit function , which is defined as the negative log of the likelihood function summed over all image pixels . once the best - fit parameters are obtained , statistical errors in the fit are estimated by monte carlo simulations .
poisson realizations of the best - fit model are simulated and fitted , and the confidence interval of each parameter is obtained from their distributions . to estimate the systematic errors , we blank out different regions in the data ( i.e. omit those pixels from the fit ) , then re - fit the model by varying only one parameter at a time while fixing the others at their best - fit values
. the changes in the best - fit parameter values thus provide estimates of the uncertainties contributed by un - modeled features in the data .
note that in addition to the single parameter systemetic errors computed here , covariance between the parameters could also lead to larger errors ( e.g. the covariance of @xmath14 with @xmath18 ) . in this study we went through the _ cxo _ data archive and identified a list of observations with apparent toroidal termination shocks in pwne ( table [ tab1 ] ) .
there are a few other objects with jets and/or possible tori ( e.g. psr b1509@xmath3658 ) , but for these cases closer examinations suggest that the pulsar motions maybe trans- or super - sonic with respect to the surrounding medium , hence static torus fits are not appropriate .
the data are cleaned up with standard reduction procedures using ciao and the background light curves are examined to filter out periods that suffered from strong particle flares .
then we remove the pixel randomization for the acis observations and apply an algorithm by @xcite , which corrects the positions of the split pixel events , to improve the spatial resolution .
finally we select the appropriate energy range and binning to optimize the s / n of the pwne , separating out the soft background from the snr and field stars .
to model the point source , a high s / n psf model is simulated with the chandra ray tracer ( chart ) at the detector source position , using the source spectrum reported in literature .
cccccc j0205 + 6449 & 3c 58 & , , & acis & 350 + j0537@xmath366910 & n157b & & acis & 50 + b0540@xmath3669 & snr [email protected] & , , , , , , & hrc & 75 + j1124@xmath365916 & g292.0 + 1.8 & & hrc & 50 + b1800@xmath3621 & & & acis & 30 + j1833@xmath361034 & [email protected] & , , , , & acis & 56 in the fitting procedure , a simulated annealing package @xcite is employed for multidimensional minimization of the figure - of - merit function , with the initial parameters chosen by eye .
after the best - fit model is obtained , we simulate 500 monte carlo poisson realizations and estimate the statistical errors from the parameter distributions accordingly .
the @xmath37 confidence interval , which corresponds to 68% confidence level , are reported . to estimate the systematic errors , we begin by blanking out the jet regions in the observations ,
since they are the most obvious un - modeled features and are present in many systems . as an example
, the image of psr b0540@xmath3669 with jet regions removed is shown in figure [ fig2 ] . the golden section search in one dimension @xcite is then employed for re - fitting the model .
this procedure gives systematic error estimates for the crab , , psrs , , and .
however for psrs and , the jets are relatively faint and well - separated from the pwne .
therefore , removing the jets alone does not change the parameter values significantly .
we believe this underestimates the systematic errors and contributions from other un - modeled features should also be considered .
therefore , we obtained another estimate by removing the point source in the data .
the results reported for these two objects are the combination of the two systematic error estimates , added in quadrature . in the following
we discuss the fits of some individual objects .
_ psr b0540@xmath3669 : _ the combined hrc exposure shows polar jet structure in the pwn .
in addition , a bright blob sw of the point source is also observed .
we have tried blanking out the blob in the fit , and found that the result is insensitive . _ psr j0205 + 6449 : _ similar to the crab pulsar , a double tori structure is observed in this pwn .
however , in this case the outer torus is very diffuse and without a clear boundary . in order to have a better constraint in the position angle
, we fit the two tori simultaneously with the same value of @xmath0 . _
_ while the pwn image in the initial _ cxo _ hrc observation showed noticeable ellipticity @xcite , the elongation was not clear in a subsequent acis exposure .
this could possibly due to the large off - axis angle degrading the psf .
therefore in this study we fitted the hrc image only .
a scheduled deep acis observation could help to resolve this structure .
_ _ a cometary trail is observed in this pwn system @xcite , which is modeled in the fit by a rectangular background region with constant brightness .
comparison with the best - fit parameters without the background provides an estimate for the systematic errors . _ _ although the 30ks acis exposure only detects a total of @xmath38200 counts ( 0.8 - 6kev ) for the point source and the pwn , the latter clearly shows extended structure with a well - defined symmetry axis .
interestingly , the direction of elongation is in general orthogonal to the proper motion claimed in @xcite .
hence , employing the torus fitting to obtain a quantitative measurement of the @xmath0 accuracy is valuable . _ _ as it s host snr is one of the calibration objects for _ cxo _ , the pulsar has been observed repeatedly .
we combined all the on - axis and in - focus acis observations in the data archive and obtain a total exposure of 56ks .
_ psr j0538@xmath362817 : _ in a short 20ks acis exposure , diffuse emission around the pulsar was fit to a simple equatorial flow .
however , while a deeper 100ks observation @xcite showed extended emission along a similar axis , it was not unambiguously possible to decide whether this was an edge - on equatorial torus or a polar jet . to be conservative , we do not include this object in the present analysis .
table [ tab2 ] lists the best - fit results with corresponding statistical and systematic errors .
we report the systematic errors for @xmath39 and the separation only , since they are the most interesting parameters .
systematic errors in other parameters have similar fractional values when compared to the best - fit parameters .
note that for the faint objects , including psrs j1124@xmath365916 , b1800@xmath3621 and j1833@xmath361034 , the detailed structures are not resolved and the large statistical errors dominate over the systematic errors .
therefore only the former are reported here .
the best - fit models for the new objects in this study are shown in figure [ fig3 ] , the rest may be found in , figure 2 and the references listed in table [ tab2 ] . lcccccccc crab ( inner ) & @xmath40 & @xmath41 & @xmath42 & 1.5 * & @xmath43 & & /@xmath44 & 1 + crab ( outer ) & @xmath45 & @xmath46 & @xmath47 & 5.9 * & @xmath48 & & /@xmath49 & 1 + vela & @xmath50 & @xmath51 & @xmath52 & 3.0 * & @xmath53 & @xmath54 & /@xmath55 & 1 + j1930 + 1852 & @xmath56 & @xmath57 & @xmath58 & @xmath59 & @xmath60 & & 1701/602 & 1 + j2229 + 6114 & @xmath61 & @xmath62 & @xmath63 & 2.5 * & @xmath64 & & 2221/1113 & 1 + b1706@xmath3644 & @xmath65 & @xmath66 & @xmath67 & 1.0 * & @xmath68 & & 2897/1221 & 2 + j2021 + 3651 & @xmath69 & @xmath70 & @xmath71 & 1.2 * & @xmath72 & @xmath73 & 235/751 & 3 + j0205 + 6449 ( inner ) & @xmath74 & @xmath75 & @xmath76 & 0.74 * & @xmath77 & & 5080/39730 & 4 + j0205 + 6449 ( outer ) & @xmath74 & @xmath78 & @xmath79 & 2.0 * & @xmath80 & & 5080/29503 & 4 + j0537@xmath366910 & @xmath81 & @xmath82 & @xmath83 & @xmath84 & @xmath85 & & 102/179 & 4 + b0540@xmath3669 & @xmath86 & @xmath87 & @xmath88 & 0.66 * & @xmath89 & & 13442/50215 & 4 + j1124@xmath365916 & @xmath90 & @xmath91 & @xmath92 & @xmath93 & @xmath94 & & 390/552 & 4 + b1800@xmath3621 & @xmath95 & @xmath96 & @xmath97 & 0.74 * & @xmath98 & & 18/88 & 4 + j1833@xmath361034 & @xmath99 & @xmath100 & @xmath101 & @xmath102 & @xmath103 & & 304/6541 & 4 as shown in the table , the fits for crab and vela are dominated by systematic errors and the statistical errors are negligible .
for the other objects , the two are comparable in @xmath0 and @xmath8 , but the systematic errors for @xmath1 are in general much larger .
this suggests the fit is most robust in obtaining @xmath0 and @xmath8 , as they are largely determined by the shape of the torus . on the other hand
, @xmath1 depends sensitively on the relative brightness of both sides of the torus ; thus , removing the jet structure can change the results significantly . as mentioned in the previous section , polar jets
are observed in many of the pulsars in this study . in general , the jets position angle is in reasonable agreement with @xmath0 axis determined from the torus geometry . since these are also relativistic , relative doppler boosting could in principle provide a second estimate of spin inclination @xmath1 and flow speed @xmath14 .
however , two effects complicate this .
first , @xmath14 might be quite different in the jets as suggested in the mhd simulations @xcite .
second , the jets are in several cases distinctly curved , and repeated exposures of crab and vela indicate that this bending in the plane of the sky changes with time .
since we expect similar curvature along the line - of - sight , and even quite small changes in inclination strongly affect the brightness ratios , interpretation of the jet boosting is more difficult than that of the tori .
one direct application of the torus fitting is measurement of the neutron star s spin axis orientation .
the alignment angle between the pulsar spin and velocity vectors provides insights into the supernova core collapse physics . for a simple picture of a kick acting at a fixed spot on the neutron star surface ,
a longer kick duration gives a better alignment as a result of rotational averaging .
@xcite performed simulations of neutrino kick models and extracted constraints on the physics parameters including the kick direction , magnitude and timescale by comparing torus fitting angles with proper motion estimates and other pulsar data . as the sample of x - ray pwne with modeled torus structure increases ,
the statistical constraints on such kick models should rapidly improve .
there are a few objects in our list which also have independent estimates for the spin geometry from radio polarimetry .
@xcite , in particular , have recently made considerable progress in measuring @xmath0 by determining the polarization at the phase of closest approach to the magnetic pole and correcting for faraday rotation to obtain the absolute position angle on the sky .
unfortunately , radio pulsar emission is complex and can switch between orthogonal modes ; this implies a @xmath104 ambiguity in the spin axis orientation .
nevertheless , careful analysis can give useful position angles .
at present only three of our pulsars have a @xmath0 from radio measurements .
@xcite observed the radio polarization for the vela pulsar and obtain a p.a . of @xmath105 . for psr b1706@xmath3644 , the result is controversial : while @xcite report a p.a . of @xmath106 ,
@xcite do not give a value , since in new observations the phase of closest approach is difficult to locate .
finally , @xcite measure the radio polarization of the pwn in psr j2229 + 6114 and suggest a symmetry axis at p.a.@xmath107 . while these three values are in reasonable agreement with the @xmath0 from the x - ray symmetry axis , more data are needed for a serious comparison .
unfortunately , interpretation of the polarimetry appears to be particularly challenging for the young pulsars showing x - ray tori ( this may be related to their powerful accelerators in relatively small magnetospheres ) .
for example , the main radio pulses in the crab appear at the same phases as the high energy emission and may indicate high - altitude emission .
a classic polarization sweep may only be found for the ` precursor ' , a low frequency component .
however , measurement of radio polarization for more objects would be very useful .
comparison with the x - ray p.a.s can reveal orthogonal emission in the radio , e.g. @xcite , and may help to calibrate the radio polarization techniques , which may then be applied to older pulsars .
psrs j2020@xmath1083651 and b1800@xmath3621 seem amenable to such analysis . in principle , for a simple dipole geometry it is also possible to measure the magnetic pole impact parameter @xmath14 ( from the polarization pa sweep rate at closest approach ) and the magnetic inclination @xmath109 ( from the overall shape of the polarization position angle sweep ) , when the polarization is well measured over a large range of pulsar phase . in practice , while @xmath14 can be fairly well determined , the @xmath109 estimates are often poor and subject to large systematic errors .
the rotating vector model ( rvm ) for such fits was developed for vela , so not surprisingly there are angles for this pulsar .
table 3 gives the values from @xcite .
however , the sum @xmath110 is in poor agreement with the x - ray inclination of @xmath111 .
it is difficult to interpret the _
arc structure seen for vela with such a small inclination .
perhaps the best approach is to adopt @xmath1 from the pwn structure , when available , and @xmath14 from radio polarization .
for vela this would imply @xmath112 .
radio ( coherent emission ) pulse profiles are often quite complex , and may depend on the dense plasma and complex field structure at their relatively low emission altitudes .
high energy ( incoherent emission ) profiles should be simpler , especially if they originate at high altitude where only dipole field components are expected to persist .
@xcite found that for the generic class of outer magnetosphere models , the @xmath113-ray pulse is usually a double peak , with a peak separation @xmath114 and phase lag @xmath9 from the radio pulse primarily determined by @xmath1 , although the actual values do depend on @xmath109 and the gap width , as well .
other models , employing emission from both hemispheres ( e.g. * ? ? ?
* ) can have more complex light curves with different phase relationships . however , there are some generic predictions of these models : @xmath113-ray peaks are strong and wide when @xmath115 , they converge as @xmath1 approaches @xmath116 , and are weak or absent for small angles . in table
[ tab3 ] , we list the observed gev pulse widths ( @xmath117 ) for crab , vela and psr b1706@xmath3644 and tentative widths for two other pulsars discovered after the end of the egret mission . the final column ( @xmath118 ) lists approximate high energy pulse widths ( or ranges ) from the contours in @xcite , figure 4 . in most cases , a range is allowed independent ( e.g. radio ) values for @xmath14 would make width predictions .
note that for crab and vela , this model tends to predict somewhat narrower pulses than those seen .
also the tentative detection of psr j2229 + 6114 has a wider pulse than predicted for its small @xmath1 .
these larger angles may be easier to accommodate in two - pole models . at present
there is sufficient latitude in the model predictions to allow a wide range of pulse shapes .
however , new @xmath113-ray observations with agile and , especially , glast should provide many high quality pulse profiles . we expect to have x - ray measurements of @xmath1 for many of these young pulsars .
if radio observations can supply @xmath14 , the geometry is determined and comparison with the @xmath113-ray data can powerfully refine ( or eliminate ) the magnetosphere models .
cccccccc crab & 124 & 61 & ? & & & 0.39 & 0 - 0.35 + vela & 131 & 64&@xmath119 & 137 ( 43 ) & -6.5 & 0.42 & 0.35 - 0.4 + j1930 + 1852 & 91&147 & & & & & 0 + j2229 + 6114 & 103 & 46&@xmath120 & & & ( @xmath380.4 ) & @xmath121 + b1706@xmath3644 & 164 & 53&@xmath122 & & & 0.25 & 0 - 0.3 + j2021 + 3651 & 45 & 79 & & & & ( @xmath380.5 ) & 0 - 0.45 + j0205 + 6449 & 90 & 92 & & & & & 0 - 0.55 + j0537@xmath366910&131 & 93 & & & & & 0 - 0.55 + b0540@xmath3669 & 144 & 93 & & & & & 0 - 0.55 + j1124@xmath365916 & 16&105 & & & & & 0 - 0.45 + b1800@xmath3621 & 44 & 90 & & & @xmath123 & & @xmath380.5 + j1833@xmath361034 & 45 & 85 & & & & & 0 - 0.55 +
in conclusion , we have applied the torus fitting technique to more pwn observations in the _ chandra _ data archive and characterized the uncertainties in the fits .
this study provides a better understanding of the systematic errors , giving quantitative estimates of the measurement uncertainties .
we argue that these robust position angle @xmath0 and inclination @xmath1 values are particularly useful for comparison with the radio and high energy pulse data . if new observations can fill in more measurements from these energy bands in table 3 , we should be able to make substantial progress in understanding the emission zones and viewing geometries of young pulsars .
this work was supported by nasa grant nag5 - 13344 and by chandra grant ar6 - 7003 issued by the chandra x - ray center , which is operated by the smithsonian astrophysical observatory for and on behalf of the national aeronautics space administration under contract nas8 - 03060 .
mori , k. , tsunemi , h. , miyata , e. , baluta , c. j. , burrows , d. n. , garmire , g. p. , & chartas , g. 2001 , in asp conf .
251 , new century of x - ray astronomy , ed .
h. inoue & h. kunieda ( san francisco : asp ) , 576 | we have applied the torus fitting procedure described in ng & romani ( 2004 ) to pwne observations in the _ chandra _ data archive .
this study provides quantitative measurement of the pwn geometry and we characterize the uncertainties in the fits , with statistical errors coming from the fit uncertainties and systematic errors estimated by varying the assumed fitting model .
the symmetry axis @xmath0 of the pwn are generally well determined , and highly model - independent .
we often derive a robust value for the spin inclination @xmath1 .
we briefly discuss the utility of these results in comparison with new radio and high energy pulse measurements . |
the entanglement in extended quantum systems and the ways to quantify it has attracted a lot of research during the last decade in different areas of theoretical physics like condensed matter , quantum information and quantum gravity ( see @xcite for reviews ) . given a quantum system in a pure state @xmath0 and assuming that its hilbert space is bipartite , i.e. @xmath1 , the entanglement entropy is an important quantity to measure the entanglement between the two parts of the system . in this manuscript we consider only spatial bipartitions , denoting by @xmath2 a generic spatial region and by @xmath3 its complement .
the @xmath2 s reduced density matrix @xmath4 is obtained by tracing over @xmath5 the density matrix @xmath6 of the entire system .
the normalisation condition is @xmath7 and the von neumann entropy of @xmath4 defines the entanglement entropy , namely [ sa def ] s_a = - ( _ a _ a ) .
the entanglement entropy @xmath8 obtained from @xmath3 s reduced density matrix can be introduced in the same way .
when @xmath9 is a pure state , we have that @xmath10 . important quantities to study the bipartite entanglement of pure states are also the rnyi entropies , which are defined as follows [ renyi entropies def ] s_a^(n ) = _ a^n , where @xmath11 is an integer parameter .
the entanglement entropy ( [ sa def ] ) can be found from the rnyi entropies ( [ renyi entropies def ] ) as @xmath12 .
this limit requires to perform an analytical continuation of @xmath13 to real values of @xmath14 close to @xmath15 . for a pure state @xmath16 for any value of @xmath14 .
when the subsystem @xmath17 is made by two regions , which can be either disjoint or adjacent , an interesting quantity to consider is the mutual information , i.e. [ mi def ] i_a_1 , a_2 = s_a_1 + s_a_2 - s_a_1 a_2 = _
n 1 i_a_1,a_2^(n ) , where @xmath18 is the corresponding combination of rnyi entropies .
the mutual information measures the total amount of correlations between the two systems @xcite .
for some quantum systems on the lattice , the entanglement entropy @xmath19 grows like the area of the boundary of the subsystem @xmath2 as its size increases @xcite .
this area law behaviour of @xmath19 has been proved for gapped systems on the lattice @xcite , but for critical systems the situation is more complicated : important exceptions are the critical systems in one spatial dimension , where @xmath19 diverges logarithmically with the length of the interval @xmath2 @xcite , and free fermions in higher dimensions @xcite .
when the continuum limit is described by a quantum field theory , an ultraviolet ( uv ) cutoff @xmath20 must be introduced and @xmath19 is a divergent quantity for @xmath21 .
the area law behaviour can occur in the coefficient of the most divergent term , which is non universal and turns out to be proportional to the area of @xmath22 ( i.e. the boundary of @xmath2 ) in these cases .
a quantum system in the continuum at criticality is described by a conformal field theory ( cft ) . considering a @xmath23 dimensional cft on a line at zero temperature and an interval @xmath2 of length @xmath24
, we have that @xmath25 , where @xmath26 is the central charge of the model @xcite . instead
, the mutual information ( [ mi def ] ) of two disjoint intervals is uv finite and it depends on the full operator content of the model @xcite . performing the replica limit to get analytic expressions for @xmath19 and @xmath27 from the ones for @xmath13 and @xmath28 can be a very difficult task ( see @xcite for a numerical approach ) . in @xmath29 dimensional cfts and for domains
@xmath2 whose boundary is smooth , the expansion of the rnyi entropies reads @xmath30 as @xmath31 , where @xmath32 is the perimeter of @xmath2 and the coefficient @xmath33 depends on the model and on the details of the uv regularisation .
the replica limit implies that @xmath34 as @xmath31 , where @xmath35 .
when the two dimensional spatial domain @xmath2 has a non smooth boundary , @xmath36 contains also a subleading logarithmic term whose coefficient is independent of the regularisation details .
such coefficient is obtained as the sum of the contributions of the vertices of the curve @xmath22 , where each term is given by a model dependent function ( corner function ) evaluated on the opening angle in @xmath2 of the corresponding vertex .
many interesting studies have been done on such corner functions in various lattice models @xcite . for a cft in @xmath29 dimensions , it has been recently found that the leading term in the expansion of the corner function as the opening angle @xmath37 is close to @xmath38 provides the constant characterising the two - point function of the stress tensor @xcite . within the context of the ads / cft correspondence ,
the prescription to compute holographically the entanglement entropy in the regime where classical gravity can be employed has been found in @xcite . in the case of ads@xmath39/cft@xmath40 ,
such holographic prescription provides also the expected subleading logarithmic divergence whenever @xmath22 contains vertices @xcite .
the mutual information ( [ mi def ] ) of disjoint domains with smooth boundaries is a uv finite quantity because the area law terms cancel .
when the separation between the regions is large with respect to their sizes , analytic results have been found @xcite . in the case of two dimensional adjacent domains
@xmath41 and @xmath42 , both the mutual information and its generalisation involving the rnyi entropies display an area law behaviour in terms of the length of the curve shared by the adjacent regions .
in particular , @xmath43 and @xmath44 as @xmath31 , where @xmath45 . the entanglement entropy and
the rnyi entropies are measures of the quantum entanglement for a bipartite system in a pure state , but this is not true when the whole system is in a mixed state .
for instance , the mutual information of a bipartite system in a thermal state is dominated by classical correlations .
another important example of mixed state is the reduced density matrix @xmath4 associated to a subsystem @xmath2 , when the entire system is in its ground state . splitting @xmath2 in two domains @xmath41 and @xmath42 , which can be either adjacent or disjoint , it is worth considering the bipartite entanglement between them .
many measures of the bipartite entanglement for a mixed state have been proposed in quantum information theory , but they are usually very difficult to compute , even for small systems . a measure which is computable also for extended systems is the logarithmic negativity @xcite . let us consider a mixed state characterised by the density matrix @xmath9 acting on a spatially bipartite hilbert space @xmath46 .
we remark that @xmath47 can be either the hilbert space characterising the whole spatial system or the one @xmath48 associated to the bipartite subsystem @xmath49 introduced above ( in the latter case @xmath50 ) .
the logarithmic negativity is defined through the partial transpose of @xmath9 with respect to one of the two parts .
considering e.g. the partial transposition with respect to @xmath42 , the matrix element of @xmath51 is given by [ partialtransp ] e_i^(1 ) e_j^(2)| ^t_2| e_k^(1 ) e_l^(2)= e_i^(1 ) e_l^(2)| | e_k^(1 ) e_j^(2 ) .
since the spectrum of the hermitian matrix @xmath51 can contain also negative eigenvalues , it is worth computing its trace norm @xmath52 .
the logarithmic negativity is defined as [ logneg def ] |^t_2| .
the logarithmic negativity can be computed also by employing a replica limit @xcite .
considering the @xmath14-th moment @xmath53 of the partial transpose and taking into account only the sequence of the even powers @xmath54 , it is not difficult to realise that ( [ logneg def ] ) can be found by performing the following analytic continuation [ replica limit neg ] = _ n_e 1 ( ^t_2 ) ^n_e . in the special case of a pure state @xmath55 acting on a bipartite hilbert space ,
the moments of the partial transpose are related to the rnyi entropies as follows @xcite [ identity pure states ] ( ||^t_2 ) ^n = \ { ll _ a_2^n_o & , + ' '' '' ( _ a_2^n_e/2 ) ^2 & , where @xmath56 is the reduced density matrix of the subsystem @xmath42 . from the relation ( [ identity pure states ] ) and the replica limit ( [ replica limit neg ] ) , one easily gets that @xmath57 .
the logarithmic negativity and the moments @xmath58 of the partial transpose are interesting quantities to compute for bipartite mixed states . in this manuscript
we focus on a particular system in its ground state , considering the mixed state given by the reduced density matrix @xmath59 of a spatial subsystem , whose corresponding hilbert space @xmath60 is bipartite . instead of the moments of @xmath61
, we find more convenient to consider the following quantity [ en def ] _
it is not difficult to show that @xmath62 . given the normalization condition of @xmath4 , the replica limit ( [ replica limit neg ] ) tells that [ replica limit e_n ] = _ n_e 1 _ n_e . for @xmath23 dimensional cfts in the ground state , the logarithmic negativity and the moments of the partial transpose have been studied for both adjacent and disjoint intervals @xcite .
this analysis has been extended also to a bipartite system at finite temperature @xcite .
the moments of the partial transpose for some fermionic systems on the lattice have been studied through a method involving correlators in @xcite and the overlap matrix in @xcite .
the logarithmic negativity has been considered also for a non vanishing mass @xcite and out of equilibrium @xcite .
other interesting numerical studies for various one dimensional lattice systems have been performed in @xcite .
the same numerical method employed to get the mutual information from the replica limit has been used to get the logarithmic negativity from the replica limit ( [ replica limit neg ] ) , since similar difficulties occur @xcite .
in two spatial dimensions , the logarithmic negativity of topological systems has been considered @xcite and recently interesting lattice analysis have been performed for both fermionic and bosonic systems @xcite .
some results have been found also in the context of holography @xcite . in this paper
we consider a two dimensional square harmonic lattice with nearest neighbour interactions in its ground state .
we focus on the regime of massless oscillators , whose continuum limit is described by the cft given by the massless scalar field in @xmath29 dimensions . in the thermodynamic limit , we study the logarithmic negativity and the quantity ( [ en def ] ) for various configurations of adjacent domains in the regime where they become large . at leading order ,
these quantities follow an area law behaviour in terms of the length of the curve shared by the adjacent regions .
this observation for the logarithmic negativity has been already done for this model in @xcite , where the configuration given by two halves of a square has been considered .
we notice that the coefficient of the area law term is related to the coefficient of the area law term in the rnyi entropies .
we study also the subleading logarithmic term , which occurs whenever the curve shared by the adjacent regions contains vertices .
such term is very interesting because it is independent of the regularisation details .
the layout of this manuscript is as follows . in [
sec : hc ] we review the method to compute @xmath19 , @xmath13 , @xmath63 and @xmath64 for this bosonic lattice . in
[ sec : area law ] we investigate the area law behaviour in the leading term of @xmath63 and @xmath64 for various configurations of large adjacent domains in the infinitely extended lattice . in
[ sec : corners ] we study the subleading logarithmic term of @xmath63 due to the occurrence of vertices in the curve shared by the adjacent domains and in [ sec : conclusions ] we draw some conclusions .
in this section we introduce the lattice model considered throughout this manuscript , its correlators in the thermodynamic limit and their role in computing the entanglement entropies , the moments of the partial transpose and the logarithmic negativity .
we consider the two dimensional square lattice made by harmonic oscillators coupled through the nearest neighbour spring - like interaction .
denoting by @xmath65 and @xmath66 the number of sites ( oscillators ) along the two orthogonal directions , such lattice contains @xmath67 oscillators .
the hamiltonian of the model reads [ ham ] h= _
1i l_x 1j l_y \ { + q^2_i , j + } , where the pair of integers @xmath68 identifies a specific lattice site .
the canonical variables @xmath69 and @xmath70 satisfy the canonical commutation relation @xmath71 = [ p_{i , j } , p_{r , s } ] = 0 $ ] and @xmath72 = \textrm{i}\,\delta_{i , r } \delta_{j , s}$ ] .
we assume periodic boundary conditions along both the spatial directions , namely @xmath73 , @xmath74 , @xmath75 and @xmath76 for a generic integer @xmath77 .
the model described by ( [ ham ] ) contains three parameters @xmath78 , @xmath79 and @xmath80 , but not all of them are independent .
indeed , by performing the canonical rescaling @xmath81{mk } q_{i , j } , \ , p_{i , j}/\sqrt[4]{mk } ) $ ] and introducing @xmath82 , the hamiltonian ( [ ham ] ) becomes [ ham2 ] h= _
1i l_x 1j l_y \ { + q^2_i , j + } . from this expression
, one can easily observe that gives the hamiltonian of a free scalar field with mass @xmath78 in two spatial dimensions discretised on a square lattice with lattice spacing @xmath83 .
the continuum limit corresponds to take simultaneously the limits @xmath84 , @xmath85 and @xmath86 , while @xmath87 and @xmath88 are kept fixed . in our lattice computations , without loss of generality , we set @xmath89 .
the hamiltonian ( [ ham ] ) can be diagonalised in a standard way , finding the following dispersion relation [ dispersion relation ] _ , where @xmath90 is a pair of integers such that @xmath91 and @xmath92 .
because of the translation invariance of the model , the zero mode with @xmath93 occurs , for which the equality holds in ( [ dispersion relation ] ) . in our analysis
we need the following vacuum correlators [ qq correlator ] & & 0 | q_i , j q_r , s| 0 = _ 0k_x < l_x 0k_y <
l_y , + [ pp correlator ] ' '' '' & & 0|p_i , j p_r , s|0 = _
0k_x < l_x 0k_y < l_y _ , which are the matrix elements of the correlation matrices @xmath94 and @xmath95 respectively ( where @xmath68 and @xmath96 are the raw and column indices respectively ) .
these matrices satisfy @xmath97 , being @xmath98 is the identity matrix .
we remark that the term in corresponding to the zero mode reads @xmath99 , which is divergent for @xmath100 .
this implies that we can not take @xmath101 in a finite lattice .
since from our computations on the lattice we would like to extract information about the model in the continuum limit , we need to consider the regime where @xmath102 , being @xmath24 the linear size of the subsystem .
thus , it is convenient to consider the thermodynamic limit , where @xmath103 and @xmath85 , while the lattice spacing @xmath83 is kept finite . in order to perform the thermodynamic limit of the correlators and
, we define @xmath104 for @xmath105 . in the thermodynamic limit @xmath106
becomes a continuous variable @xmath107 and the sum in ( [ qq correlator ] ) and ( [ pp correlator ] ) is replaced by an integration according to @xmath108 .
thus , the correlators ( [ qq correlator ] ) and ( [ pp correlator ] ) in the thermodynamic limit become respectively [ qq correlator cont ] & & 0 | q_i , j q_r , s| 0 = _ 0 ^ 2 dq_x dq_y , + [ pp correlator cont ] ' '' '' & & 0|p_i , j p_r , s|0 = _ 0 ^ 2 _
dq_x dq_y , where @xmath109}$ ] , with @xmath110 .
when @xmath111 the integral in ( [ qq correlator cont ] ) is convergent and therefore , in principle , the massless regime can be considered without any approximation .
nevertheless , in order to avoid divergent integrands , in our numerical calculations we have set @xmath112 , checking in some cases that smaller values of @xmath78 do not lead to significant changes in the final result .
following @xcite , we can compute the rnyi entropies @xmath13 for this model by considering the matrices @xmath113 and @xmath114 , which are obtained by restricting @xmath94 and @xmath95 respectively to the subsystem @xmath2 .
their size is @xmath115 , being @xmath116 the number of lattice points inside the region @xmath2 .
the matrix product @xmath117 has positive eigenvalues @xmath118 with @xmath119 and the moments of the reduced density matrix are given by [ moments rdm ] _
a^n = _ j=1^n_a
^-1 . from this expression
it is straightforward to get the rnyi entropies s_a^(n ) = _ a^n = _ j=1^n_a , while the entanglement entropy is given by s_a = _ j=1^n_a . by employing these formulas for disjoint domains on the lattice ,
one gets @xmath120 and the mutual information @xmath121 . in @xcite
it was shown that the partial transposition with respect to @xmath42 for a bosonic state corresponds to the time reversal applied only to the degrees of freedom in @xmath42 , while the remaining ones are untouched .
in particular , in @xmath42 the positions are left invariant @xmath122 everywhere , while the momenta are reversed @xmath123 if @xmath124 . given a bosonic gaussian state , like the ground state of the harmonic chain we are considering , the resulting operator after such transformation will be gaussian as well .
it is worth remarking that this is not true for fermionic systems .
for instance , for free fermions the partial transpose of the ground state density matrix can be written as a sum of two gaussian operators @xcite .
the above observations are implemented on our lattice model by introducing the following matrix _
a^t_2 = _ a_2 _ a _
a_2 , where @xmath125 is the @xmath126 diagonal matrix having @xmath127 in correspondence of the sites belonging to @xmath42 and @xmath128 otherwise . since @xmath129 , it is easy to observe that @xmath130 , as expected .
the matrix @xmath131 has also positive eigenvalues @xmath132 , but if the state is entangled some of them can be smaller than @xmath133 . from the eigenvalues
@xmath134 one gets the moments of the partial transpose of the reduced density matrix as in ( [ moments rdm ] ) for the moments of the reduced density matrix , namely ( _ a^t_2)^n = _
j=1^n_a ^-1 .
the trace norm of @xmath61 reads _
j=1^n_a ^-1 = _ j=1^n_a ( 1 , ) , which leads straightforwardly to the logarithmic negativity [ eq : neg lattice ] = _ j=1^n_a .
in the remaining part of the manuscript we discuss the numerical results obtained by employing the above lattice formulas .
in this section we consider various configurations of adjacent domains for the harmonic lattice in the thermodynamic limit described in [ sec : hc ] . for large domains ,
we show that at leading order the logarithmic negativity and @xmath64 satisfy an area law in terms of the length of the curve shared by the adjacent domains .
we observe that the coefficient of such term in these quantities is related to the coefficient of the area law term in the rnyi entropies .
we begin our analysis by considering the logarithmic negativity of two equal adjacent rectangles @xmath41 and @xmath42 which share an edge along the vertical @xmath135 axis , as shown in the inset of the left panel in fig.[fig : adjrect ] , where the adjacent domains are highlighted by blue dots and red circles .
these rectangles have the natural orientation induced by the underlying lattice , namely their edges are parallel to the vectors generating the square lattice . denoting by @xmath136 and @xmath137 the lengths of the edges along the @xmath138 and @xmath135 directions respectively , the numerical data for @xmath63 of this configuration of adjacent domains are plotted in fig.[fig : adjrect ] . between two equal rectangles ,
whose edges have lengths @xmath136 and @xmath137 , which are adjacent along the vertical edge ( inset of the left panel ) .
left : for fixed values of @xmath137 , the ratio @xmath139 reaches a constant value as @xmath136 increases .
right : for fixed and large enough values of @xmath136 , the logarithmic negativity grows linearly as @xmath137 increases ( the dashed line is obtained by fitting all the data corresponding to @xmath140 ) .
, scaledwidth=115.0% ] in the left panel we show the ratio @xmath139 as function of @xmath136 when @xmath137 is kept fixed . for any given @xmath137 , such ratio reaches a constant value when @xmath136 is sufficiently large .
this confirms the intuition that the main contribution to a quantity characterising the entanglement between two adjacent regions @xmath41 and @xmath42 should come from the degrees of freedom localized along their shared boundary , namely the curve @xmath141 . in the right panel of fig.[fig : adjrect ] , the logarithmic negativity of the same configuration is plotted as function of @xmath137 for fixed values of @xmath136 . if @xmath136 is sufficiently large , a neat linear growth can be observed .
the fact that the asymptotic value of @xmath139 depends on @xmath137 in the left panel of fig.[fig : adjrect ] is mainly due to the subleading corner contributions , which will be largely discussed in [ sec : corners ] .
these results tell us that , given two equal and large enough adjacent regions @xmath41 and @xmath42 , at leading order the logarithmic negativity @xmath63 between them increases like the length of the curve @xmath141 shared by their boundaries as their size increases
. such length will be denoted by @xmath45 throughout this manuscript .
thus , the logarithmic negativity between large adjacent domains satisfies an area law in terms of the region shared by their boundaries .
this observation has been recently done for this model also by eisler and zimbors @xcite , who have considered the logarithmic negativity between the two halves of a square as the length of its edge increases . in order to improve our analysis of the area law behaviour for
the logarithmic negativity between adjacent regions @xmath41 and @xmath42 , let us consider the six configurations of adjacent domains on the lattice shown in fig.[fig : configs1 ] , where the sites belonging to @xmath41 and @xmath42 are highlighted by blue dots and red circles .
in these examples the curve @xmath141 is not given by a simple line segment .
the domains identified by the red circles in fig.[fig : configs1 ] are convex , while the ones corresponding to the blue dots are not . and figs.[fig : negarea ] and [ fig : renyi ] ) and the corner contributions for explementary angles ( see [ sec : corners ] and fig.[fig : cornernegold ] ) . , scaledwidth=108.0% ]
it is well known that the curve separating adjacent domains on the lattice is not unique . for these configurations
we have chosen the dashed lines , which are the lines whose length has been used to get the perimeter .
the three configurations in the top panels of fig.[fig : configs1 ] are natural to define on the square lattice because their edges are parallel to the orthogonal vectors generating the lattice . instead
, the three configurations in the bottom panels of fig.[fig : configs1 ] are made by adjacent domains where the line @xmath141 either is curved or it contains a line segment which is oblique with respect to the vectors generating the lattice .
notice that a disk of given radius on the lattice could include a different number of sites depending on whether the centre of the disk is located on a lattice site or within a plaquette .
such ambiguity does not affect the leading order behaviour of the quantities that we are considering , but it could be relevant for subleading terms @xcite .
also for the logarithmic negativity of the adjacent domains shown in fig.[fig : configs1 ] we have observed the same qualitative behaviour described in the left panel of fig.[fig : adjrect ] for the equal adjacent rectangles : by keeping fixed the region corresponding to the red circles while the sizes of the region characterised by the blue dots increase , @xmath63 saturates to a constant value . in fig.[fig :
negarea ] we show some quantitative results for the logarithmic negativity of the configurations in fig.[fig : configs1 ] . in particular ,
let us consider the configuration in the top left panel , which is characterised by the lengths @xmath142 of the edges of the internal square and of the whole subregion @xmath143 respectively . in the left panel of fig.[fig : negarea ] we show @xmath144 as function of the ratio @xmath145 when @xmath146 is kept fixed and the internal square increases . for large enough @xmath146 , the area law behaviour in terms of @xmath147 is observed .
it is worth remarking that @xmath148 when @xmath149 .
this is expected because in this limit the internal convex domain becomes the whole @xmath2 .
for any fixed and large enough value of @xmath150 there is a critical size of the internal square after which the logarithmic negativity deviates from the linear growth predicted by the area law .
the numerical data tell us that such critical value of @xmath151 increases by increasing @xmath150 .
this suggests that in the continuum limit , where both @xmath151 and @xmath150 diverge but their ratio is finite , the linear behaviour occurs for any ratio @xmath152 . from the plot in the left panel of fig.[fig : negarea ] one can also notice that the ratio @xmath153 is a good regime to explore the area law behaviour even for relatively small domains .
given the latter observation , we have considered the logarithmic negativity of all the configurations in fig.[fig : configs1 ] with a ratio of @xmath154 between the size of the internal convex domain and the one of the whole subsystem @xmath2 ( for the triangle we refer to its shortest edge and for the trapezoid to its height ) . by increasing the sizes of both the domains
while their ratio is kept fixed to @xmath154 , we find the results collected in the right panel of fig.[fig : negarea ] , which nicely confirm the area law behaviour in terms of @xmath155 observed above .
notice that the different configurations in the right panel of fig.[fig : negarea ] provide linear growths with almost the same slope .
moreover , the data corresponding to the configuration in the bottom left panel of fig.[fig : configs1 ] do not provide a neat straight line , as expected whenever domains with a curved boundary on a square lattice are involved . , where @xmath156 and @xmath146 are the sizes of the edges of the red square and of the square @xmath143 respectively .
the dashed line is obtained by fitting the data corresponding to @xmath157 up to @xmath158 .
right : logarithmic negativity of adjacent domains for the configurations shown in fig.[fig : configs1 ] , which involve different shapes for the curve @xmath141 .
the sizes of the domains increase while their ratios are kept fixed .
the data correspond to configurations where the linear size of the convex domains ( highlighted by red circles ) is @xmath154 of the size of the corresponding @xmath143 .
the dashed line has been found by fitting the data obtained for the configuration in the bottom left panel of fig.[fig : configs1 ] .
, scaledwidth=115.0% ] summarising the numerical results presented above , we can conclude that at the leading order the logarithmic negativity of two large adjacent domains @xmath41 and @xmath42 on a lattice of massless harmonic oscillators with nearest neighbour spring - like interactions in the ground state satisfies an area law in terms of the length @xmath159 of the curve shared by the adjacent regions , namely [ area law neg ] = a p _ + , where the dots indicate subleading terms for large values of @xmath159 .
the area law ( [ area law neg ] ) is consistent with the fact that @xmath63 measures the bipartite entanglement between @xmath41 and @xmath42 for the mixed state characterised by the reduced density matrix @xmath160 .
the coefficient @xmath83 in ( [ area law neg ] ) is non universal , i.e. it depends on the ultraviolet details .
given the two adjacent regions @xmath41 and @xmath42 considered above , another very interesting quantity to study is their mutual information @xmath27 , which has been defined in ( [ mi def ] ) . since the area law of the entanglement entropy for large domains tells us that @xmath161 , it is straightforward to find that @xmath27 of adjacent domains satisfies an area law in terms of @xmath159 . in particular , we have that [ area law mi ] i_a_1 , a_2 = 2 p _
+ , where , as above , the dots stand for subleading terms . the moments @xmath58 of the partial transpose for integer values of @xmath14 are interesting quantities to study because they provide the logarithmic negativity through the replica limit ( [ replica limit neg ] ) @xcite
. given the configurations of adjacent domains described in
[ sec : area log neg ] , instead of considering the @xmath14-th moment of the partial transpose , we find it more interesting the ratio @xmath64 defined in ( [ en def ] ) , which also provides the logarithmic negativity through the replica limit ( [ replica limit e_n ] ) because of the normalisation condition @xmath162 .
in our model , the main reason to consider @xmath64 instead of @xmath163 occurring in ( [ replica limit neg ] ) is that , by repeating the analysis described in [ sec : area log neg ] , we find that , at leading order for large adjacent domains , @xmath64 follows an area law in terms of the length of the curve shared by the adjacent domains , i.e. [ area law renyi neg ] _
n = a_n p _ + , where the non universal coefficient @xmath164 depends on the integer @xmath14 and the dots denote subleading terms .
we recall that the rnyi entropies of our model satisfy the area law @xmath165 , where the coefficient @xmath166 is non universal as well and @xmath167 . from ( [ area law renyi neg ] ) and the area law of the rnyi
entropies it is straightforward to find the leading term of the logarithm of the moments of the partial transpose , which is given by @xmath168 thus , the quantities @xmath169 contain an area law contribution also from the boundary of @xmath170 .
since @xmath171 , such term cancels in the replica limit ( [ replica limit neg ] ) .
a similar cancellation occurs also for adjacent intervals in @xmath23 dimensional cfts . indeed , considering the divergent terms for @xmath31 , in @xmath64 only the ones giving a non trivial contribution after the replica limit survive , while @xmath163 contains also other terms @xcite , which vanish in the replica limit ( [ replica limit neg ] ) . the quantity @xmath64 for adjacent intervals in @xmath23 dimensional cfts has been studied in @xcite and for free fermions on a two dimensional lattice in @xcite .
in the left panel of fig.[fig : renyi ] we show @xmath166 as function of @xmath14 for disks and squares on the infinite lattice .
the centres of the disks have been chosen either on a lattice site ( like in the bottom left panel of fig.[fig : configs1 ] ) or in the central point of a plaquette .
as for the squares , we have considered both the ones whose edges are parallel to the vectors generating the lattice and the ones obtained by rotating of @xmath172 the previous ones ( denoted as rhombi in the plot ) . in this plot
we have @xmath173 , depending on the configuration .
a slight dependence of @xmath166 on the shape can be observed from our data points .
the asymptotic @xmath174 as @xmath175 @xcite is consistent with our numerical results . for the ising model a numerical analysis for @xmath166 as @xmath175 has been done in @xcite .
the numerical data in the left panel of fig.[fig : renyi ] have been found by employing a fitting function which includes also a logarithmic term , as it will be discussed in detail in [ sec : corners ] , but such term does not change the coefficient of the leading area law term in a significant way .
as function of @xmath14 for domains with different shapes .
centre the centres of the disks have been chosen both on the lattice site and in the centre of a plaquette . here
the edges of the square configuration are parallel to the vectors generating the lattice , while the rhombus configuration corresponds to the previous square configuration rotated by @xmath172 with respect to its centre .
right : numerical check of ( [ area law coeff a_n ] ) and ( [ area coeff neg renyi 1/2 ] ) for the configuration in the bottom left panel of fig.[fig : configs1 ] with @xmath176 .
inset : the coefficient @xmath177 as function of @xmath14 for disks ( the data are taken from the left panel with the same colour code ) .
the black curve corresponds to the best fit of the numerical data through the function @xmath178 , constrained by the condition @xmath179 . in the main plot
the red curve is @xmath180 obtained in the inset and the green one is @xmath181 . , scaledwidth=115.0% ] it is worth considering the quantity @xmath182 for the configurations of adjacent domains described in
[ sec : area log neg ] . given the area law behaviour of @xmath183 , it is easy to observe that @xmath120 displays an area law behaviour in terms of @xmath159 , namely [ area law renyi mi ] i^(n)_a_1 , a_2 = 2_n p _ + , once the configuration of adjacent domains @xmath41 and @xmath42 has been chosen , we find it interesting to compare the non universal coefficients @xmath184 and @xmath185 occurring in the area law terms of ( [ area law renyi neg ] ) and ( [ area law renyi mi ] ) respectively .
it is reasonable to expect that the area law term for @xmath64 comes from effects localised in the neighbourhood of the curve @xmath141 .
thus , considering e.g. the configurations in fig.[fig : configs1 ] where the domain identified by the red circles is entirely surrounded by the one characterised by the blue dots ( i.e. the configurations in the top left , bottom middle and bottom right panels ) , such term should be independent of the size of the domain identified by the blue dots . in the limit where this domain becomes the whole region complementary to the one identified by the red circles ,
one gets the bipartition of the ground state .
these considerations suggest us that @xmath164 in ( [ area law renyi neg ] ) is the same one occurring for a bipartition of the ground state , when the identity ( [ identity pure states ] ) can be applied .
this implies that the following relation should hold [ area law coeff a_n ] a_n = \ { ll ( 1-n_o ) _ n_o & , + ' '' '' 2 ( 1-n_e/2)_n_e/2 & .
notice that @xmath186 , as expected . by employing the relation ( [ area law coeff a_n ] ) and
the replica limit ( [ replica limit e_n ] ) , it is straightforward to find that the coefficient of the area law term in the logarithmic negativity in ( [ area law neg ] ) is equal to the coefficient of the area law term in the rnyi entropy of order @xmath187 , namely [ area coeff neg renyi 1/2 ] a = _ 1/2 . in the right panel of fig.[fig : renyi ] we show a numerical check of the relations ( [ area law coeff a_n ] ) and ( [ area coeff neg renyi 1/2 ] ) for the configuration in the bottom left panel of fig.[fig : configs1 ] . in particular , the coincidence of the data points corresponding to @xmath188 provides a check of ( [ area coeff neg renyi 1/2 ] ) .
the solid curve in the inset is obtained by fitting the data with the function @xmath189 , where the parameters are constrained by the requirement that @xmath179 .
thus , such fit has three independent parameters . as for the solid curves in the main plot ,
the red one is @xmath180 , namely the black curve found in the inset , while the green one is @xmath181 .
this analysis has been performed also for other configurations as further checks of ( [ area law coeff a_n ] ) and ( [ area coeff neg renyi 1/2 ] ) , finding the same qualitative behaviours .
in this section we consider the subleading logarithmic term in @xmath63 and @xmath64 for adjacent domains , which occurs whenever the shared curve @xmath141 contains some vertices , where its endpoints are included among them . for vertices corresponding to explementary angles , we provide some numerical evidence that the corner function of @xmath63 is given by the corner function of the rnyi entropy of order @xmath187 .
let us consider the entanglement entropy @xmath19 of a connected domain @xmath2 whose boundary contains some vertices ( see figs.[fig : configs1 ] , [ fig : configs2 ] and [ fig : configs3 ] for examples ) . for large size of @xmath2 ,
the leading term gives the area law behaviour .
the occurrence of vertices in @xmath22 provides a subleading logarithmic term which is characterised by a corner function @xmath190 as follows @xcite [ s_a corner ] s_a = p_a - ( _ ( _ i ) ) p_a + 0 < _ i , where @xmath191 is the opening angle in @xmath2 corresponding to the @xmath192-th vertex of @xmath22 and the dots denote subleading terms .
since the logarithmic term is due to the corners , we have that @xmath193 . from the general property
that @xmath194 for pure states and a bipartite hilbert space @xmath1 , we have that @xmath195 , which tells us that @xmath196 is defined for @xmath197 .
the model dependent corner function @xmath196 is universal , i.e. independent of the ultraviolet details of the regularisation . in the continuum limit ,
which is described by a @xmath29 dimensional cft , the corner function @xmath196 contains important information about the model .
for instance , recently it has been found that the constant @xmath198 entering in the asymptotic behaviour @xmath199 as @xmath200 at the leading order is related to the constant characterising the correlator @xmath201 of the underlying cft @xcite .
the corner function @xmath196 for the massless scalar has been studied by casini and huerta @xcite . in the context of holography , by employing the prescription of @xcite for @xmath19 , the corner function @xmath196 has been studied e.g. in @xcite .
we find it worth considering the mutual information ( [ mi def ] ) of two adjacent domains @xmath41 and @xmath42 when their boundaries contain some vertices . for the sake of simplicity , we focus on configurations such that either two or three curves meet at every vertex .
explicit examples are shown in figs.[fig : configs1 ] , [ fig : configs2 ] and [ fig : configs3 ] , where the adjacent domains are identified by blue dots and red circles . in the scaling limit
, when their sizes increase while the ratios among them are kept fixed , from ( [ s_a corner ] ) one finds that [ area law mi corners ] i_a_1 , a_2 = 2 p _ - ( _ ) + , where @xmath24 is a parameter characterising the common size of the adjacent regions , @xmath202 is the angle in @xmath203 and @xmath204 the angle in @xmath143 corresponding to the @xmath192-th vertex belonging to @xmath141 . in the simplest case where such vertex is not an endpoint of @xmath141 , it provides a bipartition of the angle of @xmath205 and the corresponding contribution to the sum in ( [ area law mi corners ] ) is @xmath206 .
we find it instructive to consider explicitly the examples in figs.[fig : configs1 ] and [ fig : configs2 ] .
for the configurations shown in the top left , middle and right panel of fig.[fig : configs1 ] , the sum within the parenthesis multiplying the logarithmic term in ( [ area law mi corners ] ) is given by @xmath207 , @xmath207 and @xmath208 respectively , while for the bottom middle and right panels of the same figure it reads @xmath209 and @xmath210 respectively .
as for the configurations of fig.[fig : configs2 ] , such coefficient is @xmath211 , @xmath212 and @xmath213 for the top left , middle and right panels respectively , while it is given by @xmath214 , @xmath215 and @xmath216 for the bottom left , middle and right panels respectively .
it is not difficult to get the coefficient of the logarithmic term also for the configurations in fig.[fig : configs3 ] .
let us point out that for the one in the top left panel such coefficient is vanishing . as for the rnyi entropies of domains with non smooth boundary , we have that [ renyi corner ] s^(n)_a = _
n p_a - ( _ _ n(_i ) ) p_a + , where the corner function @xmath217 depends on the order @xmath14 and it provides @xmath190 through the replica limit , i.e. @xmath218 . for the model we are dealing with , the corner function @xmath217 has been found in @xcite .
a formula similar to ( [ area law mi corners ] ) can be written for @xmath219 , by just replacing @xmath220 with @xmath166 and @xmath196 with @xmath221 .
considering the scaling limit for the configurations of adjacent domains @xmath41 and @xmath42 in figs.[fig : configs1 ] , [ fig : configs2 ] and [ fig : configs3 ] where @xmath222 and @xmath223 contain vertices , the expansion of the logarithmic negativity contains a subleading logarithmic term after the leading area law term . by analogy with the case of the entanglement entropy
, it is reasonable to expect that the coefficient of the logarithmic term is obtained by summing the contributions of the vertices occurring in @xmath222 and @xmath223 .
extracting the coefficient of such subleading logarithmic term from the lattice numerical data is a delicate task .
let us first discuss the method employed to get the numerical values presented in this section from the formulas discussed in
[ sec : hc ] .
a first rough approach could consist in fitting the numerical data by a linear term , a logarithmic term and a constant one .
nevertheless , since the logarithmic contribution is tiny with respect to the linear one , its estimation can be spoiled by the occurrence of subleading lattice effects . in order to take them into account
, we have inserted in our fitting analysis also some standard power law corrections , i.e. we have fitted our data with a function of the form @xmath224 @xcite , where @xmath24 is a characteristic length of @xmath225 .
in particular , considering the domains identified by red circles in figs.[fig : configs1 ] , [ fig : configs2 ] and [ fig : configs3 ] , in each configuration @xmath24 corresponds to the edge for squares , to the shortest edge for the triangles , to the radius for the disks and to the height for the trapezoids . for each data set , in the plots
we show the results of various fits performed in different ranges of @xmath24 , where each range is specified by the starting value @xmath226 and by the ending value @xmath227 .
we fix some maximum exponent @xmath228 and some @xmath226 by removing some initial points , which are strongly affected by lattice effects and therefore would need more corrections .
then , we plot the logarithmic coefficient obtained from different fits as a function of @xmath227 .
the parameters @xmath228 , @xmath226 and the maximum value of @xmath227 are chosen in order to get a stable result from the fits in the whole range of @xmath227 .
the logarithmic coefficient is estimated as the average of the fitted values within such range of @xmath227 . the error introduced through this procedure
is estimated by taking the maximum deviation of the data from the average within this range of stability .
an important benchmark employed to test our numerical analysis is the mutual information of adjacent domains . in particular , we have considered the coefficient of the logarithmic term in the mutual information for the configurations in the top left , bottom middle and bottom right panels of fig.[fig : configs1 ] , recovering the values of the corner function @xmath196 for @xmath229 available in the literature @xcite .
let us consider the occurrence of a subleading logarithmic term due to corners in the logarithmic negativity of some configurations of adjacent domains .
we first observe that the angles of @xmath41 and @xmath42 contributing to the logarithmic term in @xmath63 are such that at least one of their sides belongs to @xmath230 , namely only the angles whose vertices lie on the curve @xmath230 provide a non trivial logarithmic contribution .
also the endpoints of @xmath230 have to be included among such vertices .
this is expected from the guiding principle that the logarithmic negativity measures the entanglement between @xmath41 and @xmath42 . a way to check numerically
this observation is to consider e.g. the coefficients of the logarithmic terms in @xmath63 for the configurations in the top panels of fig.[fig : configs1 ] .
these configurations contain only three possible different contributions corresponding to this kind of vertices .
thus , one can easily solve the resulting linear system of three equations finding that the contribution coming from the four vertices which do not belong to @xmath230 is much smaller than the other ones ( by a factor of about @xmath231 ) . as further check that only the vertices lying on @xmath230 ( including its endpoints ) contribute to the logarithmic term of @xmath63
, we have constructed another configuration of adjacent domains as follows : starting from a configuration like the one in the top right panel of fig.[fig : configs1 ] with @xmath232 and dividing the domain corresponding to the blue dots along the diagonal with negative slope of @xmath233 , we have removed the upper triangle . in the resulting configuration @xmath234
is a triangle and the subregion identified by the red circles is a square .
comparing the coefficients of the logarithmic term for this configuration and the one for the configuration in the top right panel of fig.[fig : configs1 ] with @xmath232 , we have found the same number within numerical errors . and
@xmath235 for @xmath229 , as explained in
[ sec : corner logneg ] .
the configurations employed here are shown in the top left , bottom middle and bottom right panels of fig.[fig : configs1 ] .
the numerical values are @xmath236 , @xmath237 and @xmath238 . ]
thus , the logarithmic negativity of adjacent domains whose boundaries share a curve containing some vertices , where its endpoints are counted among them , is given by [ area law neg corners ] = a p _ - ( _ b(_i^(1 ) , _
i^(2 ) ) ) p _ + , being @xmath239 the angle corresponding to the @xmath192-th vertex of @xmath141 which belongs to @xmath203 . in ( [ area law neg corners ] )
we have assumed that either two or three curves meet at every vertex of @xmath141 , i.e. every vertex corresponds either to a bipartition ( @xmath240 ) or to a tripartition ( @xmath241 ) respectively of the angle of @xmath205 .
this assumption , which has been done also in ( [ area law mi corners ] ) , is verified for all the configurations in figs.[fig : configs1 ] , [ fig : configs2 ] and [ fig : configs3 ] . from our numerical analysis
, we find that the above considerations apply also for the quantity @xmath242 defined in ( [ en def ] ) .
thus we have [ area law renyi neg corners ] _ n = a_n p _ - ( _ b_n(_i^(1 ) , _
i^(2 ) ) ) p _ + , where the coefficient @xmath164 has been already discussed in
[ sec : area neg renyi ] and the corner function @xmath243 is related to the one occurring in the logarithmic term of ( [ area law neg corners ] ) through the replica limit ( [ replica limit e_n ] ) , namely @xmath244 . among the vertices belonging to the curve @xmath141 which contribute to the logarithmic term in ( [ area law neg corners ] ) and ( [ area law renyi neg corners ] ) ,
let us consider first the ones corresponding to pairs of explementary angles , i.e. the ones such that @xmath245 .
this kind of vertices occurs in all the panels of fig.[fig : configs1 ] except for the bottom left one , while it does not occur at all in the configurations of fig.[fig : configs2 ] .
for some values of arguments such that @xmath246 ( coloured markers in fig.[fig : cornernegnew ] ) . , scaledwidth=108.0% ] for these vertices we can make an observation similar to the one that leads to ( [ area law coeff a_n ] ) . indeed , because of the local nature of the function @xmath247 , it is reasonable to assume that these vertices provide the same contribution given in the case of a bipartition of the ground state , when ( [ identity pure states ] ) holds .
this observation leads us to propose the following relation between @xmath248 and the corner function @xmath217 entering in the rnyi entropies [ corner function pure ] b_n ( , 2- ) = \ { ll ( 1-n_o ) _
n_o ( ) & , + ' '' '' 2 ( 1-n_e/2)_n_e/2 ( ) & . by employing the replica limit ( [ replica limit e_n ] ) ,
the relation ( [ corner function pure ] ) allows to conclude that the corner function in the logarithmic negativity for this kind of vertices is equal to the corner function in the rnyi entropy of order @xmath187 , namely [ conj corner neg 1/2 ] b ( , 2- ) = _ 1/2 ( ) .
numerical checks of the relation ( [ conj corner neg 1/2 ] ) for some values of @xmath37 are shown in fig.[fig : cornernegold ] .
the values of @xmath249 and @xmath235 for @xmath250 have been found by evaluating @xmath63 and @xmath251 for the configurations shown in the top middle , bottom middle and bottom right panels of fig.[fig : configs1 ] , where the curve @xmath141 contains only the kind of vertices that we are considering .
the numerical values obtained for @xmath235 are less stable than the ones for @xmath249 .
nevertheless , they provide a reasonable check of ( [ conj corner neg 1/2 ] ) . for some pair of angles such that @xmath252 .
the data corresponding to coloured markers have been obtained by employing the configurations in fig.[fig : configs2 ] , while the data corresponding to black markers have been found by using the configurations in fig.[fig : configs3 ] and the ones in the top middle and top right panels of fig.[fig : configs1 ] .
the numerical values are reported in ( [ corner tri - vertex values ] ) . ]
an analytic expression for the function @xmath235 , which can be found by performing the analytic continuation @xmath253 of the formula for @xmath254 obtained in @xcite , is not available . given in fig.[fig : cornernegold ] and fig.[fig : cornernegnew ] . , scaledwidth=108.0% ]
let us consider the vertices corresponding to a tripartition of the angle @xmath205 , for which @xmath255 and which are the endpoints of the curve @xmath141 in the class of vertices that we are considering .
for this kind of vertices the corner function @xmath256 depends on two independent variables .
we have obtained its numerical values for some pairs of angles by employing the configurations in fig.[fig : configs2 ] .
the results are shown in fig.[fig : cornernegnew ] ( coloured solid markers ) and they are given by [ corner tri - vertex values ] lll b(/4,/4 ) = 0.0908(1 ) & b(/4,3/4 ) = 0.0614(3 ) & b(/4,5/4 ) = 0.0536(1 ) + b(/2,/2 ) = 0.0364(1 ) & b(/2 , ) = 0.0152(2 ) & b(3/4,3/4 ) = 0.0068(2 ) , where the parenthesis denote the uncertainty on the last digit . in fig.[fig : cornernegnew ] the black markers correspond to the values obtained by employing all the configurations of fig.[fig : configs3 ] and the ones in the top middle and top right panels of fig.[fig : configs1 ] .
let us remark that the coefficient of the logarithmic term in the logarithmic negativity of the configuration in the top left panel of fig.[fig : configs3 ] is non zero , while it is vanishing in the mutual information of the same configuration , as already pointed out in [ sec : corners ent ] . in principle , our numerical analysis allows to find also @xmath256 . nevertheless , in order to check the relation ( [ corner function pure ] ) we need to know the unusual corrections @xcite to the scaling in @xmath29 dimensions in order to perform a precise fitting analysis ( see e.g. @xcite ) . the continuum limit of the lattice model ( [ ham ] ) considered throughout this manuscript
is described by the massless scalar field in @xmath29 dimensions , which is a cft . denoting by @xmath20 the uv cutoff that must be introduced to regularise the model , the logarithmic negativity of adjacent domains diverges when @xmath31 .
the numerical results on the lattice discussed above tell us that the expansion of the logarithmic negativity as @xmath31 reads [ area law neg corners qft ] = - ( _ b(_i^(1 ) , _
i^(2 ) ) ) ( p _ / ) + , where the coefficient @xmath257 in front of the area law term is non universal .
a similar expression can be written for @xmath64 in ( [ en def ] ) , namely [ area law renyi neg corners qft ] _
n = _ n - ( _ b_n(_i^(1 ) , _
i^(2 ) ) ) ( p _ / ) + , where @xmath258 is non universal as well . instead , the corner functions @xmath259 and @xmath243 are independent of the uv details of the regularisation and therefore they are the most important quantities to study . in ( [ area law neg corners qft ] ) and ( [ area law renyi neg corners qft ] ) , like for their lattice versions ( [ area law neg corners ] ) and ( [ area law renyi neg corners ] ) , we have assumed that the vertices in @xmath222 and @xmath223 correspond either to a bipartition or to a tripartition of the angle of @xmath205 .
the divergent terms in the @xmath31 expansion of @xmath63 and @xmath64 are determined by local effects close to the curve @xmath141 , consistently with the intuition that the entanglement between @xmath41 and @xmath42 comes from the degrees of freedom living close to their shared boundary .
this leads to relate the coefficients @xmath257 and @xmath258 in ( [ area law neg corners qft ] ) and ( [ area law renyi neg corners qft ] ) to the area law coefficients @xmath260 and @xmath261 of @xmath19 and @xmath13 like in ( [ area law coeff a_n ] ) and ( [ area coeff neg renyi 1/2 ] ) . notice that , whenever for the @xmath192-th vertex of @xmath141 we have @xmath262 , the relations ( [ corner function pure ] ) and ( [ conj corner neg 1/2 ] ) for the corner functions @xmath263 and @xmath264 hold also in the continuum limit . instead
, whenever the vertices correspond to partitions of the angle of @xmath205 in three ( i.e. @xmath241 ) or higher number of components , we expect that the corner functions occurring in @xmath63 or @xmath64 contain new information with respect to the corner functions entering in @xmath19 or @xmath13 .
it would be very interesting to find an analytic expression for the corner function @xmath265 for any pair of angles @xmath37 and @xmath266 , where we can assume @xmath267 . by considering only the few pairs of angles that we have studied on the lattice
, we can make some observations about the behaviour of this corner function .
for instance , when the two angles are equal , the function @xmath268 is decreasing with @xmath269 $ ] and @xmath270 . moreover , for fixed @xmath271 ,
the function @xmath272 is decreasing with @xmath273 . by comparing fig.[fig :
cornernegnew ] with fig.[fig : cornernegold ] , one notices that this is not true anymore when @xmath266 is exactly @xmath274 . such behaviour is not surprising because in this limit the boundaries of @xmath41 and @xmath42 merge , @xmath275 changes abruptly and therefore a continuous behaviour of the divergent terms in the logarithmic negativity is not expected . and @xmath42 in the plane ( the yellow and cyan regions ) .
the grey part corresponds to the region @xmath3 which has been traced out .
the quantities @xmath63 and @xmath64 for the configurations in the top left and top middle panels do not contain a logarithmic term , while for the remaining ones such term is non vanishing .
the vertices corresponding to a partition of the angle of @xmath205 in three or four angles are highlighted with red circles .
, scaledwidth=115.0% ] in fig.[fig : configsqft ] we show some illustrative examples of configurations of adjacent regions in the plane .
the grey region is associated to the part @xmath3 , which has been traced out . considering @xmath63 and @xmath64 between the yellow domain and the cyan domain ,
the expressions ( [ area law neg corners qft ] ) and ( [ area law renyi neg corners qft ] ) can be employed for all the configurations in fig.[fig : configsqft ] except for the one in the bottom right panel , where the vertex highlighted by the red circle corresponds to a partition of the angle of @xmath205 in four parts . in the configurations shown in the top left and middle panels of fig.[fig : configsqft ] the curve @xmath141 is smooth ; therefore the logarithmic divergence does not occur in @xmath63 and @xmath64 .
instead , in the remaining configurations the curve @xmath141 contains vertices and the subleading logarithmic divergence occurs . as for the configuration in the top right panel ,
the coefficient of the logarithmic term of @xmath63 and @xmath64 is related to the corner functions @xmath196 and @xmath221 entering in the logarithmic term of @xmath19 or @xmath13 respectively through ( [ corner function pure ] ) and ( [ conj corner neg 1/2 ] ) .
thus , for the configurations in the top panels of fig.[fig : configsqft ] and assuming that only one scale @xmath24 occurs to determine the logarithmic term , one can construct the following uv finite quantity [ neg minus mi ] - i^(1/2)_a_1 , a_2 .
for the same configurations , also the following combinations , depending on the parity of the integer @xmath14 , are uv finite [ renyi neg minus mi ] _ n_o - i^(n_o)_a_1 , a_2 , _ n_e - ( 1- ) i^(n_e/2)_a_1 , a_2 .
the second expression in ( [ renyi neg minus mi ] ) provides ( [ neg minus mi ] ) after the analytic continuation @xmath276 . in a @xmath29 dimensional
cft , the quantities in ( [ neg minus mi ] ) and ( [ renyi neg minus mi ] ) should give non trivial scale invariant functions of the parameters characterising the adjacent domains .
for example , when the adjacent domains are given by a disk of radius @xmath277 and an annulus surrounding it whose radii are @xmath278 , the expression ( [ neg minus mi ] ) should be a model dependent function of the ratio @xmath279
. it would be interesting to compute such function analytically .
the configurations of adjacent domains shown in the bottom panels of fig.[fig : configsqft ] are more interesting because the corner functions corresponding to the vertices highlighted by the red circles is not related to the corner functions occurring in the entanglement entropies .
thus , because of such terms , we expect that the combinations in ( [ neg minus mi ] ) and ( [ renyi neg minus mi ] ) diverge logarithmically for these configurations .
in this manuscript we have investigated the logarithmic negativity @xmath63 and the moments of the partial transpose for adjacent domains @xmath41 and @xmath42 in the ground state of a two dimensional harmonic square lattice with nearest neighbour spring - like interaction .
the regime of massless oscillators in the thermodynamic limit has been considered . by exploring various configurations of adjacent domains
, we have shown that , at leading order for large domains , the logarithmic negativity and the quantity @xmath64 introduced in ( [ en def ] ) satisfy an area law in terms of the length of the shared curve @xmath141 , suggesting a relation between the coefficient of the area law term in these quantities and the coefficient of the area law term in the rnyi entropies .
a subleading universal logarithmic term occurs in @xmath63 and @xmath64 whenever the shared curve contains vertices , being its endpoints included among them .
the values of the corner function of @xmath63 have been obtained for some pairs of angles . for the vertices of @xmath141 corresponding to pairs of explementary angles
, we have proposed that the corner function of @xmath64 is related to the corner function entering in the rnyi entropies @xcite .
this relation implies that the corner function of @xmath63 for this kind of vertices coincides with the corner function of the rnyi entropy of order @xmath187 .
this statement has been supported by numerical evidences shown in fig.[fig : cornernegold ] .
as for the vertices of the curve @xmath141 corresponding to a tripartition of the angle of @xmath205 , their contribution to the logarithmic term in @xmath63 and @xmath64 is characterised by a new corner function which depends on two independent angular variables .
the numerical values of this corner function for @xmath63 have been given in ( [ corner tri - vertex values ] ) for some pair of angles ( see also fig.[fig : cornernegnew ] ) .
let us conclude with some open questions
. it would be interesting to provide further checks of ( [ conj corner neg 1/2 ] ) .
in particular , the analytic continuation to @xmath188 of the corner function found by casini and huerta @xcite should be performed .
more importantly , a method should be found to compute analytically the corner functions @xmath272 and @xmath280 for the vertices partitioning the angle of @xmath205 in three parts . by analogy with the results of @xcite obtained for the corner function of @xmath19 ,
it could be interesting to study the corner function @xmath281 for equal angles as @xmath282 . in order to extract reliable numerical results for the logarithmic term of @xmath64 from the fit of the lattice data , the unusual corrections to the scaling
must be studied , extending the analysis done by cardy and calabrese @xcite in @xmath23 dimensions .
finally , it is worth studying the corner contributions to @xmath63 and @xmath64 for other models , both on the lattice and in the continuum .
we are grateful in particular to pasquale calabrese , john cardy , horacio casini and marina huerta for useful discussions and correspondence .
we thank vincenzo alba , mario collura , viktor eisler , veronika hubeny , mukund rangamani , german sierra and zoltan zimbors for useful conversations . et
would like to thank the instituto de fisica teorica ( ift uam - csic ) in madrid for hospitality during part of this work and its support via the centro de excelencia severo ochoa program under grant sev-2012 - 0249 .
et is grateful to princeton university , mit , brandeis university and uc davis for the warm hospitality during the final part of this work .
ac acknowledges support from mineco ( grant mtm2014 - 54240-p ) and comunidad de madrid ( grant quitemad+-cm , ref .
s2013/ice-2801 ) .
ac has received funding from the european research council ( erc ) under the european union s horizon 2020 research and innovation programme ( grant agreement no 648913 ) .
et has been supported by the erc under starting grant 279391 edeqs .
99 b. groisman , s. popescu and a. winter , phys . rev . a * 72 * , 032317 ( 2005 ) ; + m. nielsen and i. chuang , _ quantum computation and quantum information _ , cambridge university press , cambridge uk , 2000 ; + m. wolf , f. verstraete , m. hastings and i. cirac , phys . rev . lett . * 100 * , 070502 ( 2008 ) . r. sorkin , tenth international conference on general relativity and gravitation , padova 4 - 9 july 1983 , contributed papers , vol .
ii , 734 , arxiv:1402.3589 ; + l. bombelli , r. koul , j. lee and r. sorkin , phys .
d * 34 * ( 1986 ) 373 ; + m. srednicki , phys .
* 71 * ( 1993 ) 666 .
m. b. plenio , j. eisert , j. dressig and m. cramer , phys .
lett . * 94 * , 060503 ( 2005 ) ; + m. cramer , j. eisert , m. b. plenio and j. dreissig , phys .
a * 73 * , 012309 ( 2006 ) ; + t. barthel , m .- c .
chung and u. schollwoeck , phys .
a * 74 * , 022329 ( 2006 ) ; + m. huerta , phys .
b * 710 * 691 ( 2012 ) . c. holzhey , f. larsen , and f. wilczek , nucl .
b * 424 * , 443 ( 1994 ) ; + c. callan and f. wilczek , phys .
b * 333 * , 55 ( 1994 ) ; + p. calabrese and j. cardy , j. stat .
p06002 ( 2004 ) ; + p. calabrese and j. cardy , j. phys .
a * 42 * , 504005 ( 2009 ) .
m. caraglio and f. gliozzi , jhep 0811:076 ( 2008 ) ; + s. furukawa , v. pasquier , and j. shiraishi , phys .
102 * , 170602 ( 2009 ) ; + p. calabrese , j. cardy , and e. tonni , j. stat .
p11001 ( 2009 ) ; + m. headrick , phys .
d * 82 * , 126010 ( 2010 ) ; + p. calabrese , j. cardy and e. tonni , j. stat .
p01021 ( 2011 ) ; + a. coser , l. tagliacozzo and e. tonni , j. stat .
p01008 ( 2014 ) .
a. kallin , k. hyatt , r. singh and r. melko , phys .
lett . * 110 * , 135702 ( 2013 ) ; + e. stoudenmire , p. gustainis , r. johal , s. wessel and r. melko , phys .
b * 90 * , 235106 ( 2014 ) ; + a. kallin , e. stoudenmire , p. fendley , r. singh and r. melko , j. stat .
p06009 ( 2014 ) ; + j .-
stphan , s. furukawa , g. misguich and v. pasquier , phys .
b * 80 * , 184421 ( 2009 ) ; + r. singh , r. melko and j. oitmaa , phys . rev .
b * 86 * , 075106 ( 2012 ) ; + s. humeniuk and t. roscilde , phys . rev .
b * 86 * , 235116 ( 2012 ) ; + t. devakul and r. singh , phys . rev .
b * 90 * , 064424 ( 2014 ) ; + n. laflorencie , d. luitz and f. alet , phys .
b * 92 * , 115126 ( 2015 ) .
n. drukker , d. gross and h. ooguri , phys .
d * 60 * , 125006 ( 1999 ) ; + t. hirata and t. takayanagi , jhep 0702:042 ( 2007 ) ; + p. fonda , l. giomi , a. salvio and e. tonni , jhep 1502 ( 2015 ) 005 ; + p. fonda , d. seminara and e. tonni , jhep 1512 ( 2015 ) 037 .
a. peres , phys .
* 77 * , 1413 ( 1996 ) ; + k. zyczkowski , p. horodecki , a. sanpera and m. lewenstein , phys .
a * 58 * , 883 ( 1998 ) ; + j. eisert and m. plenio , j. mod
. opt . * 46 * , 145 ( 1999 ) ; + j. lee , m. kim , y. park and s. lee , j. mod . opt . *
47 * , 2151 ( 2000 ) ; + g. vidal and r. f. werner , phys .
a * 65 * , 032314 ( 2002 ) ; + m. plenio , phys .
lett . * 95 * , 090503 ( 2005 ) ; + j. eisert , quant - ph/0610253 .
p. calabrese , l. tagliacozzo and e. tonni , j. stat .
p05002 ( 2013 ) ; + v. alba , j. stat .
p05013 ( 2013 ) ; + a. coser , e. tonni and p. calabrese , j. stat .
p033116 ( 2016 ) ; + a. coser , e. tonni and p. calabrese , arxiv:1511.08328 ; + c. herzog and y. wang , arxiv:1601.00678 .
h. wichterich , j. molina - vilaplana and s. bose , phys . rev .
a * 80 * , 010304 ( 2009 ) ; + j. anders and a. winter , quantum inf .
* 8 * , 0245 ( 2008 ) ; + j. anders , phys .
a * 77 * , 062102 ( 2008 ) ; + a. ferraro , d. cavalcanti , a. garcia - saez and a. acin , phys .
100 * , 080502 ( 2008 ) ; + d. cavalcanti , a. ferraro , a. garcia - saez and a. acin , phys .
a * 78 * , 012335 ( 2008 ) ; + h. wichterich , j. vidal and s. bose , phys .
a * 81 * , 032311 ( 2010 ) ; + a. bayat , p. sodano and s. bose , phys .
* 105 * , 187204 ( 2010 ) ; phys .
b * 81 * , 064429 ( 2010 ) ; + p.
sodano , a. bayat and s. bose , phys .
b * 81 * , 100412 ( 2010 ) ; + a. bayat , s. bose , p. sodano and h. johannesson , phys . rev . lett . * 109 * , 066403 ( 2012 ) ; + b. alkurtass , a. bayat , i. affleck , s. bose , h. johannesson , p. sodano , e. srensen and k. le hur , phys . rev .
b * 93 * , 081106 ( 2016 ) ; + c. chung , v. alba , l. bonnes , p. chen and a. lauchli , phys . rev .
b * 90 * , 064401 ( 2014 ) . c. castelnovo , phys .
a * 88 * , 042319 ( 2013 ) ; phys .
a * 89 * , 042333 ( 2014 ) ; + y. a. lee and g. vidal , phys . rev .
a * 88 * , 042318 ( 2013 ) ; + r. santos , v. korepin and s. bose , phys .
a * 84 * , 062307 ( 2011 ) ; + r. santos and v. korepin , j. phys .
a * 45 * , 125307 ( 2012 ) ; + x. wen , s. matsuura and s. ryu , arxiv:1603.08534 .
i. peschel and m .- c .
chung , j. phys .
a * 32 * , 8419 ( 1999 ) ; + i. peschel , j. phys .
a * 36 * , l205 ( 2003 ) ; + a. botero and b. reznik , phys .
a * 70 * , 052329 ( 2004 ) ; + i. peschel and v. eisler , j. phys .
a * 42 * , 504003 ( 2009 ) . | we study the logarithmic negativity and the moments of the partial transpose in the ground state of a two dimensional massless harmonic square lattice with nearest neighbour interactions for various configurations of adjacent domains . at leading order for large domains , the logarithmic negativity and the logarithm of the ratio between the generic moment of the partial transpose and the moment of the reduced density matrix at
the same order satisfy an area law in terms of the length of the curve shared by the adjacent regions .
we give numerical evidences that the coefficient of the area law term in these quantities is related to the coefficient of the area law term in the rnyi entropies .
whenever the curve shared by the adjacent domains contains vertices , a subleading logarithmic term occurs in these quantities and the numerical values of the corner function for some pairs of angles are obtained . in the special case of vertices corresponding to explementary angles , we provide numerical evidence that the corner function of the logarithmic negativity is given by the corner function of the rnyi entropy of order 1/2 .
[ [ section ] ] starttoctoc |
all the ab - initio calculation in this work are performed with gpaw@xcite .
the band structures of the twisted bilayers were calculated at the dft level with an lda exchange correlation functional and double - zeta polarized atomic orbitals as a basis set .
the hse06 and lda calculations for the monolayers were performed using a plane wave basis set with a cut off energy of @xmath50 ev and @xmath51 k - point grids .
we find it useful to elaborate on the color scheme choice for the twisted bilayer band structure in [ fig : bands_rotation ] .
such choice should be understood by considering that the two layers have different primitive cells and that the unfolding procedure has to be performed separately .
indeed , the unfolding for a given layer , say mos@xmath0 , not only will project the eigenstates belonging onto mos@xmath0 to their primitive cell but also the wse@xmath0 ones . as the latter
should rather be projected onto the wse@xmath0 primitive cell , we `` hide '' them by choosing a color scheme that goes from blue to white and decreases the size of the markers when going from states localized completely in mos@xmath0 to states localized completely in wse@xmath0 .
the same argument applies to the unfolding to the wse@xmath0 cell , but a red color scheme is applied instead . for lattice
matched heterostructures modeled in a minimal unit cell , it was checked that the atomic orbital basis yields the same band structure as well converged plane wave calculations . for the calculation of dielectric properties of van der waals
heterostructures we utilized dielectric building blocks available in ref . .
specifically the response function of each building block was calculated on a plane - waves basis with @xmath52 ev cut - off energy and @xmath53 k - point mesh . in order to avoid spurious interaction from artificial replica in the out - of - plane direction ,
a truncated coulomb interaction with @xmath54 of vacuum is used . the interlayer distance between the layers
are taken as average of the interlayer distance in their respective bulk form , specifically @xmath55 , @xmath56 , @xmath57 and @xmath58 .
the monolayer g@xmath1w@xmath1 calculations have been performed employing a new efficient technique@xcite that overcomes the problem of slow convergence of the band structures with k - point grid and yields well converged band gaps with @xmath51 k - points ( rather than @xmath59 using standard approaches ) .
we used an energy cut - off of @xmath60 ev for the dielectric function and sum over empty states .
the g@xmath1w@xmath1 band energies were extrapolated as @xmath61 to the infinite plane wave limit .
the mott - wannier equation in [ eq : mw ] was solved on a radial logarithmic grid ensuring numerical convergence of exciton energies up to @xmath62 ev .
the authors acknowledge support the center for nanostructured graphene ( cng ) , which is sponsored by the danish national research foundation , project dnrf58 .
the authors declare no competing financial interests .
in the main text we argued that the use of supercells is essential for a good description of the band structure of mismatched bilayers and the main differences between the bands of the isolated layers and the bilayers are consequence of charge transfer and hybrization separately . in this section of the supporting information we prove these arguments for the case of mos@xmath32wse@xmath0 based structures .
we start out by demonstrating that using a supercell is unavoidable if an accurate band structure of the mos@xmath32wse@xmath0 is needed .
this is shown in fig .
[ fig : ct_strain ] , where we plot the band structure of the strained bilayer and isolated monolayers in panel ( a ) and the corresponding unstrained structures in panel ( b ) . in both panels , the bands belonging to the isolated layers are drawn with continuous lines while the ones for the bilayers are drawn with circles .
the unstrained bilayer is constructed using a supercell and an alignment angle of @xmath63 as described in the main text , whereas for the strained bilayer we use a unit cell with the lattice parameter equal to the average of the lattice parameter of the isolated monolayers . from the figure it is evident
that straining the layers has a considerable effect both on the curvature of the bands and on their positioning with respect to vacuum .
we thus conclude that accurate band structures can not be obtained without employing supercells .
however , it is still possible to extract information about charge transfer and hybridization from the strained calculations .
indeed , we can see from fig .
[ fig : ct_strain ] that the relative difference between isolated layers and bilayer , in both panels , are practically the same . based on this consideration
we proceed the analysis of charge transfer and hybridization using the strained calculations , which are computationally more feasible .
as explained in the main text , we observe a constant shift in energy , upwards for mos@xmath0 and downwards for wse@xmath0 , and a wavevector dependent variation around @xmath6 when comparing the isolated layers to the bilayer . while the effect of hybridization is a direct consequence of the mixing among wavefunctions of the two layers , charge transfer results from the rearrangement of the electrons at the bilayer interface due to the difference in band gap centers of the two materials . from a dft calculation point of view
, it is the self - consistent procedure , in particular the change in the hartree potential in each loop , that allows the rearrangement of the electrons once the two materials are put together .
this means that performing a non - self - consistent dft calculation starting from the self - consistent ground state density of the isolated layers , would not allow for the update of the hartree potential and consequent electrons rearrangement .
the fully self - consistent band structures ( reported for comparison from fig . [ fig : ct_strain ] ( a ) ) and the non self - consistent ones are shown in fig . [ fig : noct ] , left and right panel respectively . in panel ( b ) , the isolated layers bands are now exactly on top of the bilayer ones throughout most of the brillouin zone and therefore it should be now clear that the rigid shift of the bands was a signature of charge transfer .
furthermore the alteration of the bands around the @xmath6 point has not disappeared .
this is exactly what we expected considering that hybrization is a result of the overlap of the wavefunctions which is accounted for in the non - self - consistent calculation .
[ [ validity - of - the - qeh - correction - on - g_0w_0-band - structure ] ] validity of the qeh correction on g@xmath1w@xmath1 band structure ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ to check the validity of our g@xmath42w@xmath1 approach we perform g@xmath1w@xmath1 calculations for strained mos@xmath0 and wse@xmath0 isolated layers and mos@xmath32wse@xmath0 bilayers .
the choice of strained structure is obviously imposed by the feasibility of a g@xmath1w@xmath1 calculation for the bilayers . for
the following calculation plane - wave mode has been used . in the left panel of fig .
[ fig : test_qeh ] we report the bands for the strained mos@xmath32wse@xmath0 bilayer from a g@xmath1w@xmath1 calculation ( continuous black line ) and the g@xmath1w@xmath1 approach .
the g@xmath1w@xmath1 bands for the isolated layers are also shown as reference .
as expected from the extra screening that each layer provides to the other , the intra and inter layer gaps are reduced compared to the isolated layer ones and such an effect is grasped both from the full g@xmath1w@xmath1 calculation and the g@xmath42w@xmath1 method . however , the agreement between g@xmath1w@xmath1 and g@xmath42w@xmath1 is not striking .
this is because the effect of charge transfer is still present at the g@xmath1w@xmath1 level , since the hartree potential generated by the charge rearrangement at the interface is the same as the dft one
. only self - consistency , indeed , could relieve this problem . to prove that charge transfer is still there and that it is an effect inherited from the starting lda calculation
, we evaluate the layer dependent energy shift at the @xmath8-point by comparing isolated layers and bilayers bands at the lda level and then add these shifts to the g@xmath42w@xmath1 bands .
the results are shown in the central panel of fig .
[ fig : test_qeh ] .
the agreement is nearly perfect and it supports our argument on the importance of charge transfer . as a side note , we mention that the effect of charge transfer using plane - wave mode , as opposed to lcao , is a bit lower , namely we get an increase in interlayer gap of @xmath64 ev compared to the @xmath7 ev reported in the main text . as a further proof of the validity of the g@xmath42w@xmath1 method we repeat the g@xmath1w@xmath1 for the bilayer adding @xmath65 to interlayer distance between mos@xmath0 and wse@xmath0 .
this guarantees that charge transfer and hybridization effects are negligible .
screening effects , on the other hand , are still appreciable being the coulomb coupling between the layers long range .
the bands for such a system are shown in the right panel of fig .
[ fig : test_qeh ] and it is clear that the g@xmath42w@xmath1 does a good job . | van der waals heterostructures ( vdwh ) provide an ideal playground for exploring light - matter interactions at the atomic scale .
in particular , structures with a type - ii band alignment can yield detailed insight into free carrier - to - photon conversion processes , which are central to e.g. solar cells and light emitting diodes .
an important first step in describing such processes is to obtain the energies of the interlayer exciton states existing at the interface . here
we present a general first - principles method to compute the electronic quasi - particle ( qp ) band structure and excitonic binding energies of incommensurate vdwhs .
the method combines our quantum electrostatic heterostructure ( qeh ) model for obtaining the dielectric function with the many - body gw approximation and a generalized 2d mott - wannier exciton model .
we calculate the level alignment together with intra and interlayer exciton binding energies of bilayer mos@xmath0/wse@xmath0 with and without intercalated hbn layers , finding excellent agreement with experimental photoluminescence spectra .
comparison to density functional theory calculations demonstrate the crucial role of self - energy and electron - hole interaction effects .
the use of two - dimensional ( 2d ) transition metal dichalcogenides@xcite as fundamental building blocks in ( opto)electronics has proved highly promising for the construction of ultrathin high performance devices@xcite . by reassembling different 2d crystals into van der waals heterostructures , designer materials with new and tailored properties
can be made@xcite . as for 2d monolayers@xcite ,
the optical properties of few - layer van der waals heterostructures are strongly influenced by excitonic effects@xcite as a consequence of the weak screening of the electron - hole interaction@xcite .
in addition to the intralayer excitons localized in the constituent monolayers , vdwhs can host more complex types of excitons with electrons and holes residing in distinct layers , so - called ( spatially ) indirect excitons or interlayer excitons . because of the spatial charge separation , interlayer excitons posses longer electron - hole recombination lifetimes@xcite than intralayer excitons , which make them ideal candidates for realization of bosonic many - particle states like bose - einstein condensates@xcite .
moreover , interlayer excitons are believed to play a central role in the charge separation process following photoabsorption in solar cells or photodetectors@xcite . of key importance to this process
is the exciton binding energy which quantifies the strength with which the electron and hole are bound together . due to the larger electron - hole separation
, interlayer excitons are expected to have lower binding energies than intralayer excitons . however
, a detailed understanding of interlayer excitons in vdwhs is still lacking mainly because of the highly non - local nature of the dielectric function of 2d materials which makes screening less effective at larger distances .
this is in fact the origin of the non - hydrogenic rydberg series in 2d semiconductors and the non - degeneracy of 2d excitons with different angular momentum quantum numbers@xcite .
the understanding of excitonic effects alone , however , is not sufficient for device engineering , where the knowledge of the alignment of the electronic bands of the vdwh is also crucial .
several experimental investigations have shown , for example , that an underlying type - ii band alignment is required for the formation of interlayer excitons@xcite .
however , experimental data has not been supported by consistent theoretical studies yet .
it is well known that density functional theory ( dft ) calculations are problematic when it comes to prediction of band gaps and band alignment at interfaces and do not take excitonic effects into account .
an important deficit of the dft approach , in addition to its general tendency to underestimate band gaps , is its failure to describe image charge renormalization effects that shift the energy levels of a 2d semiconductor@xcite or molecule@xcite when adsorbed on a polarizable substrates .
ideally , one should employ many - body perturbation theory like the gw approximation to obtain reliable band energies and the bethe salpeter equation ( bse ) for exciton binding energies .
however , the computational cost of such methods make them unfeasible for vdwhs containing more than a few lattice matched 2d crystals . here
, we show how to overcome these limitations , by means of our recently developed quantum electrostatic heterostructure ( qeh ) approach@xcite and accurately calculate interlayer exciton binding energies and electronic bands of vdwhs . for the excitons ,
the qeh allows us to calculate the the screened electron - hole interaction to be used in a generalized 2d mott - wannier model@xcite . for the band energies we use the qeh to modify isolated layer g@xmath1w@xmath1 calculations by including the effect of interlayer screening on the electronic levels .
remarkably we are able to predict band positioning in a vdwh at the cost of , at most , @xmath2-g@xmath1w@xmath1 monolayer calculations , with @xmath2 the number of layers in the stack . in this letter
we apply our method to the case of mos@xmath0-wse@xmath0 bilayers intercalated with a varying number of h - bn layers .
the mos2/hbn / wse2 represents a prototypical type - ii heterostructure .
its well defined atomic structure and the possibility of varying the thickness of the hbn spacer and the relative orientation of the photoactive layers , makes it an ideal platform for studying light - matter processes on atomic length scales . at the same time
, the incommensurate nature of the van der waals interfaces presents a great challenge for ab - initio calculations .
nevetheless , we show that our qeh - based methodology allows us to efficiently simulate the electronic structure of mos@xmath0/hbn / wse@xmath0 , including excitonic and self - energy effects , and accurately reproduce experimental photoluminescence spectra .
+ the main requirement for the existence of interlayer exciton is that the bottom of the conduction band and the top of the valence band in a van der waals stack are located in two different layers .
as shown in the following , this is the case of mos@xmath0-wse@xmath0 based heterostructures .
because of lattice mismatch ( see [ tab : geom ] ) , mos@xmath0 and wse@xmath0 form incommensurable heterostructures and therefore realistic band structure calculations require the use of relatively large in - plane supercells as illustrated in panel ( b ) of [ fig : bands_rotation ] . by rotating the layers with respect to each other , not only
can the dimension of the supercell be reduced , but we can also mimic more closely the experimental situation where the alignment angle between the layers is not controlled . following the procedure described by komsa et al@xcite , we set up the mos@xmath0-wse@xmath0 bilayer for two different alignment angles , specifically @xmath3 and @xmath4 , so that each layer is strained by less than @xmath5 . to be able to compare the band structure of the bilayer with the ones of the constituent isolated monolayers , we unfold the electronic bands of the supercell to the ones of the primitive mos@xmath0 and wse@xmath0 cells .
this is done by following the method described in ref . .
we stress that , because of the lattice mismatch and a non - zero alignment angle , the first brillouin zones ( bz ) of the two materials are different in size and rotated with respect to each other as shown in panel ( c ) and ( e ) of [ fig : bands_rotation ] .
this implies that the unfolding of the bands has to be performed accordingly .
the unfolded band structures , aligned with respect to vacuum level , for the two different bilayers are shown as circles in [ fig : bands_rotation ] ( d ) and ( f ) and compared to the isolated monolayers bands in continuous lines .
the band structures have been calculated using the local density approximation ( lda ) as described in the methods section and for simplicity the effect of spin - orbit coupling is not included here .
we can clearly confirm that mos@xmath0 and wse@xmath0 form a type ii heterostructure , where the top of the valence band is localized on the wse@xmath0 layer , while the bottom of the conduction band belongs to mos@xmath0 .
this implies that mos@xmath0/wse@xmath0 can host interlayer excitons as sketched in [ fig : model_sketch ] ( a ) .
furthermore , no significant difference in the band structure emerges for the two different alignment angles and thus we can conclude that the band structure of the bilayer is independent of the alignment angle . when comparing to the bands of the isolated monolayer , we can distinguish two main effects , as thoroughly shown in the supporting information : the effect of interlayer hybridization , only around the @xmath6 point , and a layer - dependent shift in energy throughout the brillouin zone . for the latter ,
we observe the mos@xmath0 bands to be shifted up in energy while the wse@xmath0 bands are shifted down , with a consequent increase in indirect gap of @xmath7 ev relative to the vacuum level - aligned isolated layers .
such an asymmetric shift is a clear signature of the formation of a dipole at the interface of the two layers as a consequence of charge rearrangement induced by the misalignment between the band gap center of the two materials .
to summarize , since hybrization is minimal and the charge transfer effect , if needed , can be accounted for just by adding a constant shift , we learn that the bilayer bands around the @xmath8-point can be directly obtained as a superposition of the constituent isolated monolayers bands .
this result is consistent with the findings in refs . which show that even for the mos@xmath0 homobilayer , expected to be particularly dependent on hybridization as it is possible to choose twisting angles corresponding to the aa and ab stacking , the band edges at the @xmath8-valley are largely unaltered .
we stress that the band edges at the @xmath8-point are the relevant ones for the formation and radiative recombination ( photoluminescence ) of interlayer excitons .
while in terms of computational cost it would be advantageous to utilize lda band structures for quantitative description of bands in vdwhs , their use is largely questionable when accuracy on band alignment and on band gaps is required . to illustrate the lda failure
, we compare the lda and g@xmath1w@xmath1 electronic bands for the isolated mos@xmath0 and wse@xmath0 monolayers in panel ( a ) of [ fig : bands_dftvsgw ] . in both approaches
we include the effect of spin - orbit coupling at a non - self - consistent level .
while the qualitative picture of a type ii band alignment is preserved within the g@xmath1w@xmath1 approximation , the band alignment and band gaps predicted by lda are wrong .
this is even more evident in panel ( b ) of [ fig : bands_dftvsgw ] , where we directly show the band edges for the isolated mos@xmath0 and wse@xmath0 layers .
one could possibly argue that the inaccuracy is due to the lda exchange correlation functional rather than the dft approach itself .
for this reason we also calculated band edges using the hse06 hybrid functional , which is known to perform well for band structures . however , as shown in panel ( b ) , hse is better than lda but still not as accurate as g@xmath1w@xmath1 .
we thus conclude that the g@xmath1w@xmath1 approximation is essential to obtain a quantitatively correct description of the band gaps and band alignment .
importantly , we notice that while dft predicts a rather large difference in band gap centers of around @xmath9 ev , g@xmath1w@xmath1 gives a much smaller difference of around @xmath10 ev .
this , together with the increased band gaps predicted by g@xmath1w@xmath1 , strongly indicates that the charge transfer and associated dipole shift of the bands , could be significantly overestimated by dft - lda .
we thus conclude that the band structure of the heterobilayer around the k - point can be obtained by combining the g@xmath1w@xmath1 band structures of the isolated layers aligned with respect to a common vacuum level and corrected for image charge screening effects ( see later ) .
+ the state of the art for describing excitonic effects from first principles is the many - body bethe - salpeter equation ( bse)@xcite . the solution of the bse is , however , computationally demanding already at the monolayer level and practically impossible for incommensurate van der waals heterostructures .
however , it is well - known that , under well defined assumptions the excitonic many - body problem can be rephrased in terms of an effective hydrogenic hamiltonian , the mott - wannier hamiltonian@xcite , which gives a satisfactory description of several excitonic properties@xcite . in the case of excitons in vdwhs ,
the motion of the electron and the hole is restricted to the in - plane direction .
this is a direct consequence of the anisotropy of layered structures , which entails that the effective masses at the k - point in out - of - plane direction are much higher than the in - plane direction . with this consideration
the hydrogenic hamiltonian reduces to a 2d problem : @xmath11f({\bf r}_\parallel)=e_\textrm{b}f({\bf r}_\parallel),\ ] ] where @xmath12 is the exciton effective mass and @xmath13 is the electron - hole interaction energy .
the exciton effective mass is evaluated as @xmath14 , where the hole and electron masses are calculated ab - initio and reported in [ tab : geom ] .
[ tab : geom ] in the case of interlayer excitons [ eq : mw ] is still valid .
indeed even if the electron and the hole are separated in the out - of - plane direction , their motion is still confined in their respective layers .
on the other hand , this spatial separation affects the screened electron - hole interaction energy , as shown below . in the specific case of mos@xmath0/wse@xmath0 heterostructures ,
the electron and hole effective masses have to be estimated from the wse@xmath0 and mos@xmath0 valence and conduction bands respectively .
we found an interlayer exciton effective mass of @xmath15 a.u . the coulomb interaction between the electron and the hole in an exciton is highly sensitive to the dielectric properties of the material@xcite . as we showed in our recent work @xcite ,
the dielectric screening of finite thickness vdwhs is strongly non local . using our recently developed quantum - electrostatic heterostructure ( qeh ) model
, we can determine the dielectric properties of these complex structures from first principles@xcite .
briefly , we first condense the dielectric response of each single layer into a `` dielectric building block '' consisting of the monopole and dipole components of the density response function of the isolated layer@xcite .
second , the dielectric building blocks are coupled together electrostatically by solving a dyson - like equation in the discrete monopole / dipole basis in order to obtain the density response function for the whole structure .
the underlying assumption of the qeh is that hybridization is weak enough that is does not affect the dielectric properties of the heterostructure .
we have found that this approximation is surprisingly good .
a collection of more than 50 dielectric building blocks together with the software for the electrostatic coupling can be found in ref . from the response function ,
the dielectric function of the heterostructure is determined and it can be used to obtain the screened electron - hole interaction energy : @xmath16 where @xmath17 ( @xmath18 ) is the electron density ( hole induced potential ) vector expressed in a basis set of monopole / dipole densities ( potentials ) .
the basis set of induced densities and potentials is also used as ( left and right ) basis functions for representing @xmath19 ( see ref . ) .
the underlying structure of a vector in the multipole basis can be readily understood .
an arbitrary density vector @xmath20 is represented as @xmath21 $ ] where @xmath22 , with @xmath23 , is the induced monopole / dipole density at the layer @xmath24 .
a completely equivalent representation can be used for the potential .
now , for the specific case of an interlayer exciton in the mos@xmath0/wse@xmath0 bilayer , the electron density vector takes the form @xmath25 $ ] , while the potential induced by the hole is @xmath26 $ ] . because the electron and hole distribution do not have dipole components , we set @xmath27 , @xmath28 and @xmath29 , @xmath30 .
it is useful to define an effective dielectric function for the electron - hole interaction as the unscreened coulomb interaction over the screened one : @xmath31 a typical signature of excitons in two - dimensional materials is the non - hydrogenic rydberg series@xcite .
the rydberg series along with the full wave - vector dependent effective dielectric screening and electron - hole interaction for the intra and inter - layer excitons in the mos@xmath32wse@xmath0 bilayer are shown in [ fig : rydbergseries ] ( a ) , ( b ) and ( c ) respectively .
the non - hydrogenic nature of the rydberg series should be clear from the comparison with the dashed line in panel ( a ) which represents the rydberg series obtained from an hydrogenic equation where the electron - hole interaction is screened by a constant dielectric function .
the first interesting characteristic to notice is that the intra and interlayer rydberg series converge towards each others for higher excited excitonic states .
this is not surprising considering that higher lying states are more delocalized and once their radius is much greater than the heterostructure thickness intra and interlayer excitons are practically indistinguishible , provided that the screening of the electron - hole interaction is comparable . as shown in panel ( b ) , the effective dielectric function is indeed the same for inter and intra - layer excitons within the region of relevant wavevectors values , i.e. values of @xmath33 smaller than the reciprocal of the exciton radius ( indicated with a vertical line in panel ( b ) for the lowest lying exciton ) .
however , it is clear from panel ( c ) that the screened interaction for the interlayer exciton is lower than for the intralayer ones , and it does not diverge for @xmath34 . this is a simple consequence of the finite electron - hole spatial separation , which guarantees that the electron and the hole are separated even for @xmath35 . to explore the effect of spatial separation even further ,
we study the case of mos@xmath0-wse@xmath0 heterostructures intercalated with h - bn . to isolate the effect of screening , we perform the same calculations with h - bn is substituted by vacuum . the results for the lowest intra and inter - layer exciton binding energies are shown in panel ( a ) and ( b ) of [ fig : distance_dep ] .
clearly the behavior of the interlayer exciton is quite similar in the two cases , meaning that the main effect of inserting hbn layers is to increase the electron - hole separation .
in contrast , the binding energy of the intralayer excitons increases in the case of increasing vacuum while it remains constant when more hbn layers are inserted .
this is because moving mos@xmath0 and wse@xmath0 apart decreases the screening whereas inserting more hbn layers increases it , leading to an overall compensation .
+ when stacking 2d layers together , the exciton binding energy is not the only quantity that is affected .
indeed , also the band gap of each of the layers in the stack is reduced due to the increased dielectric screening .
the state of the art method for properly including the effect of dielectric screening in band structures is the g@xmath1w@xmath1 method@xcite . in this many - body theory
based approach , the information of electronic screening is contained in the dynamical screened coulomb potential : @xmath36\frac{4\pi}{|{\bf g}+{\bf q}|^2}.\ ] ] computing @xmath37 is a demanding task even for simple materials and it is practically impossible for multi - layer vdwhs .
fortunately , as described elsewhere @xcite , the effect of screening on the band structure of a given layer in a van der waals stack , can be accounted for by combining the qeh model with a standard g@xmath1w@xmath1 method at the computational cost of a monolayer calculation .
the main idea is to correct @xmath38 for a given layer in the following manner : @xmath39 where @xmath40 is the correction to the head of the matrix @xmath41 that includes the extra screening coming from the neighboring layers . for a given layer ,
such a correction is efficiently calculated within the qeh model .
indeed , by using an expression equivalent to [ eq : w_q_vdwh ] , the electron - electron interaction is calculated for the isolated layer and the layer in the vdwh , then @xmath40 is obtained as the difference between the two interactions . once corrected , the screened potential can be used directly in a standard monolayer g@xmath1w@xmath1 calculation . with this approach , which we refer to as g@xmath42w@xmath1 , we are able to efficiently calculate the band positions of any vdwhs and , specifically for this work , the position of the valence band maximum and conduction band minimum of mos@xmath0 and wse2@xmath0 for a varying number of hbn layers . a verification of the g@xmath42w@xmath1 approach for the specific mos@xmath32wse@xmath0 system is provided in the supporting information . the level alignment for a bilayer mos@xmath0-wse@xmath0 obtained from the g@xmath42w@xmath1 and including spin - orbit coupling effects is shown in [ fig : model_sketch ] ( a ) in black lines . to highlight the effect of interlayer screening we reported , in the same panel , the band edges for the isolated monolayers calculated within the g@xmath1w@xmath1 approximation ( colored bars ) .
( b ) , instead , illustrates the difference in intra and interlayer gaps with respect to the isolated layers as a function of the number of intercalated hbn layers .
although the band gap renormalization is noticeable when going from monolayers to bilayer , the intercalation of h - bn does not have a considerable effect .
comparing to the variation in exciton binding energy in panel ( c ) , we observe a similar trend for the intralayer cases whereas for the interlayer case the trend is different .
this is reasonable since , differently from the interlayer exciton , there is not explicit dependence of the interlayer gap on mos@xmath0-wse@xmath0 separation . with the knowledge of band edges position and exciton binding energies we are now ready to calculate the position of the excitonic photoluminescence ( pl ) peaks in mos@xmath0/wse@xmath0 based heterostructures .
the photoluminescence signal is generated by radiative electron - hole pairs recombination .
considering that typical radiative recombination times in tmdcs are much longer than electron and hole thermalization times , we expect the exciton recombination to happen from the k - point of the conduction band in mos@xmath0 and the k - point of the valence band in wse@xmath0 .
we note that , the lattice mismatch and a non - zero alignment angle between the two layers implies a mismatch of the first bz of the two materials , as shown in [ fig : bands_rotation ] .
this means that for a radiative transition to happen the momentum mismatch has to be compensated by some other physical mechanism
. mechanisms of this kind could include phonon assisted transition , electron - electron interaction , defect scattering or breaking of momentum conservation induced by the exterior potential field generated by the neighboring layers@xcite .
here we focus on the energetics of the process , which should not be effected by the particular recombination mechanism . taking into account the type - ii band alignment ,
the position of the pholuminescence peak of the lowest bound exciton is given by : @xmath43 where @xmath44 is the interlayer electronic gap and @xmath45 the interlayer exciton binding energy .
the positions of the lowest energy photoluminescence peak for isolated layers , as well as for mos@xmath0-wse@xmath0 based heterostructures with a varying number of intercalating hbn layers are plotted in [ fig : indirect_peak ] for the free standing and supported case .
experimental photoluminescence spectra from ref .
are also reported in the same figure .
for the supported case we use @xmath46 layers of hbn to simulate the effect of a substrate .
the choice of hbn as a substrate is enforced by the qeh approach which applies only to layered materials , but it is justified by the fact that hbn has a bulk dielectric constant similar to sio@xmath0 , which is the substrate used in the experiment . the agreement with the photoluminescence peaks for mos@xmath0 it is good but for wse@xmath0 it is underestimated by around @xmath47 ev .
roughly the same constant shift is seen for the heterostructures .
this indicates that the deviation is due to a too high positioning of the wse@xmath0 valence band by the g@xmath42w@xmath1 .
however , the agreement is still highly satisfactory . indeed shifting our values by @xmath48
we can well reproduce the data for increasing number of intercalating layers as shown in the inset .
this indicates that it is the estimated indirect bandgap to be slightly off , but a difference of just @xmath47 ev is still quite accurate for our multi - step approach . according to the experimental interpretation in ref . , while the pl peak for the bilayer without hbn shows a clear interlayer exciton peak , with the intercalation of hbn the interlayer exciton signal is reduced and eventually covered by the wse@xmath0 intralayer exciton pl peak for three h - bn layers . if this is the case , it is not possible to completely validate the trend of our ab - initio pl peaks values through the actual experimental data .
anyways to demonstrate that using g@xmath1w@xmath1 for monolayer bands is strictly necessary , [ fig : indirect_peak ] shows the photoluminescence peaks obtained using a band alignment from lda and hse calculations .
it is evident that the lda dramatically underestimates the position of the photoluminescence peaks for both the isolated layer and the bilayer case .
the hse improves the band alignment significantly , but is still around @xmath10 ev below the experimental values .
furthermore even the trend of the indirect exciton peak as a function of hbn layers is reversed by lda .
this is a consequence of the strong charge transfer predicted by lda as discussed earlier . indeed
, charge transfer tends to open the indirect band gap and therefore shift the pl peaks up in energy , with a shift that decreases with increasing number of hbn layers .
this shift in energy is larger and opposite to the optical band gap reduction due to interlayer excitons , which explains why the position of the lda pl peaks descrease in energy in contrast with the g@xmath1w@xmath1 results . in conclusion
, we presented a general approach to calculate band alignment and interlayer excitons in incommensurate van der waals heterostructures .
for the mos@xmath0/wse@xmath0 heterostructure , we found that interlayer hybridization is important only around the @xmath6-point of the bz and therefore does not influence the opto - electronic properties that are governed by states around the k - point .
this implies that an accurate description of the band edge positions can be obtained from the isolated monolayer band structures aligned relative to a common vacuum level and renormalized by the polarization effect from neighboring layers . in particular , we determined the effect of interlayer polarization to account for @xmath49 reduction in the intelayer gap . on the other hand , the effect on band alignment of the formation of an interface dipole is expected to be negligible for the studied system .
we find interlayer excitons to have significant binding energies of up to @xmath10 ev and showing monotonic decrease with the layer separation .
comparison with experimental photoluminescence spectra revealed a constant redshift of the calculated lowest optical transition of around @xmath47 ev , which we ascribe to a slight overestimation of the wse@xmath0 valence band edge by g@xmath1w@xmath1 .
this indicates that for describing photoluminescence it is crucial to obtain accurate band edges at the isolated layers level .
finally , our calculations show that it is possible to obtain quantitatively accurate band- and exciton energies for rather complex vdwhs when employing proper methods , and highlight the deficiencies of standard density functional theory for band alignment problems . |
the open charm production in collisions of high energy hadrons is an appropriate class of processes to test the perturbative quantum chromodynamics ( qcd ) @xcite . a large mass of charm quark @xmath16 as a lower bound of hard energy scale @xmath17 leads to the small value of the strong coupling constant @xmath18 due to the condition @xmath19 , where @xmath20 is the asymptotic scale parameter of qcd
. production of charm quark with large transverse momentum ( @xmath21 ) is a typical multiscale hard process for which the fixed - order qcd calculations should be corrected by the fragmentation contribution to resum large logarithmic terms @xmath22 , where @xmath23 is the strong coupling constant .
nowadays , the theoretical study of inclusive @xmath0-meson hadroproduction were performed at the leading order ( lo ) for double production ( @xmath24 ) and the next - to - leading order ( nlo ) for single production in the collinear parton model ( cpm ) of qcd within the two approaches : the general - mass variable - flavor - number ( gm - vfn ) scheme @xcite , and the so - called fixed order scheme improved with next - to - leading logarithms ( fonll scheme ) @xcite . in the first one ,
realized in the refs .
@xcite , the large transverse momenta collinear logarithms are resummed through the evolution of the fragmentation functions ( ffs ) , satisfying the dokshitzer - gribov - lipatov - altarelli - parisi ( dglap ) evolution equations @xcite .
the @xmath0-meson ffs were extracted both at the lo and nlo in the fixed factorization scheme from the fit of @xmath15 data taken by the opal collaboration at cern lep1 @xcite .
opposite , in the fonll approach , the nlo @xmath0-meson production cross sections are calculated with a non - perturbative @xmath14-quark ff , that is not a subject to dglap evolution .
the fonll scheme was implemented in the refs .
@xcite and its main ingredients are the following : the nlo fixed order calculation with resummation of large transverse momentum logarithms at the next - to - leading level for heavy quark production .
the important difference of these two fragmentation approaches is in following : in the case of scale - depended gm - vfn scheme @xmath0-mesons are produced both from @xmath14-quarks and from gluons , in case of scale - independent fonll scheme @xmath0-mesons are produced only from @xmath14-quarks . at the high - energy regge limit @xmath25 , where @xmath26 is the invariant collision energy , one has new dynamic regime of particle production in the multi - regge kinematics ( mrk ) or in the quasi - multi - regge kinematics ( qmrk ) , when one particle or a group of particles are produced in the rapidity region , being strongly separated in rapidity from other particles .
radiative corrections in this regime are dominated by the production of additional hard jets .
the only one way in cpm to treat these processes is to calculate higher - order corrections , which is a challenging task for some processes even at the nlo level , such as for relevant to our study process of double charm production in the gluon fusion , @xmath27 . alternatively , to solve this problem we should change a factorization scheme from the collinear approximation of the pm to the high - energy or @xmath28-factorization @xcite and take into account a large part of the higher - order corrections which are incorporated in the transverse - momentum - dependent parton distribution functions ( pdfs ) of off - shell initial partons .
recently , the studies of single and double @xmath0-meson hadroproduction at the lhc were performed in the @xmath28-factorization approach with the scale - independent peterson @xmath14-quark ff in refs .
@xcite and with the scale - dependent @xmath14-quark and gluon ffs from @xcite in refs .
the latter are based on the parton reggeization approach ( pra ) , which is a combination of @xmath28-factorization framework with the fully gauge - invariant amplitudes with reggeized partons in the initial state . in our previous study of the single @xmath0-meson production at the tevatron and the lhc @xcite in the central rapidity region of produced mesons ( @xmath29 and @xmath30 , respectively ) in the pra we obtained quite a good agreement between theoretical calculations and the experimental data . in this work
we continue the investigation and extend it into the region of large rapidity for single and double @xmath0-meson production .
not so long ago the lhcb collaboration has provided the forward @xmath0-meson production data @xcite .
the differential cross sections on @xmath0-meson transverse momentum @xmath31 were measured for @xmath1 , @xmath2 , @xmath32 , and @xmath4 mesons in proton - proton collisions at the cern lhc at a center - of - mass energy of @xmath7 tev and with rapidities in the range of @xmath33 . nowadays ,
the recent data come from the lhc run ii and the newest measurements of @xmath0-meson production were presented by the lhcb collaboration at @xmath8 tev @xcite .
they are presented as @xmath9 distributions of @xmath0 mesons at the same rapidity region and in a wider @xmath9 range then in the lhc run i. the data are presented in three kinds : absolute double differential cross sections @xmath34 , ratios of @xmath0-meson production cross sections between different center - of - mass energies @xmath35 , and ratios of the cross sections for different mesons . not only the single but also a pair production of @xmath0 mesons is under consideration .
the first observations of double charm production were performed at the @xmath7 tev by the lhcb collaboration @xcite .
the published results include the cross sections differential in azimuthal angle difference , transverse momentum , rapidity difference and invariant mass of various combinations of @xmath0-meson pairs .
the spectra were measured at the region of large rapidities @xmath36 at the collision energy @xmath7 tev .
we continue our recent study of double @xmath0-meson production in pra @xcite testing the new contributions from reggeized quark
reggeized antiquark annihilation processes and studying different combinations of @xmath0-mesons in the measured pair production spectra .
recently , pra was successfully applied for the analysis of inclusive production of single jet @xcite , pair of jets @xcite , prompt - photon @xcite , photon plus jet @xcite , drell - yan lepton pairs @xcite , bottom - flavored jets @xcite , charmonium and bottomonium production @xcite at the tevatron and lhc .
this study has demonstrated the advantages of the high - energy factorization scheme based on pra in the descriptions of data comparing to the cpm calculations .
this paper is organized as follows . in sec .
[ sec : two ] we present the relevant reggeized amplitudes as they are obtained from the lipatov s effective high - energy action in qcd . in sec .
[ sec : three ] the formalism of our calculation in the pra and the fragmentation model are discussed . in secs .
[ sec : four ] and [ sec : five ] our results for single and double @xmath0-meson production , respectively , are presented in comparison with the experimental data and discussed . in sec .
[ sec : six ] we summarize our conclusions .
[ sec : two ] the reggeization of amplitudes at the high energy is a natural property of gauge - invariant quantum field theories , such as quantum electrodynamics @xcite and qcd @xcite . at large @xmath37
the dominant contribution to cross sections of qcd processes give mrk or qmrk parton scattering processes with the gluon or quark @xmath38-channel exchange . due to the reggeization of quarks and gluons ,
an important role is dedicated to the vertices of reggeon - particle interactions .
nowadays , they can be straightforwardly derived from the non - abelian gauge - invariant effective action for the interactions of the reggeized partons with the usual qcd partons , which was firstly introduced in ref .
@xcite for reggeized gluons only , and then extended by inclusion of reggeized quark fields in the ref .
the full set of the induced and effective vertices together with feynman rules can be found in refs .
@xcite . in our recent paper @xcite we have shown that gluon and @xmath14-quark fragmentation
give the leading contribution to the @xmath0-meson production in the pra .
the same result was also obtained in the ref .
@xcite in nlo of the cpm , as well as the light quark fragmentation turns out to be negligible . according to this statement in case of the forward production of charmed mesons in the framework of the pra we will also consider the gluon and @xmath14-quark fragmentation . the lowest order in @xmath39 parton subprocesses of pra which give a contribution to a single @xmath0-meson production via gluon fragmentation
are the following @xmath40 and @xmath41 the corresponding lowest order parton subprocesses which contribute to a single @xmath0-meson production via @xmath14-quark fragmentation are charm quark - antiquark pair production in the gluon - gluon fusion @xmath42 and in the quark - antiquark annihilation @xmath43 where @xmath44 are the reggeized gluons and @xmath45 denotes reggeized @xmath46 , @xmath47 and @xmath48 quarks .
the double @xmath49-meson production at the lowest order of pra is described by the parton subprocesses ( [ eq : rrqq ] ) and ( [ eq : qqqq ] ) of @xmath50-pair production as well as by the parton subprocesses of @xmath51-pair production @xmath52 @xmath53 in the case of @xmath6-meson production we should consider contributions from six parton subprocesses . at first , these are ( [ eq : rrgg ] ) and ( [ eq : qqgg ] ) .
there are also @xmath54 and @xmath55 subprocesses @xmath56 @xmath57 @xmath58 @xmath59 formally , the matrix elements of the @xmath60 , @xmath61 and @xmath55 subprocesses are of different order in the strong coupling constant @xmath39 , but their contributions can be of the same order because the fragmentation probabilities for gluon and @xmath14-quark fragmentation to a @xmath0-meson are related as @xmath62 . taking into account our previous study @xcite and the results of ref .
@xcite , we conclude that gluon fragmentation dominates over the @xmath14-quark fragmentation in the @xmath6-pair production at the lhc energy , so we will consider the contributions of ( [ eq : rrgg ] ) and ( [ eq : qqgg ] ) subprocesses only .
let us define four - vectors @xmath63 and @xmath64 , where @xmath65 are the four - momenta of the colliding protons , and @xmath66 are their energies .
we have @xmath67 , @xmath68 , and @xmath69 .
for any four - momentum @xmath70 , we define @xmath71 .
it is easy to see that the four - momenta of the reggeized gluons can be represented as @xmath72 where @xmath73 .
then the amplitude of gluon production in a fusion of two reggeized gluons can be presented as a convolution of the fadin - kuraev - lipatov effective prr vertex @xmath74 and polarization four - vector of final gluon @xmath75 : @xmath76 where @xmath77 , \label{amp : rrg}\ ] ] @xmath78 and @xmath79 are the color indices of the reggeized gluons with incoming four - momenta @xmath80 and @xmath81 , and @xmath82 with @xmath83 is the antisymmetric structure constant of color gauge group @xmath84 .
similarly , we can write down the amplitudes for subprocess ( [ eq : qqg ] ) with massless spinors of reggeized quark @xmath45 and antiquark @xmath85 , @xmath86 and @xmath87 respectively : @xmath88 the fadin - sherman effective vertex @xcite is given by @xmath89 .
\label{amp : qqg}\end{aligned}\ ] ] where @xmath90 are the generators of the color gauge group @xmath84 , @xmath91 and @xmath92 then the squared amplitudes have a simple form and read : @xmath93 where @xmath94 , @xmath95 . the amplitudes for the subprocesses ( [ eq : rrqq ] ) , ( [ eq : qqqq ] ) , ( [ eq : rrgg ] ) , and ( [ eq : qqgg ] ) , can be written as a convolution of corresponding effective vertices with final gluon polarization four - vectors @xmath96 in case of gluon production or with spinors @xmath86 and @xmath87 in case of initial - state reggeized quark @xmath97 and antiquark @xmath98 : @xmath99 where @xmath100 are the effective vertices of the transition of two reggeized partons into the gluon pair or charm quark - antiquark pair .
the general form of the squared amplitudes is : @xmath101 where @xmath102 and @xmath103 are functions which depend on variables @xmath48 , @xmath38 , @xmath46 , @xmath104 , @xmath105 , @xmath106 , @xmath107 , @xmath108 .
the corresponding effective vertices and matrix elements squared are too large to be presented here and can be found in an explicit form in ref .
@xcite in the case of the massless final charm quarks(antiquarks ) .
we perform our analysis in the region of @xmath109 , that allows us to use zm vfns scheme , where the masses of the charm quarks in the hard - scattering amplitude are neglected . for all the squared amplitudes of @xmath60 subprocesses we checked that in the collinear limit , i.e. @xmath110 , the reggeized squared amplitudes transform to the squared amplitudes of the corresponding parton subprocesses in cpm .
it is evident that the squared amplitudes for @xmath111 subprocesses are vanishing in the collinear limit .
[ sec : three ] in the conventional cpm the factorization theorem allows to present the cross sections of processes which take place in high - energy hadron - hadron collisions and have a one hard scale @xmath112 as a convolution of scale - dependent parton ( quark or gluon ) _ collinear _ distributions @xmath113 and squared amplitude of the hard parton scattering .
these distributions represent the density of partons carrying the longitudinal momentum fraction @xmath114 of the proton , integrated over the parton transverse momentum @xmath28 up to @xmath115 .
the evolution of this density from some scale @xmath116 , fixing a non - perturbative regime , to the typical scale @xmath117 is described by dglap @xcite evolution equations where the large logarithms of type @xmath118 ( collinear logarithms ) are summed .
the typical scale @xmath117 of the hard - scattering processes is usually of order of the transverse momentum of the produced particle , @xmath119 .
the pra gives the description of qcd parton scattering amplitudes in the region of large @xmath108 and fixed momentum transfer @xmath38 , @xmath120 ( regge region ) , with various color states in the @xmath38-channel . in the regge region
we should keep the transverse momenta of the initial partons and their virtualities .
it becomes possible introducing the unintegrated over transverse momenta parton distribution functions ( unpdfs ) @xmath121 , which depend on parton transverse momentum @xmath122 while its virtuality is @xmath123 .
the unpdfs are defined to be related with collinear ones through the equation : @xmath124 the factorization formula in the pra reads : @xmath125 where @xmath126 , @xmath127 , @xmath128 and @xmath129 are the factorization and renormalization scales respectively , @xmath130 is the unpdf of an @xmath131-parton in the initial state and @xmath132 is the cross section of the hard off - shell partonic subprocess .
the cross section ( [ eq : kttheorem ] ) is normalized to be in accordance with the cpm at the collinear limit , when @xmath110 .
according to the factorization formula ( [ eq : kttheorem ] ) the partonic cross section of the @xmath111 subprocess ( as an example we take ( [ eq : rrg ] ) ) is being a convolution of gluon production squared amplitude ( [ sqamp : rrg ] ) with unpdfs , and it can be simplified to the following formula : @xmath133 where @xmath134 is the azimuthal angle between @xmath135 and @xmath136 .
the hadronic cross section of any @xmath137 subprocess , here we show the case of heavy - quark pair production through the process ( [ eq : rrqq ] ) , can be written as follows : @xmath138 where @xmath139 , @xmath140 , @xmath141 is the azimuthal angle between @xmath142 and @xmath143 , the rapidity of the final parton with four - momentum @xmath144 can be presented as @xmath145 .
again , we have checked a fact that in the limit of @xmath146 , we reproduce the conventional factorization formula of the collinear parton model from ( [ eq : qmrkg ] ) and ( [ eq : qmrkqq ] ) .
the important ingredient of the our calculation is unpdf , which we take as one proposed by kimber , martin and ryskin ( kmr ) @xcite .
these distributions are obtained introducing a single - scale auxiliary function which satisfies the unified bfkl / dglap evolution equation , where the leading bfkl logarithms @xmath147 are fully resummed and even a major ( kinematical ) part of the subleading bfkl effects are taken into account .
the dependence on the second scale , @xmath38 , is implemented at the last step of the evolution .
this procedure to obtain unpdfs requires less computational efforts than the precise solution of two - scale evolution equations such as , for instance , ciafaloni - catani - fiorani - marchesini equation @xcite , but we found it to be suitable and adequate to the physics of processes under study .
the set of feynman diagrams corresponding to the contributions proportional to @xmath148 included into unpdfs appear if we consider the reggeization of given parton .
that makes unpdfs fully compatible with reggeized amplitudes . in our previous study
@xcite devoted to the similar processes of @xmath0-meson production we proved that they give the best description of @xmath149spectra measured at the central region of rapidity at the lhc . to describe the hadronization stage we use the fragmentation model @xcite .
so , the transition from the produced gluon or @xmath14-quark to the @xmath0 meson is described by corresponding fragmentation function @xmath150 . according the factorization theorem in qcd , in the fragmentation model
the basic formula for the single @xmath0-meson production cross section reads : @xmath151 where @xmath152 is the fragmentation function for the parton @xmath131 splitting into @xmath0-meson at the scale @xmath112 , @xmath153 is the longitudinal momentum fraction of a fragmenting particle carried by the @xmath0-meson . in the zero - mass approximation
the fragmentation parameter @xmath153 can be defined as follows @xmath154 , @xmath155 and @xmath156 are the @xmath0-meson and @xmath131-parton four - momenta , and @xmath157 . in the case of pair @xmath0-meson production
an additional integral over the momentum fraction of second parton appears .
the fragmentation formula now has the following form : @xmath158 where @xmath159 and @xmath160 are the fragmentation functions for the transitions of parton @xmath161 into @xmath0 meson with momentum fraction @xmath162 and of parton @xmath163 into @xmath164 meson with momentum fraction @xmath165 , respectively , at the scale @xmath112 . in our calculations
we use the lo ffs from ref .
@xcite , where the fits of non - perturbative @xmath1 , @xmath2 , @xmath3 , and @xmath4 ff s to opal data from lep1 @xcite were performed .
these ffs satisfy two desirable properties : at first , their scaling violations are ruled by dglap evolution equations ; at second , they are universal . as the contribution of gluon fragmentation at @xmath166
is initiated by the perturbative transition of gluons to @xmath167-pairs encountered by dglap evolution equations , the part of @xmath14-quarks produced in the subprocess ( [ eq : rrqq ] ) with their subsequent transition to @xmath0-mesons are already taken into account considering @xmath0-meson production via gluon fragmentation in the subprocess ( [ eq : rrg ] ) .
such a way , to avoid double counting , we must subtract this contribution , that can be effectively done by the lower cut of the amplitude in formula ( [ eq : qmrkqq ] ) as @xmath168 , i.e. at the threshold of the production of @xmath50-pair .
the forward rapidity region in @xmath169 collisions at the lhc became available due to the specially designed lhcb detector where the measurements of differential cross sections of @xmath1 , @xmath2 , @xmath32 , and @xmath4 mesons with @xmath33 at the @xmath170 tev were performed @xcite .
the observed data divided into 5 rapidity regions are presented together with our results obtained in the lo of pra in the fig .
[ fig:1 ] .
the sum of all contributions from subprocesses ( [ eq : rrg])-([eq : rrqq ] ) is shown as solid line .
we estimated a theoretical uncertainty arising from uncertainty of definition of factorization and renormalization scales ( @xmath171 ) by varying them between @xmath172 and @xmath173 around their central value of @xmath9 , the transverse momentum of fragmenting parton .
the resulting uncertainty is depicted in the figures by shaded bands .
we find a good agreement between our predictions and experimental data in the whole @xmath9 interval of @xmath0-meson transverse momenta within experimental and theoretical uncertainties only with one exclusion for @xmath174-mesons , when our results overestimate data by a factor 2 .
it may be connected with the bad quality of gluon or @xmath14-quark ff extracted from the @xmath15 data , see @xcite for details . in the calculations of ref .
@xcite , which were done in the similar approach with kmr unpdfs @xcite and using scale - dependent ffs kkks08 @xcite , the underestimation of experimental data from lhcb collaboration for @xmath175meson transverse momentum spectra by a factor 2 or more was found .
but , in the ref . @xcite the only @xmath176quark fragmentation into @xmath177mesons was applied .
we correct this results taking into account gluon to @xmath177meson fragmentation and obtain good agreement with the data .
we also provide theoretical calculations for @xmath177production within the lhcb acceptance at the energy @xmath178 tev .
we compare them with the most recent data from the lhcb @xcite .
the results are presented as double differential distributions @xmath179 for @xmath1 , @xmath2 , @xmath32 , and @xmath4 mesons in the same rapidity region as at the @xmath180 tev in the fig .
[ fig:2 ] .
our predictions for @xmath174 mesons became better at the @xmath8 tev than at the smaller energy and they are very close to data .
again we estimate a theoretical uncertainty varying the factorization and renormalization scales up and down by the factor @xmath181 around the central value and depict it by the shaded bands at the plot .
in addition to the differential cross sections the lhcb collaboration presented ratios between differential cross sections at 13 and 7 tev in the transverse momentum region @xmath182 gev .
@xmath183 we present our calculations of the ratio in the fig .
[ fig:3 ] .
we obtain a good agreement of our predictions with the experimental data for the @xmath1 , @xmath2 , and @xmath184 mesons at the whole ranges of their transverse momenta and rapidities . in case of @xmath174 mesons ,
the agreement is not so good because of the overestimation of data by theoretical predictions at the @xmath7 tev .
a comparative study of correlation observables in @xmath5- and @xmath6-pair production is an important test of @xmath177meson production mechanism . as it was found earlier in ref .
@xcite in case of particle plus antiparticle production ( @xmath5 ) the data from lhcb collaboration @xcite can be described well in the @xmath185factorization approach with inclusion of subprocess ( [ eq : rrqq ] ) only and using the scale - independent @xmath176quark peterson ff . working in the same manner , to describe @xmath6 pair production via @xmath14-quark fragmentation into @xmath0-meson , we should include contribution from @xmath55 subprocess ( [ eq : rrqqqq ] ) as it was done in ref .
it was obtained that predictions based on the model of @xmath0-meson production via @xmath14-quark fragmentation can not describe data for correlation observables in @xmath6-pair production absolutely and inclusion of double parton scattering ( dps ) production mechanism is needed .
the inclusion of a gluon fragmentation channel , which is the dominant one in the @xmath0-meson production at the lhc energy @xcite , allows us to describe data for @xmath6-pair production well without hypothesis about dps production mechanism @xcite .
now we start with discussion of @xmath5-pair production at the lhcb . in fig .
[ fig:4 ] , we compare our predictions with the lhcb data at the @xmath7 tev for @xmath186-pair production normalized spectra over the azimuthal angle difference @xmath187 , @xmath0-meson transverse momenta @xmath9 , rapidity distance @xmath188 and invariant mass of @xmath5 pair @xmath13 . as it is estimated , the dominating contribution is gluon - gluon fusion into @xmath50 pair with @xmath176quark fragmentation into the @xmath1 meson and with @xmath189quark fragmentation into the @xmath190 meson ( dash - dotted line ) .
the contribution from gluon - gluon fusion into two gluons when the first gluon fragments into @xmath1 meson and second gluon fragments into @xmath190 meson ( dashed line ) is lying below the leading contribution by the order of magnitude .
two next by value contributions from quark - antiquark annihilation into @xmath50 pair ( double - dot - dashed line ) or into two gluons ( triple - dot - dashed line ) are even smaller and we will ignore these ones below . in the figs .
( [ fig:5])-([fig:8 ] ) , we demonstrate that our predictions for correlation spectra of @xmath191 , @xmath192 , @xmath193 and @xmath194 are in a good agreement with the experimental data .
the production of @xmath6 pairs is mostly generated by the gluon fragmentation into the @xmath0 meson in the subprocess of gluon - gluon fusion .
the contribution of two - gluon production in the quark - antiquark annihilation is suppressed by two orders of magnitude in the same way as the contributions from the channel of @xmath195 double - pair production . in the figs .
( [ fig:9])-([fig:10 ] ) , we compare our predictions based on two - gluon fragmentation mechanism with the experimental data for correlation spectra of @xmath196 and @xmath197 , correspondingly . in the fig .
( [ fig:11 ] ) , the transverse momentum spectra for @xmath198 , @xmath199 and @xmath200 pairs are presented .
we found a good agreement between predictions obtained in the lo pra calculations and experimental data from lhcb collaboration for all @xmath6-pair correlation observables and there is no place for contribution of dps production mechanism , as it is discussed in literature .
we introduce a comprehensive study of single and pair fragmentation production of @xmath0 mesons in proton - proton collisions at the energies @xmath7 and @xmath8 tev in the forward rapidity region at the lhc in the framework of parton reggeization approach .
we use the gauge invariant amplitudes of hard parton subprocesses in the lo level of theory with reggeized gluons and reggeized quarks in the initial state in a self - consistent way together with unintegrated parton distribution functions proposed by kimber , martin and ryskin . to describe the non - perturbative transition of produced gluons and @xmath14-quarks into the @xmath0 mesons we use the universal fragmentation functions obtained from the fit of @xmath201 annihilation data from cern lep1 .
we found our results for @xmath0-meson and @xmath202-pair production to be in the good coincidence with experimental data from the lhcb collaboration .
the predictions for the @xmath203-pair production correlation spectra in the large rapidity region for the energy @xmath178 tev are also presented .
we have found that in case of single @xmath0-meson production at the lhc energies , both mechanisms , gluon and @xmath14-quark fragmentation , are important with some preference for the gluon fragmentation production . in the @xmath5-pair production we found @xmath14-quark production mechanism to be the main one , while in the @xmath6-pair production
the gluon fragmentation is a totally dominant source .
additionally , we arrive at the important conclusion that @xmath6-pair production can not be used for differentiation between single and double parton scattering approaches of high - energy qcd .
we are particularly indebted to r. maciula and a. szczurek for discussion of problems in @xmath6-pair production at the lhcb and b. kniehl and g. kramer for explaining some details of kkks08 fragmentation functions . the work of a.v .
shipilova and v.a .
saleev was supported in part by the russian foundation for basic research through the grant n 14 - 02 - 00021 and by the ministry of education and science of russia under competitiveness enhancement program of samara university for 2013 - 2020 ; the work of m.a .
nefedov and a.v .
karpishkov was supported by the ministry of education and science of russia through the contract no .
1394 . the work of a.v .
karpishkov was supported in part by the german academic exchange service ( daad ) and the ministry of science and education of the russian federation through michail lomonosov grant linie a , 2016 ( 57212599 ) .
v. n. gribov and l. n. lipatov , sov .
. phys . * 15 * , 438 ( 1972 ) [ yad . fiz .
* 15 * , 781 ( 1972 ) ] ; yu .
l. dokshitzer , sov .
jetp * 46 * , 641 ( 1977 ) [ zh . eksp .
fiz . * 73 * , 1216 ( 1977 ) ] ; g. altarelli and g. parisi , nucl . phys . *
b126 * , 298 ( 1977 ) .
e. a. kuraev , l. n. lipatov , and v. s. fadin , sov .
jetp * 44 * , 443 ( 1976 ) [ zh . eksp
. teor .
71 * , 840 ( 1976 ) ] ; i. i. balitsky and l. n. lipatov , sov . j. nucl . phys .
* 28 * , 822 ( 1978 ) [ yad .
* 28 * , 1597 ( 1978 ) ] . | the production of @xmath0 mesons in proton - proton collisions at the lhcb detector is studied .
we consider the single production of @xmath1 , @xmath2 , @xmath3 , and @xmath4 mesons and correlation spectra in the production of @xmath5 and @xmath6 pairs at the @xmath7 tev and @xmath8 tev . in case of the single @xmath0-meson production
we calculate differential cross sections over transverse momentum @xmath9 while in the pair @xmath10-meson production the cross sections are calculated over the azimuthal angle difference @xmath11 , rapidity difference @xmath12 , invariant mass of the pair @xmath13 and over the @xmath9 of the one meson from a pair .
the cross sections are obtained at the leading order of the parton reggeization approach using kimber - martin - ryskin unintegrated parton distribution functions in a proton . to describe the @xmath0-meson production we use universal scale - dependent @xmath14-quark and gluon fragmentation functions fitted to @xmath15 annihilation data from cern lep1 .
our predictions find a good agreement with the lhcb collaboration data within uncertainties and without free parameters . -2 |
quantum frequency translation ( qft ) is the process whereby two spectral modes ( or narrow bands of modes ) are swapped in the sense that the properties of their quantum states ( other than their central frequencies ) are interchanged , as in fig .
[ fig : mikefig ] @xcite .
such a process , of course , requires an external source or sink of energy , which is provided by a strong external optical field .
ideally , an arbitrary state of the narrow band of green modes will be swapped with the state of the narrow band of blue modes .
these could be coherent , squeezed , or number states , and may also be entangled with other , separate degrees of freedom , such as other modes ( e.g. , violet and yellow ) or states of atoms , quantum dots , etc . when the states of the green and blue modes are swapped , the prior entanglement should be preserved .
this paper focuses on the qft of single - photon states .
qft is defined to be noiseless , " that is , free of excess background noise .
for example , if both modes to be swapped are initially vacuum , they should remain vacuum@xmath0there are no spontaneous processes that can populate modes in this case .
the noiseless property distinguishes qft from other frequency conversion processes such as optical parametric amplification , which has been widely studied for use in classical telecom systems @xcite . in optical parametric amplification ,
power is spontaneously transferred from the pump(s ) to the weak signal fields , and without any blue input power , the signal fields can build up spontaneously .
the qft capability will play important roles in quantum information science ( qis ) , where states of a quantum memory may be transmitted to a spatially remote quantum memory by emitting and absorbing photons .
if the two memories operate using distinct frequencies of their absorption / emission lines , it is necessary to translate " the communication photons between these frequencies .
such a translation capability also allows sending the photons through low - loss optical fiber at wavelengths near 1500 nm , even in cases where the memories operate at other wavelengths .
two methods for all - optical qft have been demonstrated : three - wave mixing in a second - order nonlinear optical medium ( crystal ) @xcite and more recently four - wave mixing ( fwm ) in a third - order nonlinear optical medium ( silica fiber ) @xcite . in the case of three - wave mixing and referring to fig . [
fig : mikefig ] , the strong pump field must have frequency @xmath1 .
the configuration for fwm with two pumps , denoted @xmath2 and @xmath3 , is that of dynamic bragg scattering , in which for example , one photon is removed from pump @xmath2 and one is added to pump @xmath3 .
at the same time , one photon is removed from the green signal and one is added to the blue signal so that @xmath4 .
this contrasts with optical parametric amplification , wherein photons are removed from the pump(s ) and photons are added to both signals .
experiments have included qft of one - half of a number - correlated state @xcite and several examples of qft of single - photon states by three - wave mixing in crystals @xcite and by fwm in fiber @xcite .
the main advantage of the fwm method over the three - wave mixing method is that it allows qft between spectral bands separated by large or small shifts , whereas in practice three - wave mixing allows qft only between far - separated bands @xcite .
this paper treats the theory of fwm , but the same methods can easily be applied to three - wave mixing .
a potentially revolutionary application of qft is that it is predicted to enable the interference of two photons ( or other states of the modes ) with different colors @xcite .
this phenomenon generalizes the famous hong - ou - mandel ( hom ) interference effect , wherein two photons of like color and identical wave - packet structure impinge simultaneously on a standard beam splitter .
because of the boson nature of photons , the beam - splitter output modes both contain either zero or two photons , but never one photon in each .
this remarkable effect ( which is analogous to the exchange interaction for electrons ) enables a class of quantum information processing known as linear - optical quantum computing @xcite , including many variants @xcite .
the promise of qft is that it offers the possibility to perform linear - optical quantum computing over the rainbow . "
that is , one could use the light s color as a degree of freedom identifying optical qubits and perform linear - optical quantum computing using the generalization of the hom effect involving distinct frequencies .
this would entail creating entangled states involving many distinct frequencies ( perhaps by methods such as in @xcite ) , and manipulating such states using two - mode interference via qft .
this would require qft between modes nearby in frequency , and so fwm would be greatly preferred over three - wave mixing .
other tools proposed for linear - optical quantum computing over the rainbow " are temporal - mode filters ( pulse gates " @xcite which could be used as channel add / drop filters in an optical communications system @xcite ) and single - photon pulse shapers @xcite , based on qft by three - wave mixing . in this paper
we extend our earlier treatments @xcite to enable the explicit modeling of qft by fwm in optical fiber or by three - wave mixing in a crystal .
the treatment includes analytical and numerical components , and allows us to model realistically the effects of finite pulse durations , linear dispersion , finite phase - matching bandwidth , and complicating effects such as self - phase modulation ( spm ) and cross - phase modulation ( cpm ) @xcite .
previous theories of qft have either assumed single - mode fields @xcite , disallowing rigorous treatment of pulse propagation , or assumed that the signal pulse shapes are not altered significantly by the translation process @xcite , obscuring the accurate design of high - fidelity devices .
the main results of this study are the following : we find that for realistic fiber and laser pulse parameters , essentially 100 percent efficient qft can , in principle , be achieved .
we show that for given temporal shapes of the pump pulses , usually there are unique shapes that the two signal wave packets should have in order that they be swapped optimally by qft .
these shapes are found as the optimal modes of a singular - value ( schmidt ) decomposition of the green function for the qft process , and the shapes change significantly with changing translation efficiency .
this result provides a method for designing qft and two - photon interference with high efficiency and high fidelity .
we also show , for a given pump temporal shape , how to design signal pulse shapes to create perfect hom interference via the fwm process , as first predicted in @xcite .
the physical mechanism used to carry out frequency translation in fiber is the bragg scattering ( bs ) four - wave mixing process @xcite , pictured in fig . [
fig : bstoon ] . for convenience , we refer to the two weak signal fields as green " and blue , " although in principle their actual frequencies could correspond to visible or near - ir radiation .
two strong pump fields drive the process in which one photon from one of the pumps , pump @xmath2 in fig .
[ fig : bstoon ] , and one photon in the green field are annihilated while one photon from the other pump ( denoted @xmath3 ) and one photon in the blue field are created .
the process conserves photon number in accordance with the manley - rowe relations @xcite and can be summarized as @xmath5 , where @xmath6 represents a photon .
and @xmath7 and the signals have frequencies @xmath8 and @xmath9 , where @xmath10 .
the arrows symbolize either annihilation ( down ) or creation ( up ) of a photon in that mode .
the frequency @xmath11 is the average value of all the frequencies and @xmath12 is the frequency separation between the pump fields and also between the signal fields.,scaledwidth=40.0% ] before describing the realistic continuous - frequency case , we discuss a much simpler model in which each field is represented by a single - frequency mode .
such a model might apply to fields in a cavity , but strictly speaking not to the pulse propagation scenario in which we are interested .
the simple model has the advantage of permitting analytical solutions , thereby providing some insight into the qft process @xcite . for cases where
each field can be described by a single - frequency mode , the hamiltonian governing the process is given by @xmath13 where the creation / annihilation operators satisfy the usual commutation relations , and @xmath14 and @xmath15 are parameters quantifying the effects of phase - mismatch and nonlinearity @xcite .
the evolution is given by the spatial equations - of - motion ( with a minus sign compared to the usual time - domain heisenberg equation ) @xmath16.\ ] ] the solution of for @xmath17 as a function of position along the fiber @xmath18 for a fiber of length @xmath19 is given by @xmath20 where the transfer " functions are given by @xmath21 where @xmath22 .
the transfer functions have the relation @xmath23 , which ensures the process is unitary and conserves photon number .
mathematically , the bs translation process having frequency input and output ports " is identical to the normal beam - splitter operation on degenerate monochromatic fields @xcite . for perfect translation from green to blue , and vice versa , it must be that @xmath24 for a medium of length @xmath19 . in this case
the quantum states of the modes are completely swapped .
for example , a single green photon is replaced by a single blue photon , that is , @xmath25 , where @xmath26 denotes the state in which @xmath27 photons are in the green region and @xmath28 photons are in the blue region .
since the bs process is analogous to beamsplitter action , all of the effects that a beamsplitter can induce on a field will also occur in the bs process .
in particular , the effect analogous to two - photon hom interference will occur , where two independent photons of _ differing frequencies _ will interfere with one another and give a perfect hom dip under the appropriate circumstances @xcite .
just as in the monochromatic hom interference effect , the frequency beamsplitter " must have equal transmission and reflection coefficients , that is @xmath29 , so that it is equally likely for the blue ( green ) photon to stay in the blue ( green ) mode or be translated to the green ( blue ) mode . using the above equations it can be shown that in this case , for an input state consisting of one green photon and one blue photon , denoted by @xmath30 , the output state is given by @xmath31 for equal values of @xmath32 and @xmath33 the @xmath34 component of the state cancels to zero .
the above analysis is valid only if all the fields involved are single - mode . for fields that contain multiple - frequency components ,
as any pulsed field does , the theory becomes much more complex .
for arbitrary input , and even simple " multi - mode input such as gaussian pulses , the output can not in general be determined analytically .
however , it is possible to assign analytically a green function @xmath35 in either frequency @xmath36 or time @xmath37 that relates the input wavepacket to the output wavepacket for arbitrary input , analogous to and , for given pump spectra , fiber dispersion , and fiber length @xcite . denoting the creation operator at position @xmath18 by @xmath38 , the input @xmath39 , and output @xmath40 are related by the _ forward _ transformation @xmath41 for our purposes ,
the _ backward _ transformation is more useful : @xmath42 which gives the input operator in terms of an integral over the output operator .
a single - photon input wavepacket given by @xmath43 , where @xmath44 is the spectrum of the input wavepacket , is translated in the manner given by .
the state can thus be expressed in terms of the output operator as @xmath45 the green function is unitary and therefore satisfies the condition @xmath46 to facilitate an intuitive understanding of the transfer function , we break up frequency space into two pieces ; the range that contains the green and the range that contains the blue .
each operator or variable will be denoted by a @xmath47 " or @xmath48 " subscript .
for example , the creation operator for an green photon at a particular green frequency will be written as @xmath49 . in this case
the transfer function can be written in block form as @xcite @xmath50\ ] ] where @xmath51 refers to the evolution of an output creation operator in range @xmath52 which came from range @xmath53 .
although the @xmath35 function completely describes the evolution of arbitrary inputs , it is difficult to discover which input pulses give rise to desirable effects such as high - efficiency translation or good two - color hom interference just by examining the @xmath35 function .
however , by performing a schmidt ( singular - value ) decomposition ( svd ) @xcite on the @xmath35 function or on the sub - matrices @xmath51 , the input wavefunctions necessary to achieve these effects can be found @xcite .
for example , consider the translation of an input green field to the blue field , which is described by the green sub - matrix by @xmath54 , which is not hermitian and therefore does not admit a unitary decomposition .
we write @xmath54 in its svd form as @xmath55 where the @xmath56 are the real , positive singular values ( the generalization of eigenvalues ) , and @xmath57 and @xmath58 are the @xmath59th column vectors of the unitary matrices " @xmath60 and @xmath61 describing the decomposition , respectively . the most useful way to interpret this decomposition is that a given green wavepacket @xmath57 will be translated to the blue wavepacket @xmath58 with probability @xmath62 . hence
, only input green wavepackets @xmath57 that have associated singular values @xmath63 will translate over to the blue mode with 100 percent probability .
the @xmath64 matrix " describes an input green wavepacket scattering within the green spectral region at the output ( no translation from green to blue , although the shape of the wavepacket may change ) .
if the svd of @xmath64 is given by @xcite @xmath65 then @xmath66 is the probability that an input green wavepacket @xmath57 will scatter to an output green wavepacket @xmath67 ( the @xmath60 matrix here being the same as in the @xmath54 decomposition ) .
the singular values satisfy the relation @xmath68 , as they must to conserve photon number .
similarly , the @xmath69 and @xmath70 have their own svds , given by @xmath71 where the @xmath61 and @xmath72 matrices , and the singular values @xmath73 and @xmath74 , are the same as in and .
this highlights the interconnectedness between the various sub - green functions in that there are only four unique schmidt modes for each index @xmath59 that completely describe the process . before treating exclusively single - photon inputs ,
we emphasize that the green functions determined by our present approach can be used to treat arbitrary multi - photon inputs as well .
the svd technique can be used to discover what input wavepackets are suitable for optimal two - photon hom interference @xcite .
the condition is somewhat intuitive . since in the single - mode ,
two - color case it is necessary for both green and blue inputs to have a 50 percent probability of translating to the other mode for perfect hom interference , it is reasonable to guess this condition is also true in the multi - mode two - color case . hence for the cases in which @xmath75 , the corresponding input vectors @xmath57 and @xmath76 will show perfect two - color hom interference . alternatively
, this condition can be written as @xmath77 we refer to the quantity @xmath78 as the hom singular value , because it appears as such for a schmidt decomposition kernel designed specifically to optimize the hom interference .
( 37 ) of @xcite . )
the closer it is to unity , the better the hom interference .
there is an alternative , complimentary , method to calculate the degree of hom interference that a bs process will exhibit , for which the green function does not have to be calculated .
first , consider the case in which there are two photons , either of which could be in the green or blue region , or in some superposition of both regions .
in general , the state of this bi - photon can be written as @xmath79 must be normalized and the component states are orthonormal .
now , for input consisting of one green photon and one blue photon having input spectra of @xmath80 and @xmath81 respectively , the state is @xmath82 if the photons undergo the bs process , the input photon operators evolve according to . by switching the limits of integration and noting that @xmath83 \rightarrow a_{xy}(l,\omega_{y})$ ] , where both @xmath84 and @xmath85 could be either @xmath47 or @xmath48 and @xmath86 is a component of the spectrum at the output of the fiber , the state can be written compactly as @xmath87 \\ & \times \big [ \int d\omega_{g}^ { ' } a_{g}^{\dagger}(l,\omega_{g}^ { ' } ) a_{gb}(l,\omega_{g}^ { ' } ) + \int d\omega_{b}^ { ' } a_{b}^{\dagger}(l,\omega_{b}^ { ' } ) a_{bb}(l,\omega_{b}^ { ' } ) \big]|vac \rangle . \end{split}\ ] ] the probability , given a particular output state , for the various sub - space states to occur is given by the modulus squared of the coefficients of these states , @xmath88 . for the @xmath34 state
the coefficient is found by acting the aforementioned operators , with their accompanying functions , on the vacuum , yielding @xmath89 |1_{\omega_{g } } , 1_{\omega_{b}}\rangle\ ] ] the first term , @xmath90 , corresponds to the case where a photon initially in the green mode ( i.e. wavepacket ) scatters within the green mode at the output , and a photon initially in the blue mode scatters within the blue mode at the output .
the second term , @xmath91 , corresponds to the case where a photon initially in the green mode scatters to the blue mode at the output , and a photon initially in the blue mode scatters to the green mode at the output . since these terms are being added before their modulus squared
is taken , there is the chance that they could cancel if the terms are proportional and out of phase with one another .
this is exactly what happens in the case in which ideal hom interference occurs , leaving zero probability amplitude for the field to emerge in the @xmath34 state .
the expression for the probability for this state to occur @xmath92 is @xmath93 \big ] .
\end{split}\ ] ] this gives an expression to calculate the coincidence probability for any given input pulses without knowing the green function for the process . in the alternative case in which we launch particular schmidt modes as input signals
, we showed previously that the coincidence count rate equals @xmath94 @xcite .
therefore , in this case we have a simple relation between @xmath92 and the hom singular values ; @xmath95 .
other important quantities are @xmath96 and @xmath97 , the probabilities that two photons appear in one or the other output ports together .
the expressions for these , and their derivation can be found in appendix a.
unfortunately , it is not possible to discover the green function and its schmidt modes by examining the hamiltonian directly .
one must solve for the full propagation first , then examine the input - output relations .
this emphasizes that the schmidt modes are global properties , depending on the entire evolution .
for example , the input schmidt modes depend on the length of the medium .
the appropriate hamiltonian generates the well - known nonlinear schrodinger equation , which needs to be solved . in most realistic circumstances , taking into account actual field shapes and real fiber dispersion properties , it is not possible to solve the nonlinear schr@xmath98dinger equation for the fields analytically .
so we turn to numerical computation to determine the evolution of a completely specified input state to a completely determined output state .
the key difficulty with such an approach is that , in general , nonlinear operator equations can not be simulated by any simple numerical scheme . when the evolution is both in space and time as in our case , the problem becomes one suited for techniques such as the positive - p representation @xcite .
fortunately , in our case , although the evolution is nonlinear in the pump fields , it is linear in the weak signal fields .
this allows a simplification if we treat the pump fields as classically evolving parameters in the quantum heisenberg equations of motion for the weak signals .
then the field operators for the weak signals can be solved in terms of green functions that obey the classical evolution equations .
therefore , we solve numerically for the classical evolution of the green functions and use those solutions to calculate any quantum probability or correlation function concerning the signal fields .
an analogous procedure was used in @xcite . the usual starting point for solving for the nonlinear evolution of a pulse
is to write the total field as @xmath99 , where @xmath100 is a slowly varying envelope function , @xmath101 is the high frequency of the carrier wave , and @xmath102 is the propagation constant at @xmath101 . in that form
the evolution is governed by the nonlinear schrdinger equation @xmath103 although this is effective for single pulses , for electric fields made up of multiple , distinct pulses with significantly different frequency components , solving this form of the equation is unnecessarily computationally intensive and impractical .
instead , we solve four coupled equations , derived from the nonlinear schrdinger equation , in which terms relating to dispersion , self - phase modulation , cross - phase , and the bs process were kept . numerically , the frequency components of each field are associated with a frequency mesh pertaining only to that field .
this drastically reduces the number of calculations needed to compute the solution in that it is unnecessary to resolve the large empty bandwidth in between fields . in this scheme ,
the total electric field is the sum of the four individual fields , each having a slowly varying amplitude function @xmath104 centered around a carrier wave of frequency @xmath105 ; @xmath106 where @xmath28 denotes @xmath2 , @xmath3 , @xmath47 or @xmath48 .
in quantum theory , @xmath104 is proportional to an annihilation operator . substituting this ansatz into the nonlinear schr@xmath98dinger equation leads to , with suppressed independent variables @xmath18 and @xmath37 , the four coupled bs equations ( where @xmath107 represents the partial derivative with respect to @xmath18 ) @xmath108\\ \partial_z a_q & = i\beta_{p}(i\partial_t ; \omega_q)a_q + \left(i\gamma |a_q|^2 + 2i\gamma \sum_{k \neq q } |a_k|^2 \right ) a_q + 2i\gamma a_b^ * a_p a_g \mathrm{exp}[i \delta \beta z]\\ \partial_z a_g & = i\beta_{p}(i\partial_t ; \omega_g)a_g + \left(i\gamma |a_g|^2 + 2i\gamma \sum_{k \neq g } |a_k|^2 \right ) a_g + 2i\gamma a_p^ * a_q a_b\mathrm{exp}[-i \delta \beta z]\\ \partial_z a_b & = i\beta_{p}(i\partial_t ; \omega_b)a_b + \left(i\gamma |a_b|^2 + 2i\gamma \sum_{k \neq b } |a_k|^2 \right ) a_b + 2i\gamma a_q^ * a_p a_g \mathrm{exp}[i \delta \beta z ] , \end{split}\ ] ] where @xmath14 is the dispersive mismatch given by @xmath109 , @xmath110 and @xmath111 are the peak pump powers , and @xmath6 is the nonlinear coefficient .
the first term on the right hand side describes the effect of linear dispersion , the next two terms describe spm and cpm respectively , and the last term describes the bs four - wave mixing process .
the quantity @xmath112 is given by @xmath113 where @xmath114 is the @xmath115 derivative of @xmath116 evaluated at @xmath117 , and @xmath118 . in this scheme all the fields are propagating in the arbitrary choice of the frame of reference of pump @xmath2 . with one frame of reference
there need only be one universal time mesh .
although is for quantum field operators , we can realistically replace the pump fields @xmath119 and @xmath120 by classical functions if these fields are in strong coherent states .
we solve the first two equations for the pumps ignoring the small effects of the weak signal fields @xmath121 and @xmath122 , but retaining the spm and cpm effects of the strong pumps .
we then use these pump solutions in the second pair of equations for the weak signal fields .
because these equations are linear in the operators , we can treat them formally as classical equations . then the quantum effects can be fully accounted for by using the commutation relations when calculating quantities such as @xmath92 , as describe above . to solve the coupled equations numerically we use the split - step fourier method ( ssfm ) @xcite . for each electric field
@xmath123 we approximately solve the operator equation @xmath124 , where @xmath125 and @xmath126 are the dispersive and nonlinear differential operators , acting in the frequency and time domains respectively , which act on @xmath123 given in eq . .
the approximation is that @xmath127 is taken to @xmath128 by @xmath129 where @xmath130 and @xmath131 is the number of steps .
the nonlinear component integral is found by use of a fourth - order fixed - step runge - kutta method .
the ssfm allows one to calculate a particular output amplitude for a particular arbitrary input amplitude .
however , discovering the green function that governs the process requires solving the propagation equation for a family of orthogonal input amplitudes .
let @xmath132 be an @xmath133 matrix with each column equal to one time - domain vector from an orthogonal set , @xmath134 be an @xmath133 matrix with each column equal to the output of the @xmath135 column of @xmath132 , and @xmath35 be the green function .
in this paper we use the hermite - gaussian ( hg ) functions as the orthogonal set , where the @xmath135 function is given by @xmath136 where the characteristic time of the basis set is set by the overall time window of the mesh . due to the linear nature of the green function formalism , as expressed in , the _ forward _ green function expressed in matrix notation is @xmath137 this is strictly true only if all of the output vectors of @xmath134 can be described as some linear combination of the orthogonal vectors of @xmath132 . in practice , an orthogonal vector of
order @xmath28 will evolve to have components of both lower and higher order than @xmath28 . in the cases where the @xmath28th mode evolves to include components of higher order than @xmath59
, @xmath35 wo nt satisfy the unitary conditions of , which is often the case for the evolution of the very - high - order modes . in that case
, only the subset of orthogonal input vectors that evolve into output that can be described by all @xmath59 input vectors can be considered
valid " output , as far as determining the green function is concerned .
let @xmath138 be a @xmath139 dimensional matrix that is the sub - matrix of @xmath132 where all the input vectors evolve to output vectors ( described by the @xmath140 matrix @xmath141 , where @xmath142 ) that can be described as a linear combination of the vectors of @xmath132
. then the green function that always satisfies the unitary conditions is given by @xmath143 and will be a matrix of dimensions @xmath140 . in practice ,
the simplest way in which to determine the correct matrix subsets is by applying the unitary conditions ( calculating @xmath144 ) and seeing for what subset of vectors they hold .
when that is determined , @xmath145 can then be calculated .
finally , there is the question of fiber modeling .
as these calculations are meant to help explain past experimental results @xcite and guide future experiments using solid - core photonic crystal fiber @xcite , we desire a fiber model that fairly accurately describes photonic crystal fibers . we employ the often used step - index model in describing photonic crystal fibers @xcite , which approximates the photonic crystal fiber as a step - index fiber having a core index @xmath146 ( usually the index of bulk fused silica ) and an effective cladding index @xmath147 , where @xmath148 . here
@xmath149 is the air - filling fraction of the lattice region compromising the cladding , which along with the core radius @xmath150 , parameterizes the fiber . by measuring the modulation
instability wavelength signals produced by a fiber as a function of pump wavelength , best - fit values for @xmath149 and @xmath150 can be determined .
for a particular fiber that has been used in past experiments and is a good candidate for future experiments , this procedure has been carried out , resulting in parameter values of @xmath151 and @xmath152 @xmath153 m , with the resulting dispersion parameter @xmath154 shown in fig .
[ fig : fibdis ] .
this is the dispersion profile used for all numerical calculations in this paper .
parameter for the photonic crystal fiber used in simulations from 600 nm to 1600 nm .
the dashed line is at @xmath155 for reference.,scaledwidth=50.0% ] given a particular dispersion profile of a fiber and the central wavelengths at which translation will occur , the fiber can effectively translate input signals only with spectral widths smaller than the characteristic bandwidth of the bs process . as seen in equations in the case of low conversion efficiency ,
the state is proportional to sinc@xmath156 , which defines the dispersive mismatch of the process .
figure [ fig : fiberpm ] shows this sinc function for the fiber , central wavelengths , and length ( 20 m ) used in the simulations .
all fields were taken to be monochromatic , and the pumps where fixed at 808 nm and 845 nm .
the blue " signal field was varied from its central wavelength of approximately 649 nm ( shown in frequency on the horizontal axis ) while the green " field was varied from its central wavelength of 673 nm to conserve the energy of the process @xmath157 .
the full - width at half - maximum ( fwhm ) of the central sinc lobe is approximately 0.3 trad / s .
this number gives a rough estimate of the effective bandwidth of the bs process .
as we will see , all else being equal , input signals that have spectral widths larger than this number do not translate as efficiently as signals that have spectral widths lower than this number . for the fiber used in the simulation , where the pump wavelengths were fixed and the green " and blue " fields frequencies were varied in an energy - conserving manner .
the horizontal axis is the green " frequency variation from its corresponding central wavelength value of 673 nm .
for this plot the fiber length is @xmath158 m , and the fwhm of the function is approximately 0.3 trad / s ( @xmath159 rad / s).,width=268 ]
this section details the results of simulations for several different cases , broken up into two main categories ; cases that highlight properties relating to effective total translation from one signal field to another , and cases that highlight properties relating to good two - color hom interference . for the cases in which the green function of the process was calculated , the two pumps were identical transform - limited gaussian - shaped fields .
the pump _ intensity _ profiles had either a long " fwhm duration of 1000 ps , making them quasi - cw for most input signal fields used , or a short " fwhm duration of 70 ps . at the input to the fiber , the centers of the fields were placed at zero time .
the central wavelengths for the pump fields @xmath2 and @xmath3 were 808 nm and 845 nm , respectively , placing them in the anomalous - dispersion region .
the signals green " @xmath47 and blue " @xmath48 were at wavelengths 673 nm and 649 nm , respectively , placing them in the normal - dispersion region .
the 845-nm and 649-nm fields had nearly equal group velocities ( @xmath160 m / s ) , and the 808-nm and 673-nm fields had nearly equal group velocities ( @xmath161 m / s ) . the nonlinear coefficient @xmath6 was set to @xmath162 for all calculations .
the green functions were calculated in the hg mode basis , where the input and output field amplitudes were decomposed into a common set of hg functions .
the green function for a particular case was found by sequentially setting the green or blue input field to a single hermite - gaussian function , solving the propagation , and projecting the output pulse onto the hg basis functions .
thereby the green function is represented in the hg mode basis .
the characteristic time scale for the hg functions , which we define as the fwhm of the zeroth order hg function , used to produce the figures was found by first arbitrarily choosing a characteristic time of 40 ps , fitting the resulting absolute value of the first schmidt mode to a gaussian function , and calculating the green function again using this new characteristic time .
this time - scale choice for defining an optimal hg basis " produced green function matrices that were the least complex , some of which were nearly diagonal , and therefore the simplest to understand .
the hg basis set was centered in time at the peak of the pump and signal inputs .
parameter studies were also carried out in which the two pump pulses intensity durations were set at 1000 and 70 ps , and the input signal field durations were varied .
either the translation efficiency or hom interference properties were documented , depending on the main category of the case .
there are a few ways in which to gauge the accuracy of the numerical solutions .
while not explicitly built into the calculations , for all calculations the sum of the green and blue pulse energies ( calculated in the frequency domain ) were found to be conserved , in accordance with the manley - rowe relation @xcite for the bs process . also , there are unitary conditions on the green functions that govern the process , given by . although checking these conditions is significantly more expensive computationally than verifying the manley - rowe relations , and therefore is not feasible for studies requiring large numbers of cases like the below - mentioned parameter studies , this check was carried out for several representative cases for such studies .
the green functions were then checked against , and in these cases were found to agree well .
finally , the appropriate number of steps were taken for each calculation so that the solutions converged and were not meaningfully dependent on the exact number of steps .
figures [ fig:285.1 - 3 ] , [ fig:285.4 ] , and [ fig:285.5 - 8 ] detail the case where the pump pulses were 1000 ps long and had peak powers of 400 mw .
it was found by trial and error that translation occurred efficiently at a fiber length of 20 m for pumps with peak powers around 400 mw .
it was found that the characteristic time of the optimal hg basis was @xmath163 ps .
figure [ fig:285.1 - 3](a ) shows the absolute value of the @xmath60 matrix , as defined in , from an svd of the @xmath54 green function .
this green function corresponds to the case of translation from the green mode to the blue mode . as discussed previously ,
the columns of the @xmath60 matrix represent the input schmidt modes , so in this case the @xmath60 matrix represents the input signal field .
that is , @xmath164 the horizontal axis is the schmidt mode number @xmath59 , where order was determined by the magnitude of the corresponding singular value @xmath74 , largest to smallest , for the first 25 schmidt modes .
the vertical axis is the hg mode number @xmath28 of the schmidt modes using the same hg basis used to find the green function .
the plotted quantity is @xmath165 .
for example , the first schmidt mode of the @xmath60 matrix is mostly composed of the first hg function and almost completely described by the first four hg functions .
thus we can conclude that the temporal duration of the first schmidt mode is close to the characteristic time of the optimal hg basis , 243 ps .
matrix ( columns of which are coefficients in the hg basis for individual schmidt modes ) , and @xmath61 matrix of @xmath54 green function relating to the case with long pumps ( 1000 ps ) having peak powers of 400 mw and a fiber length of 20 m. ( a ) absolute value of the @xmath60 matrix ( input schmidt modes ) .
( b ) absolute value of the @xmath61 matrix ( output schmidt modes ) .
( c ) squared singular values @xmath166 ( red ) and @xmath167 ( blue ) indexed by the corresponding schmidt mode number.,scaledwidth=80.0% ] the output schmidt modes , which physically are the results of propagating each corresponding input schmidt mode through the qft process in the fiber , are described by the @xmath61 matrix .
the absolute value of this matrix , the column vectors of which correspond to output in the blue mode , is shown in fig .
[ fig:285.1 - 3](b ) .
the axes are the same as those in part ( a ) .
interestingly , the @xmath60 and @xmath61 matrices are qualitatively similar , although the @xmath61 matrix schmidt modes tend to be dominated by either even or odd hg functions whereas the @xmath60 matrix schmidt modes are composed of a more or less equal number of even and odd hg functions . the black circles of the red line in fig .
[ fig:285.1 - 3](c ) are the squared singular values @xmath166 of the @xmath54 green function for this case .
the circles of the blue line correspond to the squared singular values @xmath167 of the @xmath64 green function .
the @xmath166 describe the translation efficiencies of the input green schmidt modes to the output blue schmidt modes , whereas the @xmath167 describe the non - translation " efficiencies of the input green schmidt modes to the output green schmidt modes .
this corresponds to the transmission and possible reshaping of the green mode . for the manley - rowe relations to be satisfied
, the square of these singular values must add to one , which they do numerically in this case ( as well as in all other cases ) to one part in one thousand .
as is evident from the figure , many of the modes translate efficiently with the first several having efficiencies of over 90 percent .
we call this configuration non - discriminatory " in the sense that it effectively translates many different dissimilar modes efficiently .
figure [ fig:285.4 ] shows the translation efficiency as a function of length long the fiber for the first three green input schmidt modes . here
translation efficiency is defined as the number of photons in the blue mode divided by the initial number of photons in the green mode .
note that there were initially no photons in the blue mode .
the red , green , and blue lines correspond to the first , second , and third schmidt modes , respectively .
the efficiency for these modes is very sinusoidal , almost as if the signals were monochromatic and described by and .
figure [ fig:285.5 - 8 ] shows the absolute amplitudes and phases in both time and frequency of the first three input schmidt modes ( relating to the @xmath60 matrix ) of the green ( 673 nm ) mode .
the input amplitudes of the pump fields are also shown as a black dashed line . as could be guessed from fig .
[ fig:285.1 - 3](a ) , the modes look very much like hg functions , although there is some asymmetry and delay in both time and frequency .
these effects are likely the results of pump evolution , which includes convection , dispersion and cpm .
figure [ fig:285.5 - 8](a ) shows that first few schmidt modes are much narrower in time than the pump fields . looking at fig .
[ fig:285.5 - 8](b ) , which shows the phases in time , all the input schmidt modes , disregarding the @xmath168 phase jumps , are distinctly parabolic with positive curvature . for the phase convention we use , this shape implies the schmidt modes at the fiber s input are down - chirped ( frequency decreases in time ) , meaning that when they propagate through the medium , which for the signals is normally - dispersive , they will temporally compress .
that is , they are dispersion pre - compensated .
cpm also chirps the signals in the same way as linear dispersion .
figure [ fig:285.5 - 8](c ) shows that the pump fields are nearly monochromatic as compared to the schmidt modes , although all modes are well within the translation phase - matching bandwidth of the fiber .
rad / s.,scaledwidth=90.0% ] figures [ fig:51.1 - 3 ] , [ fig:51.4 ] , and [ fig:51.5 - 8 ] detail the case where the pump pulses were 70 ps long and had peak powers of 400 mw .
the goal in studying this case in relation to the former case is to highlight how using temporally shorter and spectrally broader pump pulses impacts the translation properties of the process .
it was found that the characteristic time of the optimal hg basis for this case was @xmath169 ps .
figure [ fig:51.1 - 3](a ) shows the absolute value of the @xmath60 matrix ( input green schmidt modes ) from the svd of the @xmath54 green function , while fig .
[ fig:51.1 - 3](b ) shows the absolute value of the @xmath61 matrix ( output schmidt modes ) .
the horizontal and vertical axes have the same meaning as they did in the previous case .
matrix , and @xmath61 matrix of @xmath54 green function relating to the case with short pumps ( 70 ps ) having peak powers of 400 mw .
( a ) absolute value of the @xmath60 matrix .
( b ) absolute value of the @xmath61 matrix .
( c ) squared singular values @xmath166 ( red ) and @xmath167 ( blue ) index by the corresponding schmidt mode number.,scaledwidth=80.0% ] while the absolute values of the @xmath60 and @xmath61 matrices are qualitatively similar , they are significantly different from their counterparts of the previous case .
while they start out for low mode order approximately diagonal , like in the previous case , the schmidt modes quickly become seemingly random and unordered .
this is likely due to these schmidt modes singular values [ the squared value of the singular values are shown on the red line of fig .
[ fig:51.1 - 3](c ) ] , which quickly drop to almost zero , signifying that almost no translation occurred .
since there is an infinite number of schmidt modes with singular values near zero ( practically any random combination of hg modes generally leads to a low singular value ) , even schmidt modes of relatively low schmidt mode number that have singular values close to zero can look random ( non - uniquely defined ) in the hg basis . since in this case
only the first few modes translate with significant efficiency , we call this configuration discriminatory " in the sense that only a few closely related modes effectively translate
. figure [ fig:51.4 ] shows the translation efficiency as a function of length along the fiber for the first three input green schmidt modes in the short - pumps case .
the red , green , and blue lines correspond to the first , second , and third schmidt modes , respectively .
unlike the former long - pumps case , the efficiency curves of the modes are drastically different , relating to the fact that their singular values are quite different .
figure [ fig:51.5 - 8 ] shows the absolute amplitudes and phase in both time and frequency of the first three green 673 nm input schmidt modes ( relating to the @xmath60 matrix ) .
the input amplitudes of the pump fields are also shown as a black dashed line .
qualitatively , the amplitudes for this case are similar to the amplitudes in the former case ; although asymmetric , they are roughly hg - like .
quantitatively , these amplitudes are much broader spectrally than in the former case .
the phases in time are essentially parabolas with positive curvature much larger than that of the previous case , as expected because the bandwidths are much larger .
figure [ fig : eff.param ] shows a parameter study where the pump pulses are identical gaussians with temporal fwhm durations of 1000 ps and peak powers of 400 mw . in fig .
[ fig : eff.param ] ( a ) , the maximal translation efficiency achieved along the fiber is plotted as a function of input signal pulse fwhm duration .
for relatively long " pulses in low hundreds of picoseconds , the translation efficiency is quite high due to the near - cw nature of the pump and the width of the translation bandwidth .
as the pulses become shorter , particularly around 43 ps , which corresponds to a spectral bandwidth of @xmath170 grad / s , the pulse spectral width begins to match that of the translation phase - matching bandwidth .
therefore , this dramatic drop in efficiency is likely caused by the signal spectral bandwidth being too large for the bs process to accommodate . in fig .
[ fig : eff.param](b ) , the length at which the maximal translation occurs is plotted as a function of input signal duration .
the curve has the same qualitative shape as the curve in part ( a ) ; for short signal durations , the length at which maximal translation occurs is shorter than that of signals with long durations .
this likely is due to the fact that short pulses have large spectral bandwidths , which because of cpm from the pumps , become even more spectrally broad as they evolve , making them in harder to translate .
hence the best translation occurs for these signals when they havent evolved much .
signals with long duration and small spectral width do not spread out in frequency as much as the short signals do , so they are less limited by the magnitude of the translation bandwidth and achieve good translation along the entire fiber .
figure [ fig : eff.ps70 ] show a parameter study where the pump pulses are identical gaussians with temporal fwhm durations of 70 ps and peak powers of 400 mw .
the efficiency , shown in part ( a ) , for short signal inputs behaves similarly to the previous long - pumps cases ; around 20 ps the efficiency drops sharply due to the signal having larger translation phase - matching bandwidth than the fiber . unlike the previous case , the efficiency reaches a maximum around 35 ps and slowly declines for wider input signal widths .
relatedly , it is at about this signal width that the length at which the maximal efficiency occurs , shown in part ( b ) , is equal to the length of the fiber , suggesting that the translation efficiency would be higher for a longer fiber or higher fiber for these signal widths . as the signal width becomes equal and then greater to the pumps widths
there are two main elements affecting the efficiency ; the lack of power / fiber length and the fact that the pump pulses no longer fully encompass the signal pulse in time . both of these lead to lower translation efficiency .
but it is important to note that for some signal width range , from about 25 ps to 60 ps , reasonably good translation efficiency will occur for short pump pulses with reasonable parameters such as peak pump powers of 400 mw and a fiber length of 20 m. figures [ fig:171.1 - 3 ] , [ fig:171.4 - 5 ] , and [ fig:171.6 - 9 ] detail the case where the pump pulses were 1000 ps long and had peak powers of 200 mw .
this configuration was designed to exhibit good two - color hom interference by lowering the pumps peak power so as to lessen the translation efficiency , since the hom interference effect is optimal when 50 percent , not 100 percent , translation occurs .
it was found that the characteristic time of the optimal hg basis was @xmath171 ps .
figure [ fig:171.1 - 3](a ) shows the absolute value of the @xmath60 matrix , as defined in , from an svd of the @xmath54 green function .
this green function corresponds to the case of translation from the green mode to the blue mode .
figure [ fig:171.1 - 3](b ) shows the absolute value of the @xmath61 matrix .
the horizontal and vertical axes are the same as they were the previous cases .
surprisingly , both the @xmath60 and @xmath61 matrices are nearly diagonal , meaning that the input schmidt mode set and the hg basis functions are nearly identical . in appendix
b we present a theoretical model explaining this result .
the long pump pulse of 1000 ps enables many modes to be translated with nearly similar efficiencies , as can be seen from [ fig:171.1 - 3](c ) . for this reason we describe this case as non - discriminatory " with respect to its two - color hom interference properties , since many different types of modes lead to good hom interference .
matrix , and @xmath61 matrix of @xmath54 green function relating to the case with long pumps ( 1000 ps ) having peak powers of 200 mw optimized for good hom interference .
( a ) absolute value of the @xmath60 matrix .
( b ) absolute value of the @xmath61 matrix .
( c ) squared singular values @xmath166 ( red ) and @xmath167 ( blue ) index by the corresponding schmidt mode number.,scaledwidth=80.0% ] figure [ fig:171.4 - 5 ] shows the hom singular values , as defined by , as a function of schmidt mode number , and the value of @xmath92 for the first three input schmidt modes as a function of fiber length .
we emphasize that a value @xmath92=0 means that two green or two blue photons exit the fiber , but never one green and one blue .
this is the hom interference effect as it applies to interference of photons of different colors @xcite .
figure [ fig:171.4 - 5](a ) shows clearly why this case is non - discriminatory in regards to its hom interference properties ; many orders of schmidt modes exhibit hom singular values near one . and @xmath92 values relating to the case with long pumps ( 1000 ps ) having peak powers of 200 mw .
( a ) hom singular value index by schmidt mode number .
( b ) two - photon coincidence count probability @xmath92 as a function of length for the first three input schmidt modes.the red , green , and blue lines relate to the first , second , and third schmidt modes , respectively.,scaledwidth=90.0% ] figure [ fig:171.6 - 9 ] shows the absolute amplitudes and phase in both time and frequency of the first three green input schmidt modes ( relating to the @xmath60 matrix ) .
the input amplitudes of the pump fields are also shown as a black dashed line .
since the @xmath60 matrix is nearly diagonal , the schmidt mode amplitudes are almost quantitatively identical to the hg basis functions , which stands in contrast to the case in section [ sec : longphighp ] which had the same duration pulse length but twice the pump power .
similarly to the previous cases , the phase in time is parabolic with positive curvature , showing that these inputs are also down - chirped .
matrix , and @xmath61 matrix of @xmath54 green function relating to the case with short pumps ( 70 ps ) having peak powers of 200 mw , optimized for good hom interference .
( a ) absolute value of the @xmath60 matrix .
( b ) absolute value of the @xmath61 matrix .
( c ) squared singular values @xmath166 ( red ) and @xmath167 ( blue ) indexed by the corresponding schmidt mode number.,scaledwidth=80.0% ] figures [ fig:45.1 - 3 ] , [ fig:45.4 - 5 ] , and [ fig:45.6 - 9 ] detail the case where the pump pulses were 70 ps in duration and had peak powers of 200 mw , which is close to the optimum for good hom interference .
the goal in studying this case is to contrast it with the former case that had pumps with the same peak power but were much longer in duration .
this will highlight the relative importance of having spectrally broad and temporally short pump pulses in regards to the hom interference behavior .
it was found that the optimal characteristic time of the hg basis for this case was @xmath172 ps .
figure [ fig:45.1 - 3](a ) shows the absolute value of the @xmath60 matrix from the svd of the @xmath54 green function , while fig .
[ fig:45.1 - 3](b ) shows the absolute value of the @xmath61 matrix .
the horizontal and vertical axes are the same as they were in the previous cases .
relating to the case with short pumps ( 70 ps ) having peak powers of 200 mw .
( a ) hom singular values indexed by schmidt mode number .
( b ) two - photon coincidence count probability @xmath92 values as a function of length along the fiber for the first three input schmidt modes .
the red , green , and blue lines relate to the first , second , and third schmidt modes , respectively.,scaledwidth=90.0% ] value achieved in the fiber as a function of signal pulse duration .
( b ) the length at which the minimal @xmath92 value is achieved as a function of signal pulse duration.,scaledwidth=90.0% ] value achieved in the fiber as a function of signal pulse duration .
( b ) the length at which the minimal @xmath92 value is achieved as a function of signal pulse duration.,scaledwidth=90.0% ] like the previous cases , both the @xmath61 and @xmath60 matrices are nearly diagonal for low schmidt mode number , but unlike the previous case , they quickly become seemingly random combinations of hg functions for higher - order schmidt modes .
the reason for this is the same as it was in the other case having short pump pulses ; there are many schmidt modes that have low singular values , and practically any combination of higher - order hg functions will have a low , but non - zero singular value . as shown in fig .
[ fig:45.1 - 3](c ) , only the first few schmidt modes have a singular value meaningfully above zero and will therefore have a more ordered mode structure .
figure [ fig:45.4 - 5 ] plots the hom singular values , as defined by , as a function of schmidt mode and the value of the two - photon coincidence count probability @xmath92 for the first three input schmidt modes as a function of fiber length .
figure [ fig:45.4 - 5](a ) shows that , in contrast to the previous case , the hom singular values decrease rapidly with schmidt mode number , making this case highly discriminatory " in regards to its hom interference properties ; only a few schmidt modes exhibit hom singular values significantly above zero .
figure [ fig:45.6 - 9 ] shows the absolute amplitudes and phase in both time and frequency of the first three green input schmidt modes ( relating to the @xmath60 matrix ) .
the input amplitudes of the pump fields are also shown as a black dashed line . similarly to the previous case , these first few schmidt modes are qualitatively similar to the hg basis functions , although they exhibit more asymmetry relative to their width than do the schmidt modes in the previous case . as can be seen in fig .
[ fig:45.6 - 9 ] ( c ) , the frequency amplitudes have widths that are comparable to ( and , for the second and third mode , somewhat larger than ) the translation bandwidth , which contributes to the fact that the higher - order modes experience a low translation efficiency .
the phase plots are similar to what they were in the previous cases ; the phases in time are mostly parabolic with positive curvature , implying the inputs are down - chirped .
figure [ fig : p11.param ] shows a parameter study where the pump pulses are identical gaussians with temporal fwhm durations of 1000 ps and peak powers of 200 mw . in fig .
[ fig : eff.param ] ( a ) , the minimal @xmath92 value achieved along the fiber is plotted as a function of input signal pulse fwhm duration . for short signal durations
the minimum @xmath92 value hardly changes from 1 , which is the initial value ( @xmath173 ) of @xmath92 for all modes , but for long signal durations the minimum @xmath92 value drops to nearly zero .
this is most likely due to the short signals having large spectral bandwidths and the long signals having small spectral bandwidths as compared to the translation bandwidth .
this follows because pulses with large relative bandwidth experience poor translation as compared to pulses with small relative bandwidths . in fig .
[ fig : eff.param](b ) , the length at which the minimal @xmath92 value occurs is plotted as a function of input signal duration . for short signal widths ,
the optimal length is the same as the fiber length .
this is likely due to the difficulty translating signals with large bandwidths ; it takes the entire length of the fiber to translate the small amount of the signal that is within the translation phase - matching bandwidth .
this dynamic changes once the signal width is on the order of the phase - matching bandwidth , at which point the bs process is no longer limited by bandwidth but by the shape and power of the pumps , hence the longer optimal lengths for longer signal pulses .
figure [ fig : p11.ps70 ] shows a parameter study where the pump pulses are identical gaussians with temporal fwhm durations of 70 ps and peak powers of 200 mw .
the value of @xmath92 , shown in part ( a ) , for short signal inputs behaves similarly to the previous long - pumps case : a sharp increase in @xmath92 for very short signal widths due to the signal having larger translation phase - matching bandwidth than the fiber . unlike the previous case , @xmath92 reaches a minimum around 40 ps and begins to slowly increase for wider input signal widths .
the increase for wider signal widths is likely due to signals beginning or evolving to be wider than the pumps , so that only part of the signals wavepacket could be translated .
the length at which the minimum @xmath92 occurs is shown in part ( b ) .
this is a qualitatively similar result to the previous long - pumps cases . for short signals the optimal length is equal to the length of fiber ( fixed at 20 m ) , due to the excessive bandwidth of the signal in relation to the phase - matching bandwidth .
but when the signal durations are of the same order as the pumps , the optimal length is not as constrained by the phase - matching bandwidth in the same manner , and becomes somewhat shorter than the length of the fiber .
but when the signals become as long as or longer than the pump , the overlap between all the fields in time significantly decreases , leading to less translation and raising the value of @xmath92 . in this regime
the fiber length also limits the translation , hence the reason the optimal length is equal to the length of the fiber .
this suggests that were the fiber longer the value of @xmath92 would be lower , which indeed is the case .
in order to gain further insight into the numerical simulations presented above , we developed an analytical perturbative solution , valid for not - too - high conversion efficiencies .
the main insights provided by this solution are : ( 1 ) the schmidt modes are found , to good approximation , to equal hg functions in this regime .
( 2 ) the origin of the input schmidt mode temporal delays seen in figs .
[ fig:171.6 - 9 ] and [ fig:45.6 - 9 ] can be understood as arising from differences of group velocities and the requirement for maximal interaction with the pump pulses in the fiber .
the detailed derivation is given in appendix b , the results of which are summarized here .
equations ( [ eq : ggbsvd])([eq : gbbsvd ] ) show how , in general , to express the backward green functions as sums of products of the input schmidt modes @xmath57 and @xmath76 , the output modes @xmath174 and @xmath58 , and the schmidt coefficients @xmath56 and @xmath175 .
furthermore , the diagonal green functions @xmath64 and @xmath70 can be deduced from the off - diagonal green functions @xmath54 and @xmath69 ( and vice versa ) . in appendix
b we rewrite the third and fourth lines of ( [ eq:4mode ] ) in a frame moving with the average group speed of the signals .
we include the effects of signal convection and second - order dispersion , but do not include the effects of time - dependent cross - phase modulation , which were discussed qualitatively after fig .
[ fig:285.5 - 8 ] . in this frame
the dispersion relations for the green and blue signal fields @xmath176 and @xmath177 give the associated wavenumbers , correct to second order in frequency difference . in our simulations
@xmath178 is negative , so the green mode is faster than the blue mode . for the parameters in our simulations ,
the temporal walk - off of pulses is significant , but pulse broadening by linear group - velocity dispersion is negligible .
therefore , we set @xmath179 here for the simplest analytical solution .
we assume the pumps have identical temporal intensity profiles , proportional to @xmath180 , where @xmath181 is a width parameter .
we then find that the perturbative solution for the signals ( to first order in the coupling parameter @xmath182 ) involves green functions that can be expressed in svd form , as are those in ( [ eq : ggbsvd])([eq : gbbsvd ] ) . in this perturbation theory
the schmidt modes are independent of the coupling parameter @xmath182 and pump peak intensities , but of course this will break down at higher values of @xmath182 or pump peak intensities .
notably , we find that the schmidt modes are given by hg functions in this regime .
the explicit forms of the input green and blue schmidt modes are , respectively : @xmath183 where the @xmath184 are the hg functions defined in ( 26 ) , and the temporal scale factor is @xmath185 .
likewise , the output green and blue schmidt modes are , respectively : @xmath186 note that the spectral phases in these mode solutions correspond to temporal delays or advances , which in this simple model result in both green and blue modes maximally overlapping with the pump pulses at the midpoint ( @xmath187 ) of the medium . in our simulations
@xmath178 is negative , so the green input mode is delayed and the blue input mode is advanced ( although note that in our simulations the two pumps walk off from each other , which is not the case for the analytic solution given here ) . using the same parameters as in the numerical simulations leading to figs .
[ fig:171.6 - 9 ] and [ fig:45.6 - 9 ] , the approximate input schmidt modes for the blue signal are plotted as functions of time and frequency in figs .
[ fa1 ] and [ fa2 ] .
( note that the time - domain modes are also described by hg functions . ) for conversion efficiencies up to 50% , the approximate analytical solutions predict the modes widths accurately , and their delays or advances qualitatively . for higher conversion efficiencies ,
accurate analytical solutions are not yet known .
clearly , the perturbative solutions will not hold for conversion efficiencies approaching 100% , as we see significant alterations of the modes in figs .
[ fig:285.5 - 8 ] and [ fig:51.5 - 8 ] .
in this paper we studied quantum frequency translation ( qft ) by the four - wave mixing interaction of bragg scattering ( bs ) using a realistic numerical model of photonic crystal fiber and an analytical approximate solution valid for low conversion efficiency . because bs implements a quantum operator transformation analogous to that of an ordinary beam splitter ( mode coupler ) , it is noiseless ( background free ) @xcite .
furthermore , the beam - splitter nature of bs leads to linear - optical quantum operations between weak light pulses of different central frequencies , in particular two - photon hong - ou - mandel ( hom ) interference @xcite .
the main questions addressed were : ( 1 ) under what conditions does high - efficiency qft take place ? ( 2 ) under what conditions does high - efficiency hom interference take place ? ( 3 ) given the shapes of the pump pulses , how can one design ideal signal pulse shapes to optimize the above - mentioned processes ? ( 4 ) how are these ideal signal shapes affected by the level of the conversion efficiency for the qft ?
that is , are the ideal shapes invariants for a given pump shape , or do they evolve differently at higher conversion efficiency than at low conversion efficiency ?
our modeling included realistic input pump and signal fields and effects such as convection , dispersion , self - phase modulation and cross - phase modulation .
this work extends previous theoretical and numerical work @xcite on these subjects not only through explicitly modeling of specific cases ( specific input signal wavepackets ) , but more generally through the determination of the green function that governs the bs process for an entire range of input signal wave packets . knowing the green function for a process enables the calculation of the signal input and output schmidt modes and corresponding singular values ( generalizations of eigenmodes and eigenvalues , respectively ) , found by the singular value decomposition of the green function .
knowledge of the schmidt modes and their associated singular values enables the calculation of the effect of the bs process on the input signals without the need to further solve the pulse propagation equations .
most importantly for single - photon inputs , this includes the frequency translation and two - color two - photon interference effects .
it also allows for the optimization of these effects
. in particular , conditions for achieving high translation efficiency of input signal wave packets with gaussian pumps with both long ( 1000 ps ) and short ( 70 ps ) durations at a modest peak power of 400 mw were discovered by finding the green function , and then schmidt modes , of the process .
these two cases represent the quasi - cw regime and pulsed regime , respectively , in regards to the translation phase - matching bandwidth of the process .
the long - pumps case was found to be non - discriminatory , " where many schmidt modes were translated with high efficiency .
therefore , this configuration of fiber and pumps would be useful for effective translation of many different input signal wave packets or in the case where an input packet is a superposition of the high - efficiency ones .
in contrast , the short - pump case was found to be discriminatory , " where only the first few schmidt modes experienced significant translation .
configurations of the discriminatory type would be most suited to applications where it is desirable for only a few types of wave packets ( ideally only one ) to experience significant translation , such as in a temporal - mode - selective add / drop filter @xcite .
recently it was proposed that such a filter , or pulse gate " can be implemented using frequency upconversion by sum - frequency generation in an optical crystal driven by a single strong pump pulse @xcite .
this process is analogous to the bs process studied here , with a major difference being that in upconversion the two signal fields being connected by frequency translation must be separated by an amount equal to the pump frequency .
this limits the process to large frequency separations .
the numerical scheme developed here is well suited for modeling such a process . to gain insight into the fwm bragg scattering process
, we developed an analytical perturbation solution .
we find that for low efficiency qft the schmidt modes are well approximated by a set of orthogonal hermite - gaussian functions with characteristic temporal widths determined simply by group - velocity differences and the spectral width of the pumps .
nevertheless , we emphasize that the hg modes become strongly altered for high conversion , in apparent contradiction to what is argued in @xcite .
also studied were the conditions for achieving a highly visible two - color photon interference .
we studied two pump configurations having long ( 1000 ps ) and short ( 70 ps ) durations and a peak power ( 200 mw ) close to the optimum value for a 20-m - long fiber .
similar to the high - efficiency translation cases , the long - pump case was non - discriminatory " and the short - pump cases was discriminatory " in regards to high visibility of the two - color photon interference effect .
we verified numerically that for the long - pumps configuration up to about 50 percent conversion efficiency , both the input and output schmidt mode sets are very well approximated by orthogonal hermite
gauss ( hg ) function sets . for the short - pumps configuration ,
the hg functions are reasonable approximations .
the approximate theory on which the hg functions are based allows one to predict the schmidt - mode widths and singular values as functions of the physical parameters . in conclusion ,
qft by fwm bragg scattering appears to be a powerful and robust technique for linearly interacting two quantum fields of distinct frequencies .
besides offering the ability to change the colors of single photons , it may offer other useful abilities in quantum information science , such as linear - optical quantum computing over the rainbow " of optical and near - ir frequencies .
this work was partially supported by nsf grant eccs-0802109 and by the oregon nanoscience and microtechnologies institute .
we thank s. van enk for helpful comments .
the probability for the occurrence of the @xmath188 and @xmath189 states requires care to derive due to the possibility that degenerate photons would be created in the green and blue frequency regions . in this case
the primed variables attached to the operators can not be taken over to their unprimed counterparts ( as @xmath190 to @xmath191 was for example ) .
this ambiguity can be avoided by calculating @xmath96 directly as the product of the appropriate bra and ket vectors ( similarly for @xmath97 ) . from the term corresponding to the creation of two green photons
leads to an expression for @xmath96 of @xmath192 where the variables of the conjugate wavefunction are denoted with an overhead bar .
the key to evaluating this expression is to utilize the field commutation relation . in the frequency variable
the commutation relation is @xmath193 = \delta(\omega-\omega^{'})$ ] .
hence the commutation relation to be used on the innermost operators is @xmath194 this breaks the expression into two parts , and leads to the expression , employing dummy variable notation , latexmath:[\[\label{eq : p20 } p_{20 } = \left|\int d\omega a_{gg}^{*}(l,\omega)a_{gb}(l,\omega ) \right|^2 + \int d\omega d\omega^ { ' } @xmath97 is the same when @xmath196 and @xmath197 .
the two terms of have distinct physical interpretations .
the first term , derived in part from the delta function of , is essentially the overlap of the two spectral functions @xmath198 and @xmath199 , and accounts for cases in which the two green photons are created in the same mode .
this purely quantum mechanical term arises from the bosonic nature of photons , the operators of which obey the above - mentioned commutation relation .
the second term accounts for the cases in which the two green fields are purely classical , as if these fields did not obey the quantum bosonic commutation relations .
according to ( [ eq : ggbsvd])([eq : gbbsvd ] ) , the ( backward ) constituent green functions are determined by the input schmidt modes @xmath57 and @xmath76 , the output modes @xmath174 and @xmath58 , and the schmidt coefficients @xmath56 and @xmath200 .
furthermore , the diagonal green functions @xmath64 and @xmath70 can be deduced from the off - diagonal green functions @xmath54 and @xmath69 ( and vice versa ) . in this appendix , the off - diagonal green functions and the associated schmidt modes will be determined approximately , for low conversion efficiencies .
the third and fourth of ( 24 ) govern the signal evolution in the time domain . in the frequency domain ,
the signal equations are @xmath201 where @xmath105 is an envelope frequency , @xmath202 and @xmath203 are the associated wavenumbers ( correct to second order in frequency ) , and @xmath204 is the fourier transform of the coupling term @xmath205 divided by @xmath206 . equations ( [ a1 ] ) and ( [ a2 ] ) are valid in a frame moving with the average group speed of the signals ( so @xmath207 are the relative group slownesses ) .
they include the effects of convection and second - order dispersion , but do not include the effect of time - dependent cross - phase modulation ( which chirps the signals ) .
let @xmath208 $ ] .
then the transformed amplitudes @xmath209 obey the transformed equations @xmath210
b_b(z,\omega_b)d\omega_b , \label{a3 } \\
\partial_zb_b(z,\omega_b ) & = & i\int \gb^*(z,\omega_b - \omega_g)\exp[ik_g(\omega_g)z - ik_b(\omega_b)z ] b_g(z,\omega_g)d\omega_g .
\label{a4 } \end{aligned}\ ] ] in general , the @xmath211-equations are no simpler than the @xmath212-equations , because they depend explicitly on @xmath18 . however , in the low - conversion - efficiency regime , one can replace the mode amplitudes on the right sides by the input amplitudes . in this regime , @xmath213 where @xmath214 $ ] is the kernel in ( [ a3 ] ) . note that @xmath215^*$ ] , so @xmath216 . if the pumps are gaussian in time and identical , and their relative convection ( walk - off ) is neglected , then @xmath217 , where @xmath218 is the peak pump power and @xmath73 is the ( gaussian ) pump width . in this case ,
@xmath219\exp(-\sigma^2\omega^2/2 ) , \label{a7}\ ] ] where the width parameter @xmath220 .
define the wavenumber - mismatch function @xmath221 .
then @xmath222 by combining the preceding results , and using the identity @xmath223 , one finds that the integrated kernel @xmath224\exp[-i\beta_1l(\omega_g + \omega_b)/2 - i\beta_2l(\omega_g^2 - \omega_b^2)/4 ] \nonumber \\ & & \times\exp[-\sigma(\omega_g - \omega_b)^2/2]\sinc[\beta_1l(\omega_g + \omega_b)/2 + \beta_2l(\omega_g^2 - \omega_b^2)/4 ] .
\label{a9 } \end{aligned}\]]the @xmath105 terms in the first exponential produce delays ( advances ) in the time domain , whereas the @xmath225 terms produce chirps in the frequency domain ( convolutions in the time domain ) . by making the approximation @xmath226 , where @xmath227 , and omitting dispersion ( setting @xmath179 ) , which is negligible for the parameters of our simulations , one finds that @xmath228\exp[-i\beta_1l(\omega_g + \omega_b)/2 ] \nonumber \\ & & \times\exp[-\sigma^2(\omega_g - \omega_b)^2/2 - \beta^2(\omega_g + \omega_b)^2/2 ] , \label{a10 } \end{aligned}\ ] ] where the walk - off parameter @xmath229 .
because kernel ( [ a10 ] ) is a gaussian function of frequency , it can be decomposed using the mehler identity @xcite @xmath230 = [ \pi(1 - \mu^2)]^{1/2}\sum_{n=0}^\infty |\mu|^n\psi_n(x)\psi_n(y ) , \label{a11}\ ] ] where the hermite - gaussian functions @xmath231 were defined in ( 27 ) . by comparing ( [ a10 ] ) and ( [ a11 ] ) , one finds that @xmath232 , @xmath233 and @xmath234 , where @xmath235 .
hence , kernel ( [ a10 ] ) has the singular value ( schmidt ) decomposition @xmath236 , \label{a12}\ ] ] where the schmidt coefficients @xmath237^{1/2}|\mu|^n$ ] and the ( normalized ) schmidt modes @xmath238 .
the kernel is separable when @xmath239 ( @xmath240 ) , in which case @xmath241 and @xmath242 . in this case , only one mode will be partially translated ; the others will be unaffected , creating a mode - selective filter , similar to that proposed in @xcite .
let @xmath243 be the green function that describes the effect on the output transformed amplitude @xmath244 of the input transformed amplitude @xmath245 .
then ( [ a5 ] ) implies that @xmath246 .
the associated ( forward ) green function for the original amplitudes , @xmath247 and @xmath248 , is @xmath249 .
hence , @xmath250 , \label{a13}\ ] ] from which it follows that the output ( green ) and input ( blue ) schmidt mode functions are given by @xmath251 respectively .
likewise , ( [ a6 ] ) implies that the transformed green function @xmath252 , and the original green function @xmath253 .
hence , @xmath254 , \label{a16}\ ] ] from which it follows that the output ( blue ) and input ( green ) schmidt mode functions are given by @xmath255 respectively . to obtain these
, we needed to account for the distinction between forward and backward green functions , and note that here we are propagating @xmath212 , which corresponds to the quantum operator @xmath150 , not @xmath256 . if @xmath178 is positive , the green mode is slower than the blue mode .
the preceding results show that the green input is advanced and the blue input is delayed , so their collision is centered on the midpoint of the fiber .
this location maximizes the distance over which the signals interact with the peaks of the pump pulses . in our simulations
@xmath178 is negative , so the green input is delayed and blue input is advanced . | we study quantum frequency translation and two - color photon interference enabled by the bragg scattering four - wave mixing process in optical fiber . using realistic model parameters , we computationally and analytically determine the green function and schmidt modes for cases with various pump - pulse lengths .
these cases can be categorized as either non - discriminatory " or discriminatory " in regards to their propensity to exhibit high - efficiency translation or high - visibility two - photon interference for many different shapes of input wave packets or for only a few input wave packets , respectively .
also , for a particular case , the schmidt mode set was found to be nearly equal to a hermite - gaussian function set .
the methods and results also apply with little modification to frequency conversion by sum - frequency conversion in optical crystals .
99 a. p. vandevender and p. g. kwiat , `` high efficiency single photon detection via frequency up - conversion , '' j. mod .
opt . * 51 * , 14331445 ( 2004 ) .
s. tanzilli , w. tittel , m. halder , o. alibart , p. baldi , n. gisin , and h. zbinden , `` a photonic quantum information interface , '' nature * 437 * , 116120 ( 2005 ) . c. j. mckinstrie , j. d. harvey , s. radic , and m. g. raymer , `` translation of quantum states by four - wave mixing in fibers , '' opt .
express * 13 * , 91319142 ( 2005 ) .
a. h. gnauck , r. m. jopson , c. j. mckinstrie , j. c. centanni , and s. radic , `` demonstration of low - noise frequency conversion by bragg scattering in a fiber , '' opt .
express * 14 * , 89898994 ( 2006 ) .
d. mchin , r. provo , j. d. harvey , and c. j. mckinstrie , `` 180-nm wavelength conversion based on bragg scattering in an optical fiber , '' opt .
express * 14 * , 89958999 ( 2006 ) .
h. j. mcguinness , m. g. raymer , c. j. mckinstrie , and s. radic , `` wavelength translation across 210 nm in the visible using vector bragg scattering in a birefringent photonic crystal fiber , '' photon .
. lett .
* 23 * , 109111 ( 2011 ) .
j. m. huang and p. kumar , `` observation of quantum frequency conversion , '' phys .
rev .
lett . * 68 * , 21532156 ( 1992 ) .
h. j. mcguinness , m. g. raymer , c. j. mckinstrie , and s. radic , `` quantum frequency translation of single - photon states in a photonic crystal fiber , '' phys .
rev .
lett . * 105 * , 093604 ( 2010 ) .
m. t. rakher , l. ma , o. slattery , x. tang , and k. srinivasan , `` quantum transduction of telecommunications - band single photons from a quantum dot by frequency upconversion , '' nature photon . * 4 * , 786791 ( 2010 ) .
m. g. raymer , s. j. van enk , c. j. mckinstrie , and h. j. mcguinness , `` interference of two photons of different color , '' opt
. commun . * 283 * , 747752 ( 2010 ) .
e. knill , r. laflamme , and g. j. milburn , `` a scheme for efficient quantum computation with linear optics , '' nature * 409 * , 4652 ( 2001 ) .
p. kok , w. j. munro , k. nemoto , t. c. ralph , j. p. dowling , and g. j. milburn , `` linear optical quantum computing with photonic qubits , '' rev .
mod .
phys . * 79 * , 135174 ( 2007 ) . n. c. menicucci , s. t. flammia , and o. pfister , `` one - way quantum computing in the optical frequency comb , '' phys .
rev lett .
* 101 * , 130501 ( 2008 ) .
j. l. obrien , `` quantum computing over the rainbow , '' physics * 1 * , 23 ( 2008 ) .
a. eckstein , b. brecht , and c. silberhorn , `` a quantum pulse gate based on spectrally engineered sum frequency generation , '' opt .
express * 19 * , 1377013778 ( 2010 ) j. a. salehi , a. m. weiner , and j. p. heritage , `` coherent ultrashort light pulse code - division multiple access communication systems , '' j. lightwave technol .
* 8 * , 478491 ( 1990 ) .
m. e. marhic , `` coherent optical cdma networks , '' j. lightwave technol . * 11 * , 854864 ( 1993 ) .
b. brecht , a. eckstein , a. christ , h. suche , and c. silberhorn , `` from quantum pulse gate to quantum pulse shaper - engineered frequency conversion in nonlinear optical waveguides , '' http://arxiv.org/abs/1101.6060 ( 2011 ) .
d. kielpinski , a.f .
corney , h.m .
wiseman , `` quantum optical waveform conversion '' phys .
rev lett . * 106 * , 130501 ( 2011 ) .
h. j. mcguinness , `` the creation and frequency translation of single - photon states of light in optical fiber , '' ph.d .
thesis , university of oregon , eugene , oregon ( 2011 ) .
k. inoue , `` tunable and selective wavelength conversion using fiber four - wave mixing with two pump lights , '' ieee photon
. technol . lett .
* 6 * , 14511453 ( 1994 ) .
r. w. boyd , _ nonlinear optics , 3rd ed . _
( elsevier science , amsterdam , 2008 ) .
c. j. mckinstrie , s. radic , and m. g. raymer , `` quantum noise properties of parametric amplifiers driven by two pump waves , '' opt .
express * 12 * , 50375066 ( 2004 ) .
a. ekert and p. l. knight , `` entangled quantum systems and the schmidt decomposition , '' am . j. phys . *
63 * , 415423 ( 1995 ) .
p. d. drummond and c. w. gardiner , j. phys .
a * 13 * , 23532368 ( 1980 ) . c. w. gardiner , _ quantum noise _
( springer , berlin , 1992 ) .
w. wasilewski , and m. g. raymer , `` pairwise entanglement and readout of atomic - ensemble and optical wave - packet modes in traveling - wave raman interactions , '' phys
. rev .
a , * 73 * , 063816 ( 2006 ) . w. wasilewski , a. i. lvovsky , k. banaszek , and c. radzewicz , `` pulse squeezed light : simultaneous squeezing of multiple modes , '' phys .
rev .
a , * 73 * , 063816 ( 2006 ) .
g. p. agrawal , _ nonlinear fiber optics , 4th ed . _ ( elsevier science , amsterdam , 2008 ) .
w. h. press , s. a. teukolsky , w. t. vetterling , and b. p. flannery , _ numerical recipes in fortran , 2nd ed . _
( cambridge university press , cambridge , 1992 ) .
p. s. j. russell , `` photonic crystal fibers , '' science * 299 * , 358362 ( 2003 ) .
g. k. wong , a. y. h. chen , s. w. ha , r. j. kruhlak , s. g. murdoch , r. leonhardt , j. d. harvey , and n. joly , `` characterization of chromatic dispersion in photonic crystal fibers using scalar modulation instability , '' opt .
express * 13 * , 86628670 ( 2005 ) .
f. g. mehler , `` uber die entwicklung einer funktion von beliebig vielen variablen nach laplaceschen functionen hherer ordnung , '' j. reine angew .
math . * 66 * , 161176 ( 1866 ) .
p.
m. morse and h. feschbach , _ methods of theoretical physics _ ( mcgraw - hill , new york , 1953 ) . |
one can learn a great deal about the pairing mechanism of a superconductor from the electronic properties in the presence of inhomogeneity , such as vortices and impurities .
although multiple - band superconductivity was first studied almost 50 years ago , @xcite the structure of a vortex in multiband superconductors has not been well understood , especially when impurities are present .
the best material that has been discovered so far for studying multiband superconductivity is mgb@xmath0.@xcite the consensus is that superconductivity is driven by electron - phonon interactions , and that it can be well described by a two - band model , with the `` strong '' @xmath1 band ( energy gap @xmath2 mev ) and the `` weak '' @xmath3 band ( @xmath4 mev).@xcite the two energy gaps vanish at a common transition temperature @xmath5 ( refs . ) and there is evidence of induced superconductivity in the @xmath3 band.@xcite the one unusual aspect of this material as a multiband superconductor is the effect of impurities .
the standard theory@xcite tells us that interband scattering by nonmagnetic impurities should reduce @xmath5 and the gap ratio .
mazin and co - workers,@xcite however , have shown theoretically that this does not apply to mgb@xmath0 , due to different symmetries of the @xmath1 and @xmath3 orbitals , and hence negligible interband scattering .
moreover , impurities or defects , in particular those at mg - sites which tend to occur more easily than at b - sites , affect only the @xmath3 band strongly .
indeed , many experiments have shown that the @xmath1 and @xmath3 bands are essentially in the ballistic and diffusive limit , respectively.@xcite even in samples in which the @xmath1 band is influenced by impurities substantially , two - gap superconductivity is retained.@xcite the question of whether or not the two gaps merge in dirty samples has not yet been settled.@xcite due to induced superconductivity and possibly also owing to these peculiar effects of impurities , the electronic structure around a vortex in mgb@xmath0 has been found to exhibit intriguing properties .
_ et al._,@xcite by tunneling along the @xmath6 axis , have probed the vortex core structure in the @xmath3 band .
the local density of states ( ldos ) was found to be completely flat as a function of energy at the vortex core , showing no sign of bound states .
also the ` core size ' ( @xmath750 nm ) as measured by a decay length of the zero - bias ldos turned out to be much longer than expected from @xmath8 ( @xmath710nm).@xcite moreover , the existence of two effective coherence lengths in mgb@xmath0 has been suggested by the @xmath9sr measurement of a vortex lattice.@xcite there has been , however , no experiment directly probing the @xmath1 band in the vortex state , and the electronic structure around a vortex in the @xmath1 band is yet to be determined .
theoretically , the vortex structure in a two - band superconductor has been studied in terms of two clean bands@xcite and two dirty bands.@xcite neither of these models , however , applies to many mgb@xmath0 samples , which have the clean @xmath1 and dirty @xmath3 bands .
recently , we have formulated a unique model for a multiband superconductor with both a clean and a dirty band.@xcite we have studied the effects of induced superconductivity and impurities in the weak diffusive band on the electronic properties in the strong ballistic band around a vortex . a particularly intriguing feature found in this model is the possibility of bound states near the gap edge in the ballistic band in the vortex core , in addition to the well - known caroli - de gennes - matricon bound states.@xcite such bound states do not exist in a single ballistic band , and they arise solely from coupling to the diffusive band .
our model has also been applied to study the kramer - pesch effect@xcite in coupled clean and dirty bands.@xcite it has been found that the kramer - pesch effect is induced in the dirty band , which is absent when there is no coupling with the clean band .
thus , hybridization of ballistic and diffusive bands can lead to unusual properties of a multiband superconductor , and the vortex core structure is an example in which the effects of induced superconductivity and impurities manifest clearly . in this work , we make an extensive study of the electronic structure in the vicinity of a vortex in a strong ballistic band , under the influence of hybridization with a weak diffusive band .
our model is based on coupled eilenberger and usadel equations , which are solved directly and numerically .
we assume that interband scattering by impurities is negligible and that the two bands are coupled only by the pairing interaction , as justified for typical mgb@xmath0 samples .
unique features of the order parameter , the current density , and the vortex core spectrum are examined in detail . in particular , we study the development of the gap - edge bound states as various physical parameters are varied .
although this work has been motivated by the impurity effects @xcite and the stm measurement of the vortex state in mgb@xmath0 , @xcite our model is general and applicable to any superconductor in which ballistic and diffusive bands are coupled mainly by the pairing interaction .
the paper is organized as follows . the formulation and computational details
are given in secs .
[ theory ] and [ comp ] , respectively .
results are presented in sec .
[ results ] and they are summarised and discussed in sec . [ conclusion ] . throughout the paper we use units with @xmath10 .
both the ballistic and diffusive limits of superconductivity can be described within one unified theory , the quasiclassical theory of superconductivity.@xcite the central quantity of this theory , containing all the physical information , is the quasiclassical green function , or propagator , @xmath11 . here
@xmath12 is the quasiparticle energy measured from the chemical potential , @xmath13 the quasiparticle momentum on the fermi surface of band @xmath14 , and @xmath15 is the spatial coordinate .
the hat refers to the 2@xmath162 matrix structure of the propagator in the nambu - gorkov particle - hole space . in the ballistic case ,
the equation of motion for @xmath17 is the eilenberger equation , and in the diffusive case the usadel equation .
our model is appropriate for any two - band superconductor with a clean and a dirty band ( generalisation to several bands is straightforward ) .
however , having in mind @xmath18 , for definiteness we call the two bands @xmath1 and @xmath3 bands , respectively . in the clean @xmath1 band , @xmath19 satisfies the eilenberger equation , @xcite @xmath20 + i { \bf v}_{f\sigma } \cdot { \nabla } \hat g_\sigma = \hat 0 , \label{eil}\ ] ] where @xmath21 is the fermi velocity and @xmath22 the ( spatially varying ) order parameter .
the three pauli matrices in nambu - gorkov space are denoted by @xmath23 , @xmath24 , and @xmath25 $ ] denotes the commutator . throughout this work
, we ignore the variation of the magnetic field in the vortex core , assuming a strongly type - ii superconductor ( this is justified for example for mgb@xmath0 ) . for the @xmath3 band
we assume that it is in the diffusive limit . in the presence of strong impurity scattering ,
the momentum dependence of the quasiclassical green function is averaged out , and the equation of motion for the resulting propagator @xmath26 reduces to the usadel equation,@xcite
@xmath27 + { \nabla } \frac{\mathbb{d}}{\pi } ( \hat g_\pi { \nabla } \hat g_\pi ) = \hat 0,\ ] ] with the diffusion constant tensor @xmath28 .
both ballistic and diffusive propagators are normalized according to@xcite @xmath29 a two - band superconductor with a ballistic and a diffusive band can exist only if interband scattering by impurities is weak .
we neglect in the following interband scattering by impurities , and assume that the quasiparticles in different bands are coupled only through the pairing interaction .
self - consistency is achieved through the coupled gap equations for the spatially varying order parameters in each band , @xmath30 where @xmath31 , and @xmath32 .
the coupling matrix @xmath33 determines the pairing interaction , @xmath34 is the fermi - surface density of states on band @xmath35 , and @xmath36 here @xmath37 is the upper off - diagonal ( 1,2 ) element of the matrix propagator @xmath38 , and @xmath39 is a cutoff energy .
the fermi surface average over the @xmath1-band is denoted by @xmath40 .
note that within our model the precise form of the fermi surface in the @xmath3 band is not relevant , as in the diffusive limit all necessary information is contained in the diffusion constant tensor @xmath28 . for simplicity
we have assumed in our calculation an isotropic tensor @xmath41 .
the diffusion constant defines the @xmath3-band coherence length @xmath42 . for the @xmath43-band fermi surface
we assume a cylindrical shape , as motivated by the fermi surface of mgb@xmath0 .
this allows us to treat the @xmath43 band as quasi - two - dimensional , @xmath44 , with cylindrical coordinates @xmath45 and the unit vector @xmath46 in direction of @xmath47 .
we define the coherence length in the @xmath43 band as @xmath48 .
it will be used as length unit throughout this paper .
the numerical solution of the ( nonlinear ) system of eqs .
( [ eil])([norm ] ) is greatly simplified by using the riccati parameterization of the green functions , both for the ballistic case@xcite and for the diffusive case,@xcite @xmath49 with @xmath50 and @xmath51 .
the transport equation for @xmath52 is given by,@xcite @xmath53 and for @xmath54 it is@xcite
@xmath55 we solve eqs .
( [ eil])([f2 ] ) self - consistently .
after self - consistency has been achieved for the order parameters , the ( for the @xmath1 band , angle - resolved ) ldos in each band can be calculated by @xmath56 where @xmath57 is the upper diagonal ( 1,1 ) element of @xmath38 .
the current density around the vortex has contributions from both @xmath3 and @xmath43 bands , @xmath58 the corresponding expressions are @xmath59 \tanh\left ( \frac{\epsilon}{2t}\right ) , \ ; \label{cd}\end{aligned}\ ] ] where @xmath60 is the electron charge .
the first step is to write eqs . and in a form that is independent of the cut - off energy @xmath61 and the cut - off dependent interaction matrix @xmath62 .
we diagonalize the interactions by @xmath63 where @xmath64 and @xmath65 .
the band index runs over @xmath66 , and thus we need two basis functions for a complete system , denoted by @xmath67 .
we expand the order parameter and the anomalous green functions in terms of the eigenvectors @xmath68 the eigenvectors are orthonormal .
we also introduce the vectors @xmath69 using the expansion @xmath70 with @xmath71 and @xmath72 , the gap equations are given by @xmath73 without restriction we can choose the largest eigenvalue to be @xmath74
. it can be eliminated together with the cutoff frequency in favor of the transition temperature between the normal and superconducting states , @xmath5 .
although both coupling constants @xmath74 and @xmath75 are cutoff dependent , the parameter @xmath76 defined by @xmath77 is cutoff independent .
we parameterize the two coupling constants by the two cutoff independent quantities @xmath5 and @xmath76 . near @xmath5 ,
the subdominant order parameter @xmath78 is of higher order in @xmath79 than @xmath80 is .
the bulk gaps on the two fermi surface sheets are thus determined near @xmath5 by @xmath80 only , and are given by @xmath81 we define the ratio of the bulk gaps on the two bands near @xmath5 , @xmath82 we also introduce the notation @xmath83 .
then the matrix of eigenvectors is given by @xmath84 in the homogeneous case , at zero temperature @xmath85 where @xmath39 is the usual bcs cutoff frequency of the order of the debye frequency .
using the bcs formula @xmath86 ( @xmath87 is euler s constant ) , we obtain the zero - temperature gap equations @xmath88 where @xmath89 denotes the bulk order parameter at zero temperature on band @xmath14 . from this
we can derive an equation for the ratio @xmath90 ( assuming that the gap ratio @xmath91 is real ) , @xmath92 and for the dominant zero - temperature gap @xmath93\ln |\rho_0| } .
\end{aligned}\ ] ] equation ( [ gapratio ] ) is a quadratic equation for @xmath94 , which is readily solved analytically in terms of @xmath95 , @xmath96 and @xmath76 : @xmath97 in the zero - temperature limit as given by eqs . and :
( a ) as a function of @xmath98 for @xmath99 and various values of @xmath100 , and ( b ) as a function of @xmath101 for various sets of @xmath95 and @xmath100 . for a fixed @xmath102
, @xmath103 shows nonlinear behaviour as a function of @xmath95 , where the bcs value is recovered at @xmath104 and 1 .
the bulk gap is enhanced as @xmath102 increases , and this effect is larger for larger @xmath95 and for smaller @xmath100 . ] the four material parameters @xmath105 , @xmath96 , @xmath100 , and @xmath5 completely specify the bulk behaviour of the system .
we illustrate in fig .
[ gaptc ] how the dominant bulk gap to @xmath5 ratio at zero temperature is affected by @xmath95 , @xmath106 , and @xmath100 . in ( a ) , @xmath107 determined by eqs . and
is shown as a function of @xmath95 , for @xmath99 and various values of @xmath100 . in ( b ) we compare @xmath107 as a function of @xmath102 for several values of @xmath95 , and for @xmath108 and 0.1 .
it can be seen in fig .
[ gaptc](a ) that for a fixed @xmath102 , @xmath107 exhibits nonlinear behaviour as a function of @xmath95 depending on the value of @xmath100 , while the bcs value ( 1.764 ) is recovered at @xmath104 and 1 . for @xmath109
this ratio is always larger than the bcs value , and it is always larger for @xmath109 than for @xmath110 , for fixed @xmath95 and @xmath102 . when @xmath110 , @xmath107 can be smaller than the bcs value for a certain range of @xmath95 .
the effect becomes even more pronounced for larger @xmath102 . for a fixed @xmath95 , as shown in fig .
[ gaptc](b ) , @xmath107 increases almost linearly with increasing @xmath102 , with larger slope for larger @xmath95 . here again , the enhancement of @xmath107 is always stronger for @xmath109 than for @xmath110 . for @xmath111
there is only one band and the ratio reduces to the bcs value .
in fact , the enhancement of the bulk zero - temperature gap by stronger coulomb interactions in the subdominant channel can be estimated for small @xmath112 analytically from eqs . and , which give @xmath113 and @xmath114 . more generally , at low @xmath115 for given @xmath95 and @xmath96 , the homogeneous gap over @xmath5 ratio increases with decreasing @xmath76 in both bands .
the slope at @xmath116 in fig .
[ gaptc](a ) is independent of @xmath76 and is given by @xmath117 .
an expansion for small @xmath95 or for small @xmath101 gives for both cases @xmath118 we now discuss the relation between the gap ratios near @xmath5 , given by @xmath94 , and at @xmath119 , given by @xmath95 .
when @xmath120 , we have @xmath121 . for small @xmath112 or small @xmath122
we obtain from eq .
( [ rho ] ) the approximate relation @xmath123 for @xmath124 it follows that when the coulomb repulsion dominates in the second channel ( @xmath125 ) , the magnitude of the gap ratio at @xmath126 is increased with respect to that at zero temperature , @xmath127 ; for effective attraction in the subdominant channel ( @xmath128 ) , @xmath129 . and
@xmath130 ( left - hand panel ) and the ratio @xmath131 ( right - hand panel ) in the homogeneous case as a function of temperature @xmath115 for various values of @xmath100 , for fixed @xmath132 and @xmath133 .
there is a second transition temperature associated with the @xmath3 band for @xmath134 ( see text ) .
, title="fig : " ] and @xmath130 ( left - hand panel ) and the ratio @xmath131 ( right - hand panel ) in the homogeneous case as a function of temperature @xmath115 for various values of @xmath100 , for fixed @xmath132 and @xmath133 .
there is a second transition temperature associated with the @xmath3 band for @xmath134 ( see text ) .
, title="fig : " ] in fig .
[ homoorder ] we show the solutions of the homogeneous gap equations for both @xmath43 and @xmath135 bands as a function of temperature for various values of @xmath76 , for fixed @xmath132 and @xmath136 . in the left - hand and right - hand panels of fig .
[ homoorder ] the order parameters in the two bands and the ratio of the two are plotted , respectively , as a function of temperature . as discussed above , when @xmath137 ( thick lines ) , the ratio @xmath138 stays constant and equal to @xmath121 , while for @xmath109 ( @xmath139 ) @xmath91 is larger ( smaller ) than @xmath95
. if the order parameter in the @xmath3 band has a zero slope at @xmath5 , i.e. , @xmath140 ( see , e.g. , the lowest curves in fig .
[ homoorder ] ) , then for positive @xmath76 there is a second superconducting transition@xcite associated with the @xmath3 band at temperature @xmath141 in this case the order parameter in the @xmath135 band stays zero above @xmath142 ; below @xmath142 both order parameters are nonzero , and the zero - temperature gap ratio is @xmath143 so that the relation @xmath144 holds .
for negative @xmath76 there is no second transition in the @xmath135 band , and superconductivity is purely induced for all temperatures . for small absolute values of @xmath94 , the order parameter in the @xmath135 band
is given approximately by @xmath145 this equation is correct for @xmath146 as long as @xmath147 stays small .
for @xmath148 the order parameter in the @xmath3 band is nonzero in the entire temperature range [ see fig . [ homoorder](a ) ] , being purely induced for @xmath125 , and for @xmath146 when @xmath149 .
it follows from eq .
( [ smallrho ] ) that for @xmath150 , @xmath138 is reduced from @xmath151 as temperature decreases , while for @xmath152 it increases from @xmath153 for @xmath154 [ see also fig .
[ homoorder](b ) ] .
( top ) and @xmath155 ( bottom ) as a function of the gap ratio at @xmath5 , @xmath156 , obtained from eqs . and , for @xmath109
[ ( a ) and ( c ) ] and @xmath110 [ ( b ) and ( d ) ] . while for @xmath157 there is a unique solution , for @xmath128 , for sufficiently small @xmath94 there are three solutions .
this indicates possible first - order transitions as a function of temperature , if @xmath76 for a given @xmath94 is larger than a critical value . ] for a given @xmath95 , the critical @xmath100 , above which a second transition in the @xmath3 band can be observed , is given by @xmath158 . in the case of fig .
[ homoorder ] this yields @xmath159 .
we find that for @xmath160 there is a @xmath3 phase difference between the order parameters in the two bands for @xmath146 ( the gap ratio becomes negative ) , and for @xmath161 a more complicated picture arises , with three solutions that compete , and possible first - order phase transitions .
this is demonstrated in fig .
[ drho ] , in which the zero - temperature bulk gap to @xmath5 ratio in the @xmath1 ( top ) and @xmath3 ( bottom ) band is plotted as a function of the gap ratio at @xmath5 , @xmath162 , obtained from the solution of eqs . and .
results are shown for the effective subdominant coupling @xmath76 ranging from repulsive ( left ) to attractive ( right ) .
for @xmath157 there is a unique solution for the bulk gaps at zero temperature for a given @xmath91 ( left - hand panels ) .
however , as can be seen in the right - hand panels , when @xmath110 , there are three possible solutions to the gap equations for sufficiently small @xmath91 . as these zero - temperature solutions are associated with the same single solution near @xmath5 , this indicates that first - order phase transitions can occur as temperature is varied .
this behaviour is not the topic of this paper , however , and we assume for the remainder of the paper that @xmath163 for a given @xmath95 ( or @xmath164 for a given @xmath76 ) . for mgb@xmath0 , if estimated from the _ ab initio _ values of the electron - phonon coupling strengths@xcite and the coulomb pseudopotentials,@xcite @xmath100 varies depending on which calculations are used , while some calculations yield very small @xmath100 . but using eq .
( [ rho ] ) , one can fix @xmath100 from the observed values of @xmath95 and @xmath91 although for the latter , some data points have relatively large error bars . from the experiments of refs .
, , and , we find roughly @xmath165 . among these , the measurement of ref
. has a data point at the highest temperature , @xmath166 , and @xmath100 estimated using this point is close to zero .
therefore , we assume that for mgb@xmath0 , @xmath100 is small and can be either positive or negative .
the strength of induced superconductivity in the @xmath3 band is specified in our model by @xmath100 and @xmath95 , the latter of which is about 0.3 in mgb@xmath0 .
results are presented mainly for @xmath167 .
the @xmath168 in mgb@xmath0 has been determined experimentally to be @xmath169.@xcite we show results for @xmath170 and 2 , and discuss the effects of the @xmath3-band density of states .
finally , there is another material parameter in our model , that is the ratio of the coherence lengths in the two bands , @xmath171 .
note in this regard that in defining the @xmath135-band coherence length @xmath172 , we have used the energy scale @xmath5 rather than @xmath173 .
this is motivated by the fact that , as discussed above , superconductivity in the @xmath135 band is assumed to be mostly induced by the @xmath43 band .
consequently , for the variation of the @xmath135-band order parameter determined by the self - consistency equation , eq .
( [ gapeq ] ) , the energy scale @xmath5 is relevant .
however , the _ quasiparticle motion _ in the @xmath3 band is governed by the eilenberger transport equation supplemented with an appropriate impurity self - energy @xmath174 of the order of @xmath175 , where @xmath176 is the @xmath135-band quasiparticle lifetime due to elastic intraband impurity scattering .
to determine under which conditions the diffusive approximation can be used in the @xmath3 band , the relative size of the @xmath3-band gap , @xmath177 , and the @xmath3 band impurity self - energy , @xmath174 , matters .
thus , the condition for the @xmath3 band to be diffusive so that the eilenberger equation supplemented with @xmath178 reduces to the usadel equation is @xmath179 in practise it turns out that the diffusive approximation applies quite well already when @xmath180 is only larger , not much larger , than @xmath181 . using the above definitions for @xmath182 and @xmath183 and with the diffusion constant @xmath184 , this yields @xmath185 strong electron - phonon coupling in the @xmath1 band renormalizes the fermi velocity as @xmath186,@xcite where @xmath187 is the renormalization factor and is about 2 in the zero - frequency limit.@xcite taking the _ ab initio _ fermi velocities , @xmath188,@xcite as the unrenormalized values , and using @xmath189 as found in several experiments , we estimate @xmath190 for mgb@xmath0 . as discussed in sec .
[ total_ldos ] , an estimate from the experiment of ref .
yields a value between 1 and 3 for mgb@xmath0 . to keep the discussion general , the effects of the coherence length ratio on the vortex core structure are illustrated for @xmath191 . throughout this work
, cylindrical symmetry is assumed for the two - dimensional @xmath1 band , and an isotropic diffusion tensor for the three - dimensional @xmath135 band . up to this point
, the energy gaps ( homogeneous order parameters ) in the two bands have been denoted as @xmath181 and @xmath192 . in the following ,
@xmath181 and @xmath192 represent the spatially varying order parameters .
and @xmath193 and ( b ) the ratio @xmath194 as a function of coordinate @xmath195 along the path @xmath196 for various values of @xmath100 , for @xmath197 , @xmath132 , @xmath198 , @xmath136
. the coulomb repulsion ( which reduces @xmath76 ) renormalizes the recovery lengths of the order parameters in the two bands differently . ,
title="fig : " ] and @xmath193 and ( b ) the ratio @xmath194 as a function of coordinate @xmath195 along the path @xmath196 for various values of @xmath100 , for @xmath197 , @xmath132 , @xmath198 , @xmath136 . the coulomb repulsion ( which reduces @xmath76 ) renormalizes the recovery lengths of the order parameters in the two bands differently .
, title="fig : " ] negative @xmath76 means that the effective coulomb interaction dominates over the effective pairing interaction in the second pairing channel . in this case ,
coulomb interactions are responsible for the fact that the @xmath3 band can not maintain superconductivity on its own , and it superconducts only due to its proximity to the superconducting @xmath43 band .
we first examine the effects of the coulomb interaction on the order parameters .
we consider a vortex that extends in @xmath199 direction , with the vortex centre situated at @xmath200 for each @xmath199 . in fig .
[ order1 ] , we show the order parameter magnitudes as a function of coordinate @xmath195 along the path @xmath196 for three values of @xmath76 , appropriate for attractive , repulsive and no effective interaction in the second pairing channel .
figure [ order1](a ) shows @xmath201 and @xmath193 , and fig .
[ order1](b ) their ratio .
the coulomb repulsion renormalizes the recovery lengths of the order parameters ( the characteristic length scale over which the order parameter recovers to the bulk value ) in both bands .
although superconductivity in the @xmath3 band is mostly induced by that in the @xmath43 band , it is clear from fig .
[ order1 ] that the length scales in the two bands can differ considerably if @xmath202 . for @xmath137
there is only one length scale and hence the same recovery length in both bands . when @xmath109 ( @xmath139 ) , @xmath201 is enhanced ( suppressed ) around the vortex core , resulting in a shorter ( longer ) recovery length compared to the case with @xmath137 . on the other hand , the sign of @xmath100 has opposite effects on @xmath193 , although these effects are relatively small for the parameter set for fig .
[ order1 ] .
in fact , @xmath192 ( its length scale and magnitude ) can be affected by the @xmath3-band coulomb interaction significantly for larger @xmath95 and @xmath171.@xcite as demonstrated in fig .
[ gaptc ] , the bulk gap over @xmath5 ratios in both bands are enhanced by stronger coulomb repulsion in the diffusive band , an effect also seen in fig .
[ order1](a ) . as shown in eq .
( [ lambda0 ] ) , this renormalization of the magnitude of the bulk gap has the same sign for both bands , and is very different from the renormalization in the core region seen in fig .
[ order1](a ) , that has opposite sign for the two order parameters .
the latter effect can be seen most dramatically when plotting the ratio @xmath194 , as we do in fig .
[ order1](b ) .
for @xmath202 the ratio increases by up to a factor of 2 for @xmath108 , and decreases by about the same factor for @xmath203 . when @xmath137 , @xmath194 stays constant and equal to @xmath204 .
this results from the fact that in this case @xmath205 in eq . , leading to @xmath206
. holds for all @xmath195 . and @xmath193 and ( b ) the ratio @xmath194 as a function of coordinate @xmath195 along the path @xmath196 for various density of states ratios @xmath168 , and fixed @xmath197 , @xmath132 , @xmath198 , @xmath108 .
the @xmath3-band density of states renormalizes the recovery lengths of the order parameters , and the bulk magnitudes of the order parameters are larger for larger @xmath168 . , title="fig : " ] and @xmath193 and ( b ) the ratio @xmath194 as a function of coordinate @xmath195 along the path @xmath196 for various density of states ratios @xmath168 , and fixed @xmath197 , @xmath132 , @xmath198 , @xmath108 .
the @xmath3-band density of states renormalizes the recovery lengths of the order parameters , and the bulk magnitudes of the order parameters are larger for larger @xmath168 .
, title="fig : " ] figure [ order2 ] demonstrates the effects of the fermi - surface density of states in the diffusive band on the order parameters for a typical set of material parameters and for @xmath198 . in fig .
[ order2](a ) we show the order parameter magnitudes and in ( b ) their ratio .
it can be seen in fig .
[ order2](a ) that the bulk gap over @xmath5 ratios increase with increasing density of states ratio @xmath102 , a behavior consistent with fig .
[ gaptc](b ) and with eqs . and .
from fig .
[ order2](a ) we see that a larger @xmath102 also enlarges the core area in both bands . at higher temperature (
not shown ) , @xmath102 has less effect on the order parameters .
the effects of the @xmath3-band coulomb interaction described in the last section manifest more drastically in both bands for larger @xmath102 . when comparing the shapes of the order parameter profiles in the @xmath43 band and the @xmath135 band in fig .
[ order2](a ) , there is a qualitative difference between the cases @xmath207 and @xmath208 .
nevertheless , fig .
[ order2](b ) shows that despite the different shapes of the order parameter profiles in the two bands for different values of @xmath102 , the spatial variation of the _ ratio _ @xmath194 is almost independent of @xmath102 .
this is for @xmath209 a nontrivial and rather interesting finding .
band @xmath201 as a function of coordinate @xmath195 along the path @xmath196 for various values of @xmath95 and @xmath171 , for @xmath198 and @xmath170 ; for ( a ) @xmath108 and ( b ) @xmath210 . for @xmath110
, @xmath201 is depleted around the vortex core for larger @xmath171 , and this effect is larger for larger @xmath95 .
when the coulomb repulsion dominates ( @xmath109 ) , for small @xmath95 ( e.g. , 0.1 ) , @xmath201 is enhanced near the vortex centre for larger @xmath171 .
, title="fig : " ] band @xmath201 as a function of coordinate @xmath195 along the path @xmath196 for various values of @xmath95 and @xmath171 , for @xmath198 and @xmath170 ; for ( a ) @xmath108 and ( b ) @xmath210 . for @xmath110 , @xmath201
is depleted around the vortex core for larger @xmath171 , and this effect is larger for larger @xmath95 . when the coulomb repulsion dominates ( @xmath109 ) , for small @xmath95 ( e.g. , 0.1 ) , @xmath201 is enhanced near the vortex centre for larger @xmath171 . ,
title="fig : " ] in fig .
[ order3b ] we illustrate how the order parameter profile in the @xmath3 band changes as @xmath95 and @xmath171 are varied .
we compare the two cases of ( a ) @xmath108 and ( b ) @xmath210 , leaving @xmath198 and @xmath211 fixed . as can be seen in fig .
[ order3b](b ) , for @xmath110 , @xmath201 is depleted in the core area as @xmath171 increases , and this effect is larger for larger @xmath95 . on the other hand , when the coulomb repulsion is dominant in the @xmath3 band ( @xmath125 ) , as shown in fig .
[ order3b](a ) , the profile of @xmath201 changes in a peculiar way as @xmath171 varies .
for relatively small @xmath95 [ e.g. , 0.1 in fig .
[ order3b](a ) ] , @xmath201 is enhanced around the vortex core for larger @xmath171 , while for @xmath132 @xmath201 hardly changes for different values of @xmath171 . for larger
@xmath212 ( not shown ) , however , the core region becomes larger with increasing @xmath171 in a similar way as for @xmath110 , though not so drastically .
this lets us conclude that for the case @xmath213 in fig .
[ order3b](a ) a cancellation between two opposite tendencies is at work , that renders @xmath201 seemingly insensitive to the parameter @xmath171 . as to @xmath193 , similarly as @xmath201 in fig .
[ order3b](b ) , the core area is enlarged for larger @xmath171 , more noticeably for larger @xmath95 , and for both positive and negative @xmath100 [ see fig . [ order4](b ) below ] . as a function of coordinate @xmath195 along the path @xmath196 for various values of @xmath95 and @xmath171 , for @xmath198 and @xmath170 ; for ( a ) @xmath108 and ( b ) @xmath210 . for negative ( positive ) @xmath100 ,
the ratio is larger ( smaller ) than @xmath95 around the vortex core , and the area of the enhancement ( depletion ) is larger for larger @xmath171 . as @xmath95 increases , the ratio changes more drastically as @xmath195 approaches zero . ] as discussed above , even though superconductivity is induced in the @xmath3 band due to coupling with the @xmath1 band , the characteristic length scale of the order parameter can be quite different in the two bands .
this is demonstrated in fig .
[ order3a ] for various sets of @xmath95 and @xmath171 . here
the order parameter ratio @xmath194 is plotted as a function of @xmath195 . in agreement with our discussion of fig .
[ order1](b ) , this ratio is in the vortex core larger than @xmath95 for negative , and smaller than @xmath95 for positive @xmath100 . when @xmath109 , despite the peculiar change of the @xmath201 profile as a function of @xmath171 depending on the value of @xmath95 , @xmath194 is always larger than @xmath95 in the core area , as can be seen in fig .
[ order3a](a ) .
the main conclusion of fig .
[ order3a ] is that the enhancement for @xmath109 and depletion for @xmath128 of @xmath194 is increasing with @xmath171 and @xmath95 . and @xmath193 as a function of coordinate @xmath195 along the path @xmath196 for @xmath198 , @xmath108 , @xmath170 ; for ( a ) various values of @xmath95 for @xmath197 and ( b ) different values of @xmath171 for @xmath132 . for a fixed @xmath171 , as @xmath95 increases , the core area is enhanced in both bands . for a fixed @xmath95 , for larger @xmath171 the core area
is enlarged in the @xmath1 band .
, title="fig : " ] and @xmath193 as a function of coordinate @xmath195 along the path @xmath196 for @xmath198 , @xmath108 , @xmath170 ; for ( a ) various values of @xmath95 for @xmath197 and ( b ) different values of @xmath171 for @xmath132 . for a fixed @xmath171 , as @xmath95 increases , the core area is enhanced in both bands . for a fixed @xmath95 , for larger @xmath171 the core area
is enlarged in the @xmath1 band .
, title="fig : " ] in fig .
[ order4 ] we compare the influence of the material parameters @xmath105 and @xmath171 on the order parameter profiles in the vortex core . in fig .
[ order4](a ) we vary @xmath95 for fixed @xmath197 , and in ( b ) we vary @xmath171 for fixed @xmath132 . as shown in fig .
[ gaptc ] , in general the bulk gap to @xmath5 ratio in the @xmath1 band changes in a nonmonotonous way as a function of @xmath95 . in the case of the parameter set for fig .
[ order4](a ) , however , the bulk @xmath103 increases with increasing @xmath95 for the values of @xmath95 shown ( for a fixed @xmath171 ) . at the same time , as the coupling of the two bands becomes stronger , the recovery length of @xmath214 increases [ fig .
[ order4](a ) ] and the length scales in the two bands are more different [ see also fig .
[ order3a](a ) ] . for a fixed @xmath95 the @xmath1-band order parameter
has a larger core area as @xmath171 increases [ fig .
[ order4](b ) ] .
the @xmath214 profile changes in a similar way with varying @xmath171 when @xmath110 . also seen in fig .
[ order4](b ) is that the insensitivity of the @xmath215 profile due to the cancellation effect for @xmath216 discussed above , combined with the enhanced depletion of @xmath214 , yields a seemingly paradoxical picture , namely that an increase in @xmath183 ( for fixed @xmath217 ) results in a larger core area in the @xmath43 band instead of the @xmath135 band . as a function of position @xmath195 along the path @xmath196 for @xmath197 and @xmath170 , for various values of @xmath95 and @xmath218 ; for ( a ) @xmath108 and ( b ) @xmath210 .
the difference between the length scales in the two bands and its temperature dependence are enhanced as @xmath95 increases , especially for @xmath109 .
as temperature is raised , the difference between the recovery lengths of the two order parameters becomes smaller . ]
the ratio @xmath194 as a function of coordinate @xmath195 is plotted in fig . [ order4a ] for various values of @xmath95 and @xmath218 , for ( a ) @xmath108 and ( b ) @xmath210 . the enhancement ( depletion ) and its temperature dependence of this ratio in the core area are larger for larger @xmath95 for negative ( positive ) @xmath100 .
this effect is considerable when the coulomb repulsion dominates in the @xmath3 band [ fig .
[ order4a](a ) ] .
as temperature becomes higher , however , the difference between the recovery lengths of the two order parameters is reduced , as can be seen in fig .
[ order4a ] ( see also fig . [ order5 ] ) . and ( b ) @xmath201 as a function of @xmath195 along the path @xmath196 for @xmath132 , @xmath197 , @xmath170 , @xmath108 .
the ` weak ' @xmath3 band makes the temperature dependence of the @xmath214 profile more drastic than in the single - band case . ]
figure [ order5 ] shows the temperature dependence of ( a ) @xmath193 and ( b ) @xmath201 for a typical set of material parameters . by coupling to the weak @xmath3 band ,
the suppression of @xmath214 and the widening of its core region with increasing temperature are more drastic compared with the single - band case .
the @xmath3-band order parameter is less affected by temperature for low enough @xmath218 .
for higher temperature , however , the profile of the order parameters in the two bands are similar .
the temperature dependence of the core area of the order parameter can be characterised by the vortex ` core size ' defined by@xcite @xmath219 where @xmath220 is the radial coordinate measured from the vortex centre . in a clean single - band superconductor , the order parameter exhibits the kramer - pesch ( kp ) effect,@xcite i.e. , shrinkage of the vortex core as @xmath115 is lowered , approaching zero in the zeo - temperature limit @xcite . in an @xmath221-wave superconductor , however , the core size as defined above saturates as temperature approaches zero , when there are nonmagnetic impurities.@xcite this can be understood by the broadening of the vortex core bound states that carry the supercurrent around the vortex centre .
this broadening removes the singular behaviour in the spatial variation of the order parameter in the vortex centre , and the vortex core shrinking ceases when @xmath222 becomes smaller than the energy width of the zero - energy bound states situated at the vortex centre ( in quasiclassical approximation one can neglect the small splitting of these bound states due to the caroli - degennes minigap ) .
recently , our model of coupled clean and dirty bands@xcite has been applied to study the kp effect in a two - band superconductor , and it has been shown that , by coupling to the clean band , the kp effect is induced in the dirty band.@xcite the calculation in ref
. corresponds to positive @xmath100 in our formulation .
defined by eq .
( [ coresize ] ) in the two bands for @xmath132 , @xmath197 , and @xmath170 , for ( a ) @xmath210 and ( b ) @xmath108 .
the kramer - pesch effect is induced in the @xmath3 band , although absent in a single diffusive band ( dashed - dotted curves ) . as seen from ( a )
, the core radius @xmath223 is larger in the @xmath135 band than that in the @xmath1 band when @xmath110 .
however , the opposite is true for @xmath109 , as shown in ( b ) . ] in fig .
[ order6 ] we plot @xmath224 in the @xmath1 and @xmath3 bands as a function of temperature for @xmath132 , @xmath197 , and @xmath170 , for ( a ) @xmath210 and ( b ) @xmath108 .
points are data obtained by self - consistent calculation , and curves are guides to the eye .
single - band results are also shown for clean ( dotted curves ) and dirty ( dashed - dotted curves ) superconductors , which exhibit the core shrinkage and saturation as @xmath225 , respectively .
in the two - band case , the core size in the dirty @xmath135 band ( dashed curves ) tends to zero in the zero - temperature limit , as the core size in the clean @xmath43 band ( full curves ) does . for @xmath110 , shown in fig .
[ order6](a ) , the core size in the @xmath135 band is always larger than the core size in the @xmath43 band .
there is the kp effect in the @xmath3 band also when the coulomb repulsion is dominant [ see fig . [ order6 ] ( b ) ] .
however , here the core size defined by eq . ( [ coresize ] ) is always smaller in the @xmath3 band than in the @xmath1 band .
surprisingly , with dominating coulomb interactions , the kp effect is better developed rather in the diffusive band than in the ballistic band . only at very low temperature the core size extrapolates to zero in the @xmath43 band , whereas this happens in the @xmath135 band already below @xmath226 .
we suggest that the kp effect can be used to study the strength of coulomb interactions in the effective pair interaction matrix . finally , when @xmath137 ( not shown ) , the order parameters in the two bands have the same length scale and the same @xmath223 .
-band ldos @xmath227 as a function of energy @xmath12 at various distances @xmath220 from the vortex centre ( @xmath228 from 0 to 40 with an increment of 5 ) .
the parameter values are @xmath229 , @xmath132 , @xmath198 , @xmath211 , and @xmath137 .
the @xmath227 shows no sign of localized states in the vortex core , consistent with the experiment on mgb@xmath0 ( ref .
@xcite ) . ] in fig .
[ usadel1 ] we show an example for the ldos @xmath227 as obtained from eq . in the diffusive @xmath3 band as a function of energy @xmath12 at various distances @xmath220 from the vortex centre .
at the vortex centre , @xmath230 as a function of @xmath12 is completely flat , showing no sign of localized states .
this is consistent with the experiment on mgb@xmath0,@xcite which probed the @xmath3-band ldos by tunneling along the @xmath6 axis . far away from the vortex core
, the bulk bcs density of states is recovered . for a fixed @xmath95 , the bcs density of states
is recovered at larger distances for larger @xmath171 ; while for a fixed @xmath171 , the ldos recovers to the bulk density of states further away for smaller @xmath95 .
experimentally the vortex ` core size ' obtained from tunneling spectroscopy can be defined as a measure of the decay of the zero - bias ldos as one moves away from the vortex centre .
note that this definition of the vortex core size is very different from that defined in eq .
( [ coresize ] ) and shown in fig . [ order6 ] : the two quantities in fact differ by a large amount as we discuss in the following .
figure [ usadel2 ] shows the zero - bias ldos in the @xmath3 band @xmath231 as a function of @xmath220 for ( a ) varying @xmath232 , ( b ) varying @xmath105 , ( c ) varying @xmath76 , and ( d ) varying @xmath218 ; for the combination of the remaining parameters @xmath211 , @xmath197 , @xmath132 , @xmath198 , @xmath108 . for a fixed @xmath95 , the core size defined as a decay length of the zero - bias ldos increases substantially with increasing @xmath171 , as shown in fig .
[ usadel2](a ) . on the other hand , for a fixed @xmath171 ,
the core size becomes considerably larger for weaker coupling to the @xmath1 band , i.e. , smaller @xmath95 , as presented in fig .
[ usadel2](b ) .
it can be seen in fig .
[ usadel2](c ) that the ldos profile in the @xmath3 band is barely changed for different values of @xmath100 .
the fermi - surface density of states @xmath233 also has little effect on the @xmath3-band ldos ( not shown ) . as illustrated in fig .
[ usadel2](d ) , the ldos is hardly affected by temperature over a rather large temperature range ; however , the decay length increases notably for temperatures close to @xmath5 .
[ usadel2 ] is a central result of our calculations , as it can be used directly as reference for comparison with scanning tunneling experiments , in order to determine either of the material parameters @xmath232 or @xmath105 . in particular ,
the parameter @xmath232 is otherwise difficult to access experimentally . as a function of distance @xmath220 from the vortex centre for @xmath211 ; ( a ) @xmath132 , @xmath198 , @xmath108 for various values of @xmath171 ; ( b ) @xmath197 , @xmath198 , @xmath108 for various values of @xmath95 ; ( c ) @xmath132 , @xmath197 , @xmath198 for various values of @xmath100 ; ( d ) @xmath132 , @xmath197 , @xmath108 for various temperatures . by comparing with the experimental data ( ref . )
we estimate @xmath234 for mgb@xmath0 , taking @xmath132 . ]
the decay length of the ldos can be very different in the two bands .
the length scale of the zero - bias ldos in the @xmath1 band is @xmath182 and thus shorter than that in the @xmath3 band for small enough @xmath95 or large enough @xmath232 .
the existence of two coherence length scales in mgb@xmath0 has been suggested ; @xcite i.e. , a ` core size ' ( @xmath750 nm ) much larger than that expected from @xmath8 ( @xmath710 nm ) .
theoretically , two apparent length scales in the ldos have been found also in the case of two clean bands@xcite and two dirty bands.@xcite the difference in the decay lengths comes from the fact that , although superconductivity is induced in the @xmath135 band , quasiparticle motion is governed by the effective coherence length resulting from the usadel equation ( [ usadeleq ] ) , @xmath235 ( that is longer than @xmath172 ) .
our model of a clean and a dirty band is suitable for describing mgb@xmath0 , and the values of @xmath95 , @xmath233 and @xmath100 for fig .
[ usadel2](a ) are appropriate for modelling mgb@xmath0 . using the _ ab initio _
value@xcite of @xmath236 and including the renormalization factor as mentioned above , @xmath237 nm for mgb@xmath0 ( @xmath238 k ) . by comparing fig .
[ usadel2](a ) with the zero - bias ldos measured in the experiment of ref . [ fig .
3(a ) in ref . ] , for this parameter set , we find @xmath171 is about 2 for mgb@xmath0 . as discussed in detail in the next section , there can be bound states near the gap edge in the ` strong ' ballistic @xmath1 band , which arise from coupling to the ` weak ' diffusive @xmath3 band .
such bound states exist for parameter values relevant for mgb@xmath0 , e.g. , for @xmath239 , @xmath132 , and for various temperatures , @xmath233 , and @xmath100 [ see , e.g. , fig . [ edge1 ] ] .
-band ldos @xmath240 as a function of energy @xmath12 at various distances @xmath220 from the vortex centre ( @xmath228 from 0 to 10 with an increment of 1 ) .
the parameter values are @xmath197 , @xmath132 , @xmath198 , @xmath211 , and @xmath108 . coupling to a diffusive band results in bound states at the gap edge in the clean band , in addition to the well - known caroli - de gennes - matricon bound states . ] as a function of energy @xmath12 at various distances @xmath220 from the vortex centre ( @xmath228 from 0 to 10 with an increment of 1 ) for @xmath198 ; ( a ) the single - band case , ( b ) @xmath241 , @xmath242 , ( c ) @xmath239 , @xmath132 , and ( d ) @xmath241 , @xmath132 . in the two - band cases , @xmath211 and @xmath108 .
the gap - edge bound states , which are absent in the single - band case , are enhanced ( and the number of branches increases ) for larger @xmath171 and @xmath95 . ]
we turn now to the spectral properties of the vortex core originating from the clean @xmath43 band .
we plot in fig .
[ ldose1 ] the total ( angle - averaged ) ldos @xmath243 in the ballistic @xmath1 band as a function of energy @xmath12 at various distances @xmath220 from the vortex centre .
the spectra show the well - known caroli - de gennes - matricon ( cdgm ) bound - state bands@xcite at low energies . in the single - band case
[ see fig .
[ edge1](a ) ] , at the gap edge the spectrum is suppressed and there is neither coherence peak nor bound state in the vortex centre spectrum.@xcite the new feature found in our model is the additional bound states at the gap edge , as can be seen clearly in fig .
[ ldose1 ] , which arise from coupling to a diffusive band .
while the cdgm bound states disperse strongly as a function of position , the extra bound states stay near the gap edge as @xmath220 is varied .
the self - consistency of the order parameters is essential for studying the presence or absence of these gap - edge bound states . in fig .
[ edge1 ] the total ldos near the gap edge at various distances @xmath220 is shown at temperature @xmath198 , for ( a ) the single - band case , ( b ) @xmath241 , @xmath242 , ( c ) @xmath239 , @xmath132 , and ( d ) @xmath241 , @xmath132 .
it can be seen in fig .
[ edge1](a ) that in the single - band case , there is no localized state near the gap edge at @xmath244 , while a small peak develops at the gap edge at large @xmath220 .
this is a bound state arising solely from the phase winding around the vortex . in fig .
[ edge1](b ) , additional bound states exist at the gap edge both in the core region and far away from the vortex centre . for a fixed @xmath171 , with increasing @xmath95 , gap - edge bound states are enhanced and more bound states can be created [ figs . [
edge1](b ) and [ edge1](d ) ] . in the case of fig .
[ edge1](d ) , there are two bound states close to the gap edge in the @xmath244 spectrum . for nonzero @xmath220 ,
all bound states broaden into bands due to dispersion of the bound states with momentum direction , exhibiting the typical @xmath245-like behaviour at the band edges .
this can be seen clearly in fig .
[ ldose1 ] for the main bound state crossing zero energy at @xmath244 .
the additional gap - edge bound states also show this behaviour , as can be seen , e.g. , in fig .
[ edge1](d ) for the main gap - edge bound state at small @xmath220 . for
a fixed @xmath95 , ( more ) bound states can exist for larger @xmath171 [ figs .
[ edge1](c ) and [ edge1](d ) ] . in the case of fig .
[ edge1](c ) , gap - edge bound states exist at large @xmath220 , but not in the core region .
however , in this case a coherence peak can be observed at the gap edge at small @xmath220 .
-band at the vortex centre , @xmath246 , for energy @xmath12 close to the energy gap , for @xmath247 and @xmath248 for different values of @xmath100 ; for @xmath242 ( a , b ) and 0.5 ( c , d ) and for @xmath211 ( a , c ) and 2 ( b , d ) .
bound states exist near the gap edge for @xmath100 smaller than a certain value . ]
we now study the detailed development of bound states near the gap edge in the @xmath43 band in terms of the spectrum at the vortex centre .
we start with the sensitivity of the gap - edge bound state spectrum to @xmath76 and to @xmath249 . in fig .
[ gapedge1 ] we show the @xmath1-band total ldos @xmath246 for energy @xmath12 close to the energy gap at temperature @xmath250 . to emphasize the changes we show here examples for a relatively large coherence length ratio @xmath247 .
the panels ( a)-(d ) correspond to various fixed parameter combinations , and the curves in each panel are for different values of @xmath100 .
we show results for @xmath242 ( a , b ) and 0.5 ( c , d ) , and for and for @xmath211 ( a , c ) and 2 ( b , d ) .
the coulomb interaction in the @xmath3 band also affects the gap - edge bound states .
it can be seen in fig .
[ gapedge1](a ) that , for the given set of parameters , gap - edge bound states exist for @xmath108 , but not for @xmath203 .
an increase in @xmath233 results in an increase in the energy gap , and ( more ) gap - edge bound states due to widening of the core area [ fig .
[ gapedge1](b ) ] .
increasing @xmath95 has similar effects .
it can be seen in figs .
[ gapedge1](c , d ) that the lowest branch shifts considerably to lower energy , while the energy gap becomes larger , as @xmath100 is reduced . ,
but for fixed @xmath137 , and for ( a ) various values of @xmath171 for @xmath251 and ( b ) different values of @xmath95 for @xmath241 .
gap - edge bound states develop as @xmath95 or @xmath171 is increased . ]
next we discuss the dependence on the coherence length ratio @xmath171 and on the zero temperature gap ratio @xmath95 .
there is a threshold for both parameters , above which gap - edge bound states appear .
we illustrate this in fig .
[ gapedge2 ] for @xmath248 and @xmath137 . in figs .
[ gapedge2](a , b ) @xmath171 is changed for fixed @xmath251 .
results are for ( a ) @xmath211 and 2 ( b ) . for this gap ratio ,
gap - edge bound states exist for @xmath252 for both ratios of the density of states shown . as @xmath171 increases
, the energy gap changes only slightly . on the other hand ,
the lowest bound - state energies are reduced substantially and more branches appear .
this effect is enhanced for larger @xmath233 . for @xmath241 and @xmath251 , there are three branches for @xmath253 .
figures [ gapedge2](c , d ) show the ldos for various values of @xmath95 for fixed @xmath241 for ( c ) @xmath211 and 2 ( d ) . for this parameter set , for @xmath254
there is no gap - edge bound state for @xmath255 . as @xmath95 increases
, the energy gap is enhanced and more branches appear at the gap edge , and this effect is more significant for larger @xmath233 . , but for @xmath253 and @xmath137 , for four sets of other parameters ( see text ) and various temperatures .
the energy gap is reduced significantly as temperature increases and the energies of the gap - edge bound states are shifted together with the energy gap . ] increasing temperature @xmath115 also enhances the features of gap - edge bound states .
figure [ gapedge3 ] shows the ldos for @xmath253 and @xmath137 for various @xmath115 . in ( a ) and ( c )
we compare for @xmath197 the cases @xmath132 and @xmath251 , and in ( b ) and ( d ) we fix @xmath251 and compare the cases @xmath229 and @xmath241 . as @xmath115 increases , bound states at the gap edge shift lower together with the energy gap and increase in strength . the change from @xmath198 to 0.5 in fig .
[ gapedge3 ] is substantial for @xmath251 and @xmath241 .
the results in fig .
[ gapedge1]-[gapedge3 ] illustrate that the gap edge bound states are a robust feature of our model . as a function of energy @xmath12 , for quasiparticles moving in the @xmath195 direction at various positions along the @xmath256 axis ( from @xmath257 to @xmath258 with an increment of 3@xmath182 ) .
the parameter values are @xmath241 , @xmath259 , @xmath211 , @xmath108 , with ( a ) @xmath132 and ( b ) @xmath251 .
an increase in @xmath95 results in more gap - edge bound states and enhancement of their dispersion . ] obtained from fig .
[ angle1 ] .
there are two branches of gap - edge bound states for @xmath132 .
for @xmath251 there is a third branch in the vortex core : the lowest branch is extended over large distances and its dispersion is enhanced considerably compared with the case for @xmath132 . ] the bound - state spectrum can be discussed most clearly in terms of the angle - resolved ldos spectra , obtained from eq . .
[ angle1 ] , the angle - resolved ldos @xmath260 for quasiparticles moving with impact parameter @xmath256 in the @xmath261 direction is shown as a function of @xmath12 at various positions @xmath256 ( @xmath262 ) .
the two panels are for ( a ) @xmath132 and ( b ) @xmath251 .
the position of the bound states as a function of impact parameter @xmath256 is plotted in fig .
[ peak1 ] .
the cdgm bound - state branch disperses with angular momentum , crossing the chemical potential in the vortex centre . in fig .
[ angle1](a ) additional bound states near the gap edge can be seen , which have only weak dispersion .
a close inspection reveals that there are two branches ; for @xmath263 , the lower branch for roughly @xmath264 and the higher one for @xmath265 , and similarly for negative @xmath256 for @xmath266 .
the higher branch hardly disperses with @xmath256 [ see fig . [ peak1](a ) ] . as coupling with the @xmath3 band becomes stronger , the energy gap increases . at the same time , the energies of the bound states are lowered ( for relatively small angular momenta ) , and more bound states appear near the gap edge , due to enlargement of the core area [ fig . [ angle1](b ) ] [ see also fig .
[ peak1](b ) ] . for @xmath251 , there is a third bound - state branch at the gap edge for small @xmath267 , and the lowest branch extends over large distances , with enhanced dispersion compared to that for @xmath132 . for a fixed @xmath95 , increasing @xmath171 results in a larger core area and
has similar effects as increasing @xmath95 . and
@xmath268 , as a function of distance @xmath220 from the vortex centre , for @xmath211 and @xmath137 , for various temperatures . in the left - hand panels
, we have @xmath254 , for ( a ) @xmath269 and ( b ) 3 .
the right - hand panels have @xmath251 , for ( c ) @xmath269 and ( d ) 3 .
the @xmath3-band contribution can be substantial , or even dominating , and have strong temperature dependence far outside the vortex core for large enough @xmath95 and @xmath171 . ] for a clean single - band superconductor , supercurrents around a vortex are strongly coupled to the andreev spectrum discussed in the preceding section , and are inside the core area carried mainly by these bound states.@xcite in the following we present results for the current density calculated with eqs . for our two - band model .
[ current1 ] we show contributions from the two bands to the current density around the vortex separately , for parameters @xmath211 and @xmath137 , and for various temperatures . in the left - hand panels , we have @xmath254 , and ( a ) @xmath269 and ( b ) 3 . for this weak coupling between the two bands ,
the current contribution from the @xmath3 band is negligible for @xmath269 . for @xmath270 , though still small , at low temperature ( i.e. , @xmath198 ) the @xmath3-band contribution in the bulk ( outside the vortex core , but still well within the penetration depth distance from the core ) is almost the same as that in the @xmath1 band .
as @xmath271 , the peak of @xmath268 approaches the vortex centre this is the kp effect manifest in the supercurrent .
the current density arising from the induced superconductivity in the @xmath3 band is also enhanced by decreasing temperature , but the maximum does not approach the vortex centre [ see fig . [
current1](b ) ] . in ref .
, the same trend of the @xmath3-band current density as a function of temperature was presented for parameter values appropriate for mgb@xmath0 . as @xmath95 increases
, the @xmath3-band contribution to the current density becomes considerable . in the right - hand panels of fig .
[ current1 ] , the current density contributions are plotted for @xmath251 , for ( c ) @xmath269 and ( d ) 3 . interestingly , as the coupling between the two bands is increased , the @xmath3 band starts exhibiting the kp effect in the current density .
the maximum shifts towards the vortex centre as temperature is decreased , although the ` core size ' as defined by the position of the current maximum remains larger than that in the @xmath1 band [ see figs . [ current1](c ) and ( d ) ] . as can be seen in fig .
[ current1](d ) , for relatively large @xmath95 and @xmath272 , the @xmath3-band contribution to the current density becomes substantial .
especially away from the vortex core , where the contribution of the @xmath43 band is reduced , the @xmath3-band contribution can be dominant .
another important observation is that the current density in the @xmath3 band has strong temperature dependence also far outside the core . on the contrary ,
the @xmath1-band current density is temperature - dependent only in the core area . far away from the vortex core
, we can use in eq .
( [ cd ] ) the analytical solutions to the eilenberger and usadel equations for a quasihomogeneous system , with constant magnitude and constant phase gradient of the order parameters .
we thus obtain the current - density magnitudes for a large distance @xmath220 from the vortex centre ( but small compared to the london penetration depth ) as @xmath273 clearly , the temperature dependence of @xmath274 is dominated by that of @xmath275 .
this bulk behaviour of the current density in the two bands also explains the dominance of the @xmath3-band contribution in the bulk for relatively large @xmath95 and @xmath171 . for large @xmath220 , the current - density ratio approaches
@xmath276 thus the @xmath3-band contribution to the current density dominates when @xmath277 .
this is relevant for mgb@xmath0 , for which this condition roughly reads as @xmath278 .
in conclusion , we have studied a model recently introduced by us@xcite for describing a multiband superconductor with a ballistic and a diffusive band in terms of coupled eilenberger and usadel equations .
both equations were solved directly and numerically until self - consistency of the order parameters was achieved .
we have studied the effects of induced superconductivity and impurities in the weak diffusive ( @xmath3 ) band on the order parameter , the current density , and the spectral properties of the strong ballistic ( @xmath1 ) band .
a unique feature found in our model is the existence of additional bound states at the gap edge in the ballistic band , which are absent when there is no coupling with the diffusive band .
although the two bands are coupled by the pairing interaction , the order parameters in the two bands can have very different length scales .
when the coulomb interaction dominates in the @xmath3 band , the order parameter in the @xmath1 band has a longer recovery length than that in the @xmath3 band . in this case
, the @xmath3 band exhibits the kramer - pesch effect with a ` core size ' [ defined by eq .
( [ coresize ] ) ] smaller than that in the @xmath1 band . furthermore ,
when coupled to the diffusive band , the order parameter in the ballistic @xmath1 band is suppressed by temperature more strongly than in the single - band case .
the zero - bias ldos in the two bands can have very different decay lengths , the one in the @xmath3 band being larger than that in the @xmath1 band . as the induced
superconductivity in the @xmath3 band becomes stronger , the core area widens in the @xmath1 band and ( more ) gap - edge bound states appear .
moreover , an increase in the fermi - surface density of states in the @xmath3 band results in an increase in the energy gap and the number of gap - edge bound states in the @xmath1 band . increasing the coulomb repulsion in the @xmath3 band and temperature also enlarges the core area in the @xmath1 band , which results in additional bound states at the gap edge .
the gap - edge bound states have only weak dispersion .
it is thus expected that these bound states are affected only weakly by impurity scattering within the @xmath1 band.@xcite results incorporating impurities in the strong @xmath1 band confirm this statement and will be presented in a future paper.@xcite the supercurrent density is dominated in the vortex core by the @xmath1-band contribution , and outside the core the @xmath3-band contribution can be substantial , or even dominating over the @xmath1-band current density .
the current density in the @xmath1 band shows the kramer - pesch effect in the core area . on the other hand ,
the @xmath3 band current density has strong temperature dependence also outside the core .
it exhibits the kramer - pesch effect when the coupling between the two bands is relatively strong .
stronger coulomb repulsion in the @xmath3 band enhances the current density in the @xmath3 band , while it has little effect on the @xmath1-band current density .
our model is suitable for describing mgb@xmath0 . for parameter values appropriate for mgb@xmath0
, we have found the above intriguing features in the ldos and the current density , including additional bound states near the gap edge in the @xmath1 band .
the gap - edge bound states should be affected only weakly by the strong electron - phonon interaction , as the energy of relevant phonons is much higher ( above 60 - 70 mev ) ( refs . ) .
our predictions on the spectral properties of the @xmath1 band can be tested by tunneling electrons onto the @xmath279 plane.@xcite
two of the authors ( k.t . and m.e .
) would like to thank morten r. eskildsen , boldizsr jank , juha kopu , and gerd schn for valuable discussions and contributions .
two of the authors ( k.t . and d.f.a .
) are grateful to g. w. crabtree and w. k. kwok for hospitality at the argonne national laboratory , where a part of this work was performed , and to m. iavarone , w. k. kwok , p. c. moca , and h. schmidt for discussions .
the research was supported by the nserc of canada , the canada foundation for innovation , the deutsche forschungsgemeinschaft within the cfn , the u.s .
doe , basic energy sciences ( contract no .
w-7405-eng-36 ) , the nsf ( grant no .
dmr-0381665 ) , and the american chemical society petroleum research fund .
s. v. shulga , s .- l .
drechsler , h. eschrig , h. rosner , w. e. pickett , arxiv : cond - mat/0103154 ( unpublished ) .
a. y. liu , i. i. mazin , j. kortus , phys .
lett . * 87 * , 087005 ( 2001 ) .
f. bouquet , y. wang , r. a. fisher , d. g. hinks , j. d. jorgensen , europhys .
lett . * 56 * , 856 ( 2001 ) .
h. schmidt , j. f. zasadzinski , k. e. gray , d. g. hinks , phys .
lett . * 88 * , 127002 ( 2002 ) .
m. iavarone , g. karapetrov , a. e. koshelev , w. k. kwok , g. w. crabtree , d. g. hinks , w. n. kang , e .-
choi , h. j. kim , h .- j .
kim , and s. i. lee , phys .
rev . lett . * 89 * , 187002 ( 2002 ) .
a. a. golubov , j. kortus , o. v. dolgov , o. jensen , y. kong , o. k. andersen , b. j. gibson , k. ahn and r. k. kremer , j. phys . : condens .
matter * 14 * , 1353 ( 2002 ) .
h. j. choi , d. roundy , h. sun , m. l. cohen , s. g. louie , phys .
b * 66 * , 020513(r ) ( 2002 ) ; nature ( london ) * 418 * , 758 ( 2002 ) . i. i. mazin , o. k. andersen , o. jepsen , a. a. golubov , o. v. dolgov , j. kortus , phys .
b * 69 * , 056501 ( 2004 )
. physica c * 385*(1 - 2 ) ( 2003 ) , special issue on mgb@xmath0 , edited by w. kwok , g. w. crabtree , s. l. budko , and p. c. canfield .
t. dahm , in _ frontiers in superconducting materials _ , edited by a.v .
narlikar ( springer , new york ) ( 2005 ) .
p. szab , p.
samuely , j. kacmarck , t. klein , j. marcus , d. fruchart , s. miraglia , c. marcenat , a. g. m. jansen , phys .
lett . * 87 * , 137005 ( 2001 ) .
r. s. gonnelli , d. daghero , g. a. ummarino , v. a. stepanov , j. jun , s. m. kazakov , j. karpinski , phys .
lett . * 89 * , 247004 ( 2002 ) .
z. holanov , p. szab , p. samuely , r. h. t. wilke , s. l. budko , and p. c. canfield , phys .
b * 70 * , 064520 ( 2004 ) . m. r. eskildsen , m. kugler , s. tanaka , j. jun , s. m. kazakov , j. karpinski , .
fischer , phys .
. lett . * 89 * , 187003 ( 2002 ) ; m. r. eskildsen , n. jenkins , g. levy , m. kugler , .
fischer , j. jun , s. m. kazakov , and j. karpinski , phys .
b * 68 * , 100508(r ) ( 2003 ) .
j. geerk , r. schneider , g. linker , a. g. zaitsev , r. heid , k .-
bohnen , and h. v. lhneysen , phys .
rev . lett . * 94 * , 227005 ( 2005 ) .
f. giubileo , f. bobba , a. scarfato , a. m. cucolo , a. kohen , d. roditchev , n. zhigadlo , j. karpinski , arxiv : cond - mat/0604354 ( unpublished ) .
tang , phys . rev .
b * 2 * , 129 ( 1970 ) .
a. a. golubov and i. i. mazin , phys .
b * 55 * , 15146 ( 1997 )
. i. i. mazin , o. k. andersen , o. jepsen , o. v. dolgov , j. kortus , a. a. golubov , a.b .
kuzmenko , and d. van der marel , phys .
* 89 * , 107002 ( 2002 ) .
s. c. erwin , i. i. mazin , phys .
b * 68 * , 132505 ( 2003 ) .
m. putti , v. braccini , e. galleani dagliano , f. napoli , i. pallecchi , a. s. siri , p. manfrinetti , a. palenzona , phys .
b * 67 * , 064505 ( 2003 ) .
m. putti , v. braccini , c. ferdeghini , i. pallecchi , a. s. siri , f. gatti , p. manfrinetti , and a. palenzona , phys . rev .
b * 70 * , 052509 ( 2004 ) .
m. putti , c. ferdeghini , m. monni , i. pallecchi , c. tarantini , p. manfrinetti , a. palenzona , d. daghero , r. s. gonnelli , and v. a. stepanov , phys .
b * 71 * , 144505 ( 2005 ) .
j. w. quilty , s. lee , s. tajima , a. yamanaka , phys .
* 90 * , 207006 ( 2003 ) .
m. ortolani , d. di castro , p. postorino , i. pallecchi , m. monni , m. putti , and p. dore , phys .
b * 71 * , 172508 ( 2005 ) .
a. carrington , j. d. fletcher , j. r. cooper , o. j. taylor , l. balicas , n. d. zhigadlo , s. m. kazakov , j. karpinski , j. p.
h. charmant , and j. kortus , phys .
b * 72 * , 060507(r ) ( 2005 ) . i. pallecchi , v. braccini , e. galleani dgliano , m. monni , a. s. siri , p. manfrinetti , a. palenzona , and m. putti , phys . rev .
b * 71 * , 104519 ( 2005 ) . p.
samuely , p. szab , z. holanov , s. budko , p. c. canfield , physica c * 435 * , 71 ( 2006 ) ; p. szab , p.
samuely , z. holanov , s. budko , p. c. canfield , and j. marcus , arxiv : cond - mat/0604342 ( unpublished ) .
y. wang , f. bouquet , i. sheikin , p. toulemonde , b. revaz , m. eisterer , h. w. weber , j. hinderer , and a. junod , j. phys .
* 15 * , 883 ( 2003 ) .
samuely , z. holanov , p. szab , j. kacmarcik , r. a. ribeiro , s. l. budko , and p. c. canfield , phys .
b * 68 * , 020505(r ) ( 2003 ) . h. schmidt , k. e. gray , d. g. hinks , j. f. zasadzinski , m. avdeev , j. d. jorgensen , and j. c. burley , phys . rev .
b * 68 * , 060508(r ) ( 2003 ) . m. putti , m. affronte , p. manfrinetti , and a. palenzona , phys . rev .
b * 68 * , 094514 ( 2003 ) .
r. a. ribeiro , s. l. budko , c. petrovic , p. c. canfield , physica c * 384 * , 227 ( 2003 ) .
m. angst , s. l. budko , r. h. t. wilke , and p. c. canfield , phys .
b * 71 * , 144512 ( 2005 ) .
s. tsuda , t. yokoya , t. kiss , t. shimojima , s. shin , t. togashi , s. watanabe , c. zhang , c. t. chen , s. lee , h. uchiyama , s. tajima , n. nakai , and k. machida , phys . rev .
b * 72 * , 064527 ( 2005 ) .
v. braccini _ et al .
_ , phys . rev .
b * 71 * , 012504 ( 2005 ) . p.
samuely , p. szab , p. c. canfield , s. l. budko , phys .
lett . * 95 * , 099701 ( 2005 ) .
t. klein , l. lyard , j. marcus , c. marcenat , p. szab , z. holanov , p.
samuely , b. w. kang , h - j .
kim , h - s .
lee , h - k .
lee , and s - i .
lee , phys .
b * 73 * , 224528 ( 2006 ) .
r. s. gonnelli , d. daghero , a. calzolari , g. a. ummarino , v. dellarocca , v. a. stepanov , s. m. kazakov , n. zhigadlo , and j. karpinski , phys .
b * 71 * , 060503(r ) ( 2005 ) . j. kortus , o. v. dolgov , r. k. kremer , and a. a. golubov , phys . rev
. lett . * 94 * , 027002 ( 2005 ) ; _ ibid .
_ , * 95 * , 099702 ( 2005 ) .
a. kohen , t. cren , th .
proslier , y. noat , w. sacks , d. roditchev , f. giubileo , f. bobba , a. m. cucolo , n. zhigadlo , s. m. kazakov , j. karpinski , app .
. lett . * 86 * , 212503 ( 2005 ) .
s. serventi , g. allodi , r. de renzi , g. guidi , l. roman , p. manfrinetti , a. palenzona , ch .
niedermayer , a. amato , ch .
baines , phys .
93 * , 217003 ( 2004 ) .
l. kramer and w. pesch , z. phys . * 269 * , 59 ( 1974 ) .
a. gumann , s. graser , t. dahm , n. schopohl , phys .
b * 73 * , 104506 ( 2006 ) .
a. i. larkin and y. n. ovchinnikov , zh .
fiz . * 55 * , 2262 ( 1968 ) , [ sov .
jetp * 28 * , 1200 ( 1969 ) ] .
g. eilenberger , z. phys . * 214*. 195 ( 1968 ) .
k. usadel , phys .
lett . * 25 * , 507 ( 1970 ) .
j. w. serene and d. rainer , phys .
* 101 * , 221 ( 1983 ) .
j. rammer and h. smith , rev .
phys . * 58 * , 323 ( 1986 ) .
a. i. larkin and y. n. ovchinnikov , in _ nonequilibrium superconductivity _
, edited by d. n. langenberg and a. i. larkin ( elsevier science , new york , 1986 ) , p. 493 . m. eschrig , j. a. sauls , h. burkhardt , and d. rainer , in _
high-@xmath5 superconductors and related materials - material science , fundamental properties , and some future electronic applications _ , proceedings of the nato asi , nato science partnership - sub - series 3 , vol .
86 , edited by s .-
drechsler and t. mishonov ( kluwer academic , norwell , ma , 2001 , and springer netherland , 2001 ) , pp . 413 - 446 .
y. nagato , k. nagai , and j. hara , j. low temp . phys . *
93 * , 33 ( 1993 ) , s. higashitani and k. nagai , j. phys .
64 * , 549 ( 1995 ) , y. nagato , s. higashitani , k. yamada , and k. nagai , j. low temp . phys .
* 103 * , 1 ( 1996 )
. n. schopohl and k. maki , phys .
b * 52 * , 490 ( 1995 ) ; n. schopohl , arxiv : cond - mat/9804064 ( unpublished ) .
m. eschrig , j. a. sauls , and d. rainer , phys .
b * 60 * , 10447 ( 1999 ) .
m. eschrig , phys .
b * 61 * , 9061 ( 2000 ) .
m. eschrig , j. kopu , a. konstandin , j.c .
cuevas , m. fogelstrm , and g. schn , adv . in sol .
state phys .
* 44 * , 533 , springer ( 2004 ) .
a. junod , y. wang , f. bouquet , and p. toulemonde , _ studies of high temperature superconductors _ , edited by a. v. narlikar ( nova science , commack , ny , 2002 ) vol .
* 38 * , p. 179 .
a. v. sologubenko , j. jun , s. m. kazakov , j. karpinski , and h. r. ott , phys .
b * 66 * , 014504 ( 2002 ) . e. j. nicol and j. p. carbotte , phys . rev .
b * 71 * , 054501 ( 2005 ) ; e. j. nicol , j. p. carbotte , d. j. scalapino , @xmath280 * 73 * , 014521 ( 2006 ) .
o. v. dolgov , r. k. kremer , j. kortus , a. a. golubov , s. v. shulga , phys .
b * 72 * , 024504 ( 2005 ) .
a. brinkman , a. a. golubov , h. rogalla , o. v. dolgov , j. kortus , y. kong , o. jepsen , and o. k. andersen , phys .
b * 65 * , 180517(r ) ( 2002 ) , table 1 . for the @xmath135 band
the quantity entering the diffusion constant is @xmath281/2 $ ] . for the @xmath43 band within our model
the quantity @xmath282 is relevant .
y. kato and n. hayashi , j. phys .
jpn * 70 * , 3368 ( 2001 ) .
m. ichioka , n. hayashi , n. enomoto , and k. machida , phys .
b * 53 * , 15316 ( 1996 ) .
n. hayashi , y. kato , and m. sigrist , j. low temp .
phys . * 139 * , 79 ( 2005 ) . in the work of ref .
, the diffusion constant in the @xmath3 band was assumed to be larger than that in the @xmath1 band , for the @xmath3 band to have a longer decay length of the zero - bias ldos .
experimentally it has been found that @xmath283 in dirty samples ( refs . and ) .
a. gurevich _
et al . _ ,
* 17 * , 278 ( 2004 ) .
m. eschrig , d. rainer , and j.a .
sauls , in _ vortices in unconventional superconductors and superfluids _ , edited by r.p .
huebener , n. schopohl , and g.e .
volovik , springer series in solid - state sciences ( springer , new york , 2002 ) , pp .
175 - 202 .
g. naidyuk , i. k. yanson , o. e. kvitnitskaya , s. lee and s. tajima , phys .
* 90 * , 197001 ( 2003 ) .
i. k. yanson , s. i. beloborodko , yu .
g. naidyuk , o. v. dolgov , a. a. golubov , phys .
b * 69 * , 100501(r ) ( 2004 ) .
m. iavarone , r. di capua , a. e. koshelev , w. k. kwok , f. chiarella , r. vaglio , w. n. kang , e. m. choi , h. j. kim , s. i. lee , a. v. pogrebnyakov , j. m. redwing , x. x. xi , phys .
b * 71 * , 214502 ( 2005 ) . | we study the electronic structure in the vicinity of a vortex in a two - band superconductor in which the quasiparticle motion is ballistic in one band and diffusive in the other .
this study is based on a model appropriate for such a case , that we have introduced recently [ tanaka _ et al .
_ , phys .
rev .
b * 73 * , 220501(r ) ( 2006 ) ] .
we argue that in the two - band superconductor mgb@xmath0 , such a case is realised .
motivated by the experimental findings on mgb@xmath0 , we assume that superconductivity in the diffusive band is `` weak , '' i.e. , mostly induced .
we examine intriguing features of the order parameter , the current density , and the vortex core spectrum in the `` strong '' ballistic band under the influence of hybridization with the `` weak '' diffusive band .
although the order parameter in the diffusive band is induced , the characteristic length scales in the two bands differ due to coulomb interactions .
the current density in the vortex core is dominated by the contribution from the ballistic band , while outside the core the contribution from the diffusive band can be substantial , or even dominating .
the current density in the diffusive band has strong temperature dependence , exhibiting the kramer - pesch effect when hybridization is strong .
a particularly interesting feature of our model is the possibility of additional bound states near the gap edge in the ballistic band , that are prominent in the vortex centre spectra .
this contrasts with the single band case , where there is no gap - edge bound state in the vortex centre .
we find the above - mentioned unique features for parameter values relevant for mgb@xmath0 . |
in this paper , we are interested in the numerical approximation of a randomly - perturbed system of reaction - diffusion equations that can be written @xmath0 for @xmath1 , with initial conditions @xmath2 and @xmath3 , and homogeneous dirichlet boundary conditions @xmath4 .
the stochastic perturbation @xmath5 is a space - time white noise and @xmath6 is a small parameter . in the recent article @xcite
, we have proved that an averaging principle holds for such a system , and we have exhibited an order of convergence - with respect to @xmath7 - in a strong and in a weak sense : the slow component @xmath8 is approximated thanks to the solution of an averaged equation . in this article , we analyse a numerical method of time discretization which reproduces this averaging effect at the discrete time level .
more precisely , our aim is to build a numerical approximation of the slow component @xmath8 , taking care of the stiffness induced by the time scale separation .
the heterogeneous multiscale method - hmm - procedure can be used , as it is done in @xcite for sdes of the same kind .
first we recall the general principle of such a method , which has been developped in various contexts , both deterministic and stochastic - see the review article @xcite and the references therein , as well as @xcite , @xcite , @xcite . in system ,
the two components evolve at different time scales ; @xmath8 is the slow component of the problem , while the fast component @xmath9 has fast variations in time .
we are indeed interested in evaluating the slow component , which can be thought as the mathematical model for a phenomenon appearing at the natural time scale of the experiment , whereas the fast component can often be interpreted as data for the slow component , taking care of effects at a faster time scale .
instead of using a direct numerical method , which might require a very small time step size because of the fast component , we use a different solver for each time scale : a macrosolver and a microsolver .
the macrosolver leads to the approximation of the slow component ; it takes into account data from the evolution at the fast time scale .
the microsolver is then a procedure for estimating the unknown necessary data , using the evolution at the microtime scale , which also depends on the evolution at the macrotime scale .
we emphasize a difference with the framework of @xcite : instead of analysing various numerical schemes , the infinite dimensional setting implies that we only focus on a semi - implicit euler scheme . in section [ secttheo ] , we state the two main theorems of this article : we show a strong convergence result - theorem [ cvforte ] - as well as a weak convergence result - theorem [ cvfaible ] - which are similar to the available results for sdes . compared to @xcite , we propose modified and simplified proofs leading to apparently weaker error estimates ; we made this choice for various reasons .
first , even if apparently we get weaker estimates , under an appropriate choice of the parameters the cost of the method remains of the same order .
second , the generalization of the final dimensional results would not yield the same bounds , due to the regularity assumptions we make on the nonlinear coefficients of the equations .
finally , we can extend the weak convergence result to the situation where the fast equation only satisfies a weak dissipativity assumption . in the case of a linear fast equation - when @xmath10 is equal to @xmath11 - it is well - known that the second equation in is dissipative . in the general case , we make assumptions on @xmath10 so that this property is preserved for @xmath9 - see assumptions [ strictdiss ] and [ weakdiss ] below .
the fast equation with frozen slow component - defined by in the abstract framework - then admits a unique invariant probability measure , which is ergodic and strongly mixing - with exponential convergence to equilibrium . under the strict dissipativity condition [ strictdiss ]
, we can prove that the averaging principle holds in the strong and in the weak sense ; moreover the `` fast '' numerical scheme has the same asymptotic behaviour as the continuous time equation .
if we only assume weak dissipativity of assumption [ weakdiss ] , the averaging principle only holds in the weak sense , and we can not prove uniqueness of the invariant law of the fast numerical scheme .
nevertheless , in the general setting @xcite gives an approximation result of the invariant law of the continuous time equation with the numerical method which is used to prove theorem [ cvfaible ] ; the order of convergence is @xmath12 with respect to the timestep size - the precise result is recalled in theorem [ weakdistance ] . the paper is organized as follows . in section [ sectdescri ] ,
we give the definition of the numerical scheme .
we then state the main assumptions made on the system of equations . in section [ secttheo ]
we state the two main theorems proved in this article , while in section [ commsect ] we compare the efficiency of the hmm scheme with a direct one in order to justify the use of a new method . before proving the theorems ,
we give some useful results on the numerical schemes .
finally the last two sections contain the proof of the strong and weak convergence theorems .
instead of working directly with system , we work with abstract stochastic evolution equations in hilbert spaces @xmath13 : @xmath14 with initial conditions @xmath15 , @xmath16 . to get system from , we take @xmath17 ; the linear operators @xmath18 and @xmath19 are laplace operators with homogeneous dirichlet boundary conditions - see example [ exampleab ] - and the nonlinearities @xmath20 and @xmath21 are nemytskii operators - see example [ exfg ] . the process @xmath22 is a cylindrical wiener process on @xmath13 - see section [ sectwiener ] . for precise assumptions on the coefficients ,
we refer to section [ sectassumptions ] .
we recall the idea of the averaging principle - proved in the previous article @xcite : when @xmath7 goes to @xmath11 , @xmath23 can be approximated by @xmath24 defined by the averaged equation @xmath25 the error is controlled in a strong sense by @xmath26 and in a weak sense by @xmath27 , where @xmath28 can be chosen as small as necessary , and where @xmath29 is a constant .
the averaged coefficient @xmath30 - see - satisfies @xmath31,\ ] ] where @xmath32 is the unique invariant probability measure of the fast process @xmath33 with frozen slow component - more details are given in section [ known ] .
to apply the hmm strategy , we need to define a macrosolver and a microsolver .
we denote by @xmath34 the macrotime step size , and by @xmath35 the microtime step size .
let also @xmath36 be a given final time .
the construction of the macrosolver is deeply based on the averaging principle : for @xmath37 @xmath38 can be approximated by @xmath39 . if the averaged coefficient @xmath30 was known , one could build an approximation with a deterministic numerical scheme on the averaged equation ; nevertheless in general it is not the case , and the idea is to calculate an approximation of this coefficient on - the - fly , by using the microsolver .
therefore the macrosolver is defined in the following way : for any @xmath40 , @xmath41 with the initial condition @xmath42 .
@xmath43 has to be defined ; before that , we notice that the above definition leads to a semi - implicit euler scheme - we use implicitness on the linear part , but the nonlinear part is explicit .
if we define a bounded linear operator on @xmath13 by @xmath44 , we rather use the following explicit formula @xmath45 we want @xmath43 to be an approximation of @xmath46 . the role of the microsolver
is to give an approximation of @xmath47 - the fast process with frozen slow component @xmath48 , when @xmath49 is fixed ; moreover we compute a finite number @xmath50 of independent replicas of the process , in order to approximate theoretical expectations by discrete averages over different realizations of the random variables , in a monte - carlo spirit .
therefore the microsolver is defined in the following way : we fix a realization index @xmath51 , and a macrotime step @xmath49 ; then for any @xmath52 @xmath53 as above we can give an explicit formula @xmath54 where for any @xmath55 @xmath56 .
the noises @xmath57 are defined by @xmath58 where @xmath59 are independent cylindrical wiener processes on @xmath13 .
it is essential to use independent noises at each macrotime step .
it is important to remark that this equation is well - posed in the hilbert space @xmath13 - see the conditions required when a cylindrical wiener process is used in section [ sectwiener ] - since @xmath60 is a hilbert - schmidt operator from @xmath13 to @xmath13 , under assumptions given in section [ sectassumptions ] .
the missing definition can now be written : @xmath43 is given by @xmath61 @xmath62 is the number of microtime steps that are not used in the evaluation of the average in @xmath43 , while @xmath63 is the number of microtime steps that are then used for this evaluation .
each macrotime step then requires the computation of @xmath64 values of the microsolver . at each macrotime step @xmath65
must be initialized at time @xmath66 . in our proofs
, this is not as important as in @xcite , but for definiteness we use the same method : we initialize with the last value computed during the previous macrotime step : @xmath67 , while @xmath68 .
the aim of the analysis for hmm schemes is to prove that under an appropriate choice of the parameters @xmath69 of the scheme , we can bound the error by expressions of the following kind , where @xmath28 is chosen as small as necessary : for @xmath70 , we have the strong error estimate @xmath71 and if @xmath72 is a test function of class @xmath73 , bounded and with bounded derivatives , we have the weak error estimate @xmath74 the origin of the three error terms appears clearly in the proofs - see sections [ sectfort ] and [ sectfaible ] : the first one is the averaging error , the second one is the error in a deterministic scheme with the macrotime step , and the third one is the weak error in a scheme for stochastic equations with the microtime step .
we recall that in the spde case the strong order of the semi - implicit euler scheme for the microsolver used here is @xmath75 , while the weak order is @xmath12 , while in the sde situation the respective orders are @xmath12 and @xmath76 .
the macrosolver is deterministic , so that the order is @xmath76 .
precise results for any choice of @xmath69 are given in theorems [ cvforte ] and [ cvfaible ] below , while the choice of these parameters is explained in section [ commsect ] .
as mentioned above , system satisfies an averaging principle , and strong and weak order of convergence can be given .
the hmm method relies on that idea .
the natural assumptions are basically the same as the hypothesis needed to prove those results , but must be strenghtened sometimes . the typical kind of coefficients is specified in examples [ exampleab ] and [ exfg ] . here ,
we recall the definition of the cylindrical wiener process and of stochastic integral on a separable hilbert space @xmath13 - its norm is denoted by @xmath77 or just @xmath78 . for more details , see @xcite .
we first fix a filtered probability space @xmath79 . a cylindrical wiener process on @xmath13
is defined with two elements : * a complete orthonormal system of @xmath13 , denoted by @xmath80 , where @xmath81 is a subset of @xmath82 ; * a family @xmath83 of independent real wiener processes with respect to the filtration @xmath84 : @xmath85 when @xmath81 is a finite set , we recover the usual definition of wiener processes in the finite dimensional space @xmath86 .
however the subject here is the study of some stochastic partial differential equations , so that in the sequel the underlying hilbert space @xmath13 is infinite dimensional ; for instance when @xmath87 , an example of complete orthonormal system is @xmath88 - see example [ exampleab ] .
a fundamental remark is that the series in does not converge in @xmath13 ; but if a linear operator @xmath89 is hilbert - schmidt , then @xmath90 converges in @xmath91 for any @xmath92 .
we recall that a linear operator @xmath89 is said to be hilbert - schmidt when @xmath93 where the definition is independent of the choice of the orthonormal basis @xmath94 of @xmath13 .
the space of hilbert - schmidt operators from @xmath13 to @xmath95 is denoted @xmath96 ; endowed with the norm @xmath97 it is an hilbert space .
the stochastic integral @xmath98 is defined in @xmath95 for predictible processes @xmath99 with values in @xmath100 such that @xmath101 a.s ; moreover when @xmath102;\mathcal{l}_2(h , k))$ ] , the following two properties hold : @xmath103 a generalization of it formula also holds - see @xcite . for instance ,
if @xmath104 , we can define @xmath105 we then have the following space - time white noise property @xmath106 therefore to be able to integrate a process with respect to @xmath107 requires some strong properties on the integrand ; in our spde setting , the hilbert - schmidt properties follow from the assumptions made on the linear coefficients of the equations .
first , we have to specify some properties of the linear operators @xmath18 and @xmath19 coming into the definition of system ; we assume that the linear parts are of parabolic type , with space variable @xmath108 .
we assume that @xmath18 and @xmath19 are unbounded linear operators , with domains @xmath109 and @xmath110 , which satisfy the following assumptions : [ hypab ] 1 . there
exist complete orthonormal systems of @xmath13 , @xmath111 and @xmath112 , and @xmath113 and @xmath114 non - decreasing sequences of real positive numbers such that : @xmath115 we use the notations @xmath116 and @xmath117 for the smallest eigenvalues of @xmath18 and @xmath19 .
2 . for every @xmath118 , @xmath119 is lipschitz continuous and bounded on @xmath120 $ ] , with a uniform control with respect to @xmath121 : there exists @xmath122 such that for any @xmath123 $ ] @xmath124 3 .
the sequences @xmath125 and @xmath126 go to @xmath127 ; moreover we have some control of the behaviour of @xmath126 given by : @xmath128 [ exampleab ] @xmath129 , with domain @xmath130 - homogeneous dirichlet boundary conditions : in that case @xmath131 , and @xmath132 - see @xcite . in the abstract setting
, powers of @xmath133 and @xmath134 , with their domains can be easily defined : [ defpowers ] for @xmath135 $ ] , we define the operators @xmath136 and @xmath137 by @xmath138 with domains @xmath139 on @xmath140 , the norm @xmath141 and the sobolev norm of @xmath142 are equivalent : when @xmath143 belongs to a space @xmath140 , the exponent @xmath144 represents some regularity of the function @xmath143 .
when @xmath145 , we can also define a bounded linear operator @xmath146 in @xmath13 with @xmath147 where @xmath148 . under the previous assumptions on the linear coefficients , it is easy to show that the following stochastic integral is well - defined in @xmath13 , for any @xmath92 :
@xmath149 it is called a stochastic convolution , and it is the unique mild solution of @xmath150 under the second condition of assumption [ hypab ] , there exists @xmath151 such that for any @xmath152 we have @xmath153 ; it can then be proved that @xmath154 has continuous trajectories - via the _ factorization method _ , see @xcite - and that for any @xmath155 , any @xmath156 , there exists a constant @xmath157 such that for any @xmath92 @xmath158 we now give the assumptions on the nonlinear coefficients @xmath159 .
first , we need some regularity properties : [ hypf ] we assume that there exists @xmath160 and a constant @xmath29 such that the following directional derivatives are well - defined and controlled : * for any @xmath161 and @xmath162 , @xmath163 and @xmath164 . * for any @xmath161 , @xmath162 , @xmath165 , @xmath166 . * for any @xmath161 , @xmath162 , @xmath167 , @xmath168 . * for any @xmath161 , @xmath162 , @xmath167 , @xmath169 . * for any @xmath161 , @xmath170 , @xmath171 , @xmath172 .
we moreover assume that @xmath20 is bounded .
we also need : [ hypf2 ] for @xmath173 defined in the previous assumption [ hypf ] , we have for any @xmath161 and @xmath174 latexmath:[\[\begin{gathered } we assume that the fast equation is a gradient system : for any @xmath143 the nonlinear coefficient @xmath176 is the derivative of some potential @xmath177 .
we also assume regularity assumptions as for @xmath20 .
[ hypg ] the function @xmath21 is defined through @xmath178 , for some potential @xmath179 .
moreover we assume that @xmath21 is bounded , and that the regularity assumptions given in the assumption [ hypf ] are also satisfied for @xmath21 . for @xmath21 , we need a stronger hypothesis than for @xmath20 - in order to get proposition [ fbarlip ] .
assumption [ hypf2 ] becomes : [ hypg2 ] we have for any @xmath161 , @xmath174 , @xmath180 @xmath181 finally , we need to assume some dissipativity of the fast equation .
assumption [ strictdiss ] is necessary to obtain strong convergence in the averaging principle , while assumption [ weakdiss ] is weaker and can lead to the weak convergence .
[ strictdiss ] let @xmath182 denote the lipschitz constant of @xmath21 with respect to its second variable ; then @xmath183 where @xmath184 is defined in assumption [ hypab ] .
[ weakdiss ] there exist @xmath185 and @xmath122 such that for any @xmath186 @xmath187 the second assumption is satisfied as soon as @xmath21 is bounded , while the first one requires some knowledge of the lipschitz constant of @xmath21 .
[ exfg ] we give some fundamental examples of nonlinearities for which the previous assumptions are satisfied : * functions @xmath159 of class @xmath73 , bounded and with bounded derivatives , such that @xmath178 and satisfying fit in the framework , with the choice @xmath188 .
* functions @xmath20 and @xmath21 can be * nemytskii * operators : let @xmath189 be a measurable function such that for almost every @xmath108 @xmath190 is twice continuously differentiable , bounded and with uniformly bounded derivatives .
then @xmath20 is defined for every @xmath191 by @xmath192 for @xmath21 , we assume that there exists a function @xmath10 with the same properties as @xmath193 above , such that @xmath194 . the strict dissipativity assumption [ hyplg ] is then satisfied when @xmath195 the conditions in assumption [ hypf ] are then satisfied for @xmath20 and @xmath21 as soon as there exists @xmath196 such that @xmath197 and @xmath198 are continuously embedded into @xmath199 - it is the case for @xmath18 and @xmath19 given in example [ exampleab ] , with @xmath200 .
assumptions [ hypf2 ] and [ hypg2 ] are also satisfied .
we then deduce that the system is well - posed for any @xmath6 , on any finite time interval @xmath201 $ ] . under assumptions
[ hypab ] , [ hypf ] , [ hypg ] , the nonlinearities @xmath20 and @xmath21 are lipschitz continuous , and the following proposition is classical - see @xcite : for every @xmath6 , @xmath36 , @xmath202 , @xmath203 , system admits a unique mild solution @xmath204,h)))^2 $ ] : for any @xmath205 $ ] , @xmath206
we can now state the main results : the numerical process @xmath207 defined by approximates the slow component @xmath38 of system , with strong and weak error estimates given in the theorems . if the strict dissipativity assumption [ strictdiss ] is satisfied , we can prove the following : [ cvforte ] assuming , for any @xmath208 , @xmath209 , @xmath36 , @xmath210 , @xmath211 , @xmath212 , there exists @xmath122 such that for any @xmath213 , @xmath214 , @xmath215 such that @xmath216 and @xmath217 @xmath218 under the more general weak dissipativity assumption [ weakdiss ] , we can prove the following : [ cvfaible ] assume that @xmath219 and @xmath220 , with @xmath2210,1]$ ] .
let @xmath222 bounded , of class @xmath73 , with bounded first and second order derivatives .
then with the weak dissipativity assumption , for any @xmath223 , @xmath224 , @xmath36 , @xmath210 , @xmath211 , @xmath212 , there exists @xmath122 such that for any @xmath213 , @xmath214 , @xmath215 such that @xmath216 and @xmath217 @xmath225 moreover , if we assume in theorem [ cvfaible ] , the factor @xmath226 is replaced by @xmath76 .
if we look at the estimates of theorems [ cvforte ] and [ cvfaible ] at time @xmath227 , the factor @xmath228 is of size @xmath229 .
the proofs rely on the following decomposition , which explains the origin of the different error terms : @xmath230 the numerical process @xmath231 is defined in below : it is the solution of the macrosolver with a known @xmath30 , while @xmath207 is solution of the macrosolver using @xmath43 . the continuous processes @xmath24 and @xmath23 are defined at the beginning of section [ sectdescri ] .
the first term is bounded thanks to the averaging result , using strong and weak order of convergence results - see the article @xcite .
the second term is the error in a deterministic numerical scheme , for which convergence results are classical ; we recall the estimate in proposition [ propoerrdet ] .
the third term is the difference between the two numerical approximations , and the main task is the control of this part : we show that an extension of the averaging effect holds at the discrete time level , where @xmath232 plays the role of an averaged process for @xmath48 . when we look at the theorems [ cvforte ] and [ cvfaible ] , we first remark that we obtain the same kind of bounds as in the finite dimensional case of @xcite
. however we notice some differences ; they are due both to the infinite dimensional setting and to different proofs .
first , the weak order of the euler scheme for spdes is only @xmath12 - see @xcite - while it is @xmath76 for sdes ; in the estimate , the strong order never appears , and this is one of the main theoretical advantages of the method . for completeness , we recall that the strong order is @xmath75 in the spde case - see @xcite - and @xmath12 for sdes .
in fact , at least in the strictly dissipative case , we are comparing the invariant measures of the continuous fast equation with the invariant measure of its numerical approximation - see theorem [ weakdistance ] and corollary [ weakcor ] . in the weakly dissipative case ,
we use the weak error estimates where the constants do not depend on the final time .
we need some regularity on the initial condition @xmath143 in theorem [ cvfaible ] , since we apply the weak order averaging theorem of @xcite ; nonetheless the parameter @xmath233 does not appear in the different orders of convergence , but only in the constant @xmath29 . in theorem [ cvfaible ] , we do not exactly obtain the same estimate as in @xcite , since @xmath50 does not appear : as a consequence , the estimate is not improved when @xmath50 is increasing .
since we are looking at some weak error estimate , this can seem natural ; first @xmath50 appeared in the estimate of @xcite only because of some strong error estimate used in the proof , while here to prove theorem [ cvfaible ] we always think with a weak error state of mind .
second , the corresponding term in the weak estimate of @xcite is easily controlled by @xmath34 , which already appears in another part of the error since we use a first order scheme for the macrosolver - and not a general scheme - so that there is no need for getting a better estimate .
the proofs of theorems [ cvforte ] and [ cvfaible ] are inspired by the ones in @xcite , but are different .
the strong error is analysed in a global way , as in the proof of the strong order theorem in @xcite .
moreover we do not need a counterpart of lemma @xmath234 in @xcite , and we thus think that our method is more natural . for the control of the weak error , we introduce a new appropriate auxiliary function .
as explained in the introduction , we present here some simplified results and proofs . in the stricly dissipative case , we could go further in the analysis of the error , thanks to the initialization procedure of the scheme , and to additional exponential convergence results .
but after a very technical work , due to the regularity assumptions made on the nonlinear coefficients @xmath20 and @xmath21 we would only obtain a factor @xmath235 instead of @xmath34 in the finite dimensional case - where is any @xmath28 , which can be chosen as small as necessary , and @xmath173 is defined in assumption [ hypf ] and depends on the regularity properties of the nonlinearities @xmath20 and @xmath21 .
nevertheless , the cost of the scheme remains of the same order even if we do not use those better estimates .
we consider that @xmath7 is fixed , and then the numerical method depends on several parameters @xmath236 ; we would like to give some explicit choices of the parameters for which we have a simple bound , and for which a direct scheme , where the stiffness of the system is not treated , is less efficient , for the cost defined by @xmath237 this is the total number of microtime steps for @xmath238 .
we take @xmath227 : then @xmath239 .
the time scale separation parameter @xmath7 is supposed to be very small , while we fix some tolerance @xmath240 for the error in the numerical scheme : more precisely , we want the error to be @xmath241 where @xmath242 satisfies either @xmath243 - strong error - or @xmath244 - weak error - if the parameters @xmath245 and @xmath246 are defined by the conditions @xmath247 ( strong error ) and @xmath248 ( weak error ) : the numerical error is dominant with respect to the averaging error .
we want to show that we can choose the parameters of the scheme such that each term of the estimate of the theorem [ cvforte ] are of size @xmath240 , except the first one , and such that the cost of the scheme is lower than the cost of a direct scheme .
* we first notice that the choice for @xmath34 and @xmath249 is easy : @xmath250 where @xmath243 or @xmath244 , and @xmath224 are fixed .
the choice of other parameters @xmath251 changes when we look at the strong or at the weak error . *
we first focus on the strong convergence case .
we consider the case when either @xmath252 or @xmath253 .
we obtain + [ choicestrong ] + [ cols="^,^,^ " , ] + unlike in the finite dimensional situation , the bound depends on a factor @xmath254 - where @xmath173 is linked to the regularity of the nonlinear coefficients @xmath20 and @xmath21 .
the charge of this difference could be taken by either @xmath63 or @xmath62 ;
the advantage of choosing @xmath62 is that the difference only involves a logarithmic factor on the final cost . as a consequence
, we can make no difference between the choices @xmath253 and @xmath252 .
+ we can now make a comparison with the cost coming from the use of a direct scheme : since the strong order of euler scheme for spdes is @xmath255 , the error can be bounded by @xmath256 for some constant @xmath29 . therefore to have a bound of size @xmath240 , the time step size @xmath35 much satisfy @xmath257 .
this leads to a cost @xmath258 we conclude that in this situation the hmm numerical method is better , since the ratio of the cost tends to @xmath11 when @xmath259 - under the condition @xmath243 .
* we now focus on the weak error estimate . here
@xmath50 plays no role , so that @xmath252 is the good choice ! the time steps @xmath34 and @xmath35 are still given by .
it remains to look for parameters @xmath63 and @xmath62 such that @xmath260 + once again , we need to choose @xmath62 and @xmath63 such that @xmath261 or @xmath262 is large . since exponential decrease is faster than polynomial decrease , the best choice is @xmath263 , and therefore @xmath264 , while @xmath253 .
+ we then see that @xmath265 , so that the additional factor appearing when only weak dissipativity is satisfied plays no role .
+ the corresponding cost of the scheme is then of order @xmath266 .
+ we can again compare this cost with the cost coming from the use of a direct scheme : the weak order of euler scheme for spdes is @xmath267 ; to obtain the second estimate of , we need a cost @xmath268 and again the hmm scheme is better than the direct scheme for such a range of parameters .
in this section , we just recall without proof the main results on the fast equation with frozen slow component and on the averaged equation , defined below .
proofs can be found in @xcite for the strict dissipative case , and the extension to the weakly dissipative situation relies on arguments explained below .
if @xmath202 , we define an equation on the fast variable where the slow variable is fixed and equal to @xmath143 : @xmath269 this equation admits a unique mild solution , defined on @xmath270 .
since @xmath271 is involved at time @xmath152 , heuristically we need to analyse the properties of @xmath272 , with @xmath259 , and by a change of time we need to understand the asymptotic behaviour of @xmath273 when time goes to infinity . under the strict dissipativity assumption [ strictdiss ] ,
we obtain a contractivity of trajectories issued from different initial conditions and driven by the same noise : with , for any @xmath274 , @xmath275 we have @xmath276 under the weak dissipativity assumption [ weakdiss ] , we obtain such an exponential convergence result for the laws instead of trajectories .
the proof of this result is not staightforward , and can be found in @xcite - see also @xcite for further references .
[ propoexpy1y2 ] with , there exist @xmath185 , @xmath122 such that for any bounded test function @xmath277 , any @xmath92 and any @xmath278 @xmath279 , and that the convergence to equilibrium is exponentially fast .
first , let @xmath280 be the centered gaussian probability measure on @xmath13 with the covariance operator @xmath281 - which is positive and trace - class , thanks to assumption [ hypab ] .
then @xmath282 defined by @xmath283 where @xmath2840,+\infty[$ ] is a normalization constant , is the unique probability invariant measure associated to @xmath285 .
this expression comes from the gradient structure of equation , given in assumption [ hypg ] .
second , under any of the dissipativity assumptions , the convergence to equilibrium is exponential in the following sense : [ convexp ] if we assume or , there exist constants @xmath286 such that for any bounded function @xmath287 or @xmath288 , @xmath92 and @xmath161 we have @xmath289 now we define the averaged equation .
first we define the averaged nonlinear coefficient @xmath30 : for any @xmath202 , @xmath290 using assumptions [ hypf ] , [ hypg ] and the expression of @xmath282 , it is easy to prove that @xmath30 is bounded and lipschitz continuous . under assumptions
[ hypf ] , [ hypg ] , we can easily prove the following properties on @xmath30 : [ fbarlip ] there exists @xmath291 and a constant @xmath29 such that the following directional derivatives of @xmath30 are well - defined and controlled : * for any @xmath202 , @xmath162 , @xmath292 . * for any @xmath202 , @xmath162 , @xmath293 , @xmath294 . * for any @xmath202 , @xmath174 , @xmath295 .
moreover , @xmath30 is bounded and lipschitz continuous .
the last estimate above is a consequence of assumptions [ hypf2 ] and [ hypg2 ] , and of the following fact : we have almost surely @xmath296 for any @xmath92 , and @xmath297 moreover if @xmath200 - which is the right condition in the case of linear laplace operators and of nonlinear nemytskii operators - we have @xmath298 : we thus need the restrictive condition in assumption [ hypg2 ] . even when @xmath20 and @xmath21 are nemytskii operators , @xmath30 is not such an operator in general .
then the averaged equation ( see in the introduction ) can be defined : @xmath299 with initial condition @xmath300 . for any @xmath36 , this deterministic equation admits a unique mild solution @xmath301,h)$ ] .
we can give uniform estimates on @xmath48 and @xmath65 , defined by [ schmalent ] and [ schmarapide ] : [ lentenum ] there exists @xmath122 such that we have @xmath302-almost surely @xmath303 for any @xmath304 .
the linear operator @xmath305 satifies @xmath306 ; moreover @xmath20 is bounded , so that by we almost surely have for any @xmath307 @xmath308 .
the end of the proof is then straightforward ; we also notice that @xmath29 does not depend on the final time @xmath309 .
[ rapidenum ] there exists @xmath122 - which does not depend on @xmath36 , on @xmath63 , on @xmath62 or on @xmath50 - such that for any @xmath310 , @xmath311 , @xmath304 , @xmath312 and @xmath313 , we have @xmath314 we introduce @xmath315 defined by the fast numerical scheme with no nonlinear coefficient - see - with the notation @xmath316 : for any @xmath317 , @xmath304 and @xmath312 @xmath318 with the initial condition @xmath319 .
@xmath320 is the numerical approximation - using the microsolver scheme with a step size @xmath321 - of the process defined by @xmath322 where @xmath323 .
notice that the @xmath324 are independent cylindrical wiener processes , and that @xmath325 the @xmath326 are independent realizations of @xmath327 . using theorem @xmath328 of @xcite , giving the strong order @xmath75 for the microsolver - when the initial condition is @xmath11 , with no nonlinear coefficient , with a constant diffusion term and under the assumptions made here - we get the following estimate : for any @xmath209 , there exists @xmath329 , such that for any @xmath55 , @xmath304 , @xmath313 and @xmath312 @xmath330 thanks to and , for any @xmath212
, there exists a constant @xmath331 such that for any @xmath332 , @xmath304 , @xmath313 and @xmath312 we have @xmath333 now for any @xmath312 we define @xmath334 ; it is enough to control @xmath335 , @xmath302-almost surely . by , we have the following expression : for any @xmath312 @xmath336 since @xmath21 is bounded , and using the inequality @xmath337 , we get @xmath338 therefore we have for any @xmath312 @xmath339.\ ] ] but @xmath340 .
so we get @xmath341.\ ] ] therefore @xmath342 as a consequence , we get for any @xmath304 and any @xmath312 , @xmath343 , and then @xmath344 . at the continuous time level ,
the averaging principle proved in @xcite comes from the asymptotic behaviour of the fast equation with frozen slow component , as described in section [ known ] .
the underlying idea of the hmm method in our setting is to prove a similar averaging effect at the discrete time level : we therefore study the asymptotic behaviour of the fast numerical scheme which defines the microsolver , with frozen slow component - in other words we are looking at the evolution of the microsolver during one fixed macrotime step . in section [ known ]
, we have seen that under the weak dissipativity assumption [ weakdiss ] the fast equation with frozen slow component admits a unique invariant probability measure @xmath282 . at the discrete time level
, this assumption only yields the existence of invariant laws ; to get a unique invariant law @xmath345 , we need strict dissipativity [ hyplg ] , and we obtain : [ unicinv ] under assumption [ strictdiss ] , for any @xmath55 and any @xmath202 ; the numerical scheme [ schmarapide2 ] admits a unique ergodic invariant probability measure @xmath345 .
moreover , we have convergence to equilibrium in the following sense : for any @xmath212 , there exist @xmath185 and @xmath122 such that for any @xmath346 , @xmath202 , @xmath203 , any lipschitz continuous function @xmath277 from @xmath13 to @xmath347 , and @xmath52 , we have latexmath:[\[\label{estimvitesse } we recall the notation @xmath316 for the effective time step ; the noise is defined with a cylindrical wiener process @xmath349 : @xmath350 .
if we fix the slow component @xmath202 , we define @xmath351 with the initial condition @xmath352 . with equation , we associate the transition semi - group @xmath353 : if @xmath277 is a bounded measurable function from @xmath13 to @xmath354 , if @xmath203 and @xmath52 @xmath355.\ ] ] we also denote by @xmath356 the law of @xmath357 : then @xmath358=\int_{h}\phi(z)\nu_{m , y}^{x,\tau}(dz).\ ] ] we notice that the semi - group @xmath359 satisfies the feller property : if @xmath277 is bounded and continuous , then @xmath360 is bounded and continuous .
the required tightness property is a consequence of the following estimate , which can be proved thanks to regularization properties of the semi - group @xmath361 : for any @xmath362 , @xmath55 , there exists @xmath363 such that for any @xmath364 and @xmath365 @xmath366 moreover if @xmath362 , the embedding of @xmath367 in @xmath13 is compact . the key estimate to prove uniqueness is the following contractivity property , which holds thanks to assumption [ strictdiss ] : [ theoestim ] for any @xmath212 , there exists @xmath185 such that for any @xmath346 , @xmath52 , @xmath278 , @xmath202 we have @xmath302-almost surely @xmath368 , then we have the equation @xmath369 if we take the scalar product in @xmath13 of this equation with @xmath370 , we get @xmath371 the left hand - side is equal to @xmath372 and we get @xmath373 therefore we have @xmath374 we remark that [ hyplg ] implies that for any @xmath212 , there exists @xmath185 such that if @xmath365 we have @xmath375 ; therefore @xmath376 as a consequence , there exists a unique ergodic invariant probability measure @xmath345 , which is strongly mixing .
moreover one can prove that there exists @xmath122 such that for any @xmath55 and any @xmath202 @xmath377 ; we therefore get @xmath378_{\text{lip}}\int_{h}e^{-cm\tau}|y - z|\mu^{x,\tau}(dz)\\ & \leq c[\phi]_{\text{lip}}e^{-cm\tau}.\end{aligned}\ ] ] we recall that @xmath379 denotes the invariant law of the continuous time fast equation with frozen slow component .
thanks to the fast numerical scheme , we can get an approximation result , which is proved in @xcite : with test functions of class @xmath380 , we can control the weak error for any time with a convergence of order @xmath12 with respect to the time step @xmath321 .
moreover the estimate is easily seen to be independent from the slow component @xmath143 .
we define for any @xmath72 of class @xmath380 @xmath381 [ weakdistance ] with the dissipativity condition , for any @xmath209 , for any @xmath212 , there exists @xmath286 such that for any @xmath72 of class @xmath380 , for any @xmath161 , for any @xmath346 and any integer @xmath382
c\|\phi\|_{(2)}(1+|y|^3)(((m-1)\tau)^{-1/2+\kappa}+1)\tau^{1/2-\kappa}.\ ] ] as explained in section [ invarfast ] , in general the fast numerical scheme does not admit a unique invariant probability measure , but when the strict dissipativity assumption [ strictdiss ] is satisfied , it admits a unique invariant law @xmath345 . [ weakcor ] under the assumptions of theorem [ weakdistance ]
: _ ( i ) if only is satisfied , we have for any @xmath384 @xmath385|\leq c\|\phi\|_{(2)}(1+|y|^3)(((m-1)\tau)^{-1/2+\kappa}+1)\tau^{1/2-\kappa}+cn(\phi)(1+|y|^2)e^{-cm\tau}.\ ] ] _ _ ( ii ) if moreover is satisfied , @xmath386 _ we recall that in the case of euler scheme for sdes this kind of results holds with the order of convergence @xmath76 .
we define a scheme based on the macrosolver , for theoretical purpose , in the situation when @xmath30 is known : @xmath387 we can look at the error between @xmath232 , defined by , and @xmath39 , defined by . here
quantities are deterministic , and the following result is classical - see @xcite , @xcite , or the details of the proofs in @xcite : [ propoerrdet]for any @xmath223 , @xmath388 and @xmath36 , there exists @xmath122 , such that for any @xmath389 and @xmath390 @xmath391
thanks to the following remark on the construction of the scheme , we can consider that @xmath65 corresponds to a process evaluated at time @xmath392 . the idea is to build noise processes @xmath393 by concatenation of the @xmath394 for the different @xmath49 .
[ remarknoise ] to compute @xmath65 with the microsolver , the total number of microtime steps used is @xmath395 .
it is then natural to use only one noise process @xmath396 , for each @xmath313 , and to make an evaluation involving time @xmath395 .
more precisely , we can use @xmath397 with @xmath398 being @xmath50 independent cylindrical wiener processes on @xmath13 .
then one can define for @xmath399 and for @xmath400 @xmath401 we also introduce the following notation : @xmath402 denotes conditional expectation with respect to the @xmath403-field @xmath404 we notice that @xmath48 is @xmath405-measurable , but that @xmath43 is not . the final time @xmath309 is fixed and we recall the notation @xmath406 . to simplify notations ,
we do not always precise the range of summation in the expressions below : the indices @xmath407 belong to @xmath408 , and @xmath409 belong to @xmath410 .
we recall that according to the decomposition of the error , we have to control @xmath411 the first part is controlled thanks to the strong order theorem of @xcite : for any @xmath208 , we have for any @xmath304 @xmath412 the second part is deterministic and is controlled thanks to proposition [ propoerrdet ] : @xmath413 where @xmath29 depends on @xmath414 .
it remains to focus on the third part @xmath415 .
instead of analyzing the local error like in @xcite , we adopt a global point of view , and we follow the idea of the proof of theorem @xmath416 in @xcite : for any @xmath304 @xmath417 the averaged coefficient @xmath30 is lipschitz continuous , and @xmath418 ; moreover we can define the averaged coefficient @xmath419 with respect to the invariant measure @xmath345 of the fast numerical scheme - which is unique since we assume strict dissipativity : for any @xmath202 @xmath420 the error in can then be decomposed in the following way - the idea of looking at the square of the norm in the second expression is an essential tool of the proof : @xmath421 if we can control the two last terms by a certain quantity @xmath422 , by a discrete gronwall lemma we get for any @xmath423 @xmath424 .
first , the third term in is linked to the distance between the invariant measures @xmath282 and @xmath345 - since we assume strict dissipativity for this strong estimate - which is evaluated thanks to theorem [ weakdistance ] and corollary [ weakcor ] for test functions of class @xmath380 .
since @xmath425 , using assumption [ hypf2 ] , we can apply corollary [ weakcor ] to obtain @xmath426 and summing we get @xmath427 the other term is more complicated ; in order to get a precise estimate , we expand the square of the norm of the sum .
we can then use some conditional expectations , which allow to use exponential convergence to equilibrium via theorem [ unicinv ] .
therefore we obtain the following expansion : @xmath428 _ ( i ) _ we first treat @xmath429
. by , for any @xmath430 we have @xmath431 therefore we can see that @xmath432 with the conditional expectation @xmath433 with respect to @xmath434 - see . when @xmath435 , @xmath436 and @xmath437 are independent
, so that we treat differently the cases @xmath438 and @xmath435 in the above summation , and we obtain @xmath439 for the first part , we can directly use the exponential convergence to equilibrium result of to have a bound with @xmath440 for the second part , we see that we can treat only the case @xmath441 , and we introduce the conditional expectation @xmath442 with respect to the @xmath403-field generated by @xmath434 and @xmath443 , when @xmath441 .
we get a bound with @xmath444 we therefore get @xmath445 _ ( ii ) _ we now consider @xmath446 , which corresponds to the cross - terms in the expansion of the square of the norm of the quantity @xmath447 . by the definition of @xmath448 , the general term with indices
@xmath449 in @xmath450 is bounded by @xmath451|,\end{gathered}\ ] ] using conditional expectation @xmath452 and the boundedness of @xmath20 .
using the exponential convergence result of and lemma [ rapidenum ] , we get the bound @xmath453|\leq ce^{-m\tau},\ ] ] so that the previous quantity is bounded by @xmath454 summing on @xmath449 , we can now conclude that @xmath455 then by and @xmath456 and the result of theorem [ cvforte ] now follows from .
in order to get a better bound for the weak error than for the strong error , we use an auxiliary function which is solution of a kolmogorov equation .
thanks to the averaging theorem of @xcite , which is proved under the weak dissipativity assumption , the first term above can be controlled by @xmath458 , where @xmath459 is a constant - depending on @xmath460 , for any @xmath223 . for the second term ,
since we look at the error made by using a deterministic scheme to approximate a deterministic equation , there is no difference between the strong and the weak orders - since the test function @xmath72 is lipschitz continuous ; so we again use proposition [ propoerrdet ] . for the third term
, we see that we have to control some error between two different numerical schemes , in a weak sense .
the usual strategy is to decompose this error by means of an auxiliary function satisfying some kind of kolmogorov equation .
more precisely , we use the deterministic scheme defining @xmath461 in order to define for any @xmath462 @xmath463 where we explicitly mention dependence of the numerical solution @xmath461 on the initial condition @xmath143 . moreover the second and the third estimates of this lemma
reveal some smoothing effect in the equation , due to the semi - group @xmath474 .
these results are specific to the infinite dimensional framework ; if we only use the first estimate above , and a simple one like @xmath475 we would not obtain an optimal convergence result . in the general term of
, we proceed with a taylor expansion and we get @xmath476 the second term is controlled by using the second estimate of the previous lemma [ regu_n ] . under the assumption that @xmath20 is bounded , we therefore get a @xmath477 term ; when summing over @xmath430 , we get a @xmath478 term , which is already dominated in the final estimate . for the first term
, we do not exactly follow the proof of @xcite ; we rather define auxiliary functions for @xmath462 in order to keep on looking at a weak error term : @xmath482 then if we define @xmath483 we have @xmath484 by using conditional expectation @xmath433 with respect to @xmath434 defined by , @xmath485 does not depend on @xmath486 , so that in the sequel we fix @xmath487 .
[ regpsi_nk ] for any @xmath488 and @xmath388 , there exists a constant @xmath29 , such that for any @xmath389 , any @xmath489 and any @xmath430 , the following derivatives exist and are controlled : for any @xmath161 , @xmath174 , @xmath490 , the proof relies smoothing effect of the semi - group @xmath474 .
the lemma is proved below in section [ sect_aux_proofs ] . when the strict dissipativity assumption is satisfied
, we can indeed obtain a bound without @xmath495 : we can control the distance between the invariant measures of the continuous and discrete time processes , thanks to the second part of corollary [ weakcor ] .
we use the following expression for @xmath496 : @xmath497 by definition , for any @xmath462 we have @xmath498 ; we see that the derivatives in directions @xmath499 are given by @xmath500 and @xmath501 @xmath72 is of class @xmath73 on @xmath13 , with bounded derivatives ; therefore we just need to control @xmath502 and @xmath503 .
we use the following estimates of the derivatives of @xmath30 , given in proposition [ fbarlip ] : for any @xmath202 , @xmath162 , @xmath504 , latexmath:[\[\begin{gathered } _ ( ii ) _ for any @xmath304 , we can write @xmath509 since @xmath510 when @xmath511 , and thanks to the previous estimates on @xmath512 and @xmath513 , we get @xmath514 therefore @xmath515 and a discrete gronwall lemma then yields @xmath516 _ ( ii ) _ when we look at the second - order derivative , we see that @xmath523 thanks to the last estimate of lemma [ regu_n ] , we can control the expression by @xmath524 we then notice that @xmath525 since @xmath526 and @xmath34 is bounded , and thanks to assumption [ hypf2 ] ; the other part is controlled thanks to assumption [ hypf2 ] . | we consider the discretization in time of a system of parabolic stochastic partial differential equations with slow and fast components ; the fast equation is driven by an additive space - time white noise .
the numerical method is inspired by the averaging principle satisfied by this system , and fits to the framework of heterogeneous multiscale methods .
the slow and the fast components are approximated with two coupled numerical semi - implicit euler schemes depending on two different timestep sizes .
we derive bounds of the approximation error on the slow component in the strong sense - approximation of trajectories - and in the weak sense - approximation of the laws .
the estimates generalize the results of @xcite in the case of infinite dimensional processes . |
it is of great interest to find lattice formulations of quantum chromodynamics ( qcd ) with exact chiral symmetry . in such a formulation ,
spontaneous chiral symmetry breaking and the associated spectrum of goldstone bosons ( i.e. the light mesons ) , as well as the emergence of the flavor - singlet ( @xmath0 meson ) mass via the index theorem connection between quark zero - modes and gauge field topology @xcite , can be studied directly in the lattice model without the usual need to extrapolate via chiral perturbation theory .
finding such formulations is problematic though @xcite , and was a long - standing challenge in lattice qcd .
an explicit formulation of this kind was finally found by neuberger @xcite via the overlap formulation of chiral fermions on the lattice @xcite .
it is a chirally improved version of wilson fermions , and realizes exact chiral symmetry of the ginsparg - wilson ( gw ) type @xcite .
however , its theoretical attractiveness is offset in practice by the high computational cost of implementing it in simulations . in this paper , building on insights from @xcite , a new version of overlap fermions is presented constructed from the staggered lattice fermion rather than the previously used wilson fermion .
the staggered fermion is the most computationally efficient lattice fermion formulation known , and this advantage is expected to be inherited by the new staggered version of overlap fermions . the challenge in doing this is that the overlap construction relies on specific properties of wilson fermions that are not shared by staggered fermions .
this challenge is overcome here using a new theoretical idea of interest in its own right : a wilson - like phase of the staggered fermion theory can be created by adding a certain `` flavored '' mass term . in this new phase , the number of quark flavors described by the staggered fermion is reduced from four to two , and the staggered fermion acquires wilson - like properties . the latter allow the overlap construction to work as desired .
the resulting staggered overlap fermion describes two quark flavors and has an exact _ unflavored _ gw chiral symmetry originating from the exact flavored chiral symmetry of the original staggered fermion .
however , the _ flavored _ vector and chiral symmetries of the quark pair described by the staggered overlap fermion are not exact since the corresponding symmetries of the original staggered fermion are broken by lattice effects . therefore , exact flavored vector and chiral symmetries only hold for _ pairs _ of quark flavors when each pair is described by a staggered overlap fermion . the two quark flavors described by the staggered overlap fermion can be taken as the almost massless @xmath1 and @xmath2 quarks in qcd .
then , due to the exact unflavored gw chiral symmetry , this description has all the advantages of the usual overlap fermions regarding the axial u(1 ) anomaly and index theorem connection between quark zero - modes and gauge field topology .
this is connected with some of the most subtle and interesting parts of qcd physics not only the @xmath0 mass mentioned already but also the order of the qcd phase transition at finite temperature which are also computationally highly demanding to study in lattice qcd ( see , e.g. , @xcite ) . finding a new lattice approach in which these features are not distorted by lattice effects , and which is computationally more efficient than usual overlap fermions , is therefore significant . as an indication of the desirability of this
, it can be noted that high - precision calculation of the @xmath0 mass is still an outstanding problem in lattice qcd the latest state of the art calculation , using wilson - based domain wall fermions ( a truncated version of usual overlap fermions ) , has an uncertainly of 15% @xcite .
the staggered lattice fermion field is a one component grassmann field @xmath3 , @xmath4 , describing 4 degenerate dirac fermions .
the staggered dirac operator acting on these fields is @xmath5 where @xmath6 and @xmath7 , with the parallel transporters @xmath8 given in terms of the lattice link variables @xmath9 by @xmath10 and @xmath11 .
there is an exact _ flavored _ chiral symmetry \{_55,d_st}=0 [ 3 ] where @xmath12 , given by @xmath13 , corresponds to @xmath14 in the spin - flavor interpretation of staggered fermions @xcite .
the first and second @xmath15 factors act in spinor and flavor space , respectively .
note the properties @xmath16 and @xmath17 .
direct application of the overlap construction @xcite to staggered fermions by simply replacing the kernel operator @xmath18 ( where @xmath19 is the wilson - dirac operator ) by @xmath20 does not give anything useful : the exact flavored chiral symmetry of the staggered fermion is lost , there is no gw symmetry to replace it , and there are no exact zero - modes in general .
however , as we will now show , the situation changes if , instead of a scalar mass @xmath21 , we use the following `` flavor - chiral '' mass term : m_st=_55_5,r>0 [ 5 ] here @xmath22 is the staggered fermion version of the chirality matrix , corresponding to @xmath23 up to @xmath24 discretization errors in the spin - flavor interpretation @xcite .
consequently , in the spin - flavor interpretation , m_st ~ * 1*_5+o(a ) .
[ 8 ] explicitly , @xmath22 is given as follows .
use the parallel transporters to define @xmath25 , and let @xmath26 denote the symmetrized product of @xmath27
. then @xmath28 where @xmath29 .
( explicitly , @xmath30 . )
note that @xmath22 is hermitian and commutes with @xmath12 ; therefore @xmath31 is hermitian .
the effect of this mass term in d_st - m_st [ 9 ] is to split the low - lying modes into branches with approximately definite positive and negative flavor - chirality , giving them masses @xmath32 and @xmath33 , respectively .
this is shown in the free field case in fig . 1 , taken from @xcite . spectrum of @xmath34 in the free field case @xcite.,width=124 ] in this case
, there are 16 zero - modes of @xmath35 with momenta @xmath36 , @xmath37 . on the vectorspace @xmath38 spanned by these modes we have @xmath39 , so there is a decomposition @xmath40 with @xmath41 on @xmath42 , and it can be shown that @xmath43 . hence @xmath44 on @xmath42 ; this shows how the real eigenvalues arise in the free field case . moreover , the operators @xmath45 and @xmath46 making up @xmath35 coincide on @xmath38 and can be shown to commute with @xmath47 on @xmath38 .
hence their representations @xmath48 on @xmath38 ( given in @xcite by the @xmath49 matrices @xmath50 ) , which form a representation of the euclidean dirac algebra @xcite , decompose into representations @xmath51 on @xmath42 . from this it can be seen that , from a low momentum viewpoint , the 4 dirac fermion species described by the staggered fermion are split into two pairs of dirac fermions , with the pair with positive ( negative ) flavor - chirality having mass @xmath32 ( @xmath52 ) . upon inserting the kernel @xmath34 into the overlap formula , d_so&=&(1+(d_st - m_st ) ) ,
+ & & [ 10 ] the negative mass modes are converted into the physical modes of a massless fermion , while the positive mass modes are converted into modes with heavy masses @xmath53 which decouple in the continuum limit , just as in the wilson case .
thus only two of the original four quark flavors survive in the resulting overlap fermion . furthermore , since the physical modes have approximately definite positive flavor - chirality , @xmath54 on these modes up to @xmath24 effects .
the staggered overlap dirac operator ( [ 10 ] ) satisfies the following gw relation : \{_55,d_so}=d_so_55d_so [ 11 ] note that this is now an _
unflavored _ chiral symmetry , since , as remarked , @xmath55 on the physical modes of the staggered overlap fermion .
the gw relation can be readily verified after noting that the staggered overlap operator can be expressed as d_so&=&(1+_55 ) [ 12 ] with the hermitian operator @xmath56 given by h_st=_55(d_st - m_st)=_55d_st-_5 [ 13 ] it also follows from ( [ 12 ] ) that @xmath57 , which implies that the non - real eigenvalues of @xmath58 come in complex conjugate pairs . since the only possible real eigenvalues ( which must lie on the gw circle ) are @xmath59 and @xmath60
, it follows that the fermion determinant @xmath61 is real and non - negative , as required for lattice qcd simulations .
a required property of the lattice dirac operator is locality .
it holds for @xmath58 ; this can be proved analytically by the same argument as in the wilson overlap case @xcite , using the staggered bound mentioned in @xcite , when the lattice gauge fields satisfy an admissibility condition ; the details are in @xcite .
locality of @xmath58 in realistic gauge field backgrounds has been shown numerically in @xcite . computational efficiency of @xmath58
was also studied in @xcite : preliminary numerical evidence indicates a speedup of order 2 - 3 compared to the wilson - based overlap dirac operator in realistic gauge field backgrounds .
we now verify that @xmath58 has pairs of exact chiral zero modes with index determined by gauge field topology in accordance with the 2-flavor index theorem when the lattice gauge field background is not too rough .
due to the gw relation ( [ 11 ] ) , the index is well - defined as @xmath62 where @xmath63 is the number of independent zero - modes of @xmath58 with @xmath64 chirality under @xmath12 ( which is the same as @xmath22 on these modes up to @xmath24 effects ) .
a general formula @xcite gives ( d_so)=- ( ) .
[ 16 ] we connect this to the hermitian staggered operator studied previously in @xcite , denoted by @xmath65 below : * theorem . * _ for every value of @xmath66 , the hermitian staggered operators _ h_st(m)=_55d_st - m_5 , _ st(m)=id_st - m_5 the proof of the theorem is deferred to @xcite .
it can be understood intuitively from the fact that @xmath67 arises from @xmath68 by a change of representation of the staggered sign factors @xmath69 in @xmath35 . in light of the theorem we can replace @xmath56 in ( [ 16 ] ) by @xmath67 with @xmath70
; here @xmath71 is the parameter in ( [ 5 ] ) .
but then the results of @xcite give ( d_so)=(d_st)=2q [ 18 ] in sufficiently smooth gauge field backgrounds of topological charge @xmath72 and with @xmath33 lying in a suitable range .
the last equality in ( [ 18 ] ) was confirmed in the numerical study in @xcite .
it is important to consider the symmetries of the staggered overlap fermion since these need to be sufficient for renormalizability and stability of the massless 2-flavor phase under radiative corrections . besides the replacement of the flavored chiral symmetry ( [ 3 ] ) by the gw symmetry ( [ 11 ] ) ,
the usual staggered fermion symmetries listed in @xcite all hold for the staggered overlap fermion , except for axis reversals and shift transformations . under the latter @xmath73 , and
therefore @xmath74 is invariant when these transformations are combined with the parameter flip @xmath75 in ( [ 5 ] ) .
the same invariances must hold for all counterterms that arise . hence the counterterms must be invariant under all the usual staggered fermion symmetries except ( [ 3 ] ) , and except for a possible sign change under axis reversals and shift transformations .
it is easy to show @xcite that the most general _ local _ mass - dimension 4 terms with these properties are ( up to proportionality , and modulo @xmath76 terms ) and @xmath35 commute and the second term in ( [ 18a ] ) simplifies . ]
local , all counterterms are also local , and the theory is renormalizable , provided the power - counting theorem holds @xcite .
it holds in the wilson overlap case @xcite and is plausible in the present case . assuming this , stability can be established to all orders in the loop expansion as follows . the gw relation ( [ 11 ] ) leads to _
55d_so^-1(q)+d_so^-1(q)_55 = _ 55 [ 18b ] where @xmath77 with @xmath78 being the fermion self - energy . here and in the following
, @xmath79 denotes the free field momentum representation of the staggered operator @xmath80 ; it is a linear map ( @xmath49 matrix ) @xmath81 on the zero - mode vector space @xmath38 defined earlier , with @xmath82 $ ] .
the momentum representation of @xmath12 is a constant matrix , and @xmath83 on @xmath42 where @xmath84 . in light of the preceding , at one loop
@xmath78 has the form c_1d_st(q)+c_2_55_5(q)d_st(q ) + _ 55_5(q)+ [ 18c ] up to irrelevant terms , with each @xmath85 diverging no worse than logarithmically for @xmath86 .
note that all terms in ( [ 18c ] ) map @xmath42 to itself , as does @xmath87 . from ( [ 18b])([18c ] ) and
the fact that @xmath88 on @xmath89 it is straightforward to show @xmath90 on @xmath89 .
( we omit the details ; an analogous result was found in the wilson overlap case in @xcite ) .
this implies @xmath91 in ( [ 18c ] ) , so on @xmath89 the last two terms in ( [ 18c ] ) combine to give an irrelevant term @xmath92 which vanishes @xmath93 for @xmath86 .
the first two terms in ( [ 18c ] ) act on @xmath42 as @xmath94 .
it follows that , for @xmath86 , @xmath95 on @xmath89 . the self - energy is hereby seen to have the usual effect of a logarithmically divergent wave function renormalization on the massless physical modes .
then , after renormalizing the theory at one loop , the argument above can be repeated to get the same result at two loops , and so on . in this way stability
is seen to hold to all orders in the renormalized theory .
a deeper understanding of the staggered overlap construction can be obtained by clarifying the wilson - like nature of its staggered fermion kernel . to this end
we introduce the following 2-flavor wilson - like version of the staggered dirac operator : d_sw = d_st+w_st , w_st=(1-_55_5 ) . [ 19 ] the term @xmath96 plays an analogous role to the wilson term in the wilson - dirac operator : it decouples the negative flavor - chirality modes by giving them mass @xmath60 while keeping the two positive flavor - chirality modes as the physical modes .
hence @xmath97 describes two physical quark flavors on which @xmath98 up to @xmath76 effects . in light of this
we can obtain a 2-flavor overlap fermion by taking @xmath99 with @xmath100 , @xmath101 as the kernel in the usual overlap construction . for @xmath102
this is precisely the 2-flavor staggered overlap dirac operator @xmath58 we constructed above in ( [ 10 ] ) , since @xmath103 but now we see that it can be generalized to any @xmath101 . furthermore , the role of the parameter @xmath71 in the staggered overlap construction is hereby clarified : it is analogous to the wilson parameter in the usual overlap construction based on wilson fermions .
this is surprising , since @xmath104 in ( [ 5 ] ) initially appears to be analogous to the mass parameter @xmath105 in the wilson case .
starting from wilson fermions , a domain wall fermion formulation can be constructed @xcite which gives a truncation of the overlap fermion construction @xcite . the same can now be done with staggered fermions simply by replacing d_wd_sw,_5_55 [ 26 ] in the previous wilson - based constructions .
the lattice dirac operator for the staggered domain wall fermion in 5 dimensions is then d_sdw = d_sw - m+_55_s , m= , ( 0,2 ) [ 27 ] where @xmath106 $ ] is the lattice coordinate of the 5th dimension and @xmath107 with @xmath108 and @xmath109 , @xmath110 being the forward and backward finite difference operators .
boundary conditions on @xmath111 at @xmath112 and @xmath113 are introduced analogously to the wilson - based case ; then the staggered domain wall construction is seen to be a truncation of the staggered overlap construction by essentially the same argument as in @xcite .
the details will be given in @xcite .
underlying the staggered overlap and staggered domain wall fermion constructions in this paper is a staggered version of wilson fermions obtained by adding a flavored mass term to the staggered fermion action .
the idea of considering staggered fermion with flavored mass term is not new in itself ; this was already done many years ago by golterman and smit in @xcite .
rather , the new idea here is that , for a certain choice of flavored mass term , namely the wilson term @xmath96 in ( [ 19 ] ) determined by the `` flavor - chiral '' mass term @xmath31 in ( [ 5 ] ) , one can use the staggered operator @xmath12 for the role of @xmath15 since it coincides with @xmath23 on the 2 physical flavors up to @xmath76 effects .
the significance of this is that the properties @xmath114 and @xmath115hermiticity , which are crucial in the usual overlap and domain wall constructions , continue to hold in the staggered setting with @xmath116 . this would not be true if one tried to use the direct staggered analogue @xmath22 of @xmath15 .
the flavor - chiral mass term used in this work , which reduces the 4 staggered fermion flavors to 2 flavors , is not the only possibility .
another flavor - chiral mass term which reduces the number of flavors to 1 was subsequently proposed in @xcite after the first version of this paper appeared on the arxiv .
however , it breaks more of the staggered fermion symmetries than the present proposal , namely some of the lattice rotation symmetries .
the consequences of this for radiatively generated counter - terms and fine - tuning requirements remains to be seen .
the 1-flavor version of staggered overlap fermions has the advantage that all the _ flavored _ vector and gw - chiral symmetries are exact . on the other hand , for problems where
only the _ unflavored _ chiral symmetry is important , e.g. for computing the @xmath0 mass , the present 2-flavor formulation has advantages : it preserves lattice rotation symmetry and gives `` two quarks ( the @xmath1 and @xmath2 ) for the price of one '' in lattice qcd simulations .
i thank mike creutz , philippe de forcrand , maarten golterman and herbert neuberger for discussions and/or feedback .
i also thank philippe de forcrand and his group for performing the numerical investigations in @xcite and for kindly providing me with fig . 1 .
this work was supported by the acrf tier 1 grant rg 61/10 .
some of this work was done at the national center for theoretical sciences , taiwan .
e. witten , nucl .
b 156 , 269 ( 1979 ) ; g. veneziano , nucl .
b 159 , 213 ( 1979 ) h.b .
nielsen and m. ninomiya , nucl .
b 158 , 20 ( 1981 ) ; l.h .
karsten and j. smit , nucl .
b 183 , 103 ( 1981 ) h. neuberger , phys .
b 417 , 141 ( 1998 ) ; _ ibid _ 427 , 353 ( 1998 ) r. narayanan and h. neuberger , nucl
. phys .
b 443 , 305 ( 1995 ) p.h .
ginsparg and k.g .
wilson , phys .
d 25 , 2649 ( 1982 ) p. hasenfratz _ et al .
_ , phys .
b 427 , 125 ( 1998 ) m. lscher , phys .
b 428 , 342 ( 1998 ) d.h .
adams , phys .
lett . 104:141602 ( 2010 )
p. chen _ et al .
_ , phys .
d 59:054508 ( 1999 ) n.h .
christ et al . , phys .
lett . 105:241601 ( 2010 )
golterman and j. smit , nucl .
b 245 , 61 ( 1984 ) ; m.f.l .
golterman , nucl .
b 273 , 663 ( 1986 ) p.
de forcrand , a. kurkela and m. panero , pos lattice2010:080 ( 2010 ) d.h .
adams ( in preparation ) p. hernandez _ et al .
_ , nucl .
b 552 , 363 ( 1999 ) t. reisz , nucl .
b 318 , 417 ( 1989 ) t. reisz and h.j .
rothe , nucl .
b 575 , 255 ( 2000 ) t .- w .
chiu _ et al .
_ , phys .
b 438 , 321 ( 1998 ) ; m. ishibashi _ et al .
_ , nucl .
b 576 , 501 ( 2000 ) y. shamir , nucl .
b 406 , 90 ( 1993 ) h. neuberger , phys .
d 57 , 5417 ( 1998 ) g. fleming and p.
vranas , phys .
d66 , 114503 ( 2002 ) c. hoelbling , phys .
b 696 , 422 ( 2011 ) | a new formulation of chiral fermions on the lattice is presented .
it is a version of overlap fermions , but built from the computationally efficient staggered fermions rather than the previously used wilson fermions .
the construction reduces the four quark flavors described by the staggered fermion to two quark flavors ; this pair can be taken as the up and down quarks in lattice qcd .
a domain wall formulation giving a truncation of this overlap construction is also outlined .
lattice qcd , chiral fermions , staggered fermions |
since superconductivity was discovered in the layered perovskite @xmath0 in 1994 @xcite , there has been much speculation as to the nature of the superconducting order parameter .
almost immediately , it was realized that @xmath0 might be chiral @xmath5-wave @xcite , an electronic analogue of the @xmath6-phase of @xmath7he .
however , after more than a decade and a half of intense experimental and theoretical work , basic questions is the order parameter chiral and if so , where on the three active bands of @xmath0 does it live ? remain controversial .
( for recent reviews , see refs .
@xcite , as well as the comprehensive earlier review in ref .
@xcite . )
if @xmath0 turns out to be chiral @xmath5-wave , it would be the only known example so far of this remarkable state in electronic systems .
chiral superconductivity breaks time - reversal symmetry with order parameters describing cooper pairs with nonzero magnetic moments . for chiral @xmath5-wave superconductivity ,
the time - reversal symmetry breaking triplet order parameter is _ c^_c^_-~k_x ik_y , corresponding to a nonzero angular momentum @xmath8 in the @xmath9-direction per cooper pair .
apart from the intrinsic interest in observing and understanding such an unusual quantum state , the quest to find chiral @xmath5-wave superconductors has generated enormous interest since they have been predicted to harbour majorana bound states @xcite , quasiparticles which are their own antiparticles and which may play an important role in quantum computation .
one important feature of @xmath0 that can impact the nature of the possibly chiral order parameter is the fact that it is multiband @xcite .
three atomic orbitals dominate the electronic structure of @xmath0 close to the fermi surface : ru @xmath10 , @xmath1 , and @xmath2 orbitals ( strongly hybridized with o @xmath5 orbitals ) .
the @xmath1 and @xmath2 orbitals hybridize to form the quasi - one - dimensional @xmath3 and @xmath4 bands , while the @xmath10 orbital predominates the @xmath11 band , which has an approximately isotropic two - dimensional dispersion .
spin - orbit coupling leads to some hybridization between the @xmath12 and @xmath10 orbitals , which otherwise would not hybridize @xcite .
importantly for @xmath0 , the fermi surface crosses all three of the bands resulting from this orbital structure . where does superconductivity arise amongst these three bands ?
the most widely - held viewpoint holds that superconductivity arises primarily on the @xmath11 band ( see , for instance , refs .
@xcite ) . a persistent contrarian position , however , maintains that superconductivity instead arises primarily on the @xmath3 and @xmath4 bands @xcite .
the latter model has one very appealing feature insofar as it may give rise to negligible spontaneous supercurrents at sample edges @xcite .
such edge currents have been predicted to arise in one - band chiral @xmath5-wave superconductors @xcite , but bafflingly , have never been observed @xcite .
( for a recent review of experimental probes of the order parameter in @xmath0 , see ref .
non - topological chiral superconductivity on the @xmath3 and @xmath4 bands could explain this . amongst experiments purporting to show evidence for time - reversal symmetry breaking in @xmath0 , one of the most direct probes is the kerr rotation experiment carried out in the group of kapitulnik at stanford @xcite . in this experiment , a near - infrared beam ( @xmath13ev )
is reflected off a sample of @xmath0 , the beam s axis of polarization rotating in the process .
the angle of rotation is called the _
kerr angle_. it is related to the anomalous hall conductivity , @xmath14 , by @xcite _ k ( ) = ( 4/d)[_h()()].[kerr]here , @xmath15 is the interlayer spacing and @xmath16 with @xmath17 the complex , frequency - dependent index of refraction . as we discuss in detail in sec . 2
, a nonzero hall conductivity indicates that time - reversal symmetry is broken .
if time - reversal symmetry is broken by chiral superconductivity , spectroscopic knowledge of the frequency - dependent kerr angle can tell us a lot about the chiral superconducting order parameter in @xmath0 in particular , where it lies in relation to the three active bands of @xmath0 .
( spectroscopy of leggett - like collective modes @xcite would also provide considerable insight into this matter . )
it is the point of this paper to expand on this idea and suggest ways for future experiments to make further progress in answering questions related to the nature of the order parameter . on its own , time - reversal symmetry breaking is not sufficient to have a nonzero hall conductivity in the absence of an applied magnetic field and hence , a non - zero kerr angle . as pointed out by read and green @xcite , for instance
, the hall conductivity of a translationally - invariant chiral @xmath5-wave superconductor is identically zero @xcite . since the publication of kapitulnik s group s data in 2006 @xcite
, two credible mechanisms have emerged in which chiral @xmath5-wave superconductivity in @xmath0 can give rise to a nonzero anomalous hall conductivity .
first , a purely extrinsic mechanism due to impurity scattering was suggested by goryo @xcite and lutchyn _ et al .
_ @xcite .
subsequently , an _ intrinsic _ mechanism was identified by us @xcite as well as wysokiski _ et al . _
( a previously identified intrinsic contribution from chiral p - wave collective modes @xcite is estimated to be several orders of magnitude too small to explain the experiments @xcite and will not be discussed here . ) very usefully for the questions identified above , the intrinsic mechanism gives rise to a nonzero anomalous hall conductivity if and only if there are multiple chiral order parameters spanning multiple optically active bands . for @xmath0 ,
this means that there _ must _ be superconductivity on the @xmath3 and @xmath4 bands in order for the intrinsic @xmath14 to be nonzero .
in contrast , the extrinsic mechanism contributes as well when chiral superconductivity occurs on the @xmath11 band .
thus , if it was found that the intrinsic mechanism dominated the hall conductivity , one could reliably conclude that the origin of superconductivity in @xmath0 lay in the @xmath3 and @xmath4 bands and not the @xmath11 band . in this paper
, we suggest two ways in which experiments can deduce the relative importance of the intrinsic and extrinsic mechanisms . first , the intrinsic mechanism produces a very strong spectral feature in the frequency - dependent anomalous hall conductivity ( and kerr angle ) close to twice the inter - orbital @xmath18 hopping energy , near @xmath19ev .
the detection of this feature , robust to temperature and impurity effects in clean samples of @xmath0 , would provide nearly unequivocal evidence in our opinion for strong superconductivity on the @xmath3 and @xmath4 bands .
second , by measuring the dependence of @xmath14 on the impurity concentration , knowing the dependence of the superconducting gap on this quantity as well , one could estimate the relative roles of the intrinsic and extrinsic contributions .
following our earlier work @xcite , we analyze a two - orbital model in order to make clear the physics of the anomalous hall conductivity in a multiband superconductor at finite temperatures .
we will use this model later to describe the physics of @xmath0 since , although three orbitals are needed to account for the fermi surface properties @xcite , only two orbitals the ru @xmath1 and @xmath2 ones are directly relevant for the hall conductivity .
we consider the following bcs pairing hamiltonian describing superconductivity in a two - orbital system : h = _ ( cc c^_1 & c^_2 ) ( cc _ 1 ( ) & _ 12 ( ) + _ 12 ( ) & _ 2 ( ) ) ( c c_1 + c_2 ) + _ , _ , v_,(,)c^_-c^_c_c_-.[h]here , @xmath20 is the dispersion for the bloch states constructed from the 1(2 ) orbital , @xmath21 is the inter - orbital coupling , and @xmath22 describes the pairing interaction , allowing for the possibility of inter - orbital ( @xmath23 and @xmath24 ) as well as intra - orbital ( @xmath25 ) pairing . in the discussion to follow , we use @xmath26 and @xmath27 to denote the two intra - orbital order parameters while @xmath28 is the inter - orbital order parameter .
spin labels have been suppressed . in the basis set by the spinor , @xmath29 ,
the inverse mean - field @xmath30 green s function for this model is @xmath31 here , @xmath32 are the usual @xmath33 pauli matrices , @xmath34 ( @xmath35 ) is the real ( imaginary ) part of the intra- and inter - orbital order parameters , and @xmath36 is a fermi matsubara frequency .
the two branches of the bcs quasiparticle spectrum , @xmath37 and @xmath38 , found from the solution of @xmath39 , are @xmath40 - 4\epsilon_{12}\xi_2\mathrm{re}[\delta_{12}\delta^{*}_{11}]\nonumber\\&+2\epsilon^{2}_{12}\mathrm{re}[\delta_{11}\delta^{*}_{22}]-2\mathrm{re}[\delta_{11}\delta_{22}(\delta^{*}_{12})^2]+ ( \epsilon^2_{12}-\xi_{1}\xi_2)^2\big)\big]^{1/2}\big\ } .\end{aligned}\ ] ] in the situation ( relevant for @xmath0 ) that the fermi surface crosses the two sheets formed by the bare 1 and 2 orbital dispersions and @xmath41 @xmath42 , the minimum in the _ sum _ of the two bogoliubov quasiparticle bands , @xmath43 , is close to twice the inter - orbital hopping energy : ( e_-+e_+ ) ~2_12(k_xk_y bare dispersions @xmath44 , @xmath45 both vanish ; i.e. , where the fermi surface crosses the 1 and 2 sheets . in this case , as we will see in the following , the inter - orbital hopping energy scale plays a crucial role in determining the frequency dependence of the anomalous , intrinsic hall conductivity in a multiband chiral superconductor , with multiple order parameters spanning the bands .
the optical hall conductivity @xmath46 is defined in terms of the antisymmetric part of the @xmath47-@xmath48 current correlator @xmath49 by _ h ( ) -_0.[sigmah]the total current operator in the @xmath50 direction is given by @xcite @xmath51 , where _ i ( ) = ( cc v_i,11()_2 & v_i,12 ( ) _ 2 + v_i,12 ( ) _ 2 & v_i,22()_2 ) [ v]is the @xmath30 bare current vertex ( @xmath52 is the @xmath33 identity matrix ) . in the orbital basis , @xmath53 and @xmath54 . when there are multiple orbitals present , it is possible to have a nonzero anomalous hall conductivity evaluating the current correlator at the one - loop level , and so in the following we ignore vertex corrections : _
xy(0,_m ) = e^2t_,_n[_x ( ) _ 0(,_n)_y()_0(,_n+_m)].[pi0]here , @xmath55 is a bose matsubara frequency . in the case of a single orbital ( or multiple uncoupled orbitals ) , @xmath56 is purely diagonal and commutes with @xmath57 .
consequently , @xmath58 equals @xmath59 , and the one - loop value for the hall conductivity is zero , irrespective of details such as band anisotropy , pairing symmetry , or self - energy corrections @xcite . broken
time - reversal and translational symmetries are necessary but not sufficient conditions for a nonzero anomalous hall conductivity at this level .
it follows that for superconductivity on a single orbital , vertex corrections are crucial to having a nonzero anomalous hall conductivity .
goryo @xcite and lutchyn _ et al .
_ @xcite considered impurity - scattering vertex corrections for a model of superconductivity in @xmath0 assuming superconductivity takes place predominantly on the @xmath11 band ( or @xmath10 orbital ) . for a multiorbital superconductor , on the other hand , the one - loop contribution ( [ pi0 ] ) can be nonzero , and if so , should provide a major contribution to the anomalous hall effect in a clean superconductor .
this contribution is straightforwardly evaluated by analytically continuing to real frequencies , @xmath60 , to obtain the real and imaginary parts of the anomalous hall conductivity ( [ sigmah ] ) .
the complex conductivity is given by @xmath61\times\nonumber\\ & \frac{1}{e_{+}e_{-}}\left\{\frac{1\!-\!f(e_{+})\!-\!f(e_{-})}{(e_{-}\!+\!e_{+})[(e_{-}\!+\!e_{+})^2\!-\!(\omega+i\eta)^2 ] } + \frac{f(e_{+})\!-\!f(e_{-})}{(e_{+}\!-\!e_{-})[(e_{+}\!-\!e_{-})^2\!-\!(\omega+i\eta ) ^2]}\right\ } .
\label{sigmareal}\end{aligned}\ ] ] here , @xmath62 , and @xmath63 .
@xmath64 is an infinitesimal in the clean limit .
otherwise , we can make a crude estimate of the effects of impurity scattering by using a finite @xmath65 , where @xmath66 is the impurity scattering rate . for unconventional @xmath5-wave superconductivity , superconductivity
is easily destroyed by impurity scattering and one expects to be in the clean limit , @xmath67 , for samples with the highest @xmath68 . before discussing the implications of eq .
( [ sigmareal ] ) for @xmath0 , we point out some generic features relevant for any multiorbital chiral superconductor . at the one - loop level , due to the cancellation between the two terms in eq .
( [ sigmah ] ) , the only contributions to a nonzero @xmath14 arise from time - reversal symmetry - breaking inter - orbital transitions . the factor @xmath69 in eq .
( [ sigmareal ] ) indicates that one photon vertex involves an electron transition between different orbitals , while the other vertex involves an intra - orbital transition .
the two terms in the curly brackets of eq .
( [ sigmareal ] ) indicate that , in the quasiparticle basis , only processes which create or destroy two quasiparticles on _ different _ branches ( first term ) or which scatter a quasiparticle from one branch to the other ( second term ) contribute . the rest of eq .
( [ sigmareal ] ) , the terms in square brackets together with energy denominators , are the form factors , or coherence factors , that relate the quasiparticle states to the electron orbital states . in passing from eq .
( [ pi0 ] ) to eq .
( [ sigmareal ] ) , almost all such terms cancel , leaving only two very specific terms which carry the signature of the broken time reversal symmetry of the superconducting state .
the relevant terms contain a coupling between the orbitals , either through the hopping , @xmath21 , or through intra - orbital pairing , @xmath70 , that is essential to connect an inter - orbital vertex with an intra - orbital vertex .
the signature of broken time reversal symmetry comes into these terms through a product of different intra- and inter - orbital superconducting order parameters .
a nonzero _ relative _ phase of these order parameters , _ 11,22_11-_22(^*_11_22)/|_11||_22| , for example
, then gives rise to a nonzero anomalous hall conductivity .
consequently , the existence of multiple order parameters with different complex phases spanning multiple orbitals is a necessary condition to have a nonzero anomalous hall conductivity . beyond the fact that multiple orbitals are needed in order to have a nonzero anomalous hall conductivity at the one - loop level ( i.e. , no vertex corrections ) , eq .
( [ sigmareal ] ) shows that inter-_band _
cooper pairing is also a necessary condition .
although this expression shows the hall conductivity with respect to orbital degrees of freedom , it is straightforward to apply a unitary transformation to represent this in terms of band degrees of freedom . the second line of eq
( [ sigmareal ] ) remains the same since it is expressed in terms of the bogoliubov quasiparticle basis . on the other hand , the
velocities @xmath71 and @xmath72 will be different and the term in square brackets involving the coupling @xmath21 will vanish since the bands are , by definition , uncoupled .
this leaves the remaining terms in square brackets which , in the band basis , describe inter - band pairing , @xmath73 where @xmath74 are the band indices .
thus , inter - band pairing is needed to have a nonzero anomalous hall conductivity in a chiral superconductor at the one loop level . moreover , as in the orbital picture , the inter - band order parameter can not have the same phase as at least one of the intra - band order parameters . at zero temperature , only the first term in curly brackets in eq .
( [ sigmareal ] ) contributes to the hall conductivity ; it corresponds to a process wherein a photon destroys a cooper pair and creates two quasiparticles . the imaginary part of @xmath14 , which is the absorptive part , turns on at the minimum of @xmath43 .
as discussed above , this minimum is set by the inter - orbital coupling , @xmath21 , and , in general , is expected to be a much larger energy scale than the bcs quasiparticle gap @xmath75 .
the zero temperature anomalous hall conductivity was studied in ref .
@xcite and the absorptive part exhibits sharp features due to the van hove singularites in the quasiparticle density of states near the minimum of @xmath43 .
there are two effects of finite temperature : ( 1 ) to modify the contribution from the first term in curly brackets and ( 2 ) the new contribution represented by the second term in the curly brackets . if @xmath76 , the primary effect of ( 1 ) is to replace the zero temperature values of the @xmath77 s with their finite temperature values .
in particular , the sharp features emerging due to the van hove singularities near the minimum of @xmath43 ( see figs .
[ sigmaomegateq0fig ] and [ sigmaomegateq1pt4fig ] ) will not be appreciably smoothed out at nonzero temperatures .
this is because , relative to @xmath78 , the temperature , @xmath79 , is always very small in the superconducting phase where the anomalous hall effect arises .
for this same reason , contribution ( 2 ) , due to quasiparticle scattering , is also very small .
this contribution vanishes at @xmath80 , as well as at @xmath81 , and is largest at temperatures comparable to but less than @xmath68 . for most frequencies ,
this term is smaller than contribution ( 1 ) by a factor of order @xmath82 because the fermi factors restrict the momentum integral to a very narrow region where @xmath83 .
consequently , at most frequencies , and in particular at the high frequency relevant to current experiments , the anomalous hall conductivity is dominated by the first term in the curly brackets and the main effect of finite temperature is simply to replace the zero temperature @xmath77 with @xmath84 . as noted above
, the intrinsic hall conductivity arises from inter - orbital transitions when there are order parameters of different phases living on the orbitals .
the van hove singularities represent a resonance condition for these transitions . in this way , the detection of a strong increase in the hall conductivity at frequencies on the order of the inter - orbital hopping energy would provide strong evidence in favour of multiband / multiorbital cooper pairing in a chiral superconductor , such as has been proposed in @xmath0 @xcite .
in contrast , features in the hall conductivity arising at frequencies on the order of the gap , such as impurity scattering contributions which yield a nonzero hall conductivity in a single band system @xcite , are comparatively insensitive to whether or not superconductivity is a multiband phenomenon .
having used a two - orbital model to elucidate the basic physics of the anomalous intrinsic hall effect in a multiorbital chiral superconductor , we now discuss the implications for @xmath0 .
although @xmath0 is a three - band system , in calculating the intrinsic anomalous hall conductivity , @xmath14 , an important simplification arises because , to a very good approximation , the @xmath85 and @xmath86 components of the current operator do not couple @xmath10 to the other two orbitals and , hence , only @xmath18 inter - orbital transitions contribute to its @xmath14 .
consequently , the conclusions reached from our two orbital model are still valid .
in particular , the intrinsic , anomalous hall effect in @xmath0 is only nonzero when there is complex inter - band pairing and this pairing must involve the @xmath1 and @xmath2 orbitals to a significant extent .
higher order spin - orbit coupling , of order @xmath87 ( where @xmath88 is the spin - orbit coupling parameter ; see e.g. ref .
@xcite ) , and interlayer hopping involving @xmath10 orbitals , will bring in the @xmath10 orbital , but these are are sufficiently small that they can be ignored .
consequently , the @xmath10 orbital only plays a passive role in determining the hall conductivity .
in particular , the terms outside the curly brackets in eq .
( [ sigmareal ] ) would be unchanged if we used a three - orbital model of @xmath0 , but neglected interlayer hopping and higher order spin - orbit effects .
only the quasiparticle dispersions would be modified and would renormalize our estimate of the hall conductivity somewhat .
thus , we simply ignore the @xmath10 orbital and use our two - orbital result for the hall conductivity , ( [ sigmareal ] ) , as this gives a reasonable order of magnitude estimate .
building on our previous work @xcite where we neglected next - nearest neighbour intra - orbital hopping , we now include this effect , taking @xmath89 , @xmath90 , and @xmath91 .
we use parameters from a recent lda study of the band structure of @xmath0 @xcite : @xmath92ev , @xmath93 , and @xmath94 .
we also fix the chemical potential to be @xmath95 , consistent with this study .
following ref .
@xcite , we assume purely intra - orbital pairing on the @xmath1 and @xmath2 orbitals . as a simple ansatz , we use _
11(,t ) = _ 0(t)(k_x)(k_y ) , _
22(,t ) = i_0(t)(k_y)(k_x ) , _ 12(,t)=0.[ansatz ] , defined in eq .
( [ delta0 ] ) . , scaledwidth=45.0% ] here , @xmath96 is the purely real temperature - dependent amplitude of the order parameters , assumed to be equal for both . crucially in order to obtain a nonzero hall conductivity , the two order parameters have different phases , meaning that @xmath97 and hence , @xmath14 , is nonzero .
mineev @xcite has argued that a model of the form shown in eq .
( [ ansatz ] ) , based on a quasi-1d approach , is not energetically stable in the case relevant to @xmath0 , where @xmath98 .
however , since @xmath99 , @xmath100 only affects a very small region of the fermi surface , while @xmath101 operates over the entire fermi surface . as explained by chung _
et al . _
@xcite , this allows the quasi-1d approach to remain valid even for @xmath102 .
while we use a somewhat simplified model , more realistic details , such as allowing different pairing amplitudes on the @xmath3 and @xmath4 bands would not change our results in any significant way .
_ the key property of the model is that it includes . _ as stressed in ref .
@xcite , this is a necessary condition for a non - zero intrinsic hall effect , a point also emphasized by mineev @xcite . to see the structure of the purely intraorbital order parameter , eq .
( [ ansatz ] ) in the band basis , we diagonalize the single - particle contribution to the hamiltonian , eq .
( [ h ] ) , into the @xmath3 and @xmath4 bands . in this basis , there are three order parameters : two intra - band , _ = , _ = , and one inter - band , _ = _ = , order parameters .
as expected , the inter- and intra - band order parameters have different phases and hence , the hall conductivity will be nonzero , as we already inferred from the fact that the phases of the order parameters in the orbital basis were also different .
it is interesting to note that in the hot zone " @xmath103 ( with @xmath104 ) which delineates the regions of the brillouin zone that dominate the conductivity integral in eq .
( [ sigmareal ] ) , these order parameters have approximately opposite chirality : _ _ /2and _ /2 .
turning now to the task of evaluating numerically the hall conductivity , we assume a separable potential , @xmath105 where @xmath106 , and use this to determine the amplitude @xmath96 of the order parameter .
the resulting bcs gap equation is = t_,_nk_x k_y g_0,12(,_n).[delta0]here , @xmath107 is the @xmath108 element of the @xmath30 matrix green s function , the inverse of which is given by eq .
( [ g0 ] ) .
we choose the dimensionless coupling constant ( restoring the crystalline spacing @xmath109 but keeping @xmath110 ) @xmath111 to be @xmath112 in order to have a critical temperature @xmath113 @xcite , appropriate for ultraclean @xmath0 samples .
we plot @xmath96 in fig .
[ gapfig ] .
note that the @xmath80 value , @xmath114mev , is larger than the canonical weak - coupling @xmath115-wave bcs - limiting value , @xmath116mev , because of anisotropy . in figs .
[ sigmaomegateq0fig ] and [ sigmaomegateq1pt4fig ] , we make use of the calculated @xmath96 in eq .
( [ sigmareal ] ) to plot the frequency - dependent hall conductivity at @xmath80 and @xmath117 , respectively in the ultraclean limit , @xmath118 .
there is very little difference between the two plots apart from an overall scale factor : @xmath119 .
this can be seen explicitly in the inset of figs .
[ resigmatfig ] and [ imsigmatfig ] which show the temperature dependence of the real and imaginary parts of the hall conductivity ( at @xmath13ev ) , respectively , divided @xmath120 .
one sees that this quantity is essentially independent of temperature .
this result , _
h(,t ) ^2_0(t)2t^,[propid]where the proportionality factor is independent of temperature , holds for all frequencies except those very close to the van hove singularity in the vicinity of @xmath121 . in the main panels of figs .
[ resigmatfig ] and [ imsigmatfig ] , we plot the temperature dependence of the real and imaginary parts of the conductivity . .
note the different scale as compared to fig .
[ sigmaomegateq0fig ] . ] .
note the different scale as compared to fig .
[ sigmaomegateq0fig ] . ] the fact that the hall conductivity is approximately proportional to the square of the temperature - dependent order parameter ( at least away from the van hove singularities ) accounts for its smallness . dividing the calculated hall conductivity by the relevant dimensionless energy scale @xmath122
, we see that the zero - frequency and temperature hall conductivity @xmath123 is approximately unity ( in natural units of @xmath124 ) .
it is only the smallness of the order parameter in @xmath0 that makes the anomalous hall conductivity as small as it is .
equation ( [ propid ] ) also suggests that useful information about the origin of the anomalous hall effect in @xmath0 can come from looking at the impurity concentration dependence of the hall conductivity in @xmath0 .
@xmath68 in non-@xmath115-wave superconductors is strongly suppressed by non - magnetic impurities @xcite . in @xmath0 ,
this dependence has been studied in the group of mackenzie @xcite . assuming that @xmath125 , measurement of the impurity concentration dependence of the hall conductivity ( or kerr angle , as discussed below ) in conjunction with the impurity concentration dependence of @xmath68 would help reveal the relative importance of the intrinsic and extrinsic contributions to @xmath14 .
inset : temperature dependence of the imaginary part of the hall conductivity at @xmath13ev divided by @xmath120 . ] ev .
inset : temperature dependence of the imaginary part of the hall conductivity at @xmath13ev divided by @xmath120 . ] figures [ sigmaomegateq0fig ] and [ sigmaomegateq1pt4fig ] also clearly exhibit the sharp features anticipated in sec .
iii , arising from the van hove singularities near min@xmath126\sim 2t^{\prime}\gg \delta_0 $ ] .
as discussed there , these are not at all washed out at finite temperatures since , relative to this energy scale ( @xmath127mev ) , the temperature is very small in the superconducting phase ( @xmath128mev ) .
we have also checked the effect of adding a small amount of impurity scattering , taking @xmath129 ( corresponding to @xmath130 ) and , unsurprisingly given this small broadening ( recall that @xmath131 ) , find very little difference from the results shown in figs .
[ sigmaomegateq0fig ] and [ sigmaomegateq1pt4fig ] .
notably , the resonance features remain intact .
if measured , these strong features , arising as they do from inter - orbital transitions involving the creation of bcs quasiparticles from multiple complex order parameters spanning these orbitals , would constitute strong evidence for multiband chiral superconductivity .
, of the imaginary part of the contribution , @xmath132 , to the hall conductivity from quasiparticle scattering ( the second term in curly brackets in eq .
( [ sigmareal ] ) ) . ] , of the imaginary part of the contribution , @xmath132 , to the hall conductivity from quasiparticle scattering ( the second term in curly brackets in eq .
( [ sigmareal ] ) ) . ] the second term in eq .
( [ sigmareal ] ) adds very little to the hall conductivity . for our parameters
, we find this contribution ( @xmath133 ) to be always much smaller than the contribution coming from the first term and do not find the coherence features found by wysokiski , annett , and gyrffy at low frequencies , @xmath134 @xcite .
such features only arise in our calculations by artificially inflating the order parameter ( and hence @xmath68 and crucially , the temperature ) to energies on the order of several mevs . due to the thermal fermi factor , @xmath135 , which heavily suppresses all momenta except where @xmath136 is small
, @xmath137 will always be much smaller than @xmath46 unless @xmath138 and hence , @xmath139 , is appreciable .
( note , @xmath140 and @xmath141 . ) because of the fermi factor , the largest contribution to @xmath137 will come when @xmath142 , with @xmath143 and @xmath144 . in fig .
[ secondtermfig ] , we plot the frequency - dependence of @xmath145 at @xmath146 showing this strong frequency dependence @xcite .
to better understand the temperature dependence of this contribution , in fig .
[ secondtermtfig ] , we also plot the integrated spectral weight ( defined to be positive ) s_t-^_0d^t_h ( ) = -e^2 _ [ sumrule]of @xmath147 in the clean limit @xmath148 as a function of temperature .
as expected , the low - temperature spectral weight from scattering @xmath149 bcs quasiparticles to @xmath150 is exponentially suppressed .
we now turn our attention to the kerr angle , the quantity that has been measured in experiments @xcite and which provides an indirect measure of the hall conductivity . the frequency - dependent kerr angle @xmath151 is given by eq .
( [ kerr ] ) .
( see , for instance , ref .
in addition to depending on the frequency - dependent hall conductivity , @xmath46 , the kerr angle also depends on the diagonal element of the optical conductivity tensor , @xmath152 , through its dependence on the complex index of refraction @xmath17 : ( ) = , [ alpha]where @xmath153 is equal to the square root of the component of the permeability tensor , _ ab = _ + ( 4i/ ) ( ) , [ permeability ] in the @xmath154 plane .
here , @xmath155 is the background dielectric tensor . .
the solid line shows the kerr angle for an inverse transport lifetime 0.4ev ; the dashed line corresponds to an inverse transport lifetime 0.2ev .
the large difference between the two results shows the sensitivity of the kerr angle to material parameters beyond the hall conductivity .
note the large resonant feature at @xmath156ev arising from the van - hove singularity in the hall conductivity ( see fig .
[ sigmaomegateq0fig ] ) .
, scaledwidth=50.0% ] having calculated the hall conductivity , we can combine this calculation with experimental conductivity data to determine the frequency - dependent kerr angle .
the conductivity in @xmath0 seems to be well approximated by a generalized drude expression with a frequency - dependent effective mass and transport scattering rate @xcite . for simplicity , however , we use a drude model with constant transport scattering rate @xmath11 : ( ) = - [ sigmadrude ] .this is a reasonable approximation since the transport scattering rate varies relatively slowly over a large frequency range ( see fig .
( 4b ) in ref .
@xcite ) , from approximately 2.5 ev@xmath157 at @xmath158ev to 4.5 ev/@xmath159 at @xmath160ev at low temperature .
this variation is small in comparison with that exhibited by the hall conductivity over the same range and we simply use the value at 0.8ev ( the frequency at which the kerr angle is measured in ref .
@xcite ) to compute the frequency - dependent kerr angle : @xmath161ev/@xmath159 .
the other parameters needed to calculate the kerr angle are also extracted from experiments on @xmath0 .
following ref .
@xcite we take the interlayer spacing @xmath162 , @xmath163 , and @xmath164 . using these , in fig .
[ kerrfig ] we plot the @xmath80 kerr angle @xmath151 as a function of frequency using our calculated @xmath80 hall conductivity ( see fig .
[ sigmaomegateq0fig ] ) .
the results are shown for two values of the transport scattering rate , the experimental value " of @xmath165ev/@xmath159 ( extracted at @xmath166ev ) and a somewhat smaller value of 0.2ev/@xmath159 . using the former value
, we find @xmath167nrads at @xmath168ev , while the latter value gives @xmath169nrads .
as one sees from the plot , in the vicinity of the plasma edge , the precise value of the kerr angle is very sensitive to material parameters such as the transport scattering rate . setting the scattering rate to zero , as is sometimes done , is not a good approximation because of the significant inelastic scattering at the frequencies of interest . over the frequency range of interest
, the temperature dependence of the kerr angle will track that of the hall conductivity ( see figs .
[ resigmatfig ] and [ imsigmatfig ] ) , which in turn is approximately proportional to the square of the order parameter .
in this paper , we have calculated the anomalous hall conductivity in a simple two - orbital model of a chiral superconductor , with an eye to understanding the kerr effect in @xmath0 @xcite .
although @xmath0 is a three - orbital system , we have argued that the essential physics responsible for the anomalous hall effect arises from only the ru @xmath1 and @xmath2 orbitals , since these are the only orbitals that are optically coupled .
this fact gives an important experimental handle in understanding _ where _ superconductivity arises in @xmath0 : in contrast to the impurity - scattering extrinsic contribution to the hall conductivity @xcite , there must be significant pairing on the @xmath3 and @xmath4 bands in order to have an intrinsic anomalous hall conductivity .
the smoking gun " for such superconductivity would be a large resonant feature in the hall conductivity at a frequency close to the twice the inter - orbital hopping scale @xmath170ev , as clearly seen in figs .
[ sigmaomegateq0fig ] and [ sigmaomegateq1pt4fig ] ( also in fig .
[ kerrfig ] for the kerr angle ) .
this feature is robust both to nonzero temperature as well as impurity broadening effects since both these energy scales are on the order of @xmath101 , much smaller than @xmath171 . to make contact with the kerr rotation experiment @xcite , using our two - orbital model in conjunction with parameters appropriate for @xmath0
, we have arrived at the result @xmath172nrads for the kerr angle at @xmath80 .
this is of the same order of magnitude but smaller than the experimental value of @xmath173nrads @xcite .
although the inclusion of the @xmath10 orbital would not lead to any qualitatively new physics , it can lead to a quantitative change to this prediction .
we also note that the @xmath12 hopping parameters used in our calculations were obtained from a microscopic _ three_-orbital model @xcite including spin - orbit coupling ( soc ) , which we have also excluded . as we showed in ref .
@xcite , soc does not lead to any hall effect unless there is a zeeman splitting @xmath174 between the two spin species .
its only effect otherwise is to renormalize the dispersions that enter eq .
( [ sigmareal ] ) . in comparison to our earlier work @xcite , where we predicted a @xmath80 kerr angle of @xmath175nrads using a two - orbital model without next - nearest neighbour hopping ,
the present work produces a smaller kerr angle , resulting from a smaller prediction for the hall conductivity .
this difference is due almost entirely to the form for the order parameter given in eq .
( [ ansatz ] ) , appropriate for next - nearest neighbour hopping , an effect not included in our earlier calculations .
the additional cosine factors in the order parameters lead to a significant reduction in the hall conductivity since the integrand in eq .
( [ sigmareal ] ) is dominated by a small region in the vicinity of @xmath176 ( see eq .
( [ qpmin ] ) ) .
this reduction is offset somewhat by the larger value of @xmath177 ( where @xmath178k ) used in the present work as a result of solving the bcs gap equation appropriate for our model , instead of using the @xmath115-wave bcs result @xmath179 although the kerr angle is additionally very sensitive to the low - temperature optical properties of @xmath0 , and more information about these are needed to reliably make contact with experiment , there are a few ways in which experiments should be able to distinguish between the two contributions extrinsic @xcite and intrinsic to the hall conductivity identified by theory .
first , as we have noted in this paper , the intrinsic mechanism will exhibit a strong frequency dependence in the vicinity of twice the inter - orbital hopping , around 40 mevs .
unfortunately , this seems to be outside the range of the current experimental setup at stanford @xcite .
secondly , experiments could probe the dependence of the kerr angle on the impurity concentration . to a good approximation ,
the intrinsic mechanism discussed here leads to a hall conductivity that depends on the impurity concentration only through its approximately quadratic dependence on the order parameter , eq .
( [ propid ] ) . in turn , the impurity dependence of @xmath68 , and hence , @xmath101 , has been studied @xcite .
in contrast , the extrinsic contribution to the hall conductivity @xcite will exhibit an additional , stronger dependence on the impurity concentration . in summary , a more detailed understanding of the anomalous hall conductivity in @xmath0 ,
either by probing its frequency dependence or its dependence on the concentration of non - magnetic impurities , would contribute significantly to understanding this material .
in particular , the hall conductivity is sensitive to the way in which the different bands crossing the fermi energy of @xmath0 enter into the chiral @xmath5-wave order parameter .
we hope that the present work stimulates further experimental studies of the anomalous hall effect in @xmath0 as well as in any other materials that might exhibit chiral superconductivity @xcite .
we thank steve kivelson , sri raghu , and ronny thomale for useful discussions .
this work was supported by nserc and the canadian institute for advanced research .
9 maeno y , hashimolo h , yoshida k , nishizaki s. fujita t , bednorz i g and lichtenberg f 1994 _ nature _ * 372 * 532 .
rice t and sigrist m 1995 _ j. phys . : condens .
matter _ * 7 * l643 .
baskaran g 1996 _ physica b _ * 224 * 490 .
kallin c 2012 _ rep .
phys . _ * 75 * , 042501 .
maeno y , kittaka s , nomura t , yonezawa s and ishida k 2011 _ j. phys .
jpn . _ * 81 * 011009 .
mackenzie a and maeno y 2003 _ rev .
phys . _ * 75 * 657 . read n and green d 2000 _ phys . rev .
b _ * 61 * 10267 .
puetter c m and kee h - y 2012 _ eur .
lett . _ * 98 * 27010 .
haverkort m w , elfimov i s , tjeng l h , sawatzky g a , and damascelli a 2008 _ phys .
lett . _ * 101 * 026406 .
agterberg d f , rice t m , and sigrist m 1997 _ phys .
lett . _ * 78 * 3374 .
zhitomirsky m e and rice t m 2001 _ phys .
lett . _ * 87 * 057001 .
nomura t and yamada k 2002 _ j. phys .
* 71 * 404 .
deguchi k , mao z q and maeno y 2004 _ j. phys .
jpn . _ * 73 * 1313 .
takimoto t 2000 _ phys .
b _ * 62 * r14641 .
kuroki k , ogata m , arita r and aoki h 2001 _ phys .
b _ * 63 * 060506(r ) .
annett j f , litak g , gyrffy b l and wysokiski k i 2002 _ phys .
b _ * 66 * 134514 .
raghu s , kapitulnik a and kivelson s a 2010 _ phys .
lett . _ * 105 * 136401 .
stone m and roy r 2004 _ phys .
b _ * 69 * 184511 .
kirtley j r , kallin c , hicks c w , kim e - a , liu y , moler k a , maeno y and nelson k d 2007 _ phys .
b _ * 76 * 014526 .
kallin c and berlinsky a j 2009 _ j. phys . : condens
. matter _ * 21 * 164210 .
xia j , maeno y , beyersdorf pt , fejer m m and kapitulnik a 2006 _ phys .
lett . _ * 97 * 167002 .
kapitulnik a , xia j , schemm e and palevski a 2009 _ new j. phys . _
* 11 * 055060 .
argyres p n 1955 _ phys .
rev . _ * 97 * 334 .
chung s b , raghu s , kapitulnik a and kivelson s a 2012 _ phys .
b _ * 86 * 064525 .
roy r and kallin c 2008 _ phys .
b _ * 77 * 174513 . goryo j 2008 _ phys .
b _ * 78 * 060501(r ) .
lutchyn r m , nagornykh p and yakovenko v m 2009 _ phys .
b _ * 80 * 104508 .
taylor e and kallin c 2012 _ phys .
lett . _ * 108 * 157001 .
wysokiski k i , annett j f and gyrffy b l 2012 _ phys .
lett . _ * 108 * 077004 .
yip s k and sauls j a 1992 _ j. low t. phys . _ * 86 * 257 .
although we suppress spin in our model , the hall conductivity receives equal contributions from both spins .
the current operator includes contributions from both particle and hole contributions , mimicking spin counting .
hence , the hall conductivity calculated from this spin operator correctly accounts for two spins .
rozbicki e j , annett j f , souquet j r and mackenzie a p 2011 _ j. phys . : condens . matter _ * 23 * 094201 .
note that our factor of @xmath100 is equivalent to twice @xmath180 in this reference .
mineev v i 2012 _ j. phys . soc .
_ * 81 * 093703 .
balian r and werthamer n r 1963 _ phys .
rev . _ * 131 * 1553 .
mackenzie a p , haselwimmer r k w , tyler a w , lonzarich g g , mori y , nishizaki s and maeno y 1998 _ phys .
lett . _ * 80 * 161 .
mackenzie a p , haselwimmer r k w , tyler a w , lonzarich g g , mori y , nishizaki s and maeno y 1998 _ phys . rev .
lett . _ * 80 * 3890 . for @xmath181 ,
the @xmath182 integrand is _ very _ strongly peaked about the fermi surface and it is difficult to carry out the integration to high accuracy .
although the primary feature visible at @xmath183 in fig . [ secondtermfig ] is robust , the details of the other features above this are only approximate .
katsufuji t , kasai m and tokura y 1996 _ phys .
lett . _ * 76 * 126 .
nandkishore r , levitov l s and chubukov a v 2012 _ nat .
phys . _ * 8 * 158 . | building on previous work , we calculate the temperature- and frequency - dependent _ anomalous _ hall conductivity for the putative multiband chiral superconductor @xmath0 using a simple microscopic two - orbital model without impurities .
a hall effect arises in this system without the application of an external magnetic field due to the time - reversal - symmetry breaking chiral superconducting state .
the anomalous hall conductivity is nonzero only when there is more than one superconducting order parameter , involving inter- as well as intra - band cooper pairing .
we find that such a multiband superconducting state gives rise to a distinctive resonance in the frequency - dependence of the hall conductivity at a frequency close to the inter - orbital hopping energy scale that describes hopping between ru @xmath1 and @xmath2 orbitals .
the detection of this feature , robust to temperature and impurity effects in the superconducting phase , would thus constitute compelling evidence in favour of a multiband origin of superconductivity in @xmath0 , with strong superconductivity on the @xmath3 and @xmath4 bands .
the temperature dependence of the hall conductivity and kerr rotation angle are studied within this model at the one - loop approximation . 0 * 0 * |
the spectral functions ( spfs ) of hadrons play a special role in physical observables in qcd ( see the examples in @xcite ) .
however , the lattice qcd simulations so far have difficulties in accessing the dynamical quantities in the minkowski space , because measurements on the lattice can only be carried out for discrete points in imaginary time . the analytic continuation from the imaginary time to the real time using the noisy lattice data
is highly non - trivial and is even classified as an ill - posed problem .
recently we made a first serious attempt to extract spfs of hadrons from lattice qcd data without making a priori assumptions on the spectral shape @xcite .
we use the maximum entropy method ( mem ) , which has been successfully applied for similar problems in quantum monte carlo simulations in condensed matter physics , image reconstruction in crystallography and astrophysics , and so forth @xcite . in this report
, we present the results for the pseudo - scalar ( ps ) and vector ( v ) channels at @xmath0 using the continuum kernel and the lattice kernel of the integral transform .
the latter analysis has not been reported in @xcite .
the euclidean correlation function @xmath1 of an operator @xmath2 and its spectral decomposition at zero three - momentum read @xmath3 where @xmath4 , @xmath5 is a real frequency , and @xmath6 is spf ( or sometimes called the _ image _ ) , which is positive semi - definite .
the kernel @xmath7 is proportional to the fourier transform of a free boson propagator with mass @xmath5 : at @xmath0 in the continuum limit , @xmath8 .
monte carlo simulation provides @xmath9 on the discrete set of temporal points @xmath10 . from this data with statistical noise
, we need to reconstruct the spectral function @xmath6 with continuous variable @xmath5 .
this is a typical ill - posed problem , where the number of data is much smaller than the number of degrees of freedom to be reconstructed .
this makes the standard likelihood analysis and its variants inapplicable @xcite unless strong assumptions on the spectral shape are made .
mem is a method to circumvent this difficulty through bayesian statistical inference of the most probable _ image _ together with its reliability @xcite .
mem is based on the bayes theorem in probability theory : @xmath11 = p[y|x]p[x]/p[y]$ ] , where @xmath11 $ ] is the conditional probability of @xmath12 given @xmath13 . the most probable image @xmath14 for given lattice data @xmath15
is obtained by maximizing the conditional probability @xmath16 $ ] , where @xmath17 summarizes all the definitions and prior knowledge such as @xmath18 . by the bayes theorem , @xmath19 \propto p[d|ah]p[a|h ] , \ ] ]
where @xmath20 $ ] ( @xmath21 $ ] ) is called the likelihood function ( the prior probability ) .
for the likelihood function , the standard @xmath22 is adopted , namely @xmath20= z_l^{-1 } \exp ( -l)$ ] with @xmath23 & & \mbox{\hspace{2.5 cm } } \times \ , c^{-1}_{ij } ( d(\tau_j)-d^a(\tau_j)).\nonumber \label{chi2}\end{aligned}\ ] ] @xmath24 is a normalization factor given by @xmath25 with @xmath26 .
@xmath27 is the lattice data averaged over gauge configurations and @xmath28 is the correlation function defined by the right hand side of ( [ ka ] ) .
@xmath29 is an @xmath30 covariance matrix of the data with @xmath31 being the number of temporal points to be used in the mem analysis .
the lattice data have generally strong correlations among different @xmath32 s , and it is essential to take into account the off - diagonal components of @xmath29 .
axiomatic construction as well as intuitive `` monkey argument '' @xcite show that , for positive distributions such as spf , the prior probability can be written with parameters @xmath33 and @xmath34 as @xmath35= z_s^{-1 } \exp ( \alpha s)$ ] . here @xmath36 is the shannon - jaynes entropy , @xmath37 d\omega .\nonumber\end{aligned}\ ] ] @xmath38 is a normalization factor : @xmath39 .
@xmath33 is a real and positive parameter and @xmath40 is a real function called the default model . in the state - of - art mem @xcite , the output image @xmath41
is given by a weighted average over @xmath42 and @xmath33 : @xmath43p[\alpha|dhm ] \
{ \cal d } a \
d\alpha \nonumber \\ & & \simeq \int a_{\alpha}(\omega ) \
p[\alpha|dhm ] \ d\alpha . \label{final}\end{aligned}\ ] ] here @xmath44 is obtained by maximizing the `` free - energy '' @xmath45 for a given @xmath33 .
here we assumed that @xmath46 $ ] is sharply peaked around @xmath44 .
@xmath33 dictates the relative weight of the entropy @xmath36 ( which tends to fit @xmath42 to the default model @xmath34 ) and the likelihood function @xmath47 ( which tends to fit @xmath42 to the lattice data ) .
note , however , that @xmath33 appears only in the intermediate step and is integrated out in the final result .
our lattice data show that the weight factor @xmath48 $ ] , which is calculable using @xmath49 @xcite , is highly peaked around its maximum @xmath50 .
we have also studied the stability of the @xmath51 against a reasonable variation of @xmath52 .
the non - trivial part of the mem analysis is to find a global maximum of @xmath49 in the functional space of @xmath6 , which has typically 750 degrees of freedom in our case .
we have utilized the singular value decomposition ( svd ) of the kernel to define the search direction in this functional space .
the method works successfully to find the global maximum within reasonable iteration steps .
to check our mem code and to see the dependence of the mem image on the quality of the data , we made the following test using mock data .
( i ) we start with an input image @xmath53 in the @xmath54-meson channel which simulates the experimental @xmath55 cross section
. then we calculate @xmath56 from @xmath57 using eq.([ka ] ) .
( ii ) by taking @xmath58 at @xmath31 discrete points and adding a gaussian noise , we create a mock data @xmath59 .
the variance of the noise @xmath60 is given by @xmath61 with a parameter @xmath62 , which controls the noise level @xcite .
( iii ) we construct the output image @xmath63 using mem with @xmath64 and compare the result with @xmath57 . in this test
, we have assumed that @xmath29 is diagonal for simplicity . in fig.1 , we show @xmath65 , and @xmath66 for two sets of parameters , ( i ) and ( ii ) .
as for @xmath34 , we choose a form @xmath67 with @xmath68 , which is motivated by the asymptotic behavior of @xmath42 in perturbative qcd , @xmath69 .
the final result is , however , insensitive to the variation of @xmath70 even by factor 5 or 1/5 .
the calculation of @xmath51 has been done by discretizing the @xmath5-space with an equal separation of 10 mev between adjacent points .
this number is chosen for the reason we shall discuss below .
the comparison of the dashed line ( set ( i ) ) and the dash - dotted line ( set ( ii ) ) shows that increasing @xmath71 and reducing the noise level @xmath62 lead to better spfs closer to the input spf .
we have also checked that mem can nicely reproduce other forms of the mock spfs .
in particular , mem works very well to reproduce not only the broad structure but also the sharp peaks close to the delta - function as far as the noise level is sufficiently small .
to apply mem to actual lattice data , quenched lattice qcd simulations have been done with the plaquette gluon action and the wilson quark action by the open milc code with minor modifications @xcite .
the lattice size is @xmath72 with @xmath73 , which corresponds to @xmath74 fm ( @xmath75 gev ) , @xmath76 @xcite , and the spatial size of the lattice @xmath77 fm .
gauge configurations are generated by the heat - bath and over - relaxation algorithms with a ratio @xmath78 .
each configuration is separated by 1000 sweeps .
hopping parameters are chosen to be @xmath79 0.153 , 0.1545 , and 0.1557 with @xmath80 for each @xmath81 .
for the quark propagator , the dirichlet ( periodic ) boundary condition is employed for the temporal ( spatial ) direction .
we have also done the simulation with periodic boundary condition in the temporal direction and obtained qualitatively the same results . to calculate the two - point correlation functions ,
we adopt a point - source at @xmath82 and a point - sink averaged over the spatial lattice - points .
we use data at @xmath83 for the dirichlet ( periodic ) boundary condition in the temporal direction . to avoid the known pathological behavior of the eigenvalues of @xmath29 @xcite , we take @xmath84 .
we define spfs for the ps and v channels as @xmath85 so that @xmath86 approaches a finite constant as predicted by perturbative qcd . for the mem analysis
, we need to discretize the @xmath5-integration in ( [ ka ] ) .
since @xmath87 ( the mesh size ) @xmath88 should be satisfied to suppress the discretization error , we take @xmath87 = 10 mev . @xmath89 ( the upper limit for the @xmath5 integration ) should be comparable to the maximum available momentum on the lattice : @xmath90 gev .
we have checked that larger values of @xmath91 do not change the result of @xmath6 substantially , while smaller values of @xmath91 distort the high energy end of the spectrum .
the dimension of the image to be reconstructed is @xmath92 , which is in fact much larger than the maximum number of monte carlo data @xmath93 . in fig.2 ( a ) and ( b ) , we show the reconstructed images for each @xmath81 in the case of the dirichlet boundary condition . here
we use the continuum kernel @xmath94 in the laplace transform . in these figures , we have used @xmath95 with @xmath96 for ps ( v ) channel motivated by the perturbative estimate of @xmath70 ( see eq.([cont - v ] ) and the text below ) .
we have checked that the result is not sensitive , within the statistical significance of the image , to the variation of @xmath70 by factor 5 or 1/5 .
the obtained images have a common structure : the low - energy peaks corresponding to @xmath97 and @xmath54 , and the broad structure in the high - energy region . from the position of the pion peaks in fig.2(a ) , we extract @xmath98 , which is consistent with @xmath99 @xcite determined from the asymptotic behavior of @xmath1 . the mass of the @xmath54-meson in the chiral limit extracted from the peaks in fig.2(b ) reads @xmath100 .
this is also consistent with @xmath101 @xcite determined by the asymptotic behavior .
although our maximum value of the fitting range @xmath102 marginally covers the asymptotic limit in @xmath32 , we can extract reasonable masses for @xmath97 and @xmath54 .
the width of @xmath97 and @xmath54 in fig.2 is an artifact due to the statistical errors of the lattice data .
in fact , in the quenched approximation , there is no room for the @xmath54-meson to decay into two pions . as for the second peaks in the ps and
v channels , the error analysis discussed in fig.4 shows that their spectral shape " does not have much statistical significance , although the existence of the non - vanishing spectral strength is significant . under this reservation , we fit the position of the second peaks and made linear extrapolation to the chiral limit with the results , @xmath103 for the ps ( v ) channel .
these numbers should be compared with the experimental values : @xmath104 , and @xmath105 or @xmath106 .
one should remark here that , in the standard two - mass fit of @xmath1 , the mass of the second resonance is highly sensitive to the lower limit of the fitting range , e.g. , @xmath107 for @xmath108 in the @xmath109 channel with @xmath110 @xcite .
this is because the contamination from the short distance contributions from @xmath111 is not under control in such an approach . on the other hand
, mem does not suffer from this difficulty and can utilize the full information down to @xmath112 .
therefore , mem opens a possibility of systematic study of higher resonances with lattice qcd data .
= 8.2 cm as for the third bumps in fig.2 , the spectral
shape " is statistically not significant as is discussed in fig.4 , and they should rather be considered a part of the perturbative continuum instead of a single resonance .
fig.2 also shows that spf decreases substantially above 6 gev ; mem automatically detects the existence of the momentum cutoff on the lattice @xmath113 .
it is expected that mem with the data on finer lattices leads to larger ultraviolet cut - offs in the spectra .
the height of the asymptotic form of the spectrum at high energy is estimated as @xmath114 the first two factors are the @xmath115 continuum expected from perturbative qcd .
the third factor contains the non - perturbative renormalization constant for the lattice composite operator .
we adopt @xmath116 determined from the two - point functions at @xmath117 = 6.0 @xcite together with @xmath118 and @xmath119 . our estimate in eq.([cont - v ] ) is consistent with the high energy part of the spectrum in fig.2(b ) after averaging over @xmath5 .
we made a similar estimate for the ps channel using @xmath120 @xcite and obtained @xmath121 .
this is also consistent with fig .
we note here that an independent analysis of the imaginary time correlation functions @xcite also shows that the lattice data at short distance is dominated by the perturbative continuum . in fig.3(a ) and ( b ) , the results using the lattice kernel @xmath122 are shown .
@xmath122 is obtained from the free boson propagator on the lattice .
it reduces to @xmath123 when @xmath124 .
the other parameters and boundary conditions are the same with fig.2(a , b ) .
the difference of fig.2 and fig.3 can be interpreted as a systematic error due to the finiteness of the lattice spacing @xmath125 .
the statistical significance of the reconstructed image can be studied by the following procedure @xcite . assuming that @xmath46 $ ] has a gaussian distribution around the most probable image @xmath126 , we estimate the error by the covariance of the image
, @xmath127 , where @xmath128 is a functional derivative and @xmath129 is an average over a given energy interval .
the final error for @xmath41 is obtained by averaging the covariance over @xmath33 with a weight factor @xmath130 $ ] . shown in fig.4
is the mem image in the v channel for @xmath131 with errors obtained in the above procedure .
the height of each horizontal bar is @xmath132 in each @xmath5 interval .
the vertical bar indicates the error of @xmath132 .
the small error for the lowest peak in fig.4 supports our identification of the peak with @xmath54 .
although the existence of the non - vanishing spectral strength of the 2nd peak and 3rd bump is statistically significant , their spectral `` shape '' is either marginal or insignificant .
lattice data with better quality are called for to obtain better spfs .
we have made a first serious attempt to reconstruct spfs of hadrons from lattice qcd data .
we have used mem , which allows us to study spfs without making a priori assumption on the spectral shape .
the method works well for the mock data and actual lattice data .
mem produces resonance and continuum - like structures in addition to the ground state peaks .
the statistical significance of the image can be also analyzed .
better data with finer and larger lattice will produce better images with smaller errors , and our study is a first attempt towards this goal .
there are many problems which can be explored by mem combined with lattice qcd data .
some of the applications in the baryon excited states , hadrons at finite temperature , and heavy quark systems will be reported in future publications @xcite .
we appreciate milc collaboration for their open codes for lattice qcd simulations , which has enabled this research .
our simulation was carried out on a hitachi sr2201 parallel computer at japan atomic energy research institute .
m. a. ( t. h. ) was partly supported by grant - in - aid for scientific research no .
10740112 ( no .
10874042 ) of the japanese ministry of education , science , and culture .
e. v. shuryak , rev .
* 65*,1 ( 1993 ) . m. -c .
chu , j. m. grandy , s. huang , and j. w. negele , phys .
d * 48 * , 3340 ( 1993 ) . y. nakahara , m. asakawa and t. hatsuda , hep - lat/9905034 ( phys .
d in press ) .
see the review , m. jarrell and j. e. gubernatis , phys
. rep . * 269 * , 133 ( 1996 ) .
r. n. silver et al . , phys .
65 * , 496 ( 1990 ) ; w. von der linden , r. preuss , and w. hanke , j. phys . * 8 * , 3881 ( 1996 ) ; n. wu , _ the maximum entropy method _ , ( springer - verlag , berlin , 1997 ) .
d. b. leinweber , phys .
d * 51 * , 6369 ( 1995 ) ; d. makovoz and g. a. miller , nucl .
b468 * , 293 ( 1996 ) ; c. allton and s. capitani , nucl . phys . *
b526 * , 463 ( 1998 ) ; ph .
de forcrand et al . ,
b ( proc . suppl . ) * 63a - c * , 460 ( 1998 ) .
see e.g. , j. skilling , in _ maximum entropy and bayesian methods _
j. skilling ( kluwer , london , 1989 ) , pp.45 - 52 ; s. f. gull , ibid .
pp.53 - 71 .
this formula is motivated by our lattice qcd data .
the milc code ver . 5 , + http://cliodhna.cop.uop.edu/~hetrick/milc .
y. iwasaki et al . , phys .
d * 53 * , 6443 ( 1996 ) ; t. bhattacharya et al .
6486 . l. maiani and g. martinelli , phys . lett .
* b178 * , 265 ( 1986 ) . m. gckeler et al . , nucl . phys . *
b544 * , 699 ( 1999 ) .
m. asakawa , t. hatsuda and y. nakahara , in preparation . | using the maximum entropy method , spectral functions of the pseudo - scalar and vector mesons are extracted from lattice monte carlo data of the imaginary time green s functions .
the resonance and continuum structures as well as the ground state peaks are successfully obtained .
error analysis of the resultant spectral functions is also given on the basis of the bayes probability theory . |
the mean field approximation is known to be a good way to approximate the many - body schrdinger dynamics when the number of particles is large enough ( see @xcite ) .
+ it consists in looking for the solutions to the non - linear schrdinger equation for one particle called the hartree equation .
we are interested in the density matrix associated with the wave function , this matrix satisfies the quantum liouville equation dual to the von neumann equation.the partial trace operators of this matrix , called the reduced density matrices , satisfy a hierarchy of equations .
for instance , by considering the case where the initial state for the schrdinger equation is a hartree ansatz(a product state ) which is suitable for a bosons condensate , the limit , when the number of particles @xmath0 goes to the infinity of these matrices converge in trace norm to the product of the density matrix associated with the solution to the hartree equation . and
this asymptotic density matrix satisfies the time dependent hartree equation @xcite .
+ when the particles are bosons , the suitable space for the bosons is the symmetric fock space on the phase - space .
moreover for the sake of numerical computations , a finite - dimensional phase - space will be used instead of an usual phase - space of type @xmath1 .
so here the phase - space will be @xmath2 where @xmath3 is a given integer representing the number of sites .
each particle can live in one of the @xmath3 sites .
+ for the numerical implementation , an explicit basis of the @xmath0-fold sector of the fock bosonic spaces is specified .
this basis allows the numerical computation of the full @xmath0-body quantum problem for @xmath0 large enough to validate various mean field regimes , in spite of a rapidly increasing complexity .
+ the resolution of the @xmath0-particles schrdinger equation will rely on a splitting method , one part for the free hamiltonian and the other one for the two particles interaction term . + for the simulations , the considered real bounded potential associated with the interaction term will be @xmath4 defined on @xmath5 by @xmath6 if @xmath7 and @xmath8 . + according to previous results related to the propagation of the wigner measures @xcite knowing the wigner measure at time @xmath9 determines the wigner measure at time @xmath10 and all asymptotic reduced density matrices . for many examples , like hermite states , twin fock states or states studied in quantum information theory ( see @xcite ) their wigner measure as well as the order of convergence of reduced density matrices
is known explicitly .
the evolution of the wigner measure , and consequently of the asymptotic density matrices , is evaluated after integrating numerically the mean field non linear hartree time - dependent equation . in order to preserve numerically quadratic quantities like the symplectic form ,
the latter is solved with a symplectic @xmath11 order runge - kutta method ( @xcite ) .
+ to estimate numerically the error of convergence of the reduced density matrices in the mean field limit , a discretization of a time interval @xmath12 $ ] is considered , in the examples @xmath13 is chosen , then the quantity @xmath14}\left\|\gamma^{(p)}_{n}(t)-\gamma_{\infty}^{(p)}(t)\right\|_1 $ ] is observed . here
@xmath15 denotes the time - evolved @xmath16 particles reduced density matrix for @xmath0 bosons , while @xmath17 is its theoretical limit when @xmath0 goes to the infinity , and @xmath18 denoting the trace norm .
+ for the evaluation of the order of convergence , the logarithm of the previous error estimate is drawn according to @xmath19 .
this gives a straight line whose the slope is the order of convergence in @xmath20 .
+ these numerical results agree very well and illustrate the theoretical analysis carried out in @xcite .
+ by increasing @xmath3 , we wish to approach a continuous model .
the complexity of the computations increase in the same time that @xmath3 and @xmath0 increase because of the dimension of the @xmath0-particles bosons fock space on @xmath21 which is a binomial coefficient .
the bosons fock space on an hilbert space @xmath22 is defined as @xmath23 where @xmath24 is the symmetric n - fold hilbertian tensor product of @xmath22 which is the range of the projection defined on the hilbert tensor product @xmath25 , by : @xmath26 where @xmath27 is in @xmath22 for each @xmath28 in @xmath29 $ ] and @xmath30 is the set of the permutations of @xmath31 elements . + for @xmath32 in @xmath22 and @xmath33 positive , the @xmath33-scaled annihilation and creation operators are defined for all @xmath34 in @xmath22 and @xmath31 in @xmath35 by : @xmath36 . + these operators satisfy the canonical commutation relations ( ccr ) : @xmath37&=\varepsilon \langle z_1,z_2\rangle id\ ; , & \\ [ a(z_1),a(z_2)]&=0\ ; , & \\ [ a^{*}(z_1),a^{*}(z_2)]&=0\;. \end{aligned}\ ] ] the second quantization of an operator @xmath38 or a self - adjoint operator @xmath39 in @xmath22 is defined by : @xmath40
the second quantization of @xmath41 is the number operator : @xmath42 use the following notations : @xmath43 + for @xmath44 in @xmath45 , the length of @xmath46 is written @xmath47 and the factorial of @xmath46 @xmath48 + let @xmath49 be an orthonormal basis of @xmath50 . + set @xmath51 .
then an orthonormal basis of @xmath52 can be built from this basis which is labelled by the multi - indices @xmath46 in @xmath53 such that @xmath54 .
+ with the creation operators , an orthonormal basis can be written as : + @xmath55 where @xmath56 is the vacuum of the fock space . +
then the dimension of @xmath52 is @xmath57 + and @xmath58 is the dimension of @xmath59 as a binomial number , the dimension of the @xmath0 particles bosonic sector increases rapidly but not too much , as @xmath0 increases ( see table [ dimensions ] for numerical values ) .
the complexity has to be handled carefully if we want to approach the mean field limit by taking @xmath0 large or the continuous model by taking @xmath3 large .
+ define @xmath60 the discrete laplacian operator on @xmath50 by : @xmath61 and let @xmath62 be the free hamiltonian .
+ the interaction term denoted by @xmath63 equals : @xmath64 where @xmath65 .
+ in this framework changing the value of @xmath66 add an irrelevant phase factor in the time evolved wave function . in the sequel @xmath8 is assumed .
+ or as a wick quantized operator : @xmath67 the considered linear schrdinger equation is : @xmath68 where the complete hamiltonian is defined on the bosonic fock space by : @xmath69
the energy of the hamiltonian corresponds to the symbol of the complete hamiltonian : @xmath70 while recalling our convention @xmath71 for all @xmath28 in @xmath72 .
+ the mean field equation in @xmath22 is written as : @xmath73 for each component in @xmath21 we obtain : @xmath74 @xmath75\ ] ] by writing @xmath76 where @xmath77 and @xmath16 belong to @xmath78 , it becomes : @xmath79 + a @xmath11 or @xmath80 order gauss rk method is used by using the coefficients given in @xcite to solve the hartree equation . +
a symplectic method is used to preserve the quadratic part of the energy , the symplectic form and the phase space volume .
for @xmath81 the field operator is defined by @xmath82 which is essentially self - adjoint on @xmath83 + the weyl operator is defined by @xmath84 .
+ let @xmath85 be a family of normal states on @xmath86 with @xmath87 , @xmath88 . + a measure @xmath89 is a wigner measure for this family , @xmath90 , if there exists @xmath91 , @xmath92 such that @xmath93=\int_{\mathcal{z } } e^{2i\pi\textrm{re}~\langle f , z\rangle}~d\mu(z ) \;,\ ] ] see @xcite .
+ the following result valid for separable hilbert spaces @xmath22 , apply to our finite dimensional @xmath94 .
[ th.wigdef ] if @xmath85 satisfies the uniform estimate @xmath95 \leq{c_{\delta } } < + \infty$ ] for some @xmath96 fixed , @xmath97 is not empty and made of borel probability measures ( @xmath98 separable ) such that @xmath99 . for each @xmath16 in @xmath35 ,
the reduced density matrix associated with a state @xmath100 is a trace class operator in @xmath101 defined by the duality relation : @xmath102=\frac{\tr\left[\varrho_{\varepsilon}b^{wick}\right ] } { \tr\left[\varrho_{\varepsilon}(|z|^{2p})^{wick}\right]}\end{aligned}\ ] ] where @xmath103 .
+ the asymptotic reduced density matrix associated with the wigner measure @xmath89 equals : @xmath104 in finite dimension , if the family @xmath105 satisfies @xmath106 then the @xmath107-condition ( see @xcite ) : @xmath108=\int_{\mathcal{z}}|z|^{2p}~d\mu(z)\,\ ] ] is always satisfied .
[ eq.wigcompact ] if the family @xmath105 satisfies @xmath106 with the @xmath107-condition : @xmath108=\int_{\mathcal{z}}|z|^{2p}~d\mu(z)\;,\ ] ] then @xmath109 $ ] converges to @xmath110 for all polynomial @xmath111 and @xmath112 for all @xmath113
. @xcite .
assume @xmath114 and the condition @xmath107 .
then @xmath115 .
the measure @xmath116 is the push - forward measure of the initial measure @xmath117 where @xmath118 is the hamiltonian flow associated with the equation @xmath119 after propagation of the wigner measures , for any @xmath120 , the convergence of the reduced densiy matrices is obtained at any time @xmath10 : @xmath121 @xcite .
[ th - rate ] let @xmath122 be a sequence of positive numbers with @xmath123 and such that @xmath124 is bounded .
let @xmath125 and @xmath126 be two sequences of density matrices with @xmath127 and @xmath128 for each @xmath129 .
assume that there exist @xmath130 , @xmath131 and @xmath132 such that for all @xmath129 with @xmath133 : [ init - ineq ] ^(p)_n-_^(p)_1 c_0 .
then for any @xmath134 there exists @xmath135 such that for all @xmath136 $ ] and all @xmath137 with @xmath133 , [ main - ineq ] ^(p)_n(t)-_^(p)(t)_1 c_t , where _ ^(p)(t)=_z |z^pz^p| d_t(z ) , with @xmath138 is the push - forward of the initial measure @xmath139 by the well defined and continuous hartree flow @xmath140 on @xmath22 .
to solve the mean field equation , a runge - kutta method is used .
+ let @xmath141 , @xmath142 @xmath143 be real numbers and @xmath144 .
+ an s - stage runge - kutta method with a time step @xmath145 to solve a first - order ordinary equation @xmath146 , @xmath147 is given by : @xmath148 represented as : [ cols="^,^,^,^ " , ] . in our case , the function @xmath149 corresponds to @xmath150 .
+ as a function in @xmath151 by replacing @xmath32 by @xmath152 , @xmath153 @xmath154 @xmath155 for an implicit runge - kutta method , a newton method is applied to find the coefficients @xmath156 for each step of the rk method to the function @xmath157 .
+ given the time step @xmath145 small enough , the starting point of the newton method is chosen by setting @xmath158 for all @xmath28 . + to apply the newton s method the differential of @xmath149 is computed by using the following partial derivatives of @xmath149 : @xmath159 + then the differential of @xmath160 is : [ diff.g ] @xmath161 where @xmath160 is considered as a function from @xmath162 . for a given @xmath164 in @xmath52 , the full @xmath0-body evolved state @xmath165
is computed in the basis @xmath166 .
after writing @xmath167 a modified splitting method for which the numerical error is carefully controlled ( see [ sec.error-estim ] ) , involves only multiplications by the diagonal matrix @xmath168 and the sparse matrix @xmath169 .
+ in order to handle the high complexity of the problem ( see table [ dimensions ] ) no matrix , but only vectors or the sparse matrices @xmath169 and the matrix @xmath170 are stored .
+ the complete evolution @xmath171 is computed by a composition method based on the strang splitting method : @xmath172 the @xmath11 order composition method is given by : @xmath173 where the coefficients of the method are satisfying the two equations ( see @xcite ) : [ eq.composition ] @xmath174 and are given by : @xmath175 + the numerical computation of @xmath177 , relies on the following two remarks : * the dimension of the @xmath0-fold sectors @xmath178 prevents the storage of any square matrix . *
the matrix of @xmath179 is actually non trivial sparse matrix in the basis @xmath180 .
the matrix of @xmath181 is given by : @xmath182 + we are interested in the matrix of the second quantization of the discrete laplacian on the basis of the bosons space to implement it numerically as a sparse matrix containing only @xmath183 elements whereas a full matrix contains @xmath184 .
+ then @xmath185 will be computed at each time step by a @xmath11 order taylor expansion . + this expansion is then replaced in the composition method .
+ for an operator @xmath186 , @xmath187 @xmath188 this yields : @xmath189 [ lem.deriv-annih ] for all multi - indices @xmath190 and @xmath46 the following equality holds : + @xmath191 @xmath192 by using the following separation of the variables : @xmath193 + in this space let @xmath194 be @xmath195 .
+ let us consider @xmath196 and @xmath197 , @xmath198|\omega_{i}\rangle\\ & = \varepsilon \alpha_{i}a(e_i)^{\gamma_i-1}a^*(e_i)^{\alpha_i-1}|\omega_{i}\rangle \;.\end{aligned}\ ] ] + by induction , we obtain @xmath199 + and @xmath200 + the above separation of variables leads under the condition @xmath201 to : @xmath202 for all multi - indices @xmath46 and @xmath203 , the matrix elements of @xmath204 are given by : @xmath205 where @xmath206 for @xmath46 and @xmath203 multi - indices in @xmath207 . according to and to lemma [ lem.deriv-annih ]
, we obtain : + @xmath208 and @xmath209 + then @xmath210 + numerically only the indices of the multi - indices @xmath46 and @xmath203 corresponding to the nonzero components of @xmath169 with their values , are stored in an array . + in the algorithm instead of running over the multi - indices @xmath46 or @xmath203 with a length @xmath0 ,
the multi - indices @xmath211 with a length @xmath212 are run over . and
for each @xmath28 in @xmath213 $ ] , the changes of multi - indices @xmath214 or @xmath215 are used , then the indices of the corresponding multi - indices @xmath46 and @xmath203 with length @xmath0 are looked for .
+ therefore an array composed of @xmath216 triplets of elements is numerically stored .
+ denote @xmath218 and @xmath219 .
+ by using the relations ccr : @xmath220 then @xmath221 can be rewritten as : @xmath222 + and since @xmath223 then @xmath221 is diagonal in the basis @xmath224 : @xmath225 and @xmath226 consider @xmath227 with @xmath228 and its associated homogeneous polynomial : @xmath229 let us compute the quantity @xmath230 when @xmath100 is a normal state . using an orthonormal basis of the @xmath0-fold sector @xmath231 , @xmath100 is a linear combination of operators @xmath232 .
it suffices to compute @xmath233 .
+ [ lem.trace ] set @xmath227 with @xmath228 and let @xmath34 and @xmath164 be in @xmath231 then : @xmath234 in the orthonormal basis @xmath235 of @xmath231 . given @xmath164 and @xmath34 in the n - particles bosons space and the formula [ lem.deriv-annih ] @xmath236 we can write : @xmath237 + because @xmath238 if and only if @xmath239 so @xmath240 and @xmath241 means @xmath242 .
+ the last line is obtained by a change of multi - indices by setting for each @xmath46 , @xmath243 because @xmath244 , and then @xmath245 .
+ numerically , all multi - indices of @xmath246 with length not larger than a given @xmath247 are stored in the lexicographic order . + for our algorithms , we pay attention to preserve this lexicographic order ( or reverse ) .
+ for a given @xmath248 , the list of relevant multi - indices ( with length @xmath0 ) is extracted and handled in the lexicographic order .
+ for a given @xmath249 , numerically the above summation is performed over multi - indices @xmath250 such that @xmath251 in the lexicographic order .
+ then for each @xmath250 , the multi - indices @xmath46 of length @xmath0 written as @xmath252 are looked for .
these @xmath46 are exactly the multi - indices such that @xmath244 and @xmath54 .
+ note in particular that the mapping @xmath253 preserves the lexicographic order .
+ first let us compute the matrix elements of @xmath254 in the orthonormal basis @xmath255 .
+ the matrix element corresponding to the line @xmath203 and column @xmath46 is : @xmath256 according to the duality relation of the reduced density matrices with @xmath257 , and @xmath258 + if @xmath257 , its wick quantized is : @xmath259 and then @xmath260 then owing to lemma [ lem.trace ] , all the elements of the matrices @xmath254 can be numerically computed . + in the case where the initial state is a hermite state @xmath261 , @xmath262 needs to be expanded in the orthonormal basis @xmath180 which is given by the following lemma .
[ hermite - state ] for all @xmath120 , and @xmath263 , we obtain in the basis @xmath224 : @xmath264 @xmath265 and then @xmath266 in the case where the initial state is a twin state , the following lemma is used to obtain an expansion in the basis @xmath180 .
[ twin - state ] let @xmath267 be in @xmath22 and @xmath32 the state @xmath268 such that @xmath269 .
then we obtain @xmath270 @xmath271 @xmath272 @xmath273 because of the ccr relations .
@xmath274 by setting @xmath275 , @xmath270 further the limit reduced density matrices have to be computed numerically . in order to do this
, the integration over @xmath22 of the wigner measure is discretized .
the problem is then reduced to the computation of the matrix elements of @xmath276 in the basis @xmath277 .
+ compute the matrix elements of @xmath278 , according to lemma [ hermite - state ] : @xmath279 then @xmath280 for the computation of the integral @xmath281 , the wigner measure is approximated by a convex combination of gauge invariant delta functions @xmath282 , where @xmath283 .
+ for the wigner measure associated with the hermite states , @xmath284 , and the discretization is trivial and exact .
it is not needed to be approximated because of the gauge invariance .
+ in the case of the twin states given by @xmath285 , where @xmath286 , @xmath287 , and @xmath288 , the wigner measure is @xmath289 according to @xcite , with : @xmath290 @xmath291 numerically the interval @xmath292 $ ] is discretized and @xmath139 is approximated by @xmath293 .
+ the wigner measure @xmath294 propagated at the time @xmath10 of the twin states is given by : @xmath295 where @xmath296 is solution to the hartree equation at the time @xmath10 with initial condition @xmath297 .
+ numerically it is now approximated by @xmath298 where @xmath299 solves the hartree equation . +
thus the matrix elements of @xmath300 are given by the formula : @xmath301 and the approximation of the scalar @xmath302 is given by the formula @xmath303 .
+ the matrix of @xmath304 can then be computed at any time @xmath10 numerically with a good approximation . +
the baker - campbell - hausdorff formula ( see @xcite ) allows to find the order of the composition method which is @xmath305 .
then the taylor s formula with @xmath11 order integral remainder and the cauchy inequalities are used to estimate the error . + the following proposition
gives an estimate of the composition method .
+ [ comp estimates ] let @xmath306 , @xmath307 and @xmath308 be two anti - adjoint matrices such that @xmath309 . +
then @xmath310 where @xmath311 is the composition method .
@xmath312 @xmath313 then for @xmath314 @xmath315 let us consider the holomorphic function on @xmath316 defined by : @xmath317 since the composition method is of @xmath11 order then for @xmath318 , the taylor s formula with integral remainder yields : @xmath319}\|f_{a , b}^{(5)}(t)\| \;.\end{aligned}\ ] ] by the cauchy s integral formula , we know for each @xmath320 $ ] : @xmath321 hence for all @xmath322 and @xmath323 such that @xmath324 we obtain : @xmath325 let @xmath307 and @xmath308 be such that @xmath309 . + by setting @xmath326 and @xmath327 , we obtain + @xmath328 and @xmath329 .
+ then @xmath330 the composition method is approximated by replacing @xmath185 by its @xmath11 order taylor expansion , with some normalization factor .
+ errors estimates for this modified composition method rely on the two following lemmas .
[ lem aprox split ] let @xmath331 be a normed vector space , @xmath332 , @xmath333 and @xmath334 two maps sequences from @xmath331 to @xmath331 such that for all @xmath335 : * @xmath336 is linear . *
@xmath337 for all @xmath338 . * @xmath339 for @xmath340 , @xmath341 set @xmath342 and @xmath343 with @xmath344 .
+ then we deduce @xmath345 .
let us proceed by induction on @xmath346 .
+ for @xmath347 , @xmath348 .
+ let us assume @xmath345 with the hypotheses fullfilled for @xmath349 .
+ @xmath350 since @xmath351 is linear and unitary we obtain : @xmath352 moreover @xmath353 , + then @xmath354 we obtain @xmath355 + let @xmath356 denote the @xmath11 order taylor expansion of @xmath357 around @xmath358 . + let @xmath307 be in @xmath359 , @xmath360 .
+ [ tltilde ] let @xmath361 be a vector in a normed vector space @xmath331 and let @xmath307 be an anti - adjoint operator on @xmath331 . define the application @xmath362 on @xmath331 which is non linear by : @xmath363 it preserves the norm .
+ then @xmath364 .
@xmath365 @xmath366 + the @xmath11 order error of the taylor expansion gives : @xmath367 the following proposition gives an estimate of the approximation of the composition method .
+ let @xmath307 and @xmath308 be two anti - adjoint matrices and @xmath346 an integer such that @xmath368 and @xmath369 then @xmath370 for all vector @xmath361 , where @xmath371 let @xmath361 be a normed vector . + first for @xmath372 let us estimate the error : + @xmath373 by using the fact @xmath374 is an unitary operator and lemma [ tltilde ] : @xmath375 like in the previous proof .
+ now lemma [ lem aprox split ] can be applied with @xmath376 , @xmath377 and @xmath378 .
+ then @xmath379 secondly let us estimate the error : + @xmath380 + by using proposition [ comp estimates ] with its hypotheses and the previous estimates : @xmath381 by applying that to @xmath382 and @xmath383 where @xmath384 with @xmath346 positive integer , we obtain : @xmath385 then by applying lemma [ lem aprox split ] with @xmath386 and @xmath387 , we obtain : @xmath388 by knowing that for all positive integer @xmath389 and positive real @xmath390 , @xmath391 is minimal in @xmath392 and @xmath393 , the condition @xmath394 that is @xmath395 implies @xmath396 for @xmath397 , we know according to : @xmath398 then @xmath399 when @xmath400 with @xmath401 the norm @xmath402 is bounded from above by @xmath403 independently of the number @xmath404 of particles .
+ moreover @xmath405 therefore @xmath406 . + finally by applying the last proposition with @xmath407 and @xmath408 , an error estimate is obtained for the complete evolution : @xmath409 pratically , the time step is chosen according to @xmath0 and @xmath10 so that the above error is negligeable .
for all the numerical simulations the final time is chosen to be @xmath13 , the number of time steps for the @xmath11 order runge - kutta method applied to solve the mean field equation is @xmath410 . + the loop of the number of particles is performed numerically from @xmath411 to @xmath412 particles , and only for an even number of particles .
+ in the fortran program the computations were performed by parallelizing the loop in the computation of the product sparse matrix - vector @xmath413 with openmp on 8 threads . + * results and orders of convergence for @xmath414 and @xmath415 * + for each type of states , the following graphics show for the reduced density matrices and for @xmath416 sites : * the logarithm of the error in trace norm @xmath417}\left\|\gamma^{(p)}_{n}(t)-\gamma_{\infty}^{(p)}(t)\right\|_1)$ ] according to the logarithm of the number of particles @xmath0 in the cases @xmath418 and @xmath419 . + a straight line is obtained whose the slope is the order of the error in @xmath20 .
+ these numerical experiments also valid the idea that for rather smooth but non trivial @xmath0-body bosonic system , the mean field asymptotics start to be relevant at @xmath420 .
the numerical plot agree perfectly with the theoretical results of @xcite .
* in the case @xmath418 the density of particles on each site @xmath421 given by @xmath422 for @xmath423 particles and for the mean field limit at the same times @xmath9 and @xmath424 . + * the correlations in terms of the @xmath425 and @xmath419 particles reduced density matrices , for @xmath423 particles and the mean field at the time @xmath424 . depending on the case , this plot shows with which accuracy the mean field also catches some quantum correlations . for the hermite states
@xmath262 the vector @xmath32 is given by @xmath426 .
+ for the twin states @xmath285 , @xmath427 and @xmath428 .
+ for the wq states @xmath285 , @xmath427 and @xmath428 .
+ in this case the state is given by @xmath429 , with @xmath430 , @xmath431 and @xmath432 fixed for the mean field .
+ the associated wigner measure is @xmath433 .
+ in these simulations @xmath434 .
a case when the order of convergence is equal to @xmath435.(see@xcite ) + in this case @xmath436 with @xmath437 .
+ the associated wigner measure is @xmath438 .
* class of symbols * @xcite . for any @xmath439 , define @xmath440 to be the space of homogeneous complex - valued polynomials on @xmath441 such that @xmath442 if and only if there exists a ( unique ) bounded operator @xmath443 such that for all @xmath444 : [ eq.wick ] b(z)=z^q , z^p .
@xmath445 where @xmath446 denotes the operator associated with the symbol @xmath447 according to .
the composition method based on the strang splitting with the coefficients is of @xmath11 order ( see @xcite ) .
+ dimension of @xmath448 : @xmath449 for @xmath3 in @xmath450 $ ] and @xmath0 in @xmath451 $ ] : + [ dimensions ] @xmath452 & 1 & 2 & 3 & 4 & 5 & 6 & 7 & 8 & 9 & 10 + 2 & 1 & 3 & 6 & 10 & 15 & 21 & 28 & 36 & 45 & 55 + 3 & 1 & 4 & 10 & 20 & 35 & 56 & 84&120 & 165 & 220 + 4 & 1 & 5 & 15 & 35 & 70 & 126 & 210&330 & 495 & 715 + 5 & 1 & 6 & 21 & 56 & 126 & 252 & 462 & 792 & 1287 & 2002 + 6 & 1 & 7 & 28 & 84 & 210 & 462 & 924 & 1716 & 3003 & 5005 + 7 & 1 & 8 & 36 & 120 & 330 & 792 & 1716 & 3432 & 6435 & 11440 + 8 & 1 & 9 & 45 & 165 & 495 & 1287 & 3003 & 6435 & 12870 & 24310 + 9 & 1 & 10 & 55 & 220 & 715 & 2002 & 5005 & 11440 & 24310 & 48620 + 10 & 1 & 11 & 66 & 286 & 1001 & 3003 & 8008 & 19448 & 43758 & 92378 + 11 & 1 & 12 & 78 & 364 & 1365 & 4368 & 12376&31824 & 75582 & 167960 + 12 & 1 & 13 & 91 & 455 & 1820 & 6188 & 18564 & 50388 & 125970 & 293930 + 13 & 1 & 14 & 105 & 560 & 2380 & 8568 & 27132&77520 & 203490 & 497420 + 14 & 1 & 15 & 120 & 680 & 3060 & 11628 & 38760&116280 & 319770 & 817190 + 15 & 1 & 16 & 136 & 816 & 3876 & 15504 & 54264&170544 & 490314 & 1307504 + 16 & 1 & 17 & 153 & 969 & 4845 & 20349 & 74613 & 245157 & 735471 & 2042975 + 17 & 1 & 18 & 171 & 1140 & 5985 & 26334 & 100947 & 346104 & 1081575 & 3124550 + 18 & 1 & 19 & 190 & 1330 & 7315 & 33649 & 134596 & 480700 & 1562275 & 4686825 + 19 & 1 & 20 & 210 & 1540 & 8855 & 42504 & 177100 & 657800 & 2220075 & 6906900 + 20 & 1 & 21 & 231 & 1771 & 10626 & 53130 & 230230&888030 & 3108105 & 10015005 +
the research has been supported by the anr-11-is01 - 0003 lodiquas . + the numerical computations were performed on the servers vsc of the university of vienna and magi of the university paris 13 .
the numerical illustrations were obtained with scilab .
+ i thank gilles vilmart , gilles scarella and hans peter stimming for useful discussions about numerical schemes and computers .
i also thank my advisors francis nier and norbert j. mauser for their remarks and commitment for the writing of this article .
hairer , ernst and lubich , christian and wanner , gerhard geometric numerical integration springer series in computational mathematics 31 structure - preserving algorithms for ordinary differential equations , springer - verlag ( 2002 ) | the mean field approximation is numerically validated in the bosonic case by considering the time evolution of quantum states and their associated reduced density matrices by many - body schrdinger dynamics .
the model phase - space is finite - dimensional .
the results are illustrated with numerical simulations of the evolution of quantum states according to the time , the number of the particles , and the dimension of the phase - space .
: 81s30 , 81s05 , 81t10 , 35q55 , 81p40 + : mean field limit , reduced density matrices , wigner measures , bosons fock space , second quantization . |
it is now well established that the centers of most galaxies host supermassive black holes ( bh ) with masses in the range @xmath4 @xcite . the evidence is particularly compelling for a bh of mass @xmath5 at the center of our own galaxy @xcite .
dynamical estimates indicate that , across a wide range , the central bh mass is about 0.1% of the spheroidal component of the host galaxy @xcite .
a related correlation may exist with the total gravitational mass of the host galaxy ( basically the mass of its dark matter halo ) @xcite .
an even tighter correlation is observed between the central velocity dispersion and the central bh mass @xcite .
theoretical arguments and recent observations suggest that a central bh may also exist in some globular clusters @xcite .
in particular , recent hst observations and dynamical modeling of m15 by @xcite yielded results that are consistent with the presence of a central massive bh in this cluster .
similarly @xcite have argued for the existence of an even more massive bh at the center of the globular cluster g1 in m31 . however , @xmath2-body simulations @xcite suggest that the observations of m15 and g1 could be explained equally well by the presence of many compact objects near the center without a massive bh @xcite .
when the correlation between the mass of the central bh and the spheroidal component in galaxies is extrapolated to smaller stellar systems like globular clusters , the inferred bh masses are @xmath6 , much larger than a @xmath7 stellar - mass bh , but much smaller than the @xmath8 of supermassive bh .
hence , these are called _ intermediate - mass black holes _ ( imbh ) . if some globular clusters do host a central imbh the question arises of how these objects were formed ( for recent reviews see * ? ? ?
* ; * ? ? ?
one natural path for their formation in any young stellar system with a high enough density is through runaway collisions and mergers of massive stars following core collapse . these runaways could easily occur in a variety of observed young star clusters such as the `` young populous clusters '' like the arches and quintuplet clusters in our galactic center and the `` super star clusters '' observed in all starburst and galactic merger environments ( see , e.g. , *
* ; * ? ? ?
. the pistol star in the quintuplet cluster @xcite may be the product of such a runaway , as demonstrated recently by direct @xmath2-body simulations @xcite .
a similar process may be responsible for the formation of seed bh in proto - galactic nuclei , which could then grow by gas accretion or by merging with other imbh formed in young star clusters @xcite .
further observational evidence for imbh in dense star clusters comes from recent _ chandra _ and xmm-_newton _ observations of ultra - luminous x - ray sources , which are often ( although not always ) associated with young star clusters and whose high x - ray luminosities in many cases suggest a compact object mass of at least @xmath9 @xcite , although beamed emission by an accreting stellar - mass bh provides an alternative explanation @xcite .
when they are born , star clusters are expected to contain many young stars with a wide range of masses , from @xmath10 to @xmath11 , distributed according to a salpeter - like initial mass function ( imf ) @xcite .
inspired by an early version from @xcite for the formation of supermassive bh , we show in figure [ fig_rees_diag ] a diagram illustrating the two main scenarios leading to the formation of an imbh at the center of a dense star cluster .
the early stages of dynamical evolution are dominated by the stars in the upper part of the mass spectrum . through dynamical friction ,
these heavy stars tend to concentrate toward the center and drive the system to core collapse .
successive collisions and mergers of the massive stars during core collapse can then lead to a runaway process and the rapid formation of a very massive object containing the entire mass of the collapsing cluster core .
although the fate of such a massive merger remnant is rather uncertain , direct `` monolithic '' collapse to a bh with no or little mass loss is a likely outcome , at least for sufficiently small metallicities @xcite .
an essential condition for this runaway to occur is that the core collapse must occur before the most massive stars born in the cluster end their lives in supernova explosions @xcite .
the accumulation at the center of a galaxy of many imbh produced through this runaway process in nearby young star clusters ( like the arches and quintuplet clusters in our galaxy ) provides an interesting new way of building up the mass of a central supermassive bh @xcite .
it is possible that this process of accumulation is still ongoing in our own galaxy @xcite .
these ideas have potentially important implications for the study of supermassive bh by the laser interferometer space antenna ( lisa ) , since the inspiral of imbh into a supermassive bh provides the best source of low - frequency gravitational waves for direct study of strong field gravity @xcite . in the alternative scenario where massive stars evolve and produce supernovae
before the cluster goes into core collapse , a subsystem of stellar - mass bh will be formed ( figure [ fig_rees_diag ] ) .
as demonstrated in section [ subsec_stell_evol ] , the mass loss from supernovae provides significant indirect heating of the cluster core , delaying the onset of core collapse until much later , after the stellar remnants undergo mass segregation .
the final fate of a cluster with a component of stellar - mass bh remains highly uncertain .
this is because realistic dynamical simulations for such clusters ( containing a large number of black holes _ and _ ordinary stars with a realistic mass spectrum ) have yet to be performed . for old and relatively small systems ( such as galactic globular clusters ) , complete evaporation is likely ( with all the stellar - mass bh ejected from the cluster through 3-body and 4-body interactions in the dense core ) .
this is expected theoretically on the basis of simple qualitative arguments @xcite and has been demonstrated recently by direct @xmath2-body simulations for very small systems containing only @xmath12 bh @xcite . however , for larger systems ( more massive globular clusters or proto - galactic nuclei ) , contraction of the cluster to a highly relativistic state could again lead to successive mergers ( driven by gravitational radiation ) and the formation of a single massive bh @xcite .
moreover , it has been recently suggested that , if stellar - mass bh are formed with a broad mass spectrum ( a likely outcome for stars of very low metallicity ; see @xcite ) , the most massive bh could resist ejection , even in a system with low escape velocity such as a globular cluster .
these more massive bh could then grow by repeatedly forming binaries ( through exchange interactions ) with other bh and merging with their companions @xcite .
however , as most interactions will probably result in the ejection of one of the lighter bh , it is unclear whether any object could grow substantially through this mechanism before running out of companions to merge with .
the physics of gravothermal contraction and core collapse is by now very well understood for _ single - component _ systems ( containing all equal - mass stars ) .
in particular , the dynamical evolution of an isolated , single - component plummer sphere to core collapse has been studied extensively and has become a testbed for all numerical codes used to compute the evolution of dense star clusters @xcite
. this evolution can be visualized easily using lagrange radii , enclosing a fixed fraction of the total mass of the system ( see , e.g. , fig . 3 of * ?
* and fig . 5 of ; * ? ? ?
as the system evolves , the inner lagrange radii contract while the outer ones expand .
this so - called gravothermal contraction results from the negative heat capacity that is a common property of all gravitationally bound systems @xcite . in the absence of an energy source in the cluster core , and in the point - mass limit ( i.e. , neglecting physical collisions between stars )
, the contraction of the cluster core becomes self - similar and continues indefinitely .
this phenomenon is known as core collapse and its universality is very well established ( see , e.g. , * ? ? ?
* table 1 and references therein ) . when the system contains stars with a variety of masses , the evolution to core collapse is accelerated .
this has been demonstrated by the earliest @xmath2-body and monte carlo simulations @xcite . to illustrate this behavior , and as a preview of the results presented later in sec .
4 , we show in figure [ fig_cc_compare ] the evolution of a cluster described initially by a simple plummer model containing @xmath13 stars with a broad salpeter imf between @xmath14 and @xmath15 .
core collapse occurs after @xmath16 , where @xmath17 is the initial half - mass relaxation time ( see eq .
( [ eq_rel_time ] ) below ) . in sharp contrast ,
core collapse in a single - component plummer model occurs after @xmath18 ( a well - known result ) .
thus the presence of a broad imf can dramatically accelerate the evolution of the cluster to core collapse .
this acceleration of the evolution to core collapse is due to the changing nature of energy transfer in the presence of a wide mass spectrum .
relaxation processes tend to establish energy equipartition ( see , e.g. , * ? ? ?
8.4 ) . in a cluster where the masses of the stars are nearly equal
, this can be ( very nearly ) achieved .
the core collapse is then a result of energy transfer from the inner to the outer parts of the cluster , leading to gravothermal contraction @xcite .
a large difference between the masses of the stars allows a more efficient mechanism for energy transfer . in this case
, energy equipartition would tend to bring the heavier stars to lower speeds .
however , as a result , the heavier stars sink to the center , where they tend to gain kinetic energy , while the lighter stars move to the outer halo .
this process is called `` mass segregation . '' as the mass segregation proceeds , the core contracts and gets denser , leading to a shorter relaxation time , which in turn increases the rate of energy transfer from heavier to lighter stars . in typical cases
this evolution eventually makes the heavier stars evolve _ away _ from equipartition @xcite .
the fundamental inability of the heavier stars to establish energy equipartition with the lighter stars in a system with a continuous mass spectrum is similar to the spitzer `` mass - segregation instability '' in two - component clusters .
@xcite , using analytic methods and a number of simplifying assumptions , determined a simple criterion for a two - component system to achieve energy equipartition in equilibrium .
if the mass of the lighter ( heavier ) stars is @xmath19 ( @xmath20 ) and the total mass in the light ( heavy ) component is @xmath21 ( @xmath22 ) , then spitzer s criterion can be written @xmath23 @xcite , using numerical simulations , obtained a more accurate empirical condition , @xmath24 when this stability criterion is not satisfied , energy equipartition can not be established between heavy and light stars .
@xcite noted that the equilibrium would not be achieved for realistic mass spectra , because there is always sufficient mass in high - mass stars that , through mass segregation , they can form a subsystem ( near the center of the cluster ) that decouples dynamically from the lower - mass stars .
this was later supported by more detailed theoretical studies @xcite . here
we carry out a simple calculation to show that a typical system , with a salpeter imf ( see sec .
3.2 ) extending from @xmath25 to @xmath26 , when viewed as a two - component cluster of `` light stars '' and `` heavy stars , '' does not satisfy any of the simple stability criteria .
let us separate the stars into these two components according to some arbitrary boundary : we group all stars lighter than @xmath27 in the first component , and all stars heavier than @xmath27 in the second component .
we use @xmath19 and @xmath21 to denote the values of the average and total mass in the first component , and @xmath20 and @xmath22 similarly for the second component .
we obtain , with @xmath27 in solar mass , @xmath28 when these values are used in equations ( [ eqn_spitzer ] ) and ( [ eqn_watters ] ) we see ( figure [ fig_ins ] ) that the stability criteria are almost never satisfied for any value of @xmath27 , except when it is extremely close to the maximum mass ( as expected , since the model reduces artificially to a single - component system in this limit , with all stars having the average mass ) .
@xcite , devised a criterion genuinely adapted to clusters with a continuous mass spectrum , under the _ ad - hoc _ assumption that the shape of the density distribution does not depend on the stellar mass .
he derives the following necessary condition for stability : @xmath29 for all @xmath27 ( his eq .
figure [ fig_ins ] shows that this condition can not be satisfied either ( see sec . [ subsec_mass_fcts ] ) is required for central equipartition to set in before core collapse while vishniac s criterion predicts that @xmath30 is sufficient . ] .
these results suggest that for any salpeter - like imf , one can always find a collection of stars that have large enough average and total mass that they will dynamically decouple from the lighter stars in the cluster .
once these decouple they form a subsystem with a much shorter relaxation time and consequently evolve to core collapse very rapidly .
the half - mass relaxation time ( relaxation time at the half - mass radius @xmath31 ) for a cluster of @xmath2 stars is given by ( * ? ? ?
* eq . 2.63 ) : @xmath32 where @xmath33 is the mass density and @xmath34 in the coulomb logarithm .
let us assume that a collapsing subsystem is formed by stars that constitute 1% of the total mass , and that all of them come from the uppermost part of the mass spectrum . for a salpeter imf with @xmath14 and @xmath15 , the number of stars in this subsystem is then @xmath35 . at the time of dynamical decoupling the central density of the subsystem must be comparable to the central density of the overall cluster .
therefore we conclude that the relaxation time is around three orders of magnitude smaller for the collapsing subsystem of heavy stars at the onset of instability .
so , essentially , the heavy stars will go into core collapse as soon as they start dominating the mass density near the center .
if the cluster starts its evolution with heavy stars distributed throughout , the timescale for core collapse will be determined by their mass segregation timescale .
the process of mass segregation for the heaviest stars in the cluster is driven by dynamical friction ( * ? ? ?
* section 7.1 ) .
the timescale for mass segregation and the onset of core collapse therefore depends primarily on the mass ratio between the dominant heavy stars and the lighter background stars @xcite . for simple dynamical friction in an effective two - component model
, one would then expect the core collapse time to be comparable to @xmath36 . here
@xmath27 ( @xmath37 , the upper mass limit of the imf ) should be the mass above which the imf contains a large enough number of massive stars to define a `` collapsing subsystem . ''
if this number is @xmath38 and the total number of stars @xmath39 we get @xmath40 for our standard salpeter imf and this simple analysis would then predict the correct order of magnitude for the core collapse time ( @xmath41 ; see figure [ fig_cc_compare ] ) .
note , however , that this simple dynamical friction picture corresponds to one or a few massive objects traveling through a uniform background of much lighter particles , a description that provides , at best , only a rough approximation of the real situation under consideration here ( see appendix ) . in this paper ,
the first of a series , we consider a wide variety of initial cluster models and we investigate the evolution of all systems until the onset of core collapse .
results from numerical simulations that include stellar collisions and track the growth of a massive object through successive mergers of massive stars during core collapse will be presented in a subsequent paper .
this work is in progress and preliminary results have already been reported elsewhere @xcite .
here we determine how the core collapse time is related to various initial parameters including the imf ( sec .
3.2 and 4.2 ) and the central concentration and tidal radius of the cluster ( sec . 3.1 and 4.3 ) , and we also include the possibility of initial mass segregation ( sec . 3.3 and 4.4 ) .
we then make a comparison with the stellar evolution time and derive limits on cluster initial conditions to allow the development of runaway collisions and the possible subsequent formation of a central massive bh ( sec .
we also derive from our calculations an estimate of the total mass of the stars that participate in the runaway collisions , which provides an upper limit to the final bh mass .
we do not include stellar evolution in our calculations since our aim here is to investigate dynamical processes taking place before even the most massive main - sequence stars in the cluster have evolved .
we do , however , study the effects of mass loss from stellar winds and their dependence on metallicity ( sec .
[ subsec_stell_evol ] ) . in a subsequent paper @xcite
, we will study the evolution of `` post - collapse '' star clusters in which a central imbh is assumed to have formed early on ( through the runaway collision process ) .
in particular , we will study the possibility that mass loss from the stellar evolution of the remaining massive stars ( those that have escaped the central collapse and runaway ) could disrupt the cluster ( cf . * ? ? ? * ) , thereby producing a `` naked '' imbh .
this is motivated by observations of ultraluminous x - ray sources in regions of active star formation ( e.g. , in merging galaxies ) containing many young star clusters , but with the x - ray sources found predominantly _ outside _ of those clusters @xcite .
an important factor that could affect significantly the dynamics of core collapse in young star clusters is the presence of primordial binaries @xcite or the dynamical formation of binaries through three - body interactions @xcite . as pointed out by @xcite , the formation rate of hard three - body binaries
would be accelerated in clusters with a mass spectrum @xcite . in this first paper
we do not take into account the presence of binaries in the cluster .
this is justified because we do not expect binaries to play an important role until after the onset of core collapse , when the central density increases suddenly .
however , collisions should become dominant immediately after the onset of core collapse .
our expectation , based on several previous theoretical studies of physical collisions during interactions of hard primordial binaries and for three - body binary formation , is that the presence of binaries in the core will in fact _ increase _ collision rates , thereby helping to trigger the runaway @xcite .
this is also supported by the results of direct @xmath2-body simulations showing that , in smaller systems containing @xmath42 single stars , collisions indeed occur predominantly through the interactions of three - body binaries formed at core collapse @xcite .
numerical methods for investigating the dynamical evolution of star clusters include direct @xmath2-body integration , solutions of the fokker - planck equation by direct ( finite - difference ) or monte carlo methods , and gaseous models ( for a review , see * ? ? ?
here we refer to direct @xmath2-body integrations simply as `` @xmath2-body simulations , '' direct integrations of the fokker - planck equation as `` fokker - planck ( fp ) simulations , '' and fokker - planck simulations based on monte carlo techniques as `` monte carlo ( mc ) simulations . ''
note , however , that our mc simulations , in which the cluster is modeled on a star - by - star basis , are in fact another type of `` @xmath2-body simulations . ''
each approach offers different advantages and disadvantages for understanding core collapse and massive black hole formation in star clusters with a realistic mass spectrum .
@xcite carried out @xmath2-body simulations starting from a variety of initial conditions for clusters containing up to @xmath43 stars .
they found that runaway collisions driven by the most massive stars can happen in sufficiently dense clusters .
their results apply directly to small star clusters containing @xmath44 stars .
however , in such a small cluster , any realistic imf typically contains only a very small number of massive stars .
for example , a salpeter imf with minimum mass @xmath14 and maximum mass @xmath15 contains a fraction @xmath45 of its stars above @xmath46 . the dynamical role played by massive stars
can therefore depend strongly on the total number of stars in the cluster .
in addition , in small systems , the dynamical evolution might be dominated by the random behavior or the initial conditions of just a few very massive stars .
consequently , to investigate runaway collisions in larger systems such as super star clusters or proto - galactic nuclei , realistically large numbers of stars must be used in numerical simulations . the computational time required for direct integration of an @xmath2-body system over one crossing time @xmath47 scales as @xmath48 ( @xmath49 on parallel machines ,
if one can adjust the number of cpus to @xmath2 optimally ; see @xcite ) . since the relaxation time is @xmath50 ( see sec .
[ cc_and_si ] ) , the scaling of the total cpu time of @xmath2-body simulations is nearly as steep as @xmath51 .
currently , using a state - of - the - art grape-6 board to accelerate the computations @xcite , the evolution of a cluster containing @xmath52 stars with @xmath53 and no primordial binaries can be integrated up to core collapse in about one day ( h. baumgardt , private communication ) . however , including primordial binaries or increasing the number of stars would still lead to prohibitively long computation times , especially for a parameter study where a large number of integrations are required .
a wide mass spectrum also leads to increased computation times for fp simulations and gaseous models . in these methods ,
a continuous mass spectrum is approximated by discrete mass bins . in the gaseous model ,
the most time - consuming operation of the algorithm is the inversion of large matrix whose dimension is proportional to the number of equations , itself proportional to @xmath54 .
hence the computing time increases like ( r. spurzem , private communication ) . ]
for fp codes , there is one diffusion term coupling each component to all others , leading to @xmath56 .
these steep scalings limit @xmath54 to at most @xmath57 to avoid computation times longer than a few days .
even with a small number of mass bins , fp simulations have yielded important qualitative results .
inagaki and collaborators investigated two- and multi - component systems @xcite .
they found that the energy transfer _ within _ the core is an important process for determining the onset of core collapse .
fp simulations by @xcite and @xcite show that a mass function with @xmath58 needs to be discretized into about 15 components or more to obtain an accurate value of the core collapse time .
a coarser discretization leads to an artificially slow evolution . @xcite
successfully tracked the energy transfer between different components and demonstrated the importance of this process .
@xcite used fp simulations to follow the evolution of single- and two - component systems .
he used a unique setting for his two - component clusters , with each component having a different initial spatial distribution .
his aim was to study the interaction of galactic nucleus containing mainly dark ( compact ) objects with the bulge of the galaxy . in one model he considered a structure with `` inverse initial mass segregation , '' where the central nucleus is made of objects much lighter than the normal stars composing the bulge .
he showed that this situation also leads to highly accelerated evolution as the more massive stars get trapped by the nucleus through dynamical friction and undergo rapid core collapse .
@xcite also investigated the evolution of clusters with a mass spectrum using fp simulations .
his simulations were two - dimensional ( in phase space ) and therefore he was able to study the development of velocity anisotropy .
gaseous models have the advantage of being very fast ( for @xmath59 ) but they include the greatest number of simplifying assumptions and so require independent checking . the good agreement shown with our mc simulation for a plummer model with salpeter imf in figure [ fig_cc_compare ] is encouraging .
note , however , that the gaseous model calculation shown in figure [ fig_cc_compare ] used 50 mass components .
this number of components is exceptionally high , leading to a total computation time in excess of three weeks , much longer than most mc runs .
it was chosen to be able to follow core collapse to a very advanced stage , until the most massive component dominates the collapsing core .
we note that the evolution of the gaseous model to core collapse is faster by some 30% and exhibits a more gradual contraction of the inner region
. the reasons for these small discrepancies will be investigated in future work , where additional comparisons between monte carlo and gaseous model calculations will also be presented @xcite . for the time being , suffice it to mention that for multi - mass clusters , there are basically two adjustable ( dimensionless ) parameters in the gaseous model , one setting the effective thermal conductivity and the other the timescale for energy exchange between components ( @xmath60 and @xmath61 , see @xcite ) . for the model plotted in figure [ fig_cc_compare ] , we used the standard values of these parameters ( @xmath62 and @xmath63 ) which were established for clusters with a different structure and mass spectrum and some adjustment ( preferentially through comparisons with @xmath2-body runs ) may be required .
a direct comparison between fp and gaseous models was also carried out by @xcite , for two - component clusters , and also resulted in good agreement .
recently , a hybrid code has been developed by @xcite , combining a gaseous model with mc techniques . in this approach
the single stars are represented by the gaseous model , while primordial binaries are followed with a monte carlo treatment .
a similar hybrid treatment could be applied to the problem we are studying here , but with massive ( single and binary ) stars included in the monte carlo component , and lower - mass stars represented by a gaseous model .
the solution of the fokker - planck equation with an orbit - averaged monte carlo method is an ideal compromise for the problem at hand .
it can handle a suitably large number of stars and a wide mass spectrum can be implemented with very little additional difficulty .
most importantly , as in direct @xmath2-body integrations , mc simulations can implement a star - by - star description of the cluster .
this allows the inclusion of many important processes such as collisions , binary interactions ( including primordial binaries ) , stellar evolution ( and the accompanying mass loss ) , as well as the effects of a massive central object with relative ease and much higher realism compared to direct fp simulations or gaseous models .
we will incorporate the effects of all these additional processes on cluster evolution in the subsequent papers of this series .
the mc code we have used to obtain the main results of this paper is described in detail by @xcite .
it is based on the ideas of @xcite and in many respects it is very similar to mc codes developed by @xcite , @xcite , and @xcite . in the rest of this section , we give a brief summary of the numerical method .
the main simplifying assumption is the fokker - planck approximation , in which relaxation processes are assumed to be dominated by small distant encounters rather than strong encounters with large deflections @xcite .
the dynamical evolution can then be treated as a diffusion process in phase space . following the individual interactions between the stars , as in direct @xmath2-body simulations , is computationally expensive .
however , the _ average _ cumulative effect on a star in a given amount of time can be characterized by diffusion coefficients in phase space . to compute the relaxation , the timestep for the numerical evolution has to be chosen smaller than the relaxation time . since the relaxation time is normally shortest at the center
, we choose our timestep to be a fraction of the central relaxation time .
this ensures that the relaxation is followed accurately throughout the cluster .
another important simplification in the mc method is the assumption of spherical symmetry .
the position of the particles is represented by a single radial coordinate , @xmath64 , and the velocity is represented by radial , @xmath65 , and tangential , @xmath66 , components .
the specific angular momentum , @xmath67 , and the specific energy , @xmath68 , of a star with index @xmath69 , are given by : @xmath70 where @xmath71 is the gravitational potential at a given point .
the assumption of spherical symmetry implies that the potential at all points can be computed in a time proportional to the number of particles @xmath2 , rather than @xmath48 . in hnon
s algorithm , the evolution is simulated by reproducing the effect of the cluster on each star by a single effective encounter in each timestep . at every iteration
the two integrals of the motion @xmath72 and @xmath73 characterizing the orbit of each star are perturbed in a way that is consistent with the value of the diffusion coefficients @xcite . to conserve energy ,
this perturbation is realized by a single effective scattering between two neighboring stars .
the square of the scattering angle , @xmath74 , is proportional to the timestep chosen .
we choose our timestep such that @xmath75 at the center of the cluster .
choosing too large a timestep will lead to a saturation effect and the relaxation will proceed artificially slowly . using too small a timestep , on the other hand , not only increases the computation time , but also can lead to spurious relaxation ( see the appendix for a discussion of spurious relaxation effects ) .
after the perturbation of @xmath72 and @xmath73 , the stars are placed at random positions between their apocenters and pericenters using a probability distribution that is proportional to the time spent at a given location on their new orbit . for a star with index @xmath69 , the apocenter and pericenter distances are calculated by finding the roots of @xmath76 this random placement is justified by the assumption of dynamical equilibrium , i.e. , the evolution of the system does not take place on the crossing ( or dynamical ) timescale , but rather on the relaxation timescale .
the only important point about assigning a specific position to a particle on its orbit is that its contribution to density , potential , interaction rates , etc . , has to be estimated correctly .
after all stars have been placed at their new positions , the potential is recalculated and the whole cycle of perturbation is repeated .
this method can be modified so that the timestep is a fraction of the _ local _ relaxation time @xcite . @xcite and @xcite divided the system into zones resulting in an approach intermediate between a fixed and smoothly varying timestep . dividing the system into radial zones also allows the mc method to be parallelized efficiently for use on multi - processor machines @xcite .
another possible modification uses a scaling of the units such that each particle in the simulation can represent an entire spherical shell of many identical stars rather than a single star @xcite .
it should also be noted that , although the effects of strong encounters between stars on relaxation are assumed to be negligible compared to weak , more distant encounters , they can be incorporated by estimating their rate of occurrence in a way similar to physical collisions @xcite or interactions with binary stars @xcite . our mc code has been used previously to study many fundamental dynamical processes such as the spitzer instability @xcite and mass segregation @xcite for simple two - component systems , as well as the evolution of systems with a continuous but fairly narrow mass spectrum of evolving stars @xcite .
a difficulty introduced by a broad continuous imf ( with a large @xmath77 ratio ) is the necessity of adjusting the timestep to treat correctly encounters between stars of very different masses .
when pairs of stars are selected to undergo an effective hyperbolic encounter as described above , one has to make sure that the deflection angle remains small for _ both _ stars . in situations where the mass ratio of the pair can be extreme , one has to decrease the timestep accordingly @xcite . in practice , for the simulations described here , we find that the timestep has to be reduced by a factor of up to @xmath78 compared to what would be appropriate for a cluster of equal - mass stars .
we discuss further the applicability of orbit - averaged mc methods to systems with a continuous mass spectrum and large @xmath77 ratio in the appendix .
the characteristics of core collapse and the subsequent runaway collisions depend on the initial conditions for the cluster .
these initial conditions include the total number of stars , the imf , the initial spatial distribution of the stars ( density profile and , possibly , initial mass segregation ) , and the position of the cluster in the galaxy , which determines the tidal boundary .
as we shall see , the most important initial parameters are the slope of the imf , the maximum stellar mass , and the initial degree of central concentration of the cluster density profile .
we have used a wide variety of initial conditions , both to test the robustness of our findings and to establish the dependence of our results on these parameters . as a typical reference model we use an isolated plummer sphere with a salpeter imf and stellar masses ranging from @xmath14 to @xmath15 .
we then explore variations on this model by changing the initial cluster structure , the imf or the number of stars .
we have examined three families of models : plummer and king models ( * ? ? ? * sections 2.2 and 4.4 ) , which have a core - halo structure , and @xmath79-models @xcite , which have a cusp near the center .
all these models have a characteristic radius given by @xmath80 for plummer , king , and @xmath79-models respectively . in these formulae
, @xmath81 is the half - mass radius , @xmath82 is the central density , and @xmath83 is a king model parameter .
we show the density profiles corresponding to these various models in figure [ fig_dens_prof ] . here
@xmath84 is the dimensionless central potential , related to the concentration parameter ( * ? ? ?
other useful quantities characterizing the various density profiles are given in table [ table_clust ] .
the initial model used for each of our mc simulations is listed in the second column of table [ table_models ] .
a general procedure for producing these models is given by @xcite ; a less general but simpler procedure for the plummer model is given by @xcite .
rrllllllll plummer & 1.167 & 0.532 & 0.589 & @xmath85 & 0.769 & 0.417 & 0.192 & 0.093 & 0.0437 + king @xmath86 & 0.454 & 0.534 & 1.517 & 2.568 & 0.858 & 0.670 & 0.321 & 0.110 & 0.1134 + 2 & 0.530 & 0.526 & 1.003 & 2.800 & 0.849 & 0.612 & 0.281 & 0.108 & 0.0930 + 3 & 0.652 & 0.518 & 0.749 & 3.134 & 0.839 & 0.543 & 0.238 & 0.106 & 0.0722 + 4 & 0.860 & 0.510 & 0.576 & 3.625 & 0.827 & 0.465 & 0.195 & 0.104 & 0.0523 + 5 & 1.252 & 0.504 & 0.438 & 4.362 & 0.814 & 0.382 & 0.1546 & 0.101 & 0.0348 + 6 & 2.112 & 0.503 & 0.320 & 5.471 & 0.804 & 0.293 & 0.1171 & 0.100 & 0.0205 + 7 & 4.526 & 0.511 & 0.2146 & 6.987 & 0.812 & 0.2032 & 0.0830 & 0.101 & 0.00997 + 8 & 13.742 & 0.530 & 0.1253 & 8.344 & 0.872 & 0.1211 & 0.0531 & 0.112 & 0.00368 + 9 & 55.671 & 0.558 & 0.0649 & 8.374 & 0.980 & 0.0633 & 0.0307 & 0.134 & 0.00106 + @xmath87 & @xmath85 & 0 & @xmath88 & @xmath85 & 0.805 & 0 & 0 & 0.100 & 0 + @xmath89 & @xmath85 & 0 & @xmath90 & @xmath85 & 0.851 & 0 & 0 & 0.108 & 0 [ table_cluster_prop ] since we expect the conditions leading to core collapse to depend sensitively on the imf , we have carried out simulations with a wide range of imf parameters .
however , it is generally established that , at least for the high - mass end of the spectrum , a universal imf very close to the simple salpeter - like power - law is obeyed .
this is indicated both by observations and by theoretical calculations ( * ? ? ?
* ; * ? ? ?
* and references therein ) . in our simulations
we assign the masses of individual stars using a sampling procedure .
we first choose a random number , @xmath91 , from a uniform distribution between 0 and 1 .
for a simple power - law imf with @xmath92 between @xmath93 and @xmath94 we calculate the corresponding mass using @xmath95 \right)^{1/(1-\alpha)},\ ] ] where the value @xmath96 would correspond to a salpeter imf .
in addition to simple power laws , for two models we have used the miller - scalo @xcite and kroupa @xcite imfs , which are steeper at the high - mass end of the spectrum and shallower at the low - mass end . both of these imfs can be represented by broken power laws . for the miller - scalo imf , @xmath97 and for the kroupa imf , @xmath98 where all numerical values are in solar mass .
however , for the sake of computational convenience , we prefer to implement a somewhat different parametrization of these distributions . to generate the mass spectra
corresponding to these imfs we use the functions @xcite @xmath99 and @xmath100 for the miller - scalo and kroupa mass functions , respectively . in both cases if @xmath101 or @xmath102 the result is discarded and a new random number is generated .
the values in equation ( [ eq_kroupa ] ) correspond to the choice of @xmath103 in table 10 of @xcite .
in most of our simulations we have used @xmath104 stars . for all mass functions ,
this implies the presence of many massive stars , allowing us to resolve fully the higher end of the mass spectrum .
for example , for @xmath14 and @xmath105 , with @xmath106 , we have @xmath107 stars with @xmath108 for a salpeter imf .
the results from our simulations are therefore not much affected by random fluctuations in a small number of very massive stars .
note , however , that in much smaller systems , containing perhaps only @xmath109 stars ( as in the arches and quintuplet clusters near our galactic center ) , these small number effects and random fluctuations may indeed play a dominant role in determining the dynamical fate of the few most massive stars in the system . initial mass segregation in star clusters ( i.e.
, the tendency for more massive stars to be _ formed _ preferentially near the cluster center ) is expected to result from star formation feedback in dense gas clouds @xcite or from competitive gas accretion onto proto - stars and mergers between them @xcite .
there is also some observational evidence for initial mass segregation in both open and globular clusters @xcite .
we have considered the possibility of initial mass segregation in a few of our mc simulations .
we adopt a simple prescription whereby we increase the average stellar mass within a certain radius @xmath110 . for @xmath111 , rather than sampling from an imf with fixed @xmath112 , we randomly choose between two values of @xmath93 , one that is used for the outer part of the cluster and another one that is larger .
for @xmath113 , we follow a similar procedure , this time changing @xmath94 .
the average mass within @xmath110 is larger by a factor @xmath114 and outside @xmath110 is smaller by a factor @xmath115 , with respect to a cluster without initial mass segregation .
we choose @xmath114 and @xmath115 such that the overall average stellar mass in the cluster does not change with these modifications .
changing the average stellar mass within any region of the cluster would of course in general leave the system out of dynamical equilibrium . to maintain virial equilibrium
, the mass density profile must also be preserved .
we achieve this by modifying the number density of stars appropriately .
our initial conditions for models with initial mass segregation are summarized in table [ table_ims ] .
here @xmath116 is the initial cluster mass fraction within @xmath110 .
this implies @xmath117 .
increasing @xmath116 or @xmath114 represents more extended or more pronounced initial mass segregation .
our prescription for initial mass segregation allows the formation of massive stars in the outer parts of the cluster , as well as the formation of lighter stars in the inner parts , but the more massive stars are more likely to be found near the center .
this makes our approach different from that of @xcite , who put all massive stars closer to center . for all our numerical calculations ,
we adopt the standard @xmath2-body units @xcite : we set the initial total cluster mass @xmath118 , the gravitational constant @xmath119 , and the initial total energy @xmath120 . in the tabulation of the results we also use the initial half - mass relaxation time @xmath17 , given by equation ( [ eq_rel_time ] ) , as the unit of time for comparison with other work in the literature .
the conversion to physical units is done by evaluating the initial half - mass relaxation time in years .
for example , for the plummer model , we can write @xmath121 similar expressions can also be obtained for king models and @xmath79-models by use of the quantities in table [ table_cluster_prop ] . for a single - component model ( containing equal - mass stars ) the value of @xmath122 in the coulomb logarithm
can be calculated theoretically @xcite , but for a system with a wide mass spectrum it must be determined by comparing to direct @xmath2-body integrations . @xcite carried out such comparisons and found @xmath123 . our own comparison with a recent @xmath2-body result for @xmath124 stars with a wide mass spectrum led us to adopt the value @xmath125 ( h. baumgardt , private communication ) .
for processes occurring in the central parts of the cluster , the central relaxation time is a more relevant quantity , @xmath126 where @xmath127 , @xmath128 and @xmath129 are the 3d velocity dispersion , number density and average stellar mass at the cluster center ( * ? ? ? * eq . 3.37 ) .
the @xmath2-body unit system only specifies unambiguously _ dynamical times _ , which are independent of @xmath2 . in this system ,
relaxation times are proportional to @xmath130 .
it is therefore useful to define also the so - called `` fokker - planck time unit , '' which is the @xmath2-body time unit ( @xmath131 ) multiplied by @xmath130 .
the most important physical properties of all our initial cluster models , including @xmath17 and @xmath132 , are given in table [ table_cluster_prop ] ( in @xmath2-body and fp units ) .
rllrrrr 55 & plummer & pl , @xmath133 & 0.2120 & 6.57 & 0.287 & @xmath134 + 50 & plummer & pl , @xmath135 & 0.2120 & 2.74 & 0.131 & 0.0024 + 1 & plummer & pl , @xmath136 & 0.2120 & 1.28 & 0.0899 & 0.0018 + 51 & plummer & pl , @xmath137 & 0.2120 & 0.87 & 0.0700 & 0.0020 + 2r & plummer & pl , @xmath138 & 0.2120 & 0.69 & 0.0716@xmath139 & @xmath134 + 2s & plummer & pl , @xmath138 & 0.2120 & 0.69 & 0.0702@xmath139 & @xmath134 + 2 & plummer & pl , @xmath138 & 0.2120 & 0.69 & 0.0700@xmath140 & 0.0020@xmath140 + 2b & plummer & pl , @xmath138 & 0.2120 & 0.69 & 0.0700@xmath140 & 0.0019@xmath140 + 2c & plummer & pl , @xmath138 & 0.2120 & 0.69 & 0.0706 & 0.0020 + 2d & plummer & pl , @xmath138 & 0.2120 & 0.69 & 0.0720 & 0.0018 + 52 & plummer & pl , @xmath141 & 0.2120 & 0.58 & 0.0719 & 0.0022 + 3 & plummer & pl , @xmath142 & 0.2120 & 0.48 & 0.0696 & 0.0018 + 53 & plummer & pl , @xmath143 & 0.2120 & 0.40 & 0.0834 & 0.0012 + 4 & plummer & kroupa & 0.2120 & 0.96 & 0.0858 & 0.0014 + 5 & plummer & miller - scalo & 0.2120 & 0.71 & 0.0723 & 0.0022 + 28 & plummer & pl , @xmath138 & 0.2360 & 0.72 & 0.0795 & 0.0020 + 27 & plummer & pl , @xmath138 & 0.2240 & 0.71 & 0.0760 & 0.0020 + 24 & plummer & pl , @xmath138 & 0.290 & 0.68 & 0.0664 & 0.0014 + 20 & plummer & pl , @xmath138 & 0.260 & 0.67 & 0.0786 & 0.0024 + 21 & plummer & pl , @xmath138 & 0.220 & 0.62 & 0.156 & 0.0028 + 22 & plummer & pl , @xmath138 & 0.28 & 0.56 & 0.478 & 0.0010 + 25 & plummer & pl , @xmath138 & 0.25 & 0.53 & 0.805 & 0.0016 + 23 & plummer & pl , @xmath138 & 0.22 & 0.45 & 2.20 & @xmath134 + 26 & plummer & pl , @xmath138 & 0.21 & 0.37 & 4.29 & @xmath134 + 30 & king , @xmath86 , ( i ) & pl , @xmath138 & 0.2120 & 0.69 & 0.152 & 0.0020 + 36 & king , @xmath86 & pl , @xmath138 & 0.2120 & 0.69 & 0.151 & 0.0014 + 35 & king , @xmath144 , ( i ) & pl , @xmath138 & 0.2120 & 0.69 & 0.134 & 0.0020 + 37 & king , @xmath144 & pl , @xmath138 & 0.2120 & 0.69 & 0.129 & 0.0014 + 11 & king , @xmath145 , ( i ) & pl , @xmath138 & 0.2120 & 0.69 & 0.107 & 0.0022 + 10 & king , @xmath145 & pl , @xmath138 & 0.2120 & 0.69 & 0.110 & 0.0016 + 31 & king , @xmath146 , ( i ) & pl , @xmath138 & 0.2120 & 0.69 & 0.0779 & 0.0022 + 38 & king , @xmath146 & pl , @xmath138 & 0.2120 & 0.69 & 0.0778 & 0.0022 + 32 & king , @xmath147 , ( i ) & pl , @xmath138 & 0.2120 & 0.69 & 0.0561 & 0.0024 + 39 & king , @xmath147 & pl , @xmath138 & 0.2120 & 0.69 & 0.0526 & 0.0020 + 12 & king , @xmath148 , ( i ) & pl , @xmath138 & 0.2120 & 0.69 & 0.0336 & 0.0014 + 40 & king , @xmath148 & pl , @xmath138 & 0.2120 & 0.69 & 0.0322 & 0.0020 + 33 & king , @xmath149 , ( i ) & pl , @xmath138 & 0.2120 & 0.69 & 0.0163 & 0.0014 + 41 & king , @xmath149 & pl , @xmath138 & 0.2120 & 0.69 & 0.0150 & 0.0018 + 34 & king , @xmath150 , ( i ) & pl , @xmath138 & 0.2120 & 0.69 & 0.00545 & 0.0010 + 42 & king , @xmath150 & pl , @xmath138 & 0.2120 & 0.69 & 0.00577 & 0.0012 + 13 & king , @xmath151 , ( i ) & pl , @xmath138 & 0.2120 & 0.69 & 0.00135 & 0.0010 + 43 & king , @xmath151 & pl , @xmath138 & 0.2120 & 0.69 & 0.00138 & 0.0012 + 6 & @xmath152 , hernquist & pl , @xmath138 & 0.2120 & 0.69 & @xmath153 & @xmath134 + 7 & @xmath154 & pl , @xmath138 & 0.2120 & 0.69 & @xmath155 & @xmath134 + rlrrrrrrr m01 & 0.3 & 0.69 & 1.2 & 1.094 & 0.0588 & 0.0018 + m17 & 0.3 & 0.69 & 1.5 & 1.273 & 0.0490 & 0.0020 + m02 & 0.3 & 0.69 & 1.8 & 1.522 & 0.0443 & 0.0022 + m18 & 0.3 & 0.69 & 2.1 & 1.892 & 0.0366 & 0.0022 + m04 & 0.2 & 0.55 & 1.2 & 1.053 & 0.0637 & 0.0030 + m15 & 0.2 & 0.55 & 1.5 & 1.143 & 0.0512 & 0.0026 + m05 & 0.2 & 0.55 & 1.8 & 1.250 & 0.0498 & 0.0020 + m16 & 0.2 & 0.55 & 2.1 & 1.379 & 0.0439 & 0.0022 + m06 & 0.2 & 0.55 & 2.4 & 1.538 & 0.0399 & 0.0022 + m08 & 0.1 & 0.40 & 1.2 & 1.023 & 0.0664 & 0.0016 + m13 & 0.1 & 0.40 & 1.5 & 1.059 & 0.0588 & 0.0018 + m09 & 0.1 & 0.40 & 1.8 & 1.098 & 0.0558 & 0.0016 + m14 & 0.1 & 0.40 & 2.1 & 1.139 & 0.0560 & 0.0020 + m10 & 0.1 & 0.40 & 2.4 & 1.184 & 0.0519 & 0.0018 + m11 & 0.1 & 0.40 & 3.0 & 1.286 & 0.0506 & 0.0022 + m12 & 0.1 & 0.40 & 3.6 & 1.406 & 0.0471 & 0.0016 + [ sec_results ] the initial conditions and main results of all our mc simulations are summarized in tables [ table_models ] and [ table_ims ] .
all models have initially @xmath104 stars , except for models 2r , s , b , c , and d , which have varying @xmath2 between @xmath156 and @xmath157 .
the maximum value of @xmath2 is set in practice by the available computer memory and @xmath158 corresponds to 2 gb of available memory for our code .
the run for model 2d took about two weeks of cpu time to complete on a 2.8 ghz pentium 4 linux workstation .
more typical runs for @xmath106 took around @xmath159 cpu hours . the close agreement between the outcomes of models 2r
2d confirms the expectation that our results should be independent of the number of stars in the system , at least for sufficiently large @xmath2 to avoid small - number effects in the cluster core .
as expected , all models with a salpeter - like imf and a wide mass spectrum undergo core collapse considerably faster than any single - component cluster ( cf . sec . [ cc_and_si ] ) . in figure [ fig_model2 ]
, we show the evolution of various lagrange radii , as well as the average stellar mass inside these radii , for our reference model ( model 2 ; same as shown in figure [ fig_cc_compare ] ) .
in contrast to the evolution of a single - component model , the inner lagrange radii remain almost constant until the very abrupt onset of core collapse at @xmath160 .
a more detailed view of core collapse is shown in figure [ fig_model2_zoom ] . here
the time axis has been replaced by the central potential depth , which increases monotonically in time and provides a natural stretch near core collapse .
note that the core collapse time for a particular run can be determined very accurately , to within @xmath161 . however , random fluctuations in the realization of each initial condition for a particular model lead to a much larger physical `` error bar '' on @xmath162 ( see below ) . the rapid increase of the average stellar mass inside the innermost lagrange radii seen in figures [ fig_model2 ] and [ fig_model2_zoom ] is an indication of significant mass segregation . in fact
, very significant mass segregation takes place throughout the evolution of the system .
this can be seen in the upper panel of figure [ fig_mass_segre ] , which shows the evolution of half - mass radii for stars in various mass bins ( compare with fig . 1 of * ? ? ?
it is clear that the rate of mass segregation in each mass bin , measured by the slope @xmath163 of the corresponding half - mass radius @xmath164 in that bin , is very nearly constant from @xmath165 all the way to core collapse ( the least - square straight - line fits , constrained to @xmath166 at @xmath165 , are shown in the figure )
. this rate can be positive or negative . for model
2 we find that stars more massive than about @xmath167 drift inward on average , while less massive stars drift outward . even for the most massive stars , we do _ not _ find that the mass segregation rate is proportional to the average mass in the bin ( in contrast to what would be expected for a tracer population of massive stars driven by simple dynamical friction ; cf . sec .
1.2 and @xcite ) .
instead , the following simple expression provides a good fit ( to within a few percent ) to the observed mass - dependence for model 2 : @xmath168 where @xmath169 is the average stellar mass and the best - fit parameters are @xmath170 , @xmath171 , and @xmath172 .
note that the mass segregation rate actually approaches a constant for large @xmath169 ( but of course it is unphysical to extrapolate beyond @xmath94 ) . in the lower panel of figure [ fig_mass_segre ]
we show the average mass within @xmath173 , @xmath174 and @xmath175 .
the steady increase of the average mass in each region is further indication that the mass segregation not only starts immediately , but also continues until core collapse . for a smaller number of stars , the average mass within a given radius can reach saturation ( cf .
figure 2 of * ? ? ?
our results for larger @xmath2 do not show this saturation . initially and throughout the evolution until core collapse , the cluster maintains a core - halo structure .
however , when the heaviest stars start dominating the core and the spitzer instability occurs , this initial structure is lost .
we demonstrate this behavior in figure [ fig_pot_evolution ] , where the evolution of the gravitational potential profile is shown .
we have checked that the final structure and , in particular , the formation of a cusp are not dominated by small number effects but are instead the result of many massive stars participating in the core collapse . we illustrate this in figure [ fig_pot_evolution ] where close to 200 innermost stars , which constitute 0.2% of the total mass ,
are shown explicitly .
another important point , which can be deduced from the evolution of the lagrange radii , as well as from figure [ fig_pot_evolution ] ( see the lower two curves ) , is that the outer parts of the cluster are not much affected by mass segregation and core collapse , except for the disappearance of the most massive stars . as expected , the stars in the collapsing core have a mass distribution much richer in heavy stars than the imf .
we compare the mass function at core collapse with the imf , for @xmath176 , in figure [ fig_mf_comp ] .
the mass function shown in this figure is obtained by choosing the innermost 200 stars at three different times all before and within @xmath177 of core collapse time .
mass segregation continues until the core is dominated by massive stars . at this point ,
the massive stars near the center dynamically decouple from the rest of the cluster and go into collapse as a separate subsystem .
the stars outside this subsystem act as an energy sink , and , as a result of their energy gain , expand away from the center of the cluster . in a single - component system ,
as the collapse proceeds , the subset of stars that participate in the collapse becomes smaller and smaller . here instead the stars that participate in the collapse at the edge of the core are heavier than the surrounding stars so they can more efficiently give their energy away and remain together . as a result , in systems with a wide mass spectrum and a steep enough imf
, there is a clear separation between stars in the collapsing core and stars outside the collapsing core , almost suggesting a real condensation process .
the total mass in the collapsing stars is a crucial property of these systems , clearly representing an upper limit to the mass of any bh that could form eventually through the gravitational collapse of a runaway merger remnant . to estimate the total mass in the collapsing core , @xmath178
, we proceed as follows . at every timestep we calculate and record the lagrange radii for various mass fractions . for each lagrange radius , this provides typically a few thousand data points per run . the innermost radii exhibit a great amount of noise for the lowest mass fractions .
we remove this noise in two steps .
first , we take arithmetic averages over 60 points and reduce the total number of points to @xmath179 . then for each point and using the closest 6 points we fit a cubic polynomial using least squares and we evaluate this polynomial at the corresponding time .
this way we both smooth the lagrange radii further and estimate their derivatives . when the system is going into core collapse , a decrease in the collapse rate is an indication that the results provided by our code are beginning to be dominated by numerical errors .
we stop the simulation when this happens and define the core collapse time as the last point before this behavior is observed .
we then find the innermost lagrange radius that has a positive derivative at the time of core collapse and estimate the mass of the stars that participate in core collapse as the mass enclosed by this radius .
an investigation of figure [ fig_model2_zoom ] by eye verifies that this method produces reasonable results . in practice , we track many more lagrange radii than shown in this figure , at intervals of 0.02% between 0.08% and 0.36% . because of small - number fluctuations ( and , in particular , the intrinsic noise in the innermost lagrange radii ) there is a statistical uncertainty in all the numbers quoted in table [ table_models ] .
this uncertainty is not unphysical , as real systems will be affected similarly by fluctuations in the number and specific properties of a relatively small number of massive stars . to test the robustness of our results and
estimate this uncertainty we have repeated 1020 simulations with different random seeds for model 2 and some of its variants with different numbers of stars ( models 2r , 2s , and 2b ) .
for these models the numbers in table [ table_models ] are obtained by averaging the results of the various ( physically equivalent ) runs .
the standard deviations obtained for the values of @xmath180 and @xmath181 for our typical model 2 are about 5% and 20% , respectively .
the larger uncertainty in @xmath181 is a result of the high sensitivity of this quantity to noise in the innermost lagrange radii .
we can not study the evolution of the cluster past core collapse without treating in detail the dynamics of the central stars during collapse , which is beyond the scope of this paper . in future work
we will include a detailed treatment of stellar collisions in the core ( paperii , @xcite ) .
alternatively , one could use a hybrid method and treat the central part of the cluster with a direct @xmath2-body approach @xcite .
however , the simplest approximation may be to introduce an effective boundary condition at some very small radius .
@xcite , in his model b2 , took this approach , but his implementation was not conserving energy .
the sharp onset of core collapse for our reference model , shown in figure [ fig_model2 ] , is a result of the spitzer instability .
this instability is driven by the segregation of the heaviest stars toward the center and their dynamical decoupling from the rest of the system . as indicated by the results of section [ sec_ms_cc ] ,
the ratio of maximum to average stellar mass in the imf , @xmath182 , is an important parameter setting the timescale for the onset of instability .
there are various ways to study the dependence of our results on this parameter .
one can use an imf different from salpeter , e.g. , miller - scalo or kroupa .
in addition , when using a power - law imf , changing the slope @xmath183 or the maximum mass @xmath94 will obviously alter the value of the ratio @xmath184 .
our results from simulations for a large number of models exploring these various alternative imf are presented in figure [ fig_imf_comp ] .
note that for @xmath185 our results suggest a relation @xmath186 , also obeyed by all fokker - planck models from @xcite and @xcite with @xmath187 and a number of mass components sufficient to ensure proper sampling of the imf .
however our computations extend to much higher values of @xmath188 than those works . beyond @xmath189 , a domain reached by any realistic imf
, the core collapse time approaches a constant @xmath190 .
therefore our main conclusions appear to be _ independent of the details of the imf _ as long as the number of massive stars in the system is large enough . for small values of @xmath184 , not only the timescale but also the very nature of the collapse changes .
in these systems the evolution timescale , i.e. , the relaxation time , of the subsystem that can decouple from the rest of the cluster is no longer small enough that the evolution of lighter stars can be neglected .
the lagrange radii for these models behave similarly to the single - component case , i.e. , there is no clear separation between collapsing and expanding lagrange radii .
consequently , we do not give values of @xmath181 for these models in table [ table_models ] .
@xcite , using fp simulations , also found that , for a clear decoupling , the relaxation time of the subsystem has to be short compared to that of the other stars .
one expects the core collapse time to depend strongly on the initial density profile of the cluster , and , in particular , on the central concentration .
a simple and systematic way to examine this dependence is to use different king models with varying concentration parameter or @xmath84 ( * ? ? ?
4 - 10 ) . for single - component clusters
this has been done using a variety of methods @xcite .
we have carried out a number of simulations using king models with @xmath191 as initial structure .
our simulations include both isolated clusters ( with no tidal boundary enforced , even though the initial models are truncated ) and clusters with a tidal boundary ( assuming a circular orbit in a spherical galactic potential ) .
this tidal boundary is initially chosen to be at the tidal radius of the king model and then adjusted as the cluster loses mass @xcite . in figure
[ fig_tcc_trh]a we plot the ratio of core collapse time to initial half - mass relaxation time for our simulations , along with results for single - component clusters .
the ratio @xmath180 is much smaller for systems with mass spectrum , as would be expected .
we also see that the core collapse time for these systems is independent of the presence of a tidal boundary , as expected when core collapse is driven by local processes within the core rather than by global energy transfer .
this idea is strongly supported by the results in figure [ fig_tcc_trh]b , which shows the comparison of the ratio of core collapse time to _ central _ relaxation time , for our simulations and single - component models . with respect to single - component models ,
the ratio @xmath192 for systems with a broad mass spectrum shows very little variation .
we have also run two simulations for @xmath79-models , using @xmath87 ( hernquist model ) , and @xmath154 .
initially , models with @xmath193 have vanishing central velocity dispersion and , for @xmath194 , they exhibit a central `` temperature inversion '' @xcite
. for single - component clusters , the central region at first undergoes rapid gravothermal expansion until it becomes isothermal and `` normal '' core collapse can start @xcite
. models with @xmath194 also have zero initial central relaxation time have _ infinite _
@xmath132 because their velocity dispersion rises like @xmath195 near the center @xcite . ] .
our results for king models would then suggest that core collapse induced by central mass segregation should be extremely fast , if it were not for the opposite effect of the temperature inversion . for @xmath196 , mass segregation seems to have the upper hand in the competition and it proves impossible to resolve core collapse because extreme mass segregation appears nearly instantaneously during the mc runs . models with @xmath152 and @xmath154 expand at first and then evolve to core collapse very rapidly .
the core collapse time and collapsing core mass are very hard to determine numerically for these models . in table
[ table_models ] , we give only an upper limit on their core collapse times .
obviously , these values are so short that their physical relevance is unclear .
it is hard to imagine through which process such a cluster could be created if one wants it to be virialized but without initial mass segregation because these conditions impose constraints on the formation timescale unlikely to be satisfied in real systems . in addition , note that the local dynamical time in a @xmath152 model approaches a constant non - zero value near the center , while , formally , the relaxation time goes to zero there .
this suggests that such models can only provide an approximate description of real clusters , where finite - number effects will play a key role near the center , so that the question of determining the evolution in the limit of very large @xmath2 is not well posed .
naturally we expect that any initial mass segregation in a cluster should lead to an even shorter core collapse time as the heavier stars are starting their life closer to the center on average , and therefore do not need as much time to concentrate there through mass segregation . our results from mc simulations with initial mass segregation ( sec .
3.3 ) are presented in table [ table_ims ] and figure [ fig_ims ] .
as expected , we find that @xmath180 decreases with increasing @xmath114 ( i.e. , when stars become more and more massive on average in the inner region ) , and with increasing @xmath116 ( i.e. , when the increase in average stellar mass takes place over a larger central region ) .
however , the values of @xmath181 do not appear to be affected significantly ( within expected random fluctuations from run to run ) .
the implication for bh formation through runaway collisions is that the final bh mass may not be affected by initial mass segregation , while the condition for runaway growth to occur ( see sec . 5 )
will be relaxed as the core collapse takes place even earlier .
the simulations described so far consider only the dynamical evolution of the cluster , neglecting the stellar evolution entirely . we have also carried out a number of simulations to understand how stellar evolution and the accompanying mass loss can modify the dynamical evolution .
the stellar evolution treatment we have adopted is that of @xcite , based on the approximations of @xcite .
stellar evolution introduces a new physical clock in the system , independent of relaxation .
therefore it is necessary to specify the physical scale of an initial cluster model ( e.g. , the half - mass radius in pc ) before starting a simulation so that the relaxation time @xmath17 can be calculated in years and related to the stellar evolution timescale .
the only other parameter to be specified is the metallicity @xmath197 , which plays an important role in the calculation of stellar evolution mass loss . in our treatment of stellar evolution ,
all wind mass loss rates are proportional to @xmath198 . since stellar evolution introduces two new parameters in our initial models ,
a full exploration of the initial parameter space is clearly impossible .
fortunately , we will see that the effects of mass loss on the evolution to core collapse are rather unimportant , and so a systematic study is unnecessary at this point .
we first illustrate how _ post - main - sequence _ evolution of massive stars can prevent core collapse . in figure [ fig_mass_loss ]
we show the evolution of a system similar to model 2 with @xmath199 and @xmath200 . at such low metallicity
, there is little mass loss on the main sequence , so one expects @xmath201 without evolution beyond the main sequence . however , massive stars evolve off the main sequence after about @xmath202myr , before core collapse has occurred .
once they evolve off the main sequence , they typically lose up to @xmath203 of their mass rapidly . at the end of their life , which extends about 10% beyond their main - sequence lifetime ,
these stars undergo supernova explosions and their cores collapse to bh . in the upper panel of figure [ fig_mass_loss ]
we plot the number of these bh in the cluster .
the significant mass loss during the late stages of stellar evolution causes the cluster core to expand so that collapse is prevented and the system then goes into long - term dynamical evolution .
varying @xmath197 from @xmath204 to 0.02 ( the solar value ) , we found that this reversal of core contraction occurs for any metallicity when massive stars are allowed to evolve off the main sequence .
higher metallicities of course lead to increased mass loss and an even stronger tendency for the cluster core to expand .
we have not considered the case of @xmath205 , i.e. , a cluster of pop iii stars .
these stars may collapse to bh that incorporate essentially the entire initial stellar mass @xcite .
the dynamical evolution of the cluster would then be unaffected by mass loss .
whenever a runaway is avoided , a dense cluster of relatively massive primordial bh would then form near the center .
this is an intriguing possibility to keep in mind for future consideration , although the imf of pop iii stars is essentially unknown ( but see , e.g. , @xcite ) and recent hydrodynamic simulations suggest that the first stars may actually form isolated rather than in clusters @xcite .
we have also examined the possibility that mass loss _ on the main sequence _ may already be a source of indirect heating strong enough to reverse core collapse or , at least , delay it significantly .
our results indicate that wind mass loss from main - sequence stars alone can never prevent core collapse and that the delay introduced remains very small , even for high metallicities .
it is of course always possible to fine - tune the relaxation time so that the mass loss slows down the evolution just enough to allow a few massive stars to evolve off the main sequence and stop contraction .
but such models can only represent a very small domain in the parameter space of initial conditions .
for example , in a system like model 2 with @xmath206 , the core collapse time increases only by about 5% for @xmath207 . for @xmath208 , the evolution
is just delayed long enough to prevent core collapse for @xmath207 , but not for @xmath209 .
one has to increase the relaxation time to @xmath199 for stellar evolution to prevent core collapse for all metallicities . in conclusion
, we find that post - main - sequence mass loss can be strong enough to prevent early core collapse , but that it does not significantly tighten the condition ( on @xmath17 ) for core collapse to occur .
mass loss from main - sequence stars always plays a negligible role .
it has long been realized that a mass spectrum accelerates the dynamical evolution of dense star clusters through the process of mass segregation @xcite .
this is a consequence of the statistical tendency of two - body gravitational encounters to establish energy equipartition between stars of different masses . for any realistic imf
, equipartition can never be achieved because the massive stars quickly form a separate self - gravitating subcluster ( a system with negative effective heat capacity ) as they segregate near the center by transferring kinetic energy to lighter stars @xcite .
this instability leads to a rapid core collapse : after a finite time @xmath162 , the central density of stars would actually become infinite in the absence of finite size effects ( both physical collisions between stars and the finite number of stars in the core ) . with the exception of a few recent @xmath2-body simulations @xcite , all previous investigations of core collapse in clusters with a mass spectrum
have considered a relatively narrow range of stellar masses ( @xmath210 ) .
this restriction is appropriate for old globular clusters in which stellar evolution probably had time to remove all massive stars before significant relaxation took place .
in contrast , in the present study , we ignored stellar evolution to concentrate on systems in which , by assumption , core collapse occurs before even the most massive stars ( with @xmath211 ) leave the ms . through a set of high - resolution monte carlo simulations of clusters with a variety of imf and structural parameters ( concentration , presence or absence of tidal truncation or initial mass segregation ) , we established two important results .
the first concerns the time needed to go deep into core collapse ( @xmath162 ) . as the evolution to core collapse
is driven by relaxation processes , @xmath162 can always be written as proportional to the initial half - mass or central relaxation time ( @xmath17 , @xmath132 ) .
the proportionality constant does not depend on the size of the cluster or the number of stars , as long as they are numerous enough to avoid small - number effects . for isolated clusters with an imf of realistic slope ( @xmath212 in the high - mass range ) ,
we find @xmath213 as @xmath214 increases from 1 ( single - mass ) to @xmath215 .
this result extends and agrees nicely with previous work based on fp simulations @xcite . from simple arguments about dynamical friction
, one would naively expect a linear relation with @xmath216 @xcite .
the steeper dependence found in our numerical results may be related to the fact that the instability causing the core collapse is triggered only after some critical number , @xmath217 , of sufficiently massive stars have drifted to the core by dynamical friction .
if one has to go out to radius @xmath218 to find this number of heavy stars , the required timescale will be of order @xmath219 .
for systems with increasing @xmath214 , it is reasonable to expect @xmath217 and , consequently , @xmath218 and @xmath220 to decrease . while this provides a plausible explanation , further investigations will be needed to confirm it , in particular studies making use of the gaseous model of cluster dynamics @xcite . for @xmath221 ,
the core collapse time saturates to @xmath222 a key result given that any realistic imf is likely to have @xmath223 .
this value of @xmath162 is at least two orders of magnitude shorter than the core collapse time for single - component clusters .
moreover , we find this @xmath192 ratio to hold for all the models we have considered that have a finite @xmath132 .
this includes king models ( with or without tidal truncation ) with @xmath84 ranging from 1 to 9 , a sequence along which @xmath180 decreases by more than @xmath224 .
our result that the core collapse time is set fundamentally by @xmath132 rather than @xmath17 reflects an important difference between single - component clusters and systems with a broad mass spectrum : in systems with a mass spectrum , core collapse is driven ultimately by energy transfer occurring _ locally _ in the core ; in contrast , relaxation in a single - component system is a _ global _
process taking place on all scales , resulting in a timescale comparable to @xmath225 @xcite . for plummer models or king models with moderate concentration ( @xmath226 ) ,
we find @xmath227 , in good agreement with the results of @xmath2-body simulations by @xcite and @xmath132 , @xmath225 is , by far , easier to estimate observationally . ] .
the main - sequence lifetime @xmath228 of very massive stars approaches a constant value , of about @xmath229myr , with only weak dependence on metallicity , rotation , or the mass @xmath169 of the star , as long as @xmath230 @xcite .
it exhibits little variation with @xmath169 because such massive stars are nearly eddington - limited , hence @xmath231 , where @xmath232 is the luminosity , and @xmath233constant if the fractional mass of the convective core , @xmath234 , does not depend too strongly on the stellar mass is compared to the `` usual '' relation @xmath235 , one could conclude erroneously that massive stars never play an important role in core collapse @xcite .
however , the approximate @xmath236 scaling of the stellar lifetime applies only for @xmath237 . ] .
therefore , the necessary condition for core collapse not being stopped by stellar evolution can be written @xmath238 interestingly , we also find that core collapse can not be prevented by mass loss through stellar winds on the ms
. such mass loss can only delay the evolution to core collapse very slightly . even for
high metallicities and hence strong winds , only clusters for which the core collapse time would otherwise be very close to the critical value ( i.e. , just below @xmath228 ) can see their fate changed through wind mass loss if @xmath162 increases to become slightly longer than @xmath228 .
these clusters must represent a very small domain in the parameter space of initial conditions .
our second important finding is that , as the collapse proceeds , the mass contained in the ever - shrinking core converges to a non - zero value , @xmath239 which depends only weakly on the properties of the initial cluster .
this contrasts again with single - component clusters which , exhibiting self - similar collapse , have @xmath240 as @xmath241 .
this value of @xmath181 is in remarkable agreement with the fractional mass of a dark object , presumably a massive bh , discovered at the center of galaxies ( in which case , @xmath242 stands for the mass in the spheroidal component ) and , possibly , globular clusters like m15 and g1 .
this suggests that the collapsing core may well be the progenitor of a central imbh .
however , for a galactic nucleus , the requirement that @xmath243myr would imply that , either it evolved from a cluster with short ( central ) relaxation time to its present - day state with @xmath244yr @xcite , or that the growth of the central bh was predominantly from accretion of smaller star clusters with short relaxation times ( each one carrying an imbh at its center ; @xcite ) .
m15 and g1 also have long half - mass relaxation times ( @xmath245gyr for m15 , @xcite , and @xmath246gyr for g1 , @xcite ) indicating that the process we envision could only have taken place if these clusters were born with much more centrally concentrated density profiles than observed today for a cluster , which can be estimated from observations , must be close to @xmath17 . ]
( for m15 , an _ initial _ value of @xmath247 seems required ; see fig .
[ fig_tcc_trh ] ) . however , there is no doubt that young star clusters with half - mass relaxation times shorter than the critical value of 30myr do exist .
the arches cluster near the center of our galaxy is the best established example , with @xmath248 probably shorter than 10myr @xcite , while the quintuplet cluster appears to lie very close to the critical value @xcite .
it is also possible that some super star clusters , the birthplaces of most stars in starburst environments , have sufficiently short relaxation times ( e.g. * ? ? ?
although the final fate of the stars that participate in core collapse is not completely certain , very high rates of physical collisions are expected in the core .
primordial binaries are probably unable to stop the collapse and prevent these collisions . in smaller systems like globular clusters interactions with hard binaries
may in fact increase the collision rate . in more extreme environments like ( proto-)galactic nuclei ,
the velocity dispersion is so high ( @xmath249 ) that most primordial binaries are soft and will have been disrupted before they can play a role in the collapse .
the main uncertainty is then whether these collisions , occurring at relatively high velocity , allow a massive star to gain mass in a runaway fashion , as suggested by analytical models ( based on the dependence of the cross section on the mass and neglecting collisional mass - loss ; ( * ? ? ?
* and references therein ) ) , or , on the contrary , grind it down progressively . our preliminary results from mc simulations including a realistic treatment of stellar collisions @xcite indicate that the formation of a very massive star through runaway collisions is a likely outcome .
more extensive calculations , as well as a detailed discussion of the likely fate of the merger remnant , will be presented in paper ii .
we thank holger baumgardt for discussions and for providing data from his @xmath2-body simulations .
we are grateful to rainer spurzem for making the code for gaseous models ( _ spedi _ ) available and for his assistance in using it .
piero spinnato and simon portegies zwart kindly gave us the initial condition data for the `` dynamical friction '' run presented in the appendix .
we thank k. belczynski for providing us the population synthesis code _ startrack _ and for his help in integrating it into our mc code , and to john fregeau for useful discussions .
this work was supported by nasa atp grant nag5 - 12044 and nsf grant ast-0206276 to northwestern university .
the work of mf is funded by the sonderforschungsbereich ( sfb ) 439 ` galaxies in the young universe ' ( subproject a5 ) of the german science foundation ( dfg ) at the university of heidelberg .
mf thanks the theoretical astrophysics group at northwestern university for financial support and hospitality during several visits which made this collaboration possible .
the monte carlo method has been applied previously to two - component systems ( see , e.g. , * ? ? ? * ) and to systems with a continuous mass function covering a relatively narrow range ( with @xmath250 ; see the references in section [ sec_mc_sim ] ) . however
, one may question the applicability of the method to systems with a much broader mass spectrum .
in particular , the nature of the energy transfer in these systems is different than in systems with small @xmath77 ratio ( see the discussion in sec .
in addition , the gravitational potential is assumed to change smoothly within the cluster in the mc method .
however , in a system containing a broad imf , a relatively small number of very massive stars could introduce significant discrete changes in the slope of the potential . as the positions of these steps change randomly from one iteration to the next
, one may worry that this could introduce a significant amount of spurious relaxation in the system . in this appendix
, we provide numerical tests and theoretical arguments showing that the mc method is indeed applicable to systems with a very broad mass spectrum , such as those under consideration in this paper . in mc simulations , the orbits of the stars
are calculated using a potential assumed to be spherical and smooth .
this potential is constructed numerically from a radial distribution of point masses and therefore includes random fluctuations that can lead to spurious relaxation effects @xcite . to test the importance of these effects , we carried out simulations in which we turned off physical relaxation ( by skipping the perturbation of stars ) .
ideally , in this case , the cluster should maintain a constant state indefinitely and none of its properties should change over time . in practice , however , it is sufficient to make sure that the changes are not significant compared to changes occurring in the presence of physical relaxation . in one of our tests , we let a mc simulation , for our typical model 2 , run for 3600 iteration steps , which is twice the number of steps normally required for this model to reach core collapse .
we plot various relevant quantities showing the evolution of the system in figure [ fig_spur_relax ] .
it is clear that the deviations in these quantities are much smaller than the changes due to physical relaxation , so we conclude that the effects of spurious relaxation is not significant . the energy transfer from massive to lighter stars and the segregation of massive stars to the center of the cluster
are processes similar to simple dynamical friction for a heavy test particle embedded in a background of lighter stars ( * ? ? ?
some of the processes that govern the dynamics of dense stellar systems can not be treated simply as part of two - body relaxation .
well - known examples include collisions , binary interactions , and three - body formation of binaries .
a natural question to ask is whether dynamical friction is a process that can be modeled correctly by our implementation of two - body relaxation in the mc method .
the answer is central to the applicability of the mc method to our problem .
the frictional drag on a massive object is proportional to the _ mass density _ of light background stars , and is independent of the masses of individual stars ( see * ? ? ?
* eq . 7 - 18 ) .
so the effect of dynamical friction would be unaltered , for example , if every light star were replaced by two stars of half the mass .
however , the treatment of relaxation in the mc method ( sec .
2.1 ) does depend on the mass ratio of the two particles considered and would be altered by such a replacement .
the ratio of relaxation to crossing time is given by @xmath251 which would evidently change if the stars in a system were replaced by many more stars with smaller masses .
this might suggest that the mc method can not follow dynamical friction correctly .
however , note that the relaxation time can be written ( * ? ? ?
* eq . 3 - 37 ) @xmath252 and the timescale for dynamical friction is proportional to ( * ? ? ?
* eq . 7 - 18 ) @xmath253 here , @xmath254 is the average velocity of background stars , of mass @xmath169 and number density @xmath128 , and @xmath255 is the velocity of the massive test particle , of mass @xmath256 .
when @xmath255 is comparable to @xmath254 ( which is true , for example , if the massive particle is on a circular orbit ; see below ) , we conclude that @xmath257 the first proportionality implies that dynamical friction can be treated as a relaxation process . in the second ,
we show explicitly the dependence on the total mass density @xmath33 of the background stars . this result is also established by several numerical calculations @xcite . to demonstrate that the mc algorithm is indeed able to handle simple dynamical friction , we have simulated a system consisting of a single massive object ( mass @xmath20 ) in a cluster of much lighter particles ( mass @xmath19 , number @xmath258 ) . the initial conditions are identical to those used for the @xmath259 model of @xcite , with @xmath260 , hence the total number of particles used for this test is 400001 .
@xcite were interested in the spiral - in of an imbh in a galactic nucleus .
the massive particle is initially on a circular orbit at a relatively large distance from the center , @xmath261 .
for the power - law density profile used to represent the cluster , @xmath262 , one can predict analytically the decay @xmath263 using chandrasekhar s formula for dynamical friction ( * ? ? ?
* eq . 5 ) . in figure [ fig_dyn_fric ] , we compare the spiral - in of the massive particle with the analytical solution . in the formula ,
we have set @xmath264 to fit ( visually ) the density profile of the cluster .
the initial @xmath2-body set - up was provided by @xcite and converted directly into mc particles .
the agreement between the mc run and the analytical formula is very satisfactory , especially considering that the density profile does not follow the exact power law and in fact evolves slightly during the simulation as a result of relaxation between light particles .
note that this good agreement does not validate the assumptions used in deriving chandrasekhar s formula ( e.g. , neglecting large - angle scatterings , correlations between light particles , and gradients in the properties of the cluster ) because the mc method relies basically on the same set of assumptions .
in particular , @xmath265 is a free parameter in both approaches .
what this result demonstrates is that these assumptions are correctly implemented in our mc code . | we study the early dynamical evolution of young , dense star clusters using monte carlo simulations for systems with up to @xmath0 stars
. rapid mass segregation of massive main - sequence stars and the development of the spitzer instability can drive these systems to core collapse in a small fraction of the initial half - mass relaxation time .
if the core collapse time is less than the lifetime of the massive stars , all stars in the collapsing core may then undergo a runaway collision process leading to the formation of a massive black hole . here
we study in detail the first step in this process , up to the occurrence of core collapse .
we have performed about 100 simulations for clusters with a wide variety of initial conditions , varying systematically the cluster density profile , stellar imf , and number of stars .
we also considered the effects of initial mass segregation and stellar evolution mass loss .
our results show that , for clusters with a moderate initial central concentration and any realistic imf , the ratio of core collapse time to initial half - mass relaxation time is typically @xmath1 , in agreement with the value previously found by direct @xmath2-body simulations for much smaller systems .
models with even higher central concentration initially , or with initial mass segregation ( from star formation ) have even shorter core - collapse times .
remarkably , we find that , for all realistic initial conditions , the mass of the collapsing core is always close to @xmath3 of the total cluster mass , very similar to the observed correlation between central black hole mass and total cluster mass in a variety of environments .
we discuss the implications of our results for the formation of intermediate - mass black holes in globular clusters and super star clusters , ultraluminous x - ray sources , and seed black holes in proto - galactic nuclei . |
let @xmath10 and @xmath11 be finite dimensional smooth manifolds .
let @xmath1 , @xmath12 , be @xmath13 vector fields on @xmath10 , @xmath5 real valued @xmath14 functions on @xmath11 , @xmath2 a positive number , and @xmath15 diffusions on a filtered probability space @xmath16 with values in @xmath11 and infinitesimal generator @xmath17 which will be made precise later .
the aim of this paper is to study limit theorems associated to the system of ordinary differential equations on @xmath10 , @xmath18 under the assumption that @xmath5 ` averages ' to zero .
the ` average ' is with respect to the unique invariant probability measure of @xmath7 , in case @xmath7 satisfies strong hrmander s condition , and more generally the ` average ' is the projection to a suitable function space .
we prove that @xmath9 converges as @xmath19 to a markov process whose markov generator has an explicit expression .
this study is motivated by problems arising from stochastic homogenization .
it turned out that in the study of randomly perturbed systems with a conserved quantity , which does not necessarily take value in a linear space , the reduced equations for the slow variables can sometimes be transformed into ( [ 1 ] ) .
below , in section [ section - examples ] we illustrate this by 4 examples including one on the orthonormal frame bundle over a riemannian manifold . of these examples ,
the first is from @xcite where we did not know how to obtain a rate of convergence , and the last three from @xcite where a family of interpolation equations on homogeneous manifolds are introduced .
an additional example can be found in @xcite . in all the examples , which we described in
[ section - examples ] below , the scalar functions average to @xmath20 with respect to a suitable probability measure on @xmath11 .
bearing in mind that if a hamiltonian system approximates a physical system with error @xmath2 on a compact time interval , over a time interval of size @xmath21 the physical orbits deviate visibly from that of the hamiltonian system unless the error is reduced by oscillations , it is natural and a classical problem to study odes whose right hand sides are random and whose averages in time are zero .
the objectives of the present article are : ( 1 ) to prove that , as @xmath2 tends to zero , the law of @xmath22 converges weakly to a probability measure @xmath23 on the path space over @xmath10 and to describe the properties of the limiting markov semigroups ; ( 2 ) to estimate the rate of convergence , especially in the wasserstein distance . for simplicity
we assume that all the equations are complete . in sections [ section - formula ] , [ section - uniform - estimates ] ,
[ section - weak ] and [ section - rate ] we assume that @xmath7 is a regularity improving fredholm operator on a compact manifold @xmath11 , see definition [ def - fredholm ] . in theorem [ thm - weak ] we assume , in addition , that @xmath7 has fredholm index @xmath20 .
but strong hrmander s condition can be used to replace the condition ` regularity improving fredholm operator of index @xmath20 ' . for simplicity , throughout the introduction ,
@xmath5 are bounded and belong to @xmath24 where @xmath25 is the kernel of @xmath26 , the adjoint of the unbounded operator @xmath7 in @xmath27 with respect to the volume measure . in case @xmath7
is not elliptic we assume in addition that @xmath28 or @xmath29 , depending on the result .
the growth conditions on @xmath1 are in terms of a control function @xmath30 and a controlled function space @xmath31 where @xmath32 indicates the order of the derivatives to be controlled , see ( [ bvr ] ) . for simplicity
we assume both @xmath10 and @xmath11 are compact . in section [ section - estimates ] we present two elementary lemmas , lemma [ lemma3 ] and lemma [ lemma3 - 2 ] , assuming @xmath7 mixes exponentially in a weighted total variation norm with weight @xmath33 . in section [ section - formula ]
, for @xmath7 a regularity improving fredholm operator and @xmath34 a @xmath35 function , we deduce a formula for @xmath36 where the transmission of the randomness from the fast motion @xmath15 to the slow motion @xmath37 is manifested in a martingale .
this provides a platform for the uniform estimates over large time intervals , weak convergences , and the study of rate of convergence in later sections . in section [ section - uniform - estimates ]
, we obtain uniform estimates in @xmath2 for functionals of @xmath38 over @xmath39 $ ] .
let @xmath7 be a regularity improving fredholm operator , @xmath40 , and @xmath30 a @xmath35 function such that @xmath41 , @xmath42 for some numbers @xmath43 and @xmath44 .
then , theorem [ uniform - estimates ] , for every numbers @xmath45 there exists a positive number @xmath46 such that @xmath47 is finite and belongs to @xmath48 as a function of @xmath49 .
this leads to convergence in the wasserstein distance and will be used later to prove a key lemma on averaging functions along the paths of @xmath50 . in section
[ section - weak ] , @xmath7 is an operator on a compact manifold @xmath11 satisfying hrmander s condition and with fredholm index @xmath20 ; @xmath10 has positive injectivity radius and other geometric restrictions .
in particular we do not make any assumption on the ergodicity of @xmath7 .
let @xmath51 denote @xmath52 where @xmath53 is a basis of the kernel of @xmath7 and @xmath54 the dual basis in the kernel of @xmath26 .
theorem [ thm - weak ] states that , given bounds on @xmath1 and its derivatives and for @xmath55 where @xmath56 , @xmath22 converges weakly , as @xmath57 , to the markov process with markov generator @xmath58 .
this follows from a tightness result , proposition [ tightness ] where no assumption on the fredholm index of @xmath7 is made , and a law of large numbers for sub - elliptic operators on compact manifolds , lemma [ lln ] .
convergences of @xmath59 in the wasserstein @xmath60-distance are also obtained . in section [ section - sde ]
we study the solution flows of sdes and their associated kolmogorov equations , to be applied to the limiting operator @xmath61 in section [ section - rate ] .
otherwise this section is independent of the rest of the paper .
let @xmath62 be @xmath13 and @xmath63 vector fields respectively . if @xmath10 is compact , or more generally if @xmath1 are @xmath64 vector fields , the conclusions in this section holds , trivially .
denote @xmath65 the set of functions whose derivatives up to order @xmath32 are controlled by a function @xmath30 , c.f.([bvr ] ) .
let @xmath66 be the solution flow to @xmath67 let @xmath68 and @xmath69 .
let @xmath70 and @xmath71 for every @xmath72 .
this assumption on @xmath30 is implied by the following conditions : @xmath73 , @xmath74 , where @xmath75 are constants .
let @xmath76 .
we assume , in addition , for some number @xmath43 the following hold : @xmath77 then there is a global smooth solution flow @xmath66 , theorem [ limit - thm ] .
furthermore for @xmath78 , @xmath79 , @xmath80 , and @xmath81 . for @xmath82 ,
an example of the control pair is : @xmath83 and @xmath84 .
our conditions are weaker than those commonly used in the probability literature for @xmath85 , in two ways .
firstly we allow non - bounded first order derivative , secondly we allow one sided conditions on the drift and its first order derivatives . in this regard ,
we extend a theorem of w. kohler , g. c. papanicolaou @xcite where they used estimations from o. a. oleinik- e. v. radkevi @xcite .
the estimates on the derivative flows , obtained in this section , are often assumptions in applications of malliavin calculus to the study of stochastic differential equations .
results in this section might be of independent interests .
let @xmath86 be the markov semigroup generated by @xmath61 . in section [ section - rate ] ,
we prove the following estimate : @xmath87 where @xmath88 is a constant , @xmath89 is a function in @xmath48 and @xmath90 the solution to ( [ 1 ] ) with initial value @xmath49 . the conditions on the vector fields @xmath1 are similar to ( [ control - conditions ] )
, we also assume the conditions of theorem [ uniform - estimates ] and that @xmath7 satisfies strong hrmander s condition .
we incorporated traditional techniques on time averaging with techniques from homogenization .
the homogenization techniques was developed from @xcite which was inspired by the study in m. hairer and g. pavliotis @xcite .
for the rate of convergence we were particularly influenced by the following papers : w. kohler and g. c. papanicolaou @xcite and g. c. papanicolaou and s.r.s .
varadhan @xcite .
denote @xmath91 the probability distributions of the random variables @xmath92 and @xmath93 the probability measure determined by @xmath86 .
the under suitable conditions , @xmath94 , where @xmath32 is any positive number less or equal to @xmath95 and @xmath96 denotes the wasserstein 1-distance , see [ wasserstein ] .
we contrive to impose as little as possible on the vector fields @xmath97 , hence a few sets of assumptions are used . for the examples we have in mind , @xmath11 is a compact lie group acting on a manifold @xmath10 , and so for simplicity @xmath11 is assumed to be compact throughout the article , with few exceptions . in a future study , it would be nice to provide some interesting examples in which @xmath11 is not compact . if @xmath10 is also compact , only the following two conditions are needed : ( a ) @xmath7 satisfies strong hrmander s condition ; ( b ) @xmath98 where @xmath25 is the annihilator of the kernel of @xmath26 and @xmath32 is a sufficiently large number . if @xmath7 is elliptic , ` @xmath14 ' can be replaced by ` bounded measurable ' . for the convergence condition
( a ) can be replaced by ` @xmath7 satisfies hrmander s condition and has fredholm index @xmath20 ' .
if @xmath7 has a unique invariant probability measure , no condition is needed on the fredhom index of @xmath7 . * theorem [ thm - weak ] and corollary [ convergence - wasserstein - p ] . * under the conditions of proposition [ tightness ] and assumption [ assumption - hormander ] , @xmath99 converges weakly to the markov process determined by @xmath100 where @xmath101 is the dimension of the kernel of @xmath7 which , by the assumption that @xmath7 has fredholm index @xmath20 , equals the dimension of the kernel of @xmath26 .
the set of functions @xmath102 is a basis of @xmath103 and @xmath104 its dual basis . in case
@xmath7 satisfies strong hrmander s condition , then there is a unique invariant measure and @xmath105 is simply the average of @xmath106 with respect to the unique invariant measure .
let @xmath45 be a number and @xmath30 a lyapunov type function such that @xmath107 , a function space controlled by @xmath30 .
if furthermore assumption [ assumption2-y ] holds , @xmath108 converges , on @xmath109 $ ] , in the wasserstein @xmath60-distance . + * theorem [ rate ] . *
denote @xmath110 the solution flow to ( [ 1 ] ) and @xmath86 the semigroup for @xmath61 . if assumption [ assumption - on - rate - result ] holds then for @xmath111 , @xmath112 where @xmath113 and @xmath114 is a constant increasing in @xmath115 .
similarly , if @xmath116 , @xmath117 where @xmath118 is a function in @xmath48 independent of @xmath34 and @xmath119 are increasing functions .
+ a complete connected riemannian manifold is said to have bounded geometry if it has strictly positive injectivity radius , and if the riemannian curvature tensor and its covariant derivatives are bounded
. * proposition [ proposition - rate ] . *
suppose that @xmath10 has bounded geometry , @xmath120 , and assumption [ assumption - on - rate - result ] holds .
let @xmath23 be the limit measure and @xmath121 .
then for every @xmath122 there exists @xmath123 and @xmath124 s.t . for all @xmath125 and @xmath126 , @xmath127 besides the fact that we work on manifolds , where there is the inherited non - linearity and the problem with cut locus , the following aspects of the paper are perhaps new .
( a ) we do not assume there exists a unique invariant probability measure on the noise and the effective processes are obtained by a suitable projection , accommodating one type of degeneracy .
furthermore the noise takes value in another manifold , accommodating ` removable ' degeneracy .
for example the stochastic processes in question lives in a lie group , while the noise are entirely in the directions of a sub - group .
( b ) we used lyapunov functions to control the growth of the vector fields and their derivatives , leading to estimates uniform in @xmath2 and to the conclusion on the convergence in the wasserstein topologies . a key step for the convergence
is a law of large numbers , with rates , for sub - elliptic operators ( i.e. operators satisfying hrmander s sub - elliptic estimates ) .
( c ) instead of working with iterated time averages we use a solution to poisson equations to reveal the effective operator .
functionals of the processes @xmath9 splits naturally into the sum of a fast martingale , a finite variation term involving a second order differential operator in hrmander form , and a term of order @xmath2 . from this
we obtain the effective diffusion , in explicit hrmander form .
it is perhaps also new to have an estimate for the rate of the convergence in the wasserstein distance .
finally we improved known theorems on the existence of global smooth solutions for sdes in @xcite , c.f .
theorem [ limit - thm ] below where a criterion is given in terms of a pair of lyapunov functions .
new estimates on the moments of higher order covariant derivatives of the derivative flows are also given .
we review , briefly , basic ideas from existing literature on random ordinary differential equations with fast oscillating vector fields .
let @xmath128 , where @xmath129 are measurable functions , for which a birkhoff ergodic theorem holds whose limit is denoted by @xmath130 .
the solutions to the equations @xmath131 over a time interval @xmath132 $ ] , can be approximated by the solution to the averaged equation driven by @xmath130 .
if @xmath133 , we should observe the solutions in the next time scale and study @xmath134 .
see r. l. stratonovich @xcite .
suppose for some functions @xmath135 and @xmath136 the following estimates hold uniformly : @xmath137 then under a ` strong mixing ' condition with suitable mixing rate , the solutions of the equations @xmath134 converge weakly on any compact interval to a markov process .
this is a theorem of r. l. stratonovich @xcite and r. z. khasminskii@xcite , further refined and explored in khasminskii @xcite and a. n. borodin @xcite .
these theorems lay foundation for investigation beyond ordinary differential equations with a fast oscillating right hand side . in our case
, noise comes into the system via a @xmath7-diffusion satisfying hrmander s conditions , and hence we could by pass these assumptions and also obtain convergences in the wasserstein distances . for manifold valued stochastic processes ,
some difficulties are caused by the inherited non - linearity .
for example , integrating a vector field along a path makes sense only after they are parallel translated back .
parallel transports of a vector field along a path , from time @xmath138 to time @xmath20 , involves the whole path up to time @xmath138 and introduces extra difficulties ; this is still an unexplored territory wanting further investigations . for the proof of tightness , the non - linearity causes particular difficulty if the riemannian distance function is not smooth .
the advantage of working on a manifold setting is that for some specific physical models , the noise can be untwisted and becomes easy to deal with . our estimates for the rate of convergence , section [ section - rate ] and [ wasserstein ] ,
can be considered as an extension to that in w. kohler and g. c. papanicolaou @xcite , which were in turn developed from the following sequence of remarkable papers : r. coghurn and r. hersh @xcite , j.b .
keller and g. c. papanicolaou @xcite , r. hersh and m. pinsky @xcite , r. hersh and g. c. papanicolaou @xcite and g. c. papanicolaou and s.r.s .
varadhan @xcite .
see also t. kurtz @xcite and @xcite by d. stroock and s. r. s. varadhan .
the condition @xmath139 needs not hold for this type of scaling and convergence . if @xmath140 , where @xmath141 is a stationary process with values in @xmath142 , and @xmath143 , the hamiltonian vector field associated to a function @xmath144 whose level sets are closed connected curves without intersections , then @xmath145 converge to a markov process , under suitable mixing and technical assumptions .
see a. n. borodin and m. freidlin @xcite , also m. freidlin and m. weber @xcite where a first integral replaces the hamiltonian , and also x .- m . li @xcite where the value of a map from a manifold to another is preserved by the unperturbed system . in m. freidlin and a. d.wentzell @xcite ,
the following type of central limit theorem is proved : @xmath146 converges to a markov diffusion .
this formulation is not suitable when the conserved quantity takes value in a non - linear space .
for the interested reader , we also refer to the following articles on limit theorems , averaging and homogenization for stochastic equations on manifolds : n. enriquez , j. franchi , y. lejan @xcite , i. gargate , p. ruffino @xcite , n. ikeda , y. ochi @xcite , y. kifer @xcite , m. liao and l. wang @xcite , s. manade , y. ochi @xcite , y. ogura @xcite , m. pinsky @xcite , and r. sowers @xcite .
\(1 ) i am grateful to the associate editor for pointing out the paper by c. liverani and s. olla @xcite , where random perturbed harmiltonian system , in the context of weak interacting particle systems , is studied .
their system is somewhat related to the completely integrable equation studied in @xcite leading to a new problem which we now state .
denote @xmath147 the hamiltonian vector field on a symplectic manifold corresponding to a function @xmath34 .
if the symplectic manifold is @xmath148 with the canonical symplectic form , @xmath147 is the skew gradient of @xmath34 .
suppose that @xmath149 is a completely integrable system , i.e. they are poisson commuting at every point and their hamiltonian vector fields are linearly independent at almost all points .
following @xcite we consider a completely integrable sde perturbed by a transversal hamiltonian vector field : @xmath150 suppose that @xmath151 commutes with @xmath152 for @xmath153 , then each @xmath154 is a first integral of the unperturbed system .
then , ( * ? ? ?
* th 4.1 ) , within the action angle coordinates of a regular value of the energy function @xmath155 , the energies @xmath156 converge weakly to a markov process . when restricted to the level sets of the energies , the fast motions are ellipitic
. it would be desirable to remove the ` complete integrability ' in favour of hormander s type conditions .
there is a non - standard symplectic form on @xmath157 with respect to which the vector fields in @xcite are hamiltonian vector fields and when restricted to level sets of the energies the unperturbed system in @xcite satisfies hrmander s condition , see @xcite , and therefore provides a motivating example for further studies .
finally note that the driving vector fields in ( [ 1 ] ) are in a special form , results here would not apply to the systems in @xcite nor that in @xcite , and hence it would be interesting to formulate and develop limit theorems for more general random odes to include these two cases .
\(2 ) it should be interesting to develop a theory for the odes below @xmath158 where @xmath5 depends also on the @xmath159 process .
\(3 ) it would be nice to extend the theory to allow the noise to live in a non - compact manifold , in which @xmath7 should be an ornstein - uhlenbeck type operator whose drift term would provide for a deformed volume measure .
_ notation .
_ throughout this paper @xmath160 , @xmath161 , and @xmath162 denote the set of functions from @xmath10 to @xmath25 that are respectively bounded measurable , @xmath14 with compact supports , and bounded @xmath14 with bounded first @xmath32 derivatives . if @xmath163 the letter @xmath25 will be suppressed .
also @xmath164 denotes the space of bounded linear maps ; @xmath165 denotes @xmath14 vector fields on a manifold @xmath10 .
let @xmath166 be independent real valued brownian motions on a given filtered probability space , @xmath167 denote stratonovich integrals .
in the following @xmath168 and @xmath169 are smooth vector fields , and @xmath170 is an orthonormal basis at each point of the vertical tangent spaces . to be brief
, we do not specify the properties of the vector fields , instead refer the interested reader to @xcite for details .
for any @xmath171 , the stochastic differential equations @xmath172 are degenerate and they interpolate between the geodesic equation ( @xmath173 ) and brownian motions on the fibres ( @xmath174 ) . the fast random motion is transmitted to the horizontal direction by the action of the lie bracket @xmath175 $ ] . if @xmath176 , there is a conserved quantity to the system which is the projection from the orthonormal bundle to the base manifold .
this allows us to separate the slow variable @xmath37 and the fast variable @xmath15 . the reduced equation for @xmath37 , once suitable ` coordinate maps ' are chosen , can be written in the form of ( [ 1 ] ) .
in @xcite we proved that @xmath177 converges weakly to a rescaled horizontal brownian motion .
recently j. angst , i. bailleul and c. tardif gave this a beautiful treatment , @xcite , using rough path analysis .
by theorems in this article , the above model can be generalised to include random perturbation by hypoelliptic diffusions , i.e. @xmath170 generates all vertical directions . in @xcite
we did not know how to obtain a rate for the convergence .
theorem [ rate ] , in this article , will apply and indeed we have an upper bound for the rate of convergence . as a second example
, we consider , on the special orthogonal group @xmath178 , the following equations : @xmath179 where @xmath180 is an orthonormal basis of @xmath181 , as a subspace of @xmath182 , and @xmath183 is a skew symmetric matrix orthogonal to @xmath181 .
the above equation is closely related to the following set of equations : @xmath184 where @xmath185 are two positive numbers .
if @xmath186 and @xmath187 , the solutions are brownian motions on @xmath188 . if @xmath189 and @xmath190 , the solutions are unit speed geodesics on @xmath178 .
these equations interpolate between a brownian motion on the sub - group @xmath188 and a one parameter family of subgroup on @xmath178 .
see @xcite .
take @xmath191 and let @xmath192 , what could be the ` effective limit ' if it exists ? the slow components of the solutions , which we denote by @xmath193 , satisfy equations of the form ( [ 1 ] ) .
they are ` horizontal lifts ' of the projections of the solutions to @xmath194 .
if @xmath195 is the orthogonal complement of @xmath181 in @xmath182 then @xmath195 is @xmath196-irreducible and @xmath196-invariant , noise is transmitted from @xmath197 to every direction in @xmath195 , and this in the uniform way .
it is therefore plausible that @xmath198 can be approximated by a diffusion @xmath199 of constant rank .
the projection of @xmath200 to @xmath194 is a scaled brownian motion with scale @xmath201 .
the scale @xmath201 is a function of the dimension @xmath202 , but is independent of @xmath183 and is associated to an eigenvalue of the laplacian on @xmath188 , indicating the speed of propagation . as a third example we consider the hopf fibration @xmath203 .
let @xmath204 be the pauli matrices , they form an orthonormal basis of @xmath205 with respect to the canonical bi - invariant riemannian metric .
@xmath206 denote @xmath207 the left invariant vector field generated by @xmath208 . by declaring @xmath209 an orthonormal frame
, we obtain a family of left invariant riemannian metrics @xmath210 on @xmath211 .
the berger s spheres , @xmath212 , converge in measured gromov - hausdorff topology to the lower dimensional sphere @xmath213 .
for further discussions see k. fukaya @xcite and j. cheeger and m. gromov @xcite .
let @xmath214 be a one dimensional brownian motion and take @xmath215 from @xmath216 .
the infinitesimal generator of the equation @xmath217 satisfies weak hrmander s conditions .
the ` slow motions ' , suitably sacled , converge to a ` horizontal ' brownian motion whose generator is @xmath218 , where the trace is taken in @xmath195 .
a slightly different , ad hoc , example on the hopf fibration is discussed in @xcite .
an analogous equations can be considered on @xmath219 where the diffusion coefficients come from a maximal torus .
finally we give an example where the noise @xmath15 in the reduced equation is not elliptic .
let @xmath220 and let @xmath221 be the elementary @xmath222 matrices and @xmath223 . for @xmath224 and @xmath225 , we consider the equations @xmath226 the slow components of the solutions of these equations again satisfy an equation of the form ( [ 1 ] ) .
let @xmath7 be a diffusion operator on a manifold @xmath11 and @xmath227 its transition semigroup and transition probabilities .
let @xmath228 denote the total variation norm of a measure , normalized so that the total variation norm between two probability measures is less or equal to @xmath229 . by the duality formulation for the total variation norm , @xmath230 for @xmath231
denote @xmath232 the weighted supremum norm and @xmath233 the weighted total variation norm : @xmath234 [ assumption1 ] there is an invariant probability measure @xmath235 for @xmath7 , a real valued function @xmath236 with @xmath237 , numbers @xmath238 and @xmath239 such that @xmath240 if @xmath11 is compact we take @xmath241 . in the following lemma
we collect some elementary estimates , which will be used to prove lemma [ lemma3 ] and [ lemma3 - 2 ] , for completeness their proofs are given in the appendix .
write @xmath242 .
[ lemma2 ] assume assumption [ assumption1 ] .
let @xmath243 be bounded measurable functions and let @xmath244 .
then the following statements hold for all @xmath245 . 1 .
let @xmath246 be an @xmath7 diffusion . if @xmath247 , @xmath248 2 .
let @xmath246 be an @xmath7 diffusion . if @xmath249 then @xmath250 3 .
suppose that either @xmath251 or @xmath252 .
let @xmath253 let @xmath15 be an @xmath254 diffusion .
then for every @xmath171 , @xmath255 to put assumption [ assumption1 ] into context , we consider hrmander type operators .
let @xmath256 denote lie differentiation in the direction of a vector field @xmath257 and @xmath258 $ ] the lie bracket of two vector fields @xmath257 and @xmath215 .
let @xmath259 be a family of smooth vector fields on a compact smooth manifold @xmath11 and @xmath260 .
if @xmath261 and their lie brackets generate the tangent space @xmath262 at each point @xmath263 we say that the operator @xmath7 satisfies the strong hrmander s condition .
[ lemma1 ] suppose that @xmath7 satisfies the strong hrmander condition on a compact manifold @xmath11 and let @xmath264 be its family of transition probabilities .
then assumption [ assumption1 ] holds with @xmath265 identically @xmath266 .
furthermore the invariant probability measure @xmath235 has a strictly positive smooth density w.r.t . the lebesgue measure and @xmath267 for all @xmath263 in @xmath11 and for all @xmath268 . by hrmander s theorem
there are smooth functions @xmath269 such that @xmath270 .
furthermore @xmath269 is strictly positive , see j .-
bony @xcite and a. sanchez - calle @xcite .
let @xmath271 .
thus deblin s condition holds : if @xmath272 denotes the volume of a borel set @xmath273 , @xmath274 .
we say that @xmath265 is a @xmath275 lyapunov function for the ergodicity problem if there are constant @xmath276 and @xmath277 s.t .
@xmath278 . if such a function exists , the @xmath254 diffusions are conservative .
suppose that the lyapunov function @xmath30 satisfies in addition the following conditions : there exists a number @xmath279 and @xmath280 s.t .
for every @xmath281 , @xmath282 then there exists a unique invariant measure @xmath235 such that assumption [ assumption1 ] holds , see e.g. m. hairer and j. mattingly @xcite .
we mention the following standard estimates which helps to understand the estimates in lemma [ lyapunov - ergodicity ] .
[ lyapunov - ergodicity ] let @xmath265 be a @xmath275 lyapunov function for the ergodicity problem of @xmath7 , @xmath283 is uniformly bounded in @xmath2 for @xmath2 sufficiently small .
then there exist numbers @xmath124 and @xmath43 s.t . for all @xmath268 , @xmath284 by localizing @xmath15 if necessary , we see that @xmath285 is a martingale .
let @xmath276 and @xmath277 be constant s.t . @xmath278
. then @xmath286 .
as an application we see that , under the assumption of lemma [ lyapunov - ergodicity ] , the functions @xmath287 in part ( 3 ) of lemma [ lemma2 ] satisfy that @xmath288 .
[ definition - complete ] we say that a stochastic differential equation ( sde ) on @xmath10 is _ complete _ or _ conservative _ if for each initial point @xmath289 any solution with initial value @xmath290 exists for all @xmath291 .
let @xmath292 be its solution starting from @xmath263 .
the sde is _ strongly complete _ if it has a unique strong solution and that @xmath293 is continuous for a.s .
@xmath294 . from now on , by a solution
we always mean a globally defined solution . for @xmath295
we define @xmath296 .
let @xmath297 denote their transition semigroups and transition probabilities . for each @xmath171 ,
let @xmath15 be an @xmath254 diffusion .
let @xmath298 and @xmath37 solutions to the equations @xmath299 let @xmath300 be the solution flow to ( [ sde-3 ] ) with @xmath301 .
we denote by @xmath302 the average of an integrable function @xmath303 with respect to @xmath235 .
let @xmath304 [ lemma3 ] suppose that assumption [ assumption1 ] holds .
let @xmath305 and @xmath306 .
suppose that @xmath5 are bounded .
then for any @xmath307 , @xmath308 and @xmath309 , @xmath310 here @xmath311 we first expand @xmath312 at @xmath313 : @xmath314 by part ( 3 ) of lemma [ lemma2 ] @xmath315 it remain to estimate the summands in the second term , whose absolute value is bounded by the following @xmath316 for @xmath317 $ ] , we apply part ( 3 ) of lemma [ lemma2 ] to bound the inner iterated integral , @xmath318 we bring this back to the previous line , the notation @xmath319 will be used , @xmath320 putting everything together we see that , for @xmath321 given in the lemma , @xmath322 , @xmath323 the proof is complete . in section [ uniform - estimates ] we will estimate @xmath324 and give uniform , in @xmath2 , moment estimates of functionals of @xmath37 on @xmath325 $ ] .
[ lemma3 - 2 ] assume that @xmath15 satisfies assumption [ assumption1 ] and @xmath326 are bounded . if @xmath327 and @xmath328 , then for all @xmath329 , @xmath330 where @xmath331 we note that , @xmath332 letting @xmath333 , it is clear that for @xmath334 , @xmath335 to the second term we apply lemma [ lemma3 ] and obtain the bound @xmath336 where @xmath337 adding the error estimates together we conclude the proof .
it is worth noticing that if @xmath338 is a concave function @xmath339 is again a lyapunov function .
thus by using @xmath340 if necessary , we may assume uniform bounds on @xmath341 and further estimates on the conditional expectation in the error term are expected from cauchy - schwartz inequality
. if @xmath11 is compact then @xmath342 is bounded . in corollary
[ corollary - to - lemma3 - 2 ] , we will give uniform estimates on moments of @xmath343 .
let @xmath11 be a smooth manifold of dimension @xmath202 with volume measure @xmath344 .
let @xmath345 denote the sobolev space of real valued functions over a manifold @xmath11 and @xmath346 the sobolev norm .
the norm @xmath347 extends from domains in @xmath348 to compact manifolds , e.g. by taking supremum over @xmath349 on charts . if @xmath350 , @xmath351 is the completion of @xmath352 with the norm @xmath353 where @xmath354 is usually taken as the levi - civita connection ; when the manifold is compact this is independent of the riemannian metric . and
@xmath355 if and only if for any function @xmath356 , @xmath357 in any chart belongs to @xmath351 .
let @xmath259 be a family of smooth vector fields on @xmath11 and let us consider the hrmander form operator @xmath260 .
let @xmath358 , [ x_{i_1 } , [ x_{i_2},x_{i_3 } ] ] , i_j = 0,1 , \dots , m'\}.\ ] ] if the vector fields in @xmath359 generate @xmath262 at each @xmath360 , we say that hrmander s condition is satisfied . by the proof in a theorem of hrmander ( * ? ? ?
* theorem1.1 ) , if @xmath7 satisfies the hrmander condition then @xmath361 is a @xmath362 function in every open set where @xmath363 is a @xmath364 function .
there is a number @xmath238 such that there is an @xmath365 improvement in the sobolev regularity : if @xmath361 is a distribution such that @xmath366 , then @xmath367 .
suppose that @xmath11 is compact .
then @xmath368 , the resolvents @xmath369 as operators from @xmath370 to @xmath370 are compact , and @xmath7 is fredholm on @xmath371 , by which we mean that @xmath7 is a bounded linear operator from @xmath372 to @xmath371 and has the fredholm property : its range is closed and of finite co - dimension , the dimension of its kernel , @xmath373 is finite .
the domain of @xmath7 is endowed with the norm @xmath374 .
let @xmath26 denote the adjoint of @xmath7 .
then the kernel @xmath25 of @xmath26 is finite dimensional and its elements are measures on @xmath10 with smooth densities in @xmath371 .
denote @xmath375 the annihilator of @xmath25 , @xmath376 is in @xmath24 if and only if @xmath377 for all @xmath378 .
since @xmath7 has closed range , @xmath379 can be identified with the range of @xmath7 , and the set of @xmath380 such that the poisson equation @xmath381 is solvable is exactly @xmath375 .
we denote by @xmath382 a solution .
furthermore @xmath382 is @xmath14 whenever @xmath380 is @xmath14 .
denote by @xmath383 , @xmath384 , the index of a fredholm operator @xmath7 , where @xmath385 .
if @xmath7 is self - adjoint , @xmath386 .
[ def - fredholm ] we say that @xmath7 is a regularity improving fredholm operator , if it is a fredholm operator and @xmath387 is @xmath14 whenever @xmath388 .
let @xmath389 be a family of independent real valued brownian motions .
we may and will often represent @xmath254-diffusions @xmath15 as solutions to the following stochastic differential equations , in stratonovich form , @xmath390 [ lemma5 ] let @xmath7 be a regularity improving fredholm operator on a compact manifold @xmath11 , @xmath391 , and @xmath392 .
let @xmath393 be global solutions of ( [ sde-3 ] ) on @xmath10 .
then for all @xmath394 , @xmath171 and @xmath395 , @xmath396 suppose that , furthermore , for each @xmath171 , @xmath397 , @xmath398 is finite .
then , @xmath399 firstly , for any @xmath35 function @xmath400 , @xmath401 since the @xmath326 s are @xmath35 so are @xmath402 , following from the regularity improving property of @xmath7 .
we apply it s formula to the functions @xmath403 . to avoid extra regularity conditions ,
we apply it s formula to the function @xmath404 , which is @xmath405 , and the @xmath275 functions @xmath402 separately and follow it with the product rule .
this gives : @xmath406 substitute this into the earlier equation , we obtain ( [ ito - tight ] ) .
part ( [ ito - tight-2 ] ) is obvious , as we note that @xmath407 and the stochastic integrals are @xmath408-martingales , so ( [ ito - tight-2 ] ) follows . when @xmath11 is compact , @xmath409 is bounded and the condition becomes : @xmath410 is finite , which we discuss below
otherwise , assumptions on @xmath411 is needed .
if @xmath412 is a locally bounded function such that @xmath413 we say that @xmath30 is a pre - lyapunov function .
let @xmath298 .
let @xmath97 be @xmath405 smooth vector fields on @xmath10 such that : either ( a ) each @xmath1 grows at most linearly ; or ( b ) there exist a pre - lyapunov function @xmath414 , positive constants @xmath43 and @xmath44 such that @xmath415 then the equations ( [ sde-3 ] ) are complete . in case ( a ) let @xmath416 and @xmath417 be a constant such that @xmath418 where @xmath419 denotes the riemannian distance function on @xmath10 . for @xmath263 fixed , denote @xmath420 .
by the definition of the riemannian distance function , @xmath421 this together with the inequality @xmath422 implies that for all @xmath45 , there exist constants @xmath423 depending on @xmath60 such that @xmath424 where @xmath425 . when restricted to @xmath426 , the first time @xmath427 reaches the cut locus ,
the bounded is simple @xmath428 . in case ( b ) , for any @xmath72 , @xmath429 which followed easily from the bound @xmath430 for the convenience of comparing the above estimates , which are standard and expected , with the uniform estimates of @xmath37 in theorem [ uniform - estimates ] below in the time scale @xmath21 , we record this in the following lemma .
[ standard - estimates ] let @xmath431 .
let @xmath295 , @xmath432 , @xmath433 . 1 .
if @xmath97 grow at most linearly then ( [ sde-3 ] ) is complete and there exists @xmath434 s.t .
@xmath435 2 .
if there exist a pre - lyapunov function @xmath414 , positive constants @xmath43 and @xmath44 such that @xmath436 , then ( [ sde-3 ] ) is complete .
if ( [ sde-3 ] ) is complete and there exists @xmath414 such that @xmath436 then there exists a constant @xmath119 , s.t .
@xmath437 if @xmath438 is a positive function , denote by @xmath31 the following classes of functions : @xmath439 in particular , @xmath48 is the class of continuous functions bounded by a function of the form @xmath440 . in @xmath348 ,
the constant functions and the function @xmath441 are potential ` control ' functions .
[ assumption2-y ] assume that ( i ) ( [ sde-3 ] ) are complete for every @xmath295 , ( ii ) @xmath442 is finite for every @xmath72 ; and ( iii ) there exist a function @xmath443 , positive constants @xmath43 and @xmath44 such that @xmath444 below we assume that @xmath7 satisfies hrmander s condition .
we do not make any assumption on the mixing rate .
let @xmath392 , @xmath445 , @xmath446 , @xmath447 , and @xmath448 .
we recall that if @xmath5 and @xmath7 satisfy assumption [ assumption - hormander ] then @xmath7 is a regularity improving fredholm operator .
[ uniform - estimates ] let @xmath7 be a regularity improving fredholm operator on a compact manifold @xmath11 , and @xmath449 .
assume that @xmath1 satisfy assumption [ assumption2-y ] .
then for all @xmath45 , there exists a constant @xmath450 s.t . for any @xmath432 and all @xmath125 , @xmath451 here
@xmath452 depends on @xmath453 and @xmath454 .
let @xmath432 .
we apply ( [ ito - tight ] ) to @xmath455 : @xmath456 in the following estimates , we may first assume that @xmath457 and @xmath458 are bounded .
we may then replace @xmath138 by @xmath459 where @xmath460 is the first time that either quantity is greater or equal to @xmath202 .
we take this point of view for proofs of inequalities and may not repeat it each time .
we take the supremum over @xmath461 $ ] followed by conditional expectation of both sides of the inequality : @xmath462 by the conditional jensen inequality and the conditional doob s inequality , there exists a universal constant @xmath463 depending only on @xmath60 s.t .
, @xmath464 since @xmath465 and @xmath466 , there are constants @xmath467 and @xmath468 such that @xmath469 and @xmath470 .
this leads to the following estimate : @xmath471 let @xmath472 . for @xmath125 , @xmath473 it follows that there exists a constant @xmath474 such that for @xmath475 , @xmath476 _ remark .
_ if @xmath82 , @xmath477 are vector fields with bounded first order derivatives , then @xmath478 is a pre - lyapunov function satisfying the conditions of theorem [ uniform - estimates ] , hence theorem [ uniform - estimates ] holds .
let us recall that @xmath31 is defined in ( [ bvr ] ) .
we return to lemma [ lemma3 - 2 ] in section [ section - estimates ] to obtain a key estimation for the estimation in section [ section - rate ] .
let us recall that @xmath31 is defined in ( [ bvr ] ) .
[ corollary - to - lemma3 - 2 ] assume ( [ sde-3 ] ) is complete , for every @xmath479 , and conditions of assumption [ assumption1 ] .
let @xmath480 be a locally bounded function and @xmath46 a positive number s.t . for all @xmath72 and @xmath481 , there exists a locally bounded function @xmath482 , a real valued polynomial @xmath483 such that for @xmath484 and for all @xmath125 @xmath485 let @xmath486 . if @xmath487 is a function s.t . @xmath488 and @xmath489 for all @xmath490 , then for all @xmath432 , @xmath491 here @xmath492 is a constant , see ( [ cpsi ] ) below , and @xmath493 for all @xmath494 and @xmath45 , @xmath495 more explicitly , if @xmath496 where @xmath497 are constants , then there exists a constant @xmath88 depending only on the differential equation ( [ sde-3 ] ) s.t .
@xmath498 by lemma [ lemma3 - 2 ] , @xmath499 since @xmath265 is bounded so is @xmath342 , which is bounded by @xmath500 .
furthermore @xmath501 we gather all constant together , @xmath502 it is clear that , @xmath503 since @xmath34 , @xmath504 and @xmath505 , by ( [ moment - assumption-20 ] ) , the following quantities are finite for all @xmath45 : @xmath506 furthermore since @xmath496 , @xmath507 consequently , @xmath508 , completing the proof .
below @xmath509 denotes the injectivity radius of @xmath10 and @xmath510 is the riemannian distance function on @xmath10 from a point @xmath290 .
let @xmath511 denote a point in @xmath10 .
the following proposition applies to an operator @xmath7 , on a compact manifold , satisfying hrmander s condition .
[ tightness ] let @xmath10 be a manifold with positive injectivity radius and @xmath124 .
suppose conditions ( 1 - 5 ) below or conditions ( 1 - 3 ) , ( 4 ) and ( 5 ) . 1 .
@xmath7 is a regularity improving fredholm operator on @xmath27 for a compact manifold @xmath11 ; 2 .
@xmath512 ; 3 .
suppose that for @xmath513 , ( [ sde-3 ] ) is complete and @xmath514 ; 4 .
suppose that there exists a locally bounded function @xmath30 s.t . for all @xmath125 and for any @xmath515 , and for all @xmath45 , @xmath516 where @xmath517 , @xmath518 , and @xmath519 is a natural number ;
@xmath520 are locally bounded in @xmath138 .
there exist a function @xmath443 , positive constants @xmath43 and @xmath44 such that @xmath444 6 .
for @xmath30 in part ( 4 ) or in part ( 4 ) , suppose that for some number @xmath238 , @xmath521 then there exists a distance function @xmath522 on @xmath10 that is compatible with the topology of @xmath10 and there exists a number @xmath523 such that @xmath524 is tight . by theorem [ uniform - estimates ] , conditions ( 1 - 3 ) and ( 4 )
imply condition ( 4 ) .
( a ) let @xmath525 .
let @xmath526 be a smooth convex function such that @xmath527 when @xmath528 and @xmath529 when @xmath530 .
then @xmath531 is a distance function with @xmath532 .
its open sets generate the same topology on @xmath10 as that by @xmath419 .
let @xmath402 be a solution to @xmath533 .
for any @xmath534 , @xmath535 . since @xmath536 , @xmath537
. we may apply ( [ ito - tight-2 ] ) in lemma [ lemma5 ] , @xmath538 in the above equation , differentiation of @xmath539 is w.r.t . to the second variable . by construction @xmath540
is bounded by @xmath266 and @xmath541 .
furthermore since @xmath326 are @xmath275 functions on a compact manifold , so @xmath402 and @xmath542 are bounded . for any @xmath534 , @xmath543 .
thus @xmath544 recall @xmath545 and there are numbers @xmath468 and @xmath546 s.t .
@xmath547 , so @xmath548 let @xmath549 , it is clear that @xmath550 .
we remark that , @xmath551 by the assumption , there exists a function @xmath552 s.t . @xmath553 for @xmath554 , it is better to estimate directly from ( [ sde-3 ] ) : @xmath555 where @xmath556 .
we interpolate these estimates and conclude that for some function @xmath557 and a constant @xmath43 the following holds : @xmath558 .
there is a function @xmath559 s.t .
@xmath560 in the last step we use assumption ( 4 ) on the initial value . by kolmogorov
s criterion , there exists @xmath523 such that @xmath561 are equi uniformly hlder continuous on any compact time interval .
consequently the family of stochastic processes @xmath562 is tight .
if @xmath7 is the laplace - beltrami operator on a compact riemannian manifold and @xmath235 its invariant probability measure then for any lipschitz continuous function @xmath563 , @xmath564 where @xmath565 denotes the oscillation of @xmath34 .
if @xmath7 is not elliptic we suppose it satisfies hrmander s conditions and has index @xmath20 .
the dimension of the kernel of @xmath26 equals the dimension of the kernel of @xmath7
. let @xmath566 be a basis in @xmath567 and @xmath568 the dual basis for the null space of @xmath26 .
for @xmath569 we define @xmath570 where the bracket denotes the dual pairing between @xmath408 and @xmath571 .
[ lln ] suppose that @xmath246 is a markov process on a compact manifold @xmath11 with generator @xmath572 satisfying hrmander s condition and having fredholm index @xmath20 .
then for any function @xmath573 , where @xmath574 , there is a constant @xmath119 depending on @xmath575 , s.t .
@xmath576 since @xmath577 , @xmath578 .
by working with @xmath579 we may assume that @xmath580 and let @xmath380 be a solution to @xmath581 . by hrmander s theorem , @xcite , there is a positive number @xmath365 , such that for all @xmath582 , @xmath583 the number @xmath584 where @xmath585 is related to the number of brackets needed to generate the tangent spaces .
furthermore every @xmath361 such that @xmath586 must be in @xmath351 . if @xmath587 , @xmath351 is embedded in @xmath405 and for some constant @xmath588 , @xmath589 recall that @xmath590 .
let @xmath591 be independent one dimensional brownian motions .
let @xmath246 be solutions of @xmath592 .
since @xmath34 is @xmath35 , @xmath593 we apply the sobolev estimates to @xmath380 and use doob s @xmath408 inequality to see that for @xmath594 there is a constant @xmath119 such that , @xmath595 we remark that a self - adjoint operator satisfying hrmander s condition has index zero . [ weak - convergence ] suppose that @xmath7 satisfies hrmander s condition .
in addition it has fredholm index @xmath20 or it has a unique invariant probability measure .
let @xmath574 .
let @xmath596 be such that @xmath597 for each @xmath290 and that @xmath598 .
let @xmath329 be a pair of positive numbers , and @xmath599 ; m)\to { { \mathbb r}})$ ] .
for any equi -uniformly continuous subsequence , @xmath600 , of @xmath177 that converges weakly to a continuous process @xmath601 as @xmath602 , the following convergence holds weakly : @xmath603 where @xmath604 . for simplicity
we omit the subscript @xmath202 .
the required convergence follows from lemma 4.3 in @xcite where it was assumed that ( [ lln1 ] ) holds and @xmath7 has a unique invariant measure for @xmath605 .
it is easy to check that the proof there is valid .
we take care to replace @xmath606 in lemma 4.3 there by @xmath607 .
we remark that by the regularity improving property each @xmath608 is smooth and therefore bounded . in the first part of the proof
, we divide @xmath461 $ ] into sub - intervals of size @xmath2 , freeze the slow variable @xmath609 on @xmath610 $ ] , and approximate @xmath611 by @xmath612 on each sub - interval @xmath610 $ ] .
this approximation is clear : the computation is exactly as in lemma 4.3 of @xcite and we use the uniform continuity of @xmath37 , the fact that @xmath613 and @xmath614 are finite .
the convergence of @xmath615 follows from the law of large numbers in lemma [ lln ] .
the convergence of @xmath616 is also clear and follows from the lipschitz continuity of @xmath617 in the first variable and the equi continuity of the @xmath159 path .
finally denote by @xmath618}$ ] the restriction of the path @xmath619 to the interval @xmath620 $ ] , the weak convergence of @xmath621 } ) \int_s^t \
< h(y_{u\over\epsilon}^\epsilon , \cdot ) , \pi_i\>du$ ] to the required limit is trivial , as explained in lemma 4.3 , @xcite .
[ assumption - hormander ] the generator @xmath7 satisfies hrmander s condition and has fredholm index @xmath20 ( or has a unique invariant probability measure ) . for @xmath622 , @xmath623 for some @xmath624 . if @xmath7 is elliptic , it is sufficient to assume @xmath298 , instead of @xmath55 .
[ thm - weak ] if @xmath7 , @xmath5 , @xmath625 and @xmath626 satisfy the conditions of proposition [ tightness ] and assumption [ assumption - hormander ] , then @xmath99 converge weakly to the markov process determined by the markov generator @xmath627 by proposition [ tightness ] , @xmath628 is tight .
we prove that any convergent sub - sequence converges to the same limit .
let @xmath629 be a a monotone sequence converging to zero such that the probability distributions of @xmath630 converge weakly , on @xmath631 $ ] , to a measure @xmath23 .
for notational simplicity we may assume that @xmath628 converges to @xmath632 .
let @xmath494 , @xmath633 the canonical filtration , @xmath634 the canonical process , and @xmath635}$ ] its restriction to @xmath620 $ ] . by the stroock - varadhan martingale method , it is sufficient to prove @xmath636 is a local martingale for any @xmath637 .
by ( [ ito - tight ] ) , the following is a local martingale , @xmath638 since the third term converges to zero as @xmath8 tends to zero , it is sufficient to prove @xmath639 this follows from lemma [ weak - convergence ] , completing the proof . [ convergence - wasserstein - p ] let @xmath45 be a number and suppose that @xmath107 .
then , under the conditions of theorem [ thm - weak ] and assumption [ assumption2-y ] , @xmath108 converges in the wasserstein @xmath60-distance on @xmath640;m)$ ] . by theorem [ uniform - estimates ] , @xmath641 .
let @xmath642 denote the wasserstein @xmath60 distance : @xmath643 here the infimum is taken over all probability measures on the path spaces @xmath640;m)$ ] with marginals @xmath644 and @xmath645 . note that @xmath640;m)$ ] is a banach space , a family of probability measures @xmath646 converges to @xmath605 in @xmath642 , if and only if the following holds : ( 1 ) it converges weakly and ( 2 ) @xmath647 .
the conclusion follows .
the primary aim of the section is to study the properties of @xmath648 for @xmath649 where @xmath86 is the semigroup for a generic stochastic differential equation .
these results will be applied to the limit equation , to provide the necessary a priori estimates .
theorem [ limit - thm ] should be of independent interest , it also lead to lemma [ semigroup - estimate-2 ] , which will be used in section [ section - rate ] . throughout this section
@xmath10 is a complete smooth riemannian manifold .
let @xmath183 be @xmath63 and @xmath650 be @xmath13 smooth vector fields on @xmath10 , @xmath651 independent real valued brownian motions .
let @xmath652 be the maximal solution to the following equation @xmath653 with initial value @xmath290 .
its markov generator is @xmath654 .
let @xmath655 be the drift vector field , so @xmath656 if there exists a @xmath275 pre - lyapunov function @xmath30 , constants @xmath43 and @xmath44 such that @xmath657 then ( [ limit.sde ] ) is complete .
however we do not limit ourselves to lyapunov test for the completeness of the sde .
let us denote @xmath658 and @xmath659 .
the following observation is useful .
[ lf - l2f ] let @xmath438 be locally bounded .
* suppose that @xmath660 and @xmath661
. then if @xmath662 , @xmath663 . if @xmath664 , @xmath665 where @xmath666 , not depending on @xmath34 .
* suppose that @xmath667 if @xmath78 , @xmath80 . if @xmath668 , @xmath669 where @xmath670 is a function in @xmath48 , not dependent of @xmath34 . that @xmath671 belongs to @xmath48 follows from ( [ generator-0 ] ) . if @xmath672 , @xmath673 .
for the second part we observe that @xmath674 involves at most four derivatives of @xmath34 and two derivatives of @xmath675 and @xmath676 where @xmath677 .
let @xmath678 denote the derivative flow in the direction of @xmath679 .
it is the derivative of the function @xmath680 , in probability .
moreover , it solves the following stochastic covariant differential equation along the solutions @xmath681 , @xmath682 here @xmath683 where @xmath684 is the stochastic parallel transport map along the path @xmath685 .
denote @xmath686 the norm of @xmath687 . for @xmath688 , @xmath289 and @xmath679 , we define @xmath689 by @xmath690 let @xmath691 .
its upper bound will be used to control @xmath692 .
[ y - condition ] the equation ( [ limit.sde ] ) is complete . conditions ( i ) and ( ii ) , or ( i ) and ( ii ) , below hold .
* there exists a locally bounded function @xmath480 , s.t . for all @xmath72 and @xmath126
, there exists a number @xmath693 and a polynomial @xmath483 such that @xmath694 * there exists @xmath695 and constants @xmath43 and @xmath44 such that @xmath696 * let @xmath697 . for some constant @xmath43 , @xmath698 _ remark .
_ suppose that ( [ limit.sde ] ) is complete . since @xmath699 , ( i ) implies ( i )
in fact , @xmath700 . recall that ( [ limit.sde ] )
is strongly complete if @xmath701 is continuous almost surely on @xmath132\times m$ ] for ant @xmath268 .
[ limit - thm ] under assumption [ y - condition ] , the following statements hold . 1 .
the sde ( [ limit.sde ] ) is strongly complete and for every @xmath126 , @xmath702 is @xmath703 .
furthermore for all @xmath45 , there exists a positive number @xmath704 such that @xmath705 2 .
let @xmath706 .
define @xmath707 . then @xmath708 and @xmath709 .
furthermore for a constant @xmath704 independent of @xmath34 , @xmath710 3 .
suppose furthermore that @xmath711 then , ( a ) @xmath712 ; ( b ) if @xmath662 , then @xmath713 , and @xmath714 furthermore , ( c ) @xmath715 , and @xmath716 .
4 . let @xmath717 .
suppose furthermore that @xmath718 then @xmath719 belongs to @xmath48 .
if @xmath720 , then @xmath721 .
the statement on strong completeness follows from the following theorem , see thm . 5.1 in @xcite
suppose that ( [ limit.sde ] ) is complete .
if @xmath722 is a function and @xmath723 a number such that for all @xmath268 , @xmath44 compact , and all constants @xmath201 , @xmath724 then ( [ limit.sde ] ) is strongly complete .
furthermore for every @xmath45 there exists a constant @xmath725 such that @xmath726 since @xmath675 are @xmath13 , then for every @xmath138 , @xmath702 is @xmath703 .
it is easy to verify that condition ( [ assumption - strong - compl ] ) is satisfied .
in fact , by the assumption @xmath727 . take @xmath697 then for @xmath45 , @xmath728 this proves part ( 1 ) . for part
( 2 ) let @xmath729 .
then @xmath730 is differentiable for almost every @xmath294 .
let @xmath731\to m$ ] be a geodesic segment with @xmath732 .
then @xmath733 since @xmath734 is locally bounded in @xmath290 , @xmath735 is continuous and the expectation of the right hand side converges to @xmath736 .
the left hand side clearly converges almost surely .
since @xmath737 is locally bounded the convergence is in @xmath738 .
we proved that @xmath708 . furthermore , suppose that @xmath739 , @xmath740 the latter , as a function of @xmath290 , belongs to @xmath48 .
we proceed to part ( 3a ) .
let @xmath741 and @xmath742 .
then @xmath743 satisfies the following equation : @xmath744 it follows that , @xmath745 to the first term on the right hand side we apply cauchy schwartz inequality to split the first term in the inner product and the second term in the inner product .
this gives : @xmath746 and other terms that does not involve @xmath743 .
the stratonovich corrections will throw out the extra derivative @xmath747 which does not involve @xmath743 .
the second term on the right hand side is a sum of the form @xmath748 for which only bound on @xmath749 is required , and @xmath750 the second term is bounded by @xmath751 by the assumption , there exist @xmath752 such that , for every @xmath622 , @xmath753 there is a stochastic process @xmath754 , which does not involve @xmath743 , and constants @xmath755 such that @xmath756 by induction @xmath757 has moments of all order which are bounded on compact intervals . by gronwall s inequality , for @xmath126 , @xmath758 to obtain the supremum inside the expectation , we simply use doob s @xmath759 inequality before taking expectations .
with the argument in the proof of part ( 1 ) we conclude that @xmath760 is finite and belongs to @xmath48 . _ part ( 3b ) .
_ let @xmath662 . by part ( 1 ) , @xmath761 .
let @xmath762 . by an argument analogous to part ( 3 )
, we may differentiate the right hand side under the expectation to obtain that @xmath714 hence @xmath713 .
this procedure can be iterated . _
part ( 3c ) .
_ by it s formula , @xmath763 since @xmath764 , the expectations of the stochastic integrals with respect to the brownian motions vanish .
since @xmath765 by part ( 3 ) , @xmath766 is bounded in @xmath408 .
it follows that the function @xmath767 is continuous , @xmath768 and we obtain kolmogorov s backward equation , @xmath769 .
since @xmath770 , we apply the above argument to @xmath771 , and take @xmath138 to zero in @xmath772 and obtain that @xmath773 .
this leads to the required statement @xmath774 . _
part ( 4 ) .
_ for higher order derivatives of @xmath775 we simply iterate the above procedure and note that the linear terms in the equation for @xmath776 are always of the same form . with the assumption of part ( 3 )
, we can show that for all integer @xmath60 , @xmath777 . if we assume the additional conditions that @xmath778 the conclusion of the remark follows more easily . with the assumptions of part ( 5 ) we need to work a bit more which we illustrate below .
let @xmath779 . instead of writing down all term in @xmath780 we classify the terms in @xmath780 into two classes : those involving @xmath743 and those not .
for the first class we must assume that they are bounded by @xmath781 for some @xmath43 .
for the second class we may use induction and hence it is sufficient to assume that they belong to @xmath48 .
the terms that involving @xmath743 are : @xmath782 the essential identity to use is : @xmath783 we do not need to assume that the second order derivatives @xmath784 , it is sufficient to assume that for @xmath785 and @xmath786 for all @xmath12 . with a bit of care , we check that only one sided derivatives of @xmath676 are involved
. for example we can convert it to the @xmath787 case , @xmath788 by the first term @xmath789 we mean that in place of @xmath790 plug in all terms on the right hand side of the equation for @xmath791 , after formally converting the integrals to it form .
by @xmath792 we mean the bracket of the martingale term on the right hand side of @xmath791 .
it is now easy to check that in all the terms that involving @xmath743 , higher order derivatives of @xmath1 does not appear , except in the form of @xmath793 .
assume the sde is complete .
suppose that for some positive number @xmath119 , @xmath794 then for all @xmath45 , there exists a constant @xmath704 such that @xmath795 furthermore the statements in theorem [ limit - thm ] hold for @xmath796 .
recall that @xmath797 and @xmath798 .
[ semigroup - estimate-2 ] assume assumption [ y - condition ] and @xmath799 then there exist constants @xmath800 , @xmath467 and @xmath801 depending on @xmath138 and @xmath34 and locally bounded in @xmath138 , also functions @xmath802 , @xmath803 polynomials , such that for @xmath329 , @xmath804 denote @xmath805 , the solution to ( [ limit.sde ] ) .
then for @xmath806 , @xmath807 since @xmath808 and @xmath809 , @xmath1 belong to @xmath810 , by assumption [ y - condition](i ) , @xmath811 is finite and the last term vanishes .
hence @xmath812 . by lemma [ lf - l2f ] , @xmath663 if @xmath813 .
let @xmath814 be s.t .
@xmath816 by the assumption , we see easily that @xmath817 . by assumption [ y - condition ] ,
@xmath818 and the first conclusion holds .
we repeat this procedure for @xmath819 to obtain : @xmath820 the last term also vanishes , as every term in @xmath821 belongs to @xmath48 . indeed
@xmath822 this gives , for all @xmath111 , @xmath823 let @xmath824 be numbers such that @xmath825 .
then , @xmath826 consequently , there exist a constant @xmath827 s.t .
@xmath828 completing the proof for @xmath813 and @xmath65 .
next suppose that @xmath672 .
by lemma [ lf - l2f ] , @xmath829 , and @xmath830 if @xmath116 .
here @xmath831 and do not depend on @xmath34 .
we iterate the argument above to complete the proof for @xmath116 .
if @xmath7 has a unique invariant probability measure @xmath235 and @xmath832 denote @xmath833 .
let @xmath834 .
let @xmath835 be the entries in a square root of the matrix @xmath836 .
they satisfy @xmath837 and are constants .
let us consider the sde : @xmath838 where @xmath651 are independent one dimensional brownian motions .
let @xmath839 the results from section [ section - sde ] apply .
recall that @xmath840 and @xmath15 are @xmath841 diffusions .
let @xmath842 be the solution to the sde ( [ 1 ] ) : @xmath843 with initial value @xmath290 .
[ assumption - on - rate - result ] @xmath11 is compact , @xmath844 , and @xmath845 for @xmath622 .
conditions ( 1)-(5 ) below hold or conditions ( 1 ) , ( 2 ) and ( 3 - 5 ) hold . 1 .
the sdes ( [ limit.sde-2 ] ) and ( [ sde-3 ] ) are complete .
2 . @xmath480 is a locally bounded function and @xmath46 a positive number s.t . for all @xmath72 and @xmath481
, there exists a locally bounded function @xmath482 , a real valued polynomial @xmath483 such that for @xmath484 and for all @xmath125 @xmath846 3 .
there exists a function @xmath847 s.t . for all @xmath848 , @xmath849 and @xmath850 .
4 . for @xmath30
defined above , let @xmath697 .
suppose that @xmath851 5 .
@xmath7 satisfies hrmander s conditions and has a unique invariant measure @xmath235 satisfying assumption [ assumption1 ] .
@xmath852 .
we emphasize the following : [ rate - remark-1 ] * if @xmath30 in ( 2 ) is a pre - lyapunov function , then ( [ sde-3 ] ) is complete .
furthermore @xmath853 and so ( [ limit.sde-2 ] ) is complete . * under conditions ( 1 ) , ( 2 ) and ( 4 - 5 ) , ( 2 ) holds .
see theorem [ uniform - estimates ] . also corollary [ corollary - to - lemma3 - 2 ] holds .
conditions ( 1 - 5 ) implies the conclusions of theorem [ limit - thm ] . *
if @xmath7 satisfies strong hrmander s condition , condition ( 4 ) is satisfied .
let @xmath854 be the probability semigroup associated with @xmath855 and @xmath86 the markov semigroup for @xmath61 .
recall that @xmath856 .
we recall that operator @xmath7 on a compact manifold @xmath11 satisfying strong hrmander s condition has an exponential mixing rate , so @xmath7 satisfy assumption [ assumption1 ]
. [ rate ] assume that @xmath857 and @xmath572 satisfy assumption [ assumption - on - rate - result ] .
for every @xmath111 , @xmath112 where @xmath113 and @xmath114 are constant increasing in @xmath115 .
similarly , if @xmath116 , @xmath858 where @xmath118 is a function in @xmath48 that does not depend on @xmath34 and @xmath114 are constants increasing in @xmath115 .
_ step 1 . _ to obtain optimal estimates we work on intervals of order @xmath2 , c.f .
lemma [ lemma3 ] .
let @xmath859 be a partition of @xmath631 $ ] with @xmath860 for @xmath861 and @xmath862 .
. then , @xmath864 define @xmath865 since @xmath78 , lemma [ semigroup - estimate-2 ] applies and obtain the desired estimate on the second term : @xmath866 where @xmath867 is a constant and @xmath868 a polynomial .
let @xmath497 be constants such that @xmath869 .
we apply ( [ moment - assumption-2 ] ) from assumption [ assumption - on - rate - result ] to see that for some constant @xmath870 depending on @xmath871 , @xmath872 since @xmath873 and @xmath874 , @xmath875 if @xmath34 belongs to @xmath876 , we apply lemma [ semigroup - estimate-2 ] to see that there exists a function @xmath877 , independent of @xmath34 s.t . @xmath878 hence @xmath879 the rest of the proof is just as for the case of @xmath78 .
_ let @xmath880 .
by part ( 3 ) of theorem [ limit - thm ] , @xmath881 for any @xmath268 and @xmath882 .
we will approximate @xmath883 by @xmath884 and estimate the error @xmath885 by lemma [ lf - l2f ] , @xmath79 , and we may apply lemma [ semigroup - estimate-2 ] to @xmath886 .
we have , @xmath887 recall that @xmath888 . summing over @xmath889 and
take the expectation of the above inequality we obtain that @xmath890 if @xmath672 , @xmath891 . by lemma [ semigroup - estimate-2 ] ,
@xmath887 there exist constant @xmath114 and a function @xmath892 , independent of @xmath34 , s.t .
@xmath893 here @xmath113 .
thus for @xmath672 , @xmath894 finally instead of estimating @xmath895 , we estimate @xmath896 _ step 3 . _
if @xmath111 , by theorem [ limit - thm ] , @xmath897 for any @xmath138 .
since @xmath898 , we may apply lemma [ lemma5 ] to @xmath899 and obtain the following formula for @xmath900 . @xmath901 since @xmath902 , @xmath903 , which follows the same argument as for lemma [ lf - l2f ] .
in particular , for each @xmath904 , @xmath905 the expectation of the martingale term in the above formula vanishes . for @xmath906 and @xmath907 , let @xmath908 _ step 4 .
_ we recall that @xmath909 . by theorem [ limit - thm ] ,
@xmath910 is @xmath35 .
furthermore by assumption [ assumption1 ] , the @xmath15 diffusion has exponential mixing rate .
we apply corollary [ corollary - to - lemma3 - 2 ] to each function of the form @xmath910 and take @xmath911 there exist a constant @xmath492 and a function @xmath912 such that @xmath913 where denoting @xmath914 , @xmath915 by theorem [ limit - thm ] , @xmath916 belong to @xmath917 .
furthermore @xmath918 and its first two derivatives are bounded by a function in @xmath48 which depends on @xmath34 only through @xmath919 , for some @xmath60 .
thus there are numbers @xmath920 such that for all @xmath889 , @xmath921 , for some @xmath922 . since @xmath923 , @xmath924 , we summing over @xmath889 , @xmath925 for some constant @xmath114 and some function @xmath926 in @xmath48 .
if @xmath116 , it is easy to see that there is a function @xmath927 , not depending on @xmath34 , s.t . @xmath928 .
_ finally , by lemma [ gap ] below , for @xmath929 and @xmath930 , there is a constant @xmath119 and function @xmath931 , depending on @xmath932 s.t . for @xmath933 , @xmath934 for the partition @xmath935
, we assumed that @xmath936 and @xmath937 for @xmath938 .
let @xmath334 . since @xmath939 , estimate ( [ gap-0 ] ) holds also with @xmath34 replaced by @xmath940 , and we have : @xmath941 since @xmath402 are bounded and by theorem [ limit - thm ] @xmath942 is bounded by a function in @xmath48 that does not depend on @xmath889 , for @xmath475 , each term @xmath943 is bounded by a function in @xmath48 and @xmath944 is of order @xmath945 for some function @xmath931 .
we ignore a finite number of terms in the summation . in particular we will not need to worry about the terms with @xmath946 .
since the sum over @xmath889 involves @xmath947 terms the following bound follows from ( [ final - estimate ] ) : @xmath948 here @xmath931 and may depend on @xmath34 .
the case of @xmath116 can be treated similarly .
the estimate is of the form @xmath949 where @xmath950 does not depend on @xmath34 .
we putting together ( [ rate-1 ] ) , ( [ rate-2 ] ) , ( [ rate-3 ] ) and ( [ rate-4])to see that if @xmath111 , @xmath951 where @xmath952 . if @xmath116 , collecting the estimates together , we see that there is a constant @xmath114 s.t .
@xmath953 where @xmath926 is a function in @xmath48 that does not depend on @xmath34 . by induction
the finite dimensional distributions converge and hence the required weak convergence .
the proof is complete .
[ lemma4.2 ] assume that ( [ sde-3 ] ) are complete for all @xmath513 , some @xmath124 . 1 .
@xmath7 is a regularity improving fredholm operator on a compact manifold @xmath11 , @xmath391 .
2 . there exists @xmath954 , constants @xmath955 , s.t .
@xmath956 3 .
there exists a locally bounded @xmath957 such that for all @xmath958 and @xmath268 there are constants @xmath88 and @xmath959 , with the property that @xmath960 4 . for @xmath30 in part ( 2 ) or in part
( 2 ) , @xmath961 for all @xmath958 . for @xmath806 with the property that @xmath962 for all @xmath963
, there exists a number @xmath124 s.t . for every @xmath964 ,
@xmath965 here @xmath966 and depend on @xmath967 only through @xmath968 and @xmath969 .
in particular there exists @xmath970 s.t . for all @xmath964 , @xmath971 furthermore , @xmath972 .
since the hypothesis of theorem [ uniform - estimates ] holds , if @xmath30 is as defined in ( 2 ) , it satisfies ( 2 ) .
since @xmath488 , @xmath973 is finite .
we apply lemma [ ito - tight ] : @xmath974 let @xmath975 since @xmath976 and @xmath977 , @xmath966 .
the required conclusion follows for there conditioned inequality , and hence the estimate for @xmath978 . to estimate @xmath979
, we need to involve the diffusion term in ( [ ito - tight ] ) and hence @xmath980 appears .
[ gap ] assume the conditions of lemma [ lemma4.2 ] and assumption [ assumption1 ] .
let @xmath40 . if @xmath981 is s.t .
@xmath968 , @xmath982 , @xmath983 belong to @xmath48 for all @xmath984 , then for some @xmath46 and all @xmath985 and for all @xmath986 where @xmath481 , @xmath987 where @xmath952 and @xmath114 is a constant .
if the assumptions of theorem [ rate ] holds , the above estimate holds for any @xmath988 ; if @xmath989 , we may take @xmath990 where @xmath931 .
let @xmath126 .
since @xmath991 is the highly oscillating term , we expect that averaging in the oscillation in @xmath992 gains an @xmath2 in the estimation .
we first split the sums : @xmath993 by assumption [ assumption1 ] , @xmath7 has mixing rate @xmath994 .
let @xmath995 , @xmath996 if @xmath997 , @xmath998 .
we apply theorem [ uniform - estimates ] to the functions @xmath999 . for a constant @xmath124 , @xmath1000 where @xmath1001 is a function in @xmath48 , depending on @xmath115 .
thus for @xmath1002 , @xmath1003 let us split the first term on the right hand side of ( [ 7.4 - 1 ] ) .
denoting @xmath1004 , @xmath1005 the first term on the right hand side is estimated by ( [ 7.4 - 3 ] ) .
to the second term we take the supremum norm of @xmath992 and use lemma [ lemma4.2 ] . for some @xmath1006 and @xmath1007
, @xmath1008 then for some number @xmath114 , @xmath1009 where @xmath952 .
let us treat the second term on the right hand side of ( [ 7.4 - 1 ] ) .
let @xmath1010 .
then @xmath1011 to the first term we apply ( [ horder - rate ] ) and obtain a rate @xmath1012 .
we could assume that @xmath992 averages to zero .
subtracting the term @xmath1013 does not change @xmath1014 .
alternatively lemma [ lemma4.2 ] provides an estimate of order @xmath2 for @xmath1015 .
finally , since @xmath1016 , @xmath1017 in the last step we used condition ( 2 ) and @xmath1018 is a function in @xmath48 .
we have proved the first assertion .
if the assumptions of theorem [ rate ] holds , for any @xmath988 , the following functions belong to @xmath48 : @xmath968 , @xmath982 , and @xmath983 . if @xmath989 , the above mentioned functions can be obviously controlled by @xmath1019 multiplied by a function in @xmath48 , thus completing the proof .
let @xmath1020 denotes the collection of borel sets in a @xmath1021 smooth riemannian manifold @xmath10 with the riemannian distance function @xmath419 ; let @xmath1022 be the space of probability measures on @xmath10 .
let @xmath513 where @xmath46 is a positive number .
if @xmath1023 weakly , we may use either the total variation distance or the wasserstein distance , both imply weak convergence , to measure the rate of the convergence of @xmath1024 to @xmath1025 .
let @xmath419 denotes the riemannian distance function .
the wasserstein 1-distance is @xmath1026 here @xmath1027 are projections to the first and the second factors respectively , and the infimum are taken over probability measures on @xmath1028 that couples @xmath1029 and @xmath1025 . if the diameter , @xmath1030 , of @xmath10 is finite , then the wasserstein distance is controlled by the total variation distance , @xmath1031 .
see c. villani @xcite .
let us assume that the manifold has bounded geometry ; i.e. it has positive injectivity radius , @xmath509 , the curvature tensor and the covariant derivatives of the curvature tensor are bounded .
the exponential map from a ball of radius @xmath32 , @xmath1032 , at a point @xmath263 defines a chart , through a fixed orthonormal frame at @xmath263 .
coordinates that consists of the above mentioned exponential charts are said to be canonical . in canonical coordinates ,
all transitions functions have bounded derivatives of all order . that @xmath34 is bounded in @xmath1021
can be formulated as below : for any canonical coordinates and for any integer @xmath889 , @xmath1033 is bounded for any multi - index @xmath201 up to order @xmath889 .
the following types of manifolds have bounded geometry : lie groups , homogeneous spaces with invariant metrics , riemannian covering spaces of compact manifolds . in the lemma below we deduce from the convergence rate of @xmath1024 to @xmath1025 in the @xmath1034 norm a rate in the wasserstein distance .
let @xmath419 be the riemannian distance with reference to which we speak of lipschitz continuity of a real valued function on @xmath10 and the wasserstein distance on @xmath1022 . if @xmath1035 is a random variable we denote by @xmath1036 its probability distribution .
denote by @xmath1037 the lipschitz constant of the function @xmath34 .
let @xmath1038 .
let @xmath1039 .
[ lemma - wasserstein ] let @xmath1040 and @xmath1041 be random variables on a @xmath1021 manifold @xmath10 , where @xmath938 , with bounded geometry .
suppose that for a reference point @xmath1038 , @xmath1042 is finite .
suppose that there exist numbers @xmath1043 $ ] s.t . for @xmath1044
, @xmath1045 then there is a constant @xmath119 , depending only on the geometry of the manifold , s.t .
@xmath1046 if @xmath946 , this is clear .
let us take @xmath334 and let @xmath1047 be a lipschitz continuous function with lipschitz constant @xmath266 .
since we are concerned only with the difference of the values of @xmath34 at two points , @xmath1048 , we first shift @xmath34 so that its value at the reference point is zero . by the lipschitz continuity of @xmath34 , @xmath1049 .
we may also assume that @xmath34 is bounded ; if not we define a family of functions @xmath1050 .
then @xmath1051 is lipschitz continuous with its lipschitz constant bounded by @xmath1037 .
let @xmath1052 .
the correction term @xmath1053 can be easily controlled by the second moment of @xmath1054 : @xmath1055 let @xmath1056 be a function supported in the ball @xmath1057 with @xmath1058 and @xmath1059 , where @xmath365 is a positive number and @xmath202 is the dimension of the manifold . if @xmath82 , @xmath1060 in the last step we used the assumption on @xmath1061 for the @xmath1062 function @xmath1063 . by distributing the derivatives to @xmath1064 we see that the norm of the first @xmath889 derivatives of @xmath1063 are controlled by @xmath1037
. if @xmath34 is bounded , @xmath1065 where @xmath467 is a combinatorial constant . to summarize ,
for all lipschitz continuous @xmath34 with @xmath1066 , @xmath1067 let @xmath1068 .
since @xmath334 , we choose @xmath202 with the property @xmath1069 , then for @xmath34 with @xmath1066 , @xmath1070 let @xmath365 be a positive number with @xmath1071 .
let @xmath1072 denotes the geodesic ball centred at @xmath263 with radius @xmath32 , whose riemannian volume is denoted by @xmath1073 .
there is a countable sequence @xmath1074 in @xmath10 with the following property : ( 1 ) @xmath1075 covers @xmath10 ; ( 2 ) there is a natural number @xmath25 such that any point @xmath290 belongs to at most @xmath25 balls from @xmath1076 ; i.e. the cover @xmath1076 has finite multiplicity .
moreover this number @xmath25 is independent of @xmath365 .
see m. a. shubin @xcite . to see the independence of @xmath25 on @xmath365 ,
let us choose a sequence @xmath1077 in @xmath10 with the property that @xmath1078 covers @xmath10 and @xmath1079 are pairwise disjoint .
since the curvature tensors and their derivatives are bounded , there is a positive number @xmath119 such that @xmath1080 let @xmath289 be a fixed point that belongs to @xmath25 balls of the form @xmath1081 . since @xmath1082 , the sum of the volume satisfies : @xmath1083 and @xmath1084 .
the ratio @xmath1085 depends only on the dimension of the manifold .
let us take a @xmath1021 smooth partition of unity @xmath1086 that is subordinated to @xmath1087 : @xmath1088 , @xmath1089 , @xmath1090 is supported in @xmath1091 , and for any point @xmath263 there are only a finite number of non - zero summands in @xmath1092 .
the partition of unity satisfies the additional property : @xmath1093 , @xmath1094 .
let @xmath1095 be the geodesic charts .
let @xmath1096 and let @xmath1097 denote the representation of a function @xmath380 in a chart .
@xmath1098 it is crucial to note that there are at most @xmath25 non - zero terms in the summation . by the assumption , for each @xmath1099 , @xmath1100 by construction , @xmath1101 is bounded .
there is a constant @xmath1102 that depends only on the partition of unity , such that @xmath1103 similarly for the second summation , we work with the representatives of @xmath1104 , @xmath1105 since we work in the geodesic charts the lipschitz constant of @xmath1106 are comparable to that of @xmath1037 .
let @xmath1066 .
if @xmath34 is bounded , @xmath1107 let @xmath1068 , @xmath1108 on a compact manifold , @xmath1109 can be controlled by @xmath1037 ; otherwise we use the cut off function @xmath1051 in place of @xmath34 and the estimate @xmath1110 .
choose @xmath202 sufficiently large , as before , to see that @xmath1111 .
finally we apply the kantorovich - rubinstein duality theorem , @xmath1112 to obtain the required estimate on the wasserstein 1-distance and concluding the proof . let @xmath1113 ; m)\to m$ ] denote the evaluation map at time @xmath138 : @xmath1114 .
let @xmath1115 denote the probability distribution of a random variable @xmath1035 .
let @xmath416 .
[ proposition - rate ] assume the conditions and notations of theorem [ rate ] .
suppose that @xmath10 has bounded geometry and @xmath120 .
let @xmath23 be the limit measure and @xmath121 .
then for every @xmath122 there exists @xmath123 and @xmath124 s.t . for all @xmath125 and @xmath126 , @xmath1116 by theorem [ rate ] , for @xmath116 , @xmath1117 where @xmath1118 for some function @xmath1119 . since by theorem [ uniform - estimates ]
, there exists @xmath124 such that @xmath1120 is finite , we take @xmath1121 in lemma [ lemma - wasserstein ] to be any number less than @xmath266 to conclude the proposition .
we began with the proof of lemma [ lemma2 ] , follow it with a discussion on conditional inequalities without assuming conditions on the @xmath1122-algebra concerned .
denote @xmath1123 .
firstly , if @xmath1124 and @xmath1125 , @xmath1126 next , by the markov property of @xmath246 and the assumption that @xmath1127 : @xmath1128 from this we see that , @xmath1129 this concludes ( 1 ) .
step 2 . for ( 2 ) , we compute the following : @xmath1130 we estimate the last two terms .
firstly , @xmath1131 it remains to calculate the following : @xmath1132 gathering the estimates together we obtain the bound : @xmath1133 by adding this estimate to that in part ( 1 ) , we conclude part ( 2 ) : @xmath1134 we conclude part ( 2 ) .
we first assume that @xmath1135 , then , @xmath1136 we note that for every @xmath360 , @xmath1137 . in line ( [ estimate-1 - 1 ] )
we replace @xmath1138 , @xmath138 , @xmath1139 by @xmath1140 , @xmath1141 , and @xmath1142 respectively to see the first term on the right hand side is bounded by @xmath1143 next we observe that @xmath1144 this gives the estimate for the case of @xmath306 : @xmath1145 where @xmath1146 if @xmath1147 , we split @xmath1148 and estimate the remaining term .
we use the fact that @xmath1149 , @xmath1150 finally we obtain the required estimate in part ( 3 ) : @xmath1151 thus concluding part ( 3 ) .
the following conditional inequalities are elementary .
we include a proof for a partial conditional burkholder - davis - gundy inequality for completeness .
we do not assume the existence of regular conditional probabilities .
let @xmath1152 be a continuous @xmath408 martingale vanishing at @xmath20 .
let @xmath1153 be an adapted stochastic process with left continuous sample paths and right limits .
if for stopping times @xmath494 , @xmath1154 .
then @xmath1155 let @xmath1156 and @xmath1152 is a right continuous @xmath1157 martingale or a right continuous positive sub - martingale index by an interval @xmath1158 of @xmath1159 .
then , @xmath1160 if @xmath1161 is a right continuous @xmath1157 martingale and @xmath1162 , there exists a constant @xmath1163 s.t .
@xmath1164 mark i. freidlin and alexander d. wentzell . ,
volume 260 of _ grundlehren der mathematischen wissenschaften [ fundamental principles of mathematical sciences]_. springer , heidelberg , third edition , 2012 .
translated from the 1979 russian original by joseph szcs .
nobuyuki ikeda and yoko ochi .
central limit theorems and random currents . in _
stochastic differential systems ( bad honnef , 1985 ) _ , volume 78 of _ lecture notes in control and inform .
_ , pages 195205 .
springer , berlin , 1986 .
g. c. papanicolaou , d. stroock , and s. r. s. varadhan .
martingale approach to some limit theorems . in _
papers from the duke turbulence conference ( duke univ.,1976 ) _ , pages ii+120 pp .
duke univ . ,
durham , n.c . , 1977 .
mark a. pinsky .
homogenization and stochastic parallel displacement . in _
stochastic integrals ( proc .
durham , durham , 1980 ) _ , volume 851 of _ lecture notes in math .
_ , pages 271284 .
springer , berlin , 1981 . | consider a family of random ordinary differential equations on a manifold driven by vector fields of the form @xmath0 where @xmath1 are vector fields , @xmath2 is a positive number , @xmath3 is a @xmath4 diffusion process taking values in possibly a different manifold , @xmath5 are annihilators of @xmath6 . under hrmander type conditions on @xmath7 we prove that , as @xmath8 approaches zero , the stochastic processes @xmath9 converge weakly and in the wasserstein topologies .
we describe this limit and give an upper bound for the rate of the convergence . _
# 1/-0.55ex/_#1 _ # 1 # 1|-0.24em|-0.24em|#1|-0.24em|-0.24em| l .t .
u p ams classification : 60h , 60j , 60f , 60d . |
the _ cosmic background explorer _ ( _ cobe _ ) revealed that the extragalactic background light at infrared ( ir ) wavelengths was comparable to that at optical wavelengths indicating that half of the star formation and active galactic nuclei ( agn ) emission , averaged over cosmic time , is obscured by dust ( e.g. puget et al . 1996 ; hauser & dwek 2001 ) . in the local universe ,
the volume density of dusty , infrared luminous galaxies is insufficient to account for this background .
therefore the density of such galaxies must evolve strongly with redshift ( e.g. see lagache , puget , & dole 2005 for a review ) .
in fact , imaging observations with the _ infrared space observatory _ ( _ iso _ ) and the _ spitzer space telescope _ ( hereafter _ spitzer _ ) have shown that about 70% of the comoving star formation rate density ( sfrd ) at @xmath7 is obscured by dust ( chary & elbaz 2001 ; lefloch et al .
2005 ; prez - gonzlez et al . 2005 ) . in the local universe ,
we refer to these dusty galaxies as luminous infrared galaxies ( lirgs , @xmath8)<12 $ ] ) and ultraluminous infrared galaxies ( ulirgs , @xmath9 ) \ge 12 $ ] ) where @xmath10 is the total ir luminosity from 81000@xmath11 m ( see sanders & mirabel 1996 for a review ) .
it is now becoming clear that ir luminous galaxies are a heterogenous population with selection functions at different mid- and far - infrared wavelengths revealing different subsets of this population ( daddi et al .
2005 ; lutz et al . 2005b
; yan et al . 2005 ; pope et al . 2006 ) .
perhaps the most enigmatic members of the high redshift ulirg population are the submillimeter ( submm ) galaxies ( smgs ) .
these galaxies were initially discovered in the late 1990s with the submillimetre common user bolometer array ( scuba , holland et al .
1999 ) on the james clerk maxwell telescope ( jcmt ) and then later with max - planck millimetre bolometer ( mambo ) camera on the institut de radio astronomie millimetrique ( iram ) 30-m telescope ( see blain et al .
2002 for a review ) .
they have proven to be difficult to study , due to the large beam size of submm observations and the difficulty in finding counterparts at other wavelengths ( e.g. ivison et al .
2002 ; pope et al .
nevertheless , it has been found that they are very massive systems ( swinbank et al .
2004 ; borys et al .
2005 ; greve et al . 2005 ; tacconi et al .
2006 ) at @xmath2 ( chapman et al . 2005 ; pope et al . 2006
; aretxaga et al . 2007 ) which are thought to be connected to the most massive present day galaxies via an evolutionary sequence ( lilly et al .
a crucial aspect that needs to be explored in smgs is what powers their extreme luminosities : agn activity or star formation ?
the mid - ir spectral regime is a particularly good probe of what is producing the total infrared luminosity since it has contributions from both star - formation and agn each with a unique signature at these wavelengths .
the 515@xmath11 m mid - ir spectral energy distribution ( sed ) can be decomposed into 3 main components ( e.g. sajina et al .
2007 ) : ( 1 ) emission features from what are thought to be polycyclic aromatic hydrocarbon ( pah ) molecules ( puget & leger 1989 ; allamandola et al .
1999 ) ; ( 2 ) power - law ( or warm blackbody ) emission ; and ( 3 ) extinction characterized by prominent silicate absorption features ( e.g. draine 2003 ) .
the first is understood to be powered entirely by star formation while the second is predominantly emission from the agn ( genzel et al . 1998 ) .
the second component can also contain contributions from hot dust ( very small grains ) present in the most energetic hii regions ( e.g. tran et al .
2001 ) . the third component can be found in both starbursts and agn since it just requires a dust screen around hot dust emission .
measuring the contribution from ulirgs to the global star formation rate requires dissecting their ir emission into contributions from starbursts and agn .
mid - ir spectroscopy with _
iso _ made it possible to decompose the various contributors in local ulirgs with the result that they are predominantly starburst powered , although they also show signs of agn activity ( e.g. genzel et al .
1998 ; lutz et al . 1999
; rigopoulou et al . 1999 ; sturm et al . 2000 ; tran et al .
in particular , _ iso _ studies of local ulirgs showed that at low luminosities , the ulirgs are powered mainly by a starburst but at the very highest luminosities ( @xmath12 ) , the dominant emission source is the agn ( lutz et al .
1998 ; tran et al . 2001 ) . from this
one might expect that smgs , with their extreme ir luminosities ( @xmath13 , pope et al .
2006 ) , are primarily agn - dominated .
the _ spitzer _ infrared spectrograph ( irs ; houck et al .
2004 ) , with its sensitivity and spectral resolution , can extend this effort outside the local universe , allowing much larger samples to be observed to much fainter flux limits .
the irs can detect emission features and continuum in ulirgs out to @xmath14 ( valiante et al .
the 3 mid - ir spectral components listed above have been fit to the _ spitzer _ irs spectra of ulirgs at various redshifts to determine the agn contribution to the mid - ir luminosity ( e.g. sajina et al .
although x - ray imaging is very effective at identifying the presence of an agn in a galaxy , it is subject to obscuration in extreme , compton - thick sources .
the x - ray luminosity provides a direct measurement of the power of the agn at x - ray energies but it is necessary to make assumptions about the geometry of the obscuring dust to determine the bolometric luminosity of the agn ( e.g. alexander et al .
2005 ) . furthermore
, agn indicators at uv and optical wavelengths can be very obscured in dusty smgs ( swinbank et al .
2004 ; chapman et al .
mid - ir spectroscopy , by directly detecting the thermally re - radiated emission from the obscuring dust , provides one of the most effective ways to identify agn , including compton - thick sources , and measure their contribution to the infrared luminosity ( e.g. lutz et al .
2004 ) .
two previous studies have presented irs spectroscopy of smgs .
menendez - delmestre et al . (
2007 ) presented irs spectra of 5 radio - detected , spectroscopically confirmed smgs .
they found that four smgs with lower redshifts ( @xmath15 ) showed pah emission and a composite spectrum of these four sources fits well to a scaled m82 spectrum plus a power - law component .
they conclude that these systems host both star formation and agn activity .
their fifth source at @xmath16 was found to have a more agn - type spectrum .
valiante et al .
( 2007 ) discussed irs spectra of 13 bright smgs , 9 of which are detected with the irs ( the first two spectra from this sample were presented in lutz et al .
this sample was more representative of the redshift distribution of smgs .
they also found that smgs are likely to be mainly starburst powered .
sajina et al .
( 2007 ) used irs spectra to study a sample of _ spitzer _ 24@xmath11m - selected galaxies at @xmath2 .
this sample was selected to have @xmath17jy and very red 24-to-8@xmath11 m and 24-to-@xmath18 colors ( yan et al .
75% of this sample consists of continuum - dominated sources , with some weak pah emission indicative of star formation activity .
they conclude that these sources display mid - ir spectra similar to those of the most luminous local ulirgs .
follow - up studies of these galaxies have shown that they are not generally ( sub)mm detected sources ( @xmath19mjy for @xmath20 sources , lutz et al .
2005b ) .
in this paper we present _ spitzer _ irs spectra of a sample of smgs in goods - n which have extensive multi - wavelength information from x - ray to radio wavelengths .
this allows us to produce full spectral energy distributions and to calculate the bolometric luminosity .
our _ spitzer _ irs observations go substantially deeper ( @xmath21 ) than previous work .
this enables a study of the typical blank - field submm galaxy population which have @xmath22mjy .
the main goal of our program is to decompose the mid - ir spectra into star formation and agn components , in order to infer how much of the total infrared luminosity is coming from each . from the mid - ir spectroscopy
we are also able to obtain spectroscopic redshifts and measure the luminosities and equivalent widths of individual pah lines to investigate various line diagnostics .
we compare our measurements with line strengths in other star - forming galaxies to put constraints on the evolution of smgs and the relationship between star formation and black hole growth .
this paper is formatted as follows .
our _ spitzer _ observations are described in section 2 and the data analysis is explained in section 3 .
section 4 presents the main results of the irs spectroscopy including redshifts , spectral decomposition and agn classification . in section 5
we fit the full seds of these smgs to determine their total ir luminosities .
we explore correlations between the pah and ir luminosities in section 6 .
we discuss our results in the context of the role that smgs play in galaxy evolution in section 7 .
finally , our conclusions are listed in section 8 .
we include an appendix which discusses the noise properties of our _ spitzer _ irs spectra and includes notes on individual sources .
all magnitudes in this paper use the ab system .
we use a standard cosmology with @xmath23 , @xmath24 and @xmath25 .
the _ spitzer _ irs observations presented here were taken as part of _ spitzer _ program go-20456 ( pi : r. chary ) .
the full sample consists of high redshift smgs , agn and optically faint 24@xmath11 m sources in the great observatories origins deep survey ( goods , dickinson et al .
2003 ) north field .
the goods - n field is one of the most extensively studied regions of the sky , with deep data existing across all wavebands including : _ chandra _ 2@xmath26msec x - ray observations ( alexander et al .
2003 ) ; deep _ hst _ optical imaging using the advanced camera for surveys ( acs ) in four bands , @xmath27 , @xmath28 , @xmath29 and @xmath30 ( giavalisco et al .
2004 ) ; and deep _ spitzer _ imaging at five infrared wavelengths ( dickinson et al .
, in preparation ) .
more recently , goods - n has been surveyed with irs peak - up observations at 16@xmath11 m ( teplitz et al .
2005 ) and mips imaging at 70@xmath11 m ( frayer et al .
in addition to the extensive space - based imaging campaign , many ground - based programs have also targeted this field for imaging and spectroscopy making it one of the best field for investigating the role of agn and star formation in high redshift galaxies . here
we present the irs spectroscopy for the smgs from this project .
results from the other sources will be presented in other papers .
our sample of smgs was primarily selected from sources detected in the goods - n scuba ` super - map ' with secure multi - wavelength counterparts ( borys et al .
2003 ; pope et al . 2006 ; pope 2007 ) .
we imposed a 24@xmath11 m flux cut of @xmath31jy in order to ensure that we achieve adequate signal to noise ( snr ) on the continuum in the irs spectra .
roughly half of the secure counterparts in pope et al .
( 2006 ) meet this 24@xmath11 m flux cut which yields a sample of 10 irs targets ( after removing sources which are already observed with the irs in other programs ) . to expand our sample slightly
, we also included three scuba sources from chapman et al .
these three sources are not detected at @xmath32 in the ` super - map ' but they are detected in scuba photometry observations with a lower statistical significance .
our 13 smg irs targets have 850@xmath11 m and 24@xmath11 m flux densities of 210@xmath26mjy and 2001200@xmath11jy , respectively .
this sample is expected to be representative of the submm population discovered in existing blank - field 850@xmath11 m imaging .
there is no evidence supporting the idea that smgs which are fainter than 200@xmath11jy at 24@xmath11 m are fundamentally different besides lying at slightly higher redshifts ( e.g. pope et al .
we list the positions and mid - ir fluxes for these 13 smgs in table [ tab : targets ] .
smg ids starting with gn and c are from the scuba super - map ( pope et al . 2006 ; pope 2007 ) and chapman et al .
( 2005 ) , respectively , and the 850@xmath11 m fluxes can be found in these papers .
the majority of the smgs are undetected in the deep _ spitzer _ @xmath33 m images ( see also huynh et al .
11/13 smgs have optical spectroscopic redshifts while the other two have photometric redshifts .
the redshift distribution of our sample is consistent with that of all smgs ( chapman et al .
2005 ; pope et al .
[ fig : post ] shows multi - wavelength images , which span roughly the width of the irs slit , of our targets . in some cases there
are several optical or irac sources in the aperture . for gn04 , gn07 and gn19 ,
the two irac sources are both expected to be associated with the submm emission since both components are at the same redshift ( see pope et al .
the irs spectra of these 3 smgs will have contributions from both components .
for the remaining sources , the 8 , 16 and 24@xmath11 m images show that only our primary targets are producing the mid - ir emission which we will detect with the irs .
+ fig . 1 .
continued .
the _ spitzer _ irs observations are taken in spectral staring mode which observed the target at two nod positions within the slit . we observed in low resolution ( @xmath34100 ) using the short - low 1 ( sl1 ; 7.414.5@xmath11 m ) , long - low 2 ( ll2 ; 14.021.3@xmath11 m ) and/or long low 1 ( ll1 ; 19.538.0@xmath11 m ) orders , depending on the redshift of the source .
we chose the wavelength coverage in order to include one or more of the main pah features at 6.2 , 7.7 and 11.3@xmath11 m .
integration times varied depending on the flux of the target at 24@xmath11 m .
we used blue peak - up acquisition on nearby , isolated , bright 2mass stars .
our observations were taken in april and may 2006 .
table [ tab : obs ] shows our list of targets and integration times .
we started with the s14.0.0 @xmath35 irs pipeline data[multiblock footnote omitted ] which produces 2-dimensional ( 2d ) basic calibrated data ( bcd ) files .
note that our targets are not resolved with _
spitzer _ at these wavelengths and therefore they can be treated as point sources . at mid - ir wavelengths , the dominant source of noise is the sky background , mostly from zodiacal light .
since we have long integrations of very faint targets , additional data reduction steps are required to alleviate the detector and sky noise .
the first step is to identify and clean the rogue pixels .
there are a number of hot pixels in the irs arrays ( @xmath36 ) .
rogue pixel masks were created from the campaign masks and from examining each 2d spectral image individually .
we used irsclean[multiblock footnote omitted ] to replace the values of the hot pixels by extrapolating from the surrounding pixels .
note that irsclean fails when there is a group of several hot pixels together , so we made sure to exclude these from the spectral extraction .
the fraction of pixels in the 2d files which remain unusable after irsclean is only @xmath37 .
the next step is to remove the latent charge build - up on the array . in long irs integrations
, it has been found that latent charge builds up with time , despite the detectors being reset at the end of each integration ( see teplitz et al .
this build - up is different depending on the wavelength ( @xmath38 position on the 2d spectral image)[multiblock footnote omitted ] .
we found charge build - up on the arrays for integrations of more than 1 hour in all ll observations .
for the sl1 observations , we only saw the latent charge build - up for a few targets , even though all of these observations were greater than 1 hour .
the latent charge build - up was fit by a polynomial and the temporal component subtracted off .
we explore several methods for removing the sky background including : a ) subtracting the average spectrum at each nod position ; b ) subtracting individual observations from opposite nod positions and then averaging ; c ) subtracting a ` supersky ' created from all files from that astronomical observation request ( aor ) ; and d ) subtracting a supersky created from all our data taken in a given campaign . in all methods , we mask out all sources which fall within the irs slit identified from the deep 24@xmath1 m imaging of the goods
- n field .
the supersky subtraction was carried out with and without normalizing to the median in each image .
the third method of subtracting a normalized supersky created from all observations from that aor gave the lowest residual sky noise ( and thus highest snr ) in the final spectrum therefore this is what we used to remove the sky from all our observations . after the sky was removed from the individual data files , we then coadded them using a clipped median .
the coadds were checked again for persisting hot pixels and cleaned once more using irsclean .
these steps produced two 2d spectra for each target , one for each nod position .
we used the spitzer irs custom extraction ( spice[multiblock footnote omitted ] ) to extract the 1d spectra at each nod position .
we used a narrow extraction window of 2 pixels ( @xmath3910 arcseconds for ll observations ) which was constant as a function of wavelength .
we found this produced higher snr spectra than using an extraction window of increasing width , starting with 2 pixels at the blue end . using the optimal extraction option in spice
, we found comparable results to the 2 pixel , fixed width , extraction .
for each science spectrum , we also extracted a residual sky spectrum , offset from any sources , to provide a level of the error in our final spectrum as a function of wavelength .
the uncertainty is calculated as the standard deviation of 10 pixels surrounding each wavelength in the residual sky spectrum . for each source , the 1d spectra at each nod position
were then averaged together to produce the final spectrum for each source .
further details about the final noise in our irs spectra are given in appendix [ app : rms ] . in order to calibrate the final irs spectra , we reduced standard star calibrator observations from the same campaign as our observations , in exactly the same way that we reduced the science observations .
we then extracted the spectrum for the calibrator using our 2 pixel width and the default spice width in order to determine the calibration factor as a function of wavelength .
this calibration was verified against the broad - band photometry at 8 , 16 and 24@xmath11 m and we found very good agreement ( see fig .
[ fig : cal ] ) . in order to measure the individual pah line
luminosities we used the idea ( isap ) environment in the smart software package ( higdon et al .
2004 ) . for each of the 4 main pah lines ( 6.2 , 7.7 , 8.6 and 11.3@xmath11 m ) , we isolated a region around each line and simultaneously fit the line and continuum .
both the center and the width of the line were allowed to vary .
all measured line widths were consistent within the errors with the line widths measured in irs spectra of local star forming galaxies ( smith et al .
note that the errors we calculated from the residual sky were propagated in the fitting of individual lines .
all smg line measurements used in this paper including 3@xmath40 upper limits when the line is undetected in the irs spectrum are listed in table [ tab : lines ] .
the biggest uncertainty in comparing pah line luminosities from different samples comes from differences in how the spectral lines are measured , more specifically , how the continuum is estimated for each line .
this can lead to variations in the line flux and equivalent width of up to a factor of 4 ( see sajina et al . 2007
; smith et al .
our method of measuring the line strengths gives similar results to the cubic spline continuum fits performed in brandl et al .
( 2006 , see next paragraph ) .
this method is known to underestimate the flux of the pah lines ( particularly the 7.7@xmath11 m line ) compared to methods which fit all pah lines simultaneously ( e.g. sajina et al .
2007 ; smith et al . 2007 ) since the continuum we adopt is higher .
therefore , we took care to measure the pah lines in any comparison samples in the exact same way as we did for the smgs . in this paper , we compare the smg pah line luminosities to those of other galaxy populations .
brandl et al .
( 2006 ) presented irs spectra of 22 local starburst galaxies with infrared luminosities ranging from @xmath41@xmath42 .
these galaxies are known to have little or no agn contributing to their infrared luminosity .
because the choice for estimating the continuum under the pah lines makes a significant difference to the results , we have measured the lines for the starburst galaxies in exactly the same way as we did for the smgs .
another interesting comparison sample is the local ulirg galaxies .
armus et al .
( 2007 ) presented irs spectroscopy of 10 ulirgs from the bright galaxy sample ( bgs , @xmath43mjy , soifer et al .
1987 ) and the irs spectra for a sample of over 100 ulirgs was recently presented in desai et al .
( 2007 ) and farrah et al .
the local ulirgs show a broad range of mid - ir properties and a mix of agn and starburst mid - ir emission .
however , we did not have access to the irs spectra for the full sample of local ulirgs in order to measure the line luminosities using our method therefore we could not include them in this paper .
we measured the pah lines in 10 of the brandl et al .
( 2006 ) starburst galaxies to verify that our method is consistent with their measurements . on average the luminosities of the 6.2 and 11.3@xmath11 m lines
we measure are consistent with the values listed in brandl et al .
( 2006 ) , but the 7.7@xmath11 m line luminosity is 1.5 times larger on average from our measurements .
this is probably because we are using a straight line for the continuum as opposed to a spline .
we have corrected the measurements of the 7.7@xmath11 m line luminosity for the brandl et al .
( 2006 ) starbursts for this factor of 1.5 for all plots in this paper .
for the brandl et al .
( 2006 ) starbursts we also applied a correction to the line measurements to account for the fractional flux found within the irs apertures ( see table 2 of brandl et al .
2006 ) when comparing to the unresolved smgs .
the error on the line fluxes for the brandl et al .
( 2006 ) starbursts is dominated by the calibration error which is on the order of 10% ( less than the size of the symbols in most of the plots in this paper ) .
in this section we use the mid - ir spectra to explore the nature of smgs .
we start by using the identification of pah lines to estimate redshifts .
we then proceed to decompose the mid - ir spectra into starburst and agn components ; starting first with a composite smg spectrum since the snr is higher and then proceeding to decompose each individual smg spectrum .
we then use the information from the spectral decomposition to classify the fractional contribution of agn to the mid - ir luminosity and compare this with other agn indicators such as optical spectroscopy and x - ray emission .
these results are then used to guide the fitting of the full infrared sed to constrain the total infrared luminosity in the section 5 .
[ fig : spec ] shows the rest - frame ( unsmoothed ) irs spectra for the 13 smgs .
pah emission lines are present in all sources .
despite the relatively low snr in the spectra , all smgs seem to show very similar mid - ir spectral shapes with the exception of c1 .
while this smg clearly shows pah emission , it is superimposed on a steeply rising continuum .
we will discuss this source further in section [ ap : sources ] .
continued .
continued .
given the presence of pah features in all of our sources we are able to extract redshifts from the irs spectra . in order to calculate the redshift and its associated error from the irs data we follow the following steps .
we assumed the following rest - frame wavelengths for the main pah features ; 6.22 , 7.71 , 8.61 and 11.27@xmath11 m ( draine & li 2007 ) . for each pah line , we find the center of the gaussian fit to the line from smart and calculate the redshift .
the redshift for each source is calculated by averaging the redshift estimates from each pah line .
the total error in the redshift is the quadratic sum of the deviation due to the spread in the redshift from the different pah lines and the centroid uncertainty . for sources
whose spectra only contain 1 pah line , the error is just the centroid uncertainty of that line .
the deviation in the redshift from each line is calculated as @xmath44 .
we verified the centroid uncertainties by performing monte carlo simulations .
we take a gaussian of the same height and width as each pah feature and we run 1000 trials where we add noise which has the same standard deviation as the noise in the real data . for each trial , we fit the data to get the line center and keep track of the difference between the input line center and the measured line center .
the centroid error is estimated from the standard deviation of this difference for all 1000 trials .
we repeat this for each source . with the exception of gn26
whose spectrum has a very high snr , the centroid uncertainties for individual pah features range from 0.080.25@xmath11 m which corresponds to 0.51.5 pixels .
this is roughly an order of magnitude higher than what we would expect from the instrument based on the pixel size in the case of no noise . the total error in the irs redshift ranges from 0.010.05 for these smgs .
table [ tab : redshift ] lists the new irs redshifts and the previous optical , spectroscopic or photometric , redshifts .
the new irs mid - ir redshifts are compared with the optical redshifts in fig .
[ fig : redshift ] .
two of our 13 sources seem to have very different redshifts as determined from irs and optical spectroscopy ( c2 and gn06 ) which we discuss in appendix [ ap : sources ] ) .
since the mid - ir and submm wavelengths both trace dust emission , we believe that the irs redshifts presented here are correct for the galaxies which are causing the submm emission . we can study the ensemble properties of the smg sample by averaging together the individual mid - ir spectra .
when plotted in rest - frame luminosity units , the spectra for all smgs in this sample show very little dynamic range .
therefore we are able to average the spectra without having to normalize each at a specific wavelength which can introduce additional biases in the composite spectrum .
we exclude c1 from the composite spectrum , since it shows a very different spectral shape from all the other smgs ( see section [ ap : sources ] ) .
the error on the composite spectrum is calculated from combining the errors on the individual flux measurements .
we restrict the final smg composite spectrum to areas where there are @xmath45 data files present at that wavelength which results in a wavelength coverage of @xmath39512@xmath11 m in the rest frame .
[ fig : comp ] shows our composite spectrum of 12 smgs calculated from averaging the individual spectra ( excluding c1 ) .
we detect the 6.2 , 7.7 and 11.3@xmath11 m emission lines , with a hint of the 8.6@xmath11 m emission line in the composite spectrum .
the composite spectrum also shows a very flat underlying continuum . in order to quantify the fractional contribution of star - formation and agn activity to the emission at these wavelengths ,
we fit our composite spectrum with a model containing the 3 components discussed in section [ sec : intro ] .
the pah emission is fit using two different templates : 1 ) the mid - ir spectrum of m82 ( frster schreiber et al .
2003 ) and , 2 ) the starburst composite template from brandl et al .
m82 is a prototypical star forming galaxy and its mid - ir spectrum is expected to be dominated by emission from star - forming regions .
the starburst template from brandl et al . ( 2006 ) is a composite of 13 local starburst galaxies with bright flux densities and without a strong agn component . as we show below , these templates
are composed almost entirely of the sum of the individual pah features .
the agn emission is characterized by a power law with both the normalization and slope as free parameters .
for the extinction , we obtain @xmath46 from the draine ( 2003 ) extinction curves and applied it to both the pah and continuum components separately .
the extinction curve is not just monotonic in wavelength and contains silicate absorption features , the most notable being at 9.7@xmath11 m .
our model , @xmath47 , can be expressed as @xmath48 where @xmath49 is the pah template .
we performed a @xmath50 minimization fit for each of the two pah templates and solved simultaneously for @xmath51 , @xmath52 , @xmath53 , @xmath54 and @xmath55 .
the best fit model , @xmath47 , is shown in the left panel of fig .
[ fig : comp ] as the red dashed curve .
the blue dash - dot and green dotted curves show the individual contributions from the continuum and pah components , respectively .
the best - fit pah template was the starburst composite from brandl et al .
( 2006 ) although m82 gave very similar results . with this pah template ,
the extinction needed is @xmath56 .
the slope of the power law is very shallow ( @xmath57 ) .
the continuum component accounts for 30% of the luminosity at these wavelengths ( 511.5@xmath11 m ) .
this implies that the mid - ir luminosity of smgs is dominated by star formation with a maximum mid - ir agn contribution of 30% .
to be clear about the origin of the emission in the pah template , we also fit the smg composite spectrum to individual pah line profiles and a continuum with extinction .
we include the pah lines at 5.70 , 6.22 , 7.42 , 7.60 , 7.85 , 8.33 , 8.61 , 11.23 , 11.33@xmath11 m , each modeled as a drude profile ( e.g. draine & li 2007 ) where the height of each line is treated as a free parameter . the resulting best - fit model shown in the right panel of fig .
[ fig : comp ] is very similar to the best - fit model which includes a pah template and continuum ( left panel ) . in fig .
[ fig : comp2 ] , we show the smg irs composite spectrum compared to the irs spectra of @xmath2 _ spitzer _ 24@xmath1m - selected ulirgs from sajina et al .
( 2007 ) and the local ulirg arp220 .
while arp220 is often considered the typical local analogue to high redshift smgs , it appears to have more silicate absorption and less 6.2 and 11.3@xmath11 m pah emission than the average smg .
recall that the sample in sajina et al .
( 2007 ) consists of red 24@xmath11m - selected galaxies with @xmath58mjy , which is much brighter than most of the smg population . the majority of the ulirgs in the _
spitzer_-selected sample have weak pah emission and are agn - dominated ; the green and red curves in fig .
[ fig : comp2 ] are composite spectra of the weak - pah ( i.e. agn - dominated ) _ spitzer _ 24@xmath1m - selected ulirgs with @xmath59 and @xmath60 , respectively .
the blue curve is the composite spectrum for the strong - pah _ spitzer _ 24@xmath1m - selected ulirgs and is very similar to the smg composite .
in particular the relative line strengths are consistent .
the strong - pah _ spitzer_-selected sources represent only a small fraction of the sajina et al .
( 2007 ) sample and these sources tend to have a higher detection rate in the ( sub)mm than the rest of their sample ( sajina et al . in preparation ) . from yan
et al . ( 2007 ) and sajina et al . ( 2007 ) , the surface density of the @xmath2 _ spitzer _ 24@xmath1m - selected ulirgs is 10.4 deg@xmath61 and the strong - pah sub - sample of these sources has a surface density of 2.6 deg@xmath61 .
the surface density of smgs with @xmath62mjy ( i.e. @xmath63 , pope et al .
2006 ) is 844 deg@xmath61 ( coppin et al .
while most submm - selected galaxies are much fainter at @xmath64 m , the strong - pah _ spitzer _ 24@xmath1m - selected ulirgs appear to be a very small sub - sample of the smg population .
given the good agreement between our smg composite spectrum and the spectrum of starburst galaxies , we proceed to fit the individual smg mid - ir spectra to the model given in equation [ equ : model ] . for
each of the 13 smgs we solve for the five parameters in equation [ equ : model ] and plot the best - fit model in fig .
[ fig : spec ] as the red dashed curve .
the separate starburst and agn components including extinction are shown as the green and blue curves .
for most smgs , the continuum component is zero and not plotted in fig .
[ fig : spec ] .
we note that the spectrum of gn26 is even lower than the model at long wavelengths which confirms that the agn component must be negligible in this source . from our model
fits , we derived the fraction of the mid - ir luminosity which comes from the continuum component .
the continuum component provides an upper limit to the agn contribution since hot dust present in the most energetic hii regions can also contribute to the continuum .
when the best - fit model does not include a continuum component , we calculate an upper limit by conservatively assuming that all of the emission around @xmath65 m is coming from a continuum component .
we used this fraction to classify the mid - ir spectrum and the results are listed in table [ tab : agn ] .
@xmath66 smgs are clearly starburst - dominated sources ( continuum component @xmath67 of the mid - ir luminosity ) and only @xmath0 smgs are agn - dominated ( continuum component @xmath68 of the mid - ir luminosity ) in the mid - ir .
the remaining sources have a combination of pah and continuum emission in the mid - ir .
the results for the individual galaxies are in agreement with the composite spectrum which showed a 30% agn contribution on average .
alexander et al .
( 2005 ) showed that deep x - ray observations are a powerful tool for identifying the presence of agn in smgs .
they conclude that the majority of smgs host an agn , but that the bolometric output from smgs is dominated by star formation because of the low ratio of x - ray luminosity to ir luminosity .
they also stress that some fraction of smgs will be agn - dominated .
following the same method as alexander et al .
( 2005 ) we list the x - ray classification in table [ tab : agn ] where a hard x - ray detection tells us that the x - ray emission is dominated by agn activity .
we find that the x - ray classification gives a higher fraction of agn sources ( @xmath69 ) than the mid - ir spectra ( @xmath0 ) for smgs with @xmath31jy . for the x - ray classified agn sources , the x - ray emission expected from star formation (
e.g. bauer et al . 2002 ) is less than that observed indicating that there is an agn present .
our irs observations show that this agn is not important to the mid - ir luminosity .
on the other hand , the line widths and line ratios in the uv / optical spectra more often indicate that these smgs are starburst sources ( swinbank et al .
2004 ; chapman et al .
2005 ) and appear to be insensitive to the agn activity in many sources .
c1 is classified as a starburst from the x - ray data but shows a very strong agn in the mid - ir .
this smg is likely to host a very obscured agn , the kind of system that can help explain the unresolved portion of the hard x - ray background ( e.g. brandt & hasinger 2005 ; daddi et al .
see appendix [ ap : sources ] for more discussion about c1 . combining the x - ray imaging with mid - ir spectroscopy provides the best census of agn in smgs . in pope et al .
( 2006 ) , we found that most smgs showed evidence for a stellar bump in the 4 irac channels .
we found only 4 sources which had a good fit to a simple power - law from irac+24@xmath11 m photometry .
only one of these is in our irs sample : gn04 , since the others are too faint for spectroscopy at 24@xmath11 m .
we find that the irac sed , the mid - ir spectrum and x - ray classification all converge on the conclusion that there is an energetically important agn in gn04 .
in contrast , the uv / optical spectra of this smg did not reveal the presence of an agn in it .
another approach for separating sources which contain a mid - ir agn component is through the use of _ spitzer _ color - color diagrams .
ivison et al . (
2004 ) proposed that the @xmath70 vs @xmath71 color plane could be used to identify agn - dominated smgs .
the irs spectra have allowed us to decompose the starburst and agn components in smgs therefore we are in a position to test the plot as a diagnostic . in fig .
[ fig : cc ] , we plot this _
spitzer _ color - color diagram .
the orange dashed and blue dotted curves show the positions of mrk231 ( mid - ir spectrum from rigopoulou et al .
( 1999 ) spliced with a fit to the _ infrared astronomical satellite _ , _ iras _ , photometry ) and m82 ( frster schreiber et al .
2003 ) , respectively , as a function of redshift ( redshift is indicated by the numbers along the track ) . based on these broad - band colors ,
agn - dominated sources are expected to lie along the mrk231 tracks while starburst - dominated systems should have similar colors to m82 .
sajina , lacy & scott ( 2005 ) showed that while agn - dominated sources do lie on this agn sequence , there are also just as many high redshift pah - dominated sources with high extinction in this region .
the large red dots are the smgs from this study for which we have irs spectra .
interestingly , both smgs ( c1 and gn04 ) which have a @xmath72 contribution from continuum ( agn ) emission to the mid - ir luminosity based on the irs spectra ( see table [ tab : agn ] ) lie off the starburst sequence . based on our irs results , smg which lie in the light shaded region of fig . [
fig : cc ] are starburst - dominated in the mid - ir .
smgs outside the shaded region likely have a combination of starburst and agn emission in the mid - ir or lie at the highest redshifts ( @xmath73 ) .
this division between starburst and agn - dominated sources based on irs spectra agrees with the broad - band colors .
the crosses indicate sources which are classified as x - ray agn ( alexander et al .
many of these lie along the starburst sequence and have irs spectra which are starburst - dominated .
while the x - ray emission may be telling us that an agn is present in the system , it does not seem to be a very good diagnostic of whether the agn contributes to the bolometric luminosity .
sources which have agn - dominated mid - ir spectra , like c1 , clearly show very extreme colors in this diagram however there are not enough of these sources to characterize their allowable positions in this color - color space .
the smaller dots are the rest of the smgs from the goods - n submm super - map from pope et al . ( 2006 ) .
extrapolating our irs results , the fraction of all smgs which are starburst - dominated in the mid - ir is @xmath74 .
the small open triangles are the @xmath75 bright 24@xmath1 m ulirgs from sajina et al .
( 2007 ) ; very few of these sources lie within the starburst region defined for smgs indicating that they have a wider range of mid - ir spectra .
in pope et al . ( 2006 ) , we fit the 24@xmath11 m , 850@xmath11 m and radio photometry to a suite of models to determine the total infrared luminosity and dust temperatures .
we had used the chary & elbaz ( 2001 , hereafter ce01 ) templates and modified them by adding additional extinction from the draine ( 2003 ) models .
the ce01 templates were developed to be representative of local galaxies therefore they contain an inherent luminosity - temperature ( l - t ) relation .
when fitting , we had allowed the models to scale so that we could solve for both the luminosity and the temperature independently ( i.e. a model with high luminosity but cool average dust temperature is allowed ) .
we had found that smgs fit best to scaled - up versions of lower luminosity templates which had cooler average dust temperatures than local ulirgs . in those fits
the short wavelength part of the spectrum was constrained only by the 24@xmath11 m flux . from the present study it is apparent that the 24@xmath11 m flux of smgs will vary considerably with redshift , due to the presence of the pah lines .
therefore , we have repeated the full sed fitting now that we have the mid - ir spectra .
[ fig : c2 ] demonstrates why we need to allow the ce01 templates to scale ( i.e. luminosity and temperature are both variables ) and the need for an additional extinction term . here
we plot the mid - ir to radio sed for one of our smg sources , c2 .
first we simply plot the ce01 template which best matches the submm flux only as the green dotted curve and the ce01 template which best matches the 16 and 24@xmath11 m fluxes as the dot - dash dark blue curve .
for these two models we emphasize that the ce01 templates have not been allowed to scale therefore they follow the local luminosity - temperature relation . because smgs are so luminous at mid - ir and submm wavelengths , the green and dark blue models have a warm average dust temperature ( @xmath76k ) .
the light blue dash - dot - dot curve is the best fit to all the photometry between rest - frame @xmath77 m where we have allowed the ce01 templates to scale .
while this provides a good fit to the mid - ir spectrum and the submm flux , the radio flux and @xmath10 are underestimated compared to our best fit .
we know that the ce01 templates need to be scaled up to match the photometry of smgs however as we scale them up we are introducing a greater dust mass which must then cause more extinction .
the red dashed curve is a scaled ce01 template with additional extinction from the draine ( 2003 ) extinction curve which provides the best - fit to our data .
we require both a scaling of local templates ( in this case ce01 ) and additional extinction to fit the sed of high redshift smgs .
we thus fit the irs spectrum , the 70@xmath11 m , submm and radio photometry to the ce01 templates plus draine ( 2003 ) extinction for each of the 13 smgs in this paperm flux ) . ] .
we solve for the three free parameters ( luminosity , average dust temperature and extinction ) and find the template with the lowest @xmath50 .
[ fig : sed ] shows the full sed fits . in general , the resulting fits are very good . the dark curve and dots are the points used in the fit , while the open symbols show the other photometry points ( mostly 16 and 24@xmath11 m ) which were not used in the fits but agree well with the best - fit sed .
we required a scaling of 10100 in luminosity to fit the local ce01 templates to the smg photometry .
since high redshift smgs can be well fit by modified ce01 templates which contain an inherent radio - ir correlation , this suggests that smgs also follow a similar radio - ir correlation to local galaxies ( see also kovacs et al .
2006 ) .
we integrated the full seds in fig .
[ fig : sed ] from 81000@xmath11 m to get the total infrared luminosity . values of @xmath10 are listed in table [ tab : lines ] and they are in good agreement with the values from pope et al .
( 2006 ) .
+ fig . 9 . continued .
+ fig . 9 . continued . by simply scaling to the wavelength where the sed is a maximum as an estimate of the average dust temperature ,
we obtain a median of 32@xmath26k for these 13 smgs .
this is consistent with previous estimates of dust temperatures in smgs ( chapman et al .
2005 ; kovacs et al . 2006 ; pope et al . 2006 ) .
we also found a similar trend between luminosity and temperature as found in chapman et al .
( 2005 ) which is offset to cooler temperatures from the local correlation .
knowing the shape of the far - ir sed and the total luminosity , we can also estimate the dust mass following the relation given in dunne et al .
( 2000 ) , the median dust mass for this sample is @xmath78m@xmath79 .
this is roughly an order of magnitude higher than dust masses observed in galaxies locally although the local galaxies studied are generally much less luminous . in order to quantify the effect of the extinction term on the models , we calculate @xmath10 without the extinction .
the median amount that the extinction term changes @xmath10 by is 13% ( 020% ) , and so models which do not include the extinction term will incorrectly determine @xmath10 by around 13% .
we find an average value of @xmath56 and we do nt see any correlations between the extinction and @xmath10 for the smgs given the error bars . for sources which we found to contain an agn component in the mid - ir
, we calculate the contribution to @xmath3 from the agn . using _ iso _
spectra , tran et al .
( 2001 ) found that the ratio of @xmath80/@xmath10 was about 2.5 times greater for agn than for starbursts i.e. as you extrapolate from mid - ir to far - ir wavelengths agn become less important bolometrically than starbursts . in order to quantify this for our sample , we scale the mrk231 ( an agn whose @xmath10 is known to be dominated by agn emission and not star formation , armus et al . 2007 )
template to the agn fraction ( listed in table [ tab : agn ] ) of the mid - ir luminosity and then integrate the sed from 81000@xmath11 m to get the contribution to @xmath3 from the agn component . for sources which have an upper limit to the agn contribution , we plot the scaled mrk231 in fig . [
fig : sed ] at this limit .
the values of @xmath81 from integrating under the curves and @xmath82 are given in table [ tab : lines ] .
as seen in fig .
[ fig : sed ] , only c1 has enough agn contribution in the mid - ir to make a significant contribution to @xmath10 .
therefore the intense ir luminosities seen in smgs are powered almost entirely by star formation . in section [ sec : dis ] , we discuss how this luminosity can be converted to give us the star formation rate ( sfr ) .
pah line profiles have broad wings which contribute a significant fraction of the line flux .
pah line fluxes therefore , need to be consistently measured in a uniform way before comparisons between local sources and high redshift sources can be performed .
we have measured line fluxes of all sources consistently , including the comparison sample of low redshift starbursts ( brandl et al .
2006 ) , using the technique described in section [ ap : meas ] .
the pah line luminosities and equivalent widths of the smgs including @xmath83 upper limits when the line is undetected are listed in table [ tab : lines ] .
the 6.2@xmath11 m line is the cleanest to measure and use for diagnostics since it is the least affected by silicate extinction and line blending .
we can see that @xmath5 for the smgs varies by over an order of magnitude , while @xmath4 and @xmath6 have much smaller dynamic ranges of a factor of a few . while this might seem odd
, we note that only a subset ( 4060% ) of our sample have 6.2 and 11.3@xmath11 m spectral coverage whereas 92% have coverage of the 7.7@xmath11 m emission line .
furthermore , the redshift range for the sources with 6.2 and 11.3@xmath11 m spectral coverage is very small and therefore we expect a small dynamic range in pah luminosity .
the sources which are dominated by an agn in the mid - ir , as classified by our irs spectra ( c1 and gn04 ) , both show suppressed 6.2@xmath11 m emission relative to their 7.7@xmath11 m emission .
[ fig : lpah ] shows the relative strengths of pah lines compared to local starbursts ( brandl et al .
the top panels show @xmath5 as a function of @xmath4 ( left ) and @xmath6 ( right ) .
the smaller bottom panels show the residual scatter after removing the fit shown in the upper panels ; the two larger colored error bars quantify the @xmath84 scatter seen in the starbursts and smgs .
the outlying source in the left panel of fig .
[ fig : lpah ] is c1 , the smg with the largest agn contribution .
while the smgs do not show enough dynamic range in pah luminosity to pick out any strong correlations within the population , they do fall on the relation between various pah lines established for local starburst galaxies .
the dashed lines in fig .
[ fig : lpah ] are the best - fits to the local starbursts only ( blue squares ) .
we find essentially the same relation if we fit the local starbursts and the smgs , indicating that the pah luminosities in smgs are simply scaled up ( albeit by several orders of magnitude ) from less luminous starburst galaxies .
the best - fit relations from fitting both the smgs and the local starburst galaxies are @xmath85 the smg c1 appears to lie off this relation with an excess of 7.7@xmath11 m pah emission for a given @xmath4 or @xmath6 .
rigopoulou et al .
( 1999 ) found suppressed 6.2@xmath11 m emission relative to 7.7@xmath1 m emission in iso spectra of local ulirgs compared to starbursts and they conclude that this is primarily because of extinction .
this suggests that , on average , the smg c1 considered here may contain greater extinction or a more dominant agn component than most smgs and starburst galaxies since an agn continuum will dilute the strength of the weaker 6.2 and 11.3@xmath11 m features relative to the strong 7.7@xmath11 m line .
[ fig : ew ] shows the equivalent widths of the 6.2 and 7.7@xmath11 m pah lines . the vertical dashed line at @xmath86 is often used to separate starburst - dominated and agn - dominated sources ( armus et al . 2007 ;
sajina et al .
2007 ) . with the exception of c1 , all of our sources with measurable 6.2@xmath11 m pah emission have @xmath87 , which indicates that they are starburst - dominated .
we have again plotted the local starbursts for comparison .
we found that the values in brandl et al .
( 2006 ) for ew@xmath88 were on average 1.6 times lower than our measurements of the starburst ew@xmath88 , due to differences in the way the lines were measured .
therefore we have revised all the brandl et al .
( 2006 ) points upwards in this plot .
the starbursts cover a small area of this plot , while the smgs are spread over a larger region , although with much larger error bars .
our current data do not have sufficient snr to determine if there is any genuine spread or correlation among the ews , however our error bars are sufficiently small to exclude equivalent widths of less than 0.2 in either pah line ( with the exception of c1 ) .
our smgs show similar equivalent widths to the pah - dominated ulirgs from the _ spitzer _
24@xmath1 m selected sample in sajina et al .
( 2007 ) , if we decrease their values of @xmath89 by a factor of 4 to account for the different methods for measuring the ews . with @xmath90 values of @xmath91 and @xmath56 ,
high redshift smgs lie in region 2c of the diagnostic plot in spoon et al .
sources in this region are classified as pah - dominated and , not surprisingly , m82 also falls in this region . utilizing _ iso _ spectra of starbursts , ulirgs and agn , genzel et al .
( 1998 ) used the ratio of the 7.7@xmath11 m pah line flux to the continuum flux , @xmath92 , as a diagnostic for agn contribution in the mid - ir emission .
in general , galaxies with @xmath93 are classified as starbursts and those with @xmath94 are agn - dominated . as we stress in section [ ap : meas ] , this type of analysis is strongly dependent on the specific choice made for the continuum fitting procedure .
nevertheless , we calculate @xmath95 for the composite smg and we note that all smgs in our sample satisfy the @xmath93 criterion for starburst - dominated sources .
this is true even for c1 which clearly shows a strongly rising continuum indicative of the presence of an agn .
based on fig .
[ fig : ew ] , the 6.2@xmath11 m pah line may be more useful in weeding out agn - dominated sources . in fig .
[ fig : lir ] we plot the pah line luminosities as a function of @xmath10 for smgs and local starbursts ( large panels ) . for the smgs
, we plot the @xmath96 coming from star formation ( i.e. after subtracting the component from the agn ) .
errors in @xmath10 for the smgs have been estimated from the error in the submm flux only , since this dominates ( assuming fixed ce01 templates ) . in all panels ,
the dashed line is the best - fit to the local starburst galaxies ; the smaller panels show the residual scatter around this best - fit relation .
the best - fit relation for the starburst galaxies only is consistent with the fit to both the starburst galaxies and the smgs .
the best - fit relation for both the smgs and starburst galaxies is given by @xmath97 the derived parameters are all within @xmath84 of those found by fitting only the starbursts .
the slopes are close to one indicating a direct proportionality between @xmath3 and the pah line luminosities .
the first thing to note is that the smgs lie on the relation established for local starburst galaxies extrapolated to very high luminosities .
this is what we expect if both are dominated by the same emission mechanism , namely star formation .
brandl et al .
( 2006 ) showed that the pah line luminosities are correlated with the ir luminosities within their starburst sample .
schweitzer et al . ( 2006 ) showed a similar correlation exists in a sample of qsos due to star - formation in them .
peeters et al .
( 2004 ) showed a correlation between @xmath98 and @xmath4 for a range of scales of star formation in local massive star - forming regions to normal and starburst galaxies .
the fact that the smgs lie on this relation demonstrates the reliability of our full sed fitting to derive the total infrared luminosity and confirms that smgs do have incredibly high infrared luminosities powered primarily by star formation .
note that c1 does not stick out as an outlier in these plots , since we have removed the agn component of its @xmath10 .
the second thing which stands out in fig .
[ fig : lir ] is that there is significant scatter in this relation for both the low redshift starbursts and the high redshift smgs . in the smaller panels
the two larger colored error bars quantify the @xmath84 scatter seen in the starbursts and smgs .
the scatter is similar for the smgs and starbursts and is considerably higher than would be implied by the measurement uncertainty .
other properties of individual objects , such geometry of the starburst region , spatial distribution of the dust , viewing angle relative to a merging event , clumpiness of the star formation , etc .
are likely playing a role in this scatter .
we found no correlations between the ir luminosity and equivalent widths of the pah lines in smgs .
this is consistent with what was found in brandl et al .
( 2006 ) for the local starburst galaxies , and shows that both the mid - ir continuum and the pah strength scale together with increasing @xmath10 .
we also find no correlation between x - ray luminosity and @xmath10 ( or the pah luminosities ) for the smgs , a correlation which would have been expected had the ir emission been dominated by an agn . using the kennicutt ( 1998 ) relation between @xmath10 and sfr , we plot the corresponding sfr on the top axis of fig . [ fig : lir ] .
the kennicutt ( 1998 ) relation between @xmath10 and sfr assumes a salpeter ( 1955 ) initial mass function and applies to starbursts with ages less than @xmath99myr . assuming that both the pah and far - ir emission are tracing the star formation
, we can use the kennicutt ( 1998 ) relation and the above pah line luminosity to @xmath10 correlations to determine the relation between sfr and the pah line luminosity .
we use the 6.2@xmath11 m pah line since it is the least affected by the silicate absorption and line blending .
we find that given the pah luminosity , one can calculate the sfr using : @xmath100 } \simeq 10^{-7 } \ , { \it l}_{\rm{pah,}6.2}\,[l_{\odot}].\\\ ] ] similar ( although perhaps slightly less accurate ) relations can be found for @xmath5 and @xmath6 using eq .
[ equ : lir ] .
we have shown from the mid - ir spectra of high redshift smgs that these systems are starburst - dominated and contain only a small contribution from the agn . despite , the fact that x - ray observations suggest that most smgs harbor an agn , only a small fraction of smgs ( 2/13 ) contain an agn which contributes significantly to the mid - ir luminosity . when extrapolated to the total ir luminosity ,
all smgs are dominated ( @xmath72 of @xmath10 ) by star formation and not agn activity . from table [ tab : lines ] we see that the median @xmath96 for our sample is @xmath101 , which corresponds to a sfr of around @xmath102 .
this is an extremely high rate of star formation , capable of creating a massive @xmath103 galaxy in only @xmath104years .
coupled with the low number density of smgs compared to other massive galaxies at high redshift , this leads us to suspect that smgs are a short - lived phase in the evolution of massive galaxies .
smgs are only one of several populations of high redshift galaxies which are ultraluminous in the infrared .
many @xmath2 galaxies selected in deep 24@xmath11 m surveys also qualify as ulirgs ( e.g. daddi et al .
2005 ; yan et al . 2005 , 2007 ) .
these other ulirgs are generally not detected at submm wavelengths .
what then are the fundamental differences between various populations of high redshift ulirgs ? one of the main factors which could cause a bifurcation in the population of ulirgs at high redshift is the presence or absence of a bolometrically significant agn .
the evolutionary scenario proposed by sanders et al .
( 1988 ) shows massive galaxies going through several main stages en route to becoming a massive elliptical .
it starts with an ir luminous phase most likely triggered by a massive merger . during this stage of intense star formation
, the agn is also growing .
as the agn becomes larger it begins to feedback on the galaxy , eventually quenching the star formation completely by blowing off all the dust and gas .
thus begins the qso phase , where the agn is free to dominate the emission .
eventually the qso consumes its gas and we end up with a quiescent massive elliptical galaxy . in this scenario ,
the submm emission will be at a maximum during the initial ir luminous phase ( springel , di matteo & hernquist 2005 ) . however , one uncertainty in this scenario is the timescale . in particular , how long is the ir luminous phase and how long does it take the agn to develop ? since most ulirgs at @xmath2 show a mix of agn and starburst activity ( e.g. houck et al .
2005 ; sajina et al .
2007 ) they fit into this evolutionary scenario at the stage where the agn has developed but before it has had a chance to expel all the gas and dust , since the ir luminosities are high . since smgs show less of an agn contribution , it is likely that they represent a slightly earlier phase of this scenario , before the agn has time to develop fully and is significant to the ir luminosity .
this is also consistent with the finding that smgs have cool dust temperatures ( pope et al .
2006 ) , partly because the star - formation is taking place on extended spatial scales and partly because the agn is not yet energetically important for heating the dust .
other high redshift ulirgs , like those selected at 24@xmath11 m ( e.g. sajina et al .
2007 ) , should have warmer dust temperatures ( or have overestimated infrared luminosities ) , otherwise they would be detected in the submm .
therefore , they likely represent the stage after the smg phase in this evolutionary scenario . for local ulirgs , there is evidence which shows that the luminosity of co emission decreases as the merger progresses ( rigopoulou et al .
1999 ) , which is consistent with decreasing submm emission in the later stages of the merger .
this idea is also consistent with morphological studies which show that smgs are often tidally disrupted systems ( chapman et al .
2003 ; conselice et al . 2003 ; pope et al . 2005 ) . alternatively ,
smgs and other high redshift ulirgs might appear different because of different merger progenitors .
since the number density of smgs is much less than that of all 24@xmath11 m selected ir - luminous galaxies , then this phase in the evolutionary sequence must be shorter .
we can make a rough estimate for the duration of the submm luminous phase . in the 2@xmath26gyrs between @xmath105
there are roughly 20 smgs and 500 24@xmath11m - selected galaxies ( with @xmath106 , see chary et al .
2004 ) within a portion of the goods
- n field .
assuming that all ir - luminous 24@xmath11m - selected galaxies will undergo a submm luminous stage during this period of time , we calculate the duration of the submm luminous phase to be on the order of @xmath104yrsm - selected galaxies will undergo a submm - luminous phase since smgs are known to be massive and could represent only the most extreme galaxies .
nevertheless , this calculation provides a useful reality check on the expected duration of the smg phase . ] . as mentioned above
, this timescale is what is needed to create a galaxy with a stellar mass of @xmath107 ( borys et al .
2005 ) at a rate of @xmath108 .
this timescale is consistent with those derived from gas masses of smgs ( greve et al .
_ spitzer _ irs spectroscopy has been obtained for a sample of 13 smgs brighter than 200@xmath11jy at 24@xmath11 m .
this has effectively confirmed the identification of these smg counterparts , and in some cases has provided a new redshift estimate .
the smgs show strong pah emission and , on average , only a small continuum contribution from hot dust .
we have explored several diagnostics from the mid - ir spectra of smgs to determine the level of agn contributing to the luminosity at these wavelengths .
all of them seem to converge on a picture in which smgs are starburst - dominated systems with at most a 30% contribution from an agn at mid - ir wavelengths .
their mid - ir spectra are similar to a scaled spectrum of local starburst galaxies like m82 .
the classification of an smg as agn or starburst - dominated from the mid - ir often disagrees with the classification from the x - rays . while x - ray observations are better suited for determining the presence of an agn in an smg , the mid - ir spectra can determine how important the agn is to the total infrared luminosity since it is directly detecting the hot dust emission and not subject to details about the geometry of the obscuring material .
full ir sed fits to the mid - ir data , far - infrared and radio photometry show that smgs are best - fit by scaled up versions of local ir - luminous galaxy templates with additional extinction from the draine ( 2003 ) extinction curves .
these models have cool dust temperatures ( @xmath109k ) and high @xmath10s which imply sfrs of @xmath110 assuming a salpeter initial mass function .
smgs lie on the local relation between @xmath10 and @xmath4 ( or @xmath5 or @xmath6 ) , which means that the pah line flux can be used to estimate the sfrs in these systems , albeit with large uncertainties .
equivalent widths , which have been the focus of some other studies , are much less useful in this regard .
smgs appear to be an earlier phase in the evolution of massive galaxies than other high redshift ulirgs .
the average high redshift ulirg shows warmer dust and less pah emission .
it is therefore likely that the agn in smgs has not yet grown strong enough to heat the dust substantially and destroy the pah molecules .
we have been able to put constraints on the contribution from agn activity to the restframe 512@xmath11 m mid - ir emission , and have extrapolated this to far - ir wavelengths . without more data from 30100@xmath11
m rest frame it is difficult to directly probe the emission at far - ir wavelengths .
we will be able to model this part of the spectrum in more detail once data from the _ herschel space observatory_[multiblock footnote omitted ] and scuba-2 ( holland et al .
2006 ) become available .
details of the smg phase will be understood only after the spatially resolved spectroscopy capabilities of alma[multiblock footnote omitted ] become available .
we would like to thank the referee for his / her helpful suggestions which improved the quality of this paper .
we thank anna sajina for helpful discussions and for providing composite irs spectra of their high redshift _ spitzer _
24@xmath1m - selected ulirgs .
we are very grateful to bernhard brandl for providing the irs spectra of local starburst galaxies and benjamin magnelli for providing information from the completeness simulations of the 70@xmath11 m images .
this work was supported by the natural sciences and engineering research council of canada and the canadian space agency .
dma acknowledges the royal society for support .
this work is based on observations made with the spitzer space telescope , which is operated by the jet propulsion laboratory , california institute of technology under a contract with nasa .
support for this work was provided by nasa through an award issued by jpl / caltech .
the irs was a collaborative venture between cornell university and ball aerospace corporation funded by nasa through the jet propulsion laboratory and ames research center .
[ fig : rms ] shows the final rms of our spectra as a function of the integration time and spectral order .
values from the _ spitzer _ science center sensitivity calculator ; spec - pet[multiblock footnote omitted ] are given for comparison as the dashed curves .
we confirm that the sensitivity goes as @xmath111 even for very long integrations .
the scatter in the measured values is probably dominated by the accuracy of the sky subtraction in these crowded fields . given the scatter , our observed sensitivities are consistent with those from spec - pet .
the decrease in sensitivity in the sl1 order could be related to our choice of 240 second ramps in fields with backgrounds @xmath112mjy@xmath26sr@xmath113 at 24@xmath11 m , which results in incorrect droop correction ( see the _ spitzer _ observer s manual ) .
here we describe several of our sources in more detail .
sources which are not discussed here are relatively straightforward in terms of interpreting their mid - ir spectra .
the optical and irs redshifts disagree for this smg .
the irs redshift is based on 3 pah features : 6.2 , 7.7 and 11.3@xmath11 m and the optical spectroscopic redshift of 1.865 is inconsistent with the positions of these emission features , even within the errors .
the optical spectroscopic redshift of gn06 comes from chapman et al .
( 2005 ) where they note that the optical galaxy was offset from the radio counterpart by about an arcsecond . from the images of gn06 in fig .
[ fig : post ] , we see there are two optical knots both of which are offset from the counterpart which is detected in the radio and mid - ir .
for this source , we conclude that the optical and mid - ir spectra are picking up different galaxies and the submm emission is coming from the mid - ir detected galaxy .
c2 has the largest discrepancy between the irs and uv / optical spectroscopic redshifts .
the irs redshift for this source is based on the 7.7 and 11.3@xmath11 m pah features and there is also a hint of the 8.6@xmath11 m line .
the optical spectroscopic redshift would have placed the 7.7@xmath11 m pah line right in the middle of this spectrum and therefore it is inconsistent with the irs spectrum .
the optical spectroscopic redshift comes from detections of both h@xmath114 and lyman@xmath114 emission ( swinbank et al . 2004 ; chapman et al . 2005 ) .
the images in fig .
[ fig : post ] show an extended optical galaxy .
this is a curious source , however since the mid - ir is more likely to trace the submm emission , we conclude that the irs redshift is correct for this smg .
this smg is not detected in the deep _ chandra _ 2 msec imaging of goods - n .
furthermore , the optical spectrum from swinbank et al .
( 2004 ) clearly shows h@xmath114 , nii and the sii doublet and the low nii to h@xmath114 ratio is used to classify this smg as a starburst .
however the mid - ir spectrum of c1 clearly shows a steeply rising continuum indicative of an agn emission .
this is also apparent from the extreme _
colors shown in fig .
[ fig : cc ] .
the mid - ir spectrum also shows pah emission and the ratio of pah luminosity to ir luminosity for c1 is significantly lower than those seen in local starbursts .
we conclude that this smg harbors an compton - thick agn which accounts for 44% of the total ir luminosity .
this is the only smg which contains a bolometrically significant agn .
interestingly , c1 has not been detected in co ( greve et al .
2005 ) . from the x - ray flux limit
, we can place a limit on the column density of @xmath115 .
c1 is relatively bright at 24@xmath11 m ( @xmath116mjy ) suggesting that these types of obscured agn at high redshift should be detectable in even shallow _
surveys .
one further smg deserves special discussion and that is gn39 .
this smg has two radio / mips counterparts separated by 8 arcseconds , which we refer to as gn39a and gn39b .
both are confirmed to lie at @xmath117 ( swinbank et al .
2004 ; chapman et al .
2005 ) , therefore it is likely that they are both contributing to the submm emission .
the separation corresponds to 70 kpc at this redshift .
our irs observations picked up both counterparts and , with a narrow extraction window , we were able to extract a separate spectrum of each component .
some of the flux from gn39a falls into the spectral extraction window of gn39b however a comparison of the 24@xmath11 m flux from the irs spectra to that from the mips 24@xmath11 m image confirms that only 1/3 of the flux in the irs spectrum of gn39b is from gn39a .
the extracted spectra confirm that these two sources are indeed at the same redshift and both spectra look very similar to each other and to the rest of the smgs , supporting the idea that they are both submm emitters .
furthermore , both components are x - ray detected and classified as obscured agn by alexander et al .
( 2005 ) , although the two irs spectra show a completely starburst - dominated system at mid - ir wavelengths . since these two smgs are at the same redshift , we know that they must be associated .
given the physical distance between them , larger than those seen in most local ulirgs ( murphy et al .
1996 ) , we could be seeing the early stage in the merger of two massive galaxies ( springel , di matteo & hernquist 2005 ) .
note that gn04 , gn07 and gn19 also have double radio counterparts , which are thought to be at the same redshift .
however , in these sources , the two components are separated by only 3 arcseconds and therefore they do not resolve into two separate 24@xmath11 m sources .
thus the irs spectra for these sources contain contributions from both components .
alexander d.m . , et al . , 2003 ,
aj , 126 , 539 alexander d.m . , bauer f.e . ,
chapman s.c .
, smail i. , blain a.w . , brandt w.n .
, ivison r.j . , 2005 ,
apj , 632 , 736 allamandola , l. j. , hudgins , d. m. , & sandford , s. a. 1999 , , 511 , l115 aretxaga , i. , et al .
2007 , mnras , 379 , 1571 armus , l. , et al .
2007 , , 656 , 148 bauer , f. e. , alexander , d. m. , brandt , w. n. , hornschemeier , a. e. , vignali , c. , garmire , g. p. , & schneider , d. p. 2002
, , 124 , 2351 blain a.w .
, smail i. , ivison r.j .
, kneib j .-
, frayer d.t .
, 2002 , phr , 369 , 111 borys c. , chapman s. , halpern m. , scott d. , 2003 , mnras , 344 , 385 borys c. , smail i. , chapman s.c . , blain a.w . ,
alexander d.m . ,
ivison r.j . , 2005 ,
apj , 635 , 853 brandl , b. r. , et al .
2006 , , 653 , 1129 brandt , w. n. , & hasinger , g. 2005 , , 43 , 827 capak p. , et al . , 2004 ,
aj , 127 , 180 chapman , s. c. , windhorst , r. , odewahn , s. , yan , h. , & conselice , c. 2003 , , 599 , 92 chapman s. c. , blain a. w. , smail i. , ivison r. j. , 2005 , apj , 622 , 772 charmandaris v. , et al . , 1999 , , 266 , 99 chary r. , elbaz d. , 2001 , apj , 556 , 56 ( ce01 ) chary , r. , et al . 2004 , , 154 , 80 coppin k. , et al .
, 2006 , , 372 , 1621 conselice , c. j. , chapman , s. c. , & windhorst , r. a. 2003 , , 596 , l5 cowie l.l . ,
barger a.j . , hu e.m . ,
capak p. , songaila a. , 2004 , aj , 127 , 3137 daddi e. , et al .
, 2005 , apj , 631 , l13 daddi e. , et al . , 2007 ,
apj in press , ( astro - ph/0705.2832 ) desai v. , et al . , 2007 , apj , 669 , 810 dickinson , m. , giavalisco , m. , & the goods team 2003 , the mass of galaxies at low and high redshift , 324 draine , b. t. 2003 , , 41 , 241 draine , b. t. , & li , a. 2007 , , 657 , 810 dunne l. , eales s. , edmunds m. , ivison r. , alexander p. , clements d.l . , 2000 ,
mnras , 315 , 115 farrah , d. , et al .
2007 , , 667 , 149 fischer j. , et al .
, 1999 , , 266 , 91 frster schreiber , n. m. , et al . , 2003 , , 399 , 833 frayer , d. t. , et al .
2006 , , 647 , l9 genzel , r. , et al .
1998 , , 498 , 579 giavalisco m. , et al . , 2004 , apj , 600 , l93 greve t.r .
, et al . , 2005 , mnras , 359 , 1165 hauser , m. g. , & dwek , e. 2001 , , 39 , 249 higdon , s. j. u. , et al .
2004 , , 116 , 975 holland w.s . , et al . , 1999 , mnras , 303 , 659 holland w.s . , et al . , 2006 , spie conference , 6275 , 45 houck , j. r. , et al . 2004 , , 154 , 18 houck , j. r. , et al .
2005 , , 622 , l105 huynh , m. t. , pope , a. , frayer , d. t. , & scott , d. 2007 , , 659 , 305 ivison r.j . , et al . , 2002 , mnras , 337 , 1 ivison r.j . , et al . , 2004 , apjs , 154 , 124 kennicutt r. c. , 1998 , ara&a , 36 , 189 kovcs , a. , chapman , s. c. , dowell , c. d. , blain , a. w. , ivison , r. j. , smail , i. , & phillips , t. g. 2006 , , 650 , 592 lagache , g. , puget , j .- l . , & dole , h. 2005 , , 43 , 727 le floch , e. , et al .
2005 , , 632 , 169 lilly s.j .
, eales s.a . ,
gear w.k.p . , hammer f. , le f ` evre o. , crampton d. , bond j.r . , dunne l. , 1999 , apj , 518 , 641 lutz , d. , spoon , h. w. w. , rigopoulou , d. , moorwood , a. f. m. , & genzel , r. 1998 , , 505 , l103 lutz , d. , genzel , r. , kunze , d. , rigopoulou , d. , spoon , h. w. w. , sturm , e. , tran , d. , & moorwood , a. f. m. 1999 , , 266 , 85 lutz , d. , maiolino , r. , spoon , h. w. w. , & moorwood , a. f. m. 2004 , , 418 , 465 lutz d. , valiante e. , sturm e. , genzel r. , tacconi l. j. , lehnert m. d. , sternberg a. , baker a. j. , 2005a , apj , 625 , l83 lutz , d. , yan , l. , armus , l. , helou , g. , tacconi , l. j. , genzel , r. , & baker , a. j. 2005b , , 632 , l13 menndez - delmestre , k. , et al .
2007 , , 655 , l65 murphy , t. w. , jr . , armus , l. , matthews , k. , soifer , b. t. , mazzarella , j. m. , shupe , d. l. , strauss , m. a. , & neugebauer , g. 1996 , , 111 , 1025 prez - gonzlez , p. g. , et al .
2005 , , 630 , 82 pope a. , borys c. , scott d. , conselice c. , dickinson m. , mobasher b. , 2005 , mnras , 358 , 149 pope , a. , et al .
2006 , , 370 , 1185 pope a. , 2007 , phd thesis , university of british columbia puget , j. l. , & leger , a. 1989 , , 27 , 161 puget , j .- l . , abergel , a. , bernard , j .-
p . , boulanger , f. , burton , w. b. , desert , f .- x .
, & hartmann , d. 1996 , , 308 , l5 rigopoulou d. , spoon h. w. w. , genzel r. , lutz d. , moorwood a. f. m. , tran q. d. , 1999 , aj , 118 , 2625 sajina , a. , lacy , m. , & scott , d. 2005 , , 621 , 256 sajina , a. , yan , l. , armus , l. , choi , p. , fadda , d. , helou , g. , & spoon , h. 2007 , apj , 664 , 713 salpeter e.e . , 1955 , apj , 121 , 161 sanders , d. b. , soifer , b. t. , elias , j. h. , madore , b. f. , matthews , k. , neugebauer , g. , & scoville , n. z. 1988 , , 325 , 74 sanders , d. b. , & mirabel , i. f. 1996 , , 34 , 749 schweitzer , m. , et al .
2006 , , 649 , 79 smith , j. d. t. , et al .
2007 , , 656 , 770 soifer , b. t. , sanders , d. b. , madore , b. f. , neugebauer , g. , danielson , g. e. , elias , j. h. , lonsdale , c. j. , & rice , w. l. 1987 , , 320 , 238 spoon , h. w. w. , marshall , j. a. , houck , j. r. , elitzur , m. , hao , l. , armus , l. , brandl , b. r. , & charmandaris , v. 2007 , , 654 , l49 springel , v. , di matteo , t. , & hernquist , l. 2005 , , 361 , 776 sturm , e. , lutz , d. , tran , d. , feuchtgruber , h. , genzel , r. , kunze , d. , moorwood , a. f. m. , & thornley , m. d. 2000 , , 358 , 481 swinbank a.m. , smail i. , chapman s.c . , blain a.w .
, ivison r.j . ,
keel w.c . , 2004 ,
apj , 617 , 64 tacconi l.j . ,
et al . , 2006 , , 640 , 228 teplitz h.i . , charmandaris v. , chary r. , colbert j.w . , armus l. , weedman d. , 2005 , apj , 634 , 128 teplitz , h. i. , et al .
2007 , , 659 , 941 tran , q. d. , et al .
2001 , , 552 , 527 valiante , e. , lutz , d. , sturm , e. , genzel , r. , tacconi , l. j. , lehnert , m. d. , & baker , a. j. 2007 , , 660 , 1060 yan l. , et al . , 2005 , apj , 628 , 604 yan , l. , et al .
2007 , , 658 , 778 ccrrrrrr 123653.1 + 621120 & gn31 & 12:36:53.22 & 62:11:16.7 & 0.301 @xmath118 0.007 & 0.367 @xmath118 0.006 & @xmath119 + 123635.5 + 621238 & gn26 & 12:36:34.51 & 62:12:40.9 & 0.992 @xmath118 0.012 & 0.446 @xmath118 0.005 & @xmath120 + 123701.2 + 621147 & gn17 & 12:37:01.59 & 62:11:46.2 & 0.209 @xmath118 0.006 & 0.710 @xmath118 0.008 & @xmath119 + 123622.6 + 621629 & c2 & 12:36:22.66 & 62:16:29.5 & 0.098 @xmath118 0.006 & 0.414 @xmath118 0.007 & @xmath119 + 123555.1 + 620901 & c3 & 12:35:55.13 & 62:09:01.6 & n / a & 0.374 @xmath118 0.009 & @xmath119 + 123711.1 + 621325 & gn39a & 12:37:11.37 & 62:13:31.1 & 0.097 @xmath118 0.005 & 0.537 @xmath118 0.009 & @xmath119 + 123711.9 + 621331 & gn39b & 12:37:11.97 & 62:13:25.8 & 0.051 @xmath118 0.005 & 0.225 @xmath118 0.007 & @xmath119 + 123621.3 + 621711 & gn07 & 12:36:21.27 & 62:17:08.1 & 0.048 @xmath118 0.007 & 0.370 @xmath118 0.011 & @xmath119 + 123618.7 + 621553 & gn06 & 12:36:18.33 & 62:15:50.4 & 0.036 @xmath118 0.006 & 0.330 @xmath118 0.008 & @xmath119 + 123600.2 + 621047 & c1 & 12:36:00.16 & 62:10:47.3 & 0.478 @xmath118 0.011 & 1.220 @xmath118 0.008 & @xmath121 + 123618.8 + 621008 & gn05 & 12:36:19.13 & 62:10:04.3 & 0.044 @xmath118 0.007 & 0.215 @xmath118 0.006 & @xmath119 + 123707.7 + 621411 & gn19 & 12:37:07.19 & 62:14:08.0 & @xmath122 & 0.280 @xmath118 0.015 & @xmath119 + 123616.6 + 621520 & gn04 & 12:36:16.11 & 62:15:13.5 & 0.060 @xmath118 0.006 & 0.303 @xmath118 0.007 & @xmath119 + crrrcr gn31 & & 35@xmath1232 & & 2.3 + gn26 & 15@xmath1232 & 20@xmath1232 & [ 12@xmath1232 ] & 3.3 [ 4.1 ] + gn17 & 20@xmath1232 & 12@xmath1232 & & 3.5 + c2 & & & 45@xmath1232 & 3.0 + c3 & & 31@xmath1232 & 40@xmath1232 & 4.7 + gn39a & & 20@xmath1232 & 28@xmath1232 & 3.2 + gn39b & & [ 20@xmath1232]@xmath124 & [ 28@xmath1232 ] & [ 3.2 ] + gn07 & & 35@xmath1232 & 45@xmath1232 & 5.3 + gn06 & & 35@xmath1232 & 55@xmath1232 & 6.0 + c1 & & 6@xmath1232 & 10@xmath1232 & 1.1 + gn05 & & & 55@xmath1232 & 3.7 + gn19 & & [ 20@xmath1232 ] & [ 28@xmath1232 ] & [ 3.2 ] + gn04 & & & 70@xmath1232 & 4.7 + crrrc gn31 & 21.8 & 0.935@xmath125 & 0.93 @xmath118 0.03 & + gn26 & 22.7 & 1.219@xmath126 & 1.23 @xmath118 0.01 & + gn17 & 27.7 & 1.72@xmath127 & 1.73 @xmath118 0.01 & new irs redshift + c2 & 25.5 & 2.466@xmath128 & 1.79 @xmath118 0.04 & inconsistent + c3 & 23.5@xmath129 & 1.875@xmath126 & 1.88 @xmath118 0.02 & + gn39a & 25.9 & 1.996@xmath130 & 1.98 @xmath118 0.01 & + gn39b & 26.4 & 1.992@xmath126 & 1.99 @xmath118 0.04 & + gn07 & 27.8/23.9@xmath131 & 1.988@xmath126 , 1.992@xmath130 & 1.99 @xmath118 0.02 & + gn06 & 27.8 & 1.865@xmath126 & 2.00 @xmath118 0.03 &
inconsistent + c1 & 24.8 & 1.994@xmath126 , 2.002@xmath130 & 2.01 @xmath118 0.05 & + gn05 & 24.9 & 2.60@xmath127 & 2.21 @xmath118 0.03 & new irs redshift + gn19 & 25.4/@xmath13228@xmath131 & 2.484@xmath126 , 2.490@xmath130 & 2.48 @xmath118 0.03 & + gn04 & 26.2/@xmath13228@xmath131 & 2.578@xmath126 & 2.55 @xmath118 0.01 & + ccccr gn31 & u & n / a & sb & @xmath133 + gn26 & sb & sb & sb & @xmath134 + gn17 & sb & n / a & sb & @xmath135 + c2 & agn & sb & sb & @xmath136 + c3 & agn & sb & sb+agn & 48 + gn39a & oagn & sb & sb & @xmath137 + gn39b & oagn & sb & sb & @xmath138 + gn07 & sb & sb & sb+weak agn & 18 + gn06 & sb & sb & sb+agn & 47 + c1 & sb & sb & agn+weak sb & 82 + gn05 & u & n / a & sb & @xmath139 + gn19 & oagn & sb & sb & @xmath140 + gn04 & agn & sb & agn+sb & 61 + crrrrrrrrr gn31 & ... & @xmath141 & ... & ... & @xmath142 & ... & @xmath143 & @xmath144 & 0.27 + gn26 & @xmath145 & @xmath146 & @xmath147 & @xmath148 & @xmath149 & @xmath150 & @xmath151 & @xmath152 & 3.6 + gn17 & @xmath153 & ... & ... & @xmath154 & ... & ... & @xmath155 & @xmath156 & 2.0 + c2 & ... & @xmath157 & @xmath158 & ... & @xmath159 & @xmath160 & @xmath161 & @xmath162 & 3.3 + c3 & @xmath163 & @xmath164 & @xmath165 & @xmath166 & @xmath167 & @xmath168 & 5.2 & 0.8 & 6.0 + gn39a & @xmath169 & @xmath170 & @xmath171 & @xmath172 & @xmath173 & @xmath174 & @xmath175 & @xmath176 & 5.0 + gn39b & @xmath177 & @xmath178 & @xmath179 & @xmath180 & @xmath181 & @xmath182 & @xmath183 & @xmath152 & 2.8 + gn07 & @xmath184 & @xmath185 & @xmath186 & @xmath187 & @xmath188 & @xmath189 & 6.7 & 0.2 & 6.9 + gn06 & @xmath190 & @xmath191 & @xmath192 & @xmath193 & @xmath194 & @xmath195 & 5.7 & 0.7 & 6.4 + c1 & @xmath196 & @xmath197 & @xmath198 & @xmath199 & @xmath200 & @xmath201 & 5.4 & 4.2 & 9.6 + gn05 & ... & @xmath202 & ... & ... & @xmath203 & ... & @xmath204 & @xmath152 & 4.0 + gn19 & @xmath205 & @xmath206 & ... & @xmath207 & @xmath208 & ... & @xmath209 & @xmath210 & 12.3 + gn04 & @xmath211 & @xmath212 & ... & @xmath213 & @xmath214 & ... & 4.5 & 2.1 & 6.6 + | we present deep mid - infrared spectroscopy with the _
spitzer space telescope _ of 13 submillimeter galaxies ( smgs ) in the great observatories origins deep survey north ( goods - n ) field .
we find strong polycyclic aromatic hydrocarbon ( pah ) emission in all of our targets which allows us to measure mid - ir spectroscopic redshifts and place constraints on the contribution from star formation and active galactic nuclei ( agn ) activity to the mid - ir emission . in the high signal - to - noise ratio composite spectrum of smgs
, we find that the hot dust continuum from an agn contributes at most 30% of the mid - ir luminosity .
individually , only @xmath0 smgs have continuum emission dominating the mid - infrared luminosity ; one of these smgs , c1 , remains undetected in the deep x - ray images but shows a steeply rising continuum in the mid - ir indicative of a compton - thick agn .
we find that the mid - ir properties of smgs are distinct from those of 24@xmath1m - selected ulirgs at @xmath2 ; the former are predominantly dominated by star formation while the latter are a more heterogeneous sample with many showing significant agn activity .
we fit the irs spectrum and the mid - ir to radio photometry of smgs with template spectral energy distributions to determine the best estimate of the total infrared luminosity from star formation .
while many smgs contain an agn as evinced by their x - ray properties , our multi - wavelength analysis shows that the total infrared luminosity , @xmath3 , in smgs , is dominated by star formation and not agn activity .
we find that high redshift smgs lie on the relation between @xmath3 and @xmath4 ( or @xmath5 or @xmath6 ) that has been established for local starburst galaxies .
this suggests that pah luminosity can be used as a proxy for the star formation rate in smgs .
smgs are consistent with being a short lived cool phase in a massive merger . although there appears to be an agn present in many smgs , it does not appear to have become strong enough to heat the dust and dominate the mid- or far - infrared emission . |
the extreme low temperature required for the observation of bose - einstein condensation alkali vapors is a major achievement of present day experimentation @xcite .
unfortunately , the temperature itself is not directly observable .
it has to be estimated from other , more accessible properties of the system , such as the velocity distribution of the vapor @xcite . because of the uncertainty relation between the inverse temperature and the energy of the system for a finite number of particles , it is not clear _ a priori _ how many particles
are required for a reasonable thermodynamically defined concept of temperature . in the present paper
, we do not address this fundamental question , but rather assume that the number of bosons in the system is sufficiently large to have a meaningful definition of the temperature . under this assumption ,
the experimental estimate of the temperature@xcite is based on the hypothesis that the momentum distribution function is maxwell - boltzmann like , at least for sufficiently high momentum .
this hypothesis deserves further investigation and quantitative comparison to the momentum distribution of a parabolically confined bose - einstein gas . to the best of our knowledge ,
the bose - einstein momentum distribution in a parabolic confinement potential has not been documented before . using the methods for path integration developed in @xcite
, we have been able to calculate the momentum distribution function of a gas of parabolically confined bosons with quadratic harmonic interparticle interactions .
this calculation is presented in sec .
other investigations using the same methodology @xcite indicated that for weakly interacting systems , such as @xmath0rb and @xmath1na , the interaction slightly influences the parameters involved in the model system , and therefore have only a moderate influence the momentum distribution . in section iii ,
the approximations are examined which lead to the maxwellian momentum distribution function , and a quantitative comparison is made between the temperature obtained in the bose - einstein distribution and the estimate of the temperature , obtained by fitting a maxwell - boltzmann distribution to the tail of the measured velocity distribution .
the conclusion from this analysis is that the estimate from the maxwellian deviates substantially from the temperature of the bose - einstein distribution in the presence of a bose - einstein condensate .
the anisotropic one - body potential energy @xmath2 and the two - body potential energy @xmath3 of the model system consisting of @xmath4 bosons are given by @xmath5 where @xmath6 is the position of the @xmath7th boson , @xmath8 is the mass of the bosons and @xmath9 , @xmath10 are the frequencies of the parabolic confinement in the @xmath11-plane and along the @xmath12-axis , respectively .
the model interparticle interaction is harmonic with frequency @xmath13 and can be either attractive ( ) or repulsive ( + ) .
the renormalized frequencies in the @xmath11-plane and along the @xmath12-axis will be denoted by @xmath14 and @xmath15 , respectively .
these frequencies are the eigenfrequencies of the internal degrees of freedom of the boson gas , whereas the confining frequencies @xmath9 , @xmath10 are the eigenfrequencies for the center - of - mass motion . the momentum distribution @xmath16 , giving the statistical mean number of bosons with a given momentum @xmath17 in the canonical ensemble , can be written as @xmath18 where the angular brackets indicate quantum averages for boson statistics .
these mean values can be calculated for a system at temperature @xmath19 using the path integral propagator in momentum space @xmath20 : @xmath21 the calculation of the momentum distribution , analogous to the calculation of the spatial density distribution @xmath22 in @xcite , can be summarized as follows .
first the propagator in momentum space is evaluated by a fourier transform of the corresponding propagator in position space obtained in @xcite .
this results in @xmath23 with @xmath24 where @xmath25 is the total momentum , and @xmath26 is the one - particle propagator in momentum space for a particle in an anisotropic harmonic potential .
next , the sum over permutations appearing in the expression ( [ meanval ] ) for the mean value @xmath27 is transformed into a cyclic sum ( with @xmath28 the number of cycles of length @xmath29 ) .
the cyclic summation can be performed for the generating function @xmath30 where @xmath31 is the partition sum .
the result of the cyclic summation is @xmath32 \right\ } } { 8\sinh \left ( \ell \beta \hbar w_z/2\right ) \sinh ^2\left ( \ell \beta \hbar w_{xy}/2\right ) } .\end{aligned}\ ] ] the expectation value @xmath27 for @xmath4 bosons can then be derived in closed form from the generating functional @xmath33 by collecting all the terms with the same power in @xmath34 : @xmath35 \right\ } } { 8\sinh \left ( \ell \beta \hbar w_z/2\right ) \sinh ^2\left ( \ell \beta \hbar w_{xy}/2\right ) } .
\label{expval}\end{aligned}\ ] ] finally , the momentum distribution ( [ veld ] ) is found as a fourier transform of the expectation value ( [ expval ] ) .
the result is : @xmath36 with @xmath37 .\ ] ] in figure 1 this momentum distribution is shown at different temperatures for an isotropically confined bose gas with @xmath38 and consisting of 1000 bosons .
the momentum distribution for @xmath39 is qualitatively similar to the distribution with @xmath40 shown in figure 1 . in the inset
, the value of the momentum distribution in the origin is shown as a function of the temperature .
as the temperature is lowered below the condensation temperature , the momentum distribution becomes more pronounced in the origin : the average number of bosons with momentum zero rises sharply .
the momentum distribution ( [ exact ] ) calculated in the previous section for a gas of parabolically confined bosons can be related to the maxwell distribution through a series of approximations .
first , in the case of a non - interacting ( @xmath41 ) , parabolically confined boson gas the momentum distribution ( [ exact ] ) becomes @xmath42 furthermore , according to feynman @xcite , the maxwell - boltzman statistics can be recovered by considering only the cycles of length @xmath43 in the cyclic summations of ( [ exact ] ) .
this means that permutations involving cycles of length larger than 1 play no role any more .
hence the momentum distribution function for distinguishable non - interacting particles in a parabolic potential is : @xmath44 if moreover the parabolic confinement were neglected , which corresponds to @xmath45 in ( [ vclas ] ) , one finally would be left with the maxwell - boltzmann momentum distribution for free particles : @xmath46 in experiments on bose - einstein condensed atomic vapors ( e.g. @xcite ) , the temperature of a cloud of parabolically trapped bosons was estimated by fitting the _ maxwell - boltzmann distribution _ _ _ ( [ max ] ) to the tail of the experimentally measured momentum distribution @xmath47 .
this can be done by first choosing a sufficiently large threshold momentum @xmath48 such that atoms with a momentum larger than @xmath48 do not belong to the bose - einstein condensate , and then minimizing @xmath49 with respect to the temperature @xmath50 appearing in the maxwell - boltzmann distribution .
the temperature obtained by this minimization will be denoted as @xmath51 and the procedure will be referred to as the `` maxwell - boltzmann fit procedure '' .
since the maxwell - boltzmann momentum distribution is a rather crude approximation to the momentum distribution ( [ vid ] ) of a gas of non - interacting , parabolically trapped bosons , we rather propose to determine the temperature by fitting @xmath52 to the momentum distribution ( [ vid ] ) derived in the present paper .
this is done by minimizing @xmath53 with respect to the temperature @xmath54 appearing in the momentum distribution @xmath55 for non - interacting , parabolically trapped bosons .
we will denote this procedure as the `` bose - einstein fit procedure '' and the resulting temperature as @xmath56 . in order to compare the maxwell - boltzmann fit procedure with the bose - einstein fit procedure , we applied the former on a momentum distribution simulated with @xmath55 ( @xmath57 for weakly interacting bosons ) , at a given temperature @xmath56 .
that is , @xmath58 was minimized with respect to @xmath51 , keeping @xmath59 and @xmath48 fixed .
the result @xmath51 of the minimization is compared to @xmath56 in figure 2 , where the ratio @xmath60 is shown as a function of @xmath56 at a fixed threshold momentum @xmath61 . in the bose - einstein condensed phase the discrepancy between @xmath56 and @xmath51 increases
when the number of bosons increases , whereas for @xmath62{n/\zeta \left ( 3\right ) } , $ ] the discrepancy decreases when the number of bosons is augmented , as shown in the inset of figure 2 . for temperatures well above the condensation temperature @xmath63 , the ratio @xmath64 is small , and the maxwell - boltzmann fit procedure gives a reliable temperature estimate in this high - temperature regime .
the difference between @xmath56 and @xmath51 becomes appreciable below the condensation temperature @xmath63 .
the present analysis therefore shows that @xmath60 is not negligible in the bose - einstein condensed phase and that the bose - einstein fit procedure is more appropriate than the maxwell - boltzmann fit procedure for the determination of the temperature .
in this paper , we calculated the momentum distribution function ( [ exact ] ) for a bose gas in an anisotropic parabolic trapping potential , including quadratic interparticle interactions . in the absence of interparticle interactions
, this momentum distribution simplifies to the momentum distribution function ( [ vid ] ) of an ideal boson gas in a harmonic confining potential .
the maxwell - boltzmann distribution ( [ max ] ) follows from ( [ vid ] ) by the further simplifications of neglecting the boson statistics and the parabolic confinement .
the underlying assumption in these derivations is that the number of bosons is sufficiently large to maintain the temperature concept from the thermodynamical limit .
assuming furthermore thermodynamical equilibrium , two temperature estimates were compared . since the temperature estimates used in experiments on bose - einstein condensed atomic vapors ( e.g. @xcite ) , are obtained by fitting the maxwell - boltzmann distribution _ _ ( [ max ] ) to the tail of the experimentally measured momentum distribution @xmath65 we derived the corresponding temperature @xmath51 by fitting the large - momentum tail of the maxwell - boltzmann distribution to the large - momentum tail of the momentum distribution of the bose gas ( [ vid ] ) .
this temperature estimate was compared to the temperature @xmath56 in the momentum distribution of the boson gas .
we find that @xmath51 substantially deviates from the correct temperature @xmath56 in the bose - einstein condensed phase , and hence that the bose - einstein fit procedure is more appropriate than the maxwell - boltzmann fit procedure for the determination of the temperature .
part of this work is performed in the framework of the fwo projects no .
1.5.729.94 , 1.5.545.98 , g.0287.95 , g.0071.98 , and wo.073.94n ( wetenschappelijke onderzoeksgemeenschap over `` laagdimensionele systemen '' , scientific research community of the fwo on `` low dimensional systems '' ) , the `` interuniversitaire attractiepolen belgische staat , diensten van de eerste minister
wetenschappelijke , technische en culturele aangelegenheden '' , and in the framework of the bof noi 1997 projects of the universiteit antwerpen .
two of the autors ( j.t . and f.b .
) acknowledge the fwo for financial support .
* fig.1 : * in figure 1 the average number @xmath66 of particles with a momentum @xmath67 is shown at several temperatures for a bose gas in an isotropic parabolic confinement , as function of the momentum .
units are chosen such that @xmath68 . in this set of units ,
the momentum is expressed in @xmath69 . in the inset ,
the momentum distribution at @xmath70 is shown as a function of temperature , with @xmath71{n/\zeta ( 3)}$ ] .
for this figure we have chosen @xmath38 : the momentum distribution for @xmath39 is qualitatively similar to the distribution for @xmath40 .
* fig.2 : * the relative difference between the temperature @xmath51 found by fitting a maxwell - boltzmann distribution and the actual temperature @xmath56 is shown in this figure , representing @xmath72 versus @xmath56 in units @xmath73{n/\zeta
this is shown for 500 , 1000 and 2000 non - interacting bosons . in the inset
, the dependence of @xmath72 on the number of particles is shown . | the momentum distribution function of a parabolically confined gas of bosons with harmonic interparticle interactions is derived . in the bose - einstein condensation region
, this momentum distribution substantially deviates from a maxwell - boltzmann distribution .
it is argued that the determination of the temperature of the boson gas from the bose - einstein momentum distribution function is more appropriate than the currently used fitting to the high momentum tail of the maxwell - boltzmann distribution . |
solving the schrdinger equation with time - dependent boundary conditions , including moving ones , is a very hard work and can only be done in a few cases .
@xcite and @xcite for a recent review .
very interesting effects are seen in such problems ; diffraction in time @xcite is just such an instance .
this phenomenon is characterized by quantum temporal oscillations in matter waves released from a confining region .
it was shown when a wall acting as a perfect mirror , moves with finite velocity along the direction of propagation of a beam , the visibility of the fringes is enhanced @xcite .
moshinsky s theoretical work has been extended to the case of particles which are suddenly released from a 1d box @xcite and to particles with angular momentum @xcite initially trapped in a hard spherical box .
exact solutions of the schrdinger equation for a particle in a 1d box with a moving wall have been found @xcite . using the semiclassical approximation ,
luz and cheng @xcite evaluated the exact propagator of the problem .
grosche @xcite did this task independently by means of an exact ( mid - point ) summation of the perturbation series with point - like perturbations .
the motivation of the present work is that sudden removal of the boundary is an idealized case , thus we consider a moving boundary instead of a sudden removal of the boundary .
the limit of infinite velocity of the moving boundary clearly corresponds to the sudden removal case .
possible applications to optical effects connected with moving mirrors are additionally encouraging . for these reasons
we aim to solve the schrdinger equation for a particle in a hard sphere with varying radius .
consider a particle with mass @xmath0 inside a hard sphere with a time - dependent radius @xmath1 .
the potential energy function is zero if @xmath2 and infinite otherwise .
the schrdinger equation is then @xmath3 with the boundary condition @xmath4 .
the _ instantaneous _ energy eigenfunctions and eigenvalues are respectively @xmath5 @xmath6 ; @xmath7 and @xmath8 , where @xmath9 and @xmath10 are respectively spherical bessel functions and harmonics .
@xmath11 is the @xmath12 zero of the spherical bessel function of order @xmath13 , _
i.e. , _ @xmath14 .
it must be noted that all bessel functions with @xmath15 have a zero at the origin , but to have a non - zero wave function these zeros must be excluded . using the method of `` separation of variables '' for solving the partial differential equation ( [ eq : sch ] ) , we propose the solution @xmath16 where we have used the spherical symmetry of the hamiltonian . putting eq .
( [ eq : prosol ] ) into eq . ( [ eq : sch ] ) one gets @xmath17~.\end{aligned}\ ] ] the radial part of the proposed wave - function , @xmath18 , must be zero on the shell , thus the boundary conditions on @xmath19 are @xmath20 .
now , we follow @xcite to solve the eq .
( [ eq : radial_sch_1 ] ) . by defining a new coordinate @xmath21 we get @xmath22~,\end{aligned}\ ] ] where @xmath23 and moving boundary conditions are replaced by fixed - boundary ones ; @xmath24 . when the transformation @xmath25 \varphi(s , t ) ~.\end{aligned}\ ] ] is introduced in eq .
( [ eq : radial_sch_2 ] ) , one obtains @xmath26~,\end{aligned}\ ] ] for the uniform motion of the wall , _
i.e. , _ @xmath27 .
boundary conditions on @xmath28 are @xmath29 .
defining the new time variable @xmath30 as @xmath31 eq . ( [ eq : radial_sch_3 ] ) transforms to @xmath32~.\end{aligned}\ ] ] inserting @xmath33 in ( [ eq : radial_sch_4 ] ) , one gets @xmath34~.\end{aligned}\ ] ] by introducing new variable @xmath35 , we obtain @xmath36 the solutions of this equation are spherical bessel functions @xmath37 ~.\end{aligned}\ ] ] if the radial wave - function @xmath38 is finite at the origin , @xmath39 . the requirement that @xmath40 at @xmath41 means that @xmath42 can take on only those special values @xmath43 for the uniform change of the radius with velocity @xmath44 @xmath45 where @xmath46 is the initial radius
, one has @xmath47 by using equations ( [ eq : tau(t ) ] ) , ( [ eq : shell ] ) , ( [ eq : k_values ] ) , ( [ eq : dep_change_1 ] ) and ( [ eq : indep_change_1 ] ) one obtains @xmath48 \nonumber \\ & \times & j_{l } \left ( x_{l n } \frac{r}{l(t ) } \right ) ~,\end{aligned}\ ] ] for the radial part of the wave - function .
unknown coefficient @xmath49 is determined by the normalization condition @xmath50 where @xmath51 are the solutions of the schrdinger equation ( [ eq : sch ] ) for a particle in a spherical box with a wall in uniform motion and @xmath52 . using the orthogonality of the spherical bessel functions @xcite @xmath53 ^ 2 \delta_{nm}~,\end{aligned}\ ] ] one obtains @xmath54 \nonumber \\ & & \times j_{l } \left ( x_{l n } \frac{r}{l(t ) } \right ) y_{l m}(\theta , \phi ) \nonumber\\ & \equiv & \exp \left [ i \alpha \xi(t ) \left(\frac{r}{l(t ) } \right)^2 - i x^2_{l n } \frac{1 - 1/\xi(t)}{4 \alpha } \right ] \nonumber \\ & & \times u_{l n m}({\bf{r } } , t)~,\end{aligned}\ ] ] where we have introduced new dimensionless parameters @xmath55 and @xmath56 .
functions @xmath57 vanish at @xmath58 , remain normalized as the radius changes , and form a complete orthogonal set .
the general solution of eq .
( [ eq : sch ] ) is a superposition of functions ( [ eq : norm_waves ] ) @xmath59 with time - independent coefficients @xmath60 determined from the relation @xmath61 general solution can also be expanded in terms of instantaneous eigenfunctions as @xmath62 now with time - dependent coefficients @xmath63 determined from the relation @xmath64 using eqs .
( [ eq : coef_2 ] ) and ( [ eq : geral - sol_1 ] ) and the orthogonality of spherical harmonics , one finds @xmath65 i^*_{l^{\prime } n^{\prime } n^{\prime \prime}}(t , \alpha)~,\end{aligned}\ ] ] where @xmath66 this integral is not elementary and following the procedure of @xcite , can be reduced to a combination of terms involving the ferensel integrals and derivative of legendre polynomials .
the expectation value of the energy of the particle is obtained from @xmath67 if the particle is initially in an energy eigenstate , _
@xmath68 , then @xmath69 which is not an unexpected result as quantum numbers @xmath13 and @xmath70 do not change .
numerical computations are shown in figs . [
fig : b_coefficients ] and [ fig : energy_expectation ] for a particle that is initially in the first excited state with three fold degeneracy . in this case
we have @xmath71 for the ratio of energy expectation value to the instantaneous first excited state energy .
figure [ fig : b_coefficients ] shows the squares of energy eigenfunction expansion coefficients versus @xmath72 for three different contraction rates @xmath73 . for these values of @xmath73 , it was found that series ( [ eq : relation_coefs ] ) converges for the first ten terms .
[ fig : b_coefficients ] figure [ fig : energy_expectation ] shows the ratio of the expectation value of the energy to the energy the particle would have if it remained in the first excited state @xmath74 for the sphere in contraction . here
fifteen terms in eq .
( [ eq : energy_exp_2 ] ) leads to convergency . [
fig : energy_expectation ] we have plotted dimensionless radial probability density @xmath75 in fig . [ fig : wave_r ] for a particle initially in the state @xmath76 , against dimensionless position coordinate @xmath77 at dimensionless time coordinate @xmath78 and in fig .
[ fig : wave_t ] for a particle initially in the state ( a ) @xmath79 and ( b ) @xmath80 , against dimensionless time coordinate @xmath81 at dimensionless observation point @xmath82 ; where @xmath83 and @xmath84 . in our calculations @xmath85 and @xmath86 .
@xmath87 and @xmath88 are dimensionless classical flight times from the front and back edges of the sphere to the dimensionless observation point @xmath89 for a particle in the state @xmath90 .
[ fig : wave_r ] [ fig : wave_t ]
one can construct the propagator as follows @xmath91 where @xmath92 is the time evolution operator and we have used the fact that if the particle is in the sate @xmath93 at @xmath94 , it remains in that state as the wall moves , _
i.e. , _ @xmath95 .
now , we write this equation in the form @xmath96 where we have introduced the propagator as , @xmath97 ^ 2}~ \nonumber \\ & & \times \exp \left [ \frac{i\mu u}{2\hbar } \left ( \frac{r^2}{l(t ) } - \frac{{r^{\prime}}^2}{l(t^{\prime } ) } \right ) \right ] \nonumber \\ & & \times \exp \left [ -\frac{i \hbar}{2\mu } \frac{x^2_{l n}}{a}\left ( \frac{t}{l(t ) } - \frac{t^{\prime}}{l(t^{\prime } ) } \right ) \right ] \nonumber \\ & & \times j_{l } \left ( x_{l n } \frac{r}{l(t ) } \right)~j_{l } \left ( x_{l n } \frac{r^{\prime}}{l(t^{\prime } ) } \right)~ \nonumber \\ & & \times y_{l m}(\theta , \phi)~y^*_{l m}(\theta^{\prime } , \phi^{\prime})~.\end{aligned}\ ] ] one sees when @xmath98 , eq . ( [ eq : radial_sch_2 ] ) reduces to eq .
( 4 ) of @xcite , _ i.e. , _
@xmath98 corresponds to a particle in a 1d box with the left wall at @xmath99 and the right wall in uniform motion . in order to have the relation @xmath100 in 1d , we must write 1d propagator as @xmath101 preserving just the terms with @xmath98 , ( [ eq : propagator ] ) leads @xmath102 ^ 2}~ \nonumber \\ & & \times \exp \left [ \frac{i\mu u}{2\hbar } \left ( \frac{r^2}{l(t ) } - \frac{{r^{\prime}}^2}{l(t^{\prime } ) } \right ) \right ] \nonumber \\ & & \times \exp \left [ -\frac{i \hbar}{2\mu } \frac{x^2_{0n}}{a}\left ( \frac{t}{l(t ) } - \frac{t^{\prime}}{l(t^{\prime } ) } \right ) \right ] \nonumber \\ & & \times j_0 \left ( x_{0 n } \frac{r}{l(t ) } \right)~j_0 \left ( x_{0 n } \frac{r^{\prime}}{l(t^{\prime } ) } \right)~ \nonumber \\ & & \times \frac{1}{\sqrt{4\pi } } \frac{1}{\sqrt{4\pi}}~,\end{aligned}\ ] ] where we have used @xmath103 .
the first two bessel functions are @xmath104 thus @xmath105 and @xmath106 . using these in eq .
( [ eq : 1_1d_prop ] ) , we find @xmath107 \nonumber \\ & & \times \sum_{n=1}^{\infty } \exp \left [ \frac{i \hbar}{2 \mu } \frac{n^2 \pi^2}{u } \left ( \frac{1}{l(t ) } - \frac{1}{l(t^{\prime } ) } \right ) \right ] \nonumber \\ & & \times \sin \left ( n\pi \frac{r}{l(t ) } \right ) \sin \left ( n\pi \frac{r^{\prime}}{l(t^{\prime } ) } \right)~.\end{aligned}\ ] ] now from eq .
( [ eq : 1d-3d - prop ] ) we obtain @xmath108 \nonumber \\ & & \times \sum_{n=1}^{\infty } \exp \left [ \frac{i \hbar}{2 \mu } \frac{n^2 \pi^2}{u } \left ( \frac{1}{l(t ) } - \frac{1}{l(t^{\prime } ) } \right ) \right ] \nonumber \\ & & \times \sin \left ( n\pi \frac{x}{l(t ) } \right ) \sin \left ( n\pi \frac{x^{\prime}}{l(t^{\prime } ) } \right)~.\end{aligned}\ ] ] which is exactly eq . ( 32 ) of ref .
@xcite for the propagator of a particle in a 1d box .
this equation can be written in a compact form as a combination of @xmath109 functions @xcite .
in this letter we found solutions of the schrdinger equation for a particle confined in a hard spherical trap with a moving wall at constant velocity .
we see in solutions ( [ eq : norm_waves ] ) , except for the phase factor @xmath110 which has no coordinate dependence , a coordinate - dependent phase @xmath111 $ ] appears .
it has been shown that this factor leads to an effective quantum non - local interaction with the boundary : even though the particle is nowhere near the walls , it will be affected @xcite . from fig .
[ fig : b_coefficients ] , one sees that as the velocity of the wall increases , larger amounts of energy states other than the initial one , _
@xmath74 , are mixed in .
[ fig : energy_expectation ] shows that for rapid contraction , energy expectation value increases faster than the @xmath112 increase which would be obtained in a quasistatic contraction .
these results are in agreement with the ones of ref .
@xcite obtained for a particle in an infinite square well with one wall in uniform motion .
confinement of the particle to a smaller region leads to enhancement of the energy expectation value .
this can be explained by an application of the `` old quantum theory '' @xcite or by uncertainty relations @xcite . in the process of expansion , there are two characteristic times involved : @xmath113 , over which the parameters of the system change appreciably , and @xmath114 , representing the motion of the system itself . in our calculations , @xmath115 and @xmath116 .
figure [ fig : wave_r ] shows that for @xmath117 ( @xmath118 ) , the particle , initially in the state @xmath76 , will end up in the corresponding state of the expanded well .
this process characterizes an adiabatic one for which external conditions change gradually @xcite .
while , in the opposite limit , rapidly changing conditions prevent the system from adapting its configuration during the process , hence the probability density remains almost unchanged .
noticing fig .
[ fig : wave_t ] , one sees a quasi - classical behavior in the high - energy limit @xcite as the velocity of the wall increases .
a non - monotonous increasing behavior of the density is seen for @xmath119 only when @xmath120 , while for @xmath121 a non - monotonous decreasing behavior is seen irrespective of the wall velocity .
these results are in contrast to classical mechanics . the height of the first maximums decrease with @xmath44 .
the constructive interference with the reflected components from the wall for @xmath122 leads to this enhancement
. long time behavior of the density in the given observation point , is the same for all values of the wall velocity , which is not an unexpected result noticing the behavior of functions @xmath123 at long times .
propagator of the problem was derived using the spectral decomposition .
+ + * acknowledgment * the author would like to acknowledge two anonymous referees for valuable comments on an earlier draft of the paper . financial support from the university of qom is gratefully acknowledged .
makowski a. j. , _ j. phys . a : math .
* 25 * ( 1992 ) 3419 makowski a. j. and dembinski s. t. , _ phys .
a _ * 154 * ( 1991 ) 217 doescher s. w. and rice m. h. , _ am .
j. phys . _
* 37 * ( 1969 ) 1246 del campo a. , garca - caldern g. and muga j. g. , _ phys .
_ * 476 * ( 2009 ) 1 moshinsky m. , _ phys .
rev . _ * 88 * ( 1952 ) 625 del campo a. , muga j. g. and kleber m. , _ phys .
a _ * 77 * ( 2008 ) 013608 godoy s. , _ phys .
a _ * 65 * ( 2002 ) 42111 ; + del campo a. and muga j. g. , _ europhys .
_ * 74 * ( 2006 ) 965 ; + mousavi s. v. , _
j. phys . a : math
_ * 43 * ( 2010 ) 035304 godoy s. , _ phys .
rev . a _ * 67 * ( 2003 ) 012102 arfken g. b. and weber h. j. , _ mathematical methods for physicists _
( elsevier academic press , 2005 ) da luz m. g. e. and cheng b. k. , _ j. phys . a : math .
* 25 * ( 1992 ) l1043 grosche ch .
a _ * 182 * ( 1993 ) 28 dodonov v. v. and andreata m. a. , _ phys .
a _ , * 275 * ( 2000 ) 173 ; + greenberger d. m. , _ physica b _ * 151 * ( 1988 ) 374 ; + mousavi s. v. , quant - ph/1111.3962v2 preprint , 2011 pinder d. n. , _ am .
j. phys . _
* 58 * ( 1989 ) 54 wilhelm h. e. , _ j. phys . a : math .
* 16 * ( 1983 ) 2149 griffiths d. j. , _ introduction to quantum mechanics _
( prentice hall , new jersey ) 1994 godoy s. , _ physica b _ * 390 * ( 2007 ) 112 | exact solution of the schrdinger equation is given for a particle inside a hard sphere whose wall is moving with a constant velocity .
numerical computations are presented for both contracting and expanding spheres .
the propagator is constructed and compared with the propagator of a particle in an infinite square well with one wall in uniform motion . |
we are entering an era in which numerous experiments will search for astrophysical gravitational radiation .
ground - based detectors such as ligo ( barish 2000 ) , virgo ( fidecaro et al . 1997 ) , geo600 ( schilling 1998 ) , tama ( ando et al .
2002 ) , and others focus on relatively high frequencies @xmath1 hz . these frequencies are appropriate for the final inspiral and merger of binaries with total masses @xmath2 .
space - based missions such as the laser interferometer space antenna ( lisa ; see , e.g. , danzmann 2000 ) will complement this frequency range by focusing on comparatively low frequencies , @xmath3 hz . in this range
there are a number of guaranteed " sources in the form of main sequence and white dwarf binaries , but there is much more uncertainty about the prevalence and properties of double compact binaries containing neutron stars and black holes .
such binaries are of great interest for many reasons , such as their use as probes of strong gravity and their potential to illuminate other aspects of astronomy , including the evolution of galaxies . the existence of a new class of double compact binaries is suggested by two recent lines of evidence .
first , observations of the kinematics of the central regions of some globular clusters suggests that black holes with @xmath4 may exist in their cores ( e.g. , gebhardt et al .
2000 ; damico et al .
second , x - ray observations have revealed variable , unusually high flux point sources in a number of galaxies ( e.g. , fabbiano 1989 ; fabbiano , schweizer , & mackie 1997 ; colbert & mushotzky 1999 ; zezas , georgantopoulos , & ward 1999 ; kaaret et al . 2001 ; matsumoto et al . 2001 ; fabbiano , zezas , & murray 2001 ; see colbert & ptak 2002 for a catalog of objects ) .
the rapid and strong variability of these sources indicates that they are black holes .
if their flux is not strongly beamed , then for their luminosities to be below the eddington luminosity @xmath5 erg s@xmath6 , at which radiation forces balance gravity , the masses of the brightest sources must be at least @xmath7 ( e.g. , matsumoto et al .
in addition , their off - center positions in their host galaxies indicate that in many cases the masses can not be substantially greater than @xmath8 , otherwise dynamical friction would cause the black holes to sink rapidly to the dynamical center ( kaaret et al .
2001 ; tremaine , ostriker , & spitzer 1975 ) .
recent optical observations suggest that for at least some of the sources , the x - rays may not be beamed strongly , based on recombination emission from surrounding nebulae ( pakull & merioni 2002 ) .
this implies that the total luminosity is indeed high , so that if the sources are sub - eddington ( but see begelman 2002 ) there may therefore be a significant number of @xmath9 black holes in the universe .
currently , these sources are preferentially in star forming regions or in starburst galaxies ( matsumoto et al .
2001 ; liu & bregman 2001 ) .
a number of these objects are associated with dense young stellar clusters ( e.g. , matsumoto et al .
combined with the kinematic evidence in globular clusters , this implies that black holes of this sort undergo frequent dynamical encounters .
it has been suggested that the ultraluminous x - ray sources originated due to dynamical encounters in globular clusters , then merged with their host galaxies ( miller & hamilton 2002a , b ) .
other proposals , if these objects are @xmath10 black holes instead of being strongly beamed ( king et al .
2001 ; king 2002 ; krding , falcke , & markoff 2002 ) or super - eddington by a large factor ( begelman 2002 ) , are that they form as a result of a core collapse of a massive young star cluster ( matsushita et al .
2000 ; ebisuzaki et al .
2001 ) or as remnants of population iii stars ( madau & rees 2001 ) . here
we consider the implications for gravitational wave emission if these sources are indeed intermediate - mass black holes .
independent of their detailed origin , black holes in dense stellar environments are promising sources of gravitational radiation due to the diversity of dynamical interactions that are possible ( see also benacquista 1999 , 2002a , b ; benacquista , portegies zwart , & rasio 2001 ) .
we will focus mainly on black holes in globular clusters , but will also address signals that may be evident from young stellar clusters , regardless of whether the black hole formed there or elsewhere . in 2 we give general arguments about the properties of these sources , based on the ages of globulars and the number and types of secondaries . in
3 we estimate the rates of encounters , and in 4 we present calculations of the signal to noise ratios expected for lisa from galactic globulars and from globulars in the virgo cluster , and for coalescence of black holes as observed with the ligo - ii detector . in 5
we discuss what information could be gleaned from a reasonably high signal to noise detection .
we show that both pericenter precession and lense - thirring precession are potentially measurable .
these effects may allow preliminary mapping of the spacetime around black holes and could even yield an independent measure of the distance to the virgo cluster .
we present our conclusions and discuss the uncertainties in our estimates in 6 .
here we discuss the expected nature of intermediate - mass black holes if these are grown in dense clusters .
we begin by summarizing the dynamics of dense stellar clusters and then discuss the ways in which intermediate - mass black holes can form and grow .
we then derive the approximate distributions of spin parameter and eccentricity that we expect for these sources , with implications for detectability .
the young super star clusters now being discovered in many galaxies with active star formation ( e.g. , origlia et al .
2001 ) have typical estimated masses @xmath11 and half - mass radii 10 pc ( origlia et al . 2001 ) .
the central number density is difficult to estimate , but if we take as a guide the densest young clusters in the vicinity of the milky way the central density could be up to @xmath12 pc@xmath13 ( see , e.g. , massey & hunter 1998 for observations of the dense r136 region ) .
clusters with ages less than a few tens of millions of years have a large number of o and b stars ; older than this , and those stars have evolved to compact remnants .
globular clusters are much older , @xmath14 yr . all stars with initial masses @xmath15 have evolved off the main sequence , meaning that the most massive objects present are compact remnants such as black holes , neutron stars , and massive white dwarfs .
there are also a small number of @xmath16 blue stragglers , which are main sequence stars more massive than the ordinary main - sequence turnoff that may have been rejuvenated by collisions ( lombardi et al .
2002 ; see , e.g. , hurley & shara 2002 for a recent discussion of these and other interactions in clusters ) .
a typical globular has @xmath12 stars of average mass @xmath17 , and a central number density in the range @xmath18 pc@xmath13 ( pryor & meylan 1993 ) .
roughly 40% of the globulars surrounding the milky way have @xmath19 pc@xmath13 ( pryor & meylan 1993 ) , with a density contrast between the core and the half - mass radius that formally poise them on the edge of core collapse .
it is thought that three - body interactions of primordial binaries with field stars heat the cluster , delaying collapse much longer than otherwise possible ( e.g. , goodman & hut 1989 ; hut , mcmillan , & romani 1992 ; sigurdsson & phinney 1995 ) .
a productive analogy with thermodynamics , backed up by simulations , has shown that in clusters with stars of different masses the cluster tries to evolve towards thermodynamic equilibrium , in the sense that the kinetic energies of each of the component stars are drawn from the same distribution ( see , e.g. , binney & tremaine 1987 ) .
this means that the typical speed of an object of mass @xmath20 is proportional to @xmath21 .
this causes more massive objects to sink to the core of the cluster , a process called mass segregation ( e.g. , farouki & salpeter 1982 ) .
the typical timescale for such sinking is given by @xmath22 , where for a cluster of @xmath23 stars and dynamical crossing time @xmath24 the core relaxation timescale is @xmath25 ( see , e.g. , binney & tremaine 1987 ) .
typical globulars have @xmath26 yr ( pryor & meylan 1993 ) .
this means that even small black holes with @xmath27 will sink quickly to the center of a cluster , and if an intermediate - mass black hole with @xmath28 is present it will rarely be too far from the center even in a young cluster .
however , the short relaxation time in the core and wandering of black holes of this mass due to dynamical interactions mean that there is rapid refilling of stellar orbits that interact with the black hole .
there is therefore no loss cone " problem , unlike the situation with supermassive black holes in galactic nuclei ( frank & rees 1976 ; see milosavljevi & merritt 2001 for a recent discussion ) . as a result of mass segregation , even though the absolute number of compact objects is much less than that of main sequence stars , their number density in the core can be comparable to or in excess of that of main sequence stars for core densities @xmath29 pc@xmath13 ( e.g. , sigurdsson & phinney 1995 ) .
in addition , since binaries typically have more mass than single stars , they also tend to sink to the center .
this means that three - body interactions of binaries with single stars , and possibly binary - binary interactions , can be dynamically important .
qualitatively , three - body interactions can be understood using heggie s rule ( heggie 1975 ) that hard binaries harden and soft binaries soften .
that is , if the initial binary has a binding energy greater than the typical kinetic energy of a field star ( i.e. , the binary is hard ) then a typical encounter with a field star will cause the binary to harden further .
a soft binary will instead be softened gradually until it becomes unbound .
exchanges can also occur , in which the interloper object becomes a member of the final binary .
numerous simulations show that in a strong encounter the most likely occurrence is that the final binary is composed of the two most massive of the three stars that originally interacted ( e.g. , sigurdsson & phinney 1993 ; heggie , hut , & mcmillan 1996 ) .
there has long been interest in whether hardening from three - body encounters might bring a pair of black holes close enough together that they would merge via gravitational radiation while still in the cluster .
however , each hardening delivers a recoil kick to both the binary and the single star , and if the recoil of the binary exceeds the @xmath30 km s@xmath6 escape speed from the core of a typical dense cluster ( webbink 1985 ) then the binary will be ejected before it can merge . several simulations ( kulkarni , hut , & mcmillan 1993 ;
sigurdsson & hernquist 1993 ; portegies zwart & mcmillan 2000 ) have shown that if attention is restricted to newtonian three - body encounters of @xmath31 black holes then few if any mergers will occur inside the cluster .
there may , however , be a number of mergers that happen outside the cluster after recoil ( portegies zwart & mcmillan 2000 ) .
recently , miller & hamilton ( 2002a ) showed that merger may happen inside the cluster if a black hole starts out with a greater mass , @xmath32 . in this case , recoil is much less significant and gravitational radiation predominates .
they proposed this as a mechanism for the formation and growth of intermediate - mass black holes .
even if only lighter black holes form initially , secular resonances in hierarchical triple systems formed in binary - binary encounters ( the kozai resonance : kozai 1962 ; lidov & ziglin 1976 ) may increase the eccentricity of the inner binary enough that it can merge by gravitational radiation without being ejected by three - body recoil ( miller & hamilton 2002b ) .
it is also possible that in some three - body encounters , two of the black holes will pass close enough to each other to coalesce rapidly .
this increases optimism that dense clusters may be the seat of relatively frequent mergers of compact objects .
there are , however , a number of questions still remaining .
one of these has to do with recoil due to asymmetric emission of gravitational radiation during inspiral .
calculations have yet to be done for strong gravity , but various analyses using newtonian and post - newtonian formalisms suggest that the recoil velocity could be from a few kilometers per second up to @xmath33 km s@xmath6 , depending on the mass ratio and how close the black holes get before the final plunge ( peres 1962 ; bekenstein 1973 ; fitchett 1983 ; fitchett & detweiler 1984 ; redmount & rees 1989 ; wiseman 1992 ) . for slowly - rotating black holes with mass ratios @xmath34 , the prime focus of this paper ,
the best estimates are that the recoil speed will be well below the 50 km s@xmath6 escape speed from the core ( fitchett 1983 ; fitchett & detweiler 1984 ; wiseman 1992 ) .
we will therefore assume that if an intermediate - mass black hole forms in a dense cluster it will remain in the core through multiple mergers .
black holes that have grown by the accumulation of smaller objects in a dense stellar cluster are likely to have a different rotation parameter than either stellar - mass or supermassive black holes .
stellar - mass black holes acquire some spin at birth .
subsequent accretion adds relatively little mass or angular momentum to the black hole , either because not much mass is available ( as in low - mass x - ray binaries , where the companion is usually less than 10% as massive as the black hole ; e.g. , verbunt 1993 ) or because the duration of the accretion phase is short ( as in high - mass x - ray binaries , with durations @xmath35 yr implying a total mass transfer less than @xmath36 : verbunt 1993 ) . as a result ,
the spin parameter @xmath37 , where @xmath38 is the speed of light , @xmath39 is newton s gravitational constant , and @xmath40 is the angular momentum , of stellar - mass black holes is expected to be close to its birth value .
compelling evidence of significant spin in stellar - mass black holes has emerged from study of quasi - periodic brightness oscillations ( strohmayer 2001a , b ; miller et al . 2001 ) , and spectral studies are also suggestive of high spin ( from line profiles , miller et al . 2002 ; and earlier more model - dependent fits to continuum emission , zhang , cui , & chen 1997 ) .
for supermassive black holes , the spin parameter depends on what fraction of its mass was provided by an accretion disk with a fixed orientation .
if the fraction is close to unity then the spin parameter could also be close to unity .
some evidence of such rapidly spinning supermassive black holes is emerging in the form of extremely broad fe k@xmath41 lines from a few active galactic nuclei ( e.g. , iwasawa et al .
1996 , dabrowski et al . 1997 ; wilms et al .
2001 ) .
in contrast , black holes grown by the capture of stars or compact objects in dense stellar clusters undergo a damped random walk in the evolution of their spins after the memory of the initial spin is lost .
the damping is because retrograde orbits become unstable at larger specific angular momentum than do prograde orbits , so that it is easier to decrease than to increase the spin parameter of the black hole . assuming a roughly isotropic distribution of stellar velocities in the cluster core
, the final inspiral and deposition of angular momentum will happen at random angles to the previous spin axis of the black hole .
there is evidence of net rotation in many globular clusters ( e.g. , barmby et al .
2002 ) , so that encounters may not be precisely isotropic , but this is likely to be a small effect because relaxation will drive the core distribution towards isotropy ( e.g. , einsel & spurzem 1999 ; kim et al .
2002 for fokker - planck treatments of the evolution of rotating systems ) .
if the angular momentum of the large black hole is much greater than the orbital angular momentum of a small black hole that falls into it , then the total angular momentum is roughly constant , so that @xmath42 ( hughes & blandford 2002 ) .
in contrast , if the angular momentum of the larger black hole is sufficiently small , then the orbital angular momentum of the smaller black hole can make a significant difference .
following hughes & blandford ( 2002 ) , let us define the orbital inclination @xmath43 using @xmath44 where @xmath45 is the angular momentum parallel to the spin axis of the more massive black hole and @xmath46 is the carter constant .
if the initial spin angular momentum of the massive black hole is @xmath40 , then merger with a black hole of mass @xmath47 with an orbital inclination @xmath43 and specific orbital angular momentum @xmath48 will produce a black hole of mass @xmath49 and angular momentum @xmath50 this assumes that the amount of mass - energy and angular momentum radiated away during the merger is small .
hughes & blandford ( 2002 ) show that to good accuracy the angular momentum of a test particle in the last stable circular orbit at angle @xmath43 to the prograde equatorial direction is approximately @xmath51 where @xmath52 and @xmath53 are , respectively , the specific angular momentum of a prograde equatorial and retrograde equatorial particle at the last stable orbit .
when @xmath54 this reduces to @xmath55\ ; , \ ] ] assume the probability distribution of inclinations is @xmath56 ( e.g. , @xmath57 for an isotropic distribution ) .
if the probability distribution for @xmath58 given @xmath20 , @xmath59 , is stationary in form , then we have @xmath60 q(\cos\iota)\,d(\cos\iota)\ ; , \ ] ] where @xmath61 is the change in dimensionless angular momentum caused by the accretion of an object of mass @xmath62 at an orbital inclination @xmath43 ; that is , such an object will change the dimensionless angular momentum from @xmath63 to @xmath58 .
approximate solutions to this equation may be sought in the form @xmath64 $ ] . for isotropic encounters ,
a fit to the numerical results displayed in figure 1 gives @xmath65 and @xmath66 .
figure 1 shows the distribution of spin parameters for a number of simulated black holes that started at @xmath67 and accreted objects of mass @xmath62 ( e.g. , @xmath68 and @xmath31 ) . for intermediate - mass black holes ,
the expected dimensionless angular momentum is on the order of a few tenths , approximately independent of the initial spin if @xmath69 .
the expected eccentricity for a black hole binary depends on the way in which it was formed and on the orbital frequency at which it is observed .
we will discuss each of the evolutionary paths in turn , but in brief we find that at the high frequencies observable with ground - based detectors the orbits will have circularized to high accuracy , whereas at the lower frequencies of space - based detectors the eccentricity is expected to be significant for the majority of the detectable period .
first , consider formation via three - body encounters .
as long as the time to merger is longer than the time to the next encounter , the eccentricity will be reset by each interaction . during this period
there will be relatively little gravitational radiation emitted , as the merger time is typically long . at some point
, however , an encounter leaves the binary with a relatively short merger time and thus the binary goes into uninterrupted two - body inspiral .
the precise distribution of eccentricities at the start of this phase depends on the details of the three - body interactions , but preliminary results ( gultekin & miller , in preparation ) suggest that immediately after the last three - body encounter the eccentricity tends to be fairly high , @xmath70 .
this is because the time to merger decreases rapidly with increasing eccentricity , so a chance fluctuation up to high eccentricity will allow a binary to merge before the next encounter . for moderate to high eccentricities the gravitational radiation merger time is @xmath71 ( peters 1964 ) , where @xmath20 is the total mass and @xmath72 is the reduced mass of the binary . for @xmath73
this is @xmath74 yr , where @xmath75 is the pericenter distance . for a typical merger time , after the last encounter , of a few hundred thousand to a few million years , the pericenter distance is therefore @xmath76 cm , which is hundreds to thousands of gravitational radii for an intermediate - mass black hole .
although the full period of an orbit could be relatively high ( @xmath77 s ) , the majority of the power is emitted at a frequency of the order of the angular frequency at pericenter ( see 4.1 ) , which could be several times higher than the fundamental orbital frequency .
the peak frequency of gravitational waves is then typically on the order of @xmath78 hz for the majority of the inspiral .
this is what would most likely be seen from sources in our galactic globular system , and hence the eccentricity would be expected to be high , perhaps typically @xmath70 . in later stages of inspiral
, however , the circularizing influence of gravitational radiation reduces the eccentricity substantially . a simple way to estimate the eccentricity uses the constant @xmath79 found by peters ( 1964 ) for
the lowest - order quadrupole radiation ( see glampedakis & kennefick 2002 and glampedakis , hughes , & kennefick 2002 for a discussion of how constant this expression is when higher - order terms are included )
. disregarding factors of order unity , this means that @xmath80 is roughly constant .
thus , @xmath81 remains approximately fixed as long as @xmath82 is close to unity , after which point it drops significantly . in the first stage , where @xmath83 , @xmath84 so that @xmath85 .
note that in this stage the emission is strongly dominated by the time near pericenter .
thus , gravitational waves are emitted for a few cycles near closest approach , then little emission occurs until the next pericenter passage . as a result ,
the peak frequency of the gravitational wave emission is approximately the orbital frequency at pericenter , which is nearly constant during this phase of inspiral . in the second stage , where @xmath86 and thus @xmath87 , we have @xmath88 . here
the frequency of gravitational waves is roughly twice the mean orbital frequency , or @xmath89 , so @xmath90 . in practice
, the transition between relatively rapid change in eccentricity and relatively slow change happens at surprisingly high eccentricity .
for example , suppose the binary initially has a merger time of @xmath91 yr and an eccentricity of 0.9 . at @xmath92 yr from merger ,
the eccentricity would be @xmath93 , at @xmath94 yr from merger @xmath95 , and at @xmath96 yr @xmath97 .
if the initial eccentricity is instead 0.99 , then at @xmath92 yr @xmath98 , at @xmath94 yr @xmath99 , and at @xmath96 yr @xmath100 .
the net result is that almost all the sources of this type in the lisa band are likely to have appreciable eccentricity .
in contrast , high frequency sources detectable by ground - based instruments will have much lower eccentricities , @xmath101 typically .
this means that templates based on circular orbits will be adequate for many purposes ( see martel & poisson 1999 for a discussion of the amount of signal to noise lost by nonoptimal signal processing ) .
a second formation scenario involves kozai resonances in hierarchical triple systems formed by binary - binary encounters ( miller & hamilton 2002b ) . here
the eccentricities could be much higher and the pericenter distances much closer than in the three - body scenario .
however , the eccentricities will certainly not be higher than what would produce a merger within a single orbital period of the inner binary , because the eccentricity is increased gradually and such a short merger time would terminate the increase . for a merger time of @xmath102 yr and two @xmath31 black holes , this implies a pericenter distance of @xmath103 cm . even in this very extreme case , by the time the binary reached the @xmath104 cm separation necessary to bring the frequency into the ligo - ii detector band the eccentricity would be @xmath105 . in more realistic circumstances the pericenter distance would be more than @xmath106 cm and the binary would again be nearly circular when it entered the frequency range of ground - based detectors .
the third formation scenario involves direct two - body capture by emission of gravitational radiation .
if this happens with a black hole of mass @xmath20 and one of mass @xmath47 , then the distance of closest approach needed to cause a binary of initial relative velocity @xmath107 to become bound is ( see quinlan & shapiro 1987 ) @xmath108 where @xmath109 and @xmath110 . here
@xmath111 is the speed at infinity in units of @xmath91 cm s@xmath6 . for stellar - mass or intermediate - mass black holes ,
this distance is so much larger than the radius @xmath112 cm of the last stable orbit that , again , when the frequency is high enough to detect with ground - based instruments the orbit will have circularized to @xmath113 .
note that because @xmath114 whereas the horizon radius @xmath115 , for supermassive black holes there is a point at which in order to become bound the smaller object would have to pass inside the horizon , so that a direct capture occurs .
this happens at @xmath116 , and means that above this mass one will not observe multiple orbits .
if two - body capture happens during a three - body interaction , the criterion is that the capture must occur before a slight perturbation from the third object can deflect the binary from its highly eccentric orbit .
this means that the merger time must be less than @xmath117 yr . for such short merger times there is a significant probability that the pericenter distance will be comparable to the radius of the last stable orbit .
this is in part because in the gravitational focusing regime the cross section for a closest approach @xmath81 scales as @xmath81 and not @xmath118 .
however , we expect these to form a minor fraction of the coalescences observed with ground - based detectors . in summary ,
all of the scenarios above predict that when a binary is orbiting with high frequency its eccentricity will be small enough for the use of circular orbit templates . at the frequencies detectable with lisa
, however , we expect significant eccentricities to be typical .
what is the rate at which intermediate - mass black holes are expected to merge with less massive compact objects ?
the actual rate in a given cluster is set by the frequency of encounters , which depends on number densities and effective cross sections .
however , if this rate is too high then it is self - limiting .
one upper limit is set by the supply of smaller objects ; this supply can not be exhausted much faster than the current age of the cluster .
another upper limit is set by the mass of the large black hole ; the characteristic growth time of the black hole can not be much shorter than the current cluster age .
we now consider these in turn .
the encounter rate includes contributions from two - body capture by gravitational radiation and from three - body encounters .
consider first two - body interactions .
if the smaller objects have number densities @xmath119 pc@xmath13 , and if they are in thermal equilibrium with @xmath17 main sequence stars with a velocity dispersion of @xmath120 cm s@xmath6 ( so that the velocity dispersion of the black holes is @xmath121 cm s@xmath6 ) , then the rate of two - body capture by a given large black hole is @xmath122 ( e.g. , quinlan & shapiro 1987 ; miller & hamilton 2002a )
. the typical rate can therefore be @xmath123 yr@xmath6 , depending on the cluster parameters and black hole masses .
now consider three - body interactions .
if the large black hole is initially solitary , it will acquire a companion when it interacts strongly with an existing binary , where strongly " means roughly a closest approach less than the semimajor axis @xmath124 of the binary .
the interactions are dominated by gravitational focusing , so the cross section for a close approach within distance @xmath124 is roughly @xmath125 , where @xmath126 cm s@xmath6 is the velocity at infinity . for a moderately hard binary with @xmath124=1 au ( and thus an orbital velocity of 30 km s@xmath6 ) , this cross section is about @xmath127 @xmath128 .
if 10% of stars are in binaries , their number density is about @xmath129 cm@xmath13 , so the interaction time is @xmath131=5\times 10^{12}v_6m_{100}^{-1}n_6^{-1}$ ] s. if the relative velocity of binaries with large black holes is somewhat smaller than the main sequence velocity dispersion ( due to the higher mass of binaries ) , this is typically @xmath132 yr .
once the massive black hole is in a binary , the question is how long it will take for subsequent interactions to harden it so that it merges because of gravitational radiation .
since there is not a significant loss cone for objects of this mass , the black hole can now interact with all the stars ( density @xmath133 cm@xmath13 ) , so at least at first the interaction rate is about 1 per @xmath134 yr . the fractional hardening per interaction is small because of the large mass ratio , but the eccentricity can increase substantially ( quinlan 1996 ) . depending on how the eccentricity evolves , merger within a few hundred interactions is likely to be typical . for a @xmath135 black hole
this will give rates of @xmath136 mergers per year , with lower rates for less massive black holes .
thus , the overall rate of interactions is likely to be dominated by three - body encounters , but the merger rate could have a significant contribution from two - body contributions as well because such an interaction leads to a direct merger .
a model - independent upper limit on the rate of mergers is set by the requirement that the characteristic timescale on which the smaller objects are removed from the population can not be much shorter than the age of the cluster , otherwise the population would be reduced rapidly .
for example , suppose the central black hole has a mass of @xmath137 and we consider its interactions with neutron stars , of which we assume there are @xmath96 currently in the cluster ( and at most a few hundred in the core ) . for a @xmath138 yr old cluster
the interaction rate is therefore limited to not much more than one per @xmath139 yr , and could be less . of course , if other processes are more important then the rate could be much less than these values , but the rate can not be much more .
for example , if the process of merging with one neutron star has along the way caused the ejection of 100 neutron stars ( e.g. , through three - body recoil ) , the maximum rate drops to one per @xmath106 yr .
the rate is therefore @xmath140 , where @xmath141 is the current number of small objects of a given type in the cluster and @xmath142 is the current cluster age .
similarly , the timescale of increase in the mass of the central black hole can not be much shorter than the cluster age , otherwise the massive black hole would acquire mass rapidly and grow until it exhausted its fuel . in the example
above , if neutron stars are accreted faster than one per @xmath143 yr , the black hole mass will grow well past its current @xmath137 in a hubble time .
thus , @xmath144 , where @xmath62 is the mass of the objects accreted .
this provides a guide to the most common expected interactions between an intermediate - mass black hole and other compact objects . because exchange interactions tend to favor more massive objects , the companion to an intermediate - mass black hole is likely to be among the more massive objects present in abundance in the core of the cluster .
if there are enough stellar - mass black holes ( say , tens to hundreds ) , then in dense clusters these may dominate the actual merger rate even if most of the encounters are with other types of stars . if the number of stellar - mass black holes is smaller , then neutron stars ( with perhaps @xmath145 in a cluster ; grindlay et al .
2001 ) or massive white dwarfs ( which could constitute several percent of the number of stars , or up to @xmath146 in a cluster ) may dominate the interactions . as we will see in 4 , mergers between two black holes will provide most of the signal for high - frequency ground - based detectors , whereas the more frequent interactions between a neutron star or white dwarf and an intermediate - mass black hole will likely provide most of the signal for lower - frequency space - based instruments . note that for frequencies in the majority of the lisa band white dwarfs act as point masses .
tidal disruption occurs at the roche separation @xmath147 , where @xmath148 cm is the radius of the white dwarf .
this implies a gravitational wave frequency @xmath149 hz independent of the mass of the black hole . below this frequency
white dwarfs can make a clean contribution to gravitational wave signals in the lisa band .
the instantaneous rate of mergers is governed by @xmath150 , but as the supply is decreased and the black hole mass is increased the encounter rate can change in response .
in addition to simply decreasing the number of smaller objects with which the large black hole interacts , three - body effects inject energy into the population that tend to increase the scale height of that population .
the compact objects that interact with the black hole will settle back via interactions with normal stars , but as the black hole mass increases the rate of energy injection per object increases .
this is primarily because the fractional change in binding energy per encounter scales roughly as @xmath151 ( quinlan 1996 ) .
the binding energy released before subsequent encounters actually eject the interloper object therefore scales roughly as @xmath20 .
thus , if the initial encounter rate is high it will decrease until it matches approximately the characteristic time of depletion and of mass increase . for ground - based detectors , which can see just the last few seconds at most of a merger
, it is only the overall rate that matters .
in contrast , for space - based detectors one must also know the distribution of time to merger for a given class of system . to estimate this ,
suppose that for a given type of secondaries one can establish the typical time @xmath152 between mergers .
if the probability of a given system being a time @xmath153 away from merger is given by a poisson distribution , then @xmath154 .
for example , consider a @xmath135 black hole interacting with a population of @xmath96 neutron stars of mass @xmath155 .
the maximum rate of interaction is roughly one per @xmath139 yr ; if this rate is realized in a particular cluster , the probability of that cluster having a system less than @xmath91 yr away from merger is 10% . the implied signal to noise ratio for lisa , however , depends on the eccentricity of the system , which we examine in the next section .
note that a young dense cluster of age @xmath156 yr has much less serious constraints placed on interaction rates by supply limits and mass growth timescales . however , the maximum rate is still limited by the encounter frequency itself . within @xmath157
years a massive young cluster may be able to evolve to high density in the core , without having yet ejected many compact objects by three - body processes ( portegies zwart & mcmillan 2000 ) . in principle , therefore , some young clusters could have interaction rates an order of magnitude larger than those for globulars . to be conservative , however , we will assume that the overall rate of mergers or inspirals from young clusters is at most comparable to that for globulars .
suppose that a binary of eccentricity @xmath82 and orbital period @xmath158 is a distance @xmath159 from us , and that in a coordinate frame in which the center of mass of the binary is at the origin we are in a spherical polar direction @xmath160 , where the orbit of the binary is in the plane @xmath161 .
following the notation and development of pierro et al .
( 2001 ) , the dimensionless metric perturbations in the @xmath162 and @xmath163 polarizations are @xmath164\ , , \\
h_+&={1\over{\sqrt{2}}}\left\{{3+\cos 2\theta\over{4}}\left [ 2h_{xy}\sin 2\phi+(h_{xx}-h_{yy})\cos 2\phi\right]- { 1-\cos 2\theta\over{4}}(h_{xx}+h_{yy})\right\}\ ; .\\ \end{array}\ ] ] in the adiabatic approximation ( i.e. , the assumption that the orbital parameters do not change significantly in a single orbit ) , the metric components are expressible as sums over harmonics ( peters & mathews 1963 ; peters 1964 ; pierro et al .
2001 ) @xmath165 where the harmonic components are @xmath166\ ; , \\ h_{x - y}^{(n)}&=2h_0n\left\{j_{n-2}(ne)-j_{n+2}(ne)-2e\left[j_{n-1}(ne)- j_{n+1}(ne)\right]+(2/n)j_n(ne)\right\}\ ; , \\ h_{x+y}^{(n)}&=-4h_0j_n(ne)\ ;
.\\ \end{array}\ ] ] here the @xmath167 are bessel functions and the prefactor is @xmath168 where @xmath169 is the fractional mass difference and @xmath170 $ ] . for a circular orbit ,
the power is all at the second harmonic ( @xmath171 ) and if @xmath172 the angle - averaged dimensionless strain is @xmath173 the harmonic of peak amplitude is @xmath174 ( see pierro et al .
2001 ) .
the majority of the inspiral is detectable only with low - frequency instruments such as lisa .
signal to noise estimates for lisa are complicated by the expectation of a significant background of astrophysical sources , especially white dwarf binaries ( phinney 2001 ) .
as the detector acquires data and frequency resolution becomes better due to the longer observational baseline , individual sources can be identified .
they can therefore be removed from the data stream , so that the unresolved background will effectively diminish in strength . however , until that point the true signal to noise is less than what would be computed from the instrumental curve alone , particularly for @xmath175 hz , where the white dwarf background is expected to make its strongest contribution .
there are also numerous other sources , especially coalescing supermassive black holes , that may produce a strong astrophysical background over much of the lisa frequency range ( e.g. , phinney 2001 ; hughes 2002 ) . as with the white dwarf sources ,
in the course of the lifetime of the experiment such individual sources can be accounted for in analysis , so that the effective sensitivity of lisa to other sources will improve with time . for the estimates in this section ,
we consider only the expected instrumental noise curve ( kindly provided by r. stebbins ) .
we also consider only the far - field inspiral . for inspiral effects at smaller distances , including
the effects of black hole spin , see , e.g. , finn & thorne ( 2000 ) and hughes ( 2001 ) . as is evident from 4.1
, the detectability of a signal depends strongly on its frequency distribution and amplitude , because of the strong frequency dependence of the detector sensitivity ( e.g. , danzmann 2000 ) .
in particular , the harmonic content of the signal is important ; orbits with high eccentricities have important contributions at frequencies many times the orbital frequency .
for example , an orbital frequency of @xmath176 hz is far below the most sensitive range of lisa , but for @xmath177 most of the power is in the @xmath178 hz range where the s / n=5 threshold in a one year observation is a dimensionless strain of @xmath179 . if the time to merger is more than about @xmath91 yr , then the signal to noise ratio from a source at a fixed distance in a year s integration with lisa depends on the merger time but is insensitive to other parameters such as the individual masses or the eccentricity . to show this , note that @xmath180 , where @xmath181 is the luminosity and @xmath182 is the frequency , and that at the low - frequency end the strain that gives a signal to noise ratio of 5 in a one year integration is @xmath183 for frequencies less than @xmath184 hz@xmath6 ( r. stebbins , personal communication ) . if @xmath172 and @xmath185 , then the merger time is @xmath186 and the frequency of peak emission is @xmath187 . for simplicity ,
assume that all the power in the signal is concentrated at @xmath188
. then @xmath189 the weak dependence on the masses and eccentricity arises because of competing effects .
for example , at a fixed merger time a binary with a high eccentricity will have a large semimajor axis ( and thus low fundamental frequency ) , but the frequency peak is at a high harmonic .
a binary with a low eccentricity will have a small semimajor axis and thus high fundamental frequency , but the frequency peak is at a low harmonic . in both cases
the frequency peak is about the same . at merger times @xmath190 yr , however , the signal to noise is larger for more circular binaries because the frequency peak is greater than or about the optimal frequency @xmath191 hz for lisa , and the signal to noise is greater when the signal is concentrated into few frequency bins .
the total signal to noise is the square root of the sum of the squares of the signal to noise ratios for each of the harmonics . to compute the signal to noise we therefore need to ( 1 ) pick an eccentricity and merger time , ( 2 ) calculate the frequency distribution and dimensionless strain amplitudes for the harmonics , ( 3 ) take the ratio with the expected lisa sensitivity at each frequency , and ( 4 ) sum these in quadrature for the final signal to noise .
the computation of a cumulative signal to noise plot thus requires assumptions about the distribution of eccentricities for a given merger time , although from the calculation above this only makes a significant difference if @xmath190 yr . as a sample calculation ,
let us assume that for large merger time the probability distribution of eccentricities is the thermal distribution @xmath192 that emerges from close interactions of three equal - mass objects ( heggie 1975 ) .
we expect that this will underestimate the eccentricity for large merger times , where the eccentricity may be pushed to higher values by many small interactions ( quinlan 1996 ) . at merger times
much less than the typical time between encounters , the eccentricity will decrease due to gravitational radiation .
as mentioned earlier , peters ( 1964 ) showed that to lowest order the eccentricity and semimajor axis evolve so as to keep the expression @xmath193 constant . combining this with the expression for merger time , we take as an approximation that when @xmath194 , the eccentricity is diminished by @xmath195 .
we assume that the merger time is given by a poisson distribution , @xmath196 , where @xmath152 is the typical time between mergers as defined above . given the limits of growth of the black hole and consumption of the secondaries , we use @xmath197 yr for @xmath198 and @xmath199 .
this is the minimum value of @xmath152 allowed , so if the time is greater the contours are moved to the left in figure 2 . in all cases , we fix the distance to the globular at 10 kpc , a representative distance to the globulars around our galaxy ( actual distances vary from 2.2 kpc to 122 kpc , with roughly 55% being less than 10 kpc distant ; see harris 1996 ) .
-0.3truein if these curves are roughly representative , it suggests that globulars of at least moderate density will have a significant probability of containing a high signal to noise gravitational wave source , especially if @xmath135 black holes are common . in a ten - year lisa integration the signal to noise ratio would be greater than 10 for @xmath200% of the galactic globulars that have @xmath135 black holes and a significant supply of @xmath201 compact objects
. this could be several tens of globulars around our galaxy . over ten years the strongest of the sources could have an accumulated signal to noise of several hundred .
there is also a chance to detect inspiraling intermediate - mass black hole binaries at greater distances with lisa .
for example , the virgo cluster , at a distance @xmath20216 mpc ( e.g. , graham et al .
1999 ) , contains @xmath203 galaxies and @xmath204 globulars ( using the mass fraction @xmath205 of mclaughlin 1999 ) . from the model above
, we expect there to be a few binaries within @xmath203 yr of merging , which would therefore have signal to noise ratios @xmath206 in a single year lisa integration . as we discuss in
5 , such binaries are likely to have appreciable eccentricities and detectable frame - dragging , so that there are a number of interesting strong gravity effects that may be probed . as discussed in
2 , coalescing binary black holes in dense stellar clusters may contain intermediate - mass black holes or may consist of two stellar - mass black holes ( via the kozai mechanism or due to a close pass in a two- or three - body interaction ) .
we first treat intermediate - mass black holes , then discuss the possible rates for stellar - mass black hole coalescence . by the time that a binary with an intermediate - mass black hole is in the frequency range of a ground - based instrument , it will have shrunk through a large enough factor in pericenter distance that the orbit will be nearly circular . as discussed by flanagan & hughes ( 1998a , b ) and cutler & thorne ( 2002 )
, the remaining coalescence can then be conveniently divided into inspiral ( where analytical calculations are adequate ) , merger ( where numerical simulations are mandatory because of strong gravity effects ) , and ringdown ( where analytic theory exists ) .
the boundary between inspiral and merger is somewhat arbitrary , but it can be said to be roughly where the inward radial speed increases rapidly ( see ori & thorne 2000 ) .
flanagan & hughes ( 1998a ) conservatively assume that the inspiral is ended when the orbital frequency becomes @xmath207 ; we discuss later the effects of spin on this number .
it is likely that radiation of significant energy and angular momentum in a merger phase requires that the total angular momentum of the system exceed the kerr threshold for the total mass of the system ( s. hughes , personal communication ) . from the work of pfeiffer , teukolsky , & cook ( 2000 ) ,
the orbital angular momentum at the last quasicircular orbit between two black holes ( with total mass @xmath20 and reduced mass @xmath72 ) is @xmath208 , depending on the initial spin magnitudes and orientations . even for equal masses , @xmath209
, this is not enough by itself to exceed the kerr maximum of @xmath210 , and hence the black holes must be prograde and spinning fairly rapidly for there to be a hangup phase before final merger ( s. hughes , personal communication ) . for the higher mass ratio inspirals that are the main focus here , it is probably impossible to exceed the kerr threshold unless the larger black hole is already close to maximally rotating .
it is therefore likely that there will be no merger radiation per se , but that any excess energy or angular momentum would be emitted in an extended ringdown phase .
however , in order to be complete we will include estimates of the detectability assuming that a merger - like phase releases an energy that we scale to a fraction @xmath211 of the total mass - energy of the system ( following the estimate by flanagan & hughes 1998a ) .
similarly , the ringdown energy is scaled to a fraction @xmath212 of the total mass - energy . for a given @xmath20 and @xmath72 ,
there is a luminosity distance out to which each of these phases can be detected with a given instrument with a signal to noise ratio of at least 10 . from flanagan & hughes ( 1998a ) , for the advanced ligo instrument these luminosity distances for a source at redshift @xmath213 are @xmath214 the actual formulae are more complicated but these expressions are reasonably accurate over the @xmath215 range of interest ( except for ringdown , which scales as @xmath20 above @xmath216 and @xmath21 above @xmath217 ) .
the rate of detection at s / n@xmath21810 is given by @xmath219 note that for @xmath220 , the redshift is @xmath221 at a distance of 2 gpc and @xmath222 at a distance of 3 gpc , so cosmological corrections will usually be small . here
mpc@xmath13 is the number density of globular clusters in the local universe ( as estimated by portegies zwart & mcmillan 2000 ) and @xmath224 is the rate at which smaller objects merge with black holes of mass @xmath20 in a given cluster .
the mass distribution of large black holes in clusters is @xmath225 , where @xmath226 is the total fraction of globular clusters that have intermediate - mass black holes . as a specific example , let @xmath227 yr@xmath6 and @xmath228\,m^{-1}$ ] for @xmath20 between @xmath229 and @xmath230 and zero otherwise . for @xmath231
the approximate rates are then @xmath232^{-1}\,{\rm yr}^{-1}\\ r_m&\approx 60h^3(f_{\rm tot}/0.1)\mu_{10}^2m_{\rm max,100 } [ \ln(m_{\rm max}/m_{\rm min})]^{-1}(\epsilon_m/0.1)^{3/2}\,{\rm yr}^{-1}\\ r_r&\approx 0.07h^3(f_{\rm tot}/0.1)\mu_{10}^2m_{\rm max,100}^{5/2 } [ \ln(m_{\rm max}/m_{\rm min})]^{-1}(\epsilon_r/0.03)^{3/2}\,{\rm yr}^{-1}\ ; .\\ \end{array}\ ] ] for example , if @xmath233 , @xmath234 , @xmath235 , @xmath236 , @xmath237 , and @xmath238 then @xmath239 yr@xmath6 , @xmath240 yr@xmath6 , and @xmath241 yr@xmath6 . at a different signal to noise threshold @xmath242 , these rates should be multiplied by roughly @xmath243 , modulo cosmological corrections .
note that the rapid increase in number with diminishing signal to noise ratio means that for the purposes of statistical analysis these will dominate the constraints possible with these data .
similar answers are obtained for more general power laws , @xmath244 .
note that the division of numbers between inspiral and merger is based on the conservative assumption that inspiral continues only to a frequency @xmath245 ( flanagan & hughes 1998a ) , which is the frequency at the innermost stable circular orbit for a test particle around a nonrotating black hole . as shown in 2.2 ,
the random walk process of accretion of smaller black holes is likely to produce spin parameters of order a few tenths .
for example , a @xmath246 black hole that has grown from @xmath68 by accretion of @xmath31 black holes has a mean spin parameter @xmath247 ( see figure 1 ) . at this spin the frequency is increased by a factor 1.7 for equatorial prograde orbits , implying an increase of up to a factor of several in the detection rate because the detection sensitivity increases with frequency in this range .
further analysis of this effect will be important . for the lisa instrument
, black holes in this mass range will not be detectable during merger and ringdown . from flanagan & hughes ( 1998a )
, the final inspiral could be detected at @xmath248 in a one year integration out to a luminosity distance @xmath249 thus , the expected rate of objects detectable in a one year integration is @xmath250^{-1}\,{\rm yr}^{-1}\;.\ ] ] note , however , that over a longer integration the rate goes up dramatically because the gravitational wave amplitude scales with frequency as @xmath251 whereas in this frequency range the s / n=5 threshold of lisa scales as @xmath252 .
for example , in a 10 year integration lisa would be expected to see several to tens of objects in the last phase of their inspiral .
this leads to the interesting possibility that a @xmath253 black hole coalescing with a @xmath31 black hole may be observed with lisa a few years prior to merger , then its final merger could be seen later with ground - based instruments .
the waveform observed with lisa could be projected to the final merger , so that the time , phase , and other characteristics could be anticipated and detected with ground - based gravitational wave detectors and the region could be observed simultaneously with conventional telescopes .
given that the angular resolution of lisa would at best be a few degrees for these sources ( e.g. , cutler 1998 ) , it may be necessary to employ wide - field monitoring to catch the final merger .
we note that detection of any significant electromagnetic radiation during the merger would require a profound revision of our understanding of these systems and possibly of gravitational radiation itself . in a globular cluster
the gas density is too low for there to be meaningful accretion from the interstellar medium .
if at one point there was a substantial accretion disk around either compact object , it is likely that it would have been completely disrupted or accreted by the time of the final merger .
in addition , gravitational radiation is not thought to couple significantly to an electromagnetic field , so negligible photon luminosity is expected from the final merger .
if instead there is an electromagnetic counterpart to a merger of an intermediate - mass black hole with a neutron star or stellar - mass black hole , parts of this picture must be revised .
let us now turn from interactions of intermediate - mass black holes to interactions among only stellar - mass black holes .
as before , there is an upper limit to the rate set by the timescale in which the supply of stellar - mass black holes is used up .
for example , if a cluster contains @xmath254 stellar - mass black holes , a strong upper bound on the binary black hole merger rate by any mechanism is @xmath255 yr@xmath6 assuming that the supply is not exhausted in much less than a hubble time .
the actual rate depends on details .
consider for example merger by the kozai mechanism after a binary - binary encounter ( miller & hamilton 2002b ) .
the rate is proportional to the square of the fraction of black holes in binaries .
if most black holes are in binaries then the rate could be several tens of percent of the maximum ( see miller & hamilton 2002b ) .
if the fraction in binaries is @xmath256 , then , the rate per cluster is @xmath257 yr@xmath6 .
we expect that the clusters that contribute most to the rate of stellar black hole coalescences will be those with relatively high central densities , because low - density clusters will have a low interaction rate .
let us parameterize the fraction of clusters that contribute significantly as @xmath182 , where @xmath182 is likely to be a few tenths .
note that the denser clusters are also the ones most likely to produce intermediate - mass black holes , hence it is likely that @xmath258 , but to keep the processes distinct we use different variables to represent the fractions .
the volume rate in the universe for mergers of two stellar - mass black holes via the kozai mechanism is then @xmath259 yr@xmath6 . the other way stellar - mass black holes can coalesce in a cluster is by having a near approach during a three - body encounter .
as estimated in 6 , the rate of such coalescences will be roughly 10% the rate at which black holes are ejected from the cluster by dynamic recoil . using the bounding estimates of portegies zwart & mcmillan ( 2001 ; their eqns . 8 and 9 ) , this implies a rate in the universe of @xmath260 yr@xmath6 .
thus , if @xmath261 then the kozai mechanism dominates , otherwise direct coalescence in three - body interactions is more important . from flanagan & hughes ( 1998a )
, the inspiral of a pair of @xmath31 black holes could be detected with s / n@xmath262 out to a distance 1.6 gpc .
this implies a combined rate of @xmath263 $ ] yr@xmath6 in the advanced ligo detector for two stellar - mass black holes , compared with a rate of @xmath264 yr@xmath6 for mergers of stellar - mass black holes with intermediate - mass black holes .
the high expected eccentricities of binaries in the lisa band imply that it may be possible to observe precession of the pericenter much deeper in a gravitational well than is possible for known galactic neutron star binaries .
the angle of precession in an orbital period is @xmath265\ ] ] ( e.g. , misner , thorne , & wheeler 1973 , pg .
if @xmath73 , this is @xmath266 .
the effect of precession will be to split the single frequency of the orbit into a pair with a separation that can be detected if the observation is for a time @xmath267t$ ] , where as before @xmath158 is the orbital time ( cf .
pierro et al .
2001 ) . combining factors , the required observation time to barely resolve pericenter
precession is @xmath268 for example , a @xmath201 compact object in an @xmath269 orbit around a @xmath135 black hole , with a merger time of @xmath91 yr , requires roughly two years of observation .
when the same binary has a merger time of @xmath96 yr ( and therefore has @xmath270 ) , only two days of observation are required .
this suggests that binaries in the virgo cluster , which are numerous enough for @xmath271 yr to be probable , are good candidates for observation of pericenter precession , and that such precession may also be observable in our own galactic globular system . the lense - thirring precession rate , at which the axis of the orbital plane changes , is approximately @xmath272 .
integrated over an orbit of semimajor axis @xmath124 and eccentricity @xmath82 , the average rate is @xmath273 $ ] . as is evident from the formulae in 4.1 , the signature of such precession would be a periodic change in the relative amplitudes in the two polarizations as the angle @xmath274 varies .
for precession of more than a radian to occur during an observation time @xmath275 therefore requires @xmath276 thus it will typically be difficult to detect this effect in galactic globulars unless the signal to noise ratio is so high that precession of @xmath277 radians can be detected .
however , a binary in the virgo cluster with a @xmath201 and a @xmath135 black hole with @xmath278 that will merge within @xmath96 yr will require @xmath279 yr of observation , so this is well within reach .
an even better hope of detecting lense - thirring precession lies in characterization of the inspiral waveform with a ground - based detector such as the advanced ligo detector .
the ratio of lense - thirring to orbital frequency is @xmath280 for nearly circular orbits , so if the expected tens to hundreds of orbits of inspiral are observed then the precession of the orbital plane should be evident .
the high inclination orbits expected in this scenario are sensitive to the multipole moments of the mass distribution and may therefore test the no - hair theorem ( ryan 1996 ) , although signal to noise ratios of tens are typically required for significant constraints in a single inspiral ( ryan 1997 ) .
nonetheless , statistical combination of the constraints from observations of multiple mergers , even at relatively low signal to noise , could provide interesting limits .
observations of these orbits may therefore allow initial mapping of spacetime around rotating black holes ( particular in the final merger detectable with ground - based instruments ) , a job expected to be completed with high precision by lisa observations of stellar - mass black holes being consumed by supermassive black holes in galactic centers ( hughes 2001 ) . for binaries close to the end of their inspiral , the orbital frequency could change enough during @xmath281 yr that the inspiral is detected .
an orbit with merger time @xmath282 will change its frequency by a fractional amount @xmath283 in an observation of duration @xmath275 . for this to be detected via a one cycle shift during the observation
requires that @xmath284 , or @xmath285 with some luck ( specifically @xmath286 yr ) this may be observable within the galactic globular system , but once again the virgo cluster binaries are excellent candidates , where at signal to noise ratios greater than 10 the decay of the orbit will be clear in the signal .
one can combine the three effects discussed in this section to speculate about astrophysical information that might be available from gravitational radiation .
an interesting possibility is that the distance to the virgo cluster could be estimated from gravitational wave signals alone , with an accuracy that is competitive with optical measurements . alternatively , using optically measured distances as an input , the system is overdetermined and detection of pericenter precession and orbital decay would allow strong consistency checks of the underlying formulae .
suppose that lisa is operational for ten years .
then , from the estimates in 4 , there will be several sources in virgo that are detectable with lisa with a signal to noise of s / n@xmath287 in that ten year period .
with such a high signal to noise , the modulation due to the orbit of the earth will localize the sources to within several degrees , and the membership in virgo will be based on this association . from the discussion of black hole spin and binary eccentricity in 2 ,
the typical eccentricity for such binaries would be @xmath288 , and the typical spin parameter would be @xmath289 .
thus pericenter precession , lense - thirring precession , and decay of the orbit would all be detectable .
the eccentricity would be evident from the waveform .
this means that the combinations @xmath290 ( from the frequency ) , @xmath291 ( from pericenter precession ) , and @xmath292 $ ] ( from decay of the orbit ) are all independently measurable .
combining these , @xmath124 , @xmath62 , and @xmath20 would all be determined . along with the strength of the waves at the detector
, this would yield an estimate of the distance to the cluster from the gravitational radiation alone ( for discussion of globular cluster distance determinations using gravitational waves , see benacquista 2000 ) .
it will also be possible to estimate the spin angular momentum of the larger black hole from lense - thirring precession .
dense stellar clusters are promising locations for sources of gravitational radiation .
we estimate that up to several tens of times per year , the advanced ligo detector will see the coalescence of a small black hole with a larger one ( a few to tens during inspiral , and most of the rest during merger , with both numbers dependent on black hole spin ) .
mergers of two stellar - mass black holes in clusters will likely be detectable at a rate of a few per year . as always in astrophysical scenarios there are sources of uncertainty .
the most major is the question of the number of black holes that are to be expected in a dense cluster , whether in the @xmath293 stellar - mass range or the @xmath294 intermediate - mass range .
perhaps the most easily calculable input to this question is the number of black holes that are originally produced . if the initial mass function for the cluster was the salpeter function @xmath295 above @xmath201 and flatter below @xmath201 ( e.g. , meyer et al .
2000 ) , then approximately 0.1% of stars began with @xmath296 and presumably evolved into black holes .
if the initial mass function was instead the scalo ( 1986 ) distribution , which drops off more sharply at higher masses , the fraction with @xmath296 is more like 0.05% ( portegies zwart & mcmillan 2000 ) .
current dense clusters have @xmath297 stars , and may have had several times more at birth , so the number of stars that evolved into black holes is typically @xmath203 .
there are , however , numerous ways in which black holes may be lost from the cluster .
the first is in the supernova that produced the black hole .
neutron stars are known to have significant birth kicks of tens to hundreds of km s@xmath6 ( e.g. , hansen & phinney 1997 ; fryer & kalogera 1997 ) .
the mechanism for this is still debated ( spruit & phinney 1998 ; kusenko & segre 1999 ; lai , chernoff , & cordes 2001 ) , but it is thought that similar kick velocities for black holes are much smaller ( e.g. , brandt , podsiadlowski , & sigurdsson 1995 ; podsiadlowski et al . 2002 ; see nelemans , tauris , & van den heuvel 1999 for a somewhat different perspective ) , if for no other reason than that black holes are several times more massive than neutron stars so that a fixed energy or momentum in the kick would lead to reduced speeds .
it is therefore plausible that black holes do not receive birth kicks of @xmath298 km s@xmath6 , in which case they are retained in the cluster .
a second loss mechanism involves three - body recoil .
several simulations have shown that black holes of a fixed mass @xmath293 in binaries with other such black holes tend to be ejected by three - body interactions before they can merge by gravitational radiation ( kulkarni et al .
1993 ; sigurdsson & hernquist 1993 ; portegies zwart & mcmillan 2000 ) . the actual fraction of ejected black holes can depend on the mass function of stars and other variables , but recent estimates suggest that 10% or more of the initial black holes can be retained by their clusters over a hubble time ( portegies zwart & mcmillan 2000 ) . if more massive black holes are present initially , then miller & hamilton ( 2002a ) showed that they are usually not ejected .
these can grow by mergers after multiple three - body encounters , but typically they will eject several to tens of field stars along the way . the majority of the encounters , however , will not be with black holes so this mechanism is not expected to deplete the black hole supply significantly .
there are , in addition , at least two ways in which multibody interactions can produce a merger without accompanying dynamical recoil .
one , discussed by miller & hamilton ( 2002b ) , is that binary - binary interactions can produce a stable hierarchical triple system , and if the inclination of the orbit of the tertiary to the orbit of the inner binary is in the right range then a secular kozai resonance can increase the eccentricity of the inner binary to the extent that it merges before the next encounter , without significant recoil .
the impact of this effect depends on the binary fraction and the distribution of inclinations ( see miller & hamilton 2002b ) , but this could allow the merger of some tens of percent of the original population of black holes .
the other recoilless possibility involves resonant encounters in three - body interactions , in which the three objects orbit hundreds or thousands of times before resolving into a binary and an unbound single star . if , during these orbits , two black holes pass close enough to each other that losses to gravitational radiation in a single pass cause rapid merger , then again there is no dynamical kick .
various estimates suggest that in an equal - mass three - body encounter the probability of the closest approach being less than @xmath299 times the initial semimajor axis is @xmath300 ( hut 1984 ; mcmillan 1986 ; sigurdsson & phinney 1993 ) .
two @xmath31 black holes must approach to within @xmath301 cm to merge in a year , which will happen in a given resonant encounter with a probability of a few tenths of a percent for a semimajor axis of a few astronomical units .
if it takes @xmath281 equal - mass encounters to harden a binary to the point of ejection , this suggests that several percent of binaries will merge before ejection in this fashion . combining all of the above effects
, it seems likely that tens of percent of the original black hole population will not be ejected by three - body recoil , leaving a present - day population of hundreds .
the third loss mechanism involves the merger itself . the emission of gravitational waves during inspiral and merger will be somewhat asymmetric , leading to recoil .
calculations thus far have focused on the weak - field limit .
they suggest that in the post - newtonian limit the kick scales as @xmath302 with the separation @xmath303 at the last orbit before dynamical instability . at the separation @xmath304 appropriate for test particles around slowly rotating black holes , the kick will be a few kilometers per second for the mass ratio @xmath305 that maximizes the kick ( fitchett 1983 ; wiseman 1992 ) .
binaries with large mass ratios or nearly equal masses experience less recoil ( for example , by symmetry , equal - mass binaries have no kick ) . also , @xmath303 is greater in a comparable - mass binary than in the test particle limit ( pfeiffer et al . 2000 ) , which also decreases the kick .
it is not clear what level of recoil is to be expected in the merger phase , where the radial velocity becomes rapid .
strong - field calculations are required to resolve whether this process is dominant ( perhaps kicking most merging black holes out of the cluster ) or insignificant ( if the recoil speeds are much less than the @xmath306 km s@xmath6 escape speeds from the core ) .
thus , if mergers do not kick black holes out of dense clusters , one may expect at least tens to hundreds of black holes in many current systems .
these are expected to reside primarily in the core of the cluster , where they have a greater tendency to interact with the more massive ( and hence compact ) objects in the cluster .
this may explain why no definitive examples of black hole low - mass x - ray binaries are known in the globular cluster systems of the milky way or andromeda ; such a population would not undergo mass transfer and would thus be observable only by its gravitational wave emission .
note , however , that there is a population of @xmath307 erg s@xmath6 sources in the globulars around a number of ellipticals ( angelini , loewenstein , & mushotzky 2001 ; white , sarazin , & kulkarni 2002 ; kundu , maccarone , & zepf 2002 ) .
possible differences between these systems are an important subject for future study . what about the fraction of clusters with intermediate - mass black holes ?
miller & hamilton ( 2002a ) estimate that clusters with central densities greater than @xmath204 pc@xmath13 have high enough encounter rates to produce @xmath10 black holes . in the milky way globular system , this would imply that roughly 40% of globulars could host such objects ( pryor & meylan 1993 ) .
however , to be conservative , we have adopted 10% as our fiducial value for the estimates of merger rates .
we have also been conservative in assuming that the number density and mass of globulars is the same out to 2 - 3 gpc as it is in the local universe .
as first discussed by aguilar , hut , ostriker ( 1988 ) , evaporation and tidal interactions attenuate the globular system of a galaxy . therefore
, it is possible that coalescence rates a few billion years ago were higher by up to a factor of a few than they are now ( portegies zwart & mcmillan 2000 ) , but this is highly uncertain .
the general model described here is one that can be tested and enhanced in ways both observational and theoretical . from the observational standpoint it is important to continue kinematic work on globulars to look for evidence of the velocity and density cusps that are expected signatures of black holes ( bahcall & wolf 1975 ; frank & rees 1976 ; see gebhardt et al .
2000 , damico et al .
2002 for recent results ) . further characterization of the ultraluminous x - ray sources is also important .
for example , if a mass estimate can be obtained via radial velocity measurements of a companion , this will shed new light on the nature of these objects . from the theoretical standpoint
there are several important calculations .
these include : ( 1 ) strong - gravity computations of the recoil speed of a black hole binary after merger , for different mass ratios and spins , ( 2 ) comprehensive numerical simulations of three - body interactions with high mass ratios , to represent intermediate - mass black holes , and ( 3 ) detailed numerical analysis of binary - binary encounters and the role of the kozai resonance , among others
. whatever the results of this work , there will be significant new understanding gained on many fronts , including the information that can be obtained from analysis of the waveforms of the gravitational radiation produced by black hole mergers in dense clusters .
we thank scott hughes , sterl phinney , steinn sigurdsson , kip thorne , and linqing wen for discussions about gravitational radiation and cluster dynamics , tuck stebbins for discussions and for providing lisa response curves , and chris reynolds and scott hughes for careful readings of an earlier version of this manuscript .
this paper was supported in part by nasa grant 5 - 9756 and nsf grant ast 0098436 .
fidecaro , f. et al .
1997 , general relativity and gravitational physics ; proceedings of the 12th italian conference , eds .
bassan , m. , ferrari , v. , francaviglia , m. , fucito , f. , & modena , i. , world scientific press , p.163 meyer , m. r. , adams , f. c. , hillenbrand , l. a. , carpenter , j. m. , & larson , r. b. 2000 , in protostars and planets iv , eds .
mannings v. , boss a. p. , & russell s. s. ( tucson : university of arizona press ) , p. 121 pakull , m. w. , & mirioni , l. 2002 , to appear in the proceedings of the symposium
new visions of the x - ray universe in the xmm - newton and chandra era " , 26 - 30 november 2001 , estec , the netherlands ( astro - ph/0202488 ) | recent x - ray observations of galaxies with rosat , asca , and chandra have revealed numerous bright off - center point sources which , if isotropic emitters , are likely to be intermediate - mass black holes , with @xmath0 .
the origin of these objects is under debate , but observations suggest that a significant number of them currently reside in young high - density stellar clusters .
there is also growing evidence that some galactic globular clusters harbor black holes of similar mass , from observations of stellar kinematics .
in such high - density stellar environments , the interactions of intermediate - mass black holes are promising sources of gravitational waves for ground - based and space - based detectors . here
we explore the signal strengths of binaries containing intermediate - mass black holes or stellar - mass black holes in dense stellar clusters .
we estimate that a few to tens per year of these objects will be detectable during the last phase of their inspiral with the advanced ligo detector , and up to tens per year will be seen during merger , depending on the spins of the black holes .
we also find that if these objects reside in globular clusters then tens of sources will be detectable with lisa from the galactic globular system in a five year integration , and similar numbers will be detectable from more distant galaxies .
the signal strength depends on the eccentricity distribution , but we show that there is promise for strong detection of pericenter precession and lense - thirring precession of the orbital plane .
we conclude by discussing what could be learned about binaries , dense stellar systems , and strong gravity if such signals are detected . |
denote by @xmath221 the vector space generated by the infinite series of the form @xmath222 where @xmath223 , @xmath224 .
we denote the sum above by @xmath225 . as a vector space , @xmath221 is `` a very huge '' vector space generated by the _ set _ @xmath226 .
analogously , introduce the notation @xmath227 , @xmath228 , and the notation @xmath229 for @xmath230 . now introduce a dg lie algebra @xmath231 .
first we introduce a graded vector space @xmath232 , and then @xmath231 will be a quotient space .
we set : @xmath233\oplus\bigoplus_{n\ge 2}\mathbb{c}\{\operatorname{hom}(v , v^{\otimes n})\}[-n+1]\oplus\\ \oplus\bigoplus_{m , n\ge 2}\operatorname{hom}(v^{\otimes m},v^{\otimes n})[-m - n+2]\end{gathered}\ ] ] now we introduce a prebracket on @xmath232 .
it means that it is skew - symmetric but does not obey the jacobi identity .
the idea goes back to the constructions in section 2 .
we associate generators with the strata of codimension 0 , and their compositions ( the bracket ) by the strata of codimension 1 .
then the jacobi identity follows ( after factorization ) from the equation @xmath234 where @xmath235 is the chain differential .
philosophically , it is a kind of the markl s construction from [ m1 ] , but the author can not verbalize it at the moment .
the appearance of the infinite power series @xmath225 and @xmath229 are motivated by the identity ( [ eq23.01 ] ) .
thus , the reader can say that no infinite sums appear among the strata of codimension 0 .
nevertheless , the markl s construction is an operadic construction , and when we deal with @xmath0props we need some modifications . only the following brackets are nonzero : @xmath236:=\overline{[\psi_1,\psi_2]_g}\\ & { \rm where}\ \psi_1\in\operatorname{hom}(v^{\otimes m_1},v),\ \psi_2\in\operatorname{hom}(v^{\otimes m_2},v ) , \\ & { \rm and}\ [ \psi_1,\psi_2]_g\ { \rm is}\ { \rm the}\ { \rm gerstenhaber}\ { \rm bracket},\\ { \rm ( ii)}\ \ & [ \overline{\theta_1},\overline{\theta_2}]:=\overline{[\theta_1,\theta_2]^g}\\ & { \rm where}\ \theta_1\in\operatorname{hom}(v , v^{\otimes n_1}),\ \theta_2\in\operatorname{hom}(v , v^{\otimes n_2}),\\ & { \rm and}\ [ \theta_1,\theta_2]^g\ { \rm is}\ { \rm the}\ { \rm gerstenhaber}\ { \rm cobracket,}\\ { \rm ( iii)}\ \ & [ \overline{\theta},\overline{\psi}]:=\theta\circ\psi\pm(\theta)^{\otimes m}\circledcirc ( \psi)^{\otimes n}\\ & { \rm where}\ \psi\in \operatorname{hom}(v^{\otimes m},v),\ \theta\in\operatorname{hom}(v , v^{\otimes n}),\ \circ\ { \rm and } \circledcirc\\ & { \rm are}\ { \rm the}\ { \rm compositions}\ { \rm in}\ { \rm the}\ { \rm pre-}\frac23{\rm prop}\ { \mathrm{end}}(v),\\ { \rm ( iv)}\ \ & { \rm for}\ \alpha\in\operatorname{hom}(v^{\otimes m},v^{\otimes n}),\ m , n\ge 2,\ \psi\in\operatorname{hom}(v^{\otimes m_1},v)\\ & [ \overline{\psi},\alpha]:=\sum_{j=1}^m\pm\psi(_j\circ)\alpha,\\ { \rm ( v)}\ \ & { \rm for}\ \alpha\in\operatorname{hom}(v^{\otimes m},v^{\otimes n}),\ m , n\ge 2,\ \theta\in\operatorname{hom}(v , v^{\otimes n_1})\\ & [ \alpha,\overline{\theta}]:=\sum_{i=1}^n\pm\alpha\circ_i\theta \end{aligned}\ ] ] we also suppose that the bracket @xmath237 $ ] is graded - skew - commutative .
this bracket does _ not _ obey the jacobi identity .
the jacobi identity fails for the brackets @xmath238,\overline{\psi_3}]$ ] , @xmath239,\overline{\theta_3}]$ ] , @xmath238,\overline{\theta}]$ ] , and @xmath240,\overline{\theta_2}]$ ] .
our solution is to factorize by the vector space spanned by the jacobi identities , as follows : first , factorize the space generated by @xmath225 by the space @xmath241 _ generated _ by the jacobi identities .
it means that we first consider a graded vector space @xmath242 spanned by the vectors @xmath243_g,\psi_3]_g}\mp\overline{[\psi_1,[\psi_2,\psi_3]_g]_g}\pm \overline{[\psi_2,[\psi_1,\psi_3]_g]_g}$ ] where @xmath244 for some @xmath245 .
then consider the vector space @xmath241 spanned by the elements of the form @xmath246\dots]$ ] where @xmath247 , and @xmath248 for all @xmath90 , @xmath249 .
define the space @xmath250 , @xmath224 , as the graded component of the quotient of the space spanned by all vectors @xmath225 , by the graded space @xmath241 .
it is clear that @xmath251 $ ] with the bracket ( [ eq233.3 ] ) is a graded lie algebra .
one can show that this graded lie algebra is bigger than @xmath252 $ ] .
analogously we define the space @xmath253 and the graded lie algebra @xmath254 $ ] .
now we define the quotient - spaces @xmath255 for @xmath256 .
notice , that the bracket of @xmath241 and @xmath253 with @xmath257 is _ zero without any factorization _ because it depends only on the `` linear part '' of @xmath258 .
the same is true for the product @xmath259 .
then , define @xmath260 as the graded vector space spanned by all elements of the form @xmath261\circledcirc\overline{\theta}\mp
\psi_1\circ_*(\overline{\psi_2}\circledcirc\overline{\theta})\pm \psi_2\circ_*(\overline{\psi_1}\circledcirc\overline{\theta})$ ] where @xmath262 , @xmath263 and the analogous expressions with 2 @xmath89 s and 1 @xmath133 .
here @xmath264 denotes the sum with the signs as in ( iv),(v ) in the definition ( [ eq233.3 ] ) .
next , denote by @xmath265 the graded vector space generated by the elements of the form @xmath266 where @xmath267 .
finally , denote by @xmath268 $ ] , @xmath269 , the graded component of the quotient of @xmath270)/i$ ] .
denote @xmath271\oplus\\ & \bigoplus_{n\ge 2}\overline{\operatorname{hom}(v , v^{\otimes n})}[-n+1]\oplus\\ & \bigoplus_{m , n\ge 2}\overline{\operatorname{hom}(v^{\otimes m},v^{\otimes n})}[-m - n+2 ] \end{aligned}\ ] ] the formulas for the bracket ( [ eq233.3 ] ) define a graded lie algebra structure on @xmath231 ( that means that the jacobi identity is satisfied ) .
it follows from the definitions .
it would be very interesting to specify in which sense our construction is an analog of the markl s construction [ m1 ] applied to the case of @xmath0props .
the dg lie algebra @xmath231 defined above is a `` deformation lie algebra '' ( see section 4 ) for the bialgebra @xmath4 with 0 product and 0 coproduct .
when we want to consider deformation theory for a bialgebra @xmath4 with non - zero ( co)product , we localize @xmath231 by the corresponding solution of the maurer - cartan equation .
let @xmath272 and @xmath273 are the product and the coproduct for a ( co)associative bialgebra structure on @xmath4 .
then @xmath274 satisfies the maurer - cartan equation with 0 differential : @xmath275=0\ ] ] it is clear .
now for such @xmath276 as above , we consider the dg lie algebra @xmath277 which is the same as @xmath231 but with the differential @xmath278 .
let @xmath279 , @xmath280 be two @xmath5 algebras . recall that it means that we have odd vector fields of degree + 1 @xmath281 on @xmath282 $ ] and @xmath283 on @xmath284 $ ] such that @xmath285 .
suppose we have an @xmath5 map @xmath286 .
it means , by definition , that we have a non - linear map @xmath287\to{\mathfrak{g}}_2[1]$ ] which maps the field @xmath281 to the field @xmath283 .
suppose that the map @xmath288 is a ( non - linear ) imbedding of topological spaces .
then we can say that the vector field @xmath283 is tangent to the image @xmath289)$ ] ( because it coincides with the image of @xmath281 ) .
vice versa , suppose we have an @xmath5 algebra @xmath280 , a _ graded vector space _ @xmath279 , and an imbedding @xmath287\to{\mathfrak{g}}_2[1]$ ] such that the vector field @xmath283 on @xmath280 is _ tangent _ to the image .
then we claim that there is a unique @xmath5 structure on @xmath279 which makes @xmath288 an @xmath5 map .
apply it now to the case when @xmath290 .
consider the gerstenhaber - schack space @xmath291 $ ] as @xmath279 .
consider the following non - linear map @xmath292 : the map @xmath288 maps @xmath293 $ ] identically to @xmath231 when @xmath256 . when @xmath294 , @xmath288 maps @xmath295 to @xmath296 , and for @xmath97 , @xmath288 maps @xmath297 to @xmath298 .
it is clear that we are in the assumptions above ( before the localization by the solution @xmath299 of the maurer - cartan equation ) .
it means that the quadratic vector field on @xmath300 $ ] defining the dg lie algebra structure on @xmath231 is _ tangent _ to the image of @xmath288 .
it allows us to define an @xmath5 structure on the gerstenhaber - schack space @xmath279 which makes @xmath288 an @xmath5 map .
the author hopes to construct this @xmath5 structure explicitly in the next paper .
he is not sure that the linear part of this @xmath5 structure will be the gerstenhaber - schack differential , but he is sure that in this way we obtain a more right object .
i am grateful to maxim kontsevich for sharing with me with his construction of the spaces @xmath2 . when i constructed the croc compactification of them , i discussed it many times with borya feigin , and these discussions finally explained to me the necessity of a stasheff - type compactification for the usual ( commutative ) deformation theory of ( co)associative bialgebras .
discussions on this compactification and on the corresponding lie algebra with giovanni felder were very useful for me .
i am grateful to martin markl for the reading of the first version of this paper and for his remarks and corrections .
i also would like to thank the eth ( zurich ) for the financial support and for the very stimulating atmosphere .
999 b. feigin , the conformal field theory from the view of the cohomology theory of lie algebras , icm-90 ( kyoto ) , math .
society of japan , tokyo , 1990 m. gerstenhaber , s.d .
schack , bialgebra cohomology , deformations , and quantum groups , proc .
usa , vol .
87(1990 ) , pp . 478 - 481 , m. kontsevich , deformation quantization of poisson manifolds i , preprint q - alg/9709040 m. kontsevich , operads and motives in deformation quantization , preprint math.qa/9904055 m. kontsevich , letter to m.markl , 12 november 2002 m. markl , cotangent cohomology of a category and deformations , journal of pure and appl .
algebra , 113 ( 1996 ) , 195 - 218 | we introduce a concept of @xmath0prop generalizing the kontsevich concept of @xmath1prop .
we prove that some stasheff - type compactification of the kontsevich spaces @xmath2 defines a topological @xmath0prop structure .
the corresponding chain complex is a minimal model for its cohomology ( both are considered as @xmath0props ) .
we construct a @xmath0prop @xmath3 for a vector space @xmath4 .
finally , we construct a dg lie algebra controlling the deformations of a ( co)associative bialgebra .
philosophically , this construction is a version of the markl s operadicconstruction from [ m1 ] applied to minimal models of @xmath0props . to [ [ section ] ]
the goal of this paper is to construct a deformation theory for ( co)associative bialgebras . according to general principles ,
it means that we are looking for a dg lie algebra ( or , more generally , for an @xmath5 algebra ) controlling the deformation theory of a ( co)associative bialgebra . in the case of the deformation theory of associative algebras ,
such a dg lie algebra controlling the deformations of an associative algebra @xmath6 , is the cohomological hochschild complex of @xmath6 with the gerstenhaber bracket .
( more precisely , this hochschild complex controls the deformations of the category of @xmath6-modules ) .
first of all , recall that a ( co)associative bialgebra is a vector space @xmath6 equipped with the maps @xmath7 ( the product ) and @xmath8 ( the coproduct ) .
the product is supposed to be associative and the coproduct is supposed to be coassociative . moreover , we suppose the following compatibility of them : @xmath9 for any @xmath10 .
( here in the r.h.s .
the product is the component product in @xmath11 defined as @xmath12 ) .
notice that we do not suppose the existence of unit and counit in @xmath6 .
here we meet our first difficulty : the r.h.s . of ( [ eq230.1 ] )
is of the 4th degree and not quadratic . recall that ( little bit roughly ) we associate the deformation theory with a dg lie algebra @xmath13 as follows : we consider the solutions of the maurer - cartan equation @xmath14=0\ ] ] for @xmath15 modulo the action of the gauge group associated with @xmath16 on these solutions .
( because of possible divergences in the action of the gauge group , we say instead of this direct construction that the deformation functor is a functor from the category of the artinian algebras to the category of sets )
. it is known that the gerstenhaber - schack complex [ gs ] associated with a bialgebra @xmath6 is a deformation complex of the bialgebra structure on @xmath6 .
it means that the first cohomology of this complex are isomorphic to the infinitesimal deformations on @xmath6 . to pass from the infinitesimal deformations to the global ones
, one needs to have an appropriate dg lie algebra structure on the gerstenhaber - schack complex ( or , more generally , an @xmath5-structure ) .
recall here that as a vector space , the gerstenhaber - schack complex of @xmath6 is @xmath17\ ] ] in particular , in degree 1 we have : @xmath18 .
we could expect that for some dg lie algebra structure on @xmath19 the maurer - cartan equation ( [ eq230.2 ] ) for the element @xmath20 means exactly that @xmath21 defines a new ( co)associative bialgebra structure on @xmath6 .
but it is impossible : because the r.h.s . of the equation ( [ eq230.1 ] ) is of the 4th degree in @xmath22 and @xmath23 , while the maurer - cartan equation ( [ eq230.2 ] ) is quadratic .
it means that the best we could expect is to have an @xmath5 algebra structure on @xmath19 ( which looks quite complicated ) .
this crucial observation was explained to the author by boris tsygan about 3 years ago .
now remember that the @xmath5 algebras and the dg lie algebras is more or less the same : if we have an @xmath5 algebra structure on a graded vector space @xmath4 , we necesserily have an @xmath5 isomorphic structure of pure dg lie algebra on a ( bigger ) space @xmath24 .
it means that the question of which structure we have , dg lie algebra or @xmath5 algebra , is the question of the right choice of `` generators '' .
this means , in particular , that we could expect the existence of a complex quasi - isomorphic to the gerstenhaber - schack complex ( `` with another generators '' ) and a dg lie algebra on it , which solves the deformation problem of a ( co)associative bialgebra @xmath6 .
this idea is one source of the theory developed here in this paper .
another source is the kontsevich spaces @xmath2 .
the reader can find the definition of them in section 2 of the paper .
the original kontsevich motivation when he invented these spaces was the following : the space @xmath25 is the configuration space on two independent lines , we have 2 points in each line modulo independent common shift on each line , and modulo the following action of @xmath26 on this space : for @xmath27 , we dilatate the first line with the scale @xmath28 and the second with the coefficient @xmath29 . then @xmath25 is a 1-dimensional space : we have an interval on the first line , an interval on the second ( we identify the intervals of the same length ) , and we identify such two configurations with the same product of the lengthes of the two intervals . therefore , the configuration has the only one module the product of the lengthes of the intervals .
before compactification , it is isomorphic to @xmath30 .
now we compactify the space @xmath25 to the closed interval .
the two limit configurations are shown in the figure 1 : the kontsevich s insight was that the left picture should give the left - hand side of the compatibility equation ( [ eq230.1 ] ) , while the right picture should give the right - hand side of ( [ eq230.1 ] ) .
we say `` should give '' having in mind the markl s construction in [ m1 ] , or even further , a construction of the type of kontsevich formality .
after these short remarks , we pass to our constructions .
if the reader is interested mostly in our construction of the deformation dg lie algebra , he can begin to read the paper from section 3 and to come back to the previous two sections if it is necesserily .
[ [ section-1 ] ] here we define our main technical tool@xmath0props .
the name @xmath0props indicates that this concept is a further generalization ( or simplification ) of the concept of @xmath1prop due to maxim kontsevich ( see [ k3 ] , [ mv ] ) .
we define a @xmath0prop of vector spaces , a @xmath0prop of dg vector spaces , of topological spaces , ... can be defined analogously .
a pre-@xmath0prop of vector spaces consists of the following data : * a collection of vector spaces @xmath31 defined for @xmath32 , with an action of symmetric groups @xmath33 on @xmath31 , * a collection of vector spaces @xmath34 defined for @xmath35 , with an action of the symmetric group @xmath36 on @xmath34 , * a collection of vector spaces @xmath37 defined for @xmath38 , with an action of the symmetric group @xmath39 on @xmath37 , * compositions @xmath40 , @xmath41 , * compositions @xmath42 , @xmath43 , * compositions @xmath44 , @xmath41 , * compositions @xmath45 , @xmath43 , * compositions @xmath46 , * all the compositions are equivariant with respect to the actions of the symmetric groups .
this data should obey the following properties : * the composition @xmath47 is associative for @xmath48 : in this case @xmath49 coincides with @xmath50 , * in the notations of ( 1 ) , if @xmath51 or @xmath52 , we have @xmath53 is equal to @xmath54 where @xmath55 if @xmath51 , and @xmath56 if @xmath52 _ ( the commutativity ) _ , * the property analogous to ( 1 ) for @xmath57 , * the property analogous to ( 2 ) for @xmath57 , * the analogous proprties for @xmath58 and for @xmath59 .
notice that this structure without @xmath60 and @xmath61 is exactly the kontsevich s @xmath1prop structure .
a pre-@xmath0prop is a @xmath0prop if the operations of @xmath62-type are compatible with the operations of @xmath63-type , as follows : there are extra maps @xmath64 , @xmath65 and mapd @xmath66 , @xmath67 , which are supposed to be equivariant with respect to the actions of symmetric groups .
we also suppose that these maps are isomorphisms .
then we have : * @xmath68 is equal to @xmath69 .
more precisely , let @xmath70 , @xmath71 , and @xmath72 .
then @xmath73 * the analogous compatibility with @xmath57 .
we can imagine what is a free @xmath0prop .
it consists from all `` free '' words of the following two forms : @xmath74 and @xmath75 we can draw these free elements as `` two - sided trees '' , see figure 2 : the valences of the vertices are greater or equal than 3 .
we have trees of `` two different colors '' according to the equations ( [ eq23.1])-([eq23.2 ] ) above . the definition of ( pre-)@xmath0prop is motivated by the geometry of the kontsevich spaces @xmath2 ( see the next section ) . the reader who is interested in the origin of this definition can pass directly to section 2 .
we tried to construct a compactification of these spaces and to formalize the operations among the strata .
the advantage of @xmath0props is that the chain complex of the compactification @xmath76 is a free dg @xmath0prop .
its homology @xmath0prop is exactly the @xmath0prop @xmath77 controlling the ( co)associative bialgebras .
finally , any ( co)associative bialgebra structure on a vector space @xmath4 gives a map of the pre-@xmath0props @xmath78 .
here we define the pre-@xmath0prop @xmath3 for a vector spaces @xmath4 .
we use here the notations @xmath79 , @xmath80 , and @xmath81 .
we set : @xmath82 we should now define the compositions @xmath83 , and @xmath63 .
the case of the composition @xmath84 can be schematically shown as follows ( see figure 3 ) : let @xmath85 , @xmath86 .
their composition @xmath87 is defined as @xmath88 ( here @xmath89 stands at the @xmath90th place ) .
the picture for the composition @xmath91 is the following ( see figure 4 ) : for @xmath92 and @xmath93 , their composition @xmath94 is @xmath95 we have the particular case of the product @xmath96 when @xmath97 .
denote it by @xmath98 .
by definition , the composition @xmath99 is the @xmath100th tensor pover of the composition @xmath98 .
analogously we define @xmath101 and the composition @xmath102 as the @xmath103th tensor power of the composition @xmath101 .
it remains to define the composition @xmath63 .
we define the composition @xmath104 .
suppose @xmath105 , and @xmath106 we are going to define their composition @xmath107 .
denote @xmath108 next , we define a map @xmath109 as follows : @xmath110 now we set @xmath111 it is clear that these compositions define a pre-@xmath0prop structure on @xmath3 .
m. markl communicated to the author that our composition @xmath63 is a particular case of his `` fractions '' composition [ m2 ] .
[ [ section-2 ] ] for a ( pre-)@xmath0prop @xmath112 we define an @xmath112-algebra structure on a vector space @xmath4 as a map of pre-@xmath0props @xmath113 .
we are going to construct now a @xmath0prop @xmath77 such that a @xmath77-algebra structure on @xmath4 is exactly a ( co)associative bialgebra structure on @xmath4 .
let @xmath39 be the symmetric group on @xmath103 points , and for a group @xmath114 denote by @xmath115 the dual group .
we can consider @xmath77 as @xmath0prop of sets , or , if we like , as the corresponding @xmath0prop of vector spaces ( generated by these sets ) .
we here consider @xmath77 as a @xmath0prop of sets .
later it will appear also as the homology @xmath0prop of the topological @xmath0prop @xmath76 , then we consider it as the corresponding @xmath0prop of vector spaces .
this functor replaces the direct product @xmath116 to the tensor product @xmath117 .
we set : @xmath118 the @xmath0prop maps @xmath119 and @xmath120 are the identity maps .
we define the compositions @xmath96 , @xmath57 , @xmath58 , @xmath59 and @xmath63 as follows : consider any of these compositions for the pre-@xmath0prop @xmath3 , a composition @xmath121 .
suppose that @xmath122 and @xmath123 are its arguments .
these compositions were defined in the previous subsection .
there is the @xmath124-action on @xmath125 and the @xmath126-action on @xmath127 .
suppose @xmath128 , and @xmath129 .
we are going to define the composition @xmath130 in @xmath77 .
for this , consider the composition @xmath131 .
it is clear that it is equal to the action of some @xmath132 on the product of @xmath133 and @xmath89 in @xmath134 : @xmath135 the last equation holds for any @xmath133 and @xmath89 in the corresponding components of @xmath134 , that is , @xmath136 does not depend on the choice of @xmath137 and @xmath89 .
we define the composition @xmath130 as @xmath136 .
it clear that this definition is correct , and in this way we define a @xmath0prop @xmath77 .
a map @xmath138 of pre-@xmath0props is the same that a ( co)associative bialgebra structure on @xmath4 .
first , let @xmath4 be a ( co)associative bialgebra with the product @xmath139 and the coproduct @xmath140 .
we define a map of pre-@xmath0props @xmath141 .
let @xmath142 .
we put @xmath143 . here in the formula
@xmath144 and @xmath145 are the composition powers of the coproduct and of the product , correspondingly . because of the ( co)associativity , these powers are well - defined .
next , for @xmath146 we set @xmath147 is the @xmath100th tensor power of the map @xmath148 .
analogously , using the product , we define @xmath149 on @xmath150 .
now we explain why without the compatibility ( [ eq230.1 ] ) this definition would be incorrect .
consider many identity permutations : @xmath151 , @xmath152 , @xmath153 , and @xmath154 .
than we have the following identity in @xmath77 : @xmath155 it is clear that this identity in @xmath77 follow some identity in the images of these elements @xmath156 by the map @xmath149 of pre-@xmath0props .
the reader can easily verify that this identity is exactly the compatibility ( [ eq230.1 ] ) in a ( co)associative bialgebra .
one can prove also that if the compatibility holds , the definition of @xmath149 is correct .
vice versa , suppose we have a map @xmath157 of pre-@xmath0props .
denote @xmath158 and @xmath159 .
the compatibility follows from ( [ eq232.8 ] ) .
the reader can easily find analogous identities in @xmath77 which imply the associativity of @xmath139 and the coassociativity of @xmath140 .
[ [ section-3 ] ] first of all , recall the definition of the spaces @xmath160 due to maxim kontsevich ( see also [ sh ] ) .
we show in the sequel that these spaces and its compactification introduced below play a crucial role in the deformation theory of ( co)associative bialgebras .
first define the space @xmath161 . by definition ,
@xmath162 , @xmath163 , and @xmath164 here we denote by @xmath165 and by @xmath166 two different copies of a real line @xmath167 .
next , define a 3-dimensional group @xmath168 acting on @xmath161 .
this group is a semidirect product @xmath169 ( here @xmath170 ) with the following group law : @xmath171 where @xmath172 .
this group acts on the space @xmath161 as @xmath173 in other words , we have two independent shifts on @xmath165 and @xmath166 ( by @xmath174 and @xmath175 ) , and @xmath30 dilatates @xmath165 by @xmath28 and dilatates @xmath166 by @xmath176 . in our conditions
@xmath177 , the group @xmath168 acts on @xmath161 freely .
denote by @xmath160 the quotient - space .
it is a smooth manifold of dimension @xmath178 .
we will need also a very special case of the spaces @xmath179 introduced below .
recall here our definition of the space @xmath179 ( generalizing the kontsevich space @xmath160 ) from [ sh ] : fist define the space @xmath180 . by definition ,
@xmath181 here @xmath182 are copies of the real line @xmath167 .
now we have an @xmath183-dimensional group @xmath184 acting on @xmath180 .
it contains @xmath185 independent shifts @xmath186 and _ one _ dilatation @xmath187 this group is isomorphic to @xmath188 .
we say that the lines @xmath189 ( corresponding to the factor @xmath28 ) are the lines of the first type , and the lines @xmath190 ( corresponding to the factor @xmath29 ) are the lines of the second type . denote @xmath191 we construct a compactication @xmath76 the boundary strata of which are products of the spaces @xmath192 , @xmath193 , and @xmath194 ( it allowed to be several spaces of each type ) .
let @xmath195 .
then the space @xmath196 is 1-dimensional .
it is easy to see that @xmath197 is preserved by the action of @xmath168 , and it is the only invariant of the @xmath168-action on @xmath196 .
therefore , @xmath198 .
there are two `` limit '' configurations : @xmath199 and @xmath200 .
therefore , the compactification @xmath201 $ ] .
see figure 1 .
we will construct a compactification of the space @xmath2 which is a topological @xmath0prop .
more presisely , @xmath202 the compactifications will be defined below .
first of all , let us describe all strata of codimension 1 in @xmath76 .
there are boundary strata of codimension 1 of two different types .
the first two strata are shown in figures 3 and 4 . in the picture in figure
3 @xmath203 points on the upper line move infinitely close to each other , and the `` scale '' of this infinitely small number is irrelevant ( we have in mind here the croc compactification from [ sh ] where this scale is relevant ) , @xmath204 . in figure 4
@xmath205 points on the lower line move close to each other , @xmath206 .
the remaining stratum of codimension 1 in @xmath76 ( there is the only one such stratum ) is shown in figure 5 : here in the figure all points on the upper line move infinitely far from each other _ will a finite ratio of any two among these infinite distances_. the all distances on the lower line are finite .
of course , the stratum when the points on the lower line are infinite and the points on the upper line are in finite distances from each other , is the same : one stratum can be obtained from another by the application of the element @xmath207 .
the strata in figures 3 and 4 are isomorphic to @xmath208 and @xmath209 , correspondingly .
the stratum in figure 5 is isomorphic to @xmath210 .
let us explain the last formula : the distances between the points at the upper line are infinite , but their ratios are finite .
therefore , we should count these ratios . for this
, we apply to figure 5 the transformation @xmath207 .
then the infinite distances will become finite , and then we can count the ratios .
now we claim that using the 3 operations shown in figures 3,4,5 we can obtain _ any _ limit configuration ( here by a limit configuration we mean a configuration where some distances are infinitely large or / and infinitely small ) .
moreover , we can apply the configuration in figure 5 not more than 1 time .
let us explain it : apply a transfotmation from @xmath168 such that there are no infinite distances at the lower line , and the diameter of the configuration of points at the lower line is finite ( not infinitely small ) .
we distinguish the following two cases : in the first case in the obtained configuration there are no infinitely large distances in the upper line , and in the second some distances are infinitely large .
it is clear that we can reach any limit configuration of the first type by the applying several times the degenerations shown in figure 3 and figure 4 .
analogously , in the second case , we first apply degenerations in figure 3 ( several times ) , then apply the transformation from figure 5 ( with the scale of the infinity depending on the configuration ) , and then apply several times the degenerations from figure 4 .
it is clear that in this case we can get _ any _ limit configuration .
denote the operations shown in figures 3,4,5 by @xmath96 , @xmath57 , and @xmath63 , correspondingly .
it remains to define our operations @xmath58 and @xmath59 in the definition in section 1.1 , that is , to compactify the spaces @xmath211 and @xmath212 .
notice that the space @xmath213 is isomorphic to the stasheff polyhedron @xmath214 for _ any _ @xmath103 , as well the space @xmath215 is isomorphic to the stasheff polyhedron @xmath216 for any @xmath100 .
in particular , we define our maps @xmath217 and @xmath218 as the identity maps .
furthermore , we compactify these spaces as usual in the stasheff compactification , and the compositions @xmath58 and @xmath59 are defined in the natural way .
moreover , it is clear that the formulas ( a ) and ( b ) in the definition of @xmath0prop hold .
thus , we constructed a topological @xmath0prop @xmath76 . *
the corresponding chain dg @xmath0prop ( formed by the chain complexes of the spaces in our stratification ) is a free @xmath0prop of graded vector spaces ( when we forget about the differential ) , * the corresponding homology @xmath0prop is the @xmath0prop of bialgebras @xmath77
\(i ) follows from the fact that any limit configuration can be witten as the composition of the degenerations in figures 3,4,5 in a unique way , ( ii ) follows from the fact that all spaces @xmath76 are contractible ( and , therefore , have only 0-th nontrivial homology ) . in our compactification ,
the left picture in figure 1 is @xmath219 , and the right picture is @xmath220 .
now we are ready to introduce our main object a dg lie algebra . |
low - mass star formation is believed to occur predominantly in molecular cloud cores .
the commonly accepted scenario is that dense molecular clouds are supported against self - gravity by magnetic and thermal stresses .
since the gas is only weakly ionized , the clouds are not in a strict steady state : the prevalent neutral molecules , pulled in by the force of gravity , gradually drift inward through the magnetic field ( attached to the ionized component and anchored in the cloud envelope ) in a process known as ambipolar diffusion .
when the mass - to - flux ratio in the cloud core exceeds a critical value , the core undergoes gravitational collapse to form a central protostar .
the core evolution after the collapse is initiated can in general be divided into two phases : a rapid initial dynamical contraction , which proceeds nonhomologously and eventually results in the formation of a central point mass , followed by an accretion phase , during which the central mass gradually increases and its gravitational field dominates a progressively larger region around the origin .
the evolution of star - forming molecular clouds has been studied mostly through numerical simulations ( see , e.g. , @xcite for a review ) . until recently
, these calculations were terminated before the end of the dynamical - collapse phase , basically at the point where radiative trapping near the center started to invalidate the isothermality assumption adopted in these computations .
however , @xcite ( @xcite , hereafter ck ) circumvented this problem by treating the comparatively small region of radiative trapping as a central sink cell ; this enabled them to model the collapse of nonrotating magnetic cloud cores through point - mass formation ( pmf ) and into the protostellar accretion phase .
concurrently , @xcite ( @xcite , hereafter cck ) obtained semianalytic self - similar solutions that explicitly accounted for the effects of ambipolar diffusion and successfully reproduced the main qualitative features of the numerical simulations of ck .
our goal in this paper is to extend the previous work on the post - pmf evolution of collapsing magnetic cloud cores by including the effects of rotation .
molecular line observations ( e.g. , @xcite ; @xcite ) have established that a majority of dense ( @xmath9 ) cloud cores show evidence of rotation , with angular velocities@xmath10 that tend to be uniform on scales of @xmath11 , and with specific angular momenta in the range@xmath12 .
although present , the rotation contributes only a small fraction ( typically no more than a few percent ) to the dynamical support of these cores , and self - gravity is mostly balanced by magnetic and thermal stresses . during their quasistatic contraction phase ,
the cores evidently lose angular momentum by means of magnetic braking the magnetic transfer of angular momentum to the ambient gas through torsional alfvn waves and this process also tends to align their angular velocity and large - scale magnetic field vectors ( e.g. , * ? ? ?
once dynamical collapse is initiated and a core goes into a near free - fall state , the specific angular momentum is expected to be approximately conserved , resulting in a progressive increase in the centrifugal force that eventually halts the collapse and gives rise to a rotationally supported disk .
these expectations are consistent with the results of molecular - line interferometric observations , which have probed contracting cloud cores on scales @xmath13 ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
these observations have revealed that , on scales @xmath14 , the cores have a flattened , thick - disk morphology and a velocity field that is dominated by infall motions ( with only a modest rotational component ) .
this morphology is consistent with numerical simulations of magnetically supported clouds , which have demonstrated that the gas rapidly contracts along the field lines and maintains force equilibrium along the field even during the collapse phase ( e.g. , @xcite ; @xcite ; ck ) , including in cases where the clouds are initially elongated in the field direction ( e.g. , * ? ? ?
* ; * ? ? ?
the observations have also established that angular momentum is by and large conserved in the infalling gas ( e.g. , * ? ? ? * ) and leads to the formation of rotationally supported disks on scales @xmath15 .
the pre - pmf collapse of rotating , magnetized , weakly ionized cloud cores was previously studied by @xcite .
they found that little angular momentum is lost during the dynamical - collapse phase but that the centrifugal force nevertheless does not become important prior to pmf . in the current work we focus on the post - pmf phase of the collapse , which enables us to model the formation of circumstellar disks in the context of magnetic cloud cores and to study their properties and their role in the buildup of the central protostar .
circumstellar disks have been detected in @xmath16 of pre main - sequence stars in nearby dark clouds ( e.g. , * ? ? ?
* ) , and it is likely that most of the mass assembled in a typical low - mass young stellar object ( yso ) is accreted through a disk ( e.g. , * ? ?
the study of protostellar disks is important also because they are the incubators of planetary systems , so their properties are directly relevant to the process of planet formation ( e.g. , * ? ? ?
the rotational properties of the collapsing core are evidently also of particular relevance to the question of multiple - star formation ( e.g. , * ? ? ?
in the ensuing discussion we assume that the collapsing core does not give rise to more than one stellar object ; we nevertheless briefly address the issue of core fragmentation in [ discussion ] .
the studies of ck and cck demonstrated that ambipolar diffusion , which is unimportant during the dynamical pre - pmf collapse phase , is `` revitalized '' after a point mass starts to grow in the center and leads to a decoupling of the magnetic flux from the inflowing gas .
it was found that the `` decoupling front '' propagates outward in the form of a hydromagnetic shock ( as originally suggested by @xcite ) and that this process can go a long way toward resolving the magnetic flux problem in star formation ( the several - orders - of - magnitude discrepancy between the empirical upper limit on the magnetic flux of a protostar and the flux associated with the corresponding mass in the pre - collapse core ) . by incorporating rotation into the picture
, we can study how this issue is affected by the formation of a disk .
furthermore , as our model includes angular momentum transport ( by the large - scale magnetic field ) , which is crucial for enabling mass to accumulate in the center and form a protostar ( see [ formulation ] ) , we are also in a position to address the angular momentum problem in star formation ( analogous to the magnetic flux problem , except that the quantity under consideration is the specific angular momentum ) .
the emergence of a quasi - stationary , rotationally supported disk from a nearly freely falling , collapsing core inevitably involves a strong deceleration of the inflowing matter , which almost invariably takes place in a shock .
this `` centrifugal '' shock is distinct from the ambipolar - diffusion shock mentioned above : it is typically hydrodynamic , rather than hydromagnetic , in nature , and it is generally located at a different distance from the center .
the appearance of a second shock increases the complexity of the problem , and it might seem that it could only be tackled through a multi - scale numerical simulation .
however , we have found that it is possible to treat this problem semianalytically by generalizing the self - similar solutions presented in cck to include the effects of rotation and magnetic braking .
although this treatment is somewhat less detailed than a numerical simulation , the semianalytic approach allows us to more readily explore the fairly extensive parameter space .
this approach is supported by the strong evidence from numerical simulations that a multiscale core collapse naturally leads to a self - similar evolution .
in particular , this behavior has been found to describe the effectively magnetic flux- and angular momentum - conserving _
pre_-pmf collapse phase of a rotating magnetic core seen in the aforementioned numerical simulations of @xcite ( @xcite ; @xcite ; see also @xcite and @xcite ) .
as regards the post - pmf evolution , @xcite obtained similarity solutions for flattened cloud cores that approximate previous numerical simulations of the collapse of rotating clouds through the moment when a central point mass could be expected to form .
the good correspondence between the cck solutions and the nonrotating magnetic collapse simulations of ck further strengthens our confidence in the viability of this approach . in relating this work to previously published results , we note that neither the @xcite model nor the numerical simulations with which they compare their solutions incorporate an angular momentum transport mechanism , and therefore they do not in practice give rise to a central point mass .
( nevertheless , as we discuss in [ nonmagnetic ] , their post - pmf solutions correspond to a limiting case of the more general model considered in this paper . )
it is also worth mentioning in this connection that self - similar models of the gravitational collapse of rotating , disk - like mass distributions that incorporate _ viscous _ angular momentum transport have been constructed by @xcite and @xcite .
however , the former model arbitrarily imposes an inner boundary condition that does not allow mass to accumulate at the center , whereas the latter model is restricted to the case in which the inflow speed is everywhere so low that no centrifugal shock forms .
neither of these models therefore provides a realistic framework for studying star formation in rotating molecular cloud cores .
the plan of this paper is as follows . in [ model ] we formulate the problem , discuss the adopted approximations , and describe the self - similar model . in [ results ]
we present illustrative solutions , ordered by increasing level of complexity , which isolate different aspects of the full problem .
we use the flexibility afforded by our semianalytic approach to examine the dependence of the results on the efficiency of ambipolar diffusion and magnetic braking and to elucidate the interplay between these two processes .
we give a general discussion of these solutions and of their astrophysical implications in
[ discussion ] , where we also consider the production of centrifugally driven disk outflows .
our conclusions are summarized in [ summary ] .
our basic model setup is the same as in cck .
we consider an isothermal , disklike core that contracts in the radial direction while maintaining a vertical hydrostatic equilibrium at all times . as noted in [ introduction ] , this behavior was found in numerical simulations of isothermal cores that are initially supported against gravitational collapse by a large - scale , ordered magnetic field .
nonmagnetic simulations ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) have established that rotation could further contribute to the tendency of a core to flatten as it contracts so long as the specific angular momentum is approximately conserved . in restricting gas motions in our model to the plane of the flattened core
, we preclude the possibility that matter can reach the center from above ( or below ) the disk , and this , in turn , implies that a point mass can form in the center only if there exists a mechanism for removing angular momentum from the inflowing gas .
there have also been models in which the complementary view has been adopted , viz .
, new mass is added to the protostar and its circumstellar disk only from above and below the disk plane ( e.g. , * ? ? ?
* ; * ? ? ?
these models tend to produce rotationally supported disks whose masses are comparable to that of the central star , even if angular momentum transport within the disk is taken into account ( e.g. , * ? ? ?
* ; * ? ? ?
however , most observed protostellar disks in nearby dark clouds have masses ( estimated from dust emission or , when possible , from a direct measurement of the disk velocity field ) that are significantly smaller ( @xmath17 ) than the protostellar mass ( e.g. , * ? ? ?
* ) . as we demonstrate in [ results ]
, such low - mass systems can form naturally when mass flows into the central region primarily in the disk plane .
given that this is the expected situation for magnetically regulated core contraction and that the disklike configurations predicted in this case are consistent with observations of collapsing cloud cores , we consider this approximation to be adequate for capturing the essence of protostellar - disk and star formation .
the isothermality assumption was justified in the work of ck and cck on the grounds that it generally applies so long as the density remains below @xmath18 , which for a typical nonrotating collapse is the case on scales @xmath19
. this assumption is less tenable in the present work because , when a quasi - stationary , centrifugally supported disk forms , it has a significantly higher column density at a given radius than a nearly freely falling configuration , with the result that radiative trapping already occurs on scales @xmath15 ( e.g. , * ? ? ?
irradiation by the central protostar tends to mitigate this effect and establish vertical isothermality in the outer regions of the disk , but , in turn , it leads to a moderately strong ( @xmath20 ) variation of the disk temperature with radius ( e.g. , * ? ? ?
* ) . however , since thermal stresses turn out not to play a major role in the core - collapse dynamics , the assumption of isothermality ( which is required for self - similarity ) is not likely to introduce significant inaccuracies into the results .
another requirement for self - similarity is that the ion density @xmath21 be proportional to the square root of the neutral mass density ( which is essentially equal to the total gas density @xmath22 ) . in this case , the ratio @xmath23 of the neutral ion momentum - exchange time scale @xmath24 to the nominal self - gravity free - fall time @xmath25 is independent of density and can be treated as a constant . is smaller than the one used in cck by a factor @xmath26 . ]
this behavior is found at comparatively low densities , when molecular ionization by cosmic rays ( at a rate @xmath27 ) is balanced by rapid dissociative recombination of the molecular ions and the dominant positive charges are metal ions ( formed by a charge - exchange reaction between molecular ions and neutral metal atoms and destroyed by recombination on grain surfaces ) . for grains of radius @xmath28 , a temperature @xmath29 , and @xmath30
, this scaling applies roughly at neutral particle densities between @xmath31 and @xmath32 ( e.g. , * ? ? ?
* ; * ? ? ?
these values of @xmath33 and @xmath34 characterize the outer regions of collapsing cores , typically on scales @xmath35 ( e.g. , ck ) . in this regime , @xmath36 .
the numerical models of ck extended to scales @xmath37 , where densities @xmath38 were attained . in the case of a rotationally supported disk , even higher densities are expected on these scales . at such high densities , grains are typically the dominant carriers of both positive and negative charges , and their densities also scale as @xmath39 ( e.g. , * ? ? ?
assuming , for the sake of illustration , that the charged grain distribution is dominated by a population of small ( pah - like ) singly charged particles of radius @xmath40 ( see * ? ? ?
* ) , we can deduce the charged - grain number density @xmath41 from the ionization balance relation @xmath42 where @xmath43 ( with @xmath44 ) is the reduced grain mass .
taking as fiducial values @xmath45 and @xmath46 at a distance @xmath47 from the center of a rotationally supported protostellar disk , we can neglect the factor 1 in comparison with the second term in the last parentheses on the right - hand side of equation ( [ charge ] ) ( which accounts for the electrostatic attraction between the recombining grains ) .
we then get @xmath48 . assuming that the small grains are characterized by the same rate coefficient for momentum transfer through elastic scattering with neutrals as metal ions ( @xmath49 ; see @xcite ) , we obtain @xmath50 in this density regime . ambipolar diffusion will dominate ohmic diffusivity so long as the grain hall parameter @xmath51 ( the ratio of the cyclotron frequency to the collision frequency with neutrals ) exceeds 1 . adopting a fiducial magnetic field strength at @xmath52 of @xmath53 , we find @xmath54 , which we expect to climb securely above 1 as the density continues to decrease with increasing @xmath55 .
we thus infer that both the low- and the high - density regimes of our model collapsing cores correspond to ambipolar diffusion that is amenable to self - similar scaling
. the intermediate density regime will depart from this scaling ( e.g. , * ? ? ?
* ) , and it is also likely to involve significant contributions from the hall current term in ohm s law ( e.g. , * ? ? ?
* ; * ? ? ?
nevertheless , given that our bracketing values of the parameter @xmath23 vary by only one order of magnitude ( assuming that @xmath56 lies in the range @xmath57 ) across @xmath58 orders of magnitude in density , we may expect our diffusivity parameterization to yield qualitatively meaningful results on all scales between @xmath59 and @xmath60 .
just as in cck , we take the initial ( @xmath61 ) state of our model core to coincide with the endpoint of the pre - pmf collapse phase , which is characterized by a radial scaling of the surface density @xmath62 and the vertical magnetic field component @xmath63 @xmath64 and by spatially constant radial inflow velocity @xmath65 ( where @xmath66 is the isothermal speed of sound ) and mass accretion rate @xmath67 . in the presence of rotation
, this state also exhibits a uniform azimuthal speed . since the latter is typically much smaller than the speed of sound @xmath68 , we normalize our initial state as in cck based on the nonrotational collapse simulations of ck ( @xmath69 and @xmath70 ) .
we determine the initial value of @xmath71 ( which we denote by @xmath72 ) from the expression @xmath73 ( see * ? ? ?
* ) , where @xmath74 and @xmath75 are , respectively , the uniform background angular velocity and magnetic field strength . based on the range of measured core angular velocities ( see [ introduction ] ) , @xmath72 could be a factor @xmath76 larger or smaller than the fiducial value adopted in equation ( [ v0 ] ) .
we assume that magnetic braking , which regulates the core rotation prior to its collapse , remains the dominant angular momentum transport mechanism also during the subsequent evolution .
[ a centrifugally driven disk wind may , however , dominate the angular momentum transport in the rotationally supported disk . in
[ results ] we argue that this is , indeed , a likely possibility , and in appendix [ wind ] we show that this mechanism can be incorporated into our model without significantly modifying the basic formulation .
] we adopt the approach articulated by @xcite for the pre - pmf collapse phase .
the torque per unit mass on the slab - like core ( of surface mass density @xmath62 ) can be approximated by @xmath77 , where @xmath63 is the vertical field component at the midplane and @xmath78 is the azimuthal field component at the surface @xmath79 ( assuming @xmath80 ) .
the latter can be estimated from the relation @xmath81 where @xmath82 is the magnetic flux threading the core within a radius @xmath55 , @xmath83 is the azimuthal speed of the core ions , and @xmath84 is the speed with which the torsional alfvn waves that effect the magnetic braking propagate in the external medium . equation ( [ bm ] ) is the same as equation ( 26 ) in @xcite , where @xmath84 is also treated as a constant , except that we substituted the ion speed for the bulk ( neutral ) speed to take account of the possibility that the ions ( into which the magnetic field is frozen ) are not well coupled to the neutral core component ( see @xcite and @xcite ) . the assumption of a constant @xmath84 is consistent with theoretical models of magnetically supported clouds as well as with empirical data , which indicate that @xmath85 in molecular clouds over at least 4 orders of magnitude in density ( @xmath86 ; e.g. , @xcite ) .
this assumption is required for the construction of a self - similar model . in order for the steady - state expression ( [ bm ] ) to be applicable , the alfvn travel time in the external medium over the initial radius of the cloud should be less than the evolutionary time scale @xmath87 ( which , for a rotationally supported disk , is essentially the local magnetic braking time ) .
our representative rotationally supported disk models typically have @xmath88 ( with @xmath89 as @xmath90 ) , whereas , for a nominal temperature of @xmath91 , @xmath92 .
these estimates indicate that the assumption of rapid braking should not lead to gross errors even if it is not everywhere strictly correct . in our models
we use the parameterization @xmath93 , with @xmath94 adopted as a typical value
. we will be interested in evaluating the effects of magnetic braking in regions where rotation already plays a dynamically significant role , so that @xmath95 .
we can therefore drop @xmath74 in equation ( [ bm ] ) , which will make it possible to incorporate this equation into a self - similar formulation .
the anticipated increase in @xmath96 with decreasing @xmath55 ( @xmath97 when a keplerian disk is formed ) will in general make @xmath78 the dominant field component at the surface as the central point mass is approached .
( this situation is not encountered in the pre - pmf phase ; see @xcite . )
we expect , however , that various magnetohydrodynamic instabilities ( in particular , internal kinks ) will intervene to prevent the azimuthal field component from greatly exceeding the poloidal components .
, we also do not expect the radial surface field component to greatly exceed the vertical field @xmath63 under realistic circumstances .
] we therefore impose a cap on the azimuthal field in the form @xmath98 . in our models
we usually set @xmath99 this choice , in fact , also corresponds to the typical value of @xmath100 obtained in our model for a rotationally supported , diffusive disk when ( as expected ) the vertical angular momentum transport is dominated by a centrifugally driven wind ( see [ discussion ] and appendix [ wind ] ) .
we represent the collapsing core as a thin disk surrounding a point mass @xmath4 .
we use cylindrical coordinates @xmath101 centered on the point mass , with the disk midplane given by @xmath102 .
the disk has a column density @xmath62 , mass density @xmath22 , and half - thickness @xmath103 defined by @xmath104 .
the total mass @xmath105 enclosed within the radius @xmath55 is then given by @xmath106 , and we denote its time derivative by @xmath107 .
the disk velocity field @xmath108 has radial and azimuthal components , with the latter giving rise to a specific angular momentum @xmath109 .
we assume that the disk is threaded by an open magnetic field that is symmetric with respect to the midplane [ i.e. , @xmath110 at @xmath102 and @xmath111 at the disk surfaces ( @xmath112 ) , and similarly for @xmath113 .
the magnetic flux enclosed within the radius @xmath55 is given by @xmath114 , where @xmath115 denotes the flux trapped inside the central point mass . in appendix [ diskequations ]
we write down the basic disk equations , integrate them over @xmath116 while retaining all terms of order @xmath117 , and then go on to justify the omission of some of the @xmath118 terms in the interest of greater simplification .
the result of these manipulations is the following set of equations , @xmath119
@xmath120 @xmath121 and @xmath122 which express mass , radial momentum , and angular momentum conservation , and vertical hydrostatic equilibrium , respectively . under the assumption that the magnetic field evolution is governed by ambipolar diffusion , one can regard the magnetic field lines as being frozen into the ions .
the latter move with velocity @xmath123 and drift with respect to the bulk ( neutral ) gas component at the drift velocity @xmath124 . by balancing the lorentz force on the ions with the ion
neutral collisional drag force , one can solve for the components of @xmath125 : @xmath126 we then have @xmath127 as in cck , we adopt the monopole approximations for @xmath128 and @xmath129 , @xmath130 which considerably simplify the calculations and are not expected to introduce any significant errors ( see also @xcite , @xcite , ck , and @xcite ) . for the azimuthal magnetic field component
, we use equation ( [ bm ] ) in the limit @xmath131 together with equation ( [ phidrift ] ) , and impose the bounding condition on @xmath132 discussed at the end of [ formulation ] .
we then obtain @xmath133\,.\ ] ] we introduce a similarity variable @xmath134 and a set of nondimensional flow quantities that depend only on @xmath134 : @xmath135 @xmath136
@xmath137 @xmath138 @xmath139 @xmath140 these definitions extend the set used in cck to the rotational case ( which involves the dimensionless azimuthal speed @xmath141 and specific angular momentum @xmath142 ) note , however , that the variables @xmath143 , @xmath144 , @xmath145 , and @xmath146 in the present paper were denoted by @xmath147 , @xmath141 , @xmath40 , and @xmath148 , respectively , in cck . to help relate our results to real protostellar systems
, we observe that , for our fiducial value of @xmath66 ( @xmath149 ) , @xmath150 corresponds to a distance @xmath151 ( @xmath152 ) when @xmath153 ( the characteristic age of a class 0 yso ) and to a distance @xmath154 ( @xmath155 ) when @xmath156 ( the characteristic age of a class i yso ) .
cck demonstrated that ambipolar diffusion can be incorporated into a self - similar model when @xmath157 . when rotation is present , we find that magnetic braking can similarly be included if @xmath158 .
the nondimensional model parameters that control the strength of these two effects are @xmath23 and @xmath159 , respectively ( see [ formulation ] ) .
[ a self - similar formulation is also possible if angular momentum transport is due to an `` @xmath159 viscosity , '' @xmath160 @xcite , when @xmath66 and the parameter @xmath161 ( @xmath162 ) are constant .
this was noted by @xcite in the special case where @xmath103 has the value appropriate to a keplerian disk , and by @xcite in the case when @xmath103 has the value appropriate to a self - gravitating disk ( although the latter authors required @xmath163 , rather than @xmath161 , to be a constant ) . ] with the above expressions for the similarity variables , and defining also @xmath164 for convenience , the structure equations can be rewritten in the following nondimensional , self - similar form : @xmath165 @xmath166 @xmath167 @xmath168 @xmath169 @xmath170 @xmath171 @xmath172 @xmath173 @xmath174 and @xmath175 the disk half - thickness @xmath143 can be found from the vertical hydrostatic equilibrium condition , which yields @xmath176 the solution of this quadratic equation is @xmath177\ , \ ] ] where @xmath178 and @xmath179 .
equation ( [ thicknesssolution ] ) implies that @xmath180 , @xmath181 , and @xmath182 $ ] in the limits where self - gravity , central - mass gravity , and magnetic stresses , respectively , dominate the vertical squeezing of the disk . ) , except that they omitted the magnetic terms .
their solution for @xmath143 thus has the same form as equation ( [ thicknesssolution ] ) , but with @xmath183 and @xmath184 replaced by @xmath185 and @xmath145 , respectively . ] as discussed in [ formulation ] ,
the initial ( @xmath61 ) conditions [ which in the self - similar model also represent the outer asymptotic ( @xmath186 ) values ] correspond to a collapsing core just before pmf .
thus we require @xmath187 where @xmath188 , @xmath189 , @xmath190 , and @xmath72 are constants . as discussed in
[ formulation ] , our fiducial values for @xmath188 ( @xmath191 ) and @xmath190 ( @xmath192 ) are the same as those used by cck ; they are based on the results of numerical simulations of nonrotating cores and are compatible with observations .
we similarly adopt the cck choice @xmath193 .
our reference values for the uniform initial rotation @xmath72 are based on the range of measured core angular velocities and typical cloud parameters that enter into the estimate ( [ v0 ] ) .
we have studied three distinct cases of rotational core collapse , which we present in this section in order of increasing complexity .
they are : ( 1 ) a purely hydrodynamical collapse , with no mechanism for angular momentum transport ( [ nonmagnetic ] ) ; ( 2 ) an ideal - mhd ( imhd ) collapse , which incorporates magnetic braking but does not include any magnetic field diffusivity that could prevent the buildup of a central magnetic monopole ( [ mhd ] ) ; ( 3 ) an mhd collapse that includes ambipolar diffusion ( ad ; [ ad ] ) .
the latter model is able to reproduce the basic observational features of rotationally supported protostellar disks and their central ysos .
our models have many features in common , arising from the basic interplay between the centrifugal force and gravity . initially and at large distances from the center ( i.e. , for large values of @xmath134 ) , gravity is stronger than the centrifugal force and matter falls in at a high , supersonic speed . in this regime @xmath194 , and self - gravity dominates over attraction by the central object .
a very different behavior characterizes small values of @xmath134 .
typically , the innermost region constitutes a keplerian accretion disk , where gravity approximately balances the centrifugal force and where the infall speed is low and subsonic .
the transition between the supersonic and subsonic inflow regimes is achieved through a shock , which is located at the point ( @xmath195 ) where the infalling matter encounters a centrifugal barrier .
inward of this shock the column density increases significantly , forming a dense accretion disk .
the location of the centrifugal shock roughly coincides with the _ centrifugal radius _ , which is the largest radius where the gravitational ( @xmath196 ) and centrifugal ( @xmath197 ) accelerations are in approximate balance ( e.g. , * ? ? ?
in fact , we find that the point where the equality @xmath198 is first satisfied as the matter flows in yields a value that is only a slight ( @xmath199 ) overestimate of the actual shock location in both the imhd and ad cases when ( as is the case in our typical models ) @xmath200 . the value of the centrifugal radius is sometimes estimated from the @xmath61 expressions for the surface density and angular momentum profiles
. this estimate could be rather inaccurate since it does not account for the mass accumulation at the center or for the loss of angular momentum that the infalling matter experiences even before it reaches the centrifugal radius .
we discuss more accurate estimates of @xmath195 in the following subsections .
our `` standard '' solutions , corresponding to the fiducial parameter values , are presented in [ mhdstandard ] for the imhd case and in [ adstandard ] for the ad case .
we also present solutions in the limit when the initial rotation is very fast and magnetic braking is either nonexistent ( [ nonmagnetic ] ) or comparatively inefficient ( [ mhdfast ] and [ adfast ] for the imhd and ad cases , respectively ) . under these conditions
the rotationally supported disk becomes fairly extended and its mass greatly exceeds that of the central object .
another `` extreme '' type of a solution is obtained in the opposite limit , when the initial rotation is braked so efficiently that @xmath142 is effectively reduced to zero at a finite value ( @xmath201 ) of @xmath134 . in this limit ,
too , we present solutions for both the imhd ( [ mhdstrong ] ) and ad ( [ adstrong ] ) cases .
these configurations are effectively nonrotating ( and thus resemble the cck solutions ) for @xmath202 , and they do not feature either a centrifugal radius or a keplerian accretion disk ; instead , matter accretes directly onto the central object at a supersonic speed .
this problem was already treated in detail by @xcite ; it is included here for reference and comparison , and to serve as a starting point . in the absence of magnetic fields or of any other means of angular momentum transport ,
the self - similar collapse is characterized by @xmath203 , where @xmath204 is a constant fixed by the initial conditions ( @xmath205 ) . the inner asymptotic solution ( @xmath206 ) is given by @xmath207 and @xmath208 , with @xmath209/(2\phi^2)$ ] .
numerically , we proceed by applying the initial conditions at @xmath210 and integrating the differential equations toward smaller values of @xmath134 until the expected location of @xmath195 ( see eq .
[ [ x_c_nr ] ] below ) is reached .
there we impose the isothermal shock jump condition described in appendix [ isothermalshock ] , and continue the integration until the asymptotic solution is approximately fulfilled .
the exact value of @xmath195 is obtained iteratively by enforcing the matching to this asymptotic solution to a high precision .
figures [ 1 ] and [ 2 ] show profiles of mass , column density , and radial speed for two kinds of initial rotation : a typical ( @xmath211 ) and a rapid ( @xmath212 ) one . in both cases
the centrifugal shock is very abrupt , producing a large increase in column density and a sudden decrease in inflow speed . in the case
with faster initial rotation , the effect of the centrifugal barrier is so strong that it forces the radial velocity @xmath144 to change sign below @xmath195 to @xmath213 ( indicating a backflow ) , as shown in figure [ 2 ] .
the centrifugal shock thus separates regions of infall and backflow ; it stays stationary in self - similar coordinates , implying that it moves outward in physical space .
this phenomenon was already discussed in @xcite . by trying a few other values of @xmath72 ( but keeping the other parameter values unchanged ) , we have found that @xmath214 is too small to produce a backflow but that @xmath215 is already large enough . an interesting behavior exhibited by these solutions ( and standing out particularly in fig . [ 1 ] ) is the development of a mass `` plateau '' when the infall speed starts to increase above its outer asymptotic ( @xmath216 ) value .
this behavior is generic to all except our fast - rotation solutions ( [ mhdfast ] and [ adfast ] ) and can be understood as follows . according to equations ( [ mw ] ) and ( [ mdot ] ) , @xmath217 and @xmath218 .
these equations are consistent with the outer asymptotic solution @xmath219 and @xmath220 ( eq . [ [ initialconditions ] ] ) only so long as @xmath221 ( in which case @xmath222 ) .
this 7ex inequality breaks down when @xmath134 becomes small enough that it drops below @xmath223 .
this point marks the approximate outer edge of the mass plateau , and we label it @xmath224 . its value can be estimated by setting @xmath225 although typically @xmath223 is already somewhat larger than @xmath226 at that point .
the plateau mass can then be approximated by @xmath227 for @xmath228 , @xmath229 , and hence @xmath230 and ( using eq .
[ [ msigma ] ] ) @xmath231 .
the latter inequality accounts for the appearance of the plateau in the logarithmic plots of @xmath232 vs. @xmath134 .
if @xmath233 is sufficiently small , then , in the plateau regime , @xmath223 approaches the free - fall value corresponding to @xmath234 .
hence , once @xmath223 comes to exceed @xmath134 , their ratio increases rapidly with decreasing @xmath134 , and the mass plateau is quickly established .
the derivation of equations ( [ x_pl ] ) and ( [ m_pl ] ) is independent of the degree of magnetization of the flow , and the expression ( [ m_pl ] ) therefore also provides an excellent first guess of @xmath185 in our fiducial magnetic solutions that involve point - mass formation .
these results do not , however , apply to the fast - rotation cases considered in [ mhdfast ] and [ adfast ] , where a plateau does not form because the centrifugal shock is established before @xmath134 becomes @xmath235 , and where the central mass is typically smaller than that of the disk . setting @xmath203 and @xmath236 in the expression @xmath237
, we infer @xmath238 we have used this expression to obtain the initial guess in our numerical calculation . for the parameters of figure [ 1 ] , this gives @xmath239 , which is seen to be an excellent approximation to the actual value .
this estimate is only applicable if @xmath240 , i.e. , for @xmath241 .
this condition is still marginally satisfied in the solution presented in figure [ 2 ] .
the inner asymptotic solution of this model precludes the formation of a point mass . as the azimuthal speed @xmath141 can not diverge as @xmath206
, @xmath242 must vanish at the origin . to form a central object with @xmath243 , a braking mechanism that allows @xmath244 to decrease as @xmath206
is required .
the absence of such a mechanism explains the linear decline of @xmath232 with @xmath134 at small values of @xmath134 ( which for the parameters used in figs .
[ 1 ] and [ 2 ] effectively starts already at @xmath195 ) .
it may also be expected that a braking mechanism will lead to a more gradual radial deceleration , reducing or even altogether preventing the backflow exhibited by the rapid - rotation solution shown in figure [ 2 ] . in the following subsections we examine to what extent these expectations
are met in the presence of magnetic braking . in this section
we present an imhd model for magnetic braking based on the formalism described by equations ( [ uequation])([thicknessequation ] ) , in which we set @xmath245 .
the existence of a braking mechanism leads to a qualitative change in the character of the solutions .
in contrast to the nonmagnetic solution presented in [ nonmagnetic ] , @xmath244 is not a constant in this case , and magnetic braking allows a central mass to form despite the initial presence of rotation .
flux freezing under the ideal - mhd assumption implies that @xmath246 and @xmath247 , where @xmath248 , the nondimensional mass - to - flux ratio , is a constant ( equal to its initial value @xmath189 ) .
it is also convenient to introduce the auxiliary constant @xmath249 , which is useful for describing the effect of the magnetic `` dilution '' of the gravitational field @xcite .
there are two qualitatively different sets of asymptotic relations for the inner region ( @xmath206 ) of this model .
the first set , presented in equations ( [ mhdasymptoticbegin])([mhdasymptoticend ] ) below , applies to the configurations described in [ mhdstandard ] and [ mhdfast ] .
the second set , presented in [ mhdstrong ] , applies only in the strong - braking case .
both of these sets describe the formation of a dynamically dominant central mass and a split - monopole field . however , the first set represents a rotationally supported disk , whereas the second one depicts a nonrotating , supersonic inflow .
the disk - like inner asymptotic solution is @xmath250 these equations represent an essentially keplerian disk ; in particular , the expression for @xmath142 differs from that of a purely keplerian system only by the magnetic - dilution factor @xmath251 [ which , however , remains close to 1 ( @xmath252 ) for our fiducial value of @xmath189 ] .
the @xmath142 profile corresponding to such a disk implies that @xmath244 decreases with decreasing @xmath134 . the magnetic braking required to account for this change from the initial distribution fixes the strength and radial power - law index of the magnetic field components @xmath254 and @xmath255 .
the flux - freezing condition then determines the value of @xmath145 , and the mass conservation relation , in turn , yields the expression for @xmath144 .
the behavior of the imhd solutions is determined primarily by the magnitude of the initial rotation and the strength of the magnetic braking .
we consider three qualitatively different cases : a typical collapse in which both effects are relevant over most scales ( [ mhdstandard ] ) ; a limiting case corresponding to weak braking and fast rotation ( [ mhdfast ] ) ; and another limiting case in which very strong braking is capable of eliminating rotation altogether by the time the flow reaches a finite distance from the origin , resulting in the formation of an inner nonrotating - inflow zone ( [ mhdstrong ] ) .
the method used to solve the differential equations is essentially the same as that used in [ nonmagnetic ] . for the initial guess of the value of @xmath195
we again use the estimate ( [ x_c_nr ] ) .
although magnetic braking does operate in this case , the deviation of @xmath244 from its @xmath216 asymptotic value remains relatively small in the range @xmath256 , and the approximation obtained for the nonmagnetic case remains fairly good ( as evidenced by the fact that the value of @xmath195 in the solution presented in fig . [ 3 ] differs by a factor of less than 2 from its value in fig .
the discontinuities in the variables @xmath145 , @xmath254 , and @xmath144 across the shock are estimated using the appropriate shock jump conditions . for the typical case ( [ mhdstandard ] ) the shock occurs in a region where @xmath257 and the `` magnetically squeezed shock '' jump conditions described in appendix [ magneticshock ] are applicable . in the fast - rotation case ( [ mhdfast ] ) the shock is located sufficiently far away from the origin that a generalized ( to include magnetic effects ) isothermal shock ( appendix [ generalizedisothermalshock ] ) provides a better approximation .
( the issue of shock jump conditions is not relevant to the strong - braking inflow considered in [ mhdstrong ] , since no centrifugal shock forms in that case . ) as in [ nonmagnetic ] , a successful convergence of the solution at @xmath258 to its asymptotic ( @xmath206 ) form requires a correct selection of the point @xmath195 , which is achieved iteratively .
( incorrect choices of @xmath195 produce numerical solutions with a spontaneous singularity at some point @xmath259 , similar to those found in @xcite . ) as the value of the central mass is not known before the numerical integrations start , we adopt @xmath260 ( eq . [ [ m_pl ] ] ) as an initial guess that is refined iteratively until convergence is attained . the results for this case are presented in figure [ 3 ] . for the fiducial parameters @xmath211 , @xmath261 , and @xmath262 , rotation and braking are both moderate .
magnetic braking is strong enough to prevent shock formation in the region where self - gravity dominates ( as in [ mhdfast ] ) , but it is not so strong as to prevent the shock from forming altogether ( as in [ mhdstrong ] ) .
the centrifugal shock thus occurs well inside the region where the central mass dominates the gravitational field , with @xmath200 .
the disk that forms for @xmath258 is thus essentially keplerian and has an angular momentum profile @xmath263 .
the large inclination of the magnetic field lines to the disk surface indicated by the inner asymptotic solution ( @xmath264 as @xmath265 ; see eqs .
[ [ bzmhdasymptotic ] ] and [ [ brmhdasymptotic ] ] ) suggests that the disk would likely drive a centrifugal wind from its surfaces , which would reduce @xmath185 .
we neglect this effect in the present discussion , but we consider it in [ discussion ] and appendix [ wind ] in relation to the ambipolar - diffusion fiducial solution .
the evolution of the inflow can be characterized by dividing the range of @xmath134 into four regions , which are , in order of decreasing distance from the center : [ [ self - gravitydominated - region . ] ] self - gravity dominated region . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + for the largest values of @xmath134 , the initial value of the ratio @xmath244 ( @xmath266 ) is still conserved , and @xmath232 , which is still much larger than its central value , is proportional to @xmath134 . [
[ free - fall - region . ] ] free - fall region .
+ + + + + + + + + + + + + + + + + after a transition region , where @xmath244 decreases slightly from its initial value on account of magnetic braking , both @xmath232 and @xmath142 enter a plateau ( at @xmath267 ; see eq .
[ [ x_pl ] ] ) .
they become independent of @xmath134 , with values @xmath268 and @xmath269 ( corresponding to @xmath270 ) .
these computed values can be substituted into the expression @xmath271 to provide a refinement of the estimate ( [ x_c_nr ] ) for the location of the centrifugal shock . as can be seen from a comparison of this value with the computed result ( listed in the caption to fig .
[ 3 ] ) , the approximation is excellent ( as is the correspondence between the value of @xmath272 and that of @xmath185 ) .
near the inner edge of this region , the centrifugal force starts to exceed gravity .
[ [ the - centrifugal - shock . ] ] the centrifugal shock .
+ + + + + + + + + + + + + + + + + + + + + + the outer edge of this region is where the variables @xmath145 , @xmath144 , and @xmath254 undergo the discontinuities prescribed by the shock jump conditions of appendix [ magneticshock ] . just inside the discontinuity
the column density becomes very large , but it then rapidly decreases toward its asymptotic near - keplerian value .
the infall speed decreases rapidly , but not so strongly as to produce any backflow ( i.e. , @xmath144 remains @xmath273 for all @xmath134 ) .
it is in this postshock transition zone that the angular momentum makes the final adjustment to its asymptotic , magnetically diluted , keplerian value : this sudden bout of braking is triggered by the spike in @xmath254 that results from the shock compression .
the column density spike that occurs in this region is associated with a small ( but finite ) mass , @xmath274 , which 5ex could be pictured as a massive ring located at the edge of the keplerian disk . after undergoing fast variations as it transits through this region , the flow merges seamlessly into the next , smoothly varying zone .
[ [ magnetically - diluted - keplerian - disk . ] ] magnetically diluted keplerian disk .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in this region , the asymptotic equations ( [ mhdasymptoticbegin])([mhdasymptoticend ] ) are satisfied .
the rotationally supported disk has a mass @xmath275 of @xmath185 and is in an almost perfect dynamical equilibrium , as the inflow speed is low ( @xmath276 ) .
if the initial azimuthal speed @xmath72 is high and the braking parameters @xmath159 and @xmath277 are moderate or small , then the value of @xmath142 can become quite high , which can lead to a large value of the centrifugal radius @xmath195 . in an extreme case , the centrifugal shock may occur so far away from the center that it is located within the self - gravity dominated region , where @xmath194 . figures [ 4 ] and [ 5 ] show two solutions , corresponding to a fast initial rotation ( @xmath278 ) and two different values of the braking parameter @xmath159 ( 0.01 and 0.1 , respectively ; @xmath279 in both cases ) .
the centrifugal shock in these solutions is so strong that the radial speed @xmath144 changes sign across the shock .
this is similar to the backflow observed for large values of @xmath142 in the absence of braking ( see fig . [ 2 ] ) and is consistent with the expectation that the braking must be strong enough to qualitatively change the behavior of a rapidly rotating inflow .
the backflow region in 4ex this case is clearly seen to occupy only a finite range in @xmath134 and not to reach all the way to the origin : it represents an annulus , interior to which the infall resumes .
the large values of @xmath195 result in extended , and hence fairly massive , disks . on the other hand ,
the comparatively weak braking inhibits mass accumulation at the center : the central masses obtained in these solutions ( listed in the captions to figs . [ 4 ] and [ 5 ] ) are much smaller than the mass derived for the fiducial parameters in [ mhdstandard ] ( which , in turn , is comparable to @xmath185 in the nonrotating collapse model of cck )
. the outer regions of the disks are therefore manifestly non - keplerian .
( the specific angular momentum approaches to within @xmath280 of the asymptotic , diluted - keplerian value given by eq . [ [ jmhdasymptotic ] ] only when @xmath134 decreases below @xmath281 for the solution in fig . [ 4 ] and below @xmath282 for the solution in fig . [ 5 ] . ) as can be seen from a comparison of figures [ 4 ] and [ 5 ] , a larger value of the braking parameter @xmath159 leads to a reduction in @xmath142 and results in the value of @xmath195 becoming smaller and that of @xmath185 larger .
correspondingly , the width of the backflow region is reduced and the keplerian regime extends further out .
these trends continue as the effect of rotation is diminished : by the time the parameters attain their fiducial values , the backflow region has completely disappeared and the inner asymptotic solution starts very close to @xmath195 . in summary , the collapse of fast - rotating cores that lack an exceptionally strong braking mechanism results in fairly massive disks and in comparatively low - mass central objects .
the disks are rotationally supported and largely self - gravitating , with only their innermost regions exhibiting keplerian behavior .
if braking is very efficient then the innermost region of the inflow can have effectively zero angular momentum even if the collapsing core starts out rotating .
this situation is illustrated in figure [ 6 ] , which depicts a solution with a fairly respectable initial rotation ( @xmath212 ) but with extreme values of the braking parameters ( @xmath283 ) , which give rise to a strong surface azimuthal field ( see eq .
[ [ bphiequation ] ] ) . in this case
the angular momentum is reduced effectively to zero at @xmath284 , before a centrifugal shock could occur , and thus no rotationally supported disk is formed . for @xmath202 , @xmath285 , and as @xmath206 , the inflow tends to the asymptotic form of the nonrotating magnetic solution derived by cck ( see their 3.3 ) . in this limit
the flow becomes a supersonic , magnetically diluted free fall onto the central mass : @xmath286^{1/2}\ , , \\
\sigma=\mu b_z&=&[m_c/2(1-\mu^{-2})x]^{1/2}\ , , \\ b_{r , s}=\psi / x^2&=&m_c/(\mu x^2)\ , , \\ h=2\sigma / b_{r , s}^2&=&\mu^2[2/m_c^3(1-\mu^{-2})]^{1/2}\,x^{7/2}\ ,
. \label{mhdstrongasymptoticend}\end{aligned}\ ] ] the numerical solution converges to these asymptotic values already quite close to @xmath201 ( at @xmath287 ) . equation ( [ mhdstrongasymptoticend ] ) for the disk scale height shows the strong effect of magnetic squeezing by the radial magnetic field component in the asymptotic solution .
( since the radial surface magnetic field scales as @xmath288 , it dominates the magnetic terms in both the vertical and the radial force - balance equations . )
the dependence of @xmath143 on @xmath134 ( @xmath289 ) is much stronger than in the solution presented by cck ( @xmath290 ) , in which the effect of magnetic pinching was neglected .
however , apart from this difference , these two solutions are very similar ; in particular , the value of 5ex the central mass quoted by cck ( @xmath291 ) is very close to the value that we derive ( @xmath292 ) .
we note , however , that the extremely rapid decrease of @xmath143 with radius that is implied by this solution is unlikely to be sustained in reality .
for example , when the disk becomes sufficiently thin , the radial field components on opposite sides of the disk might reconnect , resulting in an upper bound on @xmath293 .
furthermore , the decrease in @xmath143 will be arrested once the column density grows to a value for which the disk becomes opaque and the isothermal approximation breaks down .
we do not consider these caveats in more detail since ( as we point out at the beginning of [ ad ] ) we expect _ radial _ field diffusion to intervene ( possibly even before the above effects come into play ) and prevent the imhd configuration from being set up near the center .
using the @xmath206 asymptotic relations for the imhd solution given by equations ( [ mhdasymptoticbegin])([mhdasymptoticend ] ) , it can be readily verified that the nondiffusive terms in the induction equation ( [ psiequation ] ) are @xmath294 whereas the diffusive term ( associated with the magnetic tension force ) is @xmath295 , so that the latter will become dominant near the origin .
it follows that the imhd solution can not apply in the vicinity of the center in cases in which an accretion disk is predicted to form .
ambipolar diffusion must therefore be incorporated into the model , and in this subsection we describe how the nature of the solutions changes when this is done . in the case of the asymptotic solution for a strongly braked rotational infall or a nonrotational collapse ( eqs .
[ [ mhdstrongasymptoticbegin]][[mhdstrongasymptoticend ] ] ) , the nondiffusive terms in equation ( [ psiequation ] ) are still @xmath294 , but the diffusive term scales as @xmath296 and remains subdominant as @xmath206 .
the asymptotic imhd solution is thus self - consistent in this case , as we verified by checking that , in the model presented in figure [ 6 ] , the diffusive term is everywhere smaller than 10% of the flux - advection term ( assuming @xmath297 ) .
this conclusion differs from that of cck , a discrepancy that can be traced to their neglect of the magnetic squeezing term in the equation for @xmath143 .
however , as we noted in [ mhdstrong ] , real disks are unlikely to become as extremely thin as the imhd solution that incorporates this term would imply .
furthermore , the accumulation of magnetic flux at the center , which is indicated by the imhd split - monopole solution , would lead to a severe magnetic flux problem ( see [ introduction ] ) if allowed to proceed and is likely prevented from occurring by the action of _ ohmic _ diffusivity in the disk or the yso ( see * ? ? ?
the ascendancy of the ambipolar diffusion term at low values of @xmath134 means that ions and neutrals behave differently in that region .
in contrast , the initial ( @xmath216 ) conditions involve no field diffusivity and correspond to negligible drift between these two particle species .
the study of these two regimes can be facilitated by observing that , in most cases , the strong inequality @xmath298 is satisfied for almost all values of @xmath134 : the gradient of @xmath254 is never very large ( except in shocks ) , and the disk is usually very thin . under these conditions
, one can treat the flux conservation relation ( eq .
[ [ psiequation ] ] ) as a quadratic equation for @xmath254 , @xmath299 where we regard the variables @xmath300 , @xmath143 , @xmath145 , and @xmath144 as given .
equation ( [ bzquadraticequation ] ) has two real roots , @xmath301 and @xmath302 , and we have found that in most of the solutions that we constructed they are well separated ( @xmath303 ) .
this property makes it possible to obtain simple expressions for @xmath254 in the imhd ( large @xmath134 ) and ad ( small @xmath134 ) regimes .
specifically , the lower root , which is applicable in the imhd regime , can be evaluated by disregarding the quadratic term in equation ( [ bzquadraticequation ] ) : @xmath304 the vertical field component in the ad regime corresponds to the larger root and can be approximated by omitting the constant term in equation ( [ bzquadraticequation ] ) : @xmath305 the transition between the imhd and ad regimes can take place either gradually or abruptly .
the latter case occurs when the flow is super fast - magnetosonic and the flow variables are changed by passing through a shock .
the occurrence of such an ambipolar diffusion - mediated shock was originally suggested by @xcite and subsequently verified in the nonrotating collapse calculations of ck and cck , which succeeded in resolving the shock structure .
we generalize these results to the rotating - collapse case in [ adstandard ] .
the most striking feature of this shock is the abrupt jump in @xmath254 between the two roots ( [ blo ] ) and ( [ bhi ] ) ( the other flow variables are also affected , but less strongly ) .
the jump represents a large local gradient in @xmath254 , which implies that the term @xmath306 in equation ( [ psiequation ] ) can no longer be disregarded .
in fact , it is this term that mediates the transition between the two roots of equation ( [ bzquadraticequation ] ) and allows the shock structure to be resolved .
the transition , however , remains very sharp , and equation ( [ bzquadraticequation ] ) remains applicable outside a narrow neighborhood of the shock position @xmath307 .
as in the imhd solution presented in [ mhd ] , we find it convenient to distinguish between the fiducial case ( [ adstandard ] ) and the limiting cases of fast rotation ( [ adfast ] ) and strong braking ( [ adstrong ] ) . in the fiducial case
the braking is strong enough to place the centrifugal shock well within the ad shock ( @xmath308 ) .
the centrifugal shock again disappears in the strong - braking case , but an ad shock is still present , as in the nonrotating , diffusive solution obtained by cck .
however , in the fast - rotation case the centrifugal shock is established when the flow is still in the imhd regime , and hence for @xmath258 the inflow speed is low . in this case
the transition between the imhd and ad regimes occurs smoothly instead of in a shock ( with the two roots of eq .
[ [ bzquadraticequation ] ] remaining close to each other ) , although it is still mediated by the @xmath306 term in the induction equation . to solve the self - similar equations as a boundary - value problem ,
it is necessary to know the asymptotic behavior for @xmath206 . for a flow that continues to rotate all the way down to the origin ( as in the fiducial and fast - rotation solutions ) , the behavior in this limit is given by @xmath309^{1/2 } } x^{-3/2}\nonumber\\ & \equiv & \sigma_1x^{-3/2}\ , \\
\label{adbz } b_z =- b_{\phi , s}/\delta&= & [ { m_c}^{3/4}/(2\delta)^{1/2}]x^{-5/4}\ , \\ \label{adpsistandard } b_{r , s}=\psi / x^2&=&(4/3)b_z\ , \\ \label{asymptotich }
h&=&\{2/[1+(2\eta/3\delta)^2]m_c\}^{1/2}x^{3/2}\ .\end{aligned}\ ] ] this solution represents a keplerian disk in which all the field components have comparable magnitudes .
in contrast with the corresponding imhd solution with its split - monopole field configuration ( see eqs .
[ [ mhdasymptoticbegin]][[mhdasymptoticend ] ] ) , in this case the field
matter decoupling brought about by the ambipolar diffusion results in a field that is not strong enough to either dilute the keplerian rotation or to dominate the vertical compression .
[ the vertical squeezing of the disk in the asymptotic ad solution is dominated by the tidal and self - gravity forces , which contribute in the proportion @xmath310 . ]
the asymptotic solution in the strong - braking limit , which ( as in the corresponding imhd case ) involves a flow that is nonrotating near the origin , is given in
[ adstrong ] .
the solutions presented in this subsection have been obtained by using the following numerical procedure : 1 . solve a simplified system of equations by disregarding the @xmath311 terms in equations ( [ forceequation ] ) , ( [ psiequation ] ) , and ( [ thicknessequation ] ) to obtain approximate values for the five variables @xmath145 , @xmath144 , @xmath300 , @xmath142 , and @xmath254 at a selected matching point @xmath312 .
we have found that , to assure numerical stability , it is best to choose this matching point in the region where the effects of ambipolar diffusion are just starting to become dominant .
2 . integrate the full system of equations both outward and inward from the matching point . in the outward direction ,
the solution reaches out to the point @xmath210 , where it must match the prescribed initial conditions of the flow ( given by the parameters @xmath188 , @xmath190 , @xmath189 , and @xmath72 ) . in the inward direction
, the flow must match the @xmath206 asymptotic solution ( given by eqs .
[ [ asymptoticm]][[asymptotich ] ] in the fiducial and fast - rotation cases , and by eqs . [
[ adstrongasymptoticbegin]][[adstrongasymptoticend ] ] in the strong - braking case ) , which also determines the value of @xmath185 . except in the strong - braking limit
, the flow must also incorporate a centrifugal shock ( which is located inside the matching point for the fiducial solution and outside it in the fast - rotation case ) .
the discontinuities that some of the flow variables experience at @xmath195 are handled using the appropriate shock jump conditions ( appendix [ singular ] ) .
3 . iterate on the values of the five flow variables at the matching point @xmath312 as well as on those of the central mass ( @xmath185 ) and ( if applicable ) of the centrifugal - shock position ( @xmath195 ) until convergence is reached .
our `` standard '' solution , corresponding to @xmath313 , @xmath314 , @xmath315 , and @xmath262 , is presented in figure [ 7 ] . in this case
the initial rotation is not very fast and the braking is moderate ( see [ mhdstandard ] ) , so the ambipolar - diffusion shock is located further away from the center than the centrifugal shock ( @xmath316 ) . to more clearly
separate the different flow regimes , we also show a solution where @xmath72 is reduced to 0.18 ( fig . [ 8]a ) , which has the effect of reducing @xmath195 by a factor @xmath317 , and another one ( fig . [ 8]b ) where , in addition , @xmath23 is reduced to 0.7 , which has the effect of further increasing @xmath318 ( to @xmath319 ) .
we now summarize the distinguishing properties of these flow regimes .
[ [ outer - region - xx_a - ideal - mhd - infall . ] ] outer region ( @xmath320 ) : ideal - mhd infall .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + this region extends from @xmath210 , where the initial conditions are applied , to @xmath321 , the location of the resolved ambipolar diffusion shock .
this region does not differ substantially from the imhd case , with the effects of ambipolar diffusion remaining minimal .
the approximation @xmath322 ( eq . [ [ blo ] ] ) is excellent , and it is also quite accurate to set @xmath323 . the outer asymptotic solution remains roughly applicable for @xmath324 .
although the centrifugal force is still dynamically unimportant , there is some magnetic braking , especially for @xmath325 , and so the ratio @xmath244 is not exactly constant . over most of this region
the infall speed is governed by self - gravity ( @xmath194 ) , but closer to the inner edge ( @xmath326 ) , infall starts to be dominated by the central mass .
[ [ the - ambipolar - diffusion - shock - xapprox - x_a . ] ] the ambipolar diffusion shock ( @xmath327 ) .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + this shock marks the end of the imhd regime .
it can be resolved as a continuous transition , although it may contain a viscous subshock .
its most notable feature is the rapid increase in @xmath254 , which grows from essentially @xmath301 at the outer edge of the shock transition to @xmath302 ( eq . [ [ bhi ] ] ) at the inner edge .
although the inequality @xmath328 continues to hold , the @xmath311 terms in the structure equations are important in this range .
[ it is worth noting in this connection that the sharp spike exhibited by the disk half - thickness curve at the location of the shock is a consequence of the large @xmath254-gradient term in eq .
( [ thicknessequation ] ) .
the extreme narrowness of the spike ( @xmath329 ) indicates , however , that the assumption of vertical hydrostatic equilibrium is not justified at that location . in reality , the enhanced magnetic squeezing at the shock will reduce the disk thickness to only a fraction of the amount indicated by the equilibrium curve over the fluid transit time through the shock .
this apparent overshoot is , however , a highly localized phenomenon : the solution obtained by omitting the @xmath311 term in eq .
( [ thicknessequation ] ) has a very similar global structure . ] immediately behind the shock the inward acceleration is temporarily reversed , resulting in a thin overshoot layer where the flow undergoes a weak outward acceleration .
the shock structure is basically the same as that found in cck for a nonrotational flow : for our fiducial parameters , rotation is not yet dynamically significant at this location . to obtain an estimate of the value of @xmath307 ( which can be used as an initial guess in the numerical solution ) , we make use of the fact that , in the infall region that is located just downstream from @xmath307 , @xmath330 .
these approximate equalities imply @xmath331 .
we evaluate this expression at @xmath307 by using the fact that , for @xmath320 , ideal mhd is approximately valid and hence @xmath332 is not very different from its initial value @xmath189 , and that at this location @xmath232 is already close to @xmath185 so that @xmath333 . for the solutions shown in figures [ 7 ] and [ 8 ] , the vertical squeezing at @xmath307
is controlled by magnetic pinching , so @xmath334 ( see eq .
[ [ thicknesssolution ] ] ) .
these approximations yield an estimate of the shock location , @xmath335 which depends only on prespecified flow parameters .
analogous ( albeit less simple ) algebraic expressions can be written down for the cases in which the vertical squeezing is dominated by the central tidal force or by self - gravity . since the magnetic flux contained within @xmath307
is essentially the same as the flux that initially threaded the enclosed mass , it also nearly equals the flux trapped in the central split monopole of the imhd solution .
ambipolar diffusion does not destroy the accumulated flux but rather redistributes it between the origin and the shock location . as the value of @xmath307 generally increases with the diffusivity parameter @xmath23 , the approximate flux conservation within @xmath307 implies that the vertical field component ( @xmath336 ) behind the ambipolar - diffusion shock decreases with increasing @xmath23 ( in particular , it scales as @xmath337 when @xmath307 is given by eq .
[ [ x_a ] ] ) .
a comparison of figures [ 8]a and [ 8]b reveals that the deceleration of the flow in the ad shock is stronger the lower the value of @xmath23 .
this can be attributed to the inferred dependence of @xmath302 on @xmath23 , which results in higher magnetic tension and pressure forces downstream of the shock when @xmath23 is reduced . if @xmath23 becomes smaller than @xmath338 ( with the other parameters remaining unchanged ) , then the localized decrease of the infall speed inward of @xmath307 becomes large enough to reduce @xmath164 below 1 .
but @xmath339 is the singular line of the ad system of equations , corresponding to the critical ( sonic ) speed ( see appendix [ singular ] ) .
a crossing of this curve signals a discontinuity in the flow parameters , which in this case implies the presence of an unresolved viscous subshock at the downstream end of the resolved mhd shock transition ( see * ? ? ?
[ [ ad - dominated - infall - x_c - x - x_a . ] ] ad - dominated infall ( @xmath340 ) .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + over most of the region between the two shocks the centrifugal force remains dynamically unimportant , so matter moves almost in free fall , with the gravitational force dominated by the central mass .
the ions and field are only weakly coupled to the neutrals , with the ion radial speed being effectively zero ( except in the vicinity of @xmath195 ) .
the vertical field component continues to be given by @xmath341 to a very good approximation . in many respects
this zone resembles the innermost region of the nonrotating flow studied by cck , especially when it is fairly extended ( @xmath342 ) , as it is in the two solutions presented in figure [ 8 ] .
in particular , @xmath343 , @xmath344 , and ( over a somewhat narrower range ) @xmath345 . if this region is wide enough , magnetic braking can reduce @xmath142 to an essentially constant value , @xmath346 .
the mass is also well approximated by its plateau value ( see eq .
[ [ m_pl ] ] ) , @xmath347 .
2ex as we point out in [ discussion ] , @xmath185 will be reduced below the value of @xmath232 upstream of the centrifugal shock if ( as expected ) the rotationally supported disk drives a wind from its surfaces .
the numerical value of the reduction factor depends , however , on yet another model parameter ( see eq .
[ [ m_cw ] ] ) . therefore , to simplify the discussion , we have chosen not to incorporate the effect of the wind mass loss into the solutions that we present .
the centrifugal force starts to become important near the downstream end of this zone : after it comes to exceed the gravitational force , it triggers the formation of the centrifugal shock at the point @xmath348 .
in contrast to the imhd solution presented in [ mhdstandard ] , one can not approximate @xmath349 by its initial value in this case : the braking action induced by the magnetic field amplification in the ad shock completely invalidates this approximation .
a comparison of figures [ 8]a and [ 8]b illustrates this point .
when @xmath23 is reduced from 1 ( fig .
[ 8]a ) to 0.7 ( fig .
[ 8]b ) , the location of the ad shock changes by a factor of @xmath350 and the value of the central mass remains essentially the same : this is as expected from equations ( [ x_a ] ) and ( [ m_pl ] ) , respectively .
however , the decrease in @xmath23 has a very pronounced effect on the location of the centrifugal shock : as @xmath23 is reduced by a factor of @xmath351 , @xmath195 decreases by a factor of @xmath352 .
as we now demonstrate , this can be traced to the sensitive dependence of the magnetic braking term on @xmath23 , which leads to a strong reduction in @xmath142 downstream from the ad shock for even a moderate decrease in @xmath23 .
we first estimate @xmath255 by substituting @xmath254 from equation ( [ bhi ] ) into equation ( [ bphiequation ] ) and using equation ( [ mw ] ) , @xmath353 equation ( [ angmomequation ] ) then yields @xmath354 using again the approximations @xmath355 and @xmath257 that were employed in deriving the above estimate of @xmath307 , equation ( [ djdx ] ) becomes @xmath356 using the approximations @xmath357 and @xmath358 , and assuming that @xmath359 , we get @xmath360 this equation can be integrated to give @xmath361 where @xmath362 is a constant .
this result is valid between @xmath307 and @xmath195 and can be used to estimate the latter . for @xmath363 ,
the value of @xmath142 at @xmath195 may be identified with @xmath349 and can be approximated by @xmath362 .
assuming also that no significant magnetic braking takes place for @xmath320 ( consistent with our analysis of the imhd solution in [ mhdstandard ] ) , we can write @xmath364 . using equation ( [ x_a ] ) to evaluate @xmath307 , we finally obtain @xmath365 we have verified that this estimate ( which depends on the ratio @xmath318 being @xmath366 ) yields the value of @xmath195 to within 30% for both of the solutions shown in figure [ 8 ] .
the strong dependence of @xmath195 on @xmath23 that is exhibited by these solutions can be understood from the form of the expression ( [ xcadapproximation ] ) : it indicates that the effect of even a small variation in @xmath23 is magnified on account of its presence inside an exponential function .
although the precise form of this expression may be expected to differ in cases where some of the approximations we utilized do not apply , the sensitive dependence of @xmath195 on @xmath23 is probably a general feature of the model .
[ [ the - centrifugal - shock - xapproxx_c . ] ] the centrifugal shock ( @xmath367 )
. + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the shock represents a discontinuity in the variables @xmath144 , @xmath145 , and @xmath254 through which the flow changes from the nearly undiluted free fall to subsonic accretion .
we approximate the shock as a discontinuity in the variables @xmath144 and @xmath145 , which is governed by the isothermal - shock jump conditions ( appendix [ isothermalshock ] ) , followed by a very thin ( but numerically resolvable ) layer where @xmath254 increases . in this picture , the ions are decoupled from the neutrals inside the subshock , so their radial speed in the frame of the shock , and hence ( by magnetic flux conservation , see eq .
[ [ flux_conserve ] ] ) @xmath63 , remain unchanged across the subshock even as @xmath368 and @xmath62 vary .
however , the increase in the drift speed @xmath369 across the subshock enhances the collisional drag force immediately behind it and causes a local recoupling of ions and neutrals ; this , in turn , leads to a rapid decline in @xmath370 and a consequent increase in @xmath63 .
just inward of the shock there is a narrow transition zone where the flow settles down to its asymptotic keplerian structure . within this layer the variables @xmath144 , @xmath145 , and @xmath254 adjust rapidly ( with some oscillations and overshoots ) to their asymptotic values .
this implies , in particular , that , just as in the case of the ad shock , the derivative terms @xmath371 in the constituent equations can not be neglected in this zone .
the surface density in this region is significantly larger than @xmath145 in its immediate vicinity , so the layer represents a massive ring lying just outside the main body of the disk ( resembling the situation in the fiducial imhd solution considered in [ mhdstandard ] ) . in the solution shown in figure
[ 7 ] , the width of the ring is @xmath372 of @xmath195 and it encompasses @xmath373 of the total mass of the disk within @xmath195 ; the relative size of the ring in the solutions displayed in figure [ 8 ] is , however , smaller .
the variables @xmath300 and @xmath142 do not change appreciably across this region .
[ [ the - keplerian - disk - xx_c . ] ] the keplerian disk ( @xmath258 ) .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + inward of the postshock transition layer , the flow approaches the form of a magnetized keplerian accretion disk as described by equations ( [ asymptoticm])([asymptotich ] ) .
for the solutions shown in figures [ 7 ] , [ 8]a , and [ 8]b , the total mass of the rotationally supported disk ( including the ring at @xmath195 ) is , respectively , @xmath374 , @xmath375 , and @xmath376 of @xmath185 . as in the corresponding imhd case , a high value of @xmath72 and comparatively low values of @xmath159 and @xmath277 result in the centrifugal shock being located within the self - gravity dominated region ( where @xmath377 ) .
the value of @xmath195 is large enough for ideal mhd to still be applicable ( so @xmath378 ) . since the inflow speed is strongly reduced in the dense disk that forms behind the centrifugal shock , the ambipolar - diffusion time becomes shorter than the accretion time and an ad transition front is established not far from @xmath195 .
in fact , as @xmath379 is small enough for the flow to remain subsonic ( @xmath380 ) in this region , the imhd - to - ad transition is gradual rather than sharp as in the other cases considered in this subsection , where it takes the form of an ad shock .
since the centrifugal shock occurs within the ideal - mhd region , we impose the same jump conditions at @xmath195 as in the corresponding imhd case ( see [ generalizedisothermalshock ] ) .
a representative solution is shown in figure [ 9 ] .
it exhibits the same general features ( a non - keplerian outer region , a small 3ex central mass , a backflowing layer behind the centrifugal shock ) as the fast - rotation imhd solution ( see figs . [ 4 ] and [ 5 ] ) .
these characteristics are basically a consequence of the fast initial rotation , and the strong similarity with the corresponding imhd case shown in figure [ 4 ] can be understood from the fact that the centrifugal shock occurs outside the ad regime . in
[ mhdstrong ] we have found that ideal - mhd flows with strong braking can lose all their angular momentum at a finite distance from the origin . here we consider whether a similar situation can occur in the presence of ambipolar diffusion . when @xmath381 there are two inner asymptotic solutions that are mathematically self - consistent and that can be matched to the outer asymptotic solution by integrating the system of differential equations .
correspondingly , two kinds of solutions are found : one that is almost imhd in nature , and another that is ad dominated .
the imhd - like solution resembles the one obtained in [ mhdstrong ] in that @xmath254
is everywhere well approximated by the root @xmath301 of equation ( [ bzquadraticequation ] ) , there is no ad shock , and the inner asymptotic solution is given by equations ( [ mhdstrongasymptoticbegin])([mhdstrongasymptoticend ] ) ( but with @xmath248 possibly differing from @xmath189 on account of the small , but finite , ambipolar diffusion ) .
outside the point @xmath201 the azimuthal field component can differ significantly from its imhd value , but once complete braking is accomplished and both @xmath141 and @xmath255 effectively vanish , this solution basically describes an imhd nonrotating inflow and , in particular , predicts the formation of 6ex a central split - monopole field . as we remarked at the beginning of [ ad ] , such a configuration is unlikely to occur in real systems , so we henceforth concentrate on the more plausible alternative case . in the diffusive solution
, the flow passes through an ad shock at a point @xmath307 ( which can be estimated by eq .
[ [ x_a ] ] ) that typically exceeds @xmath201 .
inward of the ambipolar - diffusion shock , the poloidal field components have the same form as in the fiducial solutions , @xmath382 . as in the corresponding imhd case ,
the inner asymptotic solution satisfies @xmath383 and describes a free fall onto the central mass , which , however , in this case is _ not _ magnetically diluted but rather resembles the situation in the ad - dominated infall region of the fiducial solutions .
it is given by @xmath384 the power - law scalings of the flow variables with @xmath134 are the same as in the asymptotic solution obtained by cck for the nonrotating ad collapse . however , because of the magnetic pinching effect included in the above solution , the expression for the coefficient @xmath385 is more involved : @xmath386{6}}\right]^2\ , \\
d&=&n^{1/4}/\eta\ , \\ n&=&2/m_c\ , \\
f)^{1/3 } - 8 n ( 3/f)^{1/3}\ , \\ f&=&9d^2+[3(27 d^4 + 256 n^3)]^{1/2}\ , \end{aligned}\ ] ] with the expression for @xmath385 representing the real , positive solution of the quartic equation @xmath387 .
an example of a strongly braked solution is shown in figure [ 10 ] .
a curious feature of the flow is the value of @xmath142 as @xmath206 , which assumes a nonzero ( albeit rather small ) plateau value ( @xmath388 ) .
this puzzling behavior can be attributed to the very small magnitude of @xmath300 in this limit , which causes @xmath389 ( eq . [ [ bphiequation ] ] ) and therefore @xmath390 ( eq . [ [ angmomequation ] ] ) to be small , more than compensating for the increase in @xmath254 behind the ad shock .
the low value of the angular - momentum derivative in turn accounts for the asymptotic near - flatness of the @xmath142 curve .
( in contrast , the large central value of @xmath300 associated with the split - monopole field in the near - imhd solution guarantees that , in this case , @xmath389 and @xmath390 remain large enough near @xmath201 to make @xmath142 actually vanish there . )
the value of @xmath349 is highly sensitive to the magnitude of the parameter @xmath23 : it decreases by @xmath317 orders of magnitude when @xmath23 is decreased from 0.5 to 0.3 .
( the strong sensitivity of @xmath195 to the value of this parameter has already been noted in connection with the fiducial solutions ; see eq .
[ [ xcadapproximation ] ] . )
however , for practical purposes , @xmath349 can be regarded as being effectively zero . in particular
, the implied location of the centrifugal radius , @xmath348 , corresponds to such a small radius for any meaningful value of @xmath391 that it would invariably lie well inside the central protostar .
the model presented in this paper provides a physical framework for a detailed study of the development of star disk systems in the collapse of rotating , magnetic molecular - cloud cores .
perhaps the most noteworthy feature of the solutions presented in [ results ]
is that , for representative values of the initial cloud rotation and magnetization and with plausible estimates of the magnetic field diffusivity and of the strength of magnetic braking , they predict the formation of circumstellar disks whose properties compare quite well with those of real protostellar systems . in particular , the disks obtained for our chosen fiducial parameters are _ keplerian _ , having a mass that is only a small fraction ( @xmath17 ) of the central mass , as has been inferred for most protostellar disks in nearby dark clouds ( see [ formulation ] ) . for the representative ad solution shown in figure [ 7 ] ( corresponding to @xmath313 , @xmath314 , @xmath315 , and @xmath262 ) , the disk mass @xmath2 is less than @xmath392 of the central mass @xmath4 . for comparison ,
if the model parameters are slightly changed ( to @xmath313 , @xmath393 , @xmath394 , and @xmath395 ) , one finds @xmath396 .
the most reliable protostellar disk velocity curves have been obtained in class ii ysos ( i.e. , t tauri stars , of typical ages @xmath397 ) , which , in contrast to younger protostars , are no longer strongly obscured by dusty envelopes ( identified with the contracting cloud cores ) .
these are also the systems with the most accurate determinations of the disk mass ( e.g. , * ? ? ?
* ) : the massive envelopes surrounding class 0 ysos prevent their disk masses from being pinned down , and even in the case of class i sources the values of @xmath2 are not as well determined ( although they appear to be similar to the disk masses in class ii systems ) . it has occasionally been suggested in the literature that protostellar disks may not be keplerian during the early ( in particular , class-0 ) phase of their evolution , when the ratio of the central mass to that of the surrounding envelope is still small .
our results demonstrate that this need not be the case .
in fact , in the self - similar solutions that we derive , if the mass of the rotationally supported disk is small in comparison with the central mass at any given time , it will be so at all times .
the observed rotationally supported disks typically have sizes @xmath15 ( see [ introduction ] ) , which are reproduced by our typical solutions .
for example , the derived disk size ( given by the value of the centrifugal radius @xmath1 ) @xmath398 after the start of the collapse ( which corresponds to the class - i evolutionary phase ) is @xmath399 for the fiducial solution presented in figure [ 7 ] and @xmath400 for the other ad solution mentioned in the preceding paragraph .
there have been reports in the literature of t tauri circumstellar disks that exhibit keplerian rotation on somewhat larger scales [ e.g. , gm aur , with @xmath401 @xcite , and dm tau , with @xmath402 @xcite .
it is quite possible that these disks have achieved their large sizes already during their early evolutionary stages .
for example , they may have possessed a comparatively high value of the diffusivity parameter @xmath23 , to which the predicted magnitude of @xmath1 is particularly sensitive ( see eq . [ [ xcadapproximation ] ] ) .
alternatively , they may have formed from collapsing cores with comparatively large initial rotations and relatively inefficient braking ( see [ adfast ] ) .
it is worth noting , though , that in the self - similar model , disks with larger values of @xmath195 tend to have higher disk - to - star mass ratios and therefore exhibit progressively greater departures from a keplerian rotation law .
it is , however , also conceivable that at least part of the cause for large observed disk sizes in these highly evolved ( and no longer strongly accreting ) systems is post - formation viscous spreading of the disk ( e.g. , @xcite ; @xcite ) .
it is interesting to note in this connection that the radial scaling of the disk surface density predicted by our self - similar solution ( @xmath403 for small @xmath55 ; see eq .
[ [ adsigma ] ] ) is the same as that inferred for the `` minimum mass '' solar nebula @xcite .
although this agreement is encouraging , it is not unique to this model : in particular , the same scaling is predicted for a self - similar _ keplerian _ disk with `` @xmath159 viscosity '' @xcite and for a self - similar _ self - gravitating _ disk with a gravitational instability - induced effective viscosity @xcite .
our model disks exhibit a surface - density enhancement near their outer edges ( see [ mhdstandard ] and [ adstandard ] ) : for the fiducial ad solution shown in figure [ 7 ] , this ring has a mass @xmath404 and a width @xmath405 .
we are not aware of any observational evidence as yet for the existence of such rings in protostellar systems , but there have been several recent reports of well - defined dust rings with similar radii ( @xmath406 ) and fractional widths around older , vega - like stars ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
although the apparent confinement of the latter rings is usually attributed to the presence of solid `` satellites '' ( e.g. , planets ) , it is interesting to speculate that their origin could be traced to the @xmath62 enhancements predicted by our solutions .
an important implication of the result that typical rotating inflows tend to produce a central mass that is surrounded by a keplerian disk is that the angular momentum problem ( see [ introduction ] ) for the modeled yso is basically resolved . in particular ,
angular momentum transport can be sufficiently efficient to allow most of the inflowing mass to end up ( with effectively no angular momentum ) at the center , with the central mass dominating the dynamics well beyond the outer edge of the disk even as the inflow is still in progress .
our solutions reveal that the ad shock , even though it is usually located well outside the region where the centrifugal force becomes important , helps to enhance the efficiency of angular momentum transport through the magnetic field amplification that it induces . in practice
, the ultimate value of the protostellar angular momentum would be determined by the braking mechanism that enables the yso to reduce its angular velocity below the centrifugal - equilibrium value that characterizes the accreted disk matter ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the formation of keplerian circumstellar disks in our `` standard '' solutions is a consequence of efficient magnetic braking , which allows the mass injected at large distances to reach the center rather than pile up in the disk ( as it does in the fast - rotation solutions discussed in [ mhdfast ] and [ adfast ] ) .
the disk in these cases therefore has little influence on the mass accretion rate onto the central yso , @xmath407 , which is determined on scales much larger than @xmath1 .
in fact , as we discuss in [ nonmagnetic ] , the eigenvalue @xmath185 is typically also insensitive to the magnetic structure of the flow and is often well approximated by the `` plateau '' value @xmath272 ( eq . [ [ m_pl ] ] ) that depends only on the initial conditions .
( for our fiducial parameters , @xmath408 , implying @xmath409 . )
the likely production of a centrifugally driven wind that transports mass away from the disk surfaces ( see discussion below ) will , however , act to reduce the value of @xmath185 : as we estimate in appendix [ wind ] , the reduction factor could be as large as @xmath410 .
the presence of rotation also does not strongly affect the `` revitalization '' of ambipolar diffusion behind the ad shock , which , as we noted in [ introduction ] , can go a long way toward resolving the magnetic flux problem .
this conclusion applies even in the fast - rotation case ( see [ adfast ] ) , where the ad shock is replaced by a more gradual transition occurring within the disk .
the asymptotic @xmath206 similarity solution for the disk implies that , at any given time , the magnetic flux scales as @xmath411 ( see eq . [ [ adpsistandard ] ] ) , which represents a slower decrease with @xmath55 than in the nonrotating case ( in which @xmath412 ; see eq .
[ [ adpsistrong ] ] ) .
however , in reality , the amount of magnetic flux that ends up threading the protostar would depend on the detailed flux transport and destruction mechanisms that operate in the inner disk and the star , which may involve ohmic resistivity ( e.g. , * ? ? ?
* ) , ambipolar diffusion ( e.g. , * ? ? ?
* ) , and reconnection ( e.g. , * ? ? ?
the quasi - equilibrium disk configurations described by our solutions make it possible to investigate the stability properties of the modeled circumstellar disks .
although a full pursuit of this topic is beyond the scope of this paper , we comment briefly on the basic mhd and gravitational instabilities that may affect the disks .
the most relevant mhd instability is evidently the magnetorotational one , which is expected to develop rapidly in any differentially rotating disk that is threaded by a subthermal magnetic field ( see @xcite for a review ) .
this instability has been shown to generate turbulence that can remove angular momentum from the inflowing gas , so the determination of whether the disk is susceptible to this instability also has implications to the self - consistency of our identification of magnetic braking as the dominant angular - momentum transport mechanism in the disk .
as it turns out , our diffusive keplerian disk models are by and large magnetorotationally _
stable_. physically , this is because the coupling between the matter and the field is generally too weak to allow the instability to grow .
specifically , a necessary condition for this instability to operate in a weakly ionized , keplerian disk is @xmath413 , where @xmath414 is the keplerian angular velocity @xcite . using our self - similar variables
, this criterion can be written as @xmath415 . since the keplerian disk region in all of our typical solutions nearly coincides with
the region where the inner asymptotic solution ( eqs .
[ [ asymptoticm]][[asymptotich ] ] ) is applicable , we can use the latter to evaluate the left - hand side of the above inequality ; it is found to equal @xmath416 .
[ the asymptotic solution satisfies the weak - field condition @xmath417 postulated in the derivation of the instability criterion ; indeed , the left - hand side of the latter inequality ( which equals @xmath418 ) vanishes ( @xmath419 ) as @xmath206 . ] but the parameter @xmath277 ( defined in
[ formulation ] ) is expected to be @xmath420 , and therefore the instability condition @xmath421 is not satisfied .
interestingly , our fiducial imhd solution is also magnetorotationally stable . in this case
the strong magnetic squeezing prevents the unstable perturbations from fitting inside the disk .
formally , one requires the critical wavelength @xmath422 to be less than the disk thickness @xmath423 for the disk to be unstable . using the inner asymptotic solution given by equations ( [ mhdasymptoticbegin])([mhdasymptoticend ] ) , with the help
of which we first verify that @xmath424 again tends to zero as @xmath206 , this condition translates into @xmath425 . for our typical parameters this requires @xmath426 , but this lower bound exceeds @xmath195 ( the approximate disk outer boundary ) for this solution .
the above results support the choice of magnetic braking as the main angular momentum transport mechanism in our disk model .
this mechanism remains effective even in the presence of ambipolar diffusion because the neutral ion collision time @xmath24 , while being comparatively long , is shorter than the magnetic braking time @xmath427 . in a keplerian accretion disk @xmath428 and @xmath429
is typically much smaller than the keplerian speed ; therefore @xmath430 can be @xmath431 even if @xmath432 exceeds 1 .
this conclusion continues to hold when the angular momentum transport in the disk ( and hence the accretion time @xmath433 ) is controlled by a centrifugally driven wind rather than by the propagation of torsional alfvn waves . as we show in appendix [ wind ] , the basic formalism of our model is unchanged in this case , and the parameter @xmath277 ( which is now determined directly from the centrifugal wind solution ) is again typically small enough for the linear stability criterion @xmath434 to be satisfied over most of the disk column density .
the argument against instability is reinforced by indications from numerical simulations ( e.g. , * ? ? ?
* ) that mri - driven turbulence and the resulting angular - momentum transport are significant only when @xmath435 .
we note in this connection that even a disk in which the bulk of the material is weakly coupled is expected to possess well coupled ( @xmath421 ) surface layers from which the centrifugal wind is launched ( see appendix [ wind ] ) .
it is , however , likely that the critical wavelength for instability exceeds the density scale height in the strongly coupled regions ( see @xcite and @xcite ) , so the surface layers should typically also be magnetorotationally stable . the basic gravitational instability that we consider involves fragmentation induced by the disk self - gravity .
the relevant stability condition against axisymmetric perturbations is given by the toomre criterion @xmath436 @xcite , where @xmath437^{1/2}$ ] is the epicyclic frequency .
although this criterion is modified somewhat by magnetic effects ( @xcite ; r. krasnopolsky & a. knigl , in preparation ) , the corrections remain small for nearly keplerian disks , especially in the ad regime , and we therefore neglect them here .
we have verified that the toomre criterion is well satisfied for most of our rotationally supported disk solutions , the only exceptions being the outermost ( backflow ) regions of the fast - rotation configurations ( figs . [ 4 ] , [ 5 ] , and [ 9 ] ) , where the effect of self - gravity is comparatively large .
this result is significant in that it suggests that in typical cases protostellar disks may not fragment even during the early phases of their evolution , when the central mass is still small . in our model this behavior can be attributed to the fact that , for representative parameters , the central gravity already dominates at the location of the centrifugal shock , so that the disks are nearly keplerian from the start .
inasmuch as the inequality @xmath438 signals the onset of self - gravity induced angular - momentum transport in the disk ( e.g. , * ? ? ?
* ) , this result also serves to justify the neglect of such a transport mechanism in our model .
we note , however , that the collapsing core could in principle be unstable to fragmentation before point - mass formation , or , following pmf , in the region outside the rotationally supported disk ( e.g. , * ? ? ?
it is also conceivable that fragmentation might be initiated even before a dynamically unstable core is formed ( possibly triggered by an ad - mediated instability in the magnetically supported parent cloud ; see @xcite and @xcite ) .
another potential angular - momentum transport mechanism that was not included in our model involves centrifugally driven disk outflows .
for a cold , thin , keplerian disk that is threaded by open magnetic field lines , such an outflow could be launched if the angle between the meridional projection of the magnetic field and the rotation axis exceeds @xmath439 @xcite .
this condition is satisfied by our asymptotic disk solutions in particular , the ad inner asymptotic solution implies a radially constant inclination angle @xmath440 ( see eq .
[ [ adpsistandard]]) so our model disks can be expected to give rise to outflows of this type .
this result is encouraging in view of the fact that centrifugally driven disk outflows are a leading candidate for the origin of the bipolar outflows that are frequently observed to emanate from ysos ( see , e.g. , @xcite for a review ) . a centrifugally driven disk outflow can be incorporated into our model by taking account of the fact that the wind would be quasi - steady on the accretion time scale and should thus be well approximated by a steady - state formulation .
the condition that the centrifugal outflow control the vertical angular momentum transport from the disk is equivalent to the requirement that the steady - state wind solution pass through the alfvn critical point .
this requirement , in turn , serves to fix the value of @xmath78 ( or , equivalently , of the parameter @xmath277 ) in the disk solution . besides angular momentum
, the wind also carries away mass , resulting in a progressive decline in the mass accretion rate with decreasing disk radius and leading to lower values of @xmath441 and @xmath232 at the center .
interestingly , the radial scaling of the disk magnetic field ( @xmath442 ; see eq . [ [ adbz ] ] ) implied by the asymptotic ad disk solution for any given value of @xmath391 is the same as in the @xcite radially self - similar , steady - state wind solution .
this suggests that the latter solution can be used in conjunction with our disk model to fix the values of @xmath277 and of the @xmath185 reduction factor in the asymptotic disk solution .
we discuss the implementation of this scheme in appendix [ wind ] . inasmuch as the incorporation of a disk wind does not qualitatively change our basic results
, we have not included its effects in the model described in [ model ] so as not to unduly complicate the presentation .
it is nevertheless important to recognize that our solutions indicate that the collapse of rotating , magnetic cloud cores naturally gives rise to disks with magnetic field configurations that are conducive to the launching of centrifugally driven winds .
this conclusion is supported by the mhd core - collapse numerical simulations of @xcite , in which disk outflows that transport angular momentum through material and magnetic stresses are observed to form .
( the possibility that a disk formed in the collapse of a magnetized rotating cloud core could give rise to a centrifugally driven wind was recognized also by @xcite . however , in contrast with our model , in their picture the angular momentum transport in the disk is dominated by turbulent viscosity and the outflow is concentrated near the outer disk radius ; furthermore , they infer that the mass outflow rate depends sensitively on the efficiency of magnetic braking in the precollapse core . ) we have also presented solutions ( [ mhdstrong ] and [ adstrong ] ) in which the magnetic braking is sufficiently efficient to prevent the formation of a rotationally supported disk from the collapse of a rotating cloud core .
although we argued that the parameter combinations characterizing these solutions are not typical , it is conceivable that such situations could be realized . a possible indication that this may sometimes be the case is provided by the detection of slowly rotating ysos that show no evidence ( from excess near- and mid - ir emission ) for the presence of a circumstellar disk @xcite .
these observations are puzzling if one considers disk accretion to be a ubiquitous feature of star formation and interprets slow stellar rotation as due to a star disk interaction ( e.g. , * ? ? ?
* ; * ? ? ?
the properties of these ysos are , however , readily understood if they are the product of a strongly braked core collapse .
we study the collapse of rotating , magnetic molecular - cloud cores with the help of a self - similar model .
this model generalizes the work of cck on nonrotating cloud cores by incorporating the effects of rotation and magnetic braking ( while also taking into account ambipolar diffusion and its dependence on the magnetic tension force as in the original model ) .
we focus on the evolution after point - mass formation , noting that the pre - pmf collapse has already been studied by numerical simulations .
our approach is motivated by previous findings that simulation results for the pre - pmf collapse of rotating cores and for the pre- and post - pmf collapse of nonrotating cores are well approximated by self - similar solutions .
our semianalytic scheme allows us to examine the full range of expected behaviors and their dependence on the physical parameters .
we present and analyze solutions for rotating but nonmagnetic collapsing cores , reproducing the results of @xcite ( @xcite ) , as well as for magnetic cores in the ideal - mhd and ambipolar - diffusion limits . in these two regimes
, we distinguish between fiducial solutions , obtained for typical parameter values that correspond to moderately fast initial cloud rotation rate and moderately strong magnetic braking , and limiting cases of ( 1 ) fast rotation and ( 2 ) strong braking .
our results can be summarized as follows : * for representative parameter values , we obtain solutions that describe rotationally supported circumstellar disks of masses and sizes that are consistent with observations of ysos .
our model thus makes it possible , for the first time , to study the formation of protostellar disks in the context of a realistic scenario of star formation in magnetically supported , weakly ionized molecular cloud cores .
the outer boundary of the nearly stationary accretion disk roughly coincides with the location of the centrifugal shock , which typically occurs inward of the ambipolar - diffusion shock where the magnetic field decouples from the matter . *
our solutions indicate that it is quite possible that t tauri ( class ii ) protostellar systems , whose disk masses are typically inferred to be @xmath17 of the central mass , have had a similarly low disk - to - star mass ratio also during their earlier ( class-0 and class - i ) evolutionary phases .
* the disk configurations that we obtain exhibit surface - density enhancements near their outer boundaries ( encompassing @xmath17 of the disk mass and radius for typical parameters ) .
we speculate that these features may be the precursors of the rings detected in some vega - like systems on scales @xmath15 . *
our models allow us to elucidate the interplay between the ambipolar diffusion and magnetic braking processes .
in particular , the magnetic field enhancement behind the ad shock increases the efficiency of angular - momentum removal from the disk . in the limiting case of strong braking , essentially all the angular momentum
is removed well before the inflowing gas reaches the center .
such systems may correspond to slowly rotating ysos that show no evidence of a circumstellar disk . *
the inferred magnetic - field structures imply that the disks could drive centrifugal outflows over much of their radial extent .
we show that the steady - state , radially self - similar wind solution of @xcite can be naturally incorporated into the asymptotic ad disk solution , making it possible to study the effects of wind angular - momentum and mass removal from the disk and to better constrain the relevant parameters .
* the derived nearly - keplerian disk configurations appear to be immune to both the magnetorotational instability and to self - gravity induced fragmentation . only in the limiting case of rapid initial rotation ( which yields comparatively massive and largely non - keplerian disk solutions )
is the toomre @xmath443 parameter @xmath444 anywhere in the disk . despite the various simplifications that needed to be employed to ensure self - similarity , our solutions likely capture the main qualitative aspects of the collapse of rotating cloud cores , and it is conceivable that our numerical estimates are also not too far off .
further progress in the study of protostellar disk formation could be achieved by means of numerical mhd simulations .
we thank j. everett , w. herbst , z .- y .
li , and n. vlahakis for helpful conversations or correspondence .
we also acknowledge i. contopoulos and g. ciolek , whose insights on the nonrotating core - collapse problem ( presented in cck and ck ) have guided us in the present work . this research was supported in part by nasa grant nag 5 - 3687 and by doe under grant b341495 .
assuming axisymmetry and isothermality , we write down ( using cylindrical coordinates ) the mass , radial momentum , and angular momentum conservation relations as well as the vertical hydrostatic equilibrium condition : @xmath445 @xmath446 @xmath447 @xmath448 where @xmath128 and @xmath449 are , respectively , the radial and vertical components of the gravitational field .
we now integrate these equations over @xmath116 .
we assume that the disk is geometrically thin [ half - thickness @xmath450 and aim to get expressions that are valid to order @xmath451 . on account of the disk thinness , we approximate the density , radial velocity , azimuthal velocity , and radial gravity as being constant with height .
we thus equate the column density @xmath452 to @xmath453 and drop the last term on the right - hand side in both equations ( [ radial3d ] ) and ( [ angular3d ] ) . equation ( [ mass3d ] ) becomes @xmath454 where the last term on the right - hand side represents the mass flux in a disk outflow [ of total outflow rate @xmath455 within a radius @xmath55 ] .
we neglect this term in the calculations presented in the body of the paper , but we return in [ wind ] to consider the effect of having @xmath457 . using the solenoidal condition on the magnetic field , [ @xmath458 , which in a thin disk allows us to treat @xmath63 as being constant with height whenever
it is not explicitly differentiated with respect to @xmath116 ] as well as the assumed field symmetry , equation ( [ radial3d ] ) integrates to @xmath460\,dz \ .\end{aligned}\ ] ] taking account of the adopted vertical mass distribution , we replace the integral in the last term of equation ( [ radial3d1 ] ) by an an integral over a finite interval , @xmath461 : @xmath462\,dz & = & \frac{1}{8\pi r^2 } \frac{\partial}{\partial r } \left[r^2 \int_{-h(r)}^{h(r ) } \left(b_r^2-b_\phi^2\right)\,dz \right]\nonumber\\ & & \mbox { } - \frac{1}{4\pi } \left(b_{r , s}^2-b_{\phi , s}^2\right)\left(\frac{dh}{dr}\right)\end{aligned}\ ] ] ( see * ? ?
we assume @xmath463 $ ] , and similarly for @xmath464 .
this choice is motivated by the field configuration derived in the case of a rotationally supported thin disk in which the field is comparatively well coupled to the matter ( e.g. , * ? ? ?
although the disks obtained in our ad collapse solutions evidently correspond to the weak - coupling case ( see [ discussion ] and appendix [ wind ] ) , for which this approximation is no longer adequate ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , we nevertheless adopt the above scalings for the sake of definiteness .
we emphasize , however , that none of the dominant terms in the equations that we solve depends on the details of the vertical variation of these field components .
we thus get @xmath462\,dz & = & \frac{1}{12\pi r^2 } \frac{\partial}{\partial r } \left[r^2 h(r ) \left(b_{r , s}^2-b_{\phi , s}^2\right ) \right ] - \frac{1}{4\pi } \left(b_{r , s}^2-b_{\phi , s}^2\right)\left(\frac{dh}{dr}\right)\nonumber\\ & = & \frac{h}{12\pi r^2}\frac{\partial}{\partial r } \left[r^2(r^2 b_{r , s}^2-b_{\phi , s}^2)\right ] - \frac{1}{6\pi}\left ( \frac{dh}{dr } \right ) \left(b_{r , s}^2-b_{\phi , s}^2\right)\ , .\end{aligned}\ ] ] approximating @xmath128 by @xmath465 [ with @xmath466 if the central mass dominates ] and substituting @xmath109 , the integrated radial momentum equation becomes @xmath467 when integrating the angular momentum equation ( [ angular3d ] ) , we write @xmath468 and use the same ansatz as above for @xmath469 and @xmath470 .
we then obtain @xmath471 in integrating the vertical hydrostatic - balance equation , we take the pressure at the disk surfaces to vanish
. the midplane pressure @xmath472 is then given by @xmath473 where the first two terms on the right - hand side constitute our approximation to @xmath449 and represent , respectively , the local disk self - gravity and the tidal squeezing by the central point mass .
the ion equation of motion is approximated by @xmath474 where @xmath475 ; we assume that the disk is sufficiently weakly ionized that the neutral velocity @xmath476 is practically indistinguishable from the bulk velocity @xmath477 .
the radial component of this equation is @xmath478 which yields , upon integration over @xmath116 , @xmath479\ ] ] ( see eq . [ [ radial3d2 ] ] ) .
similarly , the azimuthal component of equation ( [ drift3d ] ) , @xmath480 gives , when integrated over @xmath116 , @xmath481 ( see eq .
[ [ angular3d1 ] ] ) .
the flux conservation equation is obtained from faraday s law ( @xmath482 ) and ohm s law in the ambipolar diffusion - dominated regime ( @xmath483 ) together with equation ( [ drift3d ] ) .
the result is @xmath484 \times{{\mathbf{b}}}\right\}\ .\ ] ] the @xmath116 component of this equation is @xmath485 \right.\right . \nonumber\\ & & \left.\left.\mbox { } -b_r \left [ b_\phi \frac{\partial b_\phi}{\partial z } + b_r \left ( \frac{\partial b_r}{\partial z } - \frac{\partial b_z}{\partial r } \right ) \right ] \right\ } \right)\ , \end{aligned}\ ] ] which implies @xmath486 integrating this over @xmath116 and rearranging , gives @xmath487 \nonumber\\ & - & \left .
\frac{1}{3}b_{r , s}^2 \left [ hb_z\left ( \frac{d}{dr}\ln(rb_{r , s } ) - \frac{d\ln h}{dr}\right)\right]\right\ } \ .\end{aligned}\ ] ] in formulating the self - similar model in [ basiceqns ]
we use a pared - down version of the above equations .
this is motivated by our finding that our basic results can be derived _ without _ including any @xmath118 terms .
the only instance in which a term of this order plays a role in our solutions is in allowing us to refine the structure of the ambipolar - diffusion shock by employing the radial gradient of @xmath63 in the radial momentum equation .
we therefore retain the combination @xmath488 $ ] ( which is proportional to the vertically integrated azimuthal component of the current density ) in equation ( [ radial3d2 ] ) but neglect the other two @xmath118 terms in that equation , since they typically do not exceed the @xmath489 term . [ the last two terms in equation ( [ radial3d2 ] ) can become much larger than the @xmath489 term in the limit @xmath490 of the imhd solution , when @xmath491 formally diverges .
however , as we argue in [ results ] , this divergence is not expected to occur in a real disk .
furthermore , the neglect of these terms is justified even in our formal solution since @xmath492 in this limit , so the radial derivative terms remain negligible in comparison with the magnetic tension ( @xmath493 ) term .
] since the same integrated current density component appears also in equations ( [ v_dr ] ) and ( [ psifull ] ) for @xmath494 and @xmath495 , respectively , we keep the combination @xmath496 $ ] in these equations too for self - consistency , but we again omit all the other @xmath118 terms .
for the same reason we keep the @xmath497 term also in the vertical force balance equation ( [ vertical3d1 ] ) , which means that we end up retaining all the @xmath118 terms in this `` master equation '' for @xmath103 .
we similarly neglect the @xmath118 terms in the angular momentum equation ( [ angular3d1 ] ) and , correspondingly , in equation ( [ viphi ] ) for @xmath498 .
the latter equation can then be substituted into equation ( [ bm ] ) to yield a purely algebraic expression for @xmath78 .
the singular lines of the flows considered in this paper can be obtained by rewriting the radial momentum equation ( [ forceequation ] ) in the form @xmath499\ .\ ] ] in the ad limit , in which the field is effectively decoupled from the bulk of the matter , we immediately infer that the singular line corresponds to the sonic speed : @xmath500 ( see cck for a discussion of the physical basis of this result ) . in the ideal - mhd limit ,
flux freezing implies @xmath501 ( see eq .
[ [ initialconditions ] ] ) , and therefore the singular line is given by @xmath502 in the region where self - gravity dominates , @xmath503 , and equation ( [ singularimhd ] ) reduces to @xmath504 , which is the result given by equation ( 41 ) in cck . ) . ] in dimensional form , equation ( [ singularimhd ] ) is @xmath505 where the last term on the right - hand side is equal to the alfvn speed in the disk .
the imhd singular line thus corresponds to the fast - magnetosonic speed , which agrees with ( and generalizes ) the statement made in cck on the basis of the self - gravity dominated limit of equation ( [ singularimhd ] ) .
when the curve describing the flow crosses and falls below the singular line in the @xmath506 plane , this generally signals the appearance of an unresolved shock discontinuity . in the present work , this is the situation invariably encountered in the vicinity of the centrifugal radius . to properly model this discontinuity , appropriate shock jump conditions need to be applied . the conservation of mass across the shock yields @xmath507 where @xmath508 ( with @xmath509 denoting the shock location ) .
when the ions and neutrals are well coupled inside the shock , the magnetic - flux conservation condition takes the form @xmath510 in the strong - coupling case , equations ( [ massflux ] ) and ( [ magneticflux ] ) imply @xmath511 across the shock : in what follows we set @xmath512 , which applies when the entire flow upstream of the shock obeys ideal mhd . under the adopted isothermal approximation ,
the jump conditions are fully specified once we integrate also the differential terms in equation ( [ shockforceequation ] ) across the discontinuity .
performing this integral requires making some assumptions about the behavior of @xmath145 , @xmath254 , and @xmath143 in the shock region . in the following subsections , we consider four cases in which algebraic jump conditions can be derived .
these results have enabled us to explicitly model the centrifugal shock in all the core - collapse solutions presented in this paper .
if one can neglect variations in the magnetic field across the discontinuity , as is the case when the centrifugal shock occurs deep inside the ad regime ( see [ adstandard ] ) , equations ( [ shockforceequation ] ) and ( [ massflux ] ) imply @xmath513 denoting the upstream and downstream sides of the shock by the subscripts 1 and 2 , respectively , one can use equations ( [ isojump ] ) and ( [ massflux ] ) to express the nontrivial ( @xmath514 ) solution in the form @xmath515 .
this is the simplest realization of the `` isothermal '' jump conditions discussed by @xcite .
if the shock occurs in the self - gravitating regime , where @xmath503 , then @xmath516 is constant across the shock , and the integration of equation ( [ shockforceequation ] ) leads to @xmath517 where @xmath518 .
this situation is encountered in the fast - rotation solutions presented in [ mhdfast ] and [ adfast ] .
in this case the upstream and downstream radial velocities ( in the shock frame ) are related by @xmath519 , which constitutes another realization of the `` isothermal '' jump conditions . remark .
[ note that , as @xmath520 , one recovers the singular - line relation ( eq .
[ [ singularimhd ] ] ) for a self - gravitating imhd flow . ]
if magnetic squeezing due to the radial field dominates gravity in the vertical force - balance equation , then @xmath521 ( see eq .
[ [ mhdasymptoticend ] ] ) , and equation ( [ shockforceequation ] ) yields @xmath522 taking into account the relation ( [ massflux ] ) , we obtain a quartic equation for @xmath523 ( whose coefficients are combinations of the upstream flow variables ) .
we can eliminate the trivial solution @xmath524 by dividing this equation by @xmath525 , which reduces it to @xmath526 where @xmath527 .
equation ( [ shockcubic ] ) is a cubic of the form @xmath528 , whose discriminant is @xmath529 , with @xmath530 and @xmath531 . for this particular cubic , the discriminant is always positive , guaranteeing that there is exactly one real root , given by @xmath532 where @xmath533 .
this solution provides a good approximation to the centrifugal shock in the fiducial imhd flow ( [ mhdstandard ] ) .
another possibility is that @xmath143 is constant across the shock , as would be the case if the central point - mass gravity controlled the squeezing in the vertical equilibrium equation ( see eqs . [
[ asymptotich ] ] and [ [ adstrongasymptoticend ] ] ) .
in this case , integration of equation ( [ shockforceequation ] ) results in @xmath534 in conjunction with the mass conservation equation ( [ massflux ] ) , this can be written as a cubic equation for @xmath523 .
after factoring out the trivial solution , one obtains a quadratic equation , @xmath535 which has exactly one positive real root . in order for this result to be applicable to a centrifugal shock in a collapsing magnetized core
, the shock must occur far enough from the center that ideal mhd is still a good approximation ( so that the magnetic term in eq .
[ [ shockforceequation ] ] continues to play a role ) .
however , in the solutions explored in this paper , the requirements that the central gravity dominate the vertical squeezing and that the imhd regime be applicable are never satisfied at the location of the centrifugal shock : in the fiducial and strong - gravity imhd cases , magnetic squeezing exceeds the tidal force in the central gravity - dominated region , whereas in the fast - rotation imhd and ad cases , self - gravity dominates at the location of the shock .
as we note in [ discussion ] , the asymptotic ambipolar - diffusion disk solution satisfies the launching condition @xmath536 for a cold , centrifugally driven wind , and , in fact , implies a radial scaling of the magnetic field components ( at any given instant of time ) that is identical to that of the time - independent , radially self - similar wind solution of @xcite ( @xcite , hereafter bp ) . once a super - alfvnic wind of this type is established , it constitutes the dominant mechanism of removing angular momentum from the disk ( at least in the vertical direction ) . furthermore ,
if the outflow is quasi steady on the accretion time scale , then the spatially self - similar wind solution can be used to determine the value of @xmath78 ( or , equivalently , of the parameter @xmath277 ) in the asymptotic self - similar ( in space and time ) disk solution .
an explicit wind solution also makes it possible to evaluate the reduction in the mass accretion rate onto the central star caused by the diversion of part of the inflowing mass into a disk outflow .
motivated by these considerations , we now show how the centrifugally driven wind solution can be incorporated into the asymptotic ad disk model . as demonstrated in [ adstandard ] ,
the numerical solution of the disk equations typically converges rapidly behind the centrifugal shock to the asymptotic form given by equations ( [ asymptoticm])([asymptotich ] ) .
this indicates that , if the model parameters do not strongly differ from the fiducial values , then the results that we obtain for the asymptotic regime should be applicable to the bulk of the rotationally supported disk that forms around the central object .
the mass loss in the bp model is constant for each decade of disk radius and can be related to the accretion rate @xmath537 at the inner edge of the disk by @xmath538 where @xmath539 is the fraction of the disk binding energy at its inner edge @xmath540 that is carried off by the wind , and where @xmath148 is the total ( kinetic plus magnetic ) specific angular momentum in the wind , normalized by the keplerian disk value @xmath541 .
the field - line constant @xmath148 is one of the three parameters that define a solution in the cold , self - similar bp wind model , the other two being @xmath542 and @xmath543 ( the normalized mass - to - magnetic flux ratio evaluated at the disk surface ) . as we noted above , for any given value of @xmath391 in the asymptotic ad disk solution
, @xmath491 is a spatial constant ( @xmath544 ; see eq .
[ [ adpsistandard ] ] ) that exceeds the launching threshold @xmath545 .
it is thus natural to set @xmath546 .
one can then evaluate the parameter @xmath547 by equating @xmath548 with the expression given by equation ( [ mdot_wind ] ) ( where the factor of 2 accounts for the two disk surfaces ) .
the result ( in terms of the nondimensional flow quantities introduced in [ selfsimilareqns ] ) is @xmath549 substituting for @xmath550 and @xmath254 from the asymptotic - solution equations ( [ asymptoticm ] ) and ( [ adbz ] ) , respectively , this becomes @xmath551 we now substitute for @xmath277 in the asymptotic disk solution the corresponding expression in the bp wind model , @xmath552 which yields @xmath553 although we started by evaluating @xmath547 , we ended up with an expression that gives @xmath148 in terms of @xmath554 .
the latter parameter was treated by bp as being distinct from @xmath148 , but we are now able to relate them by combining the disk and wind solutions . by fixing the values of @xmath555 and @xmath148 , one can derive the value of @xmath547 from the solution of the self - similar wind equations ( see fig . 2 in bp ) , and thereby obtain @xmath277 ( using eq .
[ [ delta ] ] ) . in practice , this `` recipe '' for incorporating the wind model does not alter our basic disk model since it effectively amounts to replacing one somewhat arbitrary parameter ( @xmath277 ) by another ( @xmath554 ) .
however , by combining the disk and wind solutions , it is possible to constrain the parameter values . in particular , by requiring that the flow be magnetically dominated in the vicinity of the disk ( which implies @xmath556 ) and that it become super - alfvnic at a finite height , one can infer a lower bound on @xmath148 .
based on figure 2 in bp , this lower limit is @xmath410 for @xmath546 ( which , by eq .
[ [ lambda ] ] , implies that @xmath554 must be @xmath557 ) , and as @xmath148 increases from @xmath410 to @xmath558 ( corresponding to @xmath554 increasing from @xmath559 to @xmath560 ) , @xmath277 decreases from @xmath561 to @xmath562 . in the absence of a disk outflow , the central mass eigenvalue @xmath185 is given approximately by the product @xmath563 evaluated at @xmath195 ( see eqs . [
[ mdot ] ] and [ [ asymptoticm ] ] ) .
when the mass loss between the outer edge of the disk ( @xmath564 ) and its inner edge ( @xmath565 , the protostellar radius ) is taken into account in the mass conservation equation , the value of @xmath185 is reduced to @xmath566^{-1}(-xu\sigma)_{x_c}\ , , \ ] ] where @xmath567 ( see eqs .
[ [ mass4d ] ] , [ [ mdot_wind ] ] , and [ [ lambda ] ] ) .
to the extent that deuterium burning causes the radius of the accreting protostar to increase roughly linearly with its mass ( e.g. , * ? ? ?
* ) , @xmath568 can be approximated as a constant .
estimating @xmath1 from our fiducial ad solution ( see eq .
[ [ xcadapproximation ] ] ) and @xmath569 from the mass radius relation given by @xcite , we obtain @xmath570 .
equation ( [ m_cw ] ) then implies that mass outflow from the disk surfaces could reduce the accretion rate onto the central object by a factor as large as @xmath410 ( using the minimum value of @xmath148 , which corresponds to @xmath571 ) .
more strongly magnetized outflows have lower @xmath185 reduction factors : for example , when @xmath547 decreases to @xmath572 ( corresponding to @xmath573 and @xmath574 ) , this factor declines to @xmath575 .
although the cumulative effect of the mass loss from the wind surfaces ( embodied in the factor @xmath576 ) can be significant , its impact at any given radius is comparatively small because of the large range of radii involved ( @xmath577 in the fiducial case ) .
the incorporation of a disk wind modifies the integrated self - similar disk equations only through logarithmic terms in @xmath134 , so to lowest order the power - law scalings of the asymptotic disk solution remain unchanged .
explicit models of ad - dominated , wind - driving accretion disks were previously constructed by @xcite and @xcite .
two general types of solutions were obtained , depending on the value of the neutral ion coupling parameter @xmath578 .
strongly coupled disks have @xmath579 at all heights and are characterized by the thermal pressure not being much larger than the magnetic pressure at @xmath102 , by @xmath580 starting to increase already near the midplane and generally exceeding @xmath581 at the disk surface ( which leads to a strong magnetic squeezing of the disk ) , and by the inflow speed typically being @xmath582 @xcite .
in contrast , in weakly coupled disks ( characterized by @xmath583 over the bulk of the vertical column ) the thermal pressure can be much larger than the magnetic pressure at @xmath102 , @xmath580 only starts to grow well above the midplane ( at a height where the coupling parameter finally rises above 1 ) and usually does not exceed @xmath581 at the surface ( which results in the magnetic squeezing remaining relatively unimportant ) , and the mass - averaged inflow speed is typically well below @xmath66 @xcite . as we point out in
[ discussion ] , @xmath584 is equal to @xmath416 in the asymptotic ad disk solution .
this result is compatible with the just - described behavior of @xmath129 and @xmath132 : since @xmath585 in the asymptotic solution and @xmath586 , the expression for @xmath584 implies @xmath587 [ consistent with eq . ( 4.5 ) in @xcite and eq .
( 39 ) in @xcite ] .
the values of @xmath277 derived from the bp wind solution indicate that our ad disks are likely to be weakly coupled .
this conclusion is consistent with the low magnetic - to - thermal pressure ratio implied by the asymptotic disk solution ( it is @xmath588 as @xmath90 at any given instant of time ; see eqs .
[ [ adsigma]][[asymptotich ] ] ) , which corresponds to the situation in weakly coupled disks .
the association with the weakly coupled class of solutions is further indicated by the unimportance of magnetic squeezing in the asymptotic expression for the disk half - thickness ( eq . [ [ asymptotich ] ] ) and by the subsonic inflow speed inferred in the asymptotic regime ( @xmath589 as @xmath90 and @xmath391 is held fixed ; see eq .
[ [ adu ] ] ) .
the inner regions of real protostellar disks ( on radial scales @xmath590 ) are , in fact , expected to be so weakly ionized over most of their interiors that the requisite field - line bending for launching a centrifugal wind can only occur in comparatively thin surface layers ( e.g. , * ? ? ?
as was , however , already emphasized by @xcite , the properties of such disks and of their associated outflows depend sensitively on the details of the spatial profile of @xmath584 .
further progress in developing these models will thus require a good understanding of the ionization structure in the inner regions of protostellar disks .
balbus , s. a. , & hawley , j. f. 1998 , rev .
, 70 , 1 basu , s. 1997 , , 485 , 240 basu , s. 1998 , , 509 , 229 basu , s. , & mouschovias , t. ch . 1994 , , 432 , 720 basu , s. , & mouschovias , t. ch .
1995a , , 452 , 386 basu , s. , & mouschovias , t. ch .
1995b , , 453 , 271 beckwith , s. v. w. , & sargent , a. i. 1993 , in protostars and planets iii , ed . e. h. levy & j. i. lunine ( tucson : university of arizona press ) , 521 blaes , o. m. , & balbus , s. a. 1994 , , 421 , 163 blandford , r. d. , & payne , d. g. 1982 , , 199 , 883 ( bp ) bodenheimer , p. , burkert , a. , klein , r. i. , & boss , a. p. 2000 , in protostars and planets iv , ed .
v. mannings , a. p. boss , & s. s. russell ( tucson : university of arizona press ) , 675 boss , a. p. 2000 , , 545
, l61 calvet , n. , hartmann , l. , & strom , s. e. 2000 , in protostars and planets iv , ed . v. mannings , a. p. boss , & s. s. russell ( tucson : university of arizona press ) , 377 cassen , p. m. , & summers , a. l. 1983 , icarus , 53 , 26 ciolek , g. e. , & knigl , a. 1998 , , 504 , 257 ( ck ) ciolek , g. e. , & mouschovias , t. ch . 1998 , , 504 , 280 contopoulos , i. , ciolek , g. e. , & knigl , a. 1998 , , 504 , 247 ( cck ) contopoulos , i. , & sauty , c. 2001 , , 365 , 165 crutcher , r. m. 1999 , , 520 , 706 dalessio , p. , canto , j. , calvet , n. , & lizano , s. 1998 , , 500 , 411 dent , w. r. f. , walker , h. j. , holland , w. s. , & greaves , j. s. 2000 , , 314 , 702 desch , s. j. , & mouschovias , t. ch .
2001 , , 550 , 314 dutrey , a. , guilloteau , s. , prato , l. , simon , m. , duvert , g. , schuster , k. , & menard , f. 1998 , , 338 , l63 fiedler , r. a. , & mouschovias , t. ch .
1993 , , 415 , 680 galli , d. , & shu , f. h. 1993 , , 417 , 243 goodman , a. a. , benson , p. j. , fuller , g. a. , & myers , p. c. 1993 , , 406 , 528 goodson , a. p. , & winglee , r. m. 1999 , , 524 , 159 guilloteau , s. , & dutrey , a. 1998 , , 339 , 467 hartmann , l. , calvet , n. , gullbring , e. , & dalessio , p. 1998
, , 495 , 385 hawley , j. f. , & stone , j. m. 1998 , , 501 , 758 hayashi , c. , nakazawa , k. , & nakagawa , y. 1985 , in protostars & planets ii , ed . d. c. black & m. s. mathews ( tucson : university of arizona press ) , 1100 indebetouw , r. & zweibel , e. g. 2000 , , 532 , 361 kamaya , h. , & nishi , r. 2000 , , 543 , 257 kane , b. d. , & clemens , d. p. 1997
, , 113 , 1799 koerner , d. w. , sargent , a. i. , & ostroff , n. a. 2001 , , 560 , l181 knigl , a. 1987 , , 320 , 726 knigl , a. 1991 , , 370 , l79 knigl , a. , & pudritz , r. e. 2000 , in protostars and planets iv , ed . v. mannings , a. p. boss , & s. s. russell ( tucson : university of arizona press ) , 759 li , z .- y .
1996 , , 465 , 855 li , z .- y .
1998 , , 497 , 850 li , z .- y . , & mckee , c. f. 1996 , , 464 , 373 li , z .- y . , & shu , f. h. 1997 , , 475 , 237 lin , d. n. c. , & pringle , j. e. 1987 , , 225 , 607 lin , d. n. c. , & pringle , j. e. 1990 , , 358 , 515 lovelace , r. v. e. , romanova , m. m. , & newman , w. i. 1994 , , 437 , 136 matsumoto , t. , hanawa , t. , & nakamura , f. 1997 , , 478 , 569 mestel , l. , & spruit , h. c. 1987 , , 226 , 57 mineshige , s. , & umemura m. 1997 , , 480 , 167 mouschovias , t. ch . , & ciolek , g. e. 1999 , in the origins of stars and planetary systems , ed . c. j. lada & n. d. kylafis ( dordrecht : kluwer ) , 305 mouschovias , t. ch . ,
& paleologou , e. v. 1986 , , 308 , 781 mundy , l. g. , looney , l. w. , & welch , w. j. 2000 , in protostars and planets iv , ed .
v. mannings , a. p. boss , & s. s. russell ( tucson : university of arizona press ) , 355 myers , p. c. , evans , n. j. , & ohashi , n. 2000 , in protostars and planets iv , ed . v. mannings , a. p. boss , & s. s. russell ( tucson : university of arizona press ) , 217 nakamura , f. , hanawa , t. , & nakano , t. 1995 , , 444 , 770 nakamura , f. , matsumoto , t. , hanawa , t. , & tomisaka , k. 1999 , , 510 , 274 narita , s. , hayashi , c. , & miyama , s. m. 1984 , prog .
, 72 , 1118 neufeld , d. a. , & hollenbach , d. j. 1994 , , 428 , 170 nishi , r. , nakano , t. , & umebayashi , t. 1991 , , 368 , 181 norman , m. l. , wilson , j. r. , & barton , r. t. 1980 , , 239 , 968 , n. , hayashi , m. , ho , p. t. p. , momose , m. , tamura , m. , hirano , n. , & sargent , a. i. 1997 , , 488 , 317 popham , r. 1996 , , 467 , 749 saigo , k. , & hanawa , t. 1998 , , 493 , 342 schneider , g. , et al . 1999 , , 513 ,
l127 shakura , n. i. , & sunyaev , r. a. 1973 , , 24 , 337 shu , f. h. , & li , z .- y .
1997 , , 475 , 251 stahler , s. w. 1988 , , 332 , 804 stassun , k. g. , mathieu , r. d. , mazeh , t. , & vrba , f. j. 1999 , , 117 , 2941 stassun , k. g. , mathieu , r. d. , vrba , f. j. , mazeh , t. , & henden , a. 2001 , , 121 , 1003 terebey , s. , shu , f. h. , & cassen , p. 1984
, , 286 , 529 tomisaka , k. 1996 , , 48 , 701 tomisaka , k. 1998 , , 502 , l163 tomisaka , k. 2000 , , 528 , l41 tomisaka , k. 2002 , , 575 , 306 toomre , a. 1964 , , 139 , 1217 tsuribe , t. 1999 , , 527 , 102 wardle , m. 1997 , in asp conf .
121 , accretion phenomena and related outflows , ed .
d. t. wickramasinghe , g. v. bicknell , & l. ferrario ( san francisco : asp ) , 561 wardle , m. , & knigl , a. 1993 , , 410 , 218 wardle , m. , & ng , c. 1999 , , 303 , 239 weidenschilling , s. j. 1977 , , 51 , 153 wilner , d. j. , & lay , o. p. 2000 , in protostars and planets iv , ed . v. mannings , a. p. boss , & s. s. russell ( tucson : university of arizona press ) , 509 wuchterl , g. , guillot , t. , & lissauer , j. j. 2000 , in protostars and planets iv , ed .
v. mannings , a. p. boss , & s. s. russell ( tucson : university of arizona press ) , 1081 yorke , h. w. , bodenheimer , p. , & laughlin , g. 1993 , , 411 , 274 zweibel , e. g. 1998 , , 499 , 746 | we present self - similar solutions that describe the gravitational collapse of rotating , isothermal , magnetic molecular - cloud cores .
these solutions make it possible , for the first time , to study the formation of rotationally supported protostellar disks of the type detected around many young stellar objects in the context of a realistic scenario of star formation in magnetically supported , weakly ionized , molecular cloud cores .
this work focuses on the evolution after a point mass first forms at the center and generalizes previous results by contopoulos , ciolek , & knigl that did not include rotation .
our semianalytic scheme incorporates ambipolar diffusion and magnetic braking and allows us to examine the full range of expected behaviors and their dependence on the physical parameters .
we find that , for typical parameter values , the inflow first passes through an ambipolar - diffusion shock ( at a radius @xmath0 ) , where the magnetic flux decouples from the matter , and subsequently through a centrifugal shock ( at @xmath1 ) , inward of which a rotationally supported disk ( of mass @xmath2 ) is established . by the time ( @xmath3 ) that the central mass @xmath4 grows to @xmath5 , @xmath6 , @xmath7 , and @xmath8 .
the derived disk properties are consistent with data on t tauri systems , and our results imply that protostellar disks may well be keplerian also during earlier phases of their evolution .
we demonstrate that the disk is likely to drive centrifugal outflows that transport angular momentum and mass , and we show how the radially self - similar wind solution of blandford & payne can be naturally incorporated into the disk model .
we further verify that gravitational torques and magnetorotational instability - induced turbulence typically do not play an important role in the angular momentum transport . for completeness
, we also present solutions for the limiting cases of fast rotation ( where the collapse results in a massive disk with such a large outer radius that it traps the ambipolar - diffusion front ) and strong braking ( where no disk is formed and the collapse resembles that of a nonrotating core at small radii ) , as well as solutions describing the rotational collapse of ideal - mhd and of nonmagnetic model cores . |
special relativity , as presented in today s textbooks , is a complex mathematical theory .
the 1 spatial + 1 temporal dimensional minkowski diagrams , which initially introduce the lorentz transformation , the time dilatation and the length contraction , are soon put aside in favor of an approach based on differential calculus and linear algebra .
one gets little intuitive understanding of the law of relativistic addition of velocities , and of the fact that `` magnetism is a kind of second - order effect arising from relativistic changes in the electric fields of moving charges '' @xcite .
however , by introducing only slightly more elaborate minkowski diagrams , and using geometrical derivations , one can get back the intuitive understanding , to the great delight of the physicist who still believes in the spirit of descartes philosophy .
the complex plane , like the real plane , is a two - dimensional ( 2d ) vector space .
the scalar products for the real plane ( 1 ) and for the complex plane ( 2 ) are defined as follows @xmath0 @xmath1 where @xmath2,@xmath3 and respectively @xmath2,@xmath4 are the basis vectors .
since both planes have a scalar product , one can talk about orthogonal vectors ( their scalar product is zero ) and about the magnitude of a vector ( the square root of the scalar product of a vector with itself ) .
this allows us to define the circle ( the geometric locus of the points equally spaced from a given point ) , the angle in radians ( the length , between two points of a circle of radius one , measured along the circumference ) , and the trigonometric functions sine and cosine ( the magnitudes of the projections of a radius one vector on the two coordinate axes ) . from these definitions it follows that , for both the real and the complex planes , one has the relations : @xmath5 @xmath6 ^ 2 + [ { d \over d\alpha } \cos(\alpha ) ] ^2 = 1.\ ] ] from ( 3)-(4 ) one can get the derivatives of the trigonometric functions , the taylor series expansions of sine and cosine , and then all the well known trigonometric relations .
the only detail we have to keep in mind is that the angle @xmath7 in the real plane is a real number , while in the complex plane it is a purely imaginary number , due to the non - positive definite scalar product used in the last case .
there are a few more relevant differences , which we can best point out if we represent the complex plane as an euclidean plane .
two vectors in the complex plane are orthogonal if they make the same angle with the bisecting line of the first quadrant .
the circle in the complex plane looks like a hyperbola @xcite .
not any line passing through the origin intersects the right ( or left ) branch of the hyperbola .
this means that there are pairs of lines passing through the origin to which we can not assign an angle .
however , for the triangles we will be working with , the ratio of segments behaves as if it were an angle , of negative value .
the true angle is obtained by symmetry with respect to the first bisecting line , as the pair @xmath7 and @xmath8 indicates in figure [ velocity ] .
consider a reference frame k which is moving with a velocity @xmath9 relative to another one k , and a particle moving with a velocity @xmath10 in the reference frame k. the reference frames are chosen such that their origins and the particle coincide at the space - time point o , as shown in figure [ velocity ] .
notice that @xmath11 and @xmath12 , because @xmath13 has a component only in the x direction .
the oz axis is not plotted , but is similar to the oy axis .
the question is : what is the velocity @xmath14 of the particle in the reference frame k ?
the world - line op of the particle is projected on the complex planes ( x , o , ict ) , ( y , o , ict ) , ( z , o , ict ) , ( x,o , ict ) , ( y,o , ict ) , ( z,o , ict ) , and the resulting angles from the respective projections give the components of the velocity of the particle in the two reference frames considered .
for the situation considered the planes ( x , o , ict ) and ( x,o , ict ) coincide .
it is seen from figure [ velocity ] that @xmath15 @xmath16 @xmath17 @xmath18 @xmath19 in order to express @xmath20 and @xmath21 as functions of @xmath22 and @xmath23 we need to express @xmath24 and @xmath25 as functions of @xmath7,@xmath26 and @xmath27 . in the plane ( x , o , ict ) of the lorentz
boost the addition of velocities is based on the addition of angles @xcite @xmath28 @xmath29 from ( 11 ) , by substitution of the tangents ( 5)-(9 ) , it follows that @xmath30 two rectangles , apcd and bpce , result from the projection process .
it is evident that @xmath31 @xmath32 from ( 13)-(14 ) it follows that @xmath33}.\ ] ] by substitutions of the tangents ( 5)-(8 ) and of @xmath34^{-1/2}$ ] we get @xmath35 a similar expression is obtained for the @xmath36 component .
consider two charged particles ( with charges @xmath37 and @xmath38 ) at some arbitrary positions , moving with arbitrary , but uniform , velocities .
we orient our 3d reference frame in such a way that the first particle ( which generates the field ) is initially at the origin , moving along the ox axis with velocity @xmath9 , and the vector @xmath39 connecting the two particles is in the ( x , o , y ) plane .
the angle between @xmath40 and the ox axis is @xmath25 .
the second particle ( subject to the electromagnetic field generated by the first one ) is moving with velocity @xmath14 .
a section through the ( x , o , y ) plane can be seen in figure [ eminter ] .
the first particle is at point o and the second one is at point a. the electric field ( in gaussian units ) generated by the first particle at the position of the second particle is @xcite @xmath41^{-3/2}.\ ] ] the magnetic field generated by the first particle is @xmath42 the lorentz force acting on the second particle is @xmath43 from ( 17)-(19 ) the cartesian components of the force @xcite are obtained @xmath44^{-3/2 } [ \cos ( \theta ) + \sin ( \theta ) { v_y v \over c^2}]\ ] ] @xmath45^{-3/2 } \sin ( \theta ) ( 1 - { v_x v \over c^2})\ ] ] @xmath46 the force components ( 20)-(22 ) can be obtained in a more graphical way , if we start with the coulomb force generated by a charged particle at rest . one key assumption or experimental fact
is that in a frame where all the source charges producing an electric field @xmath47 are at rest , the force on a charge @xmath48 is given by @xmath49 independent of the velocity of the charge in that frame @xcite .
the reference frame k in which the source particle is at rest is moving with velocity @xmath13 relative to the original frame k. in the reference frame k the particle at a is observed to interact with the particle at o. the distance between particles is @xmath50 , the length of the segment @xmath51 . in the reference frame k the particle at a is observed to interact with the particle at b , where the segment @xmath52 is a position vector @xmath53 parallel to the plane ( x,o , y ) .
the following construction gives the position of point b : the segment @xmath54 is parallel to oy and intersects the ox axis at e , whereas the segment @xmath55 is parallel to ox and intersects the world - line @xmath56 at b. @xmath57 projects the point b on the ox axis at d. relative to k , the particle at b exerts a radial coulomb force on the particle at a. this force ( in gaussian units ) is @xmath58 where @xmath59 $ ] . the key point in getting the force @xmath60 in the reference frame k is to notice that the force , in any reference frame considered , is given by the projection on the real 3d space of that frame of the 4-force @xmath61 ( which is a minkowski - space vector ) , that is @xmath62 @xmath63 where @xmath64 and @xmath65 .
we will obtain the 4-force @xmath61 from its real and imaginary components ( @xmath66 and @xmath67 ) in the reference frame k , then we will decompose the same 4-force into its real and imaginary components ( @xmath68 and @xmath69 ) in the reference frame k. the lorentz force we are looking for is just @xmath70 . from ( 23)-(25 ) it follows that @xmath71 to get the imaginary component @xmath67 we use the orthogonality between the 4-force and the 4-velocity , @xmath72 , where the 4-velocity is @xmath73 .
the orthogonality condition leads to @xmath74 @xmath75 where the radial component of the velocity is @xmath76 the components of the force @xmath60 in the reference frame k are given by the projection of the 4-force @xmath61 on the 3d real space of k. an easy way to do this is to notice that we can decompose @xmath66 ( which has the direction of the segment @xmath52 ) and @xmath67 ( which has the direction of the segment @xmath77 ) into sums of 4-vectors , each of the 4-vectors being parallel to one of the axes of the reference frame k : @xmath78 @xmath79 because these expansions do not involve any component along the oz axis , this simply means that @xmath80 .
the projections of the 4-force on the ox and oy axes are @xmath81 @xmath82 the lengths of the various segments needed above are as follows : @xmath83 @xmath84 @xmath85 @xmath86 @xmath87 @xmath88^{1/2 } = r'\ ] ] we also notice that @xmath89 .
the force components in ( 32)-(33 ) become @xmath90^{3/2 } } \nonumber \\ & & + i { \gamma(v ' ) \over \gamma(v ) } { q_1 q_2 \over r^2 } { v'_{\rm rad } \over c } { \sin ( -\alpha ) \over \cos ^2(\alpha ) [ 1 + \tan ^2(\alpha ) \sin ^2(\theta)]}\end{aligned}\ ] ] @xmath91^{3/2}}.\ ] ] we can also calculate @xmath92^{1/2}}\ ] ] @xmath93^{1/2}}.\ ] ] if the velocity of the particle at a has the components @xmath94 , as measured in the reference frame k , and k is moving with the velocity @xmath95 relative to k , then the particle will have the following components of the velocity ( compare with equations ( 12 ) and ( 16 ) ) in the reference frame k @xmath96 with these components we find that @xmath97 and the radial velocity ( 29 ) becomes @xmath98^{1/2}}.\ ] ] substituting @xmath99 and @xmath100 in ( 40)-(41 ) , and also using the fact that @xmath101 , @xmath102 and @xmath103 , we finally obtain the components in ( 20)-(21 ) .
we have presented a geometrical calculation of the relativistic addition of velocities , and of the electromagnetic interaction between two uniformly moving charged particles .
the geometrical approach used is an elegant , more intuitive and alternative way of obtaining these important results of special relativity .
we hope our work will usefully complement other pedagogical efforts @xcite centered on minkowski space diagrams .
purcell e m , _ electricity and magnetism _
( new york : mcgraw - hill , 1985 ) p 238 minkowski h , _ space and time _ , reprinted in einstein a , lorentz h a , weyl h , minkowski h , _ the principle of relativity _
( new - york : dover , 1952 ) einstein a , _ the meaning of relativity _ ( princeton : princeton university press , 1988 ) p 37 landau l and lifchitz e , _ th ' eorie du champ _
( moscow : mir , 1966 ) p 119 donnelly r and ziolkowski r w , electromagnetic field generated by a moving point charge : a fields - only approach , am . j. phys .
* 62 * ( 10 ) , october 1994 , p 916 rosser w g v , _ classical electromagnetism via relativity _
( new york : plenum press , 1968 ) p 32 jackson j d , _ classical electrodynamics _ ( new york : wiley , 1975 ) p 580 mermin d , an introduction to space - time diagrams , am .
* 65 * ( 6 ) , june 1997 , p 476 mermin d , space - time intervals as light rectangles , am .
* 66 * ( 12 ) , december 1998 , p 1077 saletan e j , minkowski diagrams in momentum space , am . j. phys .
* 65 * ( 8) , august 1997 , p 799 | this paper presents an intuitive , geometrical derivation of the relativistic addition of velocities , and of the electromagnetic interaction between two uniformly moving charged particles , based on 2 spatial + 1 temporal dimensional minkowski diagrams .
we calculate the relativistic addition of velocities by projecting the world - line of the particle on the spatio - temporal planes of the reference frames considered .
we calculate the real component of the electromagnetic 4-force , in the proper reference frame of the source particle , starting from the coulomb force generated by a charged particle at rest .
we then obtain the imaginary component of the 4-force , in the same reference frame , from the requirement that the 4-force be orthogonal to the 4-velocity .
the 4-force is then projected on a real 3 dimensional space to get the lorentz force . |
in the last two decades , extensive theoretical and experimental works have focused on the study of liquid helium at negative pressures and ultra - low temperatures .
@xcite the equation of state of this material has been measured very accurately within the density range in which is stable , and extrapolated to the region of negative pressures .
@xcite on the theoretical side , quantum monte carlo methods have allowed for precise and explicit simulation of metastable liquid helium , producing results which are in remarkable good agreement with experiments .
@xcite a quantity of central interest in all these studies is the spinodal pressure @xmath3 , that is , the pressure at which the bulk modulus , @xmath5 , and sound velocity vanish and thus the liquid becomes unstable against long wavelength density fluctuations . in liquid
@xmath0he @xmath3 amounts to @xmath2 bar , while in liquid @xmath6he to @xmath7 bar .
@xcite analogous studies performed in solid helium are almost non - existent in the literature .
@xcite at @xmath8 k solid helium becomes stable at pressures larger than @xmath9 bar , thus certainly the realm of negative pressures lies well below such a threshold .
nevertheless , analysis of the mechanical instability limit in helium crystals turns out to be a topic of fundamental interest .
in contrast to liquids , solids can sustain shear stresses and because of this ordinary fact the definition of the spinodal density in crystals differs greatly from the one given above . in the particular case of solid @xmath0he ,
previous attempts to determine @xmath3 have been , to the best of our knowledge , only tentative .
@xcite the energy of a crystal under a homogeneous elastic deformation is @xmath10 where @xmath11 are the elastic constants and @xmath12 a general strain deformation ( both expressed in voigt notation ) . the specific symmetry of a crystal determines the number of independent elastic constants which are different from zero . in the case of hexagonal crystals ( e.g. , hcp @xmath0he )
these are five : @xmath13 , @xmath14 , @xmath15 , @xmath16 and @xmath17 .
the conditions for mechanical stability in a crystal follow from the requirement that upon a general strain deformation the change in the total energy must be positive .
it can be shown that in hcp crystals subjected to an external pressure @xmath18 , these conditions readily are @xcite @xmath19~ \nonumber \\ c_{11 } - c_{12 } - 2p & > & 0~[{\rm c2}]~ \nonumber \\ \left ( c_{33 } - p \right ) \left ( c_{11 } + c_{12 } \right ) - 2 \left ( c_{13 } + p \right)^{2 } & > & 0~[{\rm c3}]~. \label{eq : conditions}\end{aligned}\ ] ] the spinodal pressure , or limit of mechanical stability , in a crystal then is identified with the point at which any of the three conditions above is not fulfilled .
we must note that the bulk modulus of an hexagonal crystal can be expressed as @xcite @xmath20 thus the conditions at which the bulk modulus ( and sound velocities ) vanishes in general must not coincide with the corresponding limit of mechanical stability .
this reasoning is in fact very different from the usual @xmath3 analysis performed in liquids . in this article
we present a computational study of the energetic and elastic properties in solid @xmath0he at pressures below @xmath9 bar , based on the diffusion monte carlo method .
our first - principles calculations allow us to determine with precision the limit of mechanical stability in this crystal , which we estimate to be @xmath1 bar ( that is , much larger in absolute value than the one found in the fluid ) .
moreover , we show that , in contrast to liquid helium , the pressure variation of the sound velocities near @xmath21 do not follow a power law of the form @xmath4 with @xmath22.@xcite the organization of this article is as follows . in the next section , we provide the details of our computational method and calculations . in sec .
[ sec : results ] , we present our results and comment on them . finally , we summarize our main findings in sec . [
sec : conclusions ] .
in dmc the time - dependent schrdinger equation of a @xmath23-particle system is solved stochastically by simulating the time evolution of the green s function propagator @xmath24 in imaginary time @xmath25 . in the @xmath26 limit ,
sets of configurations ( walkers ) @xmath27 render the probability distribution function @xmath28 , where @xmath29 is the true ground - state wave function and @xmath30 a guiding wave function used for importance sampling . within dmc , virtually exact results ( i.e. , subjected to statistical uncertainties only )
can be obtained for the ground - state energy and related quantities .
@xcite we are interested in studying the ground - state of hcp @xmath31he , which we assume to be governed by the hamiltonian @xmath32 where @xmath33 is the mass of a @xmath0he atom and @xmath34 the semi - empirical pairwise potential due to aziz .
@xcite it is worth noticing that the aziz potential provides an excellent description of the he - he interactions at low pressure .
@xcite the guiding wave function that we use in this study , @xmath35 , reproduces both the crystal ordering and bose - einstein symmetry .
this model wave function was introduced in ref . [ ] and reads @xmath36 where the index in the second product runs over the lattice position vectors . in previous works
we have demonstrated that @xmath35 provides an excellent description of the ground - state properties of bulk hcp @xmath0he @xcite and quantum solid films .
@xcite here , we adopt the correlation functions in eq .
( [ snjtrial ] ) of the mcmillan , @xmath37 $ ] , and gaussian , @xmath38 $ ] , forms , and optimize the corresponding parameters with variational monte carlo ( vmc ) [ in order to simulate the metastable solid we performed one unique vmc optimization at @xmath39 bar ] .
the technical parameters in our calculations were set in order to ensure convergence of the total energy per particle to less than @xmath40 k / atom .
the value of the mean population of walkers was @xmath41 and the length of the imaginary time - step ( @xmath42 ) @xmath43 k@xmath44 .
we used large simulation boxes containing @xmath45 atoms in all the cases .
statistics were accumulated over @xmath46 dmc steps performed after system equilibration , and the approximation used for the short - time green s function @xmath47 is accurate to second order in @xmath48 .
@xcite the computational strategy that we followed for calculation of the elastic constants @xmath11 is the same than explained in refs .
[ ] , thus we address the interested reader to them .
) are shown for comparison .
( _ _ b__ottom : ) calculated bulk modulus expressed as a function of volume .
the region in which @xmath5 vanishes is augmentated on the inset . ] in fig .
[ eos ] , we show the calculated energy in solid helium expressed as a function of volume ( solid symbols ) .
we fitted our results to the polynomial curve @xmath49^{2 } + c \left [ \left ( \frac{v_{0}}{v } \right ) - 1 \right]^{3}~ , \label{eq : dmcenergy}\ ] ] and found as best parameters @xmath50 k , @xmath51 @xmath6 , @xmath52 k , and @xmath53 k. in comparison to the analogous energy curve obtained in the liquid ( see fig . [ eos ] ) , @xcite @xmath54 displays larger slopes , which translates into larger negative pressures , at volumes close to equilibrium ( i.e. , @xmath55 ) . in the same figure , we enclose the bulk modulus , @xmath56 , that is calculated directly from @xmath54 ( solid line ) .
the symbols which appear therein correspond to estimations obtained with eq.([bulkmodulus ] ) , thereby excellent consistency between our energy and @xmath57 calculations is demonstrated .
we find that the bulk modulus of helium vanishes at volume @xmath58 @xmath6 and pressure @xmath59 bar .
in what follows , we present our analysis of the elastic properties and mechanical stability in solid @xmath0he and clarify whether @xmath60 and @xmath61 are meaningful physical quantities .
[ cij ] shows how the calculated elastic constants in solid helium change as a function of volume . as it is appreciated therein , all five @xmath57 decrease with increasing volume .
in particular , @xmath14 and @xmath15 become zero at @xmath62 ( @xmath63 bar ) and @xmath64 @xmath6 ( @xmath65 bar ) .
we performed power - law fits to our @xmath66 results in order to render continuous elastic constant functions . by doing this
, we were able to represent the three conditions of mechanical stability in hcp helium [ see eq.([eq : conditions ] ) ] within the volume interval of interest .
these results are enclosed in fig .
[ mechcond ] and show that condition @xmath67 , i.e. , [ c3 ] in eq.([eq : conditions ] ) , is violated at @xmath68 @xmath6 and @xmath1 bar .
thus , we identify @xmath69 with the limiting state at which solid @xmath0he is mechanically stable .
we note that these conditions , although similar , are not equal to @xmath70 at which the bulk modulus vanishes [ as it was already expected from eqs.([eq : conditions])-([bulkmodulus ] ) ] . ) ] expressed as a function of volume .
the region in which condition 3 is not accomplished is augmentated on the inset . ]
sound velocities in solids can be either longitudinal or tranverse and depend on the direction of propagation . in crystals with hexagonal symmetry two main propagation modes
are identified , one along the @xmath71-axis and the other contained in the basal plane .
the relationships between the elastic constants and sound velocities in hcp crystals are @xcite @xmath72 along the @xmath71-axis , and @xmath73 in the basal plane . in fig .
[ soundvel ] , we plot the six sound velocities calculated in solid helium as a function of pressure ( some of them are coincident ) together with an appropriate average of them , @xmath74 .
@xcite we observe that none of the sound velocities vanishes at the spinodal pressure @xmath3 ( marked in red in the figure ) , as it was already expected from eqs.([eq : conditions])-([eqn : cvel])-([eqn : basalvel ] ) . neither any sound velocity becomes zero at @xmath61 , the pressure at which the bulk modulus vanishes .
these results differ greatly from the usual conclusions drawn on the mechanical stability in liquids .
recently , it has been suggested that the variation of the sound velocities near the spinodal density in solid @xmath0he could follow a power law of the form @xmath75 where @xmath22 , in analogy to what is observed in the liquid .
@xcite however , in light of the results shown in fig .
[ soundvel ] such an hypotheses must be rejected since none of the sound velocities becomes zero at @xmath3 .
furthermore , we performed fits of the form @xmath76 to our results and found that parameter @xmath77 was always different from @xmath78 .
for instance , in the @xmath74 case we found @xmath79 as the optimal value .
therefore , we must conclude that the propagation of sound waves in helium crystals at the verge of mechanical instability differs radically from that predicted in liquid helium under similar conditions .
-axis and in the basal plane [ see eqs.([eqn : cvel])-([eqn : basalvel ] ) ] expressed as a function of pressure .
the region in which the crystal becomes mechanically unstable is filled with red color . ]
we have presented a rigorous and precise computational study of the energy and elastic properties of solid helium at conditions in which is metastable .
we have determined the limit of mechanical stability in this crystal , i.e. , @xmath1 bar , and the variation of the sound velocities at pressures close to it .
overall , we demonstrate that solid and liquid helium behave radically different in the vicinity of their spinodal state limits .
30 j. a. nissen , e. bodegom , l. c. brodie , j. s. semura , phys .
b * 40 * , 6617 ( 1989 ) .
h. j. maris , s. balibar , m. s. pettersen , j. low temp .
phys . * 93 * , 1069 ( 1993 ) .
h. j. maris , j. low temp .
* 98 * , 403 ( 1995 ) .
h. j. maris , d. o. edwards , j. low temp .
phys . * 129 * , 1 ( 2002 ) .
j. boronat , j. casulleras , and j. navarro , phys . rev .
b * 50 * , 3427 ( 1994 )
. h. j. maris , j. low temp .
phys . * 155 * , 290 ( 2009 ) . h. j. maris , j. low temp
. phys . * 158 * , 485 ( 2010 ) .
g. v. sinko and n. a. smirnov , j. phys . : condens .
matter * 14 * , 6989 ( 2002 ) .
g. grimvall _ et al .
mod . phys . * 84 * , 945 ( 2012 ) . c. cazorla , y. lutsyshyn , and j. boronat , phys .
b * 85 * , 024101 ( 2012 ) .
h. j. maris , phys . rev .
* 66 * , 45 ( 1991 ) .
r. barnett , p. reynolds , and w. a. lester jr . , j. comput . phys . * 96 * , 258 ( 1991 ) .
j. casulleras and j. boronat , phys .
b * 52 * , 3654 ( 1995 ) . r. a. aziz , f. r. w. mccourt , and c. c. k. wong , mol . phys . * 61 * , 1487 ( 1987 ) . j. boronat and j. casulleras , phys . rev .
b * 49 * , 8920 ( 1994 ) . c. cazorla and j. boronat , j. phys . : condens
. matter * 20 * , 015223 ( 2008 ) . c. cazorla , g. astrakharchick , j. casulleras , and j. boronat , new journal of phys . * 11 * , 013047 ( 2009 ) . c. cazorla and j. boronat , phys .
b * 77 * , 024310 ( 2008 ) . c. cazorla , g. astrakharchick , j. casulleras , and j. boronat , j. phys . : condens .
matter * 22 * , 165402 ( 2010 ) .
m. c. gordillo , c. cazorla , and j. boronat , phys .
b * 83 * , 121406(r ) ( 2011 ) .
j. boronat and j. casulleras , phys .
b * 49 * , 8920 ( 1994 ) . s. a. chin , phys
a * 42 * , 6991 ( 1990 ) . c. cazorla , y. lutsyshyn , and j. boronat , phys .
b * 87 * , 214522 ( 2013 ) .
r. rota , y. lutsyshyn , c. cazorla , and j. boronat , j. low temp .
phys . * 168 * , 150 ( 2012 ) . | we present a first - principles study of the energy and elastic properties of solid helium at pressures below the range in which is energetically stable .
we find that the limit of mechanical stability in hcp @xmath0he is @xmath1 bar , which lies significantly below the spinodal pressure found in the liquid phase ( i.e. , @xmath2 bar ) .
furthermore , we show that the pressure variation of the transverse and longitudinal sound velocities close to @xmath3 do not follow a power law of the form @xmath4 , in contrast to what is observed on the fluid . |
in species sampling problems , one is interested in the species composition of a certain population ( of plants , animals , genes , etc . ) containing an unknown number of species and only a sample drawn from it is available .
the relevance of such problems in ecology , biology and , more recently , in genomics and bioinformatics is not surprising . from an inferential perspective , one is willing to use available data in order to evaluate some quantities of practical interest .
the available data specifically consist of a so - called _ basic sample _ of size @xmath0 , @xmath1 , which exhibits @xmath2 distinct species , @xmath3 , with respective frequencies @xmath4 , where clearly @xmath5 .
given a basic sample , interest mainly lies in estimating the number of new species , @xmath6 , to be observed in an additional sample @xmath7 of size @xmath8 and not included among the @xmath9 s , @xmath10 .
most of the contributions in the literature that address this issue rely on a frequentist approach ( see @xcite for reviews ) and only recently an alternative bayesian nonparametric approach has been set forth ( see , e.g. , @xcite ) .
the latter resorts to a general class of discrete random probability measures , termed _
species sampling models _ and introduced by j. pitman in @xcite . given a nonatomic probability measure @xmath11 on some complete and separable metric space @xmath12 , endowed with the borel @xmath13-field @xmath14
, a ( proper ) species sampling model on @xmath15 is a random probability measure @xmath16 where @xmath17 is a sequence of independent and identically distributed ( i.i.d . )
random elements taking values in @xmath12 and with probability distribution @xmath11 , the nonnegative random weights @xmath18 are independent from @xmath17 and are such that @xmath19 , almost surely . in the species sampling context , the @xmath20 s act as species tags and @xmath21 is the random proportion with which the @xmath22th species is present in the population . if @xmath23 is an exchangeable sequence directed by a species sampling model @xmath24 , that is , for every @xmath25 and @xmath26 in @xmath14 one has @xmath27=\prod_{i=1}^n \tilde p(a_i)\ ] ] almost surely , then @xmath23 is termed _ species sampling sequence_. besides being an effective tool for statistical inference , species sampling models have an appealing structural property established in @xcite .
indeed , if @xmath23 is a species sampling sequence , then there exists a collection of nonnegative weights @xmath28 such that @xmath29 , for any vector of positive integers @xmath30 with @xmath31 , and @xmath32=p_{k_n+1,n}(n_1,\ldots , n_{k_n})p_0 ( \cdot)+\sum_{j=1}^{k_n } p_{j , n}(n_1,\ldots , n_{k_n } ) \delta_{x^*_j } ( \cdot),\ ] ] where @xmath33 is a sample with @xmath34 distinct values @xmath35 .
statistical applications involving species sampling models for different purposes than those of the present paper are provided , for example , in @xcite .
the bayesian nonparametric approach we undertake postulates that the data are exchangeable and generated by a species sampling model .
then , conditionally on the basic sample of size @xmath0 , inference is to be made on the number @xmath36 of new distinct species that will be observed in the additional sample of size @xmath8 .
interest lies in providing both a point estimate and a measure of uncertainty , in the form of a credible interval , for @xmath36 given @xmath1 .
since the conditional distribution of @xmath36 becomes intractable for large sizes @xmath8 of the additional sample , one is led to studying its limiting behaviour as @xmath8 increases .
such asymptotic results , in addition to providing useful approximations to the required estimators , are also of independent theoretical interest since they provide useful insight on the behaviour of the models we focus on .
the only discrete random probability measure for which a conditional asymptotic result , similar to the one investigated in this paper , is known , is the two - parameter poisson
dirichlet process , shortly denoted as @xmath37 . according to @xcite ,
a @xmath37 process is a species sampling model characterized by=-1 @xmath38=0 with @xmath10 , @xmath39 and @xmath40 . in this case , @xcite provide a result describing the conditional limiting behaviour of @xmath36 . in the present paper ,
we focus on an alternative species sampling model , termed _ normalized generalized gamma process _ in @xcite .
as we shall see in the next section , it depends on two parameters @xmath39 and @xmath41 and , for the sake of brevity , is denoted by @xmath42 .
moreover , it is characterized by @xmath43 for any @xmath44 , where @xmath45 is the incomplete gamma function .
the @xmath42 process prior has gained some attention in the bayesian literature and it has proved to be useful for various applications such as those considered , for example , in @xcite . it is to be noted that the @xmath46 does not feature a posterior structure that is as tractable as the one associated to the @xmath47 process ( see , e.g. , @xcite ) . nonetheless , in terms of practical implementation , it is possible to devise efficient simulation algorithms that allow for a full bayesian analysis within models based on a @xmath42 prior .
see @xcite for a review of such algorithms . in the present manuscript
, we will specify the asymptotic behaviour of @xmath36 , given the basic sample , as @xmath8 diverges and highlight the interplay between the conditional distributions of the @xmath37 and the @xmath42 processes . since the posterior characterization of a @xmath42 process is far more involved than the one associated to the @xmath47 process ,
the derivation of the conditional asymptotic results considered in this paper is technically more challenging .
this is quite interesting since it suggests that it is possible to study the limiting conditional behaviour of @xmath36 even beyond species sampling models sharing some sort of conjugacy property .
for example , one might conjecture that the same asymptotic regime , up to certain transformations of the limiting random variable , should hold also for the wide class of gibbs - type priors , to be recalled in section [ section2 ] .
an up to date account of bayesian nonparametrics can be found in the monograph @xcite and , in particular for asymptotic studies , @xcite provides a review of asymptotics of nonparametric models in terms of `` frequentist consistency . ''
yet another type of asymptotic results are obtained in @xcite .
the outline of the paper is as follows . in section [ section2 ]
, one can find a basic introduction to species sampling models and a recollection of some results in the literature concerning the asymptotic behaviour of the number @xmath48 of distinct species in the basic sample , as @xmath0 increases .
section [ section3 ] displays the main results , whereas the last section contains some concluding remarks .
let us start by providing a succinct description of completely random measures ( crm ) before defining the specific models we will consider and which can be derived as suitable transformations of crms .
see @xcite for an overview of discrete nonparametric models defined in terms of crms .
suppose @xmath49 is a random element defined on some probability space @xmath50 and taking values on the space @xmath51 of boundedly finite measures on @xmath15 such that for any @xmath52 in @xmath14 , with @xmath53 for @xmath54 , the random variables @xmath55 are mutually independent
. then @xmath49 is termed _ completely random measure _ ( crm ) .
it is well - known that the laplace functional transform of @xmath49 has a simple representation of the type @xmath56=\mathrm { e}^{-\psi(f)},\ ] ] where @xmath57 \nu(\mathrm{d}s , \mathrm{d } y)$ ] for any measurable function @xmath58 such that @xmath59 almost surely and the measure @xmath60 on @xmath61 is known as the lvy intensity of @xmath49 .
see , for example , @xcite .
since a crm is almost surely discrete , any crm can be represented as @xmath62 with independent random jump locations @xmath17 and heights @xmath63 . for our purposes , it is enough to focus on the special case of @xmath60 factorizing as @xmath64 , which implies independence of the locations @xmath20 s and jumps @xmath65 s in the above series representation .
furthermore , @xmath66 can be taken to be nonatomic and finite , the latter ensuring almost sure finiteness of the corresponding crm .
now , if @xmath67 for any @xmath68 , one can define a random probability measure on @xmath12 as @xmath69 this family of random probability measures is known from @xcite as homogeneous normalized random measure with independent increments , a subclass of the general class of normalized processes introduced in @xcite .
note that an @xmath12-valued exchangeable sequence @xmath23 generated by @xmath24 as in ( [ eqnrmi ] ) is a species sampling sequence . here
we focus on a specific example where the crm defining @xmath24 in ( [ eqnrmi ] ) is the so - called _ generalized gamma process _ @xcite that is characterized by @xmath70 with @xmath39 and @xmath71 . in this case , @xmath72 \alpha ( \mathrm{d}x ) \label{eqlaplexpon}\ ] ] for any measurable function @xmath73 such that @xmath74 . in the sequel
the model will be reparameterized , without loss of generality ( see , e.g. , @xcite ) , by setting @xmath75 and @xmath66 as a probability measure .
the corresponding crm will be denoted by @xmath76 . henceforth , the random probability measure @xmath24 obtained by normalizing @xmath76 as in ( [ eqnrmi ] ) coincides , in distribution , with the @xmath77 process prior .
an important special case arises when @xmath78 , since @xmath79 reduces to the @xmath13-stable process , which plays a key role within the paper .
for example , it is worth noting that @xmath80 can also be defined as an exponential tilting of @xmath81 , for any @xmath82 .
specifically , if @xmath83 is the probability distribution of @xmath81 on @xmath84 and @xmath85 is a probability measure on @xmath86 that is absolutely continuous with respect to @xmath83 and such that @xmath87 then @xmath85 coincides with the probability distribution of @xmath76 . in a similar fashion
, one can also define the @xmath37 process as a polynomial tilting of @xmath81 , for any @xmath40 .
indeed , one introduces another probability measure @xmath88 that is still absolutely continuous with respect to @xmath89 and whose radon nykodim derivative is @xmath90^{-\theta}\ ] ] for any @xmath39 and @xmath40 . if @xmath91 is the random measure with probability distribution @xmath92 above , then @xmath93 coincides , in distribution , with a @xmath37 process .
see @xcite . the different tilting structure featured by the normalized generalized gamma process and the two parameter poisson dirichlet process
will be reflected by the limiting results to be illustrated in the paper .
it is also worth to recall that both the @xmath42 and the @xmath37 processes can be seen as elements of the general class of _ gibbs - type _ nonparametric priors introduced in @xcite .
gibbs - type priors represent the most tractable subclass of species sampling models .
they are characterized by a parameter @xmath94 and a collection of non - negative quantities @xmath95 that satisfy the forward recursive relations @xmath96 these @xmath97 s define the predictive weights that characterize a species sampling sequence governed by a gibbs - type prior .
indeed , one has @xmath98 for each @xmath44 .
the fundamental simplification involved in ( [ eqweightsgibbs ] ) is that the probability of observing a `` new '' or an `` old '' species depend on the sample size and on the number of already observed distinct species but not on their frequencies : this crucially simplifies explicit calculations .
turning to the two specific processes introduced before , in accordance with ( [ eqpesi0 ] ) , the @xmath37 process identifies a gibbs - type prior with @xmath99 whereas , in accordance with ( [ eqpesi ] ) and ( [ eqpesi1 ] ) , a @xmath42 prior is also of gibbs - type with @xmath100 and @xmath101 as shown in @xcite , a normalized crm is a gibbs - type prior ( with @xmath39 ) if and only if it is a @xmath42 process .
this result also motivates the focus of the paper on the @xmath77 process , which clearly has a prominent role .
the result on the limiting behaviour of @xmath36 to be determined in the next section parallels known results for the unconditional case where one aims at determining the asymptotics of @xmath34 as the sample size @xmath0 increases and connects to the conditional asymptotics displayed in @xcite for the @xmath37 process . in order to describe the result for the unconditional case , let @xmath102 be the random total mass of a @xmath13-stable crm and denote by @xmath103 its density function which satisfies @xmath104 for any @xmath105 .
moreover , let @xmath106 be the random total mass of @xmath107 process and recall that its law can be obtained by exponentially tilting the probability distribution of @xmath108 as in ( [ eqexpontilt ] ) .
in particular , if @xmath109 then its density function , with respect to the lebesgue measure on @xmath110 , coincides with @xmath111 and one has that @xmath112 according to the terminology introduced by @xcite , the random variable @xmath113 is the so - called @xmath13-_diversity _ of the exchangeable random partition induced by a @xmath42 process prior .
see also definition 3.10 in pitman @xcite .
note that a similar result holds true for the @xmath37 process .
indeed , if @xmath114 so that its probability distribution is obtained by polynomially tilting the probability distribution of @xmath108 as in ( [ eqpolytilt ] ) and @xmath115 admits density function @xmath116 then one has @xmath117 see @xcite , theorem 3.8 .
these results are somehow in line with the fact that the @xmath42 and the @xmath118 processes are distributionally equivalent to normalized random measures that are obtained by an exponential and a polynomial tilting , respectively , of a @xmath13-stable crm as highlighted in ( [ eqexpontilt ] ) and in ( [ eqpolytilt ] ) .
finally , note that a combination of @xcite , theorem 12 , and @xcite , proposition 13 , shows that the unconditional asymptotic results in ( [ eqdiversity ] ) and ( [ eqdiversitypd ] ) can be extended to the whole class of gibbs - type priors .
see also @xcite for another contribution at the interface between bayesian nonparametrics and gibbs - type random partitions .
as mentioned before , inference on @xmath36 is of great importance since it provides a measure of species richness of a community of plants / animals or of a cdna library for gene discovery . the key quantity for obtaining posterior inferences
is given by the probability distribution @xmath119 $ ] for @xmath120 . by virtue of predictive sufficiency of the number @xmath34 of distinct species observed among the first @xmath0 data @xmath33 , in @xcite it has been shown that in the @xmath77 this distribution coincides with @xmath121 \nonumber \\[-8pt ] \\[-8pt ] \nonumber & = & \frac{\mathscr{g}(m , k;\sigma ,-
n+j\sigma)}{(n)_m } \frac{\sum_{l=0}^{n+m-1}{n+m-1\choose l}(-1)^{l}\beta^{l/\sigma } \gamma ( j+k-{l}/{\sigma};\beta)}{\sum_{l=0}^{n-1}{n-1\choose l}(-1)^{l}\beta^{l/\sigma}\gamma(j-{l}/{\sigma};\beta)}\end{aligned}\ ] ] for @xmath122 , with @xmath123 denoting the non central generalized factorial coefficient .
see @xcite for a comprehensive account on generalized factorial coefficients .
expression ( [ posteriordistinct ] ) can be interpreted as the `` posterior '' probability distribution of the number of distinct new species to be observed in a further sample of size @xmath8 .
now , based on ( [ posteriordistinct ] ) , one obtains the expected number of new species as @xmath124=\sum _ { k=0}^{m}kp_{m}^{(n , j)}(k),\ ] ] which corresponds to the bayes estimator of @xmath125 under quadratic loss .
moreover , a measure of uncertainty of the point estimate @xmath126 can be obtained in terms of @xmath127-credible intervals that is , by determining an interval @xmath128 with @xmath129 such that @xmath130 \geq\alpha$ ] .
the interval @xmath128 of shortest length is then typically referred to as highest posterior density interval .
the main advantage of the distribution ( [ posteriordistinct ] ) is that it is explicit .
however , since the sum of incomplete gamma functions can not be further simplified , its computation can become overwhelming even for moderately large sizes of @xmath0 and @xmath8 .
this fact represents a major problem in the frequent practical situations in which the size of the additional sample of interest is large . for instance , in genomic applications one has to deal with relevant portions of cdna libraries which typically consist of millions of genes .
hence , it is natural to study the asymptotics for @xmath125 , given @xmath48 , as @xmath131 , in order to obtain approximations of ( [ posteriordistinct ] ) and , consequently , also of ( [ posteriormean ] ) and of the corresponding highest posterior density intervals .
indeed , if one is able to show that a suitable rescaling of @xmath125 , given @xmath48 , converges in law to some random variable , one can use the probability distribution of this limiting random quantity in order to derive the desired approximations .
the statement of the main result in the paper involves a positive random variable @xmath132 whose density function is , for any @xmath133 , @xmath134 and we @xmath135 to denote a beta random variable with parameters @xmath136 .
moreover , set @xmath137 , with @xmath138 and @xmath139 independent , and denote by @xmath140 the density function of @xmath141 .
[ limitresult ] if @xmath23 is a species sampling sequence directed by a @xmath42 process prior , conditional on @xmath142 one has @xmath143 as @xmath131 , where @xmath144 is a positive random variable obtained by exponentially tilting the density function of @xmath141 , namely @xmath145 the first part of the proof exploits a martingale convergence theorem along the same lines of @xcite , theorem 3.8 . in particular , let us start by computing the likelihood ratio @xmath146 where @xmath147 , @xmath148 is the conditional probability distribution of a normalized generalized gamma process with parameter @xmath149 given @xmath48 and , by virtue of @xcite , proposition 1 , @xmath150 for any integer @xmath151 and @xmath152 hence , @xmath153 is a @xmath154-martingale and by a martingale convergence theorem , @xmath155 has a @xmath156 almost sure limit , say @xmath157 , as @xmath131 . clearly , we have that , where @xmath158 denotes the expected value with respect to @xmath159 .
let now @xmath160 be a sequence of i.i.d .
random variables having a negative exponential distribution with parameter @xmath161 . moreover ,
suppose the @xmath162 s are independent of @xmath163 .
set @xmath164 and note that , conditionally on @xmath163 , @xmath165 has gamma distribution with expected value @xmath166 .
we can then rewrite @xmath155 as follows @xmath167.\end{aligned}\ ] ] from the strong law of large numbers , @xmath168 as @xmath169 and conditionally on @xmath163 . using the dominated convergence theorem , we have @xmath170 & \approx&\frac{\gamma(k_{n})}{\sum_{l=0}^{n-1}{n-1\choose l}(-1)^{l}\beta^{l/\sigma}\gamma(k_{n}-{l}/{\sigma};\beta ) } \biggl(1-\frac{\beta^{1/\sigma}}{(k_{n}+k_{m}^{(n)})^{1/\sigma } } \biggr)^{n+m-1}\\[-2pt ] & \approx&\frac{\gamma(k_{n})}{\sum_{l=0}^{n-1}{n-1\choose l}(-1)^{l}\beta^{l/\sigma}\gamma(k_{n}-{l}/{\sigma};\beta ) } \exp\biggl\{-m\frac{\beta^{1/\sigma}}{(k_{m}^{(n)})^{1/\sigma}}\biggr\}\end{aligned}\ ] ] as @xmath131 .
since @xmath171 almost surely ( with respect to @xmath156 ) , then there exists some positive random variable , say @xmath172 such that @xmath173 almost surely ( with respect to @xmath154 ) . in order to identify the probability distribution of @xmath172 , note that it must be such that @xmath174 = \frac{1}{\gamma(k_{n})}\int_{\beta}^{+\infty}y^{k_{n}-1}\biggl(1-\frac { \beta^{1/\sigma}}{y^{1/\sigma } } \biggr)^{n-1}\mathrm{e}^{-y } \,\mathrm{d}y.\ ] ] since @xmath175 , we have to prove that @xmath176 , that is , that the density function of @xmath172 coincides with @xmath177 \\[-9pt ] \nonumber & & { } \times\int_{z^{-\sigma}}^{+\infty}\frac{1}{v}v^{{n}/{\sigma } -1-{1}/{\sigma}}f_{\sigma}(v^{-{1}/{\sigma } } ) \biggl(\frac{z^{-\sigma}}{v}\biggr)^{k_{n}-1}\biggl(1-\frac{z^{-\sigma } } { v}\biggr)^{{n}/{\sigma}-k_{n}-1 } \,\mathrm{d}v.\qquad\end{aligned}\ ] ] so we simply have to show that the laplace transform of the density function in ( [ density1 ] ) is given by ( [ condition ] ) . by a simple change of variable , @xmath178
, the previous density reduces to @xmath179 and by the change of variable @xmath180 @xmath181 the laplace transform of @xmath182 is then given by @xmath183 & = & \frac{\sigma\gamma(n)}{\gamma(k_{n})\gamma ( { n}/{\sigma } -k_{n})}\\[-2pt ] & & { } \times \int_{0}^{\infty}\mathrm{e}^{\beta^{1/\sigma}z}z^{-n}\\[-2pt ] & & \hspace*{28pt } { } \times\int_{0}^{1}y^{-n/\sigma + 1/\sigma+k_{n}-1}(1-y)^{n/\sigma - k_{n}-1}f_{\sigma}(zy^{1/\sigma } ) \,\mathrm{d}y \,\mathrm{d}z \\[-2pt ] & = & \frac{\sigma\gamma(n)}{\gamma(k_{n})\gamma ( { n}/{\sigma } -k_{n})}\int_{0}^{1}y^{-n/\sigma+1/\sigma+k_{n}-1}(1-y)^{n/\sigma -k_{n}-1}\\[-2pt ] & & \hspace*{99pt } { } \times\int_{0}^{\infty } \mathrm{e}^{\beta^{1/\sigma}z}z^{-n}f_{\sigma}(zy^{1/\sigma } ) \,\mathrm{d}z\,\mathrm{d}y\\[-2pt ] & = & \frac{\sigma\gamma(n)}{\gamma(k_{n})\gamma ( { n}/{\sigma } -k_{n } ) } \int_{0}^{1}y^{k_{n}-1}(1-y)^{n/\sigma - k_{n}-1}\\[-2pt ] & & \hspace*{99pt } { } \times \int_{0}^{\infty}\mathrm{e}^{(\beta / y)^{1/\sigma}h}h^{-n}f_{\sigma } ( h)\,\mathrm{d}h \,\mathrm{d}y\\[-2pt ] & = & \frac{\gamma({n}/{\sigma})}{\gamma(k_{n})\gamma ( { n}/{\sigma}-k_{n})}\int _ { 0}^{1}y^{k_{n}-1}(1-y)^{n/\sigma - k_{n}-1}\\[-2pt ] & & \hspace*{99pt}{}\times \frac{\sigma\gamma(n)}{\gamma({n}/{\sigma } ) } \int_{0}^{\infty}\mathrm{e}^{(\beta / y)^{1/\sigma } h}h^{-n}f_{\sigma } ( h)\,\mathrm{d}h \,\mathrm{d}y.\end{aligned}\ ] ] according to the well - known gamma identity , we can write @xmath184 obtaining @xmath185 & & \qquad{}\times \frac{\sigma}{\gamma({n}/{\sigma } ) } \int_{0}^{+\infty}u^{n-1}\int_{0}^{+\infty}\mathrm{e}^{-h ( { \beta ^{1/\sigma}}/{y^{1/\sigma}}+u ) } f_{\sigma}(h)\,\mathrm{d}h \,\mathrm { d}u\\[-2pt ] & & \quad=\frac{\gamma({n}/{\sigma})}{\gamma(k_{n})\gamma ( { n}/{\sigma}-k_{n})}\int_{0}^{1}y^{k_{n}-1}(1-y)^{n/\sigma -k_{n}-1}\\[-2pt ] & & \qquad\times \frac{\sigma}{\gamma({n}/{\sigma})}\int_{0}^{+\infty } u^{n-1 } \mathrm{e}^{-({\beta^{1/\sigma}}/{y^{1/\sigma } } + u)^{\sigma } } \,\mathrm{d}u\\[-2pt ] & & \quad= \frac{\gamma({n}/{\sigma})}{\gamma(k_{n})\gamma ( { n}/{\sigma}-k_{n})}\int_{0}^{1}y^{k_{n}-1}(1-y)^{n/\sigma -k_{n}-1}\\[-2pt ] & & \qquad { } \times\frac{1}{\gamma ( { n}/{\sigma})}\int_{\beta}^{+\infty}z^{n/\sigma-1}\biggl(1- \biggl(\frac{\beta}{zy}\biggr)^{1/\sigma}\biggr)^{n-1}\mathrm{e}^{-z } \,\mathrm{d}z \,\mathrm{d}y\\[-2pt ] & & \quad=\frac{1}{\gamma(k_{n})\gamma({n}/{\sigma}-k_{n})}\\[-2pt ] & & \qquad{}\times\int _ { \beta}^{+\infty}\mathrm{e}^{-z } \\[-2pt ] & & \qquad{}\hspace*{36pt}\times\int _ { 0}^{z}w^{k_{n}-1}(z - w)^{n/\sigma - k_{n}-1}\biggl(1-\biggl(\frac{\beta } { w}\biggr)^{1/\sigma}\biggr)^{n-1 } \,\mathrm{d}w \,\mathrm{d}z\\[-2pt ] & & \quad=\frac{1}{\gamma(k_{n})\gamma({n}/{\sigma}-k_{n})}\int _
{ \beta}^{\infty}w^{k_{n}-1}\biggl(1-\biggl(\frac{\beta}{w } \biggr)^{1/\sigma}\biggr)^{n-1}\\[-2pt ] & & \hspace*{102pt}\qquad { } \times\int_{w}^{+\infty}\,\mathrm { e}^{-z}(z - w)^{n/\sigma - k_{n}-1 } \,\mathrm{d}z \,\mathrm{d}w \ ] ] which corresponds to ( [ condition ] ) .
finally , since the probability measures @xmath186 and @xmath187 are mutually absolutely continuous , almost sure convergence holds true with respect to @xmath186 , as well .
in order to deduce the @xmath188-law of @xmath189 , it is sufficient to exploit a change of measure suggested by @xmath190 and by the fact that @xmath191 this completes the proof .
it is worth stressing that the limit random variable in the conditional case is the same as in the unconditional case but with updated parameters and a rescaling induced by a beta - distributed random variable .
the density of @xmath144 in ( [ result ] ) can be formally represented as @xmath192 \\[-8pt ] \nonumber & & { } \times\frac{\gamma(n)}{\gamma(j)\gamma({n}/{\sigma}-j ) } z^{j-1}\int_{z}^{+\infty}v^{-1/\sigma}(v - z)^{n/\sigma -j-1}f_{\sigma } ( v^{-1/\sigma})\,\mathrm{d}v,\end{aligned}\ ] ] where we recall that @xmath103 is the density function of a positive stable random variable and , then , coincides with a density function of the random total mass of a @xmath13-stable crm @xmath193 .
theorem [ limitresult ] can be compared with an analogous result recently obtained in @xcite , proposition 2 , for the @xmath194 process , where it is shown that the number of new distinct species @xmath125 induced by the @xmath195 process is such that @xmath196 as @xmath131 , where @xmath197 and the random variables @xmath198 and @xmath199 are independent .
this can be paralleled with the unconditional limit since it is known that @xmath200 , almost surely , as @xmath201 . see , for example , @xcite , theorem 3.8 .
note that a normalized @xmath13-stable process coincides , in distribution , with both a @xmath202 and a @xmath203 process .
hence , it is no surprise that the two limits ( [ result ] ) and ( [ limittwoparameter ] ) are the same , in distribution , when @xmath204 .
another interesting case is represented by the normalized generalized gamma process with parameter @xmath205 which yield to the so - called normalized inverse - gaussian processes @xcite . in particular , for the @xmath206 process the density @xmath207 in ( [ density ] ) is known explicitly and the previous expression can be simplified to @xmath208 since the above described limiting distributions can not be easily handled for practical purposes , it is useful to devise a simulation algorithm . in this respect , one can adapt , similarly to @xcite , an exact sampling algorithm recently devised by @xcite for random variate generation from polynomially and exponentially tilted @xmath13-stable distributions .
this will allow to sample the limiting random variables @xmath209 and @xmath144 corresponding to the @xmath194 and to the @xmath42 case , respectively .
indeed , note that @xmath210 is a scale mixture involving a beta random variable @xmath211 and a positive random variable @xmath199 .
the latter is such that its transformation @xmath212 admits density function of the form @xmath213 which is precisely the density function of a polynomially tilted @xmath214-stable distribution .
therefore , random variate generation from @xmath210 can be easily done by independently sampling from a beta random variable with parameter @xmath215 and from a random variable with density function ( [ polynomialtilted ] ) by means of the algorithm devised in @xcite .
we refer to @xcite for an alternative sampling algorithm for @xmath209 via augmentation .
similar arguments can be applied in order to sample from the limit random variable @xmath144 .
indeed , observe that @xmath144 is characterized by a density function proportional to @xmath216 with @xmath140 being the density function of the random variable @xmath217 .
therefore , in order to sample from the distribution of @xmath144 one can apply a simple rejection sampling .
in particular , the sampling scheme would work as follows 1 .
generate @xmath218 .
2 . sample @xmath219 according to devroye s algorithm .
3 . set @xmath220 .
sample @xmath221 from a uniform on the interval @xmath222 .
1 . if @xmath223 set @xmath224 .
2 . if @xmath225 restart from ( 1 ) . in this final section ,
we provide a result that gives an interesting representation of the key random variable @xmath226 . to this end
, we need to provide a representation for the posterior laplace transform of the total mass of the @xmath13-stable crm @xmath81 or , equivalently , of the unnormalized @xmath202 or @xmath227 processes .
indeed one has [ poststable ] let @xmath228 be a species sampling sequence directed by a normalized @xmath13-stable process prior and suppose that the sample @xmath33 is such that @xmath229
. then @xmath230 = \frac{1}{\gamma(j)}\int_{\lambda^\sigma}^\infty y^{j-1}\biggl(1-\frac { \lambda^{1/\sigma}}{y^{1/\sigma}}\biggr)^{n-1 } \mathrm{e}^{-y } \,\mathrm{d}y\ ] ] for any @xmath105 .
set @xmath231 . since the joint distribution of @xmath232 , also known as exchangeable partition probability function ( see @xcite ) , of a normalized @xmath13-stable process coincides with @xmath233=\sigma^{j-1}\gamma(j ) \prod_{i=1}^k(1-\sigma)_{n_i-1}/\gamma ( n)$ ] , one has @xmath234&= & \frac{\gamma(n)}{\sigma^{j-1 } \gamma(j ) \prod_{i=1}^k(1-\sigma ) _ { n_i-1 } } \frac{1}{\gamma(n)}\\ & & { } \times\int_0^\infty u^{n-1}\mathrm{e}^{-(\lambda+u)^\sigma } \sigma^j \prod_{i=1}^k \frac{\gamma(n_i-\sigma)}{\gamma(1-\sigma ) } ( u+\lambda ) ^{-n_i+\sigma } \,\mathrm{d}u\end{aligned}\ ] ] and a simple change of variable @xmath235 yields the representation in ( [ eqpostlapl ] ) .
proposition [ poststable ] allows one to draw an interesting comparison between unconditional and conditional limits of the number of distinct species . as we have already highlighted in section [ section2 ] , the probability distribution of the @xmath13diversities for the @xmath42 process and the @xmath194 process arise as a power transformation ( involving the parameter @xmath13 ) of a suitable tilting of the probability distribution of @xmath236 .
we are now in the position to show that a similar structure carries over when one deals with the conditional case .
resorting to the notation set forth in theorem [ limitresult ] , let @xmath237 to be a random variable whose law coincides with the probability distribution of the conditional total mass @xmath108 of a @xmath214-stable process given a sample of size @xmath0 containing @xmath34 distinct species .
hence , from the laplace transform ( [ condition ] ) in the proof of theorem [ limitresult ] one can easily spot the following identity @xmath238 let now @xmath159 and @xmath239 be the conditional probability distributions of , respectively , the @xmath13-stable @xmath240 and the generalized gamma @xmath241 processes . according to theorem [ limitresult ] , the probability distributions @xmath159 and @xmath242 are mutually absolutely continuous giving rise to the conditional counterpart of the identity ( [ eqexpontilt ] ) , that is , @xmath243 for any @xmath39 and @xmath82 .
in particular , if we denote by @xmath244 the random variable whose probability distribution is obtained by exponentially tilting the probability distribution of @xmath237 as in ( [ tiltingesponenziale ] ) , then one can establish that @xmath245 in other terms , one can easily verify that the probability distribution of the limit random variable @xmath144 in ( [ result ] ) can be also derived by applying to the probability distribution of @xmath246 the same transformation characterizing the corresponding unconditional case . in a similar fashion , one can also derive the conditional counterpart of the identity ( [ eqpolytilt ] ) for the two parameter poisson dirichlet process . indeed
, according to @xcite , proposition 2 , one can introduce a probability measure @xmath247 on @xmath248 whose radon nikodm derivative with respect to the dominating measure @xmath159 is given by @xmath249^{-\theta}\vspace*{-1pt}\ ] ] for any @xmath39 and @xmath40 with @xmath250 being the probability measure of the random measure @xmath251 conditional on the sample .
in particular , if we denote by @xmath252 the random variable whose probability distribution is obtained by polynomially tilting the probability distribution of @xmath237 as in ( [ tiltingpolinomiale ] ) , then one can easily verify that @xmath253 this suggests that the probability distribution of the limiting random variable @xmath210 in ( [ limittwoparameter ] ) can also be derived by applying to the probability distribution of @xmath254 the same transformation characterizing the corresponding unconditional case .
the identities ( [ exptiltngg ] ) and ( [ exptilttwoparam ] ) represent the conditional counterparts of the identities ( [ prima ] ) and ( [ seconda ] ) , respectively , given a sample containing @xmath34 distinct species .
hence , in the same spirit of @xcite , proposition 13 , we have provided a characterization of the distribution of the limiting random variables @xmath144 and @xmath210 in terms of a power transformation ( involving the parameter @xmath13 ) applied to a suitable tilting for the conditional distribution of the total mass of the @xmath13-stable process . in particular , the identities ( [ exptiltngg ] ) and ( [ exptilttwoparam ] ) characterize the distribution of the limit random variables @xmath144 and @xmath210 via the same transformation characterizing the unconditional case and applied to an exponential tilting and polynomial tilting , respectively , for a scale mixture distribution involving the beta distribution and the @xmath13-stable distribution . to conclude , there is a connection between the prior , and posterior , total mass of a @xmath13-stable crm that we conjecture can be extended to any gibbs - type random probability measure and will be object of future research .
the authors are grateful to an associate editor and a referee for valuable remarks and suggestions that have lead to a substantial improvement in the presentation .
this work is partially supported by miur , grant 2008mk3afz , and regione piemonte . | in bayesian nonparametric inference , random discrete probability measures are commonly used as priors within hierarchical mixture models for density estimation and for inference on the clustering of the data .
recently , it has been shown that they can also be exploited in species sampling problems : indeed they are natural tools for modeling the random proportions of species within a population thus allowing for inference on various quantities of statistical interest . for applications that involve large samples ,
the exact evaluation of the corresponding estimators becomes impracticable and , therefore , asymptotic approximations are sought . in the present paper ,
we study the limiting behaviour of the number of new species to be observed from further sampling , conditional on observed data , assuming the observations are exchangeable and directed by a normalized generalized gamma process prior .
such an asymptotic study highlights a connection between the normalized generalized gamma process and the two - parameter poisson
dirichlet process that was previously known only in the unconditional case . |
advances in nonlinear functional analysis and the rich theory of linear distributed parameter systems have led to a growing of body of work on nonlinear infinite - dimensional models .
for instance , in a 2nd - order evolution framework ( especially , wave , elastodynamics , or thin plates with no rotational inertia terms ) for an appropriate elliptic operator @xmath6 a linear equation with viscous damping for an unknown @xmath7 may be expressed as @xmath8 with @xmath9 .
we will focus on the evolution on a bounded domain and under suitable homogeneous boundary conditions .
a nonlinear refinement on the dissipative term may take the form of a feedback law @xmath10 .
stability properties of such models have been extensively analyzed . in an infinite - dimensional sitting
such a nonlinear feedback may change the topology of the problem and uniform stability becomes reliant on the regularity of solutions ( for example , see @xcite ) . a more general scenario would account for coefficients that depend on the solution itself :
@xmath11 assuming the well - posedness of an associated initial - boundary value problem can be resolved , if the term @xmath12 is not guaranteed to be strictly positive on a _ fixed _ appropriately configured set , then analysis of stability becomes much more involved since the region where the dissipation is active now evolves with the solution and may not always comply with the requirements of the geometric optics .
the case when this coefficient vanishes at zero displacement , namely , @xmath13 , will be referred to as _ degenerate _ damping .
such a degeneracy naturally arises when investigating energy decay of higher - order norms .
for example , the natural energy space for a semilinear wave problem @xmath14 is @xmath15 and @xmath16 . with more regular initial data one can consider behavior of higher - order energy norms , namely for @xmath17 .
one approach would be to differentiate the pde in time which via the substitution @xmath18 leads to a degenerately damped problem @xmath19 a particular example can be observed in the relation between maxwell s system and the ( vectorial ) wave equation . for a given medium
denote the electric permittivity by @xmath20 , magnetic permeability by @xmath21 and conductivity by @xmath22 .
then maxwell s system reads @xmath23 with @xmath24 . on a bounded domain ,
subject to the _ electric wall _
boundary conditions , and for scalar - valued @xmath21 , @xmath20 , @xmath22 with positive lower - bounds , the term @xmath25 exponentially stabilizes this system @xcite . in a more accurate nonlinear conduction model the coefficient @xmath22 may depend on the intensity of the electric field @xmath26 . if we consider , for example , @xmath27 for @xmath28 , then differentiating the first equation in time and combining with the equation for @xmath29 gives @xmath30 for example , taking , @xmath31 gives @xmath32 where @xmath33 is the normalized vector @xmath26 .
the term @xmath34 has features of the viscous dissipation in this second - order equation , but nonlinear conductivity augments it with a degenerate coefficient @xmath35 .
the study of stability for the above models is much more delicate than in the situations where the damping , even nonlinear , depends on the time - derivative only .
weighted energy methods from basic energy laws to carleman estimates ( e.g. @xcite)have been successfully used to derive stabilization and observability inequalities for distributed parameter systems . however
, these methods typically rely on the properties of the coefficients to ensure that suitable geometric optics conditions are satisfied and the control effect suitably propagated " @xcite across the physical domain .
one can sometimes dispense with geometric optics requirements for smooth enough initial data @xcite , yet even then the support of the control / damping term must contain a subset that is time - invariant ( and with any time - dependent coefficients being non - vanishing , e.g. , as in @xcite ) . in turn , the analysis of control - effect propagation when the coefficients themselves depend on the solution and possibly go to zero wherever and whenever the solution does would require new techniques .
the following semilinear model , if recast in a higher - dimensional setting becomes highly non - trivial even when just regarding local wellposedness . in a one - dimensional framework
the nonlinearity is more tractable , but the rigorous stability analysis has long been open .
we focus on an elastic string with a _ degenerate damping _ , namely a dissipative term whose coefficient depends on and may vanish with the amplitudes : @xmath36 fixed at the end - points @xmath37 and with a prescribed initial configuration at @xmath38 : @xmath39 the initial data @xmath40 live in the natural function spaces revisited below .
function @xmath41 is assumed to be continuous non - negative , hence the term @xmath42 a priori should provide some form of energy dissipation in the model .
the scenario of interest is when @xmath43 as @xmath44 , essentially causing the dissipative effect to deteriorate at small amplitudes .
we will focus on the polynomial case @xmath45 satisfying the locally lipschitz estimate @xmath46 existence and uniqueness of weak _ finite - energy _ solutions to was proven in @xcite by means of galerkin approximations .
the advantage of a 1d framework is that the displacement function is absolutely continuous , hence topologically @xmath42 is still in @xmath47 , as in the case of the corresponding linear model .
however , in higher - dimensional analogues this embedding property is lost and proving existence becomes a markedly more complex task .
first , fractional damping exponents were considered in order to ensure that the damping term is bounded with respect to the finite energy topology @xcite .
arbitrary damping exponents were subsequently examined in @xcite . due to
the loss of regularity solutions had to be characterized via a variational _ inequality _ and established by a rather technical application of kakutani s fixed point theorem .
on the other hand , stability analysis even in one dimension poses a challenge that has been open for a number of years . despite the gain in regularity , attributed to sobolev embeddings ,
the key difficulty now is that energy estimates require some sort of information on the region where the damping is supported . in both the magnitude and the support of the damping coefficient
evolve with the geometry of the state , rendering all standard techniques inapplicable .
it is plausible to assume that some sort of a logarithmic uniform decay rate can be verified , possibly by combining ode techniques ( e.g. @xcite ) with pointwise carleman - type estimates .
another , though a rather weak , tentative indication of this outcome would be the uniform stability of the corresponding finite - dimensional analogue ( see the appendix ) .
yet the situation in infinite dimensions turns out rather different .
the goal of this article is to examine analytically and numerically stability properties of the dynamical system associated to : * establish global persistence of regularity in solutions with smooth initial data .
besides theoretical interest such a result is useful to justify the convergence estimates for numerical approximations . *
prove that a polynomial degeneracy in the damping of the form yields a system that is * not uniformly stable*. * present a numerical scheme that indicates the loss in decay rates .
such observations had been performed first and , in fact , served as a motivation for the theoretical results presented here .
the notation employed throughout the paper is summarized in section [ sec : notation ] .
the two main results on well - posedness and stability are stated in section [ sec : main ] .
several auxiliary technical definitions used in the proofs can be found in section [ sec : aux ] .
local and global wellposedness are verified respectively in sections [ sec : local ] and [ sec : global ] .
they draw upon two regularity lemmas proved earlier in section [ sec : regularity ] .
the proof of the lack of uniform stability is the subject of section [ sec : nonuniform ] .
numerical results are the subject of section [ sec : numerics ] .
the appendix contains results pertaining to the ode analog of the considered problem , namely , a damped harmonic oscillator with the damping coefficient dependent on the displacement .
this section serves as a quick reference for the basic notation used thorough the paper with some of the symbols revisited and discussed in more detail later in the text .
henceforth @xmath48 will denote the norm on a normed space @xmath49 . for the space @xmath50 we will use @xmath51 with the corresponding inner product denoted by @xmath52 .
we will also frequently involve the sobolev space @xmath53 associated to an equivalent inner product and norm @xmath54 the bilinear form @xmath55 will indicate the pairing of @xmath56 and its continuous dual @xmath57 .
we will also frequently use spaces of the form @xmath58;x ) { \quad\text{or}\quad}l^p(0,t ; x),\ ] ] which will be abbreviated respectively as @xmath59 looking ahead , for the one - dimensional dirichlet laplacian operator @xmath6 ( discussed below ) let us introduce the space @xmath60 equipped with the natural graph norm .
for example , @xmath61 ; { \mathscr{d}}(a^{1/2 } ) ) \cap c^1([0,t ] ; { { l^2(\omega)}})$ ] indicates the standard regularity in space and time for a weak solution to a linear wave equation . in turn
@xmath62 does the same for a strong solution to such an equation .
we will also be using spaces @xmath63 thus @xmath64 is the natural finite energy state space for a linear wave problem and @xmath65 denotes the domain of the corresponding evolution generator .
relation @xmath66 will occasionally be used to indicate that @xmath67 for a constant @xmath68 which only depends on the main parameters of the system , e.g. , the size of the domain @xmath69 or the exponent @xmath70 of the coefficient in . for convenience
let us summarize some of the fundamentals .
consider the operator @xmath71 which is positive , self - adjoint with compact resolvent , and has eigenvalues @xmath72 with the corresponding eigenfunctions @xmath73 the @xmath74 s form an orthonormal basis for @xmath50 .
for @xmath75 we can define fractional powers of @xmath6 : @xmath76 with @xmath77 the eigenfunctions @xmath78 form an orthogonal basis for every @xmath79 , @xmath75 .
some of the fractional powers can be identified with sobolev spaces , e.g. @xmath80 ' = h^{-1}({\omega})\,.\ ] ] since in our situation the model is one - dimensional , then trivially no issues in these identifications arise in regard to the regularity of the domain .
operator @xmath6 also corresponds to the riesz isomorphism @xmath81 , and for @xmath82 we have @xmath83 the natural finite - energy state space associated with the evolution driven by
is @xmath84 if we set @xmath85 we can recast this problem as an evolution equation @xmath86 for the skew - adjoint operator @xmath87 @xmath88 with @xmath89 we will also consider smoother solutions for which we define @xmath90 with the associated graph norm given via @xmath91|^2 + |{\mathcal{f}}_k[u_1]|^2\right)\ , .
\end{split}\ ] ] here @xmath92 denotes the @xmath93-th fourier coefficient with respect to the hilbert basis of @xmath94 given by the eigenfunctions @xmath95 of @xmath6 .
note also that in one - dimension the following continuous injection holds @xmath96 for @xmath97 if we adopt the notation convention @xmath98 .
we start with a formalized notion of a solution to the class of pde systems of the form . the following slightly more abstract formulation will help streamline the subsequent discussion .
let @xmath99\!\!\!\;\big]}}$ ] be the shorthand for the initial - boundary value problem : @xmath100 with the indicated derivatives taken in the sense of distributions , and subject to boundary conditions @xmath101 and initial data @xmath102 [ def : weak ] suppose @xmath103 and for some @xmath104 , @xmath105
. then we say a function @xmath106 is a weak solution to @xmath107\!\!\!\;\big]}}$ ] on interval @xmath108 $ ] if a. @xmath109 , b. for any @xmath110 the scalar map @xmath111 is absolutely continuous ( hence a.e .
differentiable ) on @xmath108 $ ] . c. for any @xmath112 @xmath113 [ def : regular ] a function @xmath114 on @xmath108\times { \omega}$ ] will be described as * regular of order * @xmath115 if it is continuously differentiable in time with the following regularity @xmath116 in classical terminology , weak solutions correspond to order @xmath117 and strong solutions to order @xmath118 .
suppose @xmath7 has the weak regularity ( regular of order @xmath117 ) . then according to the ( 1d ) sobolev embedding @xmath119 for @xmath120 , the function @xmath121 is well - defined as an element of @xmath122 .
in fact we will generalize this statement for the purposes of subsequently analyzing more regular solutions .
[ prop : f - regularity ] let @xmath123 for @xmath124 .
if @xmath125 , then @xmath126 \in c_t^j { \mathscr{d}}(a^{\frac{n - s - j}{2 } } ) , { \quad\text{for}\quad}s , j\in { \mathbb{n}}\cup\{0\},\quad s+j\leq n\,.\ ] ] in addition , @xmath127_{k=1,\ldots , j-1 } ( 1 + \|{\partial}_t^j \dot{z}\|_{c_t{{l^2(\omega)}}}),\quad j\leq n,\ ] ] where @xmath128 is a polynomial in @xmath129 variables .
[ example-1 ] due to a variety of spaces and indices involved in the statement of proposition [ prop : f - regularity ] , it is helpful to look at a basic example .
take @xmath130 , so @xmath131 and consider the regularity order @xmath132 .
then the condition @xmath133 reads @xmath134 in particular , @xmath135 , which corresponds to @xmath136 square - integrable derivatives , first three of which satisfy zero boundary conditions .
we have , for example , @xmath137 & = & 30 \dot{z } \ddot{z}^2 + 20 \dot{z}^2 \dddot{z } + 20 z \ , \ddot{z } \ , \dddot{z } + 10 z \dot{z } { \partial}_t^4z + z^2 { \fbox{$\displaystyle { { \partial}_t^4 \dot{z}}$}}\,.\end{aligned}\ ] ] thus , for instance , @xmath138 can be estimated using a polynomial of @xmath139 bounds on the functions @xmath140 , @xmath141 , , @xmath142 , and one term involving the @xmath50 norm of the fifth derivative of @xmath140 in time or , equivalently , @xmath143 .
this is precisely the conclusion of . likewise ,
if we consider , say , the @xmath144-nd derivative in space and @xmath144-nd in time to @xmath145 , @xmath146 in we get : @xmath147 = 2\dot{z}^3 + 6z \dot{z } \ddot{z}+z^2 \dddot{z}\ ] ] @xmath148= & 12\dot{z}_x^2\dot{z}+12z_x\ddot{z}\dot{z}_x+12z_x\dot{z}\ddot{z}_x+2z_x^2\dddot{z}+4z\dddot{z}_xz_x+12z\ddot{z}_x\dot{z}_x+6\dot{z}z_{xx}\ddot{z}+6\dot{z}^2\dot{z}_{xx}\\ & + 2z\dddot{z}z_{xx}+6z\ddot{z}\dot{z}_{xx}+6z\dot{z}\ddot{z}_{xx}+z^2{\fbox{$\displaystyle { \dddot{z}_{xx}}$ } } \end{split}\ ] ] we have that @xmath149 $ ] has zero trace as follows from and the zero boundary condition on @xmath140 .
moreover , the highest - order term @xmath150 in @xmath151 $ ] can be bounded in @xmath50 since by we have @xmath152 .
the rest of the terms are in fact in @xmath139 .
this confirms that @xmath153\in c_t { { l^2(\omega)}}$ ] in agreement with . note that if we had , say , @xmath154 then in the same context we would need to prove that @xmath155 \in c_t { \mathscr{d}}(a)$ ] instead of just @xmath156 .
first of all , @xmath154 would give @xmath157 implying that @xmath158 have zero traces .
hence so do their time - derivatives and then it is immediately follows that @xmath155 $ ] satisfies zero trace condition . and
taking two more derivatives in space in yields the highest - order term @xmath159 which is in @xmath50 from @xmath160 that implies @xmath161 .
thus , @xmath162 and whereas other summands in @xmath163 $ ] belong to @xmath139 .
we conclude that if @xmath154 , then @xmath155 $ ] ( is in @xmath50 ) has zero trace and integrable second derivative , again in accordance with .
let @xmath164 and first let s show that @xmath165 belongs to @xmath166 . for @xmath167
, @xmath165 is a polynomial in the following variables @xmath168_{k=0,1,\ldots , j } \ ] ] that is affine with respect to the highest - order derivative @xmath169 , which is at least in @xmath170 by the assertion that @xmath125 ( recall and plug @xmath171 ) .
we can just bound coefficients of @xmath172 using the fact that @xmath56 embeds continuously into @xmath173 .
thus the bound follows .
next , @xmath174 embeds into @xmath175 for @xmath176 , so all of the terms in @xmath177 are @xmath178 except for the highest order term @xmath179 . since @xmath177 is affine with respect to that term with continuous coefficients , then @xmath180 it follows that @xmath181 to strengthen this regularity to @xmath182 we must verify the boundary conditions . since @xmath183 coincides with the @xmath184 functions that have zero trace , then it is sufficient to show that @xmath185 that is , we can show that every derivative of total order ( time @xmath186 space ) up to @xmath187 of @xmath188 vanishes on the boundary .
the asserted regularity @xmath189 implies that @xmath190 has zero trace for any @xmath191 and any @xmath93 .
since any @xmath192-order ( space and time together ) derivative of @xmath188 involves at most @xmath193-st order terms in @xmath140 , then readily follows .
thus @xmath194 because @xmath195 is by definition a bounded operator on @xmath196 , then for @xmath197 we have @xmath198 confirming .
relying on ( a special case of ) proposition [ prop : f - regularity ] we formulate the notion of solution to .
we say @xmath114 is a weak solution to on @xmath108 $ ] if it is a weak solution to @xmath107\!\!\!\;\big]}}$ ] with @xmath199 ( which is in @xmath200 by proposition [ prop : f - regularity ] for @xmath201 ) .
[ definition - energy ] for a function @xmath202 define the quadratic energy functional of order @xmath203 by @xmath204 with the above definitions in mind , the new results of interest are : [ thm : well - posed ] let @xmath41 be as in with exponent @xmath205 , @xmath124 .
suppose for some integer @xmath206 the initial condition satisfies @xmath207 then there exists a unique weak solution @xmath114 , regular of order @xmath203 ( definition [ def : regular ] ) , of on @xmath108 $ ] for any @xmath104 .
moreover @xmath208 , that is , @xmath209 for @xmath210 [ ex : square ] if a solution to with @xmath211 has initial data given by the eigenfunctions of @xmath6 ( every @xmath79 ) , then @xmath212 for every @xmath28 .
[ thm : unstable ] the dynamical system generated by on the state space corresponding to weak solutions @xmath213 is non - accretive , but is * not * uniformly stable .
specifically , the energy functional @xmath214 is continuous non - increasing ; however , for any constants @xmath215 and any time @xmath216 there exists an initial datum @xmath217 such that the corresponding solution trajectory on @xmath218 $ ] does not intersect @xmath219 .
however , for any @xmath220 the lower - order norm @xmath221 decays to zero as @xmath222 with the decay rate uniform with respect to bounded sets of initial data .
in addition , @xmath223 is strictly monotonically decreasing on any interval where @xmath223 is positive .
the question of _ strong _ stability of the dynamical system generated by on the finite energy space @xmath64 remains open , as far as we are aware .
that is , for any given solution can we prove that @xmath224 ?
for convenience let us summarize a few classical results that can be easily verified , for example , using separation of variables . [
lem : linear ] consider the problem @xmath225\!\!\!\;\big]}}$ ] with @xmath226 .
then there exist a group @xmath227 of linear operators on @xmath64 such that @xmath228 determines a weak solution to this problem on every @xmath108 $ ] , @xmath104 .
the group is explicitly given by @xmath229 { \displaystyle}-\sqrt{{\lambda}_k } \sin(\sqrt{{\lambda}_k } t ) & \cos(\sqrt{{\lambda}_k } t ) \\[.05 in ] \end{bmatrix}}{\begin{bmatrix } { \mathcal{f}}_k[u_0]\\[.2in]{\mathcal{f}}_k[u_1 ] \end{bmatrix } } : { \begin{bmatrix } e_k \\[.2 in ] e_k \end{bmatrix}}\ ] ] where @xmath230 are the eigenvalue - eigenfunction pairs for the operator @xmath6 defined in , and @xmath92 is the @xmath93-th fourier coefficient with respect to the hilbert basis @xmath231 for @xmath50 : @xmath232 moreover , @xmath233 is a unitary operator on the space @xmath234 with respect to the norm given by , as follows by direct calculation using .
the next result is likewise well - known : [ prop : linear - problem ] consider the problem @xmath235\!\!\!\;\big]}}$ ] with @xmath236 if @xmath237 , this problem possesses a unique weak solution @xmath238 .
then the continuous mapping @xmath239 for energy functional satisfies @xmath240\,.\ ] ] in particular , from the gronwall estimate it readily follows @xmath241\,.\ ] ] moreover , if @xmath242 and @xmath243 , then @xmath244 and @xmath245 ( e.g. see ( * ? ? ? * thm .
2.1 , p. 229 ) ( instead of @xmath246 " ) . in the second half of that theorem , which is the one we cite ,
it is strengthened to @xmath247 for strong solutions . ] ) suppose @xmath114 is a weak solution of on @xmath108 $ ] .
then for any test - function @xmath248 with @xmath249 $ ] , it satisfies the variational identity @xmath250 let @xmath251)$ ] , then for any @xmath112 we have from @xmath252 let @xmath78 be the orthonormal basis for @xmath50 consisting of the eigenfunctions of @xmath6 . given @xmath253
we can represent it as @xmath254 for @xmath255)$ ] , @xmath256 .
this series that converges to @xmath257 in @xmath258 and its time - derivative @xmath259 converges to @xmath260 in @xmath261 . by applying identity to finite sums @xmath262 and passing to the limit @xmath263 we recover .
the existence of finite energy solutions to is known .
well - posedness for _ regular solutions _ , however , requires more work and relies on the connection between the smoothness in space and smoothness in time as summarized by the diagram below : @xmath264 [ cols="^,^,^ " , ] the purpose of this subsection is to furnish this connection which can be loosely outlined as follows : a weak solution @xmath114 of is regular of order @xmath203 on @xmath108 $ ] if and only if the @xmath203-th order energy is bounded on @xmath108 $ ] .
we split this claim into two propositions . in order not to keep track of how the structure of @xmath42 changes after differentiation we introduce the following , somewhat abstract property : [ def : reg - dependence ] we say two functions @xmath265 have * order @xmath203 dependence *
if for every @xmath266 the regularity ( * if * it holds ) @xmath267 * would imply * that @xmath268 where @xmath269 is a non - negative integer .
it is helpful to note : * if @xmath270 have order @xmath203 dependence , they trivially have order @xmath271 dependence for any @xmath272 . *
if @xmath270 have order @xmath203 dependence , then for @xmath273 , @xmath274 the functions @xmath275 and @xmath276 have order @xmath277 dependence .
again , it is helpful to consider an example .
let @xmath278 .
then it is not hard to check that @xmath270 have order @xmath203 dependence . for instance take @xmath279 and assume @xmath280 . as was shown via and in example [ example-1 ] @xmath281 continuously in time .
in particular , @xmath282 .
[ prop : regular - differentiable]suppose that @xmath114 is a weak solution on @xmath108 $ ] to @xmath107\!\!\!\;\big]}}$ ] with @xmath283 and @xmath284 .
assume functions @xmath285 have order @xmath203 dependence for some @xmath97 .
suppose , in addition , @xmath286 with @xmath287 , then @xmath288 in other words , @xmath289 .
moreover , @xmath290 for a constant @xmath291 dependent only on @xmath203 , @xmath292 , @xmath293 and being continuous monotone increasing with respect to the parameters @xmath292 , @xmath293 . throughout the argument below , the norms in the considered spaces can be ( inductively ) estimated in terms of @xmath292 and @xmath293 , thus ultimately verifying .
we will focus in detail on proving the claimed regularity . * cases @xmath294 , @xmath295 . * by assumption we always have @xmath296 and @xmath297 which takes care of the case @xmath294 .
if we in addition assume @xmath298 , then the equation @xmath299 implies that @xmath300 . since @xmath301 on the boundary then @xmath302 .
conclude : @xmath303 for @xmath304 proceed by induction : suppose the result of this proposition holds for @xmath305 .
assume holds .
then let us show .
* case @xmath304 . *
because @xmath296 , then condition implies @xmath306 . as a special case ( using @xmath307 instead of maximal @xmath203 )
, we have @xmath308 moreover , functions @xmath309 have a fortiori @xmath193-st order dependence , so the induction hypothesis gives @xmath310 next , by assumption , we also have @xmath311 to show that @xmath312 , it remains to verify that for @xmath313 we have @xmath314 .
to this end introduce @xmath315 since @xmath309 have @xmath203-th order dependence , then @xmath140 and @xmath316 have @xmath193-st order dependence ( see definition [ def : reg - dependence ] ) . as we already know ,
@xmath317 , and via the @xmath193-st order dependence of @xmath309 , we have @xmath318 for @xmath191 . because @xmath304 , then @xmath319 ; likewise @xmath304 also tells us that @xmath114 is at least in @xmath320 , so @xmath321 . consequently ,
applying @xmath322 to the equation for @xmath114 , we conclude that @xmath323 is a weak solution to @xmath324\!\!\!\;\big]}}$ ] with @xmath325 and @xmath326 . by
we also have @xmath327 the induction hypothesis now states that @xmath328 it implies that @xmath329 as desired . from here , along with , it follows that @xmath208 which completes the proof of the implication
case @xmath330 case @xmath203 " for the induction argument .
the next result complements the previous proposition , demonstrating that the same regularity @xmath331 of solutions can be inferred from a priori differentiability in space , rather than in time .
[ prop : differentiable - regular ] suppose @xmath114 is a weak solution on @xmath108 $ ] of @xmath107\!\!\!\;\big]}}$ ] with @xmath283 .
assume @xmath309 have @xmath203-th order dependence ( definition [ def : reg - dependence ] ) for some @xmath97 . if @xmath114 is regular of order @xmath203 , i.e. , @xmath332 , then @xmath333 and @xmath334 with @xmath335 dependent only on @xmath203 and @xmath336 and continuous monotone increasing with respect to parameter @xmath336 .
in the course of the proof , the bound can be traced inductively to depend only on @xmath337 . to keep the exposition concise
the argument will focus on the regularity .
if @xmath294 , the claimed regularity simply matches that of weak solutions . for @xmath295 we are given @xmath302 and @xmath338 .
solving @xmath339 verifies that @xmath340 .
proceed by induction .
fix @xmath304 , suppose the statement holds for all @xmath341 and assume @xmath337 is finite .
a fortiori it follows that @xmath342
. then @xmath317 by the induction hypothesis .
define @xmath343 , then using @xmath317 it follows as in the proof of proposition [ prop : regular - differentiable ] that @xmath140 is a weak solution to @xmath344\!\!\!\;\big]}}$ ] with @xmath345 .
moreover from the assumption @xmath346 we also have that @xmath347 , i.e. , is regular of order @xmath307 .
thus by the induction hypothesis @xmath348 equivalently @xmath349 the only remaining step from here to proving @xmath208 is to show that @xmath350 . to this end
define @xmath351 then by @xmath352 since @xmath317 and by the @xmath193-st order dependence ( implied by @xmath203-th order dependence ) of @xmath114 and @xmath188 , we deduce from that @xmath353 .
so @xmath354 confirming that @xmath355 as desired .
thus @xmath208 completing the induction argument .
it is known that possesses unique solutions @xcite .
here we extend this result to regular solutions as well .
first we formulate it for local in time solutions .
[ thm : local ] suppose @xmath356 for a non - negative integer @xmath203
. then there exists @xmath357 such that system has a local unique solution that is regular of order @xmath203 on interval @xmath108 $ ] .
* step 1 : the spaces . * note that for @xmath358 @xmath359 moreover , for @xmath360 we have @xmath361 with the temporary notational convention @xmath362 .
* step 2 : contraction mapping .
* let @xmath227 be the semigroup for the linear wave equation with data @xmath363 .
note that if @xmath140 is regular of order @xmath203 , then by proposition [ prop : f - regularity ] we would get @xmath364 with this in mind introduce the operator @xmath365 on @xmath366 , @xmath367 for @xmath249 $ ] we have @xmath368 from the local lipschitz property of @xmath41 we obtain @xmath369 + \left[f({\tilde}{\phi}_0)\phi_1 -f({\tilde}{\phi}_0){\tilde}{\phi}_1\right ] \\
= & \phi_1 m(\phi_0,{\tilde}{\phi}_0)(\phi_0 - { \tilde}{\phi}_0 ) + f({\tilde}{\phi}_0)(\phi_1 - { \tilde}{\phi}_1)\,.\ , \end{split}\ ] ] now we will rely on the fact that a priori @xmath370 gives : @xmath371 estimate , @xmath372\right|_{_0}\\ \leq & \left| \phi_1 m(\phi_0,{\tilde}{\phi}_0)(\phi_0 - { \tilde}{\phi}_0 ) + f({\tilde}{\phi}_0)(\phi_1 - { \tilde}{\phi}_1 ) \right|_{_0}\\ & + \left| \sum_{i+j+k = n } c_{ijk}\,{\partial}_x^i\phi_1 \cdot { \partial}_x^j m(\phi_0,{\tilde}{\phi}_0)\cdot { \partial}_x^k(\phi_0 - { \tilde}{\phi}_0 ) \right|_{_0 } + \left| \sum_{i+j = n } d_{ij}\ , { \partial}_x^i f({\tilde}{\phi}_0)\cdot { \partial}_x^j ( \phi_1 - { \tilde}{\phi}_1)\right|_{_0}\\ \leq & { \mathcal{p}}_1\left\ { \|\phi_0\|_{c^{n}({\overline}{{\omega } } ) } , \|{\tilde}{\phi}_0\|_{c^{n}({\overline}{{\omega}})}\right\ } \cdot \|\phi_1\|_{w^{n,2}({\omega } ) } \cdot \|\phi_0 - { \tilde}{\phi}_0 \|_{c^{n}({\overline}{{\omega } } ) } + { \mathcal{p}}_2\left\ { \|{\tilde}{\phi}_0\|_{c^{n}({\overline}{{\omega}})}\right\ } \cdot \| \phi_1-{\tilde}{\phi}_1\|_{w^{2,n}({\omega } ) } \end{split}\ ] ] where @xmath373 are a polynomials in the indicated variables with positive coefficients .
now , from @xmath374 so continuing we get @xmath375 because the group @xmath233 is unitary on @xmath376 then yields @xmath377 if @xmath378 and @xmath379 come from a bounded set , then choosing @xmath380 small enough implies that @xmath381 is a contraction .
* step 3 : invariance . * finally , we also need @xmath365 to map the the ball @xmath382 in @xmath366 into itself . according to for that it suffices that @xmath336 satisfy @xmath383 for which we can impose @xmath384
now the contraction mapping principle implies the claimed local unique solvability .
the global existence result is based on a priori bounds on the energy .
[ prop : extension ] let @xmath41 be as in with exponent @xmath205 , @xmath124 .
assume @xmath114 is a weak solution that is regular of order @xmath203 , defined on some right - maximal interval @xmath385 .
suppose @xmath386 and that there exists a continuous function @xmath387 on @xmath388 such that @xmath389 then @xmath390 . by proposition [ prop : f - regularity ] , @xmath114 and the term @xmath391 have @xmath203-th order dependence . recall that energy @xmath392 controls the @xmath366 norm of the solution ( in fact , it controls the @xmath393 norm , but the continuity with values in @xmath394 is implied by the definition of regular solution ) .
then there is a constant @xmath335 dependent on @xmath395 such that @xmath396 for any @xmath397 .
hence by theorem [ thm : local ] any initial data of the form @xmath398 and @xmath399 for @xmath400 can be extended to a solution that exists for another @xmath401 time units , independently of @xmath400 .
hence @xmath402 can not be finite .
[ lem : energy - regular ] let @xmath41 be as in with exponent @xmath205 , @xmath124 .
suppose @xmath114 is a weak solution to on @xmath108 $ ] .
if @xmath114 is regular of order @xmath203 , then for @xmath403 we have @xmath404 , @xmath405 and @xmath406 , { \partial}_t^k \dot{u})_{_0 } = e^{(k)}_u(0)\,.\ ] ] proposition [ prop : f - regularity ] confirms that @xmath188 and @xmath407 have @xmath203-th order regularity dependence .
invoke proposition [ prop : regular - differentiable ] to conclude that @xmath208 .
it certainly implies that @xmath408 for @xmath403 . and by the @xmath203-th order regularity dependence of @xmath309 we also have @xmath409 .
define @xmath410 then @xmath411 is a weak solution to @xmath412\!\!\!\;\big]}}$ ] with @xmath413 and @xmath414 .
now call upon the energy identity for @xmath411 to obtain for @xmath114 .
let @xmath114 be a weak solution on interval @xmath108 $ ] . because @xmath415 is non - negative by assumption on @xmath41 , then in the case @xmath294 by lemma [ lem : energy - regular ] we have @xmath416 } e_u(t ) \leq e_u(0)\,.\ ] ] since the bound is independent of @xmath417 , then by proposition [ prop : extension ] @xmath114 extends globally to @xmath418 .
arguing by induction , suppose that @xmath419 is a global solution that is regular of order @xmath203 , and at the same time is _ local _ regular of order @xmath420 on a maximal interval @xmath421 .
we want to show that it is also global of order @xmath420 .
it helps to preview the result on the example of @xmath422 and by extending from weak to regular of order @xmath118 solutions .
suppose @xmath114 is a global weak solution that is , in addition , regular of order @xmath423 on maximal interval @xmath421 .
for such a solution by proposition [ lem : energy - regular ] we have the
@xmath118-st order " energy identity @xmath424 \ddot{u } = e^{(1)}_u(0)\,.\ ] ] @xmath425 we can bound @xmath426 norms of @xmath114 and @xmath427 in terms of the finite energy @xmath223 .
the latter is bounded by constant @xmath428 ( but , in fact , any continuous on @xmath388 upper bound on @xmath223 would do , which is how the general argument works ) . in turn , the term @xmath429 is a part of the first - order energy .
so we have @xmath430 from here , gronwall s estimate gives us an asymptotically growing continuous upper bound on @xmath431 .
then proposition [ prop : extension ] ensures that @xmath114 is regular of order @xmath118 globally , that is @xmath390 .
now onto the actual proof of global existence . by the @xmath203-th order global regularity
there is a continuous monotone increasing function @xmath432 @xmath433 by @xmath434st order regularity and via lemma [ lem : energy - regular ] for any @xmath435 we also have @xmath436 use estimate of proposition [ prop : f - regularity ] to claim that there is a polynomial @xmath437 such that for each @xmath438 @xmath439 then @xmath440 . in particular
, @xmath441 gronwall s inequality now provides @xmath442 by the assumption on @xmath443 , the newly defined mapping @xmath444 is continuous on @xmath388 .
then proposition [ prop : extension ] ensures that the solution @xmath114 is global , regular of order @xmath420 , which completes the proof by induction .
this step also completes the proof of theorem [ thm : well - posed ] .
the additional regularity @xmath208 follows if we first use proposition [ prop : f - regularity ] , which establishes @xmath203-th order dependence of @xmath114 and @xmath42 , and then invoke proposition [ prop : regular - differentiable ] which translates @xmath203-th order regularity into @xmath331 .
this section is devoted to the proof of theorem [ thm : unstable ] .
the heart of the argument is that the energy decay slows down with the frequency of solutions , independently of their total finite energy .
let @xmath256 and consider initial condition of the form @xmath445 note that @xmath446 for any @xmath75 . by direct calculation
we have @xmath447 hence for such initial data , any constant that may be estimated in terms of @xmath448 is bounded uniformly in @xmath449 .
we continue working with the initial data @xmath450 as in . pick the anti - derivative through the origin of @xmath41 ( as in ) @xmath451 due to @xmath452 ( hence @xmath453 ) being bounded independently of @xmath449 , we have for some constant @xmath68 * independent of @xmath93 in * the estimate @xmath454 let function @xmath455 solve the linear elliptic boundary value problem @xmath456 it satisfies the standard elliptic estimate : @xmath457 for @xmath458 and @xmath68 * independent of @xmath93*. we do have @xmath459 because @xmath460 .
in fact , @xmath461 is an eigenfunction , and it is easy to show ( as in proposition [ prop : f - regularity ] ) that @xmath462 we then have @xmath463 with this function @xmath455 at hand , consider the nonlinear primitive " problem @xmath464 with homogeneous dirichlet boundary data @xmath465 and initial condition @xmath466 note the following properties of the system :
1 . because @xmath467 is continuous ,
monotone increasing vanishing at @xmath117 , then is a semilinear wave problem with monotone damping .
the data @xmath468 come from the domain of the associated nonlinear generator ; in this 1d setting it coincides with the domain @xmath65 of the corresponding linear group .
this well - posedness result is based on monotone operator theory ( * ? ? ?
3.7 , p. 306 ) .
+ in particular , we have for any @xmath104 @xmath469 moreover , given the 1d embeddings it follows that @xmath470 so @xmath471 satisfies and has @xmath472 . by proposition [ prop : linear - problem ]
we have @xmath473 2 .
the velocity @xmath474 satisfies @xmath475 and is a weak solution to @xmath476 in addition , @xmath477 and because @xmath478 , @xmath479 are continuous with values in @xmath50 , then @xmath480 according to the way @xmath481 was constructed in . _
thus the primitive problem is the velocity potential for our original problem .
namely the time derivative @xmath482 is the solution @xmath483 to our original degenerately damped system with initial data @xmath484 .
solutions to decay uniformly to zero at the rate that can be estimated explicitly in terms of the exponent @xmath205 of @xmath41 .
this is a fundamental example of a dissipative problem with full interior damping .
the decay rate can be assessed using , for example , the ode characterization of @xcite ( see ( * ? ? ?
* coro . 1 , p. 1770 ) for more details ; in particular function @xmath485 there must satisfy @xmath486 ) .
we get , as @xmath222 that @xmath487 it follows that the ( lower - order ) norm latexmath:[$|u^{(k)}(t)|_{_0}^2 = @xmath222 with the decay rate uniform with respect to the finite energy @xmath489 . by interpolation between @xmath50 and @xmath490
we conclude uniform a decay ( but at slower rates , where @xmath491 is modified by the interpolation exponent ) for every norm @xmath492 with @xmath493 .
+ note that uniform stability of the original system would have required us to include the case @xmath494 , which as we are about to show can not happen .
* summarizing : * @xmath495 is precisely the solution @xmath483 to @xmath496\!\!\!\;\big]}}$ ] .
any fractional @xmath497 sobolev norm of @xmath483 for @xmath493 decays uniformly with respect to @xmath489 .
in addition , we have the estimate @xmath498 for @xmath499 independent of @xmath500 , hence * independent of @xmath93 * in .
in particular , the lower order norm @xmath501 of the solution to our original problem @xmath502\!\!\!\;\big]}}$ ] is controlled by its initial lower - order norm @xmath503 * independently of @xmath93*. now we are going to use the smallness of the @xmath94 norm of @xmath504 to show that its @xmath505 norm can not decay too fast .
let @xmath506 be the unique solution to _ linear homogeneous _
problem @xmath507\!\!\!\;\big]}}$ ] . from the energy identity and by the choice of @xmath461 in we get @xmath508 define @xmath509 then @xmath510 solves @xmath511\!\!\!\;\big]}}$ ] .
the energy identity for @xmath140 gives @xmath512 note that * @xmath513 ( since @xmath514 ) * independently of @xmath93 * in .
* given @xmath41 as in we have @xmath515 . via 1-dimensional embeddings and interpolation for @xmath516 @xmath517 so @xmath518 because @xmath519 independently of @xmath93 in , then by we get for any @xmath418 @xmath520 for @xmath521 * independent of @xmath93 * in .
at this point we finally expand the definition of @xmath522 to get ( see ) @xmath523 plugging these observations into the estimate for @xmath524 we obtain : [ lem : decay ] for @xmath525 let @xmath483 denote the weak solution on @xmath108 $ ] to the problem @xmath526\!\!\!\;\big]}}$ ] and @xmath506 be the solution to linear homogeneous problem @xmath507\!\!\!\;\big]}}$ ]
. then the difference @xmath527 solves @xmath511\!\!\!\;\big]}}$ ] and for any @xmath528 it satisfies @xmath529\,,\ ] ] with @xmath530 * independent of * @xmath93 . at this point for brevity
let us suppress the superscript
@xmath531 " stemming from the choice of the parameter in the definition of initial data .
let @xmath532 be as in lemma [ lem : decay ] .
suppose for a moment that @xmath533 for some @xmath438 .
then via @xmath534 we have @xmath535 to apply this estimate , pick any @xmath536 and find @xmath537 such that @xmath538 ( e.g. , if @xmath539 ) .
fix @xmath528 , then there is @xmath540 large enough so that for initial condition @xmath541 yields a solution whose energy satisfies @xmath542\,,\ ] ] for @xmath543 as in lemma [ lem : decay ] .
consequently , by , @xmath544\,.\ ] ] however , the initial condition @xmath545 had energy @xmath118 ( again , independently of @xmath93 ) .
hence the family of initial conditions @xmath546 with the associated solutions @xmath483 , resides in a bounded ball ( of radius @xmath547 ) in @xmath64 , yet the corresponding solutions do not decay to zero uniformly in the topology of @xmath64 .
thus the associated dynamical system on @xmath64 is not uniformly stable .
this argument demonstrates theorem [ thm : unstable ] for @xmath548 and @xmath549 .
the general case follows merely by attaching a factor of @xmath550 to @xmath461 and choosing a potentially smaller @xmath551 in the last step of the argument .
for a weak solution @xmath114 of @xmath552\!\!\!\;\big]}}$ ] with @xmath226 the functional @xmath553 is non - increasing .
we can not presently claim whether solutions decay to @xmath117 strongly , however , it is possible to show that the energy @xmath223 is _ strictly _ monotonically decreasing for non - trivial solutions . consider a weak solution @xmath114 on @xmath108 $ ] and suppose @xmath554 on @xmath555 $ ] , then from the energy identity follows that @xmath556 thus @xmath557 a.e . in @xmath558 . since it is equivalent to @xmath559 ,
then we conclude that @xmath560 in @xmath50 for a.e .
@xmath561 : @xmath562 moreover , since @xmath563\times { \overline}{{\omega}})$ ] , then we have @xmath564 for every @xmath565 . by the continuity of the solution
this is only possible if @xmath566 for @xmath565 .
then @xmath567 and we arrive at an equilibrium solution which has to be trivial .
this observation completes the proof of theorem [ thm : unstable ] .
the strategy for the proof of instability was largely prompted by numerical observations described below .
the numerical implementation presented here treats the case @xmath568 of : @xmath569 with @xmath570 and given initial data @xmath571 solution was discretized in space via a ritz - galerkin method .
the dynamic problem could be analyzed explicitly using a discretization in time and a runge - kutta scheme , though , rigorous justification of convergence becomes more delicate .
another approach is to approximate the successive approximations of theorem [ thm : local ] which , when exact , are guaranteed to converge , at least over small time intervals .
the iterates correspond to _ linear _ inhomogeneous pde problems that are resolved using a hybrid scheme : a. for relatively short times find solutions using an approximation of semi - discrete ritz - galerkin method by discretizing time - integrals in the variation of parameter formula .
if the error in numerical integration is small , then this approach enjoys an explicit convergence estimate ( for smooth solutions and over finite time intervals ) essentially proportional to the space discretization parameter @xmath572 .
b. for larger times , collect the last @xmath93-points of the semi - discrete approximation and resolve the rest of the iteration using a multi - step method ( adams - bashforth ) .
thus , we begin with some initial guess @xmath573 } , \dot{u}_{[0]})$ ] and proceed to solve inhomogeneous linear problems @xmath574 } - { \partial}_{xx } u_{[k ] } = f_{[k-1]}\ ] ] on the space @xmath575 with the forcing term from the preceding iteration @xmath576 } = - u_{[k-1]}^2 \dot{u}_{[k-1]}\,.\ ] ] the constants in the estimates and could potentially be determined explicitly in this case ( by following the proof with specific @xmath577 and @xmath578 in the definition of @xmath41 ) which would yield an explicit bound on the lipschitz constant @xmath579 " of the contraction mapping in terms of @xmath380 . in turn , given this constant @xmath580 if @xmath581}-u_{[k-1]}\|_{c_t { \mathscr{h}}^0 } < { \varepsilon}\ ] ] then the absolute error between @xmath93-th iteration and the true solution is no more than @xmath582 .
if we use a ritz - galerkin scheme with element size @xmath572 to find an approximate solution @xmath583}$ ] to linear inhomogeneous problem , then for @xmath572 small , e.g. see ( * ? ? ?
13.1 , p. 202 ) , and for simplicity taking the initial conditions to be the more accurate ritz projections of the initial data , we get @xmath581 } - u_{[k , h ] } \|_{c_t{\mathscr{h}}^0 } \leq c h { \int_0^t}|\ddot{u}_{[k]}(s)|_2 ds\,.\ ] ] this estimate of course requires sufficiently regular solutions . as theorem [ thm : well - posed ] and example [ ex : square ] show , in order to have the @xmath584 regularity on @xmath585}$ ] it suffices to have initial data in the space @xmath586 for example , the demonstrated numerical results below use displacement and velocity proportional to the eigenfunctions of @xmath6 , which are smooth .
consider an equipartitioned mesh of subintervals of length @xmath572 and the standard piecewise linear nodal basis @xmath587 , with @xmath588 .
let @xmath589 denote the restriction of the linear evolution generator @xmath590 to the subspace @xmath591 of @xmath592 spanned by @xmath593 . by @xmath594 denote the elliptic ritz projection on the subspace of @xmath56 and let @xmath595 stand for the corresponding @xmath94 projection . given a candidate approximation @xmath596 } = { \begin{bmatrix } u_{[k-1,h]}\\ \dot{u}_{[k-1,h ] } \end{bmatrix}}\ ] ] we compute the forcing @xmath597}(t , x ) = u_{[k-1,h]}^2(t , x)\dot{u}_{[k-1,h]}(t , x)\ ] ] the coefficient vector @xmath598}$ ] of the projection of @xmath599}$ ] is obtained in terms of the coefficients @xmath600 ( which for this choice of basis functions @xmath601 form a very sparse tensor with only 3 distinct values ) . the initial guess used to calculate @xmath602}$ ]
is the constant solution : @xmath603}(t , x ) = ( { \mathcal{r}}_h^{1 } u_0(x))^2({\mathcal{r}}_h^{0 } u_1(x ) ) { \quad\text{for all}\quad}t\geq 0.\ ] ] let @xmath604 denote the projection of the initial data @xmath605 .
we obtain a semi - discrete approximation of the original system for unknown coefficient vector @xmath606}$ ] : @xmath607 } ' = -{\mathbb{a}}_h { \mathbf{y}}_{[k , h ] } + { \mathbf{f}}_{[k-1,h]},\qquad { \mathbf{y}}(0 ) = { \mathbf{y}}_0\,.\ ] ] for relatively short times we can invoke @xmath608}(s)ds\ ] ] which is in turn discretized over time scale @xmath609 with @xmath610 . according to @xmath611}(s)ds\,.\ ] ] at each step
only the integral over @xmath612 $ ] needs to be computed . for this purpose only several values of the matrix exponentials @xmath613 are needed in order to apply the newton - cotes rule on sub - interval @xmath614 $ ] , specifically : @xmath615 where @xmath578 is number of points used for newton - cotes formula ( e.g. , boole s or simpson s @xmath616th ) .
these @xmath617 matrices need to be computed just once and only depend on the time - step , but not the total number of these steps . in turn , the vectors @xmath618 have to be found for each @xmath619 .
but since @xmath604 is fixed , these can be more accurately determined using scaling and truncated taylor series approximation @xcite . for simulations over larger time - scales we can use the last few values : @xmath620 to initialize a linear @xmath70-step method , e.g. , @xmath136-step adams - bashforth to efficiently obtain solution on the interval @xmath621
$ ] . as before ,
let @xmath95 be the eigenfunction @xmath622 for @xmath6 with eigenvalue @xmath623 .
then for initial data @xmath624 for constants @xmath625 , @xmath626 , the solution of can be reduced to a dissipative ode using the ansatz @xmath627 plugging it into equation yields @xmath628 this identity would be implied if for each @xmath629 function @xmath378 solves the 2nd - order nonlinear ode @xmath630 @xmath631 it corresponds to a first - order nonlinear system : @xmath632 function @xmath633 is smooth with respect to the components of @xmath634 and to variable @xmath635 , which now acts as a parameter .
this ode system has global differentiable solutions , moreover since @xmath636 is smooth , in fact , analytic in @xmath635 then local solutions are differentiable with respect to @xmath635 ( * ? ? ?
3.1 , p. 95 ) . because the initial data is smooth then by theorem [ thm : well - posed ] the unique solution @xmath114 is , among other things , in @xmath637 .
consequently @xmath114 must coincide with the solution to the ansatz .
in turn , is a dissipative @xmath638 system of odes and can be approximated by a runge - kutta scheme .
to get some quantitative estimate on the absolute error of solutions found in section [ particulars ] , at least for initial data of the form , one can consider a piecewise linear interpolation of and then calculate the energy - norm difference from the finite - element solution .
the accompanying figures and data demonstrate some of the numerical results .
the initial data is considered of the form @xmath639 which permits to compare the finite element solutions to the pointwise runge - kutta solutions described in section [ sec : pointwise - rk ] .
figure [ fig:1 ] shows the point - value of displacement @xmath640 next to the displacement value at the same @xmath641 for the corresponding initial boundary value problem _ with linear damping_. figure [ fig:2 ] presents numerical estimates of the energy for solutions obtained by ritz - galerkin finite element scheme and successive approximations .
the graphs indicate that the energy decay deteriorates as the frequency of the initial data goes up while the initial finite - energy remains fixed ( @xmath642 independently of @xmath93 ) , thus illustrating the lack of uniform which was rigorously confirmed by theorem [ thm : unstable ] .
the initial data are of the form with zero initial velocity .
the indicated errors are obtained by comparing each finite - element solution to a piecewise - linear interpolant of the corresponding piecewise rk solution .
figure [ fig:3 ] uses multi - step extensions of the same solutions shown in figure [ fig:2 ] to a larger time - scale using ( 5-step ) adams - bashforth method .
it also includes the decay of the @xmath50-norm @xmath501 for these solutions .
the research of george avalos and daniel toundykov was partially supported by the national science foundation grant dms-1211232 . the numerical analysis presented in this
work and partial theoretical results were obtained during a research experience for undergraduates ( reu ) program on applied partial differential equations in summer 2013 at the university of nebraska - lincoln .
this reu site was funded by the national science foundation under grant nsf dms-1263132 .
the reu project would not have been possible without assistance of tom clark .
to complement the analysis of the infinite - dimensional model it is also interesting to examine the related finite - dimensional version of a degenerately damped harmonic oscillator : @xmath643 @xmath644 for @xmath645 , @xmath124 .
rewrite it as a first order evolution problem @xmath646 henceforth , let @xmath647 denote an equivalent norm on @xmath648 : @xmath649 because @xmath41 is smooth and non - negative , then classical ode results guarantee that solutions are unique , exist globally and satisfy @xmath650 below we present stability results which contrast their infinite - dimensional analogues discussed earlier . in particular , the finite - dimensional system is uniformly stable , while for distributed - parameter version the strong stability is open while uniform stability has been proven false .
this is a direct consequence of lasalle s invariance principle with the lyapunov function given by the equivalent norm : @xmath651 .
we only need to check what kinds of trajectories reside in the invariant set of the system : @xmath652 so for @xmath653 @xmath654 @xmath655 in particular , either @xmath656 or @xmath657 .
now suppose that a solution trajectory @xmath658 resides in @xmath26 .
if at some @xmath438 we have @xmath659 , then by the continuity in time , it is nonzero on some interval @xmath660 . on that interval
we must have @xmath661 by the property of @xmath26 .
but then on that interval , from equation we get a contradiction since the solution has to be constant and therefore zero .
proceed by contradiction .
assume for a bounded set @xmath662 there exists some @xmath663 , a bounded sequence of initial data @xmath664 , and a sequence of corresponding times @xmath665 with @xmath666 such that @xmath667 extract a convergent subsequence of initial data , reindexed again by @xmath203 .
let @xmath668 denote this limit point and @xmath669 be the corresponding solution . by lemma [ lem : lasalle ] @xmath670 in particular , there exists @xmath380 such that for @xmath671 we have @xmath672 .
because the non - linearity @xmath673 is locally lipschitz on @xmath648 , and the system is non - accretive , then there exists @xmath674 such that for any @xmath675 if @xmath676 , the corresponding solutions satisfy @xmath677\,.\ ] ] since @xmath678 , we can find @xmath679 .
next , because @xmath680 converge to @xmath681 , then for @xmath682 we can find find @xmath680 so that @xmath683 and , consequently , @xmath684 \,.\ ] ] then @xmath685 . because the system is non - accretive , then @xmath686 for all @xmath671 , and in particular it holds for @xmath687 .
so @xmath688 which contradicts the choice of @xmath680 .
roberto triggiani and peng - fei yao .
carleman estimates with no lower - order terms for general riemann wave equations . global uniqueness and observability in one shot .
, 46(2 - 3):331375 , 2002 .
special issue dedicated to the memory of jacques - louis lions .
irena lasiecka , roberto triggiani , and x. zhang .
global uniqueness , observability and stabilization of nonconservative schrdinger equations via pointwise carleman estimates .
i. @xmath689-estimates .
, 12(1):43123 , 2004 .
irena lasiecka , roberto triggiani , and x. zhang .
global uniqueness , observability and stabilization of nonconservative schrdinger equations via pointwise carleman estimates .
ii . @xmath690-estimates . , 12(2):183231 , 2004 .
roberto triggiani and xiangjin xu .
pointwise carleman estimates , global uniqueness , observability , and stabilization for schrdinger equations on riemannian manifolds at the @xmath689-level . in _ control methods in pde - dynamical systems _ , volume 426 of _ contemp .
_ , pages 339404 .
soc . , providence , ri , 2007 .
igor chueshov , irena lasiecka , and daniel toundykov .
long - term dynamics of semilinear wave equation with nonlinear localized interior damping and a source term of critical exponent . , 20(3):459509 , 2008 .
matthias eller and daniel toundykov .
carleman estimates for elliptic boundary value problems with applications to the stabilization of hyperbolic systems .
, 1(2):271296 , 2012 .
evolution equations and control theory .
philip hartman . , volume 38 of _ classics in applied mathematics_. society for industrial and applied mathematics ( siam ) , philadelphia , pa , 2002 . corrected reprint of the second ( 1982 ) edition [ birkhuser , boston , ma ; mr0658490 ( 83e:34002 ) ] , with a foreword by peter bates . | presented here is a study of well - posedness and asymptotic stability of a degenerately damped " pde modeling a vibrating elastic string . the coefficient of the damping may vanish at small amplitudes thus weakening the effect of the dissipation .
it is shown that the resulting dynamical system has strictly monotonically decreasing energy and uniformly decaying lower - order norms , however , is _ not uniformly stable _ on the associated finite - energy space .
these theoretical findings were motivated by numerical simulations of this model using a finite element scheme and successive approximations . a description of the numerical approach and sample plots of energy decay are supplied .
in addition , for certain initial data the solution can be determined in closed form up to a dissipative nonlinear ordinary differential equation .
such solutions can be used to assess the accuracy of the numerical examples .
[ pageinit ] @xmath0 department of computing & mathematical sciences , california institute of technology , ca 91125 + @xmath1 department of mathematics , university of nebraska - lincoln , ne 68588 + @xmath2 carroll college , mt 59625 + @xmath3 school of public health , university of michigan , mi 48104 + @xmath4 department of mathematics & statistics , texas tech university , tx 79409 + @xmath5 _ corresponding author : _ [email protected] |
let @xmath3 be a set of closed subspaces of a hilbert space @xmath11 endowed with the strong operator topology inherited from the identification of closed subspaces @xmath12 with their self - adjoint projections @xmath13:{{\mathcal{h}}}\to { { \mathcal{k}}}$ ] .
if @xmath3 is finitely parametrised in the sense that @xmath14 : \lambda \in m \subseteq { { \mathbb{r}}}^n\}\ ] ] with @xmath15 a topological manifold , then @xmath3 may in fact be homeomorphic to @xmath15 . furthermore @xmath3 may admit a certain local unitary description and an associated smooth structure under which it is a diffeomorph of a differentiable manifold in @xmath16 . natural examples of such manifolds of projections are provided by grassmannian manifolds and their submanifolds . also , operators @xmath17 in the first cowen - douglas class @xcite , @xcite for a complex connected domain @xmath18 in @xmath19 provide diverse realisations ( even for @xmath20 ) of domains in @xmath21 , namely @xmath22 : \omega \in \omega \subseteq { { \mathbb{r}}}^{2m}\}.\ ] ] however our primary motivating examples derive from invariant subspaces for semigroups of unitary operators , such as the semigroup @xmath23 associated with jointly irreducible one parameter unitary groups satisfying the weyl commutation relations . furthermore ,
such subspaces and their complementary spaces are generally infinite dimensional .
these examples motivate the consideration of general subspace manifolds as formulated in definitions [ def : cinfsubman ] , [ def : locunitsubman ] and [ def : beurlingsubman ] .
the embracing realm we consider is the traditional one allied to operator function theory , namely the set of subspaces of @xmath5 of the form @xmath24 where @xmath0 is the hardy space for the upper half plane , @xmath25 is unimodular , and @xmath26 is the space of functions supported on a measurable set @xmath27 .
we examine subspace manifolds of the form @xmath28 where @xmath29 is a finite dimensional real vector space of real - valued functions , we analyse limits of projections and we identify the associated closed topological manifolds . our approach and the ensuing identifications give a unified explanation for various so called `` strange limits '' of projections .
these include the special cases considered by katavolos and power @xcite , @xcite which were derived by ad hoc arguments leaning on operator algebra methods .
specifically we show by direct methods that the space of functions @xmath30 has subspace manifold @xmath31 whose closure is homeomorphic to the closed unit disc , while for the space @xmath32 the manifold @xmath33 has closure homeomorphic to the closed unit ball in @xmath34 .
in contrast , the closures of cowen - douglas projection manifolds are generally one point compactifications . a consequence of the limit projection analyses in @xcite
, @xcite is that a reducing invariant subspace for the weyl semigroup @xmath35 , or for the @xmath36 unitary semigroup ( with @xmath37 and @xmath38 , turns out to be a strong operator topology limit of a sequence of purely invariant projections , that is , a limit of those with no reducing part .
we obtain in theorem [ thm : purelimits ] a similar phenomenon for the multiplication semigroup @xmath39 acting on @xmath40 .
equivalently , translating to the circle @xmath41 , we show that each reducing invariant subspace for the bilateral shift , which corresponds to a measurable subset of @xmath41 , is a strong operator topology limit of beurling projections @xmath42 $ ] .
we are not aware of any other proof , direct or indirect , of this seemingly classical fact .
our principal tool is theorem [ thm : psi ] which , for a function @xmath43 in a certain admissible class , identifies the limit of the projections @xmath44 $ ] , as @xmath45 , with the projection @xmath46 $ ] .
the proof makes use of methods of stein @xcite for the analysis of oscillatory integrals .
we go on to show that for quite general @xmath8 dimensional spaces @xmath29 of admissible functions the subspace manifold @xmath47 has closure homeomorphic to the closed unit ball in @xmath16 .
moreover , in these cases the two - component subspace manifold @xmath48 has closure , denoted @xmath49 , which is homeomorphic to an @xmath8-sphere .
the @xmath9-sphere @xmath50 is the so - called fourier - plancherel sphere of @xcite .
( see figure [ fig : fps ] and example [ eg : fps ] . )
it is natural to consider how such hilbert space manifolds may be classified geometrically , that is , up to unitary equivalence , and in the examples of section [ sec:4 ] and section [ sec:5.3 ] we distinguish a number of distinct 2-spheres and 3-spheres . in section [ sec:5 ]
we consider the sphere @xmath51 and its hyperbolic variant , the 3-sphere @xmath52 , from the point of view of their unitary automorphism symmetries .
the analysis here exploits the nontrivial foliations induced by the natural order on projections .
it is shown that the group @xmath53 of unitaries which act bijectively on @xmath51 is generated by the set @xmath54 that is , by two 1-parameter unitary groups , the scalar circle group and the fourier - plancherel transform @xmath55 .
in particular , this group contains the dilation semigroup @xmath56 .
a similarly detailed description is given for @xmath57 and this leads to the unitary automorphism group identification as a certain double semidirect product @xmath58 the manifolds @xmath47 , _ beurling manifolds _ in our terminology , may be regarded as smooth in the strict , locally unitary , sense that a neighbourhood of a subspace @xmath12 is given by the local action on @xmath12 of a certain unitary group representation of @xmath16 . furthermore the fourier - plancherel 2-sphere @xmath59 is remarkable in being smooth in this way at all points except the poles @xmath60 .
it of interest then to identify similar compact projection manifolds which are smooth off a finite set . in this regard we see in section [ sec:5.3 ] that this is not generally the case for other polynomial @xmath9-spheres @xmath49 . as we have intimated above our considerations
lie entirely in the realm of operator function theory tied to the hardy space for the line .
however the oscillatory integral methods are expected to be effective in multivariable function spaces and for higher rank settings in noncommutative harmonic analysis .
this should lead to the identification of other closed subspace manifolds with interesting topology and geometry .
we wish to thank alexandru aleman , gordon blower , jean esterle , aristides katavolos , alfonso montes rodriguez and donald sarason for their interest and communications concerning strange limits .
in this section we give some definitions and examples .
let @xmath11 be a separable hilbert space , @xmath61 the set of self - adjoint projections and @xmath62 the set of unitary operators .
we shall routinely identify a closed subspace @xmath12 with its associated orthogonal projection , denoted @xmath13 $ ] .
[ def : cinfsubman ] ( i ) a _ topological subspace manifold _ in @xmath63 of dimension
@xmath8 is a set @xmath64 , considered with the relative strong operator topology , which is locally homeomorphic to @xmath16 .
@xmath65 projection manifold in @xmath63 _ ( or _ @xmath65 subspace manifold _ ) is a topological subspace manifold @xmath3 of dimension @xmath8 together with an atlas of charts @xmath66 ( with open domains and ranges covering @xmath3 ) for which the coordinate functions @xmath67 ( with nonempty domain ) are @xmath65 and such that for each chart @xmath68 with domain @xmath69 there is a dense subspace @xmath70 of _ @xmath65 vectors _ in @xmath11 ; that is , for @xmath71 , the function @xmath72 is @xmath65 on @xmath69 .
in fact we bypass the technicalities of ( ii ) in definition [ def : cinfsubman ] in view of the fact that the smooth subspace manifolds we consider have a stronger locally unitary structure as in the following formal definition .
[ def : locunitsubman ] a _ locally unitary subspace manifold _ of dimension @xmath8 in @xmath63 is a topological subspace manifold @xmath3 such that for each @xmath73 in @xmath3 there is a strong operator topology neighbourhood in @xmath3 of the form @xmath74 : \lambda \in { \mathbb{b}}_n\}\ ] ] where @xmath75 is a strong operator topology continuous representation which is a homeomorphism of @xmath76 into @xmath62 .
it is well - known that such a representation @xmath77 does possess a dense subspace of @xmath65 vectors ; see @xcite , for example .
we shall consider beurling subspace manifolds @xmath3 , given formally in the next definition , together with their complement completions , by which we mean the strong operator topology closure of @xmath7 .
[ def : beurlingsubman ] a _ beurling subspace manifold _ of dimension @xmath8 is a topological subspace manifold @xmath78 of the form @xmath79 : \lambda \in { { \mathbb{r}}}^n\}\ ] ] where @xmath80 is a weak star continuous representation of @xmath16 as unimodular functions , so that @xmath3 is locally homeomorphic to @xmath16 by the single chart @xmath81 $ ] .
if @xmath43 is a non - constant real continuous function on the line then the projections @xmath82 $ ] , for @xmath83 , give a one dimensional topological manifold .
when @xmath84 the closure in @xmath85 adds the subspaces @xmath86 and @xmath5 and the complement completion , @xmath87 say , is topologically a circle .
on the other hand for @xmath88 we shall see that the closure of @xmath89 adds @xmath90 $ ] and @xmath91 $ ] and that the complement completion @xmath92 is a locally unitary @xmath93 subspace manifold diffeomorphic to the circle .
we see later in example [ eg : fps ] that @xmath92 and @xmath87 are , respectively , the equator and a great circle of the fourier - plancherel sphere .
the subspaces @xmath94 , for @xmath95 and @xmath96 , form a subspace manifold @xmath3 which arises in the analysis of the invariant subspaces for the weyl semigroup @xmath97 where @xmath98 is the translation unitary @xmath99 .
indeed it was shown in @xcite that the invariant subspaces of @xmath35 are the spaces @xmath100 for @xmath101 , together with @xmath102 for @xmath103 in @xmath104 and , moreover , that the latter subspaces are in the closure of @xmath3 .
extending the parameter range of @xmath105 to include @xmath106 and taking the complement completion one obtains the fourier - plancherel sphere .
volterra circle _ @xmath107 consists of the subspaces @xmath102 , @xmath108 for @xmath109 and is unitarily equivalent to the great circle @xmath87 via the fourier - plancherel transform .
consider now the three dimensional subspace manifold @xmath110 which is the beurling manifold @xmath33 for the space @xmath111 in section [ sec:4 ] we give a new direct proof that @xmath112 is homeomorphic to a closed 3-ball and that @xmath113 is a topological 3-sphere . in a finite dimensional hilbert space
a locally unitary manifold can be viewed as a submanifold of a grassmannian manifold and it is instructive to note the following simple examples .
let @xmath114 be real , let @xmath115 be the unitary operator @xmath116 in @xmath117 , let @xmath118 be the unit vector @xmath119 and define @xmath120 : t\in { { \mathbb{r}}}\}$ ] .
the function @xmath121 $ ] is periodic if and only if @xmath114 is rational , in which case @xmath122 is a locally unitary manifold which is homeomorphic to a circle . on the other hand
if @xmath114 is irrational then the vectors @xmath123 return infinitely often to any neighbourhood of a particular such vector .
it follows that the relative strong operator topology does not agree with the usual topology of @xmath124 and that @xmath122 is not a locally unitary manifold .
more generally , let @xmath3 be a connected locally unitary @xmath125-dimensional subspace manifold on a hilbert space @xmath11 , with neighbourhood @xmath126 : t\in ( -1,1)\}\ ] ] for the projection @xmath73 .
assume moreover that @xmath127 with @xmath128 a bounded self - adjoint operator .
the derivative of the continuous projection - valued function @xmath129 is the operator @xmath130 where @xmath131 . in particular ,
the unitary equivalence class of the self - adjoint operator @xmath132 is an invariant for @xmath3 for unitary equivalence .
this underlines the fact that unitary equivalence here is a strong form of geometric equivalence for subspace manifolds . in particular , using this derivative invariant , one can soon see that the circle manifolds @xmath133 are unitarily equivalent if and only if the rational numbers @xmath134 coincide .
the beurling subspace manifolds given above might be more precisely specified as euclidean beurling subspace manifolds as there are many other interesting @xmath65 projection manifolds associated with hardy spaces and unimodular functions .
we do not develop this here but we note some fundamental examples . for @xmath135 let @xmath136 : { \underline}{\lambda } = ( \lambda_1 , \dots , \lambda_n ) \in u^n\}\ ] ] where @xmath137 is a blaschke factor inner function with zeros , possibly repeated , at points @xmath138 in the upper half plane @xmath139 . then @xmath140 is a @xmath65 projection manifold in @xmath141 . also , @xmath142 consists of codimension 1 projections and is locally unitary with respect to a representation of the mbius group rather than the euclidean group .
the closure of @xmath142 adds one extra projection , namely @xmath143 $ ] , and is a topological subspace manifold homeomorphic to the @xmath9-sphere , realised as the one point compactification of @xmath139 .
manifolds analogous to these , with finite or cofinite dimensional spaces , may be defined also for bergman hilbert spaces , and , more generally , in the setting of hermitian holomorphic vector bundles associated with operators in the cowen - douglas theory @xcite .
for example , with weighted bergman hilbert spaces in place of @xmath0 one obtains projection manifold realisations of the unit disc with one point compactification closures . that these are unitarily inequivalent was essentially shown in @xcite by the construction of curvature invariants for hermitian holomorphic vector bundles .
this reveals once again the geometric strictness of unitary equivalence .
we note the following alternative curvature free approach to this and in subsequent sections we find , similarly , that we do not need to consider curvature .
however , it goes without saying that it would be interesting and useful to define curvature invariants for general projection manifolds .
let @xmath144 and let @xmath145 be the weighted bergman hilbert space of holomorphic functions in the unit disc that are square integrable with respect to @xmath146 , where @xmath147 is area measure . for @xmath148 in @xmath149 let @xmath150 be the inner function @xmath151 and let @xmath152 be the projection manifold @xmath153 : \lambda \in { { \mathbb{d}}}\}.\ ] ] the range of the complementary projection @xmath154^\bot$ ] is one - dimensional and is spanned by the function @xmath155 .
( see zhu @xcite , for example . ) a standard argument with eigenvectors ( see theorem 3.6 of thomson @xcite for example ) shows that an operator which leaves invariant all these subspaces is necessarily multiplication by the complex conjugate of an @xmath156 function .
thus if @xmath157 is a unitary with @xmath158 then we may deduce that @xmath159 gives an isomorphism between the respective @xmath156 multiplication algebras . composing with a unitary automorphism of @xmath160 which effects the inverse mbius automorphism on the range algebra
we thus obtain a unitary operator @xmath161 between the bergman spaces which intertwines multiplication : @xmath162 .
thus if @xmath163 then @xmath164 . for @xmath165
this is contrary to @xmath161 being isometric .
for @xmath166 the projection manifolds @xmath152 and @xmath167 are unitarily equivalent if and only if @xmath168 .
we now develop methods that will be useful for identifying the closures of various beurling subspace manifolds . [ prop : sot - limit ] let @xmath169 be a union of intervals of @xmath124 .
a sequence of projections @xmath170 converges in the strong operator topology to @xmath171 if and only if \(i ) @xmath172 for every compact interval @xmath173 , and \(ii ) @xmath174 for every compact interval @xmath175 .
necessity is clear .
suppose that ( i ) and ( ii ) hold , and let @xmath27 and @xmath55 be disjoint compact intervals . if either is a subset of @xmath176 then @xmath177 as @xmath178 . on the other hand ,
if @xmath179 then for @xmath180 , @xmath181 as @xmath182 , and we again conclude that @xmath183 as @xmath184 .
since the unit ball of @xmath185 is compact and metrisable in the weak operator topology , there is a positive contraction @xmath186 which is a weak cluster point of @xmath187 , say @xmath188 weakly .
now @xmath189 for disjoint compact intervals @xmath190 which are each contained in either @xmath169 or its complement ; an approximation argument shows that this remains the case whenever @xmath27 and @xmath55 are disjoint bounded measurable sets .
approximation by simple functions yields @xmath191 for every @xmath192 , and similarly @xmath193 for all @xmath194 .
it follows that @xmath195 for some @xmath196 with @xmath197 and @xmath198 .
if @xmath27 is a compact interval which is contained in @xmath169 , then @xmath199 , which forces @xmath200 to be equal to @xmath125 almost everywhere on @xmath27 .
so @xmath201 .
thus the weak limit of @xmath202 is a projection , @xmath171 , from which it follows that @xmath203 as well .
thus every subsequence of @xmath204 has a subsubsequence whose limit in the strong operator topology is @xmath171 .
since the unit ball of @xmath185 is metrisable in this topology , this shows that @xmath205 strongly .
we now determine conditions under which we can identify the strong operator topology convergence @xmath205 for a sequence of beurling projections @xmath206 $ ] , where @xmath207 and @xmath169 depends on the derivative of @xmath208 .
a function @xmath209 is _ admissible _ if the following subsets of @xmath124 are discrete ( that is , they have no accumulation points ) : 1 .
the set @xmath210 of points at which @xmath43 fails to be twice continuously differentiable ; 2 .
@xmath211 ; and 3 .
the set @xmath212 consisting of the points at which @xmath213 is not locally constant . for example
, non - constant rational functions and the trigonometric functions are easily seen to be admissible , as is the map @xmath214 .
let @xmath215 be the unitary fourier - plancherel transform defined by @xmath216 the hardy space @xmath0 is equal to @xmath217 , so @xmath218={\operatorname{ad}}(m_{e^{ik\psi}}f^*)m_{\chi_{(0,\infty)}}\ ] ] where @xmath219 for a unitary operator @xmath157 and @xmath220 , and @xmath221 is the multiplication operator on @xmath5 corresponding to a function @xmath222 , which is unitary if @xmath200 is unimodular .
in particular , if @xmath223 $ ] then @xmath224 it is therefore of interest to find an estimate for the fourier - plancherel transform @xmath225 on the positive half - line .
this is done in the next lemma for admissible functions @xmath43 , using an integration by parts in the spirit of @xcite , chapter viii .
[ lem : osc ] let @xmath43 be admissible and let @xmath226 be a compact interval of positive length such that @xmath227 and @xmath228 .
let @xmath229 be the function @xmath230 where @xmath231 and let @xmath232 .
then for @xmath233 and almost every @xmath234 , @xmath235 . since @xmath212 is discrete , finitely many points in @xmath212 lie in the interior of @xmath236 , say @xmath237 .
we also write @xmath238 for the boundary points of @xmath236 so that @xmath239 $ ] .
since @xmath227 , the function @xmath43 is twice continuously differentiable on @xmath236 and for @xmath234 , @xmath240 & = \big| \int_s \frac { i}{k(z+\psi'(x))}\frac{d}{dx}\big(e^{-ik(xz+\psi(x))}\big)\,dx\big| \displaybreak[0]\\[6pt ] & = \frac1{k}\big| \big [ \frac{e^{-ik(xz+\psi(x ) ) } } { z+\psi'(x ) } \big]_{\lambda_1}^{\lambda_{n } } - \int_s \frac{d}{dx}\big(\frac1{z+\psi'(x)}\big ) e^{-ik(xz+\psi(x))}\,dx \big|\\[6pt ] & \le \frac1{k } \big ( \delta(z)+ \int_s \big| \frac d{dx}\big(\frac{1}{z+\psi'(x)}\big)\big| \,dx \big ) . \end{aligned}\ ] ] let @xmath241 . for @xmath242 ,
the quantity @xmath243 is constant , say @xmath244 .
so @xmath245_{\lambda_j}^{\lambda_{j+1}}\\ & \le ( n-1)\delta(z ) .
\end{aligned}\ ] ] the result follows .
[ thm : psi ] let @xmath246 be a sequence of positive numbers with @xmath207 as @xmath184 , let @xmath43 be admissible and let @xmath206 $ ] for @xmath247 . then @xmath248 $ ] as @xmath184 , where @xmath249
let @xmath250 and @xmath251 .
let @xmath236 be a compact subinterval of @xmath252 of positive length which does not intersect @xmath210 .
we will show that @xmath253 . since @xmath254 , @xmath255 where we have made the change of variables @xmath256 .
we apply lemma [ lem : osc ] : @xmath257 as @xmath184 , where @xmath258 and @xmath259 are defined as in lemma [ lem : osc ] .
let @xmath260 , let @xmath261 $ ] and let @xmath186 be the conjugation operator @xmath262 for @xmath263 . observe that @xmath264 . applying the argument of the previous paragraph to @xmath200 instead of @xmath43 shows that @xmath265 whenever @xmath17 is a compact subinterval of @xmath266 .
hence @xmath267 .
we have shown that @xmath253 when @xmath236 is a compact subinterval of @xmath268 and @xmath269 when @xmath17 is a compact subinterval of @xmath266 .
since @xmath210 and @xmath270 are discrete sets , the same conclusion holds for @xmath236 and @xmath17 arbitrary compact subintervals of @xmath271 and @xmath272 respectively . now proposition [ prop : sot - limit ] completes the proof .
the theorem enables us to compute immediately a wide variety of strange limits @xmath273 , so - called because while every nonzero function in the range of @xmath170 has full support , those for @xmath73 itself are supported in a proper measurable set .
for example , @xmath274 { \stackrel{\textsc{sot}}{\to}}[{{{l^{2}({{\mathbb{r}}}_+)}}}]$ ] as @xmath182 , and @xmath275 { \stackrel{\textsc{sot}}{\to}}{{l^{2}([\alpha,\beta])}}$ ] if the roots @xmath276 of the equation @xmath277 are real , and has limit @xmath278 otherwise .
we also remark that when @xmath43 is admissible , @xmath279= \big(\!\operatorname*{{\text{\sc sot}}-lim}_{n\to\infty}[e^{-in\psi}{{{h^{2}({{\mathbb{r}}})}}}]\big)^\perp.\ ] ] while admissible functions are adequate for our applications one can perhaps partially relax this constraint .
however , we are unaware of a general formula for the limit when @xmath43 is real , measurable and locally bounded . the next corollary will play a part in the proof of theorem [ thm : sot - closures ] .
[ cor : psi_n ] let @xmath246 , @xmath43 and @xmath280 be as above , let @xmath281 be a sequence of admissible functions and let @xmath282 $ ] .
suppose that the set @xmath283 is discrete .
let @xmath284,\ \alpha<\beta\}\ ] ] and suppose that the quantity @xmath285 is finite for every @xmath286 . if @xmath287 uniformly on @xmath236 for every interval @xmath286 then @xmath288 $ ] strongly as @xmath184 .
choose a compact subinterval @xmath289 .
let @xmath290 pick @xmath8 sufficiently large that @xmath291 ; then @xmath292 .
writing @xmath293 lemma [ lem : osc ] applies as before to show that @xmath294 since @xmath295 , this shows that @xmath253 .
the remainder of the proof proceeds as in theorem [ thm : psi ] .
beurling s characterisation of invariant subspaces for the bilateral shift operator when transferred to the setting @xmath5 amounts to the identification of @xmath296 with the disjoint union @xmath297 here @xmath298 denotes the lattice of closed invariant subspaces for a family of operators @xmath299 , and @xmath300 is the set of invariant subspaces @xmath301 which are purely invariant in the sense that the intersection of the subspaces @xmath302 for @xmath303 is trivial .
we now use the methods of this section to show that @xmath304 .
this seems to be a previously unobserved feature in the classical setting which may well have a wider manifestation .
however , the authors are unaware of any general results of this nature . [
lem : noacc ] let @xmath305 denote lebesgue measure on @xmath124 . if @xmath306 is measurable and @xmath307 then there is a countable disjoint union of open intervals @xmath308 such that @xmath309 is discrete and @xmath310 .
fix @xmath311 and write @xmath312 and @xmath313 . using elementary properties of lebesgue measure , we can find a set @xmath314 such that @xmath315 are disjoint open subintervals of @xmath316 and @xmath317 .
pick @xmath318 so that @xmath319 and let @xmath320 .
now @xmath321 so @xmath322 .
repeat for each @xmath311 and let @xmath323 .
then @xmath324 and @xmath309 is discrete , since @xmath325 $ ] is finite for each @xmath8 .
[ thm : purelimits ] if @xmath169 is any measurable subset of @xmath124 then there is a sequence of projections @xmath282 $ ] where @xmath326 and each @xmath281 is a real - valued function such that @xmath327 .
let @xmath307 .
by lemma [ lem : noacc ] , we can find a countable disjoint union of open intervals @xmath328 such that @xmath329 is discrete and @xmath330 .
the function @xmath331 satisfies @xmath332 and @xmath333
, so @xmath334 is admissible .
let @xmath335 $ ] . by theorem [ thm : psi ] , @xmath336 , and we also have @xmath337 strongly as @xmath338 .
let @xmath339 be a metric inducing the strong operator topology on the unit ball of @xmath185 and let @xmath247 .
choose @xmath340 such that @xmath341 and then choose @xmath342 so that @xmath343 satisfies @xmath344 .
now @xmath345 so @xmath205 strongly as @xmath184 .
we now obtain sufficient conditions under which beurling subspace manifolds @xmath47 have closures , in the strong operator topology , which are compact . using this
we construct various @xmath8-spheres and @xmath8-balls in @xmath85 . at the end of the section
we pose some further lines of enquiry .
let @xmath346 be an @xmath8-tuple of functions @xmath347 .
we write @xmath348 for @xmath349 , and @xmath350 the @xmath8-tuple @xmath351 is _ admissible _ if 1 .
the set @xmath352 is linearly independent over @xmath124 ; 2 .
every nonzero function in @xmath353 is admissible ; and 3 .
@xmath354 for each compact set @xmath355 . we will also write @xmath356 and remark that this is equal to @xmath357 and is plainly discrete . given an admissible @xmath8-tuple @xmath351 , let @xmath358 be the map @xmath359 $ ] .
observe that @xmath360 is strongly continuous , since if @xmath361 in @xmath16 then @xmath362 uniformly on compact subsets of @xmath363 and so @xmath364{\stackrel{\textsc{sot}}{\to}}{\operatorname{ad}}(m_{e^{i\langle f,\lambda\rangle}})[{{{h^{2}({{\mathbb{r}}})}}}]=\theta(\lambda)\ ] ] as @xmath365 .
we write @xmath366 .
we will shortly see that @xmath367 is a beurling subspace manifold .
[ thm : sot - closures ] given an admissible @xmath8-tuple @xmath351 , the closure of the range of @xmath360 in the strong operator topology is @xmath368 : \psi\in{{\mathcal{s}}}_f\setminus\{0\}\big\}.\ ] ] let @xmath369 be a sequence in @xmath16 and let @xmath370 be the corresponding sequence of projections .
passing to a subsequence , we may assume that @xmath369 converges to a vector @xmath371 as @xmath365 .
if @xmath372 actually lies in @xmath16 then @xmath373 by the continuity of @xmath360 . otherwise , if @xmath374 then @xmath375 as @xmath365
let @xmath376 . passing to a subsequence
, we may assume that @xmath377 for some @xmath378 independent of @xmath318 , and that @xmath379 for some @xmath380^n$ ] .
since @xmath381 , this limit @xmath382 is nonzero .
now @xmath383 $ ] where @xmath384 , and if @xmath385 then @xmath386 uniformly on compact subsets of @xmath363 . by corollary [ cor : psi_n ] , @xmath387 $ ] .
conversely , if @xmath388 then by theorem [ thm : psi ] , @xmath389=\operatorname*{{\text{\sc sot}}-lim}_{n\to\infty}[e^{in\psi}{{{h^{2}({{\mathbb{r}}})}}}]\in\overline{{{\mathcal{m}}}({{\mathcal{s}}}_f)}. \qedhere\ ] ] [ defn : sim ] given an admissible @xmath8-tuple @xmath351 , let @xmath390 be the equivalence relation defined on @xmath391 by @xmath392 if @xmath393 or @xmath394 here @xmath395 and @xmath305 is lebesgue measure on @xmath124 .
we write @xmath396 for the corresponding topological quotient space .
[ prop : compact - quotients ] the topological space @xmath396 is homeomorphic to @xmath397 . in particular , @xmath112 is compact .
let @xmath360 be the continuous map @xmath398 defined above .
observe that @xmath360 is injective : for @xmath399 if and only if @xmath400 which implies that the function @xmath401 is constant modulo @xmath402 almost everywhere .
such a function can not be admissible , so it is zero ; by linear independence , @xmath393 .
let @xmath403 be a homeomorphism of the form @xmath404 where @xmath405 is a homeomorphism .
consider the injective continuous map @xmath406 .
we extend this to @xmath391 by defining @xmath407 for @xmath408 ; this limit exists by theorem [ thm : psi ] .
the extended map is also continuous by corollary [ cor : psi_n ] and surjective by theorem [ thm : sot - closures ] . since @xmath409 if and only if @xmath392 , it follows that @xmath200 induces a homeomorphism from the compact space @xmath396 to the hausdorff space @xmath112 .
this proof shows that @xmath410 is a homeomorphism when @xmath351 is admissible , and so @xmath367 is indeed a beurling subspace manifold . determining the precise nature of the quotient space @xmath396 seems difficult in general .
however , the following shows that there are no surprises when @xmath411 .
let @xmath412 be any set of closed , pairwise disjoint intervals , each contained in @xmath413 .
then there exists a continuous increasing surjection @xmath414\to[0,1]$ ] such that @xmath415 if and only if @xmath416 for some @xmath417 .
we may assume without loss of generality that each @xmath418 is of the form @xmath419 $ ] with @xmath420 .
observe that @xmath412 must be countable since for each @xmath421 , the set @xmath422\in{{\mathcal{i}}}:b - a > n^{-1}\}$ ] is finite .
enumerate @xmath412 in decreasing size , so that @xmath423 where @xmath424 for @xmath421 .
assign values @xmath425 on @xmath426 to create a partially defined increasing function @xmath427 on @xmath428 as follows .
set @xmath429 and @xmath430 .
then set @xmath431 and @xmath432 or @xmath433 , according to whether the new interval @xmath434 is to the left or the right of @xmath435 .
continue `` interpolating dyadically '' in this way , so that @xmath436 is chosen as the mean of @xmath437 and @xmath438 where @xmath439 ( respectively , @xmath440 ) is the point of @xmath441 immediately to the left ( respectively , right ) of @xmath442 .
let @xmath301 be the closure in @xmath443 $ ] of @xmath444 .
let @xmath157 be the complement of @xmath301 , an open set in @xmath443 $ ] and thus a union of open intervals , which we call `` gaps '' .
we call a gap `` good '' if it is of the form @xmath445 where @xmath446 $ ] and @xmath447 $ ] are intervals in @xmath412 , and `` bad '' otherwise .
we can extend @xmath427 to a good gap @xmath448 by linear interpolation between @xmath425 and @xmath449 . on the other hand , a bad gap @xmath450 must have at least one of its end points a limit of endpoints of intervals @xmath417 , so @xmath427 can be extended to the closure @xmath451 as a non - decreasing function in a unique way , namely by being constant on @xmath451 .
so we get a continuous increasing function @xmath427 defined on @xmath443 $ ] .
we now correct the constancy on the intervals @xmath451 by forming @xmath452 where @xmath453 and @xmath454 is the continuous map on @xmath443 $ ] which takes the value @xmath278 to the left of @xmath450 , the value @xmath455 to the right and is linear on @xmath450 .
( none of these @xmath454 spoil constancy on the @xmath426 . )
continuity of @xmath456 is clear .
finally , let @xmath457 with @xmath458 so that @xmath459 .
[ lem : circle - quotients ] let @xmath412 be a set of closed , pairwise disjoint proper arcs in @xmath41
. then @xmath460 is homeomorphic to @xmath461 .
we may assume that @xmath462 does not belong to any set in @xmath412 .
transfer @xmath412 to @xmath443 $ ] in the obvious manner and apply the previous lemma to obtain an increasing continuous surjection @xmath414\to [ 0,1]$ ] such that @xmath415 if and only if @xmath463 for some @xmath417 .
the map @xmath464 given by @xmath465 for @xmath466 $ ] and @xmath467 induces a continuous bijection @xmath468 , which is a homeomorphism since @xmath460 is compact . [
prop : admissible - pairs ] for every admissible pair @xmath469 , the closure in the strong operator topology of the beurling subspace manifold @xmath470 is homeomorphic to @xmath461 .
let @xmath390 be the equivalence relation on @xmath461 of definition [ defn : sim ] , with equivalence classes @xmath471:\lambda\in{{\overline{{{\mathbb{d}}}}}}\}$ ] .
observe that @xmath472 and @xmath392 if and only if @xmath473 .
also , the equivalence classes are connected : if @xmath392 then the shorter of the two arcs joining @xmath372 to @xmath382 is contained in @xmath474=[\mu]$ ] .
we show that the equivalence classes are also closed .
if there is a non - trivial equivalence class , then we can make a different choice of @xmath351 and @xmath475 without changing the set @xmath476 to arrange that @xmath477 , and so also that @xmath478 .
any remaining equivalence classes are of the form @xmath474 $ ] where @xmath479 with @xmath480 ; we may assume that @xmath481 and @xmath482 .
let @xmath483 and let @xmath484 .
then @xmath474=[\lambda]_+\cap[\lambda]_-$ ] where @xmath485_+ & = \big\ { \mu : \{f'>0\}\cap(\ { h>\alpha \ } { \mathbin{\text{\footnotesize$\bigtriangleup$}}}\ { h>-\mu_1/\mu_2\ } ) \text { is null } \}\text { and } \\
[ \lambda]_- & = \big\ { \mu : \{f'<0\}\cap(\ { h<\alpha \ } { \mathbin{\text{\footnotesize$\bigtriangleup$}}}\ { h<-\mu_1/\mu_2\ } ) \text { is null } \}.
\end{aligned}\ ] ] here and below we employ abbreviations of the form @xmath486 to mean the set @xmath487 where @xmath488 and @xmath73 is a predicate depending on a real parameter . note that for any constant @xmath318 , the set @xmath489 is null ; for if not , then @xmath490 takes the value @xmath278 on a non - null set , so @xmath491 is not admissible , whence @xmath492 ; but @xmath351 and @xmath475 are independent . for @xmath493 ,
let @xmath494 be the unit vector @xmath495 .
each @xmath474_+$ ] is clearly connected .
it is also closed , for if @xmath496 and @xmath497_+$ ] then @xmath498 is equal to the union of the null sets @xmath499 and @xmath500 , so is null ; hence @xmath501 , and by a similar argument @xmath502 , lie in @xmath474_+$ ] .
the class @xmath474_-$ ] is closed for the same reasons , and hence each class @xmath474 $ ] is a closed subarc of @xmath41 .
now proposition [ prop : compact - quotients ] and lemma [ lem : circle - quotients ] complete the proof . given an admissible @xmath8-tuple @xmath351 , let @xmath503 denote the _ complement completion _ of @xmath367 ; that is , the closure of @xmath504 in the strong operator topology .
it is not hard to see that @xmath367 and @xmath505 are disjoint , and the boundaries of these sets are equal by theorem [ thm : sot - closures ] .
from this it follows that provided @xmath506 is homeomorphic to @xmath507 , the set @xmath503 is homeomorphic to @xmath508 since it is homeomorphic to the union of two copies of @xmath507 joined at their boundaries .
[ eg : fps ] the _ fourier - plancherel sphere _ is the set of projections @xmath509 by proposition [ prop : admissible - pairs ] or the direct arguments of @xcite , we see that @xmath506 is homeomorphic to @xmath510 , and as observed in @xcite , the order structure of @xmath59 is that of a union of continuous nests which meet only at @xmath278 and @xmath511 and @xmath59 is homeomorphic to a 2-sphere on which the fourier - plancherel transform @xmath55 acts as a quarter - rotation . in particular , @xmath512 is a locally unitary subspace manifold : we clearly have a locally unitary structure on @xmath513 , and we can use @xmath55 to transfer this structure to the remaining subspaces .
[ eg:2dim ] let @xmath351 be the admissible pair @xmath514 .
an easy extension of @xcite , lemma 5.1 shows that every projection @xmath515 lies in a continuum of non - commuting continuous nests which intersect only in @xmath516 .
a simple calculation reveals that the equivalence relation @xmath390 has only two non - trivial equivalence classes and these are antipodal closed quarter - circles , so @xmath517 is homeomorphic to @xmath518 and the boundary projections are @xmath519,\ a\in[0,\infty]\}.\ ] ] now although @xmath503 is homeomorphic to a 2-sphere , @xmath520 is not locally unitary . for
if there were a strong operator topology neighbourhood of @xmath521 $ ] of the form @xmath522:\lambda\in{\mathbb{b}}^2\}\ ] ] for some unitary - valued representation @xmath523 of @xmath524 , then @xmath525 would intersect @xmath367 and so contain a projection @xmath526 $ ] for some point @xmath527 such that every neighbourhood of @xmath73 contains two non - commuting projections which are comparable with @xmath73 .
applying @xmath528 , we see that @xmath525 must contain two non - commuting projections which are comparable with @xmath73 .
however , all the projections in @xmath503 which are comparable with @xmath73 all commute since they are of the form @xmath529 $ ] for some @xmath530 . if we take @xmath531 then @xmath532 and it is not hard to check that the corresponding equivalence relation @xmath390 on @xmath533 has two non - trivial equivalence classes : @xmath534 which correspond to @xmath278 and @xmath511 respectively when we identify the quotient space @xmath535 with @xmath506 .
these equivalence classes are closed and so @xmath506 is homeomorphic to @xmath533 .
if @xmath73 is a projection in @xmath367 then we claim that the projections @xmath536 $ ] for @xmath537 are the only non - trivial elements of @xmath506 which are comparable with @xmath73 . to see this ,
recall that no proper subspace of the form @xmath529 $ ] is comparable with @xmath73 by the f. & m. riesz theorem , and if @xmath538 and @xmath539 then @xmath540 and so @xmath541 is a nonzero continuous inner function .
this must be of the form @xmath542 for a unimodular constant @xmath543 and @xmath544 ( see @xcite ) , which verifies the claim .
on the other hand , the boundary of @xmath506 consists of the projections @xmath529 $ ] where @xmath27 is either an interval or the complement of an interval . as in the previous example
, it follows that the topological 3-sphere @xmath503 can not be locally unitary away from @xmath545 since the local order structure changes on the boundary of @xmath506 .
[ eg : hyp - sphere ] let @xmath546 .
then @xmath547 so if @xmath548 then @xmath549 .
the equivalence relation for @xmath351 on @xmath533 is therefore identical to the relation considered in the previous example , and @xmath506 is again homeomorphic to @xmath533 .
the order structure differs however , since @xmath506 contains the set @xmath550 from example [ eg:2dim ] .
we call the set @xmath551 the _ hyperbolic sphere_. this was first considered in @xcite , section 7
. we remark that as in example [ eg:2dim ] , @xmath552 can not be locally unitary .
we can now easily establish the compactness of the `` extended hyperbolic lattice '' @xmath553 considered in @xcite ; this fact was alluded to but not proven there . for @xmath554 , let @xmath555 be the two - valued function taking the value @xmath125 on @xmath556 and @xmath543 on @xmath557 .
then @xmath553 may be succinctly described as the set of projections @xmath558 now @xmath559 is homeomorphic to the compact space @xmath560 and @xmath553 is a continuous image of @xmath561 which is compact , so @xmath553 is also compact .
the equivalence relation @xmath390 need not have a finite number of nontrivial equivalence classes . for example , consider the admissible triple @xmath562 so that if @xmath548 then @xmath563 .
a simple analysis of cases reveals that the nontrivial equivalence classes for @xmath390 in the upper hemisphere @xmath564 are @xmath565 where @xmath566 .
the class @xmath567 corresponds to the projection @xmath568\in\overline{{{\mathcal{m}}}({{\mathcal{s}}}_f)}$ ] .
since points of the form @xmath569 for @xmath570 form a straight line in the plane @xmath571 it follows that @xmath567 is the geodesic on @xmath572 joining @xmath573 to @xmath574 where @xmath575 . in the @xmath576 hemisphere
the nontrivial equivalence classes are the sets @xmath577 which correspond to @xmath578 $ ] .
it is easy to see that the quotient space @xmath535 is homeomorphic to @xmath533 ; indeed , we may choose a homeomorphism which contracts each geodesic @xmath567 to the point @xmath574 and extend this to all of @xmath533 is a straightforward manner .
we again conclude that @xmath506 is homeomorphic to @xmath533 .
we do not know if @xmath506 , or equivalently @xmath396 , is homeomorphic to @xmath507 for every admissible @xmath8-tuple @xmath351 .
it is natural to try to emulate the argument of proposition [ prop : admissible - pairs ] , and it is not hard to show that the equivalence classes @xmath474 $ ] of @xmath390 satisfy the following conditions : 1 .
@xmath474=\{\lambda\}$ ] if @xmath579 ; 2 .
@xmath580=-[\lambda]$ ] and @xmath580\cap[\lambda]=\emptyset$ ] for @xmath581 ; 3 .
if @xmath582 then @xmath583 for every @xmath584 on the geodesic in @xmath585 joining @xmath372 to @xmath382 ; and 4 . if @xmath586 where @xmath587 and @xmath588 are convergent sequences in @xmath585 , then @xmath582 .
however , for @xmath589 we have been unable to identify the quotient space @xmath396 for such an equivalence relation .
the order structure of the 2-sphere @xmath59 and the 3-sphere @xmath559 can be viewed as providing an inherent foliation .
we exploit this structure in the next section in the determination of their unitary automorphism groups . on the other hand
we see in lemma 5.8 that the 2-spheres determined by monomial pairs @xmath590 , for @xmath591 have a trivial order structure supported in the common boundary of @xmath3 and @xmath592 .
the fourier - plancherel sphere seems to be a particularly distinguished example amongst these 2-spheres .
furthermore its equator , @xmath92 yields an interesting compact @xmath125-dimensional subspace manifold which is locally unitary and is probably not ( periodically ) unitary . it would be interesting to determine other ( unitarily inequivalent ) subspace manifolds of this form
given a set of projections @xmath593 , the unitary automorphism group of @xmath594 is @xmath595 as usual , @xmath596 is the map @xmath597 . in this section
we compute the unitary automorphism groups of the fourier - plancherel sphere , the hyperbolic sphere and the extended hyperbolic lattice .
these projection manifolds inherit a relatively rich order structure from @xmath85 which we are able to exploit .
in contrast we show in section [ sec:5.3 ] that many other polynomial 2-spheres are essentially rigid .
further operator algebra related to the two main examples can be found in @xcite , @xcite , @xcite .
recall the definition of the fourier - plancherel sphere @xmath59 from example [ eg : fps ] .
the following notation from @xcite is convenient : @xmath598 we also introduce notation for some nests contained in @xmath59 : the _ analytic nest _ is @xmath599:\lambda\in{{\mathbb{r}}}\}\cup\{0,i\}$ ] and the _ volterra nest _ is @xmath600:t\in{{\mathbb{r}}}\}\cup\{0,i\}.\ ] ]
if we write @xmath601 then @xmath602 and the order structure that @xmath59 inherits from @xmath85 is such that if @xmath603 with @xmath604 then @xmath605 and @xmath606 unless @xmath607 for some nest @xmath525 in this union .
it is easy to see that @xmath608 , @xmath609 and @xmath610 all lie in @xmath611 for @xmath612 , as does the fourier - plancherel transform @xmath55 since @xmath613 by @xcite , theorem 7.1 .
we first show that @xmath610 may be expressed solely in terms of @xmath614 and @xmath55 .
let us write @xmath618 for the operation of convolution with a function @xmath619 , defined on the schwartz space @xmath620 ; that is , @xmath621 for @xmath622 , let @xmath623 denote the square root of @xmath624 with non - negative real part .
let @xmath625 be the alternate fourier transform defined on @xmath620 by @xmath626 observe that @xmath627 and that @xmath628 . in section xi.1 of
@xcite it is shown that @xmath629 or , writing @xmath630 and rearranging , @xmath631 observe that @xmath632 for @xmath633 .
hence for @xmath634 , @xmath95 and @xmath635 , @xmath636 equating these expressions for @xmath637 and using the density of @xmath620 in @xmath5 gives @xmath638 now @xmath639 and @xmath640 commutes with @xmath641 for any @xmath642 , since @xmath643 is even and @xmath644 .
so @xmath645 using this and setting @xmath646 completes the proof .
let @xmath648 and let @xmath87 , @xmath107 and @xmath649 denote the `` great circles '' @xmath650 the map @xmath159 preserves orthogonality and the order structure on @xmath59 , so it must permute these great circles .
if @xmath651 , then @xmath652 fixes @xmath87 ; if @xmath653 for some @xmath654 , then @xmath655 fixes @xmath87 .
so we may assume that @xmath656 . if @xmath657 , then @xmath658 for some @xmath659 .
if @xmath660 then by ( [ eq : fourier ] ) , @xmath661 and @xmath662 , so @xmath663 satisfies @xmath664 and @xmath665 .
so we may assume that @xmath666 and @xmath656 .
suppose that case ( iii ) holds .
we claim that @xmath672 for some @xmath660 .
to see this , let @xmath673 be the set of nests @xmath674 so that @xmath59 is the union of all nests in @xmath673 .
since @xmath157 is unitary , it maps nests onto nests and so induces a bijection of @xmath673 .
let @xmath92 be the `` equator '' of @xmath59 , @xmath675 here @xmath676 is the set of complex conjugates of functions in @xmath0 , which is equal to @xmath677 @xcite .
the set @xmath92 contains exactly one subspace from each nest in @xmath673 , so the action @xmath678 , @xmath679 of @xmath159 on @xmath92 determines the action of @xmath159 on @xmath673 .
moreover , @xmath159 is a homeomorphism between @xmath92 and @xmath680 , and @xmath92 is itself homeomorphic to the circle @xmath41 .
let us give @xmath673 the topology induced by the topology on @xmath92 .
the bijective action of @xmath159 on @xmath673 is then a homeomorphism .
it follows that the closed connected set @xmath681 = \bigcup_{s\geq 0}{{\mathcal{n}}}_s\cup \vnest\ ] ] must be mapped by @xmath159 onto @xmath682=\bigcup_{s\geq 0}{{\mathcal{n}}}_s^\perp\cup \vnest^\perp\cup \bigcup_{s\in { { \mathbb{r}}}}{{\mathcal{n}}}_s\cup \vnest\\[6pt ] \text{or}\quad [ \vnest,\anest^\perp ] = \vnest\cup \bigcup_{s\leq 0}{{\mathcal{n}}}_s^\perp .
\end{gathered}\ ] ] if @xmath683=[\anest^\perp,\vnest]$ ] then there is some @xmath660 such that @xmath684 . since @xmath685 and @xmath157 is unitary , @xmath686 which is impossible by the f. & m. riesz theorem .
so @xmath683=[\vnest,\anest^\perp]$ ] and so @xmath672 for some @xmath660 .
since @xmath685 , it follows from @xcite , chapter 17 that there exist a unimodular function @xmath687 and an order - preserving almost everywhere differentiable bijection @xmath688 such that @xmath689 where @xmath690 is the unitary composition operator corresponding to @xmath475 .
thus @xmath691 for some @xmath692 .
moreover , @xmath693 , so @xmath694 for some @xmath695 .
since @xmath696 for @xmath697 , @xmath698 taking orthogonal complements , we see that @xmath699 where @xmath700 is the unimodular function @xmath701 so @xmath702 must be constant almost everywhere . but
@xmath703 as @xmath704 , so this is impossible .
so we are reduced to case ( i ) : @xmath159 fixes both the analytic nest and the volterra nest , and so is a unitary automorphism of @xmath705 , the fourier binest algebra . by @xcite , lemma 4.1 , @xmath706 for some @xmath707 .
now apply lemma [ lem : lang - vt ] .
it can be shown that , modulo scalars , this automorphism group is isomorphic to the semidirect product @xmath708 . the isomorphism is implemented by the map sending @xmath709 to @xmath710 , @xmath711 and @xmath712 respectively .
we refer the reader to @xcite for the details .
recall the definitions of the hyperbolic sphere @xmath559 and the extended hyperbolic lattice @xmath714 from example [ eg : hyp - sphere ] . for @xmath715 ,
let @xmath716 and for @xmath717 let @xmath718 a typical projection in @xmath719 is @xmath720 { \quad\text{where}\quad } ( s,\lambda,\mu)\in{{\mathbb{r}}}^3.\ ] ] if @xmath721 then it is shown in @xcite that @xmath722 for @xmath717 .
we further define operators @xmath723 these are the unitary composition operators corresponding to the symmetries @xmath724 and @xmath725 , respectively .
the linear span of the set of functions @xmath726 for @xmath727 ( or @xmath728 ) is dense in @xmath0 ( or in @xmath676 , respectively ) . applying @xmath729 and @xmath730 to these sets
reveals that @xmath731 and @xmath732 .
it is easy to see that all of these operators are unitary automorphisms of @xmath553 , and if we fix @xmath733 then we obtain unitary automorphisms of @xmath559 .
we will show that in each case , these operators generate the whole unitary automorphism group .
[ lem : gamma ] let @xmath734 be a conformal automorphism of the upper half plane @xmath139 . for each nonzero @xmath570 ,
the subspace @xmath735 has zero intersection with each subspace in @xmath719 unless @xmath734 is either of the form @xmath736 for some @xmath737 or @xmath738 for some @xmath739 .
suppose that @xmath740 .
let @xmath741 be a nonzero function in this intersection , where @xmath742 .
multiplying this equation by @xmath743 if @xmath744 and by @xmath745 if @xmath746 and writing @xmath747 gives @xmath748 for nonzero functions @xmath749 .
observe that @xmath543 takes at most four values , since @xmath750 applying the f. & m. riesz theorem to the function @xmath751 reveals that @xmath543 must be constant almost everywhere ; in particular , since @xmath752 , no three of these intersections can have nonzero lebesgue measure . since @xmath734 induces a conformal automorphism of the upper half plane , either @xmath753 with @xmath737 and @xmath754 or @xmath755 with @xmath756 and @xmath739 . applying the condition in the previous paragraph forces @xmath736 with @xmath737 or @xmath757 with @xmath739 .
[ thm : autgroups ] ( i ) the unitary automorphism group of @xmath553 is equal to the union @xmath758 where @xmath759 ( ii ) the unitary automorphism group of @xmath559 is equal to the union @xmath760 where @xmath761 \(i ) we exploit the order structure of @xmath553 , given in proposition 5.2 of @xcite .
suppose that @xmath157 lies in @xmath762 , the unitary automorphism group of @xmath553 .
let us write @xmath3 for the set @xmath763:s,\lambda,\mu\in{{\mathbb{r}}}\}\ ] ] and let @xmath764 be the topological boundary of @xmath3 , which by theorem [ thm : sot - closures ] is the set of projections in @xmath553 of the form @xmath529 $ ] . observe first that @xmath764 must be mapped onto itself by @xmath159 .
this will follow if we can show that the set @xmath765 may be intrinsically described as the union of all non - commutative sublattices of @xmath553 which are order - isomorphic to the slice @xmath766:\lambda,\mu\in{{\mathbb{r}}}\}$ ] and whose closure contains @xmath545 . writing @xmath767:\lambda,\mu\in{{\mathbb{r}}}\}\ ] ] for @xmath768 , observe that each of the slices @xmath769 , @xmath770 has this property .
hence this union contains @xmath765 . on the other hand ,
suppose that @xmath771 is such a non - commutative sublattice and that @xmath772 . since @xmath773 there are two continuous nests @xmath774 contained in @xmath775 which do not commute with one another such that @xmath776 .
now @xmath73 is of the form @xmath777 $ ] and all the non - trivial projections @xmath778 in @xmath553 which satisfy @xmath779\le r$ ] are all contained in @xmath764 by the f. & m. riesz theorem , so @xmath780 . but all projections in @xmath764 commute , so we obtain a contradiction and @xmath781 must be empty . hence @xmath782 .
observe that @xmath783 and that the terms in this union are the components of @xmath765 , which @xmath159 must therefore permute .
if @xmath784 , then we may replace @xmath157 with @xmath785 to ensure that @xmath786 , and then replacing @xmath157 by @xmath787 for suitable @xmath788 we can arrange that @xmath789 , and so also that @xmath790 .
since @xmath782 , @xmath159 maps projections in @xmath791 to projections in @xmath791 and so induces an automorphism of @xmath792 .
von neumann s theorem of @xcite shows that this is necessarily induced by a borel isomorphism @xmath734 .
( see also nordgren @xcite ) .
since @xmath792 is maximal abelian it follows readily that @xmath793 for some unimodular function @xmath200 where @xmath794 is the unitary composition operator for @xmath734 .
moreover , @xmath157 induces an automorphism of @xmath795 , since if @xmath796 then @xmath797 and so @xmath798 it follows from the beurling - lax theorem @xcite that @xmath799 . the same argument applied to @xmath800 shows that the map @xmath801 is indeed an automorphism of @xmath795 .
hence @xmath734 induces a conformal automorphism of the upper half plane .
fix @xmath659 .
since @xmath790 , the subspace @xmath802 must lie in @xmath3 .
by lemma [ lem : gamma ] , either @xmath803 or @xmath804 for some @xmath615 .
hence @xmath805 .
multiplying by @xmath806 reduces to the case @xmath807 . since @xmath808 ,
we have @xmath809 and so @xmath200 is constant almost everywhere .
\(ii ) following the first paragraph of the proof of ( i ) with @xmath559 in place of @xmath553 , we may assume that @xmath810 satisfies @xmath782 .
now @xmath811 has two components , which @xmath159 must permute .
we may again assume that @xmath790 by multiplying by @xmath730 if necessary .
the remainder of the proof proceeds as above .
we see that @xmath812 is isomorphic to the double semidirect product @xmath813 the map sending @xmath814 to @xmath815 , @xmath816 and @xmath817 respectively for @xmath818 is a homomorphism onto @xmath812 with kernel @xmath819 . finally , we consider isomorphisms between @xmath9-spheres of the form @xmath820 where @xmath821 we write @xmath822 for the topological boundary of @xmath823 .
this can be easily computed using the results of section [ sec:4 ] : [ lem : bdys ] let @xmath824 be nonzero integers and let @xmath543 be the residue class of @xmath825 in @xmath826 , which we identify with @xmath827 .
then @xmath828 depends only on @xmath543 .
in fact @xmath829 , the volterra circle , @xmath830,\ e=[-a , a],\ a\in [ 0,\infty]\}\ ] ] and if @xmath831 then @xmath832 given @xmath841 , there is a unique function , @xmath842 whose nontangential limit boundary value function @xmath843 is equal to @xmath844 almost everywhere .
suppose that @xmath845 is analytic on an open disc @xmath157 with @xmath846 for some @xmath847 , and that @xmath844 agrees with @xmath848 almost everywhere on @xmath849 .
then @xmath850 on @xmath851 .
indeed @xmath118 and @xmath852 both have restrictions in @xmath853 and their boundary functions agree on a set of positive measure , so the conclusion follows from the f. & m. riesz theorem together with the riemann mapping theorem .
( see also fisher @xcite . )
we apply this principle to @xmath854 .
if @xmath844 is inner and @xmath855 with @xmath856 , consider the analytic function @xmath857 , @xmath858 .
since @xmath852 is analytic on each of the open discs @xmath157 which meet the right half line and the union of these discs contains @xmath139 , we conclude that @xmath859 on @xmath139 . however , it is routine to check that this function is bounded in the upper half plane only under the stated conditions .
suppose that @xmath864 . by the f. & m. riesz theorem , @xmath865 .
suppose without loss of generality that @xmath866 ; if this is not the case , then apply the unitary automorphism @xmath730 induced by @xmath725 , which maps @xmath867 onto @xmath823 , to make it so .
if @xmath868 then there exist @xmath869 such that @xmath870 where these subspaces are the ranges of @xmath73 and @xmath862 , respectively . let @xmath871 where @xmath872 and @xmath873 , so that @xmath874 .
taking complex conjugates yields @xmath875 , so @xmath876 and thus @xmath877 .
so @xmath878 it is well - known that a unimodular function which preserves @xmath0 must be constant almost everywhere , so @xmath879 , which would imply that @xmath880 , an obvious contradiction .
so @xmath881 , say @xmath882,\quad q= [ e^{i(\alpha x^m+\beta x^n)}{{{h^{2}({{\mathbb{r}}})}}}].\ ] ] now @xmath871 leaves @xmath0 invariant , so is an inner function .
hence @xmath879 by lemma [ lem : noblaschke ] and @xmath883 , a contradiction .
the conclusion of lemma [ lem : polysphere - order ] holds nontrivially precisely when @xmath884 are not both even and it follows that the boundary @xmath822 is a unitary invariant for @xmath833 and also for the balls @xmath823 .
furthermore , lemma [ lem : bdys ] shows that @xmath822 generates @xmath792 precisely when @xmath884 are not both odd . when both these conditions prevail we can classify the spheres and balls by an argument similar to that of theorem [ thm : autgroups ] .
in fact we expect that a somewhat deeper analysis will show that in general the ( unordered ) set @xmath885 is a complete unitary invariant .
[ lem : halfplane ] let @xmath886 be nonzero integers with @xmath887 such that @xmath888 and @xmath889 . if @xmath890 induces a conformal automorphism of the upper half plane and @xmath891 are real constants such that @xmath892 where @xmath893 then @xmath894 without loss of generality , suppose that @xmath895 and @xmath824 .
either @xmath753 where @xmath737 and @xmath754 , or @xmath755 for @xmath896 and @xmath739 .
suppose first that @xmath753 ; without loss of generality , we may take @xmath897 . since @xmath898 is
then continuous and so constant on @xmath899 it follows that @xmath900 and so @xmath898 is constant on @xmath124 .
the equation @xmath901 holds almost everywhere .
considering the coefficient of @xmath902 gives @xmath903 and @xmath904 , so we suppose that @xmath905 .
differentiating gives @xmath906 if @xmath907 then we set @xmath908 to deduce that @xmath909 is algebraic , and set @xmath68 equal to any other algebraic number to see that @xmath910 is also algebraic .
simple arguments show that the same holds if @xmath911 and @xmath912 or if @xmath913 and @xmath914 .
now equate the constant terms in the original expression : @xmath915 since the right hand side is algebraic , @xmath916 . by counting repeated roots
, it now follows that @xmath917 .
if on the other hand @xmath755 then @xmath898 is continuous and so constant on @xmath918 , so @xmath919 and @xmath920 almost everywhere for @xmath921 and for @xmath922 . since the left hand side is locally unbounded only at @xmath923 and has limit @xmath278 as @xmath924 , we must have @xmath925 and @xmath920 almost everywhere .
it only remains to consider the order of growth and decay at @xmath278 and @xmath926 to see that @xmath927 and @xmath928 .
let @xmath936 with @xmath937 .
consider the subspace @xmath938 .
since @xmath159 preserves the order structure , it must map @xmath822 onto @xmath939 by lemmas [ lem : bdys ] and [ lem : polysphere - order ] . by composition with @xmath725 if necessary , we may assume that @xmath940 and then translating by the `` obvious '' inner automorphisms of @xmath941 , that @xmath789 .
let @xmath942 be the residue class of @xmath825 in @xmath826 , which by assumption is in @xmath943 .
observe that by lemma [ lem : bdys ] , the von neumann algebra generated by @xmath944 is the multiplication algebra @xmath792 , and the only possibilities for the algebra @xmath945 are @xmath946 the latter algebra has uniform multiplicity @xmath9 .
since @xmath159 sends projections in @xmath947 to projections in @xmath948 , it induces an isomorphism between @xmath792 and @xmath299 .
spatial isomorphisms preserve multiplicity , so in fact @xmath949 .
now @xmath159 is an isomorphism @xmath950 and it follows that @xmath793 where @xmath222 is unimodular and @xmath794 is a unitary composition operator with symbol @xmath734 , a borel isomorphism .
exactly as in the proof of theorem [ thm : autgroups ] , @xmath734 induces a conformal automorphism of the upper half plane . since @xmath159 is a homeomorphism with @xmath951 it maps the two components @xmath823 and @xmath867 of @xmath952 to @xmath941 and @xmath953 respectively .
in particular , there exist real @xmath891 such that @xmath954 hence the hypotheses of lemma [ lem : halfplane ] are satisfied and the result follows .
j. e. thomson .
invariant subspaces for subnormal operators . in _ surveys of some recent results in operator theory ,
volume 171 of _ pitman res .
notes math .
_ , pages 241259 .
longman sci .
tech . , harlow , 1988 . | the hardy space @xmath0 for the upper half plane together with a unimodular function group representation @xmath1 , @xmath2 , gives rise to a manifold @xmath3 of orthogonal projections for the subspaces @xmath4 of @xmath5 . for classes of admissible functions
@xmath6 the strong operator topology closures of @xmath3 and @xmath7 are determined explicitly as various @xmath8-balls and @xmath8-spheres .
the arguments used are direct and rely on the analysis of oscillatory integrals ( stein @xcite ) and hilbert space geometry . some classes of these closed projection manifolds are classified up to unitary equivalence . in particular the fourier - plancherel @xmath9-sphere and the hyperbolic @xmath10-sphere of katavolos and power @xcite appear as distinguished special cases admitting nontrivial unitary automorphisms groups which are explicitly described . |
free probability ( see e.g. @xcite ) is a branch of operator algebras that was invented by voiculescu for the purpose of studying properties of free group factors .
later voiculescu discovered in @xcite that free probability has also an application to the behavior of eigenvalues of non - commutative polynomials in independent large random matrices .
this is one of the most striking success of free probability . when applying free probability to random matrices , the standard assumption is that a family of hermitian matrices @xmath0 has a _ ( joint ) limiting distribution _ as @xmath1 , meaning that for any non - commutative @xmath2-polynomial @xmath3 in @xmath4 variables , @xmath5 admits a finite limit as @xmath1 where @xmath6 is the normalized trace such that @xmath7 .
then voiculescu s result @xcite ( see also @xcite ) states that if @xmath8 is a haar unitary random matrix , then with probability one , the enlarged family @xmath9 also has a joint limiting distribution almost surely as @xmath1 , and in the limit , @xmath8 becomes free from @xmath10 .
furthermore , haagerup - thorbjrnsen @xcite , male @xcite and collins - male @xcite obtained versions of voiculescu s results in the context of operator norm convergence .
what they proved is that the family of matrices @xmath9 admits a _ strong ( joint ) limiting distribution _ as @xmath1 , which means that it has a joint limiting distribution as stated before , and in addition , for any non - commutative self - adjoint @xmath2-polynomial @xmath3 in @xmath11 non - commuting variables , the random matrix @xmath12 has no _ outliers _ , i.e. no eigenvalues outside the limiting support of the spectrum . before the above results in free probability ,
several ` single random matrix models ' were known to have strong limiting distributions ; for example , this is the case for wigner matrices under some assumptions , and in particular for gaussian unitary ensembles ( gue ) and wishart matrices ( see ( * ? ? ?
* theorem 2.1.22 and bibliographical notes ) ) . on the other hand , in the last 10 years
, random matrix models that do not have strong limiting distributions have become fashionable .
the literature is abundant .
we refer for examples to @xcite and in particular to the pioneering work of baik - ben arous - pch @xcite and pch @xcite where outliers of finite rank deformations of wishart matrices and of gues have been studied , respectively .
the purpose of this paper is to investigate random matrix models similar to those studied in @xcite , but our models admit _ purely discrete spectra _ when the dimension tends to infinity .
in other words , our model has , as a limiting spectrum , the eigenvalues of a selfadjoint compact operator on a hilbert space .
the precise definition of our model is as follows . for simplicity ,
the dependence on @xmath13 being dropped , let @xmath14 be a family of @xmath15 deterministic matrices which has a limiting joint distribution with respect to the _ non - normalized trace _
@xmath16 , i.e. for any non - commutative @xmath2-polynomial @xmath3 without a constant term , the following limit exists : _
n _ n(p(a_1 , , a_k ) ) . let @xmath17 be a family of @xmath15 deterministic matrices which has a limiting joint distribution with respect to @xmath6 .
let @xmath18 be an @xmath15 haar unitary ( we can treat several independent haar unitaries , but for now we restrict to a single haar unitary ) .
we investigate the limiting eigenvalues of [ model ] p(a_1 , , a_k , u b_1 u^ * , , u b_u^ * ) , where @xmath3 is a selfadjoint non - commutative @xmath2-polynomial @xmath19 satisfying @xmath20 .
we show that the eigenvalues of our model converge pointwise to deterministic eigenvalues of a compact operator almost surely .
we also compute the limiting eigenvalues explicitly in terms of the limiting joint distributions of @xmath21 and of @xmath22 for several polynomials @xmath3 including _
i=1^k x_i y_i x_i^ * , _ i=1^k y_i x_i y_i^*,x y + y x , ( x y -y x ) .
our model is closely related to the recent work of shlyakhtenko where asymptotic infinitesimal freeness was proved for @xmath23 and @xmath24 .
we strengthen this result with a self - contained proof , and then show the almost sure asymptotic convergence .
shlyakhtenko s work is devoted to the study of spiked models where both discrete spectra and continuous spectra appear , studied by baik - ben arous - pch @xcite , pch @xcite and others . on the other hand ,
our research is devoted to purely discrete spectra . in our discrete spectrum model , we are able to handle completely general polynomials of random matrices and can show the almost sure convergence of eigenvalues , while general polynomials are still difficult to treat if both continuous and discrete spectra are involved . on methodology
, our model needs a new method outside the standard techniques in free probability , since our model in distribution in the usual sense : _
n _ n(p(a_1 , ,
a_k , u b_1 u^ * , , u b_u^*)^p ) = 0,p . to analyze the discrete spectrum ,
we develop the moment method with respect to the _ non - normalized trace _ @xmath16 .
the most important point is that the convergence of moments with respect to @xmath16 as @xmath1 implies the pointwise convergence of eigenvalues . together with this moment method ,
the _ weingarten calculus _ developed in free probability @xcite enables us to compute moments with respect to @xmath16 and prove the pointwise convergence of eigenvalues .
when continuous and discrete spectra are mixed , it is not obvious if our method can somehow be extended .
this paper is organized as follows .
after this introduction , section [ sec : cv - eigenvalues ] gathers preliminary materials in order to handle eigenvalue distributions of non - commutative random variables that are compact operators . in section [ sec : cyclic - monotone ]
we introduce the notion , central to this paper , of _ cyclic monotone independence _ which is a special case of infinitesimal freeness , in the framework of a non - commutative probability space with a tracial weight .
section [ sec : rmt ] contains the main result : almost sure asymptotic cyclic monotone independence .
finally , section [ sec5 ] provides examples and applications .
in this section , we consider convergence of eigenvalues in a general setting .
first , we introduce an order and the classification of eigenvalues to prove theorems in this section .
after then , we obtain characterization of convergence of eigenvalues for schatten class operators from a viewpoint of the non - normalized trace .
the results play crucial roles to handle eigenvalues using non - commutative probability theory .
we use the following notations in this paper ( in particular in this section ) .
1 . @xmath25 : the set of real - valued bounded continuous functions on @xmath26 that vanish in a neighborhood of @xmath27 .
2 . @xmath28}$ ] : the supremum norm on @xmath29 $ ] , @xmath30 .
@xmath31 : the set of functions @xmath32 that are infinitely differentiable .
@xmath33 : the trace on a separable hilbert space @xmath34 .
when @xmath35 , @xmath33 is denoted by @xmath16 .
@xmath6 : the normalized trace @xmath36 on @xmath37 .
@xmath38 : the set of @xmath39-schatten class operators on a separable hilbert space @xmath34
@xmath40 : the @xmath39-schatten norm .
if @xmath41 is a selfadjoint compact operator with eigenvalues @xmath42 then @xmath43 .
it is useful to regard eigenvalues as a multiset . for a selfadjoint compact operator @xmath41 ,
we denote by @xmath44 the multiset of its eigenvalues .
the disjoint union of multisets counts the multiplicity , e.g. @xmath45 .
let us make remarks about this definition .
* we may also view @xmath46 as a positive @xmath47 measure on @xmath48 with the following properties : ( a ) @xmath49 for each @xmath50 , ( b ) @xmath51 if the dimension of the underlying hilbert space is @xmath52 , ( c ) @xmath53 for all @xmath54 .
* alternatively , we may view @xmath44 as the collection of all real sequences @xmath55 tending to zero , quotiented by the equivalence relation @xmath56 iff there exists a permutation @xmath57 of @xmath58 such that @xmath59 for all @xmath13 .
* whenever needs be , we extend the notion of eigenvalues abstractly to selfadjoint elements of a @xmath60-algebra with a tracial weight even if the weight is not a trace on a separable hilbert space .
this is defined in section [ sec : cyclic - monotone ] and is related to section [ sec5 ] . in order to discuss the convergence of eigenvalues , it is useful to order them in a nice way . 1 .
we say that a sequence of real numbers @xmath61 is _ properly arranged _ if @xmath62 for all @xmath63 .
2 . let @xmath61 be a properly arranged sequence of real numbers .
we arrange properly the set @xmath64 and denote it by @xmath65 .
similarly , @xmath66 denotes the proper arrangement of @xmath67 . from now on ,
we always arrange eigenvalues properly unless specified otherwise .
the properly arranged eigenvalues of a selfadjoint compact operator @xmath41 on a separable hilbert space are denoted by @xmath68 .
if the dimension of the hilbert space is finite then we understand that the index @xmath69 stops at the dimension . instead of
@xmath70 we use the notation @xmath71 for the properly arranged positive and negative parts of the eigenvalues .
[ def : cv - in - eigenvalues ] let @xmath72 be selfadjoint compact operators on separable hilbert spaces @xmath73 , respectively .
we say that _ @xmath74 converges to @xmath41 in eigenvalues _ if @xmath75 for any @xmath63 and @xmath76 .
if a sequence stops at a finite @xmath69 then infinitely many @xmath27 s are to be added in the end .
convergence in eigenvalues is denoted by @xmath77 it seems also natural to define the convergence @xmath78 in eigenvalues by _
k_i(a_k)=_i(a ) i , but this is not good .
for example if @xmath79 and @xmath80 , then @xmath78 in eigenvalues , but the convergence does not hold in the sense of @xmath81 .
note that the set of `` eigenvalue distributions '' \{\{_i}_i1 : _
i_i=0 , } is metrizable by [ metric ] d(\{_i}_i1 , \{_i}_i1 ) : = _ i=1^
+ _ i=1^ , where infinitely many @xmath27 s are to be added in the end if the sequence @xmath82 or @xmath83 ends at a finite @xmath69 .
this metric is compatible with the convergence in eigenvalues .
[ convergence1 ] let @xmath84 be selfadjoint compact operators on separable hilbert spaces @xmath85 respectively
. then the following are equivalent . 1 .
[ c10 ] @xmath74 converges to @xmath41 in eigenvalues ( cf . definition [ def : cv - in - eigenvalues ] ) .
[ c20 ] @xmath86 for any @xmath87 .
[ c30 ] @xmath86 for any @xmath88 .
let @xmath89 and @xmath90 for simplicity . recall that we arrange the eigenvalues in the way @xmath91 and @xmath92 , and similarly for @xmath93 .
take @xmath95 then @xmath96 on @xmath97 for some @xmath98 .
only finitely many eigenvalues of @xmath41 are contained in @xmath99 , say @xmath100 and @xmath101 .
convergence in eigenvalues implies that there exists @xmath102 such that @xmath103 for all @xmath104 .
this implies that @xmath105 for all @xmath104 and @xmath106 .
similar facts hold for negative eigenvalues . therefore , for sufficiently large @xmath4 we have _ h_k(f(a_k))&=_i=1^f(_i^+(k ) ) + _
i=1^m f(_i^-(k ) ) + & _ i=1^f(_i^+)+ _ i=1^m f(_i^- ) = _ h(f(a ) ) .
@xmath94 : obvious . @xmath94 .
if we take @xmath88 such that @xmath107 on @xmath26 , @xmath108 for @xmath109 and @xmath110 for @xmath111 , then @xmath112 . so for sufficiently large @xmath4 we have @xmath113 and hence @xmath74 has no eigenvalues in @xmath114 .
this implies that the eigenvalues of @xmath115 are uniformly bounded , i.e. they are contained in a common interval @xmath116 $ ] .
if @xmath117 does not converge to @xmath118 then there exists a subsequence @xmath119 , a real number @xmath120 $ ] such that @xmath121 and @xmath122 as @xmath123 .
we derive a contradiction below .
let @xmath124 .
* case @xmath125 .
we take a nonnegative function @xmath88 such that @xmath126 for @xmath127 and @xmath128 for @xmath129 .
then all eigenvalues of @xmath41 lie outside the support of @xmath130 .
therefore @xmath131 for large @xmath132 , but @xmath112 , a contradiction .
* case @xmath133 .
we take a nonnegative function @xmath134 such that @xmath135 for @xmath136 and @xmath137 for @xmath138 .
then @xmath139 for large @xmath132 , but @xmath140 , a contradiction .
thus we conclude that @xmath117 converges to @xmath118 as @xmath141 .
then we go to the induction : suppose that @xmath142 as @xmath141 for every @xmath143 and suppose that @xmath144 does not converge to @xmath145 .
then as before there exists a subsequence @xmath119 , a real number @xmath146 $ ] such that @xmath147 and @xmath148 as @xmath123 .
let @xmath149 .
* case @xmath150 .
we take a nonnegative function @xmath88 such that @xmath126 for @xmath151 and @xmath128 for @xmath152 .
then @xmath153 for large @xmath132 , but @xmath154 , a contradiction .
* case @xmath155 .
we take a nonnegative function @xmath134 such that @xmath135 for @xmath156 and @xmath137 for @xmath157 .
then @xmath158 for large @xmath132 , but @xmath159 , a contradiction .
thus @xmath160 converges to @xmath145 as @xmath141 .
by induction we conclude that @xmath161 converges to @xmath162 as @xmath141 for every @xmath163 .
similarly we can prove the convergence of @xmath164 to @xmath165 .
we define a notion of the distributional convergence of a tuple of compact operators , which is motivated from definition [ def : cv - in - eigenvalues ] and proposition [ convergence1 ] .
[ def : cv - compact ] given selfadjoint compact operators @xmath166 on separable hilbert spaces @xmath167 respectively , we say that @xmath168 _ converges in compact distribution _ to @xmath169 with respect to @xmath170 as @xmath1 if for every function @xmath171 , @xmath172 and tuple @xmath173 we have that _
n_h_n(f_i_1(a_i_1(n ) ) f_i_p(a_i_p(n ) ) ) = _ h(f_i_1(a_i_1 ) f_i_p(a_i_p ) ) . in noncommutative probability
, the moment method is an important tool to prove weak convergence of probability measures . here , we show a counterpart for eigenvalues .
[ approx ] let @xmath174 and let @xmath172 .
if @xmath175 and @xmath176 on @xmath177 $ ] , then for any @xmath178 , there exists a polynomial @xmath3 such that @xmath179 , @xmath180}<\ep$ ] and @xmath181 for @xmath182 $ ] . without loss of generality
we may assume that @xmath183 .
we only consider @xmath184 for simplicity ; the general case is similar . by weierstrass approximation
we can find a polynomial @xmath185 such that @xmath186}<\ep/8 $ ] .
this implies that @xmath187 , so the polynomial @xmath188 satisfies that @xmath186}<\ep/4 $ ]
. then we define @xmath189 .
note that @xmath190 does not have a constant term . for @xmath191
$ ] one has that @xmath192 .
then we define @xmath193 , which satisfies @xmath194 .
then for any @xmath195 $ ] one has that @xmath196 . for @xmath197 $ ]
, we have & = 2|x| _ |z||x| |r(z)| z x^2 . on the last line we have used the fact that @xmath198 for @xmath199 $ ] and @xmath200 } \leq \ep/4 $ ] . [ prop2.6 ]
let @xmath84 be selfadjoint operators on separable hilbert spaces @xmath85 respectively .
suppose that there exists @xmath172 such that @xmath201 and @xmath202 for all @xmath203 .
suppose that @xmath204 as @xmath141 for any integer @xmath205 .
then @xmath74 converges to @xmath41 in eigenvalues .
we may assume that @xmath39 is an even integer since @xmath206 .
by proposition [ convergence1 ] , it suffices to show that for any bounded @xmath207 function @xmath130 such that @xmath208 on @xmath177 $ ] for some @xmath98 , it holds that @xmath209 @xmath210 . if the dimensions of some of @xmath211 are finite , then the following proof is still available by adding infinitely many 0 s to the eigenvalues , so let us assume that the dimensions of hilbert spaces are all infinite .
let @xmath212 be the properly arranged eigenvalues of @xmath213 , respectively .
let @xmath214 and suppose that @xmath215 ; otherwise @xmath213 are all zero elements .
it is easy to see by chebyshev s inequality that [ chebyshev ] # \{i : |_i(k)| > } = # \{i : _
i(k)^p > ^p } , and similarly , @xmath216 therefore if we define @xmath217\in\n$ ] then @xmath218 for all @xmath203 and @xmath219 .
since @xmath220 , the eigenvalues of @xmath213 are all contained in the interval @xmath221 $ ] . given @xmath178 , by lemma [ approx ]
we can find a polynomial @xmath3 of the form @xmath222 such that @xmath223}<\ep / i_0 $ ] and @xmath224 for @xmath182 $ ] .
then for all @xmath203 we have that [ eqa ] & |_i=1^i_0 f(_i(k ) ) - _
i=1^i_0 p(_i(k))| + |_i=1^i_0 p(_i(k ) ) - _ i=1^i_0 p(_i)| + & + |_i=1^i_0 p(_i ) - _ i=1^i_0 f(_i)| + & 2 + |_i=1^i_0 p(_i(k ) ) - _ i=1^i_0 p(_i)| .
we note here that _
i = i_0 + 1^|p(_i(k))| ^-p _ i = i_0 + 1^|_i(k)|^p ^-p a_k_p^p , and similarly , @xmath225 .
therefore , [ eqb ] |_i=1^i_0 p(_i(k ) ) - _
i=1^i_0 p(_i)| 2 + |_i=1^ p(_i(k ) ) - _ i=1^ p(_i)| .
since @xmath226 for all @xmath227 and the coefficients of @xmath228 of @xmath3 are zero , there exists @xmath229 such that [ eqc ] |_i=1^ p(_i(k ) ) - _
i=1^ p(_i)| for @xmath230 . from ,
, we have that |_i=1^f(_i(k ) ) - _ i=1^f(_i)| 5 , k k_0 , the conclusion .
[ cor unique ] let @xmath231 be selfadjoint operators on separable hilbert spaces @xmath232 , respectively , such that @xmath233 and @xmath234 for some @xmath172 . if @xmath235 for every integer @xmath227 then @xmath236 . actually in proposition [ prop2.6
] , we need not a priori assume the existence of a limiting operator @xmath41 .
[ prop2.7 ] let @xmath237 and @xmath238 be separable hilbert spaces .
let @xmath239 be selfadjoint .
suppose that the limit @xmath240 exists for every integer @xmath227 . then there exist a separable hilbert space @xmath34 and a selfadjoint operator @xmath241 such that @xmath74 converges to @xmath41 in eigenvalues , and moreover , @xmath242 for at least @xmath243 . 1 .
it is useless to assume that @xmath244 , because then the conclusion will only be @xmath245 .
2 . we can not conclude that @xmath246 .
for example we may take as eigenvalues of @xmath74 @xmath247 then @xmath248 for all @xmath63 , but @xmath249 .
let @xmath250 be the properly arranged eigenvalues of @xmath74 , and let @xmath251 .
since @xmath252 are uniformly bounded by @xmath253 , there exists a subsequence @xmath254 of @xmath255 such that @xmath256 converges to some @xmath257 .
then we can take a subsequence @xmath258 of @xmath254 such that @xmath259 converges to some @xmath260 , and continue this procedure .
let @xmath261 and @xmath262 . by construction @xmath263 for each @xmath264 .
therefore @xmath265 for every @xmath63 .
then @xmath42 is properly arranged since so is @xmath250 . by fatou s lemma _ i=1^_i^p _
. therefore , there exists a separable hilbert space @xmath34 and a selfadjoint @xmath241 such that @xmath266 .
note that @xmath267 is absolutely convergent for @xmath227 since @xmath268 is increasing on @xmath269 .
we want to show that @xmath270 for @xmath271 . for any @xmath178 , by chebyshev s inequality ( now we use @xmath178 instead of @xmath98 )
, there exists @xmath272 such that @xmath273 for all @xmath274 , and hence @xmath275 . for @xmath276
we have & |_n - _ h_k(a_k^n)| + | _ i=1^i_0(_i(k)^n - _ i^n ) | + _ i = i_0 + 1^(|_i(k)|^n + |_i|^n ) + & |_n - _ h_k(a_k^n)| + | _ i=1^i_0(_i(k)^n - _ i^n ) | + ^n - p
( _ i = i_0 + 1^(_i(k)^p + _
i^p ) ) .
the first two terms converge to 0 if we put @xmath277 and let @xmath278 tend to infinity .
the last term is bounded by @xmath279 , so we get @xmath280 and hence @xmath281 for @xmath243 . by corollary [
cor unique ] ( applied to now @xmath282 ) , the limit eigenvalues @xmath42 do not depend on the choice of the subsequence @xmath283 .
since the space of eigenvalue distributions is metrizable ( see ) , @xmath74 converges in eigenvalues to @xmath41 .
our aim is to analyze asymptotic behavior of discrete eigenvalues of random matrices . for this we would like to abstract the notion of the non - normalized trace @xmath33 on a hilbert space @xmath34 .
so we define a non - commutative measure space that replaces a state in a non - commutative probability space with a _
weight_. a _ non - commutative measure space _ is a pair @xmath284 where @xmath285 is a ( unital or non - unital ) @xmath2-algebra over @xmath286 .
let @xmath287 be a tracial weight meaning that * @xmath287 is defined in a ( possibly non - unital ) @xmath2-subalgebra @xmath288 of @xmath285 and @xmath289 is linear , * @xmath287 is positive , i.e. @xmath290 for every @xmath291 , * @xmath292 for all @xmath293 , * @xmath294 for all @xmath295 .
moreover , if @xmath285 is unital , @xmath296 and @xmath297 then we call @xmath284 _ a non - commutative probability space_. let @xmath284 be a non - commutative measure space and let @xmath298 .
the _ distribution _ of @xmath299 is the family of _ ( mixed ) moments _ \{(a_i_1^_1 a_i_p^_p ) : p1 , 1 i_1,
,i_pk , ( _ 1 , , _
p)\{1,}^p}. given non - commutative measure spaces @xmath300 and elements @xmath301 , @xmath302 , we say that @xmath169 has _ the same distribution _ as @xmath303 if ( a_i_1^_1a_i_p^_p)= ( b_i_1^_1b_i_p^_p ) for any choice of @xmath304 , @xmath305 and @xmath306 .
the distribution of @xmath299 is a _ trace class distribution _ if there exist a separable hilbert space @xmath34 and @xmath307 ( trace class operators ) such that the distribution of @xmath299 is the same as that of @xmath308 . in this case
we define the eigenvalues of a self - adjoint @xmath2-polynomial @xmath309 by the eigenvalues of @xmath310 .
let @xmath311 be non - commutative measure spaces and @xmath312 , @xmath313 .
we say that @xmath314 _ converges in distribution _ to @xmath169 if _
n(a_i_1(n)^_1a_i_p(n)^_p ) = ( a_i_1^_1a_i_p^_p ) for any choice of @xmath304 , @xmath305 and @xmath306 .
if the distributions of selfadjoint elements @xmath315 and @xmath291 are trace class then we can define the concept of _ convergence in eigenvalues _ in a natural way ( cf .
definition [ def : cv - in - eigenvalues ] ) , i.e. @xmath316 for every @xmath63 and @xmath76 .
let @xmath317 be non - commutative measure spaces .
suppose that @xmath318 have trace class distributions and that @xmath319 converges in distribution to @xmath169 .
then for any selfadjoint non - commutative @xmath2-polynomial @xmath3 without a constant term , @xmath320 converges in eigenvalues to @xmath309 .
this follows from proposition [ prop2.6 ] .
we introduce a kind of independence in a non - commutative probability space with a tracial weight @xmath321 . for @xmath2-subalgebras @xmath322 of @xmath323 such that @xmath324 , let @xmath325 be the ideal generated by @xmath285 over @xmath326 .
more precisely , _ ( ) : = span\{b_0 a_1 b_1 a_n b_n : n , a_1 , , a_n , b_0 , , b_n } , which is a @xmath60-subalgebra of @xmath323 containing @xmath285 .
[ construction ] let @xmath321 be a non - commutative probability space with a tracial weight @xmath287 . 1 .
let @xmath327 be @xmath2-subalgebras of @xmath323 such that @xmath328 .
we say that the pair @xmath329 is _ cyclically monotonically independent _ or more simply _ cyclically monotone _ with respect to @xmath330 if * @xmath331 ; * for any @xmath332 and any @xmath333 , we have that @xmath334 2
. given @xmath312 and @xmath335 , the pair @xmath336 is _ cyclically monotone _ if @xmath337 is cyclically monotone .
note that we _ do not _ assume that @xmath338 contains the unit of @xmath323 .
this definition is similar to monotone independence of muraki @xcite , but the difference is @xmath339 the factor @xmath340 is to be replaced by @xmath341 in the monotone case . let @xmath342 be a non - commutative measure space and let @xmath343 be a non - commutative probability space . let @xmath344 be the algebraic free product of @xmath285 and @xmath326 , @xmath345 being identified with the unit @xmath346 , @xmath347 where @xmath348 is a direct sum decomposition as a vector space .
we define a linear functional @xmath349 on an ideal of @xmath350 by @xmath351 for @xmath352 , @xmath353 .
we call @xmath354 the _ cyclic monotone product _ of @xmath287 and @xmath355 .
the cyclic monotone product @xmath356 is a tracial positive linear functional on @xmath357 and hence @xmath358 becomes a non - commutative measure space .
let @xmath359 and @xmath360 where @xmath361 .
the traciality follows from ( ) ( x y ) & = ( a_1 a_2 a_n a_1 a_m ) ( b_0b_m ) ( b_n b_0 ) _
i=1^n-1 ( b_i)_i=1^m-1 ( b_i ) + & = ( a_1 a_m a_1 a_2 a_n ) ( b_mb_0 ) ( b_0 b_n ) _
i=1^n-1 ( b_i)_i=1^m-1 ( b_i ) + & = ( ) ( y x ) . for the positivity it is enough to prove @xmath362 when x= _ i=1^n_i b_i,0 a_i,1b_i,1a_i,2b_i,2a_i , m(i)b_i , m(i ) , where @xmath363 , @xmath364 and @xmath365 .
by the positivity of @xmath287 we have that @xmath366 then we have @xmath367 for any @xmath368 .
thus the @xmath15 matrix a= ( ( a_i , m(i)^*a_i,1^ * a_j,1a_j , m(j)))_i , j=1^n is positive definite .
we also show that the matrices @xmath369 are positive definite .
it is easy to show that @xmath370 is positive definite .
the matrix @xmath371 is also positive definite since & _ i , j=1^n_j(b_i,1^*)(b_j,1 ) ( b_i , m(i)-1^ * ) ( b_j , m(j)-1 ) ( b_i , m(i)^*b_j , m(j ) ) + & = ( ( _ i=1^n _ i ( b_i,1 ) ( b_i , m(i)-1 ) b_i , m(i ) ) ^ * ( _ i=1^n _ i ( b_i,1 ) ( b_i , m(i)-1 ) b_i , m(i ) ) ) 0 for any vector @xmath372 the schur product @xmath373 is also positive definite . now using the definition of cyclic monotone product , we have @xmath374 therefore we obtain @xmath362 .
if we define a tracial state @xmath375 on @xmath350 by @xmath355 on @xmath326 and @xmath27 otherwise , we get a triple @xmath376 .
then @xmath377 is cyclically monotone with respect to @xmath378 .
this construction is a universal one .
when applying to random matrices we need the asymptotic version of independence .
let @xmath379 be non - commutative probability spaces with tracial weights .
let @xmath380 and @xmath381 .
the pair @xmath382 is _ asymptotically cyclically monotone _ if there exist a non - commutative probability space @xmath383 with a tracial weight , @xmath312 and @xmath335 such that 1 .
@xmath384 is cyclically monotone ; 2 . for any non - commutative @xmath2-polynomial @xmath385
such that @xmath386 we have @xmath387 and @xmath388 biane , goodman and nica introduced a kind of freeness related to type b noncrossing partitions in @xcite and then belinschi and shlyakhtenko formulated it as infinitesimal freeness , which is freeness with respect to a parametrized state @xmath389 up to the order @xmath390 @xcite ( see also fvrier and nica s work @xcite ) .
more precisely , let @xmath391 be a family of traces on an unital @xmath2-algebra @xmath323 .
let @xmath392 and suppose that @xmath393 is differentiable in the sense that the limit (x):= _ t0 exists for every @xmath394 .
moreover , suppose that @xmath2-subalgebras @xmath327 are free with respect to @xmath393 for every @xmath395 , then we have a computation formula such as @xmath396 for @xmath397 and @xmath398 .
putting @xmath399 we get these formulas for @xmath355 , and moreover if we take the derivative regarding @xmath400 at 0 we get the formulas @xmath401 these computation formulas give some kind of universal formulas for mixed moments .
conversely , given @xmath402 we can define infinitesimal freeness .
let @xmath323 be a unital @xmath2-algebra , let @xmath355 be a tracial state on @xmath323 and let @xmath403 be a tracial linear functional which satisfies @xmath404 .
the triple @xmath405 is called an _ infinitesimal non - commutative probability space _ or non - commutative probability space of type b. [ defif ] let @xmath406 be an infinitesimal non - commutative probability space .
let @xmath285 and @xmath326 be @xmath2-subalgebras of @xmath323 , which may not contain the unit of @xmath323 .
let @xmath407 for @xmath408 .
we say that @xmath285 and @xmath326 are infinitesimally free if for any @xmath409 , @xmath410 and @xmath333 _ t((a_1 - _ t(a_1)1_)(b_1-_t(b_1)1_)(a_2-_t(a_2)1 _ ) ( b_n-_t(b_n)1 _ ) ) = o(t ) . in the above definition @xmath393
may not be positive , but it does not matter in defining infinitesimal freeness .
our cyclic monotone independence is a special case of infinitesimal freeness , which was essentially proved in @xcite .
[ propif ] let @xmath406 be an infinitesimal non - commutative probability space and let @xmath322 be its @xmath2-subalgebras such that @xmath411 and @xmath412 . then @xmath322 are infinitesimally free if and only if @xmath329 is cyclically monotone with respect to @xmath413 .
we assumed the positivity of the weight and the state in the definition of cyclic monotone independence , but we may drop it .
the conclusion of proposition [ propif ] is to be understood in this generalized setup .
suppose that @xmath322 are infinitesimally free .
let @xmath414 .
since @xmath322 are free with respect to @xmath393 up to @xmath390 , it can be shown that [ eq329 ] _
t(a_1b_1a_n b_n)= _ t(a_1
a_2a_n ) _ t(b_1 ) _
t(b_2)_t(b_n ) + r(t ) + o(t ) , where @xmath415 is the sum of monomials such that every monomial in @xmath415 contains at least two factors of the form @xmath416 .
for the proof apply the formula for products of free random variables ( * ? ? ?
* theorem 14.4 ) .
our assumption implies that @xmath417 , and so @xmath418 . comparing the coefficients of @xmath400 in we get the formula (a_1b_1a_n
b_n)= (a_1 a_2a_n ) ( b_1 ) ( b_2)(b_n ) , implying the cyclic monotone independence . for the converse direction ,
we check definition [ defif ] .
we have the decomposition & _ t((a_1 - _ t(a_1)1_)(b_1-_t(b_1)1_)(a_2-_t(a_2)1 _ ) ( b_n-_t(b_n)1 _ ) ) + & = ( ( a_1 - _ t(a_1)1_)(b_1-_t(b_1)1_)(a_2-_t(a_2)1 _ ) ( b_n-_t(b_n)1 _ ) ) + & + t ((a_1 - _ t(a_1)1_)(b_1-_t(b_1)1_)(a_2-_t(a_2)1 _ ) ( b_n-_t(b_n)1 _ ) ) + & = : j_1 + t j_2 . by the assumption @xmath412
, we can see that j_1 & = ( ( - _ t(a_1)1_)(b_1-_t(b_1)1_)(-_t(a_2)1 _ ) ( b_n-_t(b_n)1 _ ) ) + & = ( -1)^n t^n (a_1 ) (a_n ) ( ( b_1-_t(b_1)1_)(b_2 -_t(b_2)1 _ ) ( b_n-_t(b_n)1 _ ) ) + & = o(t^2 ) .
note that when @xmath419 , we can show @xmath420 since @xmath421 .
for @xmath422 , since @xmath423 we can see that [ eq330 ] j_2 = (a_1 ( b_1-_t(b_1)1_)a_2 a_n ( b_n-_t(b_n)1_))+o(t ) . by cyclic monotonicity we obtain [ eq331 ] & (a_1 ( b_1-_t(b_1)1_)a_2 a_n ( b_n-_t(b_n)1 _ ) ) + & = (a_1 a_2 a_n ) ( ( b_1)-_t(b_1 ) ) ( ( b_n)-_t(b_n ) ) + & = (a_1 a_2 a_n ) ( -t)^n (b_1)(b_2 ) (b_n ) + & = o(t ) . combining and we obtain @xmath424 .
therefore , @xmath425 . for later use
we define a notion of the convergence of elements in non - commutative probability spaces to elements in an infinitesimal non - commutative probability space .
[ def convergenceif ] suppose that @xmath406 is an infinitesimal non - commutative probability space , @xmath426 are non - commutative probability spaces and @xmath427 .
we say that the tuple @xmath428 _ converges in infinitesimal distribution _ to @xmath429 if 1 .
@xmath430 converges in distribution to @xmath431 , 2 . for any non - commutative @xmath2-polynomial @xmath432 ,
n = (p(c_1 , , c _ ) ) .
in this section , we shall introduce and prove asymptotic cyclic monotone independence of haar invariant random matrices which have limiting compact distributions and random matrices which have limiting distributions with respect to the normalized trace .
theorem [ acm ] is essentially equivalent to ( * ? ? ?
* lemma 3.1 , lemma 3.2 ) by shlyakhtenko .
we supply a proof of this theorem using the _ weingarten calculus_. actually in our proof we can remove the assumption of some norm boundedness in lemma 3.1 of shlyakhtenko , and so we can unify the proofs of ( * ? ? ?
* lemma 3.1 , lemma 3.2 ) .
moreover , we can prove the almost sure asymptotic independence and almost sure convergence of discrete eigenvalues , still without the norm boundedness in the trace class setting .
we generalize the results to the compact setup , but then we need the norm boundedness .
first , we introduce the tool called the weingarten calculus , summarizing results in @xcite ( see also the references @xcite and ( * ? ? ?
* lecture 23 ) ) .
let @xmath433 be the symmetric group acting on a finite set @xmath434 and in particular let @xmath435 be the symmetric group @xmath436 .
the identity of @xmath437 is denoted by @xmath438 .
let @xmath439 be an expectation in a probability space and let @xmath440 be a normalized haar unitary random matrix , i.e. the law of @xmath18 is the normalized haar measure on the unitary group @xmath441 .
let @xmath442 be a linear map of @xmath443 defined by ( a ) = .
let @xmath444 be the canonical basis of @xmath445 $ ] and let @xmath446 be a linear map from @xmath443 to @xmath445 $ ] defined by ( a ) = _ s_k _ m_n()^k ( ( ) ^a ) _ , where @xmath447 is the natural representation ( ( ) ) ( v_1 v_k ) : = v_^-1(1)v_^-1(k ) , v_i^n . for @xmath448 and @xmath449 , we define the weingarten function @xmath450 as ( , n ) = .
it follows from ( * ? ? ?
* eq . ( 9 ) ) that the function @xmath451 is a linear combination of the characters of irreducible representations of @xmath435 , and so @xmath452 .
the conjugate classes of a symmetric group are determined by the structure of cycle decomposition . since @xmath57 and @xmath453
have the same structure of cycle decomposition , we have @xmath454 for every @xmath455 .
what is important is asymptotics of the weingarten function for large @xmath13 .
let @xmath456 be the catalan number _ p:= , p . for @xmath455 let @xmath457 be the length function , that is , the minimal number of transpositions to express @xmath57 as the product of them .
let @xmath458 be the cycle decomposition of @xmath455 and then let [ moeb ] ( ) : = _ i=1^ ( ) ( -1)^|c_i|_|c_i| .
note that @xmath459 and @xmath460 .
then it is known that ( * ? ? ?
* corollary 2.7 ) [ eq : decay ] ( , n ) = n^-k-||(()+o(n^-2 ) ) .
the weingarten function can obviously be regarded as the element of @xmath445 $ ] = _ s_k ( , n)_. it was shown in @xcite that for all @xmath461 [ wg1 ] ( a ( b ) ) = ( a ) ( b ) . in this paper
we consider @xmath462 where @xmath463 . for a cycle @xmath464 ,
let @xmath465 be the product @xmath466 and let @xmath467 be the product of traces @xmath16 according to the cycle decomposition @xmath468 , _
( a_1 , , a_k)= _ i=1^ ( ) _ n(a_c_i ) . for example
if @xmath469 then @xmath470 . note that this notation is well - defined thanks to the cyclic property of the trace .
we can then show that _ m_n()^k ( ( ) ^a ) = _ ( a_1 , , a_k )
. then reads , for every @xmath471 , [ wg2 ] & + & = _ _ _ 1(a_1 , ,
a_k ) _ _ 2(b_1 , , b_k ) ( _ 3,n ) .
this formula is the main tool of our analysis below .
we prove asymptotic cyclic monotone independence on average in the trace class setup .
note that we can replace @xmath472 with @xmath473 in the following asymptotic results .
[ acm ] let @xmath474 be an @xmath15 haar unitary and @xmath475 , @xmath476 be @xmath15 random matrices .
suppose that 1 .
[ a11 ] @xmath477 converges in distribution to a @xmath4-tuple of trace class operators as @xmath1 , 2 .
[ b11 ] @xmath478 converges in distribution to an @xmath479-tuple of elements in a non - commutative probability space as @xmath1 , 3 .
@xmath480 are independent . then the pair @xmath481 is asymptotically cyclically monotone with respect to @xmath482 .
we may assume that @xmath483 since otherwise we may add @xmath27 s to @xmath484 s or the identity matrices to @xmath485 s .
for simplicity @xmath486 is abbreviated to @xmath487 .
note that it suffices to show that the expectation [ eq222 ] factorizes following cyclic monotone independence , since a general monomial can be written in the form @xmath488 where @xmath489 and @xmath490 . by the traciality
, the distribution of @xmath491 with respect to the non - normalized trace is the same as that of @xmath492 , so every monomial reduces to .
also , a concrete construction of a limiting non - commutative probability space with a tracial weight can be given by the cyclic monotone product in definition [ construction ] .
thus , it suffices to show the factorization of the expectation .
setup @xmath493 to be the circular permutation , i.e. @xmath494 . in this context , by the weingarten formula we have & + & = _ ( _ 3,n ) .
next we make a decay analysis .
since @xmath495 has a limiting trace class distribution , then the leading behavior of @xmath496 is @xmath497 .
since @xmath498 has a limiting distribution with respect to the normalized trace , the behavior of @xmath499 is @xmath500 .
finally , we know that @xmath501 behaves as @xmath502 . therefore , for a triple @xmath503 such that @xmath504 , the contribution of the summand is @xmath505
. therefore , the asymptotics is driven by the summands for which @xmath506 i.e. @xmath507 .
this forces @xmath508 , and there is only one such summand . as a conclusion ,
= _ i=1^k + o(n^-1 ) .
some further analysis shows that @xmath509 interestingly , this is quite different from classical random matrix models inspired from free probability theory @xcite , where the covariance behaves rather like @xmath510 .
as it is classically known , a behavior of @xmath510 , summable in @xmath13 , allows to prove the almost sure convergence of the traces of random matrix models . here , since we do not have the behavior @xmath510 , we need to investigate other functionals . from now on
, we assume that the matrices @xmath511 are deterministic , but by conditioning @xmath512 to be constant , we can generalize the results to the case when @xmath511 are random matrices independent of the haar unitary @xmath18 such that @xmath513 and @xmath514 almost surely converge in distributions to deterministic elements . for example
this allows us to take @xmath515 to be a gue @xmath516 , and in this case we do not need to rotate @xmath516 by a haar unitary since @xmath516 is rotationally invariant .
[ acm2 ] let @xmath517 be @xmath15 deterministic matrices and let @xmath518 be a haar unitary random matrix .
suppose that 1 .
[ a12 ] @xmath513 converges in distribution to a @xmath4-tuple of trace class operators as @xmath1 , 2 .
[ b12 ] @xmath514 converges in distribution to a @xmath4-tuple of elements in a non - commutative probability space as @xmath1
. then @xmath519 \right|^4\right]=o(n^{-2}).\ ] ] we prove a more general result , namely , instead of taking @xmath4 matrices @xmath484 and @xmath4 matrices @xmath487 , we take @xmath520 for each satisfying the same assumption
. moreover , for notational simplicity @xmath486 is abbreviated to @xmath487 .
let @xmath521 , \end{aligned}\ ] ] which are complex - valued random variables .
we prove that [ eq : power4 ] & = o(n^-2 ) .
this implies the lemma if we define @xmath522 such that @xmath523 .
the last condition @xmath524 can be realized by taking @xmath525 , since & = _
n((a_1 b_1 a_k
b_k)^*)= _ n(a_k^ * b_k-1^ * a_k-1^ * a_2^ * b_1^ * a_1^ * b_k^ * )
. we shall denote by @xmath526 the interval @xmath527 by @xmath528 the cyclic permutation @xmath529 of @xmath530 and by @xmath531 the permutation @xmath532 of @xmath533 .
thanks to the weingarten formula we have formulas to compute the moments @xmath534 $ ] . if we expand this mixed moment , we obtain @xmath535 ( products of ) expectations . our notation is @xmath536=\sum_{a \subset \{1,2,3,4\ } } \be_a , \\
\be_a= ( -1)^{\ # a } \be\!\left[\prod_{i\in a } x_i\right ] \prod_{i \in a^c}\be [ x_i ] .
\end{aligned}\ ] ] for example , our notation means @xmath537
x_3 x_4]$ ] . in this specific example
, the weingarten formula boils down to a sum over permutations in @xmath538 for the evaluation of @xmath539 $ ] , and those in @xmath540 for the evaluation of @xmath541 $ ] .
it follows from the weingarten formula that for each @xmath542 there exists a number @xmath543 such that _ a=___1(a_1 , , a_4k)__2(b_1 , , b_4k)f_a ( _ 1,_2,n ) . note that @xmath544 may be omitted but it is written for clearer understanding .
the number @xmath545 is either zero , or a product of ( two , three or four ) weingarten functions .
again , instead of going through heavy notation that would perhaps not ease the understanding , let us describe @xmath543 in the generic case @xmath546 .
recall that @xmath547\be[x_1 x_3 x_4]$ ] .
the product of weingarten formulas for @xmath548 and for @xmath549 only gives permutations of @xmath550 , and so , for @xmath551 such that @xmath552 , f_a ( _ 1,_2,n ) = - ( _ 3| i_2,n)(_3| i_1 i_3
i_4,n ) , + , + 0 , .
the general case is alike , and hence we can define a function @xmath543 for every subset @xmath553 .
this gives us the expression & = _ _ _ 1(a_1 ,
, a_4k)__2(b_1 , , b_4k )
f ( _ 1,_2,n ) , where [ coefficients ] f ( _ 1,_2,n)= _ a\{1,2,3,4 } f_a ( _ 1,_2,n ) . as seen above , @xmath554 is a signed sum of products of wg functions on various permutation groups . by assumption , @xmath555 as @xmath1 and @xmath556 . from it follows that _ _ 2(b_1 , ,
b_4k ) f ( _ 1,_2,n)=o(n^-|_2|-|_3| ) .
we show that actually [ eq : decay2 ] _ _ 2(b_1 , ,
b_4k ) f ( _ 1,_2,n)=o(n^-2 ) , or equivalently f ( _ 1,_2,n)=o(n^-4 k + |_2|-2 ) for each fixed @xmath557 ( such that @xmath558 ) . given @xmath551 such that @xmath559 , we define an equivalence relation on @xmath560 : @xmath561 if there exists @xmath562 such that @xmath563 .
since this group contains @xmath531 , every interval @xmath526 must be a subset of some equivalence class .
then , the permutations @xmath503 associate a set partition @xmath564 of @xmath565 such that the subsets of @xmath566 , _
i p_1 i_i,_i p_2 i_i , , are exactly the equivalence classes generated by the actions of @xmath503 .
a subset @xmath553 of @xmath565 also associates the set partition @xmath567 of @xmath565 .
then only @xmath545 for which @xmath568 is coarser than or equal to @xmath569 contributes to @xmath130 in the sum and the other @xmath545 s are zero .
we distinguish several cases according to the set partition associated to @xmath503 .
@xmath570 , or equivalently , the group @xmath571 acts on @xmath572 transitively . in this case , @xmath545 is zero unless @xmath573 , and so @xmath574 . we have to exclude @xmath575 , as the condition @xmath552 contradicts transitivity
. we can also exclude @xmath576 .
indeed this means that one is the identity , and the other one is a transposition , and again the condition @xmath552 is incompatible with a transitive action .
therefore , @xmath577 , and so @xmath578 .
@xmath569 is a pair partition , e.g. @xmath579 . in this case
again we have @xmath580 and from a similar reasoning we must have @xmath581 and hence @xmath578 . in the other cases some nice cancellation occurs between wg functions . 1 .
@xmath569 has two blocks with cardinality 1 and 3 , say @xmath582 .
the equivalence classes are @xmath583 and @xmath584 .
the only indices @xmath553 s for which @xmath545 is non - zero are @xmath585 . by inspection
we see that @xmath586 by and the multiplicativity of moebius functions , we obtain @xmath587 thus @xmath588 2 .
@xmath569 has 3 blocks , say @xmath589 .
the indices @xmath553 s for which @xmath545 is non - zero are @xmath590 . we see that @xmath591 3 .
@xmath592 , namely , every @xmath526 is invariant under the actions of @xmath503 . in this case
@xmath545 contribute to @xmath130 for all the 16 subsets @xmath593 .
by multiplicativity , the dominant contribution to @xmath130 is the sum of 16 mbius functions @xmath594 multiplied by @xmath595 .
exactly half of them have minus signs , so they cancel .
thus @xmath596 .
this concludes the proof of .
[ acm3 ] let @xmath597 @xmath476 be @xmath15 deterministic matrices and let @xmath518 be a haar unitary random matrix .
suppose that 1 .
[ a13 ] @xmath513 converges in distribution to a @xmath4-tuple of trace class operators as @xmath1 , 2 .
[ b13 ] @xmath598 converges in distribution to an @xmath479-tuple of elements in a non - commutative probability space as @xmath1 .
the pair @xmath599 is asymptotically cyclically monotone almost surely with respect to @xmath600 . for notational convenience
we write @xmath486 simply as @xmath487 . from the arguments in the proofs of theorem [ acm ] , and it suffices to show that when @xmath601 _ n_n(a_1 b_1 a_k b_k ) = _ n _ n(a_1a_k ) _ i=1^k _ n _ n(b_i ) one may use the standard borel - cantelli argument , but a simpler argument is possible . by lemma [ acm2 ] and by monotone convergence
we have < , and so _
n=1^|_n(a_1b_1a_kb_k)-|^4 < which implies that @xmath602\right|=0 $ ] a.s . by theorem [ acm ] we know that _ n= _
n(a_1a_k ) _ i=1^k _ n _ n(b_i ) , so we get the conclusion . proposition [ prop2.7 ] implies the following .
note that we can only conclude that the limiting operator is hilbert schmidt .
[ convergence ev ] under the assumptions of theorem [ acm3 ] , for any selfadjoint @xmath2-polynomial @xmath385 such that @xmath603 , the hermitian random matrix @xmath604 converges in eigenvalues to a selfadjoint hilbert schmidt operator almost surely .
the limiting eigenvalues can be computed by using cyclic monotone independence .
examples of the limiting eigenvalues are computed in section [ sec5 ] .
our result implies the almost sure version of shlyakhtenko s asymptotic infinitesimal freeness .
[ if2 ] in addition to the assumption of theorem [ acm3 ] , we assume that @xmath22 converges in infinitesimal distribution to an @xmath479-tuple of elements in an infinitesimal non - commutative probability space ( see definition [ def convergenceif ] ) .
then for any @xmath2-polynomial @xmath3 in the unital noncommutative @xmath2-polynomial ring @xmath605 $ ] the limits @xmath606 exist almost surely , and thus @xmath406 is an infinitesimal non - commutative probability space .
moreover , @xmath607 are infinitesimally free with respect to @xmath413
. we decompose @xmath608 where @xmath609 and @xmath610 such that @xmath611 . by theorem [ acm3 ]
@xmath612 converges almost surely to a finite real number , so @xmath613 .
hence @xmath614 converges almost surely .
since @xmath598 converges in infinitesimal distribution then @xmath615 converge too .
therefore the limits @xmath616 exist .
let @xmath617_0 $ ] ( not containing the unit ) and let @xmath618 $ ] ( containing the unit )
. then @xmath412 since , as we saw , @xmath613 .
since @xmath329 is cyclically monotone by theorem [ acm3 ] , @xmath322 are infinitesimally free by proposition [ propif ] . through
the calculation in @xcite shlyakhtenko suggested that infinitesimal freeness is applicable to outliers , but the following question seems not solved yet . combining the calculation of shlyakhtenko @xcite and our almost sure convergence ( and other ideas if needed ) , is it possible to rigorously prove the phase transition phenomena of outliers found by baik et al .
@xcite , pch @xcite , and more generally by benaych - georges and nadakuditi @xcite and belinschi et al .
@xcite ? actually , we do not need trace class distributions in order to obtain the almost sure convergence of eigenvalues , the compact setup is enough . in this section
we denote by @xmath619 the operator norm on @xmath37 .
[ enhancedacm ] let @xmath620 , @xmath476 be deterministic @xmath15 matrices and @xmath474 be an @xmath15 haar unitary such that 1 .
@xmath621 are hermitian , 2 .
[ bounded1 ] @xmath513 converges in compact distribution to a @xmath4-tuple of compact operators @xmath622 as @xmath1 ( see definition [ def : cv - compact ] ) , 3 .
@xmath598 converges in distribution to an @xmath479-tuple of elements in a non - commutative probability space as @xmath1 , 4 .
[ bounded ] @xmath623 for every @xmath624 .
let @xmath625 be a selfadjoint @xmath2-polynomial with selfadjoint variables @xmath626 such that @xmath20 .
then @xmath627 converges in eigenvalues to a deterministic compact operator almost surely .
we may assume that @xmath601 .
for simplicity @xmath628 is abbreviated to @xmath629 . by assumption
, every @xmath630 converges in eigenvalues .
we can then find some sequence @xmath631 such that @xmath632 and @xmath633 let @xmath634 be a continuous function on @xmath26 such that @xmath634 is non - decreasing and f_j(x)= 0 , & |x| < _ j+1 , + x , & |x| > _
j. let @xmath635 be the truncations @xmath636 respectively , so that they are finite rank operators and in particular trace class operators . by the definition of convergence in compact distribution
, @xmath637 converges in distribution to the trace class operators @xmath638 as @xmath1 . under such circumstances , for each fixed @xmath639 we apply corollary [ convergence ev ] to the pair @xmath640 thus , the random eigenvalues of the polynomial @xmath641 , denoted by @xmath642 , converge to some deterministic eigenvalues @xmath643 as @xmath1 : _
n(_i^(j))^(n ) = ( _ i^(j))^,i , j
. it is follows by functional calculus that @xmath644 after the use of several triangular inequalities , we can show by , and the assumption that the random variables @xmath645 converge to @xmath27 almost surely as @xmath123 .
let @xmath646 be the random eigenvalues of @xmath647 . by weyl s inequality for eigenvalues (
* corollary iii.2.6 . )
, we have @xmath648 the second inequality gives us @xmath649 for @xmath650 , and so @xmath651 is a cauchy sequence and has a limit @xmath82 as @xmath123 for each fixed @xmath69 .
the first inequality gives us _ n|_i^(n)-_i^| & _
n|_i^(n)-(_i^(j))^(n)| + _ n|(_i^(j))^(n)-(_i^(j))^| + & + |(_i^(j))^-_i^| + & _ j+|(_i^(j))^-_i^| by letting @xmath123 we get the almost sure convergence @xmath652 as @xmath1 .
finally we prove that @xmath653 as @xmath654 , so the limiting eigenvalues correspond to a compact operator .
we denote by @xmath655 the singular values of a compact operator @xmath656 .
note that in our notation of proper arrangement , for a selfadjoint operator @xmath656 it holds that @xmath657 .
note also that singular values satisfy @xmath658 and @xmath659 which can be proved by the mini - max principle ( * ? ? ?
* corollary iii.1.2 and problem iii.6.2 ) .
suppose that the polynomial @xmath3 is of the form p(a_1 , ,
a_k , b_1 , , b_k)=_=1^m x _ , where @xmath660 is a monomial containing some @xmath661 .
take @xmath178 . by assumption
, there exists some @xmath272 such that _ n , 1 p k|_i(a_p)| /m , i i_0 .
let @xmath662 be the monomial @xmath660 with the matrix @xmath661 removed from it .
then we get @xmath663 since @xmath664 and @xmath665 then @xmath666 is finite almost surely too .
this shows that @xmath667 converges to @xmath27 as @xmath654 almost surely , and so @xmath668 .
theorems [ acm ] , [ acm3 ] , [ enhancedacm ] and corollaries [ convergence ev ] , [ if2 ] can be generalized to the case when several independent haar unitaries are involved .
proofs are just to combine our results of asymptotic cyclic monotonicity with asymptotic freeness between @xmath487 s .
for example theorem [ acm ] can be generalized as follows .
[ gacm ] let @xmath669 be @xmath15 random matrices and @xmath670 be independent haar unitary random matrices .
suppose that 1 .
[ a14 ] @xmath477 converges in distribution to a @xmath4-tuple of trace class operators as @xmath1 , 2 .
[ b14 ] for each @xmath69 , @xmath671 converges in distribution to a @xmath4-tuple of elements in a non - commutative probability space as @xmath1 , 3 .
the families @xmath672 are independent . then the pair @xmath673 is asymptotically cyclically monotone with respect to @xmath482 .
we take a haar unitary @xmath18 independent of all @xmath674 and @xmath675 .
let @xmath676 . by ( * ?
* theorem 3.1 ) it follows that @xmath677 are asymptotically free with respect to @xmath678 , so @xmath679 converges in distribution to a @xmath680-tuple of elements in a non - commutative probability space as @xmath1 .
theorem [ acm ] implies that the pair @xmath681 is asymptotically cyclically monotone with respect to @xmath482 .
since @xmath682 has the same distribution as @xmath683 , we conclude that the pair @xmath684 is also asymptotically cyclically monotone with respect to @xmath482 .
the same technique allows us to generalize lemma [ acm2 ] to the several haar unitaries case , and so we obtain the almost sure convergence , namely the generalization of theorem [ acm3 ] .
[ gacm3 ] let @xmath669 be @xmath15 deterministic matrices and @xmath670 be independent haar unitary random matrices .
suppose that 1 .
[ a1 ] @xmath513 converges in distribution to a @xmath4-tuple of trace class operators as @xmath1 , 2 .
[ b1 ] for each @xmath69 , @xmath685 converges in distribution to a @xmath4-tuple of elements in a non - commutative probability space as @xmath1 , then the pair @xmath673 is asymptotically cyclically monotone with respect to @xmath600 almost surely .
corollary [ convergence ev ] is generalized in the following form .
[ g convergence ev ] under the assumptions of theorem [ gacm3 ] , for any selfadjoint @xmath2-polynomial @xmath686 such that @xmath687 , the hermitian random matrix @xmath688 converges in eigenvalues to a selfadjoint hilbert schmidt operator almost surely .
the limiting eigenvalues can be computed by asymptotic cyclic monotonicity of @xmath689 and asymptotic freeness of @xmath690 .
corollary [ if2 ] and theorem [ enhancedacm ] can be similarly generalized , the explicit statements of which are omitted .
we also mention that we can obtain the above results if we take @xmath691 where @xmath692 are independent gues normalized so that each @xmath693 converges in distribution . in this case
we may remove the haar unitaries @xmath675 since the independent gues provide independent haar unitaries when diagonalized . in the case of a single haar unitary , considering the pair @xmath694 is equivalent to considering the pair @xmath599 .
however , in the several haar unitaries case , the pair @xmath695 becomes rather trivial .
[ trivial ] let @xmath696 be @xmath15 deterministic matrices and @xmath670 be independent haar unitaries .
suppose that 1 .
[ a1 ] @xmath513 converges in distribution to a @xmath4-tuple of trace class operators as @xmath1 , 2 .
[ b1 ] @xmath697 converges in distribution to a @xmath4-tuple of elements in a non - commutative probability space as @xmath1 , 3 .
@xmath698 for every @xmath699 for any tuple @xmath700 such that at least two of them are distinct , we have the almost sure convergence @xmath701 we may suppose that @xmath702 .
by @xmath703 we denote the expectation with respect to @xmath704 , leaving @xmath705 unchanged ( that is , the conditional expectation onto the @xmath57-field generated by @xmath706 ) .
we show that almost surely [ eq54 ] ^u_1=o(n^-2 ) . by the cyclic property of the trace and by the obvious property @xmath707 , the lhs of equals [ eq55 ] ^u_1[_n(c_1 u_1 d_1u_1^ * c_u_1 d_u_1^ * ) _ n(c_+1 u_1 d_+1u_1^ * c_2 u_1 d_2 u_1^ * ) ] , where @xmath708 s are products of @xmath709 and @xmath710 with @xmath711 and @xmath712 .
at least one of the matrices @xmath713 must have a factor @xmath714 for some @xmath715 and similarly for @xmath716 . let @xmath717 be the permutation @xmath718 . by the weingarten formula , the quantity equals _ _ _ 1(c_1 , , c_2)__2(d_1 , ,
d_2 ) ( _ 3,n ) . by assumption @xmath719 and by inspection @xmath720 . by @xmath501 behaves as @xmath721 .
thus , we obtain the behavior @xmath510 of . by taking the sum @xmath722 and by using the conditional monotone convergence theorem
we obtain the conclusion . therefore , a monomial of @xmath723 is trivial with respect to @xmath16 if it contains at least two distinct factors from @xmath724 .
the nontrivial case is only when one factor from @xmath724 appears in each monomial ; for instance _
b_i u_i a_i u_i^ * b_i^*. we compute the eigenvalues of this model in proposition [ enhancedacm3 ] .
the previous theorem [ enhancedacm ] has an interesting consequence for some subalgebras of @xmath726 , the @xmath725-algebra of compact operators on a separable hilbert space .
indeed , let @xmath285 be a non unital @xmath727-subalgebra of the compact operators @xmath726 .
without loss of generality , one may assume that it is generated by a countable sequence of selfadjoint trace class operators @xmath728 .
let @xmath729 be the @xmath60-algebra generated by the operators @xmath728 . by construction , it is dense in @xmath285 .
let @xmath326 be a separable unital @xmath727-algebra and @xmath355 be a normal tracial faithful state on it . here again , we assume without loss of generality that it is generated by a countable sequence of selfadjoint operators @xmath730 .
we form abstractly @xmath731 the @xmath60-algebra generated by monomials ( words ) in @xmath732 with the condition that at least one @xmath733 appears in the word and there is no non - trivial relation between the @xmath733 s and the @xmath734 s . clearly , @xmath735 .
besides , by the results developed earlier in this paper , for any @xmath736 , we can define a norm @xmath737 under the mild assumption that @xmath738 have microstates in @xmath739 . indeed , it is possible to take matrix models @xmath740 for @xmath734 ( by the microstates assumption ) and matrix models @xmath741 for @xmath733 ( by truncating by a finite rank projection ) .
let @xmath742 be a random haar unitary matrix independent from any preexisting @xmath57-algebra . recall that @xmath743 is a linear combination of words of @xmath744 .
replacing @xmath734 by @xmath745 and @xmath733 by @xmath741 in @xmath743 we define @xmath746 .
we can define a random norm @xmath747 to be the square root of the largest eigenvalue of @xmath748 .
theorem [ enhancedacm ] implies the almost sure convergence of eigenvalues , and therefore the almost sure convergence of @xmath747 .
we denote by @xmath737 this limit .
a limit of @xmath725-norms is clearly a @xmath725-norm .
therefore we can complete @xmath731 into a @xmath725-algebra @xmath749 , and we have @xmath750 . note that unless @xmath326 is trivial , @xmath751 , and there is an embedding @xmath752 even if @xmath285 is @xmath726 itself .
note that in principle , this construction depends on the truncation .
however , following the main results of this paper , we can check that it does not depend on the truncation , and therefore @xmath737 is well - defined .
we believe that this construction is new and hopefully of its own interest , as it somehow mimics @xmath725-algebraic constructions of free products in the case of free probability .
however we do not have applications of this construction at this point .
we believe that the assumption that the existence microstates for @xmath753 is not necessary , but at this point , we do not see how to carry the proof without this assumption .
we provide explicit discrete eigenvalues of some polynomials of random matrices . by results in section [ sec : rmt ] ,
the essential part is the computation of the eigenvalues of polynomials of cyclically monotone elements .
we need a combinatorial lemma to compute the eigenvalues of anti - commutator .
[ lem1 ] for every integer @xmath754 we have @xmath755}{m } , \qquad m=0,1,\dots , [ n/2],\ ] ] where @xmath756 for integers @xmath39 .
if @xmath757 the identity is easy , so we assume that @xmath758 .
the desired formula can be written in the form _
p=0^ ( -1)^p = ( -1)^m-1 , m=1 , , [ ( r+1)/2 ] . in order to prove this we use the identity = _ p=0^x^p .
we observe that [ eq92 ] & = ( _ p=0^(-x)^p ) ( _ q=0^x^p ) + & = _ r=0^(_p=0^ ( -1)^p ) x^r .
the left hand side can be computed as [ eq93 ] & = + & = ( -1)^m-1(1+x ) _
p=0^x^2 p-1 + & = ( -1)^m-1_p=0^ ( x^2 p-1+x^2 p ) + & = ( -1)^m-1_r=0^ x^r .
the comparison of and implies the conclusion .
[ compute ] let @xmath321 be a non - commutative probability space with a tracial weight .
let @xmath759 and let @xmath760 .
suppose that @xmath761 has a trace class distribution with respect to @xmath287 and that the pair @xmath762 is cyclically monotone in @xmath321 . 1 .
[ a ] suppose that @xmath763 are selfadjoint .
let @xmath764 .
then @xmath765 where @xmath766 is viewed as an element of @xmath767 .
[ d ] suppose that @xmath768 are selfadjoint .
then @xmath769 3 .
[ b ] suppose that @xmath231 are selfadjoint .
let @xmath770 . then @xmath771 4 .
[ c ] suppose that @xmath231 are selfadjoint .
. then @xmath773 ) = ( r \ev(a ) ) \sqcup ( -r \ev(a ) ) .\ ] ] our proofs are rather direct combinatorial arguments , but it may be possible to prove and by generalizing nica and speicher s computation of the distributions of commutators and anti - commutators of free random variables @xcite to the infinitesimal free case .
let @xmath774 . since @xmath775 is symmetric nonnegative definite matrix , we can define @xmath776
. then @xmath777 .
for @xmath778 we have @xmath779 where @xmath780 .
corollary [ cor unique ] implies that @xmath781 .
the claim follows from @xmath782 and from corollary [ cor unique ] .
we may assume that @xmath783 since the general case can be covered by approximation .
let @xmath784 .
then ( ( a b+ b a)^n)=_(i_1 , , i_n)\{1,2}^n ( c_i_1 c_i_n ) .
given @xmath785 the set of _ ascents _ is defined by ( ) : = \{1 m n ( i_m , i_m+1 ) = ( 1,2 ) } with the convention that @xmath786
. then ( c_i_1 c_i_n ) = ( a^n ) ( b)^n ^2 # ( ) , where @xmath787 .
let @xmath788 be the set of noncrossing partitions of @xmath789 such that every block has cardinality 1 ( singleton ) or 2 ( pair block ) , each singleton has depth 0 or 1 and each pair block has depth 0 .
to each @xmath790 such that @xmath791 we associate @xmath792 in the following procedure ( with the convention that @xmath793 ) : * if @xmath794 and @xmath795 then @xmath796 is a singleton of @xmath797 ; * if @xmath798 and @xmath799 then there exists a unique minimal integer @xmath800 such that @xmath801 .
then @xmath802 is a pair block of @xmath797 .
( 80,10 ) ( 0,0)(1,1)10 ( 10,10)(1,0)20 ( 30,10)(1,-1)10 ( 40,0)(1,0)20 ( 60,0)(1,1)10 ( 70,10)(1,0)10 @xmath803 ( 80,20 ) ( 0,0)(0,1)10 ( 0,10)(1,0)30 ( 10,0)(0,1)5 ( 20,0)(0,1)5 ( 30,0)(0,1)10 ( 40,0)(0,1)10 ( 50,0)(0,1)10 ( 60,0)(0,1)10 ( 60,10)(1,0)20 ( 70,0)(0,1)5 ( 80,0)(0,1)10 ( 80,10 ) ( 0,0)(1,0)20 ( 20,0)(1,1)10 ( 30,10)(1,0)10 ( 40,10)(1,-1)10 ( 50,0)(1,1)10 ( 60,10)(1,-1)10 ( 70,0)(1,0)10 @xmath804 ( 80,20 ) ( 0,0)(0,1)10 ( 10,0)(0,1)10 ( 20,10)(1,0)20 ( 20,0)(0,1)10 ( 40,0)(0,1)10 ( 30,0)(0,1)5 ( 50,0)(0,1)10 ( 60,0)(0,1)10 ( 50,10)(1,0)10 ( 70,0)(0,1)10 ( 80,0)(0,1)10 the correspondence @xmath805 is bijective and @xmath806 . then _ ( c_i_1 c_i_n ) & = ( a^n ) ( b)^n _ _ 1,2;1(n)^2 # ( ) + & = ( a^n ) ( b)^n_m=0^[n/2 ] _ ^2 m + & = ( a^n ) ( b)^n_m=0^[n/2 ] ^2 m + & = ( a^n ) ( b)^n .
we then compute the sum over @xmath807 such that @xmath808 .
let @xmath809 be the flip of @xmath810 and @xmath811 .
then the map @xmath812 defines an involution on @xmath813 whose restriction to @xmath814 is a bijection onto @xmath815 .
moreover , it holds that ( c_i_1 c_i_n ) = ( c_i_1^ c_i_n^ ) .
thus _ ( c_i_1 c_i_n ) = _ ( c_i_1 c_i_n ) .
therefore we conclude that @xmath816 so corollary [ cor unique ] implies the conclusion .
let @xmath817 as before . then ( ( ) ^n)=_(i_1 , , i_n)\{1,2}^n ^n ( d_i_1 d_i_n ) .
we show that @xmath818)^n)=0 $ ] when @xmath13 is odd .
if we apply the involution @xmath819 defined in the proof of , then [ involution ] ( d_i_1d_i_n ) = ( -1)^n ( d_i_1^d_i_n^ ) . therefore if @xmath13 is odd then ( ( ) ^n ) & = _
( i_1 , , i_n)\{1,2}^n ^n ( d_i_1 d_i_n ) + & = _ ^n ( d_i_1 d_i_n)+ _ ^n ( d_i_1^ d_i_n^ ) + & = 0 .
hereafter we may assume that @xmath13 is even .
let @xmath820 be the set of inner singletons of @xmath821 , that is , the set of singletons with depth 1 .
given @xmath785 , it holds that ( d_i_1 d_i_n ) = ( -1)^#(())+ # ( ( ) ) ( a^n)(b)^n ^2 # ( ( ) ) , where @xmath253 is the number defined in the proof of . for fixed integers @xmath822
, we have that # \{_1,2;1(n)#()=m , # ( ) = } = , which is to be understood as 1 when @xmath823 and 0 when @xmath824 .
this is because there are @xmath825 ways to place @xmath479 unlabeled singletons inside @xmath278 labeled pair blocks , and then there are @xmath826 ways to place the other @xmath827 unlabeled singletons outside the @xmath278 labeled pair blocks ( the number of ways to distribute @xmath828 unlabeled balls into @xmath829 labeled boxes ) .
thus & _ ^n ( d_i_1 d_i_n ) + & = ^n ( a^n)(b)^n_m=0^n/2 ( -^2)^m _ = 0^n-2 m ( -1)^ , and by lemma [ lem1 ] , we have that _ ^n ( d_i_1 d_i_n ) & = ^n ( a^n)(b)^n_m=0^n/2 ( -^2)^m + & = ( -1)^n/2 ( a^n)(b)^n ( 1-^2)^n/2 + & = ( a^n ) r^n . since we know , we get ( ( ) ^n ) = 2 r^n ( a^n ) for even @xmath13 . if @xmath830 are the eigenvalues of @xmath41 then for every nonnegative integer @xmath13 @xmath831)^n ) = ( r\lambda_1)^n + ( r \lambda_2)^n + \cdots + ( -r\lambda_1)^n + ( -r \lambda_2)^n+\cdots,\ ] ] showing the conclusion by corollary [ cor unique ]
. the above computation immediately gives the corresponding results on random matrices converging in distribution to trace class elements .
we can moreover show the results in the compact setup ( i.e. the same assumptions as in theorem [ enhancedacm ] ) by approximation .
[ enhancedacm2 ] let @xmath832 be deterministic @xmath15 matrices and @xmath474 be an @xmath15 haar unitary such that 1 .
@xmath621 are hermitian , 2 .
[ bounded1 - 2 ] @xmath513 converges in compact distribution to a @xmath4-tuple of compact operators @xmath622 as @xmath1 ( see definition [ def : cv - compact ] ) , 3 .
@xmath514 converges in distribution to a @xmath4-tuple of elements in a non - commutative probability space as @xmath1 , 4 .
[ bounded-2 ] @xmath623 for every @xmath833 . in this case
let @xmath834 , @xmath835 and @xmath836 .
the following statements hold true
. 1 . [ a2 ] we have @xmath837 where @xmath766 is viewed as an element of @xmath838 .
[ d2 ] suppose that @xmath839 are hermitian .
then @xmath840 3 .
[ b2 ] suppose that @xmath841 and @xmath842 is hermitian .
let @xmath843 then @xmath844 4 . [ c2 ] suppose that @xmath841 and @xmath842 is hermitian .
. then @xmath846 ) = ( r \ev(a_1 ) ) \sqcup ( -r \ev(a_1 ) ) \text{~a.s.}\ ] ] when @xmath763 are trace class operators and @xmath847 converges in distribution to @xmath622 , we can modify to allow @xmath484 to be non - hermitian as in theorem [ compute ] .
the last statement follows from theorem [ compute ] and corollary [ convergence ev ] . for the statements in the compact setup , we only show since the other cases are similar .
we reuse the notations and proof of theorem [ enhancedacm ] .
in particular @xmath486 is abbreviated to @xmath487 .
now the polynomial @xmath3 is p(x_1 , , x_k , y_1 , , y_k ) = _
i=1^k y_i x_i y_i^*. recall that @xmath848 from the result in the trace class setup ( i.e. the last statement ) , we have the identity [ eq567 ] \{_i^(j)}_i1 = ( ( a_1^(j ) , , a_k^(j ) ) ) .
now we define another sequence of eigenvalues \{_i}_i1:= ( ( a_1 , , a_k ) ) .
our goal is to demonstrate that @xmath849 for every @xmath63 .
for this we use the inequality & + |(_i^(j))^(n ) - ( _ i^(j))^| + | ( _ i^(j))^- ( _ i)^| .
we proved in the proof of theorem [ enhancedacm ] that the first term on the rhs converges to 0 as @xmath1 and proved in that the second term is bounded by @xmath850 uniformly on @xmath13 .
the third term converges to 0 as @xmath1 from the result in the trace class setup .
taking @xmath851 we get |_i^-(_i)^| _ j| ( _ i^(j))^- ( _ i)^| . by
it now suffices to show that _ j ( ( a_1^(j ) , ,
a_k^(j ) ) ) = ( ( a_1 , , a_k ) ) , which follows from the fact _
j ( a_1^(j ) , , a_k^(j ) ) - ( a_1 , , a_k ) = 0 and weyl s inequality ( * ? ? ?
* corollary iii.2.6 ) which bounds the difference of eigenvalues by the operator norm of the difference of elements .
when several independent haar unitaries are involved we can still compute the eigenvalues .
for example we obtain the following . [ enhancedacm3 ] under the assumptions of theorem [ enhancedacm2 ] , we take independent haar unitaries @xmath852 . 1 .
[ a6 ] let @xmath834 . then @xmath853 2 . [ b6 ]
. then we have @xmath855 suppose that @xmath513 converges in distribution to trace class operators ; the general compact case is proved by approximation . by corollary [ convergence ev ] ,
the computation formula is obtained by the asymptotic cyclic monotonicity of @xmath856 and then the asymptotic freeness of @xmath857 , @xmath858 , @xmath859 .
this implies that we only need to replace the covariance matrix @xmath775 in theorem [ enhancedacm2 ] with @xmath860 .
the limiting eigenvalues are given by those of @xmath861 .
since @xmath862 , the conclusion follows easily .
by proposition [ trivial ] , for each @xmath863 we have _ n((_i=1^k b_i u_i
a_i u_i^ * b_i^*)^ ) & = _ i=1^k _
n((b_i u_i a_i u_i^ * b_i^*)^)+o(1 ) + & = _
i=1^k _ n((a_i ( u_i^ * b_i^*b_i u_i))^)+o(1 ) . by the cyclic monotonicity of the pair @xmath864
( see theorem [ acm3 ] ) , we have the almost sure convergence _ n_n((_i=1^k b_i u_i
a_i u_i^ * b_i^*)^ ) & = _ i=1^k _
n_n(a_i^ ) _ n_n(b_i^*b_i)^ + & = _
i=1^k _ h ( ( _ i a_i)^ ) , and the conclusion follows by proposition [ prop2.6 ] . in this subsection , we give numerical illustrations of main theorems of this paper .
let @xmath865 be a family of independent @xmath15 non - selfadjoint gaussian random matrices , that is , each @xmath528 has entries that are independent identically distributed with the standard complex normal distribution .
we consider a sample covariance matrix @xmath866 and a diagonal matrix @xmath867 . from theorem [ acm ] , the matrices @xmath749 and @xmath656 are asymptotically cyclically monotone .
table [ moments ] shows numerical simulations of mixed moments and moments decomposed by cyclic monotone independence when @xmath868 .
.moments ( @xmath868 ) [ cols="^,^,^,^",options="header " , ] [ moments2 ]
all authors were supported by jsps kakenhi grant number 26800048 , 15k17549 and 15k04923 , respectively .
b. collins was supported by nserc discovery and accelerator grant , and anr grant stoq .
capitaine , m. , donati - martin , c. and fral , d. the largest eigenvalues of finite rank deformation of large wigner matrices : convergence and nonuniversality of the fluctuations . _ ann .
_ 37 ( 2009 ) , no .
1 , 147 . collins , b. moments and cumulants of polynomial random variables on unitary groups , the itzykson - zuber integral , and free probability .
not_. 2003 , no .
17 , 953982 . | we study the eigenvalues of polynomials of large random matrices which have only discrete spectra .
our model is closely related to and motivated by spiked random matrices , and in particular to a recent result of shlyakhtenko in which asymptotic infinitesimal freeness is proved for rotationally invariant random matrices and finite rank matrices .
we show the almost sure convergence of shlyakhtenko s result .
then we show the almost sure convergence of eigenvalues of our model when it has a purely discrete spectrum .
we define a framework for analyzing purely discrete spectra and develop the moment method for the eigenvalues of compact ( and in particular schatten class ) operators .
we give several explicit calculations of purely discrete eigenvalues of our model . |
the gell - mann , or `` decontraction '' formula is a simple prescription designed to determine a deformation of a lie algebra that is `` inverse '' to the inn - wigner contraction @xcite .
this formula expresses elements of `` decontracted '' algebra in terms of elements of the contracted one . since , by a rule , various properties of the contracted algebra are much easier to explore ( e.g. construction of representations @xcite , decompositions of a direct product of representations @xcite , etc . ) , this formula found its place , as a useful and simple tool , even in some textbooks @xcite and in the mathematical encyclopedia @xcite .
there is a myriad of inn - wigner lie algebra contraction applications arising in various parts of theoretical physics .
just to mention a few ranging from contractions from the poincar algebra to the galilean one , and from the heisenberg algebras to the abelian ones of the same dimensions ( a symmetry background of a transition processes from relativistic and quantum mechanics to classical mechanics ) to those of contractions from ( anti-)desitter to the poincar algebra , and various cases involving virasoro and kac - moody algebras .
a recent study of an affine gauge gravity theory in @xmath7 @xcite is heavily related to the @xmath3 algebra contraction w.r.t .
its @xmath8 subalgebra , and the representations of the relevant algebras .
the main drawback of the gell - mann formula is its limited validity .
there is a number of references dealing with the question when this formula is applicable @xcite .
the formula is best studied in the case of ( pseudo ) orthogonal algebras
@xmath9 contracted w.r.t .
their @xmath10 and/or @xmath11 subalgebras , i.e. on the group level for @xmath12 or @xmath13 , where , loosely speaking , the gell - mann formula works very well @xcite .
moreover , the case of ( pseudo ) orthogonal algebras is the only one where this formula is valid for ( almost ) all representations @xcite .
there were some attempts to generalize the gell - mann formula @xcite , that resulted in a construction of relatively complicated polynomial formulas for the `` decontracted '' algebra operators , applicable to complex simple lie algebras @xmath14 with respect to decomposition @xmath15 .
in this work we are generally interested in gell - mann s formula for the @xmath16 algebras , that are contracted w.r.t
. their maximal compact @xmath2 subalgebras .
note that , due to mutual relations of the @xmath16 and @xmath1 algebras , one can convey the results obtained for the @xmath16 algebras to the corresponding ones of the @xmath1 algebras .
there are some subtleties in that process that will be considered below .
the gell - mann formula is , in this case , especially valuable for the problem of finding all unitary irreducible representations of the @xmath16 algebras in the basis of the @xmath17 and/or @xmath18 groups generated by their @xmath2 subalgebras .
finding representations in the basis of the maximal compact subgroup @xmath17 of the @xmath19 group , is mathematically superior , and it suites well various physical applications in particular in nuclear physics , gravity , physics of p - branes @xcite etc .
moreover , this framework opens up a possibility of finding , in a rather straightforward manner , all matrix elements of noncompact @xmath19 generators for all representations .
unfortunately , the original gell - mann formula is , in that respect , limited to some classes of ( multiplicity free ) representations only . as already stated , the gell - mann formula , except in the cases of ( pseudo ) orthogonal algebras , is not generally valid by itself , and its validity depends on the representations of the algebra as well
. therefore , in the case of the @xmath19 groups , i.e. their @xmath16 algebras , one is faced , in addition to the pure algebraic features , with a problem of studying the matters that are relevant to characterize representations as well : ( i ) the group topology properties , and ( ii ) the non trivial multiplicity of the @xmath19 , and @xmath16 representations in the @xmath17 , and @xmath2 basis , respectively .
both features are rather subtle for @xmath20 .
note that , in the case of the @xmath16 algebras , due to a fact that the generalization of the gell - mann formula obtained below depends on the algebra representation features , we deviate from the standard lie algebra deformation approach .
the @xmath19 group can be decomposed , as any semisimple lie group , into the product of its maximal compact subgroup @xmath21 , an abelian group @xmath22 and a nilpotent group @xmath23 .
it is well known that only @xmath24 is not guaranteed to be simply - connected .
there exists a universal covering group @xmath25 of @xmath21 , and thus also a universal covering of @xmath26 : @xmath27 . for @xmath20 ,
@xmath19 has double covering , defined by @xmath28 the double - covering of the @xmath17 subgroup . the universal covering group @xmath29 of a given group @xmath30 is a group with the same lie algebra and with a simply - connected group manifold . a finite dimensional covering , @xmath31 exists provided one can embed @xmath31 into a group of finite complex matrices that contain @xmath18 as subgroup . a scan of the cartan classical algebras points to the @xmath32 groups as a natural candidate for the @xmath19 groups covering .
however , there is no match of the defining dimensionalities of the @xmath32 and @xmath18 groups for @xmath20 , @xmath33 } = dim(spin(n))$ ] , except for @xmath34 . in the @xmath34 case
, one finds that the orthogonal subgroup of the @xmath35 and @xmath36 groups is @xmath37 and not @xmath38 . for a detailed account of the @xmath39 case cf
thus , we conclude that there are no covering groups of the @xmath19 , @xmath40 groups defined in finite - dimensional spaces .
an explicit construction of all @xmath41 irreducible representations , unitary and nonunitary multiplicity - free spinorial @xcite , and unitary non - multiplicity - free @xcite , shows that they are infinite - dimensional .
the universal ( double ) covering groups , @xmath42 , @xmath20 of the @xmath43 , @xmath20 group are groups of infinite complex matrices .
all their spinorial representations are infinite dimensional . in the reduction of this
representations w.r.t .
@xmath18 subgroups , one finds @xmath18 representations of unbounded spin values .
the @xmath44 groups are compact , with a simply connected group manifold , thus being its own universal coverings .
the @xmath17 subgroups are embedded into the @xmath44 groups as @xmath45-dimensional matrices , and this embedding does not allow nontrivial ( double ) covering of @xmath17 within @xmath44 . as a consequence , in the reduction of the @xmath44 unitary irreducible representations
one finds the tensorial @xmath17 representations only .
an inspection of the unitary irreducible representations of the @xmath31 , @xmath46 groups @xcite shows that they have , as a rule , a nontrivial multiplicity of the @xmath18 , @xmath47 subgroup representations .
it is well known , already from the case of the @xmath48 representations in the @xmath49 subgroup basis , that the additional labels required to describe this nontrivial multiplicity can not be solely related to the group generators themselves .
an elegant solution , that provides the required additional labels , is to work in the group manifold of the @xmath17 maximal compact subgroup , and to consider an action of the group both to the right and to the left . in this way one obtains , besides the maximal compact subgroup labels , an additional set of labels to describe the @xmath17 subgroup multiplicity .
all unitary irreducible representations of the @xmath50 and @xmath51 groups are classified , and various relevant explicit expressions are known @xcite .
it turns out that an answer to the question of the gell - mann formula generalization can be effectively read out from these known closed form expressions of representations of noncompact generators in both @xmath52 and @xmath53 cases .
such a generalization then , as a rule , has an overall validity for all representations .
we study the known representations of the noncompact @xmath54 and @xmath55 generators in the maximal compact subgroup basis , and infer the sought for expressions for the corresponding gell - mann formula . on the basis of these results
, we turn to the case of the @xmath56 generators , whose unitary irreducible representations are not known completely . as a result , we obtain a single closed expression that generalizes the gell - mann formula for the @xmath3 algebra w.r.t . its maximal compact @xmath57 subalgebra .
the @xmath16 algebra operators , i.e. the @xmath19 group generators , can be split into two subsets : @xmath59 , @xmath60 operators of the maximal compact subalgebra @xmath2 ( corresponding to the antisymmetric real @xmath61 matrices , @xmath62 ) , and the , so called , sheer operators @xmath63 , @xmath64 ( corresponding to the symmetric traceless real @xmath61 matrices , @xmath65 ) . the @xmath16 commutation relations , in this basis , read : & = & i(_acm_bd + _ adm_cb - _ bcm_ad - _ bdm_ca ) , + [ m_ab , t_cd ] & = & i(_act_bd + _ adt_cb - _ bct_ad - _ bdt_ca ) , [ mtcommutator ] + [ t_ab , t_cd ] & = & i(_acm_db + _ adm_cb + _ bcm_da + _ bdm_ca ) . [ ttcommutator ] the @xmath1 algebra operators can be split likewise w.r.t .
its @xmath2 subalgebra into @xmath59 and @xmath66 , @xmath64 .
the @xmath66 and @xmath63 operators are mutually related by @xmath66 = i @xmath63 , and the @xmath67 $ ] differs from ( [ ttcommutator ] ) by having an overall plus sign on the right - hand side .
the inn - wigner contraction of @xmath0 with respect to its maximal compact subalgebra @xmath2 is given by the limiting procedure : u_ab _ 0 ( t_ab ) , [ limit ] which leads to the following commutation relations : & = & i(_acm_bd + _ adm_cb - _ bcm_ad - _ bdm_ca ) + [ m_ab , u_cd ] & = & i(_acu_bd + _ adu_cb - _ bcu_ad - _ bdu_ca ) [ mucommutator ] + [ u_ab , u_cd ] & = & 0 .
therefore , the inn - wigner contraction of @xmath0 gives a semidirect sum @xmath68 algebra , where @xmath69 is an abelian subalgebra ( ideal ) of `` translations '' in @xmath70 dimensions .
the gell - mann formula , which is a prescription to provide an `` inverse '' to the contraction , ( [ limit ] ) , in this case reads : t_ab = u_ab + [ c_2(so(n ) ) , u_ab ] , [ gell - mann_original]where @xmath71 denotes the second order casimir operator of the @xmath2 algebra , @xmath72 , while @xmath73 is an arbitrary ( complex ) parameter and @xmath74 is a ( real ) normalization constant that depends on @xmath45 . in order to make use of the gell - mann formula to obtain @xmath16 representations ,
the first necessary step is to construct representations of the contracted algebra
. representations of the @xmath2 generators @xmath59 are well known .
there are two properties that characterize representations of the @xmath75 operators : ( i ) the @xmath75 operators transform w.r.t the @xmath76 dimensional representation of @xmath2 ( [ mucommutator ] ) , i.e. as a symmetric second order @xmath2 tensors ( in young diagram notation @xmath77 ) , and ( ii ) @xmath75 mutually commute .
these two requirements are met by expressing the @xmath75 operators in terms of the , so called , wigner @xmath78-function ( the @xmath17 group matrix elements expressed as functions of the group parameters ) : u_ab ~d^()_(cd)(ab)(g^-1 ( ) ) , [ uisd]@xmath79 being an @xmath17 element parameterized by @xmath80 , pairs of indices @xmath81 and @xmath82 label the @xmath17 matrix elements , while @xmath83 form a basis of the @xmath77 representation space . taking inverse of @xmath14 in ( [ uisd ] ) insures the correct transformation properties .
the contracted @xmath68 and @xmath16 algebras are represented in a space of square integrable functions over the @xmath18 group ( in accord with the @xmath19 topological properties ) , with a standard invariant haar measure : @xmath84 .
harish - chandra proved @xcite that this space is rich enough to contain all possible representations ( up to equivalence ) of the @xmath31 group , i.e. @xmath16 algebra .
the @xmath75 operators act multiplicatively on this space , while the @xmath2 subalgebra operators act , in a standard way , via a left group action : @xmath85 this representation space is highly reducible , however this fact is of no relevance for the present considerations . the @xmath75 expressions ( [ uisd ] ) fulfill , straightforwardly , both required properties : commute mutually , as being ordinary functions of @xmath80 , and transform under @xmath77 when acting to the right on a ket vector @xmath86 ( characterized accordingly by the @xmath81 indices )
. however , the bra vector @xmath87 , standing to the left of @xmath88 , can be an arbitrary vector from the @xmath77 representation , thus providing an additional set of labels of the @xmath75 , and accordingly the @xmath63 , operators .
it turns out , as it will be seen below , that these additional labels play an important role in the @xmath16 unitary irreducible representations description , in particular in characterizing nontrivial @xmath18 subgroup multiplicity .
a natural orthonormal basis in the @xmath18 representation space is given by properly normalized functions of the @xmath18 representation matrix elements:\{| l@l j & + k & m > d^j_km(g()^-1 ) d|g ( ) } , = _
jj_kk_mm , [ naturalbasis]where @xmath89 is an ( normalized ) invariant haar measure , and @xmath90 are the representation matrix elements @xmath91 here , @xmath92 stands for a set of @xmath18 irreducible representation labels , while @xmath93 and @xmath94 labels numerate representation basis vectors .
an action of the @xmath2 operators in this basis is well known , and it can be written in terms of the clebsch - gordan coefficients of the @xmath18 group as follows , = = _ jj c. [ maction ] the matrix elements of the @xmath75 operators in this basis are readily found to read : & = & + & = & d_km^j * ( ) d^_(cd)(ab ) ( ) d_km^j ( ) d [ uaction ] + & = & c c a closed form of the matrix elements of the whole contracted algebra @xmath68 representations is thus explicitly given in this space by ( [ maction ] ) and ( [ uaction ] ) .
let us now , equipped with the knowledge about the contracted algebra representation , consider the gell - mann formula in the cases of the @xmath95 and @xmath96 algebras . in the @xmath95 algebra case , the maximal compact subgroup of the corresponding @xmath50 group is @xmath97 , and a basis of the @xmath95 representation space
is given by the well known set of vectors , \
{ | l@l j & + k & m > , j = 0 , 12 , 1 , 32 , ; |k| , |m| j } .
[ sl3 basis ] the traceless symmetric tensor @xmath77 transforms under a five - dimensional @xmath49 , i.e. @xmath97 , representation of @xmath98 .
one can make use of an arbitrary vector from this representation to evaluate the @xmath75 operators expressions .
we take the simplest , however for our purposes adequate , realization of the @xmath75 operators , @xmath99 , @xmath100 , i.e. in the spherical @xmath49 basis , @xmath101 , @xmath102 .
the gell - mann formula ( [ gell - mann_original ] ) yields now : t
_ = d^2_0 + i,= 0,1,2[gell - mannsl(3 ) ] the matrix elements of the shear operators @xmath103 are given by the following expression : l@l j & + k & m | t _
| l@l j & + k & m = ( + i ( j(j+1)- j(j+1 ) ) ) c c [ tmatrixsl(3 ) ] the shear operators ( [ gell - mannsl(3 ) ] ) satisfy the relation ( [ mtcommutator ] ) by a construction .
however the relation ( [ ttcommutator ] ) is not a priory granted , and it must be checked to hold .
an explicit calculation shows that this relation does not hold in general .
it turns out , that the commutation relations of the @xmath95 algebra , as realized by ( [ maction ] ) and ( [ tmatrixsl(3 ) ] ) , hold only for the representation subspaces characterized by @xmath104 , and provided @xmath105 , i.e. for \ { | l@l j & + 0 & m > , j = 0 , 12 , 1 , 32 , , |m| j } , therefore , the gell - mann formula is valid only for the @xmath95 representations defined in the hilbert spaces over the @xmath106 coset space .
these representations are the multiplicity free ones w.r.t .
the compact @xmath107 subalgebra , since the label @xmath93 , which counts the @xmath107 representations multiplicity ( i.e. in the physical terms the spin @xmath92 degeneracy ) is fixed .
the gell - mann prescription ( [ gell - mannsl(3 ) ] ) does not work in the general case ; a comparison with the complete classification of the @xmath95 representations @xcite reveals that the @xmath95 representations with nontrivial multiplicity , as well as the spinorial multiplicity free representations , can not be obtained in this way ( @xmath108 implies that @xmath92 must take strictly integer values ) .
moreover , a detailed analysis shows that this conclusion hold irrespectively of the concrete choice @xmath109 that we have made . in the @xmath96 algebra case ,
the maximal compact subgroup of the corresponding @xmath51 group , is @xmath110 .
one possibility of choosing a basis for the @xmath111 representations space corresponds to the @xmath111 algebra decomposition @xmath112 .
the @xmath111 representation space basis is now @xmath113 , where pairs @xmath114 and @xmath115 specify vectors of the two @xmath49 representations .
a basis of the @xmath96 representation space ( [ naturalbasis ] ) is then given by vectors \ { | l@l j_1 j_2 & + k_1 k_2 & m_1 m_2 > , j_i = 0 , 12 , , |k_i| , |m_i| j_i , i = 1,2 } .
[ sl4basis_1 ] similarly , as in the @xmath116 case , the gell - mann formula yields : t__1_2 = d^11_00_1_2 + [ c_2(so(4 ) ) , d^11_00_1_2],= 0,1;i=1,2 .
[ gell - mannsl(4_1)]the corresponding matrix elements of the @xmath117 operators read : & & l@l j_1j_2 + k_1k_2 & m_1m_2 | t__1_2 | l@l j_1 j_2 & + k_1 k_2 & m_1 m_2 + & & = ( + i ( j_1(j_1+1)+j_2(j_2+1)-j_1(j_1 + 1)-j_2(j_2 + 1 ) ) ) + & & c c c c. however , analogously as in the @xmath95 case , the commutation relations of the noncompact operators close correctly only for the subspaces ( [ sl4basis_1 ] ) characterized by @xmath118 , and thus for @xmath119 , @xmath120 being integers . the corresponding @xmath96 representations are the multiplicity free ones , while the representation spaces are hilbert spaces over the @xmath121 coset space ( @xmath122 denoting the double cover of @xmath123 ) .
additional @xmath96 multiplicity free representations can be obtained by working in the @xmath110 representation spaces characterized by the subgroup chain @xmath124 , with basis vectors : \ { | c j_1 j_2 + j + m > , j_i = 0 , 12 , ;
|j_1 -j_2| j j_1 + j_2 ; |m| j ; i = 1,2 } .
[ so4basis_2 ] the @xmath96 representation space is then defined by basis vectors : \ { | c@c j_1 j_2 & + k & j + k & m > , j_i = 0 , 12 , , |j_1 - j_2| k , j j_1 + j_2 , |m| j } .
[ sl4basis_2 ] the corresponding gell - mann formula expression for the ( noncompact ) shear generators reads t_j = d - [ c_2(so(4 ) ) , d ] , j=0,1,2 ; ||j , [ gell - mannsl(4_2)]and yields the correct commutation relations in the subspace of ( [ sl4basis_2 ] ) for @xmath125 , only .
the corresponding @xmath96 representations are the multiplicity free ones , defined in symmetric spaces over the @xmath126 coset space .
these representations have only the `` diagonal '' @xmath111 content , since the condition @xmath127 implies @xmath128 . note that , neither the representations with multiplicity , nor the spinorial multiplicity free @xmath96 representations ( cf .
@xcite ) can be obtained by making use of the expressions given by ( [ gell - mannsl(4_1 ) ] ) or ( [ gell - mannsl(4_2 ) ] ) .
it is clear from these examples , that the gell - mann formula has a limited scope when applied to the @xmath16 , @xmath46 cases .
it is neither valid as an operator expression , nor it holds for a generic @xmath16 , @xmath46 representation space .
one can make use of the gell - mann formula in some subspaces of the most general representation space , only .
more precisely , the expressions ( [ gell - mannsl(3 ) ] ) , ( [ gell - mannsl(4_1 ) ] ) and ( [ gell - mannsl(4_2 ) ] ) yield the noncompact algebra operators in the symmetric spaces over @xmath129 , @xmath130 and @xmath126 , respectively . note , that this result is in agreement with the hermann theorem @xcite stating that the gell - mann formula certainly works in the symmetric spaces @xmath131 if @xmath24 is a simple compact subgroup from a cartan decomposition of the starting group @xmath30 ( here @xmath132 ) and if there exists some @xmath133 which is invariant under the action of @xmath134 . unfortunately
, this theorem does not give the necessary conditions for the gell - mann formula to hold .
it turns out that we can shed some light on the question of a validity of the gell - mann formula in the @xmath16 , @xmath135 algebra case . in doing that , we recall first some relevant facts about the @xmath17 , @xmath135 group manifold and its representations .
let us consider an action of the @xmath17 representation on the lower left - hand side labels ( quantum numbers ) of the basis vectors given by ( [ naturalbasis ] ) .
first , let us make use of the operator k _
g^ d_^m _ , = 1,2, ,n
[ koperator]where @xmath136 is the cartan metric tensor of @xmath17 ) . the @xmath137 operators behave exactly as the rotation generators @xmath138 , it is only that they act on the lower left - hand side indices of the basis ( [ naturalbasis ] ) : = = _ jj
c. [ kaction ] due to the fact that mutually contragradient @xmath17 representations are equivalent , the @xmath137 operators are directly related to the `` left '' action of the @xmath17 subgroup on @xmath139 : @xmath140 .
the operators @xmath137 and @xmath138 mutually commute , however , the corresponding casimir operators match , i.e. @xmath141 .
whereas the @xmath16 operators @xmath59 are invariant under this left action , the shear operators @xmath63 , constructed by using the gell - mann formula prescription , are not .
the transformation properties of the shear operators @xmath63 are inherited from the corresponding contracted operators @xmath75 , which have additional nontrivial transformation properties described by the @xmath137 operator labels ( of the right - hand side vector of the @xmath17 matrix in ( [ uisd ] ) ) .
consequently , a commutator of two such shear operators has a nontrivial properties under the @xmath142 group ( generated by the @xmath137 operators ) .
it is not an @xmath142 scalar ( unlike @xmath138 ) , and therefore not bound to close upon the @xmath16 commutation relations .
this is precisely the reason why the shear generators ( [ gell - mannsl(3 ) ] ) , ( [ gell - mannsl(4_1 ) ] ) and ( [ gell - mannsl(4_2 ) ] ) do not satisfy the commutation relations ( [ ttcommutator ] ) in a generic representation space over the @xmath17 group manifold . in particular cases ,
it is possible to make a restriction to such subspace of the representation space , where only the invariant part of the @xmath143 $ ] commutator survives that is proportional to the @xmath2 operators representation in that subspace . for example , in the @xmath95 case above , a restriction was made to the subspace of @xmath108 , where the @xmath143 $ ] commutator piece proportional to @xmath144 vanishes ; likewise for the commutators of the shear generators ( [ gell - mannsl(4_1 ) ] ) and ( [ gell - mannsl(4_2 ) ] ) .
the above analysis raises the question of whether it is possible to modify the gell - mann formula by adding some terms proportional to the generators of the left @xmath142 group , that cancel the unwanted terms and `` fix '' the @xmath16 algebra commutation relations .
such a generalization of the gell - mann formula in the @xmath95 case can be read out directly from the known matrix elements of the @xmath95 representations with multiplicity @xcite : t _ = d^2_0 + [ c_2(so(3 ) ) , d^2_0 ] + i(d^2_2-d^2_-2)k_0 + ( d^2_2+d^2_-2 ) , [ generalized sl3 gm formula]@xmath102 , and where @xmath73 and @xmath145 are the @xmath95 group representation labels .
the additional terms to the `` original '' gell - mann formula secure that the @xmath146 operators satisfy the commutation relation ( [ ttcommutator ] ) in the entire representation space . note that , there are two @xmath95 representation labels @xmath73 and @xmath145 , matching the algebra rank , contrary to the case of the original gell - mann formula whose single free parameter can not account for the entire representation labeling .
the additional label @xmath93 ( @xmath147 , @xmath148 ) describes the nontrivial multiplicity in @xmath92 . the generalized expression ( [ generalized sl3 gm formula ] ) contains the original formula ( [ gell - mannsl(3 ) ] ) as a special case : by restricting the representation space to the subspace of @xmath104 , and choosing @xmath149 one arrives at the multiplicity free representations that were obtained by using the expression ( [ gell - mannsl(3 ) ] ) .
moreover , the generalized gell - mann formula allows one to obtain some @xmath95 multiplicity free representations that can not be reached by making use of the original formula ( [ gell - mannsl(3 ) ] ) .
for example , the choice @xmath150 , and @xmath151 @xcite in a new basis of vectors ( linear combinations of basis vectors with different @xmath93 values ) , c \{| c 12 + m > = | c@ c 12 & + 12 & m > + | c@ c 12 & + -12 & m > , + | c 52 + m > = | c@ c 52 & + 52 & m > + | c@ c 52 & + 12 & m > + | c@ c 52 & + -12 & m > + | c@ c 52 & + -52 & m > , + | c 92 + m > = | c@ c 92 & + 92 & m > + | c@ c 92 & + 52 & m > + | c@ c 92 & + 12 & m
> + | c@ c 92 & + -12 & m > + | c@ c 92 & + -52 & m > + | c@ c 92 & + -92 & m > , } . yields a representation space without multiplicity that is closed under the action of the @xmath146 operators .
this is a basis of a spinorial @xmath95 unitary irreducible ( @xmath92 content are half - odd integers ) representation space , where the original gell - mann formula does not apply .
the gell - mann formula can similarly be generalized in the case of the @xmath96 algebra .
again , by extracting from the known matrix elements of the @xmath96 representations with multiplicity @xcite , we find : & & t__1_2 = i(d^11_00_1_2 + [ c_2(so(4 ) ) , d^11_00_1_2 ] + & & + _ 1 ( d^11_11_1_2 + d^11_-1 - 1_1_2 ) + ( d^11_11_1_2 - d^11_-1 - 1_1_2)(k^10_00+k^01_00 ) + & & + _ 2 ( d^11_-11_1_2 + d^11_1 - 1_1_2 ) + ( d^11_-11_1_2 - d^11_1 - 1_1_2)(k^10_00-k^01_00 ) ) , [ generalized gell - mannsl(4_1 ) ] where @xmath152 . as the rank of the @xmath153 algebra is three , there are precisely three representation labels @xmath73 , @xmath154 , and @xmath155 ( if complex , only three real are independent ) . as in the @xmath156 case ,
the generalized formula reduces , for certain values of the labels , in a representation subspace defined by @xmath118 to the original gell - mann formula ( [ gell - mannsl(4_1 ) ] ) .
it is not a straightforward matter to see that the formula ( [ gell - mannsl(4_2 ) ] ) follows from the generalized gell - mann formula .
however , the generalized gell - mann formula for the @xmath96 algebra can be expressed in an equivalent form as follows : & & t_j = _ 1 d - [ c_2(so(4 ) ) , d ] + & & + _ 2 d + i d ( k+k ) - i d ( k+k ) [ generalized gell - mannsl(4_2 ) ] + & & + _ 3 ( d + d ) + i(d - d ) ( k+k ) , where @xmath157 , @xmath158 , and provided @xmath159 , @xmath160 , and @xmath161 .
derivation of ( [ gell - mannsl(4_2 ) ] ) is now obvious for @xmath162 . in a parallel to the @xmath156 algebra case
, the generalized gell - mann formula for the @xmath96 algebra case holds for all representations , irrespectively of the @xmath163 representations multiplicity .
it is now straightforward to write down the generalized gell - mann formula expressions for the @xmath164 and @xmath163 algebras , thus obtaining the generators of the @xmath165 and @xmath166 factor groups .
@xmath164 : @xmath167 = @xmath168 , @xmath169 , where @xmath146 is given by ( [ generalized sl3 gm formula ] ) .
@xmath163 : @xmath170 = @xmath171 , @xmath172 , i.e. @xmath173 , @xmath174 , @xmath175 , where @xmath176 and @xmath177 are given by ( [ generalized gell - mannsl(4_1 ) ] ) and ( [ generalized gell - mannsl(4_2 ) ] ) , respectively .
we have shown above that the original gell - mann formula expressions for the @xmath16 and @xmath1 algebras do not satisfy the @xmath143 $ ] commutation relations ( [ ttcommutator ] ) on a pure algebraic level . however , by setting the gell - mann formula existence question into a group representation framework , it is possible to restrict representation spaces and thus achieve a closure of the @xmath143 $ ] commutator ( @xmath178 being given by the gell - mann formula expression ) .
moreover , we have shown , by extracting information from the known results about the @xmath16 , @xmath46 representations , that there exist a generalization of the gell - mann formula for @xmath16 , @xmath46 , which is valid for all representation spaces .
an important role , in that process , was played by the @xmath24 operator ( [ koperator ] ) . in the following , we make use of the gell - mann formula generalization for @xmath16 , @xmath46 , and a peculiarity of the @xmath57 algebra representation labels to follow the @xmath111 @xmath179 @xmath180 labeling features .
let us recall first some basic @xmath57 algebra representation notions .
the @xmath57 algebra is of rang two , and its irreducible representations are labeled by a pair of labels @xmath181 , resembling the @xmath111 labeling .
the complete labeling of the representation space vectors can be achieved by making use of the subalgebra chain : @xmath57 @xmath182 @xmath111 @xmath179 @xmath180 @xmath182 @xmath183 .
the basis of the @xmath57 algebra representation space can be taken as in @xcite : \ { | cc j_1 & j_2 + j_1 & j_2 + m_1 & m_2 > , j_i = 0 , 12 , ; j_1 j_2 ; |m_i| j_i , i=1,2 } .
[ so5basis]the admissible values of @xmath119 and @xmath120 , within an irreducible representation @xmath184 are given in @xcite .
now , the basis of the @xmath57 algebra , i.e. the @xmath6 group , representation space vectors ( [ naturalbasis ] ) is given as follows : \ { | cccc j_1 & j_2 & & + k_1 & k_2 & j_1 & j_2 + k_1 & k_2 & m_1 & m_2 > } .[sl5basis ] the ten @xmath57 algebra operators , generating the adjoint representation of @xmath6 , transform , in notation ( [ so5basis ] ) , under the representation @xmath185 .
their @xmath111 subalgebra representation content is : @xmath185 @xmath186 @xmath187 @xmath188 @xmath189 @xmath188 @xmath190 .
the shear operators transform under the @xmath191-dimensional @xmath57 irreducible representation @xmath192 of @xmath57 which contains @xmath193 , @xmath194 and @xmath195 representation upon reduction to @xmath111 : @xmath196 the original gell - mann formula is not applicable , again , in the whole space spanned by ( [ sl5basis ] ) , but only in the symmetric spaces @xmath197 and @xmath198 , with appropriate choices of the @xmath199 labels for the contracted operators @xmath200 . neither representations with multiplicity , nor spinorial representations of the @xmath3 algebra
can be obtained in this way .
we made an educated guess , based on the structure of the generalized gell - mann formula for the @xmath96 algebra , when adding possible terms to the generalized gell - mann formula in the @xmath3 case .
we omit the terms proportional to @xmath201 altogether , as well as the terms proportional to @xmath202 , @xmath203 , @xmath204 , and @xmath205 .
the @xmath143 \subset m$ ] commutation relations condition ( [ ttcommutator ] ) , together with a knowledge of the @xmath57 clebsch - gordan coefficients for @xmath206 @xcite , finally yields the sought for generalized gell - mann formula expression for the @xmath3 algebra shear operators : rl t = & _ 1 d + i[c_2(so(5 ) ) , d ] + & + i(_2 d + 12[c_2(so(4)_k ) , d ] + & - d ( _ 1 + k - k)- d ( _ 1 - k + k ) + & + d ( _ 2 + k + k)+ d ( _ 2 - k - k ) ) , [ sl5ggm ] where @xmath207 , @xmath208 , @xmath209 , the representation labels @xmath210 and @xmath155 are arbitrary ( complex ) parameters ( four real are independent ) , and @xmath211 denotes the quadratic casimir operator of the left action @xmath212 algebra .
naturally , the same result ( [ sl5ggm ] ) is obtained , when searching for the generalized gell - mann formula expression , by starting with all possible additional terms proportional to @xmath213-functions and demanding ( [ ttcommutator ] ) , though by a much more tedious calculation .
the @xmath4 algebra elements , as given by the generalized gell - mann formula , are the @xmath214 operators ( [ maction ] ) , generating the @xmath5 subgroup of the @xmath215 group , and the @xmath216 operators ( [ sl5ggm ] ) , generating the @xmath217 factor group .
contrary to the generalized gell - mann formula for the @xmath16 , @xmath135 algebras , where we started from the known matrix elements of the shear operators , here in the case of the @xmath3 algebra , we are in a position to obtain , for the first time , the shear operators matrix elements for a generic representation space starting from the generalized gell - mann formula expression ( [ sl5ggm ] ) .
the matrix elements of the @xmath3 shear ( noncompact ) operators read : c = c + ( ( _ 1 + i(j_1(j_1 + 2 ) + j_2(j_2 + 1 ) - j_1(j_1 + 2 ) - j_2(j_2 + 1 ) ) ) c + + i(_2 + k_1(k_1 + 1 ) + k_2(k_2 + 1 ) - k_1(k_1 + 1 ) - k_2(k_2 + 1 ) ) c + - i(_1 + k_1 - k_2 ) c - i(_1 - k_1 + k_2 ) c + + i(_2 + k_1 + k_2 ) c + i(_2 - k_1 - k_2 ) c ) , [ tmatrixgglsl(5 ) ] where @xmath218 is the dimension of the @xmath57 irreducible representation characterized by @xmath219 @xcite . to sum up , the matrix elements of the ( noncompact ) shear operators @xmath178 ( [ tmatrixgglsl(5 ) ] ) , together with the known matrix elements of the ( compact ) @xmath2 operators @xmath214 ( [ maction ] ) , define an action of the @xmath3 algebra on the basis vectors ( [ sl5basis ] ) of representation spaces of the maximal compact subgroup @xmath6 of the @xmath220 group .
this result is general due to a corollary of harish - chandra @xcite that explicitly applies to the case of the @xmath3 algebra .
the gell - mann formula , beyond the case of ( pseudo ) orthogonal algebras , is not valid in general as a pure algebraic expression .
its applicability can be broaden , by utilizing it in certain cases , provided some lie algebra representation conditions are met . as for the @xmath16 algebras ,
contracted w.r.t .
their @xmath2 subalgebras , the algebraic expression of the gel - mann formula matters generally for the multiplicity free representations only .
it was demonstrated in this work that one can generalize the gell - mann formula for the @xmath16 , @xmath221 algebras to be valid for a generic representation space . a brief account required by a description of the @xmath16 , @xmath221 representation spaces , that heavily depends on the corresponding group topology properties and the @xmath2 subgroup multiplicity , is given . in the @xmath156 and @xmath153 cases we inferred , starting from a suitable existing expressions of the algebra operators representations with non - trivial multiplicity , a generic generalized gell - mann formulas .
these formulas offer a new starting point for mathematical physics investigations , as well as for various physical applications . by analyzing the structure of the gell - mann formula for the @xmath16 , @xmath221 cases , and making use of the specific features of the @xmath57 algebra representations
, we obtained the generalized gell - mann formula for the @xmath3 and @xmath4 cases .
note that this formula , that is valid for all representation hilbert spaces , is characterized precisely by a right number ( algebra rank ) of the representation labels . as a first and most precious application , based on this generalized gell - mann formula
, we obtained for the first time a closed form of the generic expressions of all matrix elements of the @xmath3 noncompact generators .
a distinct feature of our generalized gell - mann formula approach is that the resulting expression goes beyond the standard notion of a deformation of the contracted algebra , as it depends on additional operators not belonging , however directly related , to the contracted algebra .
due to this fact , our generalization of the gell - mann formula is remarkable simple ( compared to complicated polynomial expressions appearing in some other approaches to generalize the gell - mann formula ) , nevertheless establishing a direct relation between representations of the contracted and original algebras .
this work was supported in part by mntr , project-141036 .
one of us , i.s . , would like to acknowledge hospitality and useful discussions at the institute for nuclear research and nuclear energy in sofia ( bulgaria ) during his visit as early stage researcher supported by the fp6 marie curie research training network `` forces - universe '' mrtn - ct-2004 - 005104 . | the so called gell - mann formula expresses the lie algebra elements in terms of the corresponding inn - wigner contracted ones . in the case of @xmath0 and @xmath1 algebras contracted w.r.t .
@xmath2 subalgebras , the gell - mann formula is generally not valid , and applies only in the cases of some algebra representations .
a generalization of the gell - mann formula for @xmath3 and @xmath4 algebras , that is valid for all representations , is obtained in a group manifold framework of the @xmath5 and/or @xmath6 group .
pacs : 02.20.sv , 02.20.qs ; msc2000 : 20c33 , 20c40 ; |
the interaction of a neutral atom with a material surface is a problem profusely addressed in the literature @xcite . in most of the approaches
the atom is taken in a stationary state with respect to ( w.r.t ) the time of observation . for distances greater than the relevant atomic transition wavelengths
this interaction is referred to as _ retarded casimir - polder ( cp ) interaction _ , while for distances much shorter than those wavelengths it is referred to as _ van der waals _ or _ non - retarded cp interaction_. it is well known in cavity - qed that vacuum rabi oscillations are generated between two non - degenerate atomic states as a result of the strong coupling of the atom to a single resonant cavity mode @xcite . in this letter
we show that rabi oscillations between two degenarate atomic states can also be induced by the cp interaction of the atom with a reflecting surface .
in contrast to the cavity - qed vacuum rabi oscillations , all the vacuum modes participate on the resultant _ cp induced rabi oscillations_. the net effect of the cp interaction on non - degenerate atomic states is an additive level shift @xcite . in particular , the resonant component of the cp interaction gives rise to an effective renormalization of the transition frequencies @xcite . on the contrary , on ( quasi)degenerate atomic states
the cp interaction may cause the admixture of neighboring states ( eg .
it is in this case that , under the conditions investigated in this letter , the atom oscillates coherently between the two ( quasi)degenerate states with a well defined rabi frequency proportional to the strength of the cp interaction . in order to prove this we study the time - evolution of the wave function of an atom initially prepared in a coherent superposition of ( quasi)degenerate states .
we illustrate this phenomenon with the computation of the cp induced rabi frequency on a @xmath0rb atom as it interacts with a reflecting surface and oscillates between two zeeman sublevels of its ground state .
the resultant variation of atomic angular momentum is provided by the vacuum photons which mediate the interaction with the surface .
in order to show the rabi oscillations induced by the cp interaction of a reflecting surface on an atom , we compute the time evolution operator , @xmath1 , restricted to the two atomic states of our interest , @xmath2 .
we assume @xmath3 .
later , we will consider the limiting case in which @xmath2 forms a degenerate doublet w.r.t .
the time of observation , @xmath4 . in the following ,
we describe the fundamentals of the calculation .
an atom in free space with eigenstates @xmath5 is described by the free hamiltonian given by @xmath6 while the hamiltonian of the free electromagnetic ( em ) field is @xmath7\nonumber\\ & = & \sum_{\mathbf{k},\mathbf{\epsilon}}\hbar\omega(a_{\mathbf{k},\mathbf{\epsilon}}a^{\dagger}_{\mathbf{k},\mathbf{\epsilon}}+1/2),\end{aligned}\ ] ] where @xmath8 and @xmath9 are the electric and magnetic vacuum fields respectively , @xmath10 is the photon frequency , and the operators @xmath11 and @xmath12 are the creation and annihilation operators of photons with momentum @xmath13 and polarization @xmath14 respectively . in order to simplify matters the center of mass of the atom is considered fixed at a given location @xmath15 such that , in application of the born - oppenheimer approximation , we can disregard the dynamics of the external atomic degrees of freedom .
we expand the interaction potential of the atom with the em field in a multipolar series and truncate that series at the magnetic dipole order @xcite , @xmath16 in these equations @xmath17 and @xmath18 are the atomic electric and magnetic dipole operators respectively and @xmath19 , @xmath20 are the electric and magnetic vacuum field operators at the location of the atom . in the following
we adopt an approach based on the fluctuation - dissipation theorem ( fdt ) according to which the atom is treated as a quantum system while the conducting surface is regarded as a classical object concerning its interaction with the em field , and we restrict ourselves to zero temperature .
the quantum electric and magnetic fields in the interaction potential @xmath21 , @xmath8 , @xmath9 , are schrdinger operators and the em vacuum state , @xmath22 , is dressed by the electric current fluctuations on the surface @xcite .
by dressed we mean that the linear response of the em field i.e . ,
its green function , includes the scattering with the surface .
the electric and magnetic fields can be decomposed respectively in terms of @xmath23 modes as @xmath24,\nonumber\\ \mathbf{b}(\mathbf{r})&=&\int_{0}^{\infty}\textrm{d}\omega[\hat{\mathbf{b}}(\mathbf{r};\omega)+\hat{\mathbf{b}}^{\dagger}(\mathbf{r};\omega)],\nonumber\end{aligned}\ ] ] where @xmath25 @xmath26 $ ] and @xmath27 @xmath28 $ ] are the creation and annihilation electric ( magnetic ) field operators of photons of energy @xmath29 at the position of the atom , @xmath15 .
the quadratic vacuum fluctuations of @xmath25 and @xmath30 satisfy the fdt relations at zero temperature @xcite , @xmath31},\nonumber\\ \langle\tilde{0}|\hat{\mathbf{b}}(\mathbf{r};\omega)\otimes\hat{\mathbf{b}}^{\dagger}(\mathbf{r}';\omega)|\tilde{0}\rangle&=&\frac{\hbar}{\pi\epsilon_{0}c^{2}}\nonumber\\ & \times&\im{[\nabla_{r}\wedge\mathbb{g}(\mathbf{r},\mathbf{r}';\omega)\wedge\nabla_{r'}}],\nonumber\end{aligned}\ ] ] where both @xmath15 and @xmath32 lie on the r.h.s . of the surface where the atom is placed
, @xmath33 denotes the imaginary part and @xmath34 is the green function of the maxwell equation for the em field , @xmath35\mathbb{g}(\mathbf{r},\mathbf{r}';\omega ) = \delta^{(3)}(\mathbf{r},\mathbf{r}')\mathbb{i},\nonumber\\ z , z'&>&0.\label{maxwelleq}\end{aligned}\ ] ] in this equation @xmath36 and @xmath37 are the relative electric permittivity and magnetic permeability tensors respectively , and the atom s position lies to the right of the surface , @xmath38 [ fig.[fig3](@xmath39 ) ] .
the green function can be decomposed into a free - space component and a scattering component .
the contribution of the free space term to the casimir energy is the ordinary free - space lamb - shift that we consider included in the bare values of the atomic transition frequencies . )
atom ( in red ) placed at @xmath15 in front of a plate perpendicular to the @xmath40 axis , with relative permittivity @xmath41 and permeability @xmath42 different to unity .
( @xmath43 ) feynman diagram of a multiple reflection process which contributes to @xmath44 in ( quasi)degenerate conditions , @xmath4 .
this diagram is proportional to the product of two one - particle - irreducible non - additive phase shift factors and one additive phase shift factor , @xmath45 dissipative factors are omitted here for brevity .
( @xmath46 ) feynman diagram of a multiple reflection process which contributes to @xmath47 in ( quasi)degenerate conditions .
this diagram is proportional to the product of three one - particle - irreducible non - additive phase shift factors , @xmath48 .
the states @xmath49 are intermediate atomic states.,width=294,height=313 ]
we consider @xmath21 as a small perturbation w.r.t . @xmath50 and compute the time evolution operator in the subspace @xmath2 , @xmath1 , as a power series on @xmath21 . in the first place
, the unperturbed time - evolution operator of the atomic states @xmath5 , including @xmath51 and @xmath52 , is @xmath53 on the other hand , free one - photon states evolve according to the evolution operator associated to @xmath54 , @xmath55 if the atomic wave function at @xmath56 is @xmath57 , at a later time @xmath58 it reads @xmath59|g\rangle\nonumber\\ & + & [ _ { e}\textrm{u}_{g}(t)a_{g}(0)+\ : _ { e}\textrm{u}_{e}(t)a_{e}(0)]|e\rangle.\nonumber\end{aligned}\ ] ] straightforward application of time - dependent perturbation theory @xcite projected on the subspace @xmath60 yields the following expression for the time - evolution operator , @xmath61\cdot[\mathbb{u}^{at}_{0}\otimes\mathbb{u}^{\gamma}](t)\nonumber\\ & \cdot & \textrm{t - exp}\int_{0}^{t}\textrm{d}t[\mathbb{u}^{at}_{0}\otimes\mathbb{u}^{\gamma}]^{\dagger}(t)\cdot w\cdot[\mathbb{u}^{at}_{0}\otimes\mathbb{u}^{\gamma}](t)\nonumber\\ & \cdot&[(|g\rangle\langle g|+|e\rangle\langle e|)\otimes|\tilde{0}\rangle\langle\tilde{0}|],\label{texp}\end{aligned}\ ] ] where the operators flanking @xmath62\cdot$]t - exp@xmath63 account for the projection onto the subspace @xmath64 . at leading order
the above expression contains terms proportional to the level shifts of the states @xmath51 and @xmath52 , @xmath65 and @xmath66 respectively ; and crossed terms proportional to @xmath67 and @xmath68 , where the tilded states are states dressed by the action of @xmath21 .
we will refer to the first kind of terms as _ additive _ and to the latter kind as _ non - additive _ , since only the integration of the former consists of their simple addition to the atomic energy levels .
it can be verified that in the non - degenerate case , @xmath69 , the contribution of the non - additive terms is of the order of @xmath70 and thus neglegible , so that the net effect of the cp interaction reduces here to a shift of the energy levels of the atom in the diagonal components of @xmath71 . on the contrary , in the ( quasi)degenerate case , @xmath4
, the non - additive terms become as relevant as the additive ones . the diagrams which weight the most in the perturbative series , both in the diagonal and in the off - diagonal components of @xmath71 , are those in which the atom transits through intermediate states with an only virtual photon before arriving repeatedly to the states @xmath72 and/or @xmath73 with no photon . typical diagrams which contribute to diagonal and off - diagonal components of @xmath71
are depicted in figs.[fig3](@xmath43 ) and ( @xmath46 ) respectively .
their summation yields series of products of one - particle - irreducible factors , with real and imaginary ( dissipative ) components . following the nomenclature explained earlier , we distinguish between additive terms , @xmath74 , @xmath75 ; and non - additive terms , @xmath76 , @xmath77 , with @xmath78 where we have used @xmath79 , @xmath80 .
analogous expressions hold for @xmath81 , @xmath82 , @xmath83 and @xmath84 with the substitution @xmath85 .
the single superscript @xmath86 in the expressions for @xmath87 and @xmath88 denotes the reference frequency for the transitions within the sums .
double subscripts , @xmath89 , @xmath90 , @xmath91 or @xmath92 , denote the bra and ket states in the quantum amplitudes . from the diagrams of fig.[fig3 ]
we observe that each factor @xmath93 flips the state of the atom from @xmath52 to @xmath51 , while each transposed factor produces an opposite flip .
this is analogous to the action of the two raman lasers which drive the rabi oscillations of an atom @xcite . in the rotatory frame , taking the rotatory - wave - appriximation , the well - known effective rabi hamiltonian reads , @xmath94 the effect of the factors @xmath95 and @xmath96 is analogous to that of @xmath97 and @xmath98 respectively , except for the fact that in the latter case the dissipative terms break the time reversal symmetry . as a matter of fact , @xmath71 can be recast in the familiar form @xcite , @xmath99,\nonumber\\ _
{ g}\textrm{u}_{e}(t)&=&-ie^{-i(\tilde{\omega}_{g}-\tilde{\delta}/2)t}\frac{|\omega|}{\omega_{r}}\sin{(\omega_{r}t/2)},\nonumber\\ _ { e}\textrm{u}_{e}(t)&=&e^{-i(\tilde{\omega}_{e}+\tilde{\delta}/2)t}[\cos{(\omega_{r}t/2)}+i\frac{\tilde{\delta}}{\omega_{r}}\sin{(\omega_{r}t/2)}],\nonumber\\ _ { e}\textrm{u}_{g}(t)&=&-ie^{-i(\tilde{\omega}_{e}+\tilde{\delta}/2)t}\frac{|\omega|}{\omega_{r}}\sin{(\omega_{r}t/2)},\label{ur}\end{aligned}\ ] ] with the parameters so defined , so defined is not the complex conjugate of @xmath100 because of the damping factors . ]
we propose here a simple setup which illustrates the phenomenon of interest .
we study the oscillations of a @xmath0rb atom between two degenerate zeeman sublevels of its ground state , @xmath102 and @xmath103 , at zero temperature in the vicinity of a perfectly reflecting metallic surface .
the cp induced level shifts of zeeman states on a trapped electron have been studied recently in ref.@xcite . for reasons that will be clear later , we must take the axis of quantization parallel to the surface .
if the surface is perpendicular to the axis @xmath40 , we choose the quantization axis along the @xmath104 direction .
the non - additive phase shift term is largely dominated by the electric dipole interaction . making use of the fdt we find , as a function of the dyadic green s function @xcite , @xmath105\cdot\langle i|\mathbf{d}|e\rangle\},\label{dege2}\end{aligned}\ ] ] where the first term corresponds to the off - resonant component ( @xmath106 is the complex frequency ) and
the second one is the resonant component . in the basis of hyperfine states labeled with the quantum numbers @xmath107 and @xmath108 , it is convenient to express the dipole moment operator in the spherical basis , @xmath109 where all the dipole transition matrix elements are real numbers . in this basis
we find , @xmath110 where the frequency and position dependence of the dyadic green function components have been omitted .
the off - diagonal terms of the green function are identically zero .
we find for a perfectly conducting reflector @xcite , @xmath111 straightforward application of selection rules @xcite in eq.([gs ] ) implies that only the term of eq.([gs ] ) proportional to @xmath112 may survive , which explains why the axis of quantization must be taken parallel to the surface . )
feynman diagram of the non - additive term @xmath113 .
initially , virtual photons of left handed circular polarization , @xmath114 , are created at the time the atom is in state @xmath51 .
later , @xmath114 photons get reflected off the surface turning into right handed circularly polarized , @xmath115 .
finally , @xmath115 photons are annihilated at the time the atom gets to state @xmath52 .
( @xmath43 ) schematic representation of the virtual transitions which drives a @xmath0rb atom through coherent oscillations between two zeeman sublevels of its ground state , @xmath102 and @xmath103 .
virtual photons of opposite circular polarization mediate the transition of @xmath51 and @xmath52 , @xmath116 and @xmath117 transitions respectively , to a series of common intermediate states states @xmath118 in ( @xmath39) which are hyperfine zeeman sublevels of @xmath119 ( in green ) and @xmath120 ( in red ) with @xmath121 .
the hyperfine levels with @xmath122 are considered as the reference levels .
the transition frequencies are @xcite , @xmath123thz , @xmath124thz , and the hyperfine intervals , @xmath125ghz , @xmath126ghz , @xmath127ghz.,width=313,height=453 ] in the calculation of @xmath113 it suffices to consider the d1 and d2 line transitions , @xmath128 and @xmath129 respectively . in fig.[fig4](@xmath43 ) it appears depicted all possible virtual transitions from the states @xmath51 and @xmath52 which contribute to @xmath113 and @xmath82 . in all of them the intermediate states
are hyperfine zeeman sublevels of @xmath119 and @xmath120 with @xmath121 , and the virtual photons created and annihilated during the transitions possess opposite circular polarization , @xmath130 and @xmath131 respectively [ see fig.[fig4](@xmath39 ) ] .
the latter implies a net variation of @xmath132 in the atomic angular momentum .
important is the fact that a non - zero value of @xmath113 needs the inclusion of retardation in the electric casimir interaction , even at submicron distances . the reason being that the following relation holds , for @xmath133 , @xmath134 therefore , non - zero hyperfine intervals in the states @xmath135 , @xmath136 , are necessary for @xmath113 not to vanish . inserting the formula of eq.([gs ] ) into eq.([dege ] ) we get , @xmath137 where @xmath138 and @xmath139 , with @xmath140 being the transition frequency from the hyperfine level @xmath107 of @xmath141 to the states @xmath51 , @xmath52 . at
leading order in @xmath142 we find @xmath143,\label{derrier}\end{aligned}\ ] ] where @xmath144 are the reduced dipole moment matrix elements of the d@xmath145 and d@xmath146 transition lines respectively . given by eqs.([dege],[dege2],[derrier ] ) in the ranges @xmath147 @xmath148nm@xmath149 nm and @xmath150 @xmath151m@xmath152 m .
the continuous lines correspond to the case where @xmath51 and @xmath52 belong to the ground state @xmath153 .
the dashed lines correspond to the case where @xmath51 and @xmath52 belong to the rydberg state @xmath154.,width=325 ] tabulated data of all the unknowns in eq.([derrier ] ) can be found in ref.@xcite .
fig.[fig5 ] contains the plot of @xmath155 as a function of the distance to the surface .
in particular , for @xmath51 and @xmath52 in @xmath153 , the cp induced rabi frequency is @xmath156hz at 40 nm and @xmath156mhz at 270 nm approximately .
it is worth stressing that the symmetry of the setup yields the same additive shifts for the states @xmath51 and @xmath52 , keeping this way their degeneracy .
in particular , the electric cp interaction provides a level shift between 0.2mhz and 70 mhz within the interval 40nm@xmath157 nm , which guarantees the perturbative nature of the calculation ; and the magnetic interaction of the net magnetic moments of the states @xmath51 and @xmath52 , @xmath158 respectively , provides them with a common shift in the range 0.01hz-4.3hz which raises them over the state @xmath159 .
this might be a potential problem for the observation of rabi oscillations since both @xmath51 and @xmath52 can decay into the state @xmath159 through an m1 transition with the same rate , @xmath160 .
this rate depends on the actual conductivity of the reflecting surface through its skin depth , @xmath161 . for the reasonable assumption that @xmath162 nm we find @xcite , @xmath163hz(40nm@xmath164 at @xmath165 nm and @xmath166mhz(270nm@xmath164 at @xmath167 nm . in order to complete a rabi cycle , @xmath168 is necessary , which is guaranteed by the initial assumption @xmath162 nm within the interval 40nm@xmath157 nm .
in addition , we note that similar values are obtained for the non - additive dissipative terms , @xmath169 , which means that their contribution to @xmath170 can be discarded in good approximation .
finally , we observe that through a rabi semi - cycle , @xmath171 , the total angular momentum of the atom , @xmath172 , varies an amount @xmath132 along the quantization axis @xmath104 . for an atom in the state @xmath51 at @xmath56 , it holds @xmath173 .
more specifically , during a rabi semi - cycle the nuclear spin varies from @xmath174 to @xmath175 while the electronic spin varies in the oppossite sense , from @xmath176 to @xmath177 along @xmath104 , such that @xmath178 .
this variation of angular momentum is provided by the vacuum photons which mediate the interaction with the reflecting surface as explained in fig.[fig4](@xmath39 ) . at a theoretical level ,
conservation of total angular momentum implies that , in order to compensate for this variation , the angular momentum of the reflecting surface must vary an equal amount in opposite sense .
it explains why the axis of quantization must be taken parallel to the surface as the surface is seen as an infinite plane from the atom .
the rotational invariance of an infinite plane around its perpendicular axis the @xmath40 axis in our case would prevent otherwise the exchange of angular momentum between the surface and the atom in this direction .
an explicit calculation of the angular momentum of the surface would require a microscopical model for the interaction between the surface and the em field .
we finish this section with a remark on rydberg states . at first glance
it could be thought that greater values of @xmath170 would be obtained in the oscillation between zeeman sublevels of rydberg states with large principal number @xmath179 .
a study of the scaling behaviour as a function of @xmath179 of the dipole moment matrix elements and resonant frequencies which enter the expression of @xmath180 @xcite reveals that much greater values of @xmath170 can only be obtained at distances for which the cp interaction enters the retarded regime w.r.t .
the d@xmath145 and d@xmath146 transtions of the ground state , while it remains in the non - retarded regime w.r.t .
the transition between neighboring rydberg states .
this is illustrated in fig.[fig5 ] , where the rydberg state is chosen to be @xmath154 .
the calculation of @xmath170 is analogous to the one presented above for the zeeman sublevels of the ground state @xcite . at distances greater than @xmath181 nm , the values of @xmath170 are much larger for oscillations between zeeman sublevels of the rydberg state .
however , those values are still too small to be observed due to the short lifetimes of the excited states @xcite .
105 casimir h.b.g . and
polder d. _ _ phys .
rev.__**73 * * ( 1948 ) 360 .
wylie j.m . and sipe j.e .
a__**30 * * ( 1984 ) 1185 ; _ _ phys .
a__**32 * * ( 1985 ) 2030 .
scheel s. and buhmann s.y .
_ _ acta phys . slov.__**58 * * ( 2004 ) 675 .
and ducloy m. _ _ eur .
j. d__**40 * * ( 2006 ) 343 .
henkel c. , ptting s. and wilkens m. _ _ appl .
phys . b__**69 * * ( 1999 ) 379 .
buhmann s.y .
, trung - dung h. , knll l. and welsch d.g .
rev . a__**70 * * ( 2004 ) 052117 .
haroche s. _ cavity quantum electrodynamics _ in _ fundamental systems in quantum optics _
, les houches session liii , elsevier science publishers ( 1992 ) .
donaire m. _ _ phys .
rev . a__**85 * * ( 2012 ) 052518 .
ribeiro s. and scheel s. arxiv:1406.0172 ( 2014 ) .
power e.a . and thirunamachandran t.
lond . a__**372 * * ( 1980 ) 265 .
agarwal g.s .
rev . a__**11 * * ( 1975 ) 230 .
sakurai j.j .
_ advanced quantum mechanics _
additon - wesley ( 1994 ) .
jaynes e.t . and cummings f.w .
_ _ proc . ieee__**51 * * ( 1963 ) 89 .
r. bennett and c. eberlein , _ _ phys .
rev . a__**89 * * ( 2014 ) 042107 .
donaire m. _ et al .
_ in preparation .
steck d.a .
_ rubidium 87 d line data _ , available online at http://steck.us/alkalidata ( revision 2.0.1 , 2 may 2008 ) .
tauschinsky a. , newell r. , van linden van den heuvell h.b . and spreeuw r.j.c .
rev . a__**87 * * ( 2013 ) 042522 .
branden d.b .
_ et al . _
b__**43 * * ( 2010 ) 015002 . | we show that the casimir - polder interaction may induce coherent oscillations between degenerate atomic states .
we illustrate this effect by computing the casimir - polder induced rabi frequency on a @xmath0rb atom as it interacts with a reflecting surface .
the atom oscillates between two zeeman sublevels of its ground state undergoing a periodic exchange of angular momentum with the vacuum photons . |
_ `` the tension between giving away your information- to let people know what you have to offer- and charging them for it to recover your costs is a fundamental problem in the information economy . '' _ carl shapiro and hal r. varian in information rules : a strategic guide to the network economy , harvard business school press ( 1998 ) .
+ the advent of the internet as a big playground for resource sharing among selfish agents with diverse interests , and the emergence of web as a giant platform for hosting information has raised a plethora of opportunities for commerce , as well as , a plenty of new design , pricing and complexity problems .
one good example of a multi - billion dollar industry evolved as a consequence of web is the sponsored search advertising ( ssa ) , making fortunes for internet search giants such as google and yahoo ! , and has got tremendous attention in academia recently , due to various interesting research problems originated as a result of this continuously growing industry@xcite .
one of the most important concern for such an industry or in general for the information economy is the _
`` pricing problem''_. for example , for the goods like an ad - slot in the ssa which has no intrinsic value and is perishable , it is not clear what price should it be sold for .
similarly , for a digital good , where the cost of reproduction is negligible , the standard way of pricing based on the production cost does not work .
therefore , auctions are becoming a popular pricing mechanism in electronic commerce as they automatically adjust prices to market conditions , and specifically prices gets adjusted according to its value to the consumers rather than to the production costs .
auction theory has a pretty impressive literature @xcite- from the lovely vickrey auction to the sponsored search auctions@xcite and the auctions of digital goods@xcite .
the literature has so far focused mostly on the design of mechanisms that takes the revenue or the efficiency as a yardstick .
this is perfectly logical as these are two very important metrics from the viewpoint of the seller / auctioneer and the society respectively .
however , the scenarios where the _ capacity _ , which we define as _ `` the number of bidders the auctioneer wants to have a positive probability of getting the item '' _ , is a fundamental concern are ubiquitous in the information economy . for instance , in the sponsored search auctions or in online ad - exchanges , the true value of an ad - slot for an advertiser is inherently derived from the conversion - rate , which in turn depends on whether the advertiser actually obtained the ad - slot or not , which in turn depends on the capacity . in general , the same holds true for all information goods / digital goods . in the present paper , our goal is to first motivate _ capacity _ as a fundamental metric in designing auctions in the information economy and then to initiate study of such a design framework for some simple and interesting scenarios . in section [ mot ] , we motivate the capacity as a fundamental and interesting additional metric on the top of revenue / efficiency for mechanism design . in section [ single ] ,
we start with the _ capacity _ constrained framework for selling an indivisible item .
we propose a simple way to incorporate capacity constraints via designing mechanisms to sell probability distributions .
we show that such optimal probability distributions could be identified using a linear programming approach , objective being revenue , efficiency or a related function .
further , we define a quantity called _ price of capacity _ to capture the tradeoff between capacity and revenue / efficiency and derive upper bounds on it . in section [ ssa ] ,
we discuss the case of sponsored search auctions and also note that the auctioneer controlled probability spikes based auctions suggests a new model for sponsored search advertising , where a click is sold directly and not indirectly via allocating impressions . in section [ conc ] ,
we conclude with a list of research directions for future work , inspired by the present paper .
[ [ experience - goods ] ] * experience goods : * + + + + + + + + + + + + + + + + + + + + a bidder might not know her true valuation for the item unless she acquires it sometimes meaning that the true value is inherently derived from the actual acquisition of the item .
such a good is called an _ experience good_@xcite .
experience goods are ubiquitous in the information economy as clearly all information goods are experience goods@xcite . sometimes , a particular good
might also act as an experience for another good .
for example , a particular song from a singer might act as an experience for another song of that singer . therefore , in the auction of an experience good , the values for some of the bidders can be known to them as they might have experienced it from an earlier purchase of this or a related item , and for the other bidders the values are still unknown and they have to simply guess this value if they are participating in the auction . for the second kind of bidders , even if their values might turn out to be pretty high , their guesses might not be high enough to actually acquire the item when revenue / efficiency is the only goal .
moreover , they might be _ loss - averse_@xcite and would not bid a high value at all , due to the potential risk involved if the item does not turn out to be of high value to them . therefore , it is important that such bidders be given a chance to acquire the item , and consequently capacity becomes a fundamental metric in designing the mechanisms to achieve this goal .
[ [ two - fold - exploration - in - sponsored - search - auctions ] ] * two - fold exploration in sponsored search auctions : * + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + first , the ad - slots in the ssa are necessarily experience goods for a new advertiser and the estimates for the advertisers getting lower ranked slots is also generally poor as they hardly get any clicks .
the value of an ad - slot is derived from the clicks themselves ( i.e. rate of conversion or purchase given a click ) , and therefore , unless the bidder actually obtains a slot and receives user clicks , there is essentially no means for her to estimate her true value for the associated keyword .
second , even if all the true valuations are known to the corresponding bidders , for each bidder the ssa involves a parameter called _ quality score _ of the bidder which is defined as the expected clickability of the bidder for the associated keyword if she obtains a slot .
this parameter is also not known a priori and the auctioneer needs to estimate it .
certainly , a model that automatically allows one to estimate these key parameters ( i.e. click - through - rates and true values ) is desirable .
indeed , some mechanisms to incorporate explorations for estimating such important parameters has started to appear in literature@xcite .
capacity as an additional metric can provide a generic framework for designing such exploration based mechanisms . [ [ avoiding - over - exposure - in - online - advertising ] ] * avoiding over - exposure in online advertising : * + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + typically , the online ad - exchanges such as right media or doubleclick convince their advertisers that their ads will not be over exposed to users . one way of avoiding such an over - exposure could be via increasing capacity .
[ [ uncertainty - and - switching - costs ] ] * uncertainty and switching costs : * + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + let us consider a production company @xmath0 buying raw materials ( multiple units of a good ) from providers @xmath1 and @xmath2 via a reverse - auction and @xmath0 is uncertain about the time these providers might take to deliver the raw material to @xmath0 .
the providers are very likely to lie about the delivery times and it is hard to incorporate delivery times in designing the auction . if the goal of the auctioneer ( i.e. the company @xmath0 ) is cost minimization , it will buy the raw material from the provider with the minimum ask ( assume that the ask of @xmath3 is smaller than that of @xmath4 which is smaller than that of @xmath2 ) . now if the provider @xmath3 lied about the delivery time at least for a significant fraction of the total required units , @xmath0 s production gets delayed .
if @xmath0 wants to switch to some other provider , run another auction and buys from @xmath4 , now it will buy at a higher cost and their is still a delay in @xmath0 s production as @xmath4 will take its time in delivery too . the time taken by @xmath4 could actually be smaller than that of @xmath3 s for the remaining units , however , still there is a delay .
moreover , such delay might persist further as @xmath4 could also lie about its delivery time .
it might have been better if @xmath0 would buy not only from @xmath3 to start with and give @xmath5 a chance as well .
therefore , one way to reduce such delay times could be via increasing _
capacity _ as per our definition .
there is a single _ indivisible _ item for sale .
there are @xmath6 bidders interested in the item .
the bidder @xmath7 has a value @xmath8 for this item , for some bidders @xmath8 s are the actual true values while for some others these are just crude estimates / guesses . ] .
the item is sold via an auction on an experiment designed by the seller / auctioneer .
the experiment has @xmath9 outcomes @xmath10 with associated probabilities @xmath11 , where @xmath12 . therefore
, the item is essentially sold via an auction of the probability spikes @xmath11 wherein the auctioneer can choose these probability spikes in advance or adaptively based on bidders reports so as to achieve some defined goal such as maximizing her profit or efficiency or to accommodate a wider pool of bidders .
bidders bid on the experiment by reporting bids @xmath13 s to indicate their respective values of the item .
_ at most one probability spike is assigned to each bidder .
_ thus there are effectively two steps in this auction model .
* stage 1 ( _ commit / compete _ ) : the bidders report their bids @xmath13 s and by way of using some mechanism , the auctioneer assigns the probability spikes to them and decides corresponding payments to be made by them . let us call a bidder a _ prospective winner _ if she was assigned one of the probability spikes .
* stage 2 ( _ win or lose _ ) : the experiment is performed .
if the outcome of the experiment is @xmath14 , then the _ prospective winner _ assigned to the spike @xmath15 is declared the _ winner _ , and is given the item .
further , the auctioneer could choose various payment schemes such as - * _ betting : _ every _ prospective winner _ is charged its payment decided in _
compete / commit _ stage irrespective of whether she will be a _ winner _ or not .
* _ pay - per - acquisition _ : a bidder is charged the amount decided in _
compete stage _ only when she is a _ winner _
i.e. only when she actually acquires the item
. the above model can also be interpreted as selling of a single _ divisible _ item in terms of specified fractional bundles , the bundles corresponding to the probability spikes . without loss of generality ,
let us assume that @xmath16 .
further , for notational simplicity let @xmath17 .
let @xmath18 be the allocation rule and @xmath19 be the payment rule decided in _
compete stage_. thus , for @xmath20 , the spike @xmath21 is assigned to the bidder @xmath22 and for @xmath23 , the bidder @xmath22 is not assigned any spike .
further , for @xmath20 , @xmath24 is the expected payment to be made by the bidder @xmath22 and @xmath25 otherwise
. therefore , the expected utility of the bidder assigned to spike @xmath21 is given by @xmath26 for @xmath20 and is zero otherwise .
for the sake of simplicity , let @xmath27 , then the famous _ vcg _ mechanism ranks the bidders by their bids ( and true values @xmath8 s as being a truthful mechanism ) and charges them their respective opportunity costs .
that is , @xmath22 is the bidder with the @xmath21th maximum bid and @xmath28 for @xmath20 and zero otherwise . if the payment is done via the _ betting _ model then this is the amount charged to bidder @xmath21 in the _ compete / commit _ stage .
if the payment is done via _ pay - per - acquisition _ and @xmath21 is the _ winner _ then she is charged an amount @xmath29 , and therefore , her expected payment is still @xmath24 as @xmath15 is the probability that she wins .
therefore , the auctioneer s revenue is @xmath30 let @xmath31 be the spike gaps , then @xmath32 .
the condition @xmath33 translates to @xmath34 and @xmath16 translates to @xmath35 therefore , @xmath36 and clearly @xmath37 as _ vcg _ ranks by true values .
[ vcgrev ] the revenue of the auctioneer in the _ vcg _ mechanism for selling probability spikes @xmath11 can be expressed as @xmath38 where @xmath39 for all @xmath40 , @xmath41 and @xmath37 for all @xmath42 .
further , the efficiency for the _ vcg _ mechanism is @xmath43 further , we have @xmath44 [ vcgeff ] the efficiency in the _ vcg _ mechanism for selling probability spikes @xmath11 can be expressed as @xmath45 where @xmath39 for all @xmath40 , @xmath41 and @xmath37 for all @xmath42 .
we say that a linear function @xmath0 of spike - gaps @xmath46 s is _ gap - wise monotone _ if @xmath47 , where @xmath48 s do not depend on gaps @xmath46 s and @xmath49 for all @xmath50 .
a mechanism for selling probability spikes is called _ gap - wise monotone _ if the revenue of the auctioneer at the prescribed equilibrium point is _ gap - wise monotone _ and is called _ strongly gap - wise monotone _ if the social value ( i.e. efficiency ) at the prescribed equilibrium point is _ gap - wise monotone _ as well . therefore , from lemma [ vcgrev ] and lemma [ vcgeff ] we obtain the following theorem .
[ vcg ] the _ vcg _ mechanism for selling probability spikes is _ strongly gap - wise monotone_. define , @xmath51 .
* walrasian equilibrium : * let @xmath52 be an allocation and @xmath53 be a payment rule , then @xmath54 is called a walrasian equilibrium if for all @xmath55 , @xmath56 . following @xcite , it is not hard to establish the following lemma .
let @xmath52 be an allocation and @xmath53 be a payment rule , then @xmath54 is a walrasian equilibrium iff it is efficient .
therefore , at a walrasian equilibrium , bidders are ranked according to their values and efficiency can be written as in the case of vcg i.e @xmath57 where @xmath58 for all @xmath59 .
let @xmath54 be a walrasian equilibrium for selling probability spikes then the efficiency at this equilibrium is gap - wise monotone .
this means that optimal efficiency is always gap - wise monotone .
further , the optimal omniscient auction ( i.e. when the auctioneer knows everyone s true value @xmath8 s ) extracts a revenue equal to @xmath57 , where @xmath60 for all @xmath61 . therefore , the optimal revenue of omniscient auction is also gap - wise monotone . in this section ,
we develop a linear programming approach to identify optimal probability spikes subject to the capacity constraints in terms of spikes gaps , where the objective is a gap - wise monotone function . for such functions ,
it is simpler to put the constraints in terms of spike - gaps than in terms of spikes themselves , however , it wo nt be hard to see that a similar approach can also be developed if we put the constraints in terms of the spikes , as well as , in the case of functions more general than the gap - wise monotone .
for the sake of simplicity we omit any such details .
let @xmath0 be a _ gap - wise monotone _ function and @xmath62 be a generic set of parameters with the property that @xmath63 and let us consider the following linear programming problem in variables @xmath46 s , @xmath64 the dual problem is @xmath65 and the _ kkt _ conditions are @xmath66 and therefore an optimal solution is @xmath67 @xmath68 as it can be checked to satisfy the _
_ conditions .
the optimal value is @xmath69 clearly , the maximum of the optimal solution @xmath70 over parameters @xmath71 s is attained when @xmath72 for all @xmath50 and in that case @xmath73 now , note that the primal optimal variables @xmath46 s do not depend on the quantities @xmath48 s at all .
therefore , so long as @xmath0 is _ gap - wise monotone _ , the optimal solution to the primal remains the same as in equation [ optsol ] .
it is quite intuitive as the best possible spike allowed by the capacity constraints is assigned to the best possible bidder ( i.e. @xmath74 ) , and all other spikes are the minimum possible as per the constraints .
let @xmath47 be a gap - wise monotone function and the spike - gaps @xmath46 s satisfy conditions [ gap - ep ] , [ epsilonjs0 ] and [ epsilonjs ] , then the optimal choice of spike - gaps are given by equation [ optsol ] and the optimal value of @xmath0 is given by equation [ h - opt ]
. given parameters @xmath71 , let us define the _ capacity _ as @xmath75 now consider the parameters @xmath76 satisfying the properties [ epsilonjs0 ] and [ epsilonjs ] such that @xmath77 .
given a gap - wise monotone function @xmath47 , we claim that such @xmath78 s satisfying properties [ epsilonjs0 ] and [ epsilonjs ] can always be obtained from @xmath79 s such that @xmath80 as long as @xmath81 , meaning that the _ capacity _ can always be increased without any loss in optimal value as long as we do not shoot over the absolute optimum @xmath82 .
we have , @xmath83 now we can always choose @xmath78 s satisfying properties [ epsilonjs0 ] and [ epsilonjs ] by taking suitable @xmath84 , @xmath85 and @xmath86 .
in particular , taking @xmath87 , @xmath85 and @xmath88 does the job .
an interesting case to consider is when @xmath89 for all @xmath90 and @xmath72 otherwise . and
in this case we can increase the capacity without loss in optimal value by taking @xmath91 and @xmath92 otherwise , where @xmath93 now let us define @xmath94 then it is clear that @xmath95 and therefore , as long as @xmath96 capacity can always be increased without any loss in the optimal value as discussed above , however there is a strict decrease in the optimal value if we wish to increase capacity from @xmath97 to @xmath98 .
we can naturally define a parameter which we call _ price of capacity _ as follows : @xmath99 thus , _ price of capacity _ is the worst possible loss in optimal value while increasing capacity from @xmath97 to @xmath98 . again let us consider the case when all non - zero @xmath100 , then @xmath101 often our goal will be to maximize efficiency or revenue subject to the capacity constraints , and consequently such a loss may not be considered good . therefore , its really a price that we are paying for increasing capacity
as we discussed in the section [ mot ] , one nice motivation for the study of capacity as a metric for mechanism design comes from the sponsored search advertising .
we first describe the formal ssa model .
formally , in the current models , there are @xmath102 slots to be allocated among @xmath6 ( @xmath103 ) bidders ( i.e. the advertisers ) .
a bidder @xmath7 has a true valuation @xmath8 ( known only to the bidder @xmath7 ) for the specific keyword and she bids @xmath13 . the expected _ click through rate _ ( ctr ) of an ad put by bidder @xmath7 when allocated slot @xmath21 has the form @xmath104 i.e. separable in to a position effect and an advertiser effect .
@xmath105 s can be interpreted as the probability that an ad will be noticed when put in slot @xmath21 and it is assumed that @xmath106 for all @xmath107 and @xmath108 for @xmath109 .
@xmath110 can be interpreted as the probability that an ad put by bidder @xmath7 will be clicked on if noticed and is referred to as the _ relevance_(quality score ) of bidder @xmath7 .
the payoff / utility of bidder @xmath7 when given slot @xmath21 at a price of @xmath111 per - click is given by @xmath112 . as of
now , google as well as yahoo ! use schemes closely modeled as rbr(rank by revenue ) with gsp(generalized second pricing ) .
the bidders are ranked in the decreasing order of @xmath113 and the slots are allocated as per this ranks . let the @xmath114 be the bidder allocated to the slot @xmath7 according to this ranking rule , then @xmath114 is charged an amount equal to @xmath115 per - click .
this mechanism has been extensively studied in recent years@xcite .
the solution concept that is widely adopted to study this auction game is a refinement of nash equilibrium called _ symmetric nash equilibria(sne ) _ independently proposed by varian@xcite and edelman et al@xcite . for notational simplicity ,
let @xmath116 , then under this refinement , the revenue of the auctioneer at equilibrium is given by @xmath117 in this section , we discuss how to incorporate the capacity constraints in the keyword auctions being currently used by google and yahoo!. we understand that there could be several ways for doing so , however , we consider a very simple and intuitive way of incorporating the capacity constraints via probability spikes as follows : * the first @xmath118 slots are sold as usual to the @xmath118 high - ranked bidders . * the last slot ( i.e. the @xmath102th slot ) is sold via probability spikes @xmath11 among the @xmath9 bidders ranked @xmath102 through @xmath119 .
@xmath9 is chosen to accommodate as many more bidders as the auctioneer wants .
* there is a single combined auction for both of the above . clearly , the single combined auction is equivalent to the keyword auction with @xmath120 slots with position based ctrs taken as @xmath121 i.e. @xmath122 for @xmath123 , @xmath124 for @xmath125 and @xmath126 otherwise .
the revenue of the auctioneer at _ sne _ is @xmath127 now , maximizing @xmath128 as a function of @xmath15 s is equivalent to maximizing the function @xmath129 note that @xmath0 may not be gap - wise monotone , however , it is not hard to see that the similar linear programming analysis as in the section [ genopt ] can be done to compute the price of capacity in the present scenario of keyword auctions as well .
we omit the details .
[ [ selling - clicks - via - auctioneer - controlled - probability - spikes - a - new - model - for - sponsored - search - advertising ] ] selling clicks via auctioneer - controlled probability spikes : a new model for sponsored search advertising : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the design framework in section [ single ] , suggests a new model for ssa , where the clicks are sold directly and not indirectly via allocating impressions . in the usual pay - per - click model , a click is assigned to one of the advertisers who has been allocated an impression in the page currently being viewed by the user ; thus , the user s collective experience determines the probability that an impression winner would be the beneficiary of the click , and having set up the impressions and the user experience , the auctioneer does not actively control the probability with which a click will be allocated to a bidder .
the impressions are sold and clicks are only stochastically related to impressions , even though an advertisers is charged only when she obtains a click . in the new model ,
a click could be considered as the indivisible item being sold via probability spikes .
instead of putting ads directly in the slots , the auctioneer could put some categories / information related to the specific keyword as a link .
an auction based on probability spikes is run whenever a user clicks on this link and the user is directly taken to the landing page of the winning advertiser .
in the present paper , our main goal was to motivate _
capacity _ as a fundamental metric in designing auctions in the information economy and then to initiate study of such a design framework for some simple and interesting scenarios such as single indivisible item and sponsored search advertising
. however , there are myriad of other interesting scenarios where the capacity - enabled framework should be interesting to study . for example , the auctions of digital goods@xcite , combinatorial auctions @xcite for selling information goods in bundles , double auctions and ad - exchanges@xcite etc .
further , probably the most important question that remains to be addressed is to identify the best way of putting capacity constraints and to see how generic the approach via probability spikes could be made .
we can consider the most general case of selling any set of items e.g. heterogeneous , homogeneous , indivisible , or divisible or any combination their of .
let @xmath130 be the set of all possible allocations .
further , the elements of @xmath131 are named such that @xmath132 , where @xmath133 is auctioneer s preference over allocations . then , these set of items can be sold with capacity constraints via probability spikes @xmath134 s with @xmath135 , wherein @xmath136 is enforced with probability @xmath134 .
the authors thank dork alahydoian , sushil bikhchandani , gunes ercal , himawan gunadhi and adam meyerson for insightful discussions .
the work of s.k.s was partially supported by his internship at netseer inc .
, los angeles .
g. aggarwal , a. goel , r. motwani , truthful auctions for pricing search keywords , ec 2006 .
u. birchler and m. butler , information economics , routledge , 2007 .
t. borgers , i. cox , m. pesendorfer , v. petricek , equilibrium bids in sponsored search auctions : theory and evidence , technical report , university of michigan ( 2007 ) . s. bikhchandani and j. m. ostroy , from the assignment model to combinatorial auctions . in combinatorial auctions
mit press 2006 .
p. crampton , y. shoham , and r. steinberg , combinatorial auctions , mit press , 2006 .
gabrielle demange , david gale , and marilda sotomayor , multi - item auctions , jour .
political economy , 94 , 863 - 872 , 1986 .
b. edelman , m. ostrovsky , m. schwarz , internet advertising and the generalized second price auction : selling billions of dollars worth of keywords , american economic review 2007 .
a. v. goldberg , j. d. hartline , a. wright , competitive auctions and digital goods , soda 2001 .
a. v. goldberg , j. d. hartline , a. r. karlin , m. saks , a. wright , competitive auctions , games and economic behavior , vol .
55 , pp:242 - 269 , 2006 . v. krishna , auction theory , academic press , san diego , 2002 .
j. kalagnanam and d. parkes , auctions , bidding and exchange design . in : simchi - levi ,
wu , shen : supply chain analysis in the ebusiness area , kluwer academic publishers , 2003
. s. lahaie , an analysis of alternative slot auction designs for sponsored search , ec 2006 .
s. lahaie , d. pennock , revenue analysis of a family of ranking rules for keyword auctions , ec 2007 .
p. milgrom , putting auction theory to work , cambridge university press , 2004 .
a. mehta , a. saberi , u. vazirani , v. vazirani , adwords and generalized on - line matching , focs 2005 .
p. nelson , information and consumer behaviour , the journal of political economy , vol 78 , pp:311 - 329 , 1970 . s. pandey , and c. olston , handling advertisements of unknown quality in search advertising , nips 2006 .
s. k. singh , v. p. roychowdhury , m. bradonji , and b. a. rezaei , exploration via design and the cost of uncertainty in keyword auctions , preprint ( available at http://arxiv.org/abs/0707.1053 ) .
s. k. singh , v. p. roychowdhury , h. gunadhi , and b. a. rezaei , capacity constraints and the inevitability of mediators in adword auctions , wine 2007 ( to appear ) .
s. k. singh , v. p. roychowdhury , h. gunadhi , and b. a. rezaei , diversification in the internet economy : the role of for - profit mediators , preprint ( available at http://arxiv.org/abs/0711.0259 ) .
l. s. shapley and m. shubik , the assignment game i : the core , int .
j. game theory 1 , no . 2 , 111 - 30 , 1972 .
c. shapiro and h. r. varian , information rules : a strategic guide to the network economy , harvard business school press , 1998 .
a. tversky ; d. kahneman , loss aversion in riskless choice : a reference - dependent model , the quarterly journal of economics , vol .
( nov . , 1991 ) , pp .
1039 - 1061 .
h. varian , position auctions , to appear in international journal of industrial organization .
j. wortman , y. vorobeychik , l. li , and j. langford , maintaining equilibria during exploration in sponsored search auctions , wine 2007 ( to appear ) . | the auction theory literature has so far focused mostly on the design of mechanisms that takes the revenue or the efficiency as a yardstick .
however , scenarios where the _ capacity _ , which we define as _ `` the number of bidders the auctioneer wants to have a positive probability of getting the item '' _ , is a fundamental concern are ubiquitous in the information economy .
for instance , in sponsored search auctions ( ssa s ) or in online ad - exchanges , the true value of an ad - slot for an advertiser is inherently derived from the conversion - rate , which in turn depends on whether the advertiser actually obtained the ad - slot or not ; thus , unless the capacity of the underlying auction is large , key parameters , such as true valuations and advertiser - specific conversion rates , will remain unknown or uncertain leading to inherent inefficiencies in the system .
in general , the same holds true for all information goods / digital goods .
we initiate a study of mechanisms , which take capacity as a yardstick , in addition to revenue / efficiency .
we show that in the case of a single indivisible item one simple way to incorporate capacity constraints is via designing mechanisms to sell probability distributions , and that under certain conditions , such optimal probability distributions could be identified using a linear programming approach .
we define a quantity called _ price of capacity _ to capture the tradeoff between capacity and revenue / efficiency .
we also study the case of sponsored search auctions .
finally , we discuss how general such an approach via probability spikes can be made , and potential directions for future investigations . |
the nature of scalar mesons continues to be an intriguing problem in qcd .
experimentally , there are well established scalar mesons with isospin @xmath2 , the @xmath3 , @xmath22 with isospin @xmath23 , the @xmath24 meson , and with isospin @xmath7 , the @xmath25-mesons at 980 , 1370 @xcite and 1500 mev from gams , crystal barrel @xcite , wa102 @xcite and bes @xcite . besides these resonances
, there are different experimental indications @xcite , especially from bes @xcite , e791 @xcite , focus @xcite , kloe @xcite , snd @xcite , cmd2 @xcite , belle @xcite , wa102 @xcite and @xmath26 scattering data @xcite for some other scalar states , with @xmath7 , the @xmath12 , @xmath27 and @xmath18 , and with @xmath23 , the @xmath28 . the real quark and/or gluon contents of these states are not fully understood , which the interpretation using effective theories most of them based on a linear realization of chiral symmetry can not clarify . in the following , we shall focus on the tests of the @xmath5 , @xmath1 and gluonium natures of these scalar mesons by confronting the recent experimental data with some qcd predictions based on qcd spectral sum rules ( qssr ) complemented with some low - energy theorems ( let ) @xcite and lattice calculations @xcite .
+ these channels are expected to be simpler as we do not expect to have any mixing with a gluonium .
if one assumes that these states are @xmath5 mesons , one can naturally associate them to the divergence of the vector currents : a_0(980)&&_v^_|ud(m_u - m_d):|u(i)d : , + ( 840)&&_v^_|us(m_u - m_s):|u(i)s : . within the qssr approach
, the properties and implications of these mesons can be studied from the two - point correlator : _
|uq(q^2)=id^4x e^iqx0|t_v^_|uq(x)_v^_|uq(0)^|0 .
since the pioneering work in @xcite , numerous authors have used and improved the analysis of the previous correlator for the extraction of the running @xmath30-@xmath31 mass - difference and strange quark masses @xcite .
the improvements come from the inclusion of higher order terms in the pt qcd series @xcite ; the inclusion of the @xmath32 term @xcite which mimics the uv renormalon effects and which is also an alternative to the direct instanton effects where the later is not under a good quantitative control due to the incertainties of the instanton size and widths , ... ; the treatment of the spectral function using new @xmath33 phase shift data @xcite .
the obtained value of the @xmath30-@xmath31 mass - difference is consistent with some other determinations , while the extraction of the @xmath3 mass from the sum rule @xcite is perfectly consistent with the data .
recent analysis @xcite has lead to a value of the strange quark running mass which is consistent with some other qssr determinations from different channels @xcite .
all these features support the @xmath5 meson assignements for the @xmath3 and the @xmath28 mesons .
however , the apparent _ wrong _ splitting of the @xmath3 and the @xmath28 mesons is intriguing . here
, we investigate this analysis using the ratio of exponential laplace / borel sum rules : ( ) = -dd . using the pt series to order @xmath34 , including the np condensates of dimension 6 and the new 1/@xmath35 terms
, we give the prediction for the @xmath3 in fig .
[ fig : scalar1]a ) and for the @xmath28 in fig .
[ fig : scalar1]b ) using a nwa .
we use the most recent value of @xmath36(2 gev)=@xmath37 mev compiled in @xcite .
one can see from these figures that the method reproduces the _ wrong _ splitting of the two mesons .
the reason is that the @xmath38 breaking effects increase the value of the sum rule optimization scale compared to the one of the @xmath39 and then emphasize the contribution of the dimension six condensates , which is a vital correction in the analysis . indeed , from the analytic expression of the sum rules
, one can qualitatively extract the approximate mass formula:@xmath40 and @xmath41@xmath40 mass splittings @xcite .
] m^2 _ & & m^2_a_0 + 2_s^2- 8 ^ 2m_s|ss_0 + & + & 32 140881 ^ 3_s|ss^2-|uu^2 ^ 2_0 + & -&13m^2__^2_0 , where all different parameters including the @xmath39 mass are evaluated at the sum rule optimization scale @xmath42 ; @xmath43 @xcite indicates the deviation from the vacuum saturation of the four - quark condensate ; @xmath44 measures the @xmath38 breaking of the quark condensate @xcite . as shown in the above formula , the @xmath38 breaking corrections are relatively small , and the four - quark condensate tends to decrease the @xmath6 mass . compared to the value of the @xmath6 mass given in fig .
[ fig : scalar1]b ) using nwa , the finite width correction reduces the mass by about 20 mev , where we have used the width of 310 mev from bes @xcite .
one should mention that at the value of @xmath45 , the ope in powers of @xmath46 converges , while the radiative corrections to the parton graph , though large in individual sum rules , remain small in the ratio of moments @xmath47 , as these corrections tend to compensate each others , justifying the uses of the result at this relativley low scale . from the previous analysis ,
we deduce : m_a_0930 mev and m_920 mev , with about 10% error , in good agreement with recent data @xcite .
parameters @xcite used as in put a value of @xmath48 and @xmath49 much higher than here and gives higher value of the @xmath6-mass . ] + * the decay constants * + the decay constant @xmath50 of the @xmath39 normalized as : 0_v^_|uda_0 f_a_0m^2_a_0 , in the same way as @xmath51 mev has been estimated several times in the literature @xcite : f_a_0(1.60.5 ) mev , where a better accuracy is claimed in @xcite .
using @xmath38 symmetry and the almost degeneracy of the @xmath39 and @xmath6 masses , we expect to have with a good accuracy : 40 . * the hadronic couplings * + the @xmath39 and @xmath6 hadronic couplings have been obtained using either a vertex sum rule @xcite or / and @xmath38 symmetry rotation @xcite . the leading order vertex sum rule results are : g_a_0k^+k^- & & 8 ^ 23m_s|ssm^2_kf_k1 - 2r3 gev + g_a_0k^+k^-g_k^+^- & & e^-(m_k^2-m_^2)_01 - 2r 1.17 , where we have used @xcite : @xmath52 , @xmath53 and @xmath54 gev@xmath55 .
we expect an accuracy of about 20% ( typical for the 3-point fonction sum rules ) for these estimates . using the @xmath38 relation : g_a_0g_a_0k^+k^-
one obtains : a_0 & & g_a_0 ^ 216m_a_01-m^2_m^2_a_0 + & & 84 mev , in agreement with the range of data from 50 to 100 mev given by pdg @xcite . using the previous value of the @xmath6 coupling
, one can deduce : [ eq : kappawidth ] ( k)(k^+-)104 mev , which is about a factor 4 smaller than the present data @xcite , but is a typical value for the width of a @xmath5 state . + * the @xmath56 widths * + the @xmath56 width of the @xmath39 has been evaluated using vertex sum rules within the @xmath5 and four - quark assignements of this meson , with the result @xcite : a_0(|qq)(0.3~2.1 ) kev , and @xcite : a_0(4q)(2~5)10 ^ -4 kev , to be compared with the data of ( @xmath57 ) kev compiled in @xcite . due to the inaccuracy of the qssr predictions , we shall definitely use , in the following , the measured value of the @xmath39 width as an input for our theoretical predictions of the @xmath0 meson - width . +
* concluding remarks * + @xmath58 the previous analysis shows that the mass and widths of the @xmath3 are well described by a @xmath5 assignement of this meson .
+ @xmath58 a qssr analysis of the four - quark assignement @xcite of these states gives predictions which reproduce the experimental mass of the @xmath3 @xcite , like do the lattice calculations @xcite .
the result for the hadronic coupling @xmath59 in the four - quark scenario depends crucially on the operators describing the @xmath39 and can range from 1.6 gev @xcite to ( 58 ) gev @xcite .
however , the prediction for the @xmath60 can agree with the data @xcite depending on the size of the operator mixing parameter .
therefore , an eventual selection of the two approaches will be a precise measurement of @xmath59 or / and a lattice measurement of the decay constant which should depend linearly on the light quark mass in the @xmath5 scheme , but is a constant in the four - quark one .
+ @xmath58 another problem arises when one computes the @xmath56 width using vertex sum rules .
the ratio of the widths in the two approaches is @xcite : _
4q/_|qq(1~2)10 ^ -3 , which is of the order of @xmath61 , indicating that the four - quark assignement prediction is too small contrary to some claims in the literature .
+ @xmath58 for the @xmath6 meson , the @xmath5 assignement can reproduce quite well its _
wrong _ splitting with the @xmath39 , but fails to reproduce its large experimental width [ about a factor 4 smaller ( see eq .
( [ eq : kappawidth ] ) ] , while the four - quark one gives a width of about a factor 2 smaller @xcite .
the failure of the two separate approaches ( @xmath5 and four - quark assignement ) may suggest that the quark content of the @xmath6(841 ) is more complex than navely expected : it can be a mixing between a @xmath5 and a four - quark states , or it can come from a large interference of the @xmath5 ground state with non - resonant backgrounds .
further tests are needed for clarifying its nature .
the isoscalar scalar states are especially interesting in the framework of qcd since , in this anomalous @xmath9 channel , their interpolating operator is the trace of the energy - momentum tensor : _
^=(_s ) g^2+_i [ 1+_m(_s ) ] m_i|_i_i , where @xmath62 is the gluon field strengths , @xmath63 is the quark field ; @xmath64 and @xmath65 are respectively the qcd @xmath66-function and quark mass - anomalous dimension ( @xmath67 and @xmath68 for @xmath69 flavours ) . in the chiral limit @xmath70 , @xmath71 is dominated by its gluon component @xmath72 , like is the case of the @xmath10 for the @xmath11 axial - anomaly , explaining why the @xmath10-mass does not vanish like other goldstone bosons for @xmath70 . in this sense , it is natural to expect that these @xmath7 scalar states are glueballs / gluonia or have at least a strong glue component in their wave function .
this gluonic part of @xmath73 should be identified with the @xmath9 term @xcite in the expression of the effective lagrangian based on a @xmath74 linear realization of chiral symmetry ( see e.g. @xcite ) . +
* unmixed @xmath7 scalar @xmath5 mesons * + we shall be concerned with the mesons @xmath75 and @xmath76 mesons associated respectively to the quark currents : j_2=m:1(|uu+|dd ) : and j_3=m_s:|ss : . from the good realization of the @xmath77 flavour symmetry ( @xmath78 and @xmath79 ,
one expects a degeneracy between the @xmath39 and @xmath75 states : m_s_2m_a_0930 mev , while its hadronic coupling is @xcite : [ eq : s2pipi ] g_s_2^+^-|uu_0 e^m^2_2_0/22.46 gev . corresponding to : ] . ] & & ( s_2^+^-)120 mev , using @xmath38 symmetry , one can also deduce : [ eq : s2kk ] g_s_2k^+k^-g_s_2^+^-1.23 gev .
the @xmath75 @xmath56 width can be deduced from the one of the @xmath80 obtained previously , through the non - relativistic relation ( ratio of the square of quark charges ) : [ eq : gamma ] _ s_2 _
a_0(0.70.2 ) kev .
the mass of the mesons containing a strange quark is predicted to be @xcite : m_s_3/m_1.030.02 m_s_3948 mev , if one uses @xmath81 mev , mev . ] while its coupling to @xmath82 is @xcite : [ eq : s3kk ] & & g_s_3k^+k^-(2.70.5 ) gev .
these results suggest that the nave @xmath5 assignement of the @xmath83 does not fit the data .
+ * gluonia masses and decay constants * + @xmath58 these states have been explicitly analyzed in @xcite using qssr of the two - point correlator : _
s(q^2)=16id^4x e^iqx0|t^_(x)^_(0)^|0 , in the chiral limit @xmath84 , for the observables : _ n()=-_0^dt t^n e^-t im_s(t ) , and : _
n , n+1()=-dd , where @xmath85 . for @xmath86 ,
the sum rule is sensitive to the subtraction constant : [ eq : psi0 ] _
s(0)=-16_1_s g^2 , fixed from let @xcite , where @xmath87 gev@xmath88 @xcite .
one has found in @xcite that , due to @xmath89 , the subtracted sum rule ( ssr ) @xmath90 is more weighted by the high - energy behaviour of the spectral integral than the unsubtracted sum rule ( usr ) @xmath91 , which motivated the introduction of 2 resonances ( below and above 1 gev ) for solving the controversial results obtained in the past .
the results of the analysis using the standard ope by retaining higher order pt series and the lowest dimension condensates are @xcite : f__b & & ( 884116 ) mev , + f_g & & ( 390145 ) mev , corresponding to : [ eq : massglue ] m__b1 gev and m_g(1.50.2 ) gev .
@xmath92 has been obtained in @xcite using a least square fit of @xmath93 .
its decay constant has been obtained using a least square fit of @xmath93 or / and a stability criterion of @xmath94 @xcite .
channel has given an estimate of the @xmath10 parameters and of the @xmath11 topological charge @xcite .
] the mass of the 2nd resonance has been fixed from @xmath93 @xcite and its decay constant comes from @xmath90 @xcite .
we shall see that @xmath95 is the one which can be compared with the present lattice value of about 1.6 gev in the quenched approximation @xcite , while the @xmath8 mass will be a @xmath10-like meson expected to couple strongly with goldstone boson pairs ( _ huge ozi violation _ ) @xcite and playing a role in the saturation of the @xmath9 two - point correlator subtraction constant @xmath89 .
it can only be tested using lattice with dynamical fermions ( see e.g. @xcite for the inclusion of the disconnected part of the scalar propagator ) . + @xmath58 one can also notice that a possible effect of the radial excitation of the @xmath96 can be obtained by matching the radial excitation contribution with the qcd continuum .
assuming its mass to be around 1.4 gev , one can deduce @xcite : f__b(139~224 ) mev , while a weaker bound of about 500 mev has been allowed in @xcite .
+ @xmath58 the effect of the @xmath32 term , to the previous results , which is an alternative of the direct instanton contribution has been shown to be small @xcite , though this term is necessary for solving the sum rule scale hierarchy of the gluonia channels compared to the usual @xmath5 mesons .
+ @xmath58 a recent qssr analysis of the same gluonium correlator using gaussian sum rules and including instantons @xcite confirms the previous mass values obtained in eq .
( [ eq : massglue ] ) , but not the results in ref .
@xcite , where it is argued that the presence of the direct instantons solve the controversial results noticed in @xcite between the subtracted @xmath86 sum rule with the other @xmath97 unsubtracted ones , without the need of two resonances . in our normalization , the results in @xcite are : m_s=(1.250.2 ) gev , f_s= ( 30.3 ) gev .
the mass value does not contradict the ones in eq .
( [ eq : massglue ] ) as it is about the mean value of the two resonances ones , while the decay constant leads to @xcite : ( j / s ) 1.510 ^ -2 , which is about 10 times higher than the one of the @xmath98 and which is already excluded by the bes @xcite and some other data .
this fact may signal some eventual internal inconsistencies in the treatment of the instanton contributions .
+ @xmath58 an upper bound on the gluonium mass has been also obtained in @xcite : m_g(2.160.22 ) gev , using the positivity of the moment @xmath99 .
this bound has been strengthened in @xcite to 1.7 gev , which can not be an absolute bound because of the inclusion of the qcd continuum model and of its related uncertainties for its derivation . + * gluonia widths to @xmath26 * + @xmath58 for this purpose , we consider the gluonium-@xmath26 vertex : v(q^2)=_1|^_|_2 , q = p_1-p_2 , where : @xmath100 . in the chiral limit @xmath101
, the vertex obeys the dispersion relation : v(q^2)=_0^ v(t ) , which gives the 1st nv sum rule @xcite : _
s_b,_b , gg_sf_s 0 . using the fact that @xmath102 @xcite , one obtains the 2nd nv sum rule @xcite : _
s_b,_b , gg_sf_s / m^2_s=1 . to a first approximation , we assume a @xmath8-dominance in the 2nd nv sum rule , while in the 1st sum rule , there is a matching between the @xmath8 and all higher mass glueball contributions , which we replace by an effective @xmath103 mass of about 1.4 gev .
then , one obtains : [ eq : sigmapipi ] g__b^+^-g__bk^+k^- ( 3.2~6.8 ) gev .
a complete matching in the 1st nv sum rule would lead to an effective coupling : g__b^+^-12 gev , when using @xmath104 mev .
an interpretation of this value of @xmath105 is unclear but it is expected to parametrize all higher states contributions to the 1st sum rule . if one uses a resonance mass of about 1.4 gev , one would obtain a very broad @xmath106 which can mimic the _ red dragon _ proposed earlier @xcite , but can not manifest as a peak , making its identification with a true resonance difficult .
different strategies for extracting the @xmath19 parameters from the continuum background has been discussed in @xcite leading to a width of the @xmath25 compiled in pdg @xcite .
for definiteness , we shall use in the following , the value of @xmath107 about 250 mev obtained in @xcite within a factor two accuracy .
this leads to : g__b^+^-g__bk^+k^-(4.20.7 ) gev .
* @xmath8 as an @xmath10-like meson * + @xmath58 one can check that the let for @xmath89 given in eq .
( [ eq : psi0 ] ) is almost saturated by the contribution of the lowest mass @xmath108 meson : _
s(0)2m^2__b f__b^2 1.5 gev^4 , compared to the let value of 1.6 gev@xmath88 .
this property is very similar to the contribution of the @xmath10 in the topological charge ( subtraction constant of the anomalous @xmath11 gluonium two - point correlator @xcite),which explains why it is not degenerated with the pion at finite @xmath109 .
+ @xmath58 the @xmath8 large coupling to pseudoscalar pairs , through ozi violating process , can also be compared with the affinity of the @xmath10 to couple to ordinary mesons , making it as an ambidextre gluonium - meson state + @xmath58 like in the @xmath11 sector , the quenched lattice simulations obtain a higher glueball mass which is not the @xmath10 mass . the @xmath10-mass and decay constant are only measured from a lattice calculation of the @xmath11 topological charge @xcite .
we expect that the same situation occurs in the @xmath9 channel , where a quenched lattice gives a scalar gluonium mass of about 1.5 gev , while the @xmath8 parameters can be obtained from the measurement of the scalar correlator subtraction constant @xmath89 including dynamical fermions . + * @xmath110 widths into @xmath111 and @xmath112 * + @xmath58 analogous low - energy theorem @xcite gives : _
1|^_|_1= 2m^2__1 , where @xmath113 is the unmixed @xmath114 singlet state of mass @xmath115 0.76 gev @xcite .
writing the dispersion relation for the vertex , one obtains the nv sum rule : _
s_b,_b , gg_s_1_1f_s= 2m^2__1 , which , by assuming a @xmath116-dominance of the vertex sum rule , leads to : [ coup ] g_g_1_1(1.2~1.7 ) . introducing the physical " @xmath10 and @xmath117 through : ~_p _ 1-_p _ 8 + ~_p _ 1+_p _ 8
, where @xcite @xmath118 is the pseudoscalar mixing angle , one can deduce : ( g)(5~10 ) mev .
the previous scheme is also known to predict ( see nv and @xcite ) : r0.22 , g_g_pg_g , compared with the gams data @xcite @xmath119 .
this result can then suggest that the @xmath120 seen by the gams group is a pure gluonium , which is not the case of the particle seen by crystal barrel @xcite which corresponds to @xmath121 . + * gluonia widths into @xmath21 * + within our scheme , we expect that the @xmath21 are mainly @xmath122-waves initiated from the decay of pairs of @xmath8 . using : _
b|^_|_b= 2m^2__b , and writing the dispersion relation for the vertex , one obtains the sum rule : _
i=_b,_b , gg_s_b_bf_s= 2m^2__b .
we use @xmath123 gev , @xmath124 gev and the observed @xmath125 width into @xmath21 of about @xmath126 mev @xcite ( @xmath122-wave part ) . neglecting , to a first approximation , the @xmath8 contribution to the sum rule
, we can deduce : |g_g_b_b|1.3~3.7 , where the first ( resp .
second ) value corresponds to @xmath127 gev ( resp .
this leads to the width into @xmath128 of about @xmath129 mev , much larger than the one into @xmath112 and @xmath130 .
this feature is satisfied by the @xmath110 state seen by gams , crystal barrel and wa102 @xcite .
however , the previous approaches show the consistency in interpreting the @xmath110 seen at gams as an almost " pure gluonium state ( ratio of the @xmath130 versus the @xmath112 widths ) , while the state seen by the crystal barrel and wa102 , though having a gluon component in its wave function , can not be a pure gluonium because of its prominent decays into @xmath112 and @xmath131 . +
* gluonia widths into @xmath56 * + these widths have been derived in @xcite by identifying the @xmath56-glue - glue box diagram with the scalar @xmath56 lagrangian where the quarks in the internal have been taken to be non - relativistic . in this way , one has obtained : g_f_m^2__u , d , s q_i^2/m_i^4 , where @xmath132 and @xmath133 are the quark charge and constituent masses .
this leads to : ( _ b)(0.03~0.08 ) kev , + ( _b)(0.01~0.03 ) kev , + ( g)(0.3~0.6 ) kev .
alternatively , one can use the trace anomaly to order @xmath134 in order to deduce : 0|14(_s ) g^2|_1_2 - 0| r f_1^f_2^|_1_2 , where @xmath135 , @xmath136 is the qed coupling .
this relation gives : _
f_ig_ir . from this relation and using the previous values of @xmath137 and @xmath138 , one can deduce : _
g(1~6 ) kev , which is quite inaccurate but still consistent with the previous determination . +
* comments * + comparing the above results , especially the predicted widths , with the experimentally observed candidates , it is likely that the @xmath96 and some of its radial excitations have a lot of glue in their wave functions . as a consequence , a quarkonium - gluonium decay mixing scheme has been proposed in the @xmath7 scalar sector .
@xcite , for explaining the observed spectrum and widths of the possibly wide @xmath139 gev ) and the narrow @xmath13 . where the resulting @xmath140 width of about 4 kev is much bigger than generally expected and which needs to be confirmed by some other data .
we plan to come back to this point in the future . ]
@xmath58 bes data suggest that the @xmath96(600 ) is produced in the ozi forbidden @xmath141 process @xcite , which can indicate the large amount of glue in its wave function .
its production from the ozi allowed @xmath142 processes , is expected to be due to its quark component , while the relative small branching ratio in the ozi allowed @xmath143 process relative to @xmath144 can be due to an interference between the @xmath82 amplitude from the gluon and quark components of the @xmath96 .
+ @xmath58 in the same way , the @xmath13 is produced in the ozi violating @xmath145 @xcite and @xmath146 @xcite processes which may also indicate its gluonium component , while its production from @xmath147 can signal a strong @xmath148 component in its wave function .
we shall keep in mind these results for building the mixing scheme . + * the meson - gluonium mixing scheme * + @xmath58 we assume that the observed states come from the mixing between the gluonium @xmath8 and quark @xmath149 and @xmath150 bare states :- matrix formalism ( see e.g. @xcite ) . due to the large error in our mass predictions
, we neglect , to a first approximation , some possible shifts on the masses which can be induced by the decay processes mentioned in @xcite . ]
[ eq : mixing1 ] + f_0 + = & + - & + _ b + s_2+_s s_3 where @xmath151 for a @xmath38 singlet and @xmath152 for an su(3 ) octet .
+ @xmath58 in @xcite , the mixing angle @xmath153 has been fixed from the analysis of the predicted decays of the hypothetical bare states @xmath154 and of the observed meson @xmath155 . using the predictions : @xmath156 in eq .
( [ eq : gamma ] ) , @xmath157 kev @xcite , and the recent data @xmath158 kev @xcite , one can deduce : ( 4515)^0 , where the @xmath159 width is suppressed as ( 2/25 ) compared to the one of the @xmath75 .
+ @xmath58 using , for definiteness , as inputs the theoretical predictions given in eqs ( [ eq : s2pipi ] ) and ( [ eq : sigmapipi ] ) , we predict the couplings : g_^+^-&&(5.31.8 ) gev , + g_f_0^+^-&&(1.81.3 ) gev , in reasonnable agreement with the ones from the data : g^exp_^+^-&&(3.01.5 ) gev , + g^exp_f_0^+^-&&(1.50.3 ) gev , corresponding to : ^exp((600 ) ) & & 481 mev , + ^exp(f_0(980 ) ) & & ( 7030 ) mev , @xmath58 in order to predict the mixing parameter @xmath160 , we fit the experimental @xmath161 width from bes @xcite : 2.050.15 , and we use the theoretical predictions given in eqs ( [ eq : s2pipi ] ) , ( [ eq : s3kk ] ) and ( [ eq : sigmapipi ] ) .
then , we obtain : _
s3.0 , and g_k^+k^-/ g_^+^-2 .
further data are needed for improving and testing this result .
+ @xmath58 one should note that using only the constraint from the @xmath162 and @xmath163 hadronic widths , one would obtain : [ eq : scheme2 ] _
s=16 ^ 0 , and _
s-1.4 , indicating that the @xmath96 is an almost pure gluonium and the @xmath25 a @xmath5 @xmath38 octet , which is similar to the scheme in @xcite . however , the result in eq .
( [ eq : scheme2 ] ) would give a too high value of @xmath164 and does not explain the ozi violating production of the @xmath25 in @xmath165 and @xmath166 radiative decays .
+ * comments on alternative approaches * + @xmath58 one should note that a four - quark qssr analysis gives @xcite : [ eq : 4quark ] g_f_0k^+k^- & & g_a_0k^+k^-(1.60.1 ) gev + g_f_0^+^-&&(0.470.05 ) gev , where the absolute values differ from the ones given in @xcite and then question the realibility of the results obtained there .
therefore , lattice calculations of these couplings become mandatory .
( [ eq : 4quark ] ) leads to : ( f_0(980 ) ) & & 7 mev , which is too small compared with the range @xmath167 mev given by the data @xcite .
+ @xmath58 alternative approaches based on @xmath168 loop , @xmath168 molecules and four - quark states can predict value of @xmath169 in agreement with the data @xcite,@xcite , but , most of them , do not give a satisfactory prediction for the @xmath170 width and their production from ozi violating decays . on the other , it would be interesting to see the connection of these effective approaches with the quark - gluon picture used here . + * tests from @xmath165 and @xmath166 radiative decays . *
+ @xmath58 these decays are known to be a gluonium filter .
the production of a gluonium @xmath122 from @xmath165 radiative decays can be approximated by @xcite : & ( j / s ) ^4 ^ 4 + & f^2_s , where @xmath171 gev is the charm constituent quark mass . in our scheme ,
the @xmath96 is mostly a gluonium .
therefore , one expects the branching ratio : b(j/)1910 ^ -5 .
extrapolating the previous exppression to the @xmath166-meson and using @xmath172 mev , one obtains : b()1210 ^ -5 , which , despite the crude approximation used , compares quite well with the kloe @xcite data : b(^0 ^ 0)10.90.30.510 ^ -5 . *
tests from @xmath173 semileptonic decays * + this section has been discussed in details in @xcite and will be only sketched in the following .
+ @xmath58 @xmath174 * meson productions * + if the scalar mesons were simple @xmath5 states , the semileptonic decay width could be calculated quite reliably using qssr , where the relevant diagram is a quark loop triangle .
several groups @xcite predict all form factor to be : @xmath175 yielding , for @xmath176 mev , a decay rate : ( ds_2 l)=(83)10 ^ -16 gev , which is , unfortunately , even in high stastistics experiments , at the edge of observation since the decays into an isoscalar are ckm - suppressed due to the @xmath177-@xmath30 transition at the weak vertex .
+ @xmath58 * scalar gluonium or / and @xmath148 productions * + semi - qualitative but model independent results for the production of gluonium have been given in @xcite ( see also @xcite ) : + * * the only way to obtain a non - ckm suppressed isoscalar is to look at the semileptonic decay of the @xmath178-meson , where the light quark is a strange one and an isoscalar @xmath179 or / and gluonium state can be formed .
+ * * if the @xmath148 state is relatively light ( @xmath180 gev ) , which might be the natural partner of the ( @xmath181 ) often interpreted to be a @xmath182 in the literature , then , one should produce a @xmath183 pair through the isoscalar @xmath148 state .
the non - observation of this process will disfavour the @xmath5 interpretation of the @xmath96 and @xmath25 mesons .
+ * * if a gluonium state is formed it will decay with even strength into @xmath26 and a @xmath184 pairs .
therefore a gluonium formation in semileptonic @xmath178 decays should result in the decay patterns : d_s_b d_s_bk |k , with about the same rate up to phase space factors .
the observation of the semileptonic @xmath26 decay of the @xmath178 by e791 @xcite is a sign for glueball formation . and
using @xmath168 loop has been given in @xcite .
however , the same assumption for @xmath25 but using qssr leads to a negative conclusion @xcite . ] + * * using , e.g. , the result in @xcite , the one for light @xmath185 quarkonium production behaves as : ~|v_cq|^2g^2_f m^5_c |f_+(0)|^2 . * * the @xmath186 production , can be obtained from the @xmath187 behaviour of the @xmath188 box diagram . using dispersion techniques similar to the one used for @xmath189 processses @xcite , one obtains , assuming a @xmath178 and @xmath8-dominances @xcite : ~|v_cs|^2g^2_f , where @xmath190 is by definition proportional to @xmath191 .
then , one deduces : ~1|f_+(0)|^2 ^ 2 , which is @xmath192 for @xmath193 gev .
+ @xmath58 * comments * + these semi - quantitative results indicate that the gluonium production rate can be of the same order as the @xmath5 one contrary to the nave perturbative expectation ( @xmath194 suppression ) , which is a consequence of the ozi - rule violation of the @xmath8 decay .
-decays have been discussed in @xcite .
] however , it also shows that , due to the ( almost ) universal coupling of the @xmath8 to goldstone boson pairs , one also expects a production of the @xmath184 pairs , which can compete with the one from @xmath148 quarkonium state , and again renders more difficult the identification of the such @xmath148 state if allowed by phase space .
qssr does not have a precise systematic framework for extracting the properties of the radial excitations , except the approximate value of the mass indicated by the value of the qcd continuum threshold at which the mass of the ground state has been optimized . in order to check this result ,
one often uses ( if available ) empirical observations of the splittings between the radial excitations and the ground state , or / and arguments based on the ( linear ) regge trajectories . for this reason ,
the discussions which we shall give below will be very qualitative . + * the @xmath2 and @xmath195 mesons * + the @xmath22 and the @xmath196 are almost degenerated indicating the restoration of the @xmath38 flavour symmetry where the @xmath38 breakings behaves like @xmath197 and @xmath198 for the radial excitations .
these scalar radial excitations are also almost degenerated with the pseudoscalar @xmath199 and @xmath200 also indicating the restoration of the spontaneous breaking of the @xmath201 chiral symmetry broken by the quark condensate at higher scale . within such observations
, one can expect that these scalar states are @xmath5 states . + * the @xmath7 scalar mesons * + there are proliferations of these states from the data @xcite : @xmath202 and @xmath18 . using the previous symmetry restorations
, one may expect that the 1st radial excitation of the @xmath203 and @xmath148 unmixed states should be in the range of 1400 mev , which are in the vicinity of the @xmath19 and @xmath204 being good experimental candidates .
the 2nd radial excitations are expected to be in the range of the @xmath205 where good experimental candidates are the @xmath27 and @xmath18 . within these observations in mind , we shall interpret the different data given by bes @xcite : + @xmath58 the @xmath19 and @xmath27 are respectively produced through the ozi forbidden @xmath141 and allowed @xmath206 processes @xcite , while the latter is also produced from the glueball filter process @xmath207 .
these features can indicate that they can have an important gluonium component in their wave function . to a first approximation and iteration
, we assume that these two states result from the mixing : [ eq : mixing2 ] f_0(1370 ) + f_0(1710 ) + = & + - & + _b + s_2+_s s_3 which is a replica of the mixing among ground states in eq .
( [ eq : mixing1 ] ) .
we fix the coupling of the @xmath5 radial excitations by assuming that they behave like the one of the pion and @xmath199 , namely : g_s_2,3|pp^n g_s_2,3|pp , where @xmath208 ; @xmath209 .
then , we deduce : g_s_2^+^- & & 2g_s_2k^+k^-4.8 gev , + g_s_3k^+k^- & & 6.4 gev . using the experimental input from bes @xcite : ( f_0(1370 ) & & ( 26540 ) mev , + ( f_0(1370|kk)(f_0(1370)&&(0.080.08 ) , one can fix the two angles to be : _s75 ^ 0 _s-0.55 .
the solution indicates that the @xmath19 contains more @xmath210 than a gluonium , while the @xmath27 contains more glue and @xmath148 than @xmath210 , which is in line with general trends . from this analysis
, we predict : g^th_f_0(1710)^+^-2.8 gev , + g^th_f_0(1710)k^+k^-4.3 gev , which lead to the widths : 90 mev , + 176 mev .
the results are of the order of the bes data , which are respectively @xmath211 mev and 125 mev @xcite , but are not yet quite satisfactory . as the @xmath212 contains more glue after the mixing , it is natural that it will mix with the glueball @xmath213 with the theoretical parameters obtained previously .
+ @xmath58 therefore , in the 2nd step analysis , we consider that the observed @xmath27 and @xmath204 result from the mixing between @xmath212 and @xmath213 .
+ @xmath58 the @xmath204 is produced in the gluonium filter process @xmath214 , and it is needed for improving the fit of the ozi violating @xmath141 process , but has a small effect in the @xmath215 process @xcite .
on the other , one expects from previous sections that the @xmath213 couplings to @xmath26 and @xmath168 are small as the vertex sum rule is almost saturated by the @xmath8 , but we do not have a precise quantitative control of such couplings .
however , one can check that the model can not explain simultaneously the small @xmath216 and @xmath217 widths .
+ @xmath58 a solution to this problem may be given by the mixing of the previous two states with the @xmath18 and some 2nd radial excitations of the @xmath5 states expected to be in this range of energy .
has chiral coupling to pairs of goldstone bosons using perturbative qcd arguments .
however , the validity of this result has been questioned in @xcite due to instanton contributions . ]
the @xmath18 may contain more gluon in its wave function as it is produced in the glueball filter channel @xmath218 and in the ozi violating @xmath141 process , while it also decays into @xmath21 .
the smallness of the @xmath219 width may come from a destructive interference between the gluonium and @xmath5 states width is based on the restoration of chiral symmetry which favours the previous ozi violation process production of the @xmath18 @xcite .
however , the author does not explain the absence of its @xmath82 decay .
motivated by the various recent experimental progresses for producing scalar mesons , we have proposed new results and updated previous predictions of @xcite : + @xmath58 the unusual _ wrong _ splitting between the @xmath3 and @xmath4 being respectively a @xmath220 and @xmath221 mesons can be understood from qssr using the present value of @xmath222 .
+ @xmath58 the maximal meson - gluonium mixing for the @xmath12 and @xmath13 appears to be supported by the present data .
+ @xmath58 the @xmath223 , @xmath224 , @xmath12 and @xmath13 mesons appear to complete the scalar nonet below 1 gev .
+ @xmath58 the productions of the @xmath225=0 mesons above 1 gev through the ozi violating @xmath226 , @xmath15 decays and @xmath16 glueball filter processes may indicate that most of these @xmath7 mesons have important gluonium in their wave functions , where a simple meson - gluonium mixing scheme can explain the general features of the data .
+ @xmath58 our results suggest that the @xmath19 is mostly @xmath5 , while the @xmath17 and @xmath18 have significant gluonium component in their wave functions .
tests of these results can be provided by the measurements of the pure gluonium @xmath20 and @xmath21 specific @xmath11 decay channels .
+ we conclude that , after about a 1/4 century study , we still remain with more questions than answers on the true nature of scalar mesons .
it is a pleasure to thank jean - marc richard for some comments on the preliminary draft .
99 pdg2005 , s.e .
eidelman et al . , _ phys .
* b 592 * ( 2004 ) 1 . for a review
prior 2001 , see ee e.g. : l. montanet , _ nucl .
* 86 * ( 2000 ) 381 ; talk given at the gribov memory , wsc 2001 ; u. gastaldi , rencontres de physique de la valle daoste , 27 feb-4 march ( 2000 ) .
wa102 collaboration , d. barberis et al .
* b 479 * ( 2000 ) 59 [ hep - ex/0003033 ] .
bes collaboration , j. z. bai et al .
, hep - ex/0307058 ; m. ablikim et al . , hep - ex/0406038 ; hep - ex/0409007 ; hep - ex/0411001 ; for a review , see e. g. s. fang , hep - ex/0509034 .
e791 collaboration , i. bediaga , talk given at qcd05 , montpellier 4 - 8 july 2005 ; _ nucl .
_ * 121 * ( 2003 ) 104 .
focus collaboration , a. massafferi , talk given at qcd05 , montpellier 4 - 8 july 2005 .
kloe collaboration , a. aloisio et al .
* b 537 * ( 2002 ) 21 ; * b 536 * ( 2002 ) 209 .
snd collaboration , m.n .
achasov et al . , _ phys .
* b 479 * ( 2000 ) 53 ; * b 485 * ( 2000 ) 349 .
cmd2 collaboration , r.r .
akhmetshin et al .
* b 462 * ( 1999 ) 380 .
belle collaboration , m. nakao , talk given at qcd05 , montpellier 4 - 8 july 2005 . j. pelaez and f.j .
yndurain , hep - ph/0510216 ; r. kaminski , hep - ph/0510218 .
anisovich , hep - ph/0510409 ; v.v .
anisovich and a.v .
sarantsev , _ eur .
j. _ * a 16 * ( 2003 ) 229 ; v.v .
anisovich , d.v .
bugg and a.v .
sarantsev , _ phys .
rev . _ * d 58 * ( 1998 ) 111503 . for reviews , see e.g. : s. narison , qcd as a theory of hadrons , _ cambridge monogr .
* 17 * ( 2002 ) 1 [ hep - ph/0205006 ] ; _ world sci .
notes phys .
_ * 26 * , ( 1989 ) 1 - 529 ; _ riv
cim . _ * 10 n2 * ( 1987 ) 1 ; _ phys .
* 84 * ( 1982 ) 263 . for reviews ,
see e.g. , s. narison , _ nucl .
* 121 * ( 2003 ) 131 ; _ nucl .
* 96 * ( 2001 ) 244 ; _ nucl .
* 23 b * ( 1991 ) 280 ; _ nucl .
phys . _ * a 675 * ( 2000 ) 54c ; hep - ph/0009108 .
s. narison , _ nucl .
* b 509 * ( 1998 ) 312 ; _ nucl
* 64 * ( 1998 ) 210 .
s. narison , _ z. phys .
_ * c 22 * ( 1984 ) 161 ; _ phys .
lett . _ * b 125 * ( 1983 ) 501 .
c. morningstar and m. j. peardon , nucl - th / 0309068 ; g. bali et al .
* d 62 * ( 2000 ) 054503 ; a. hart and m. teper , _ phys .
_ * d 65 * ( 2002 ) 34502 ; c. michael et al .
, hep - lat/9909036 .
t. kunihiro et al . , hep - ph/0310312 .
s. narison , n. paver , e. de rafael and d. treleani , _ nucl .
phys . _ * b 212 * ( 1983 ) 365 . for recent reviews , see e.g. : s. narison , hep - ph/0202200 , _ nucl .
( proc . suppl . )
_ * b 86 * ( 2000 ) 242 .
k. chetyrkin and c. sturm , talk given at qcd05 , montpellier 4 - 8 july 2005 .
k. chetyrkin , s. narison and v.i .
zakharov , _ nucl .
* b 550 * ( 1999 ) 353 ; s. narison , _ phys .
lett . _ * b 300 * ( 1993 ) 293 . for reviews ,
see e. g. v. a. zakharov , hep - ph/0509114 ; f.v .
gubarev , m.i .
polikarpov and v. i. zakharov , hep - th/9812030 ; s. narison , hep - ph/0508259 .
m. jamin , j. a. oller and a. pich , eur .
j. * c 24 * ( 2002 ) 237 .
s. narison , hep - ph/0510108 .
s. narison , _ nucl .
* b 182 * ( 1981 ) 59 .
t. schaefer and e. v. shuryak , _ rev .
* 70 * ( 1998 ) 423 .
dosch and s. narison , _ nucl .
* 121 * ( 2003 ) 114 .
g. launer , s. narison and r. tarrach , _ z. phys .
_ * c 26 * ( 1984 ) 433 ; r.a .
bertlmann et al . ,
_ z. phys . _
* c 39 * ( 1988 ) 231 ; y. chung et al . ,
_ z. phys . _ * c 25 * ( 1984 ) 151 ; h.g .
dosch , m. jamin and s. narison , _ phys .
* b 220 * ( 1989 ) 251 .
s. narison , _ phys .
* b 216 * ( 1989 ) 191 .
k. maltman , hep - ph/9906267 . s. narison and n. paver , _ phys
. lett . _ * b 158 * ( 1985 ) 153 .
s. narison , _ phys .
* b 175 * ( 1986 ) 88 .
a. bramon and s. narison , _ mod .
lett . _ * a 4 * ( 1989 ) 1113 .
jaffe , _ phys .
* d 15 * ( 1977 ) 267 ; * d 17 * ( 1978 ) 1444 .
latorre and p. pascual , _ jour .
_ * g11 * ( 1985 ) l231 .
m. alford and r.l .
jaffe _ nucl .
* b 509 * ( 1998 ) 312 .
brito , f.s .
navarra , m.nielsen , m.e .
bracco , _ phys .
* b 608 * ( 2005 ) 69 .
p. di vecchia and g. veneziano , _ nucl .
_ * b 171 * ( 1981 ) 253 .
see e.g. : kyoto workshop on the @xmath96 meson , kek proceedings 2000 - 4 .
d. black , m. harada and j. schechter , _ nucl .
* 121 * ( 2003 ) 95 ; j. ellis and j. lanik , _ phys .
* b 150 * ( 1985 ) 289 ; h. gomm et al .
rev . _ * d 33 * ( 1986 ) 801 ; d. ebert et al .
, hep - ph/0007131 ; a. boduylov and v. novozhikov , desy report 90 - 151 ( 1990 ) ; g. mennessier , _ z. phys .
_ * c 16 * ( 1983 ) 241 .
s. narison and g. veneziano , _ int
phys . _ * a 4 , 11 * ( 1989 ) 2751 .
novikov et al . , _ nucl .
* b 191 * ( 1981 ) 301 .
s. narison , _ phys .
* b 361 * ( 1995 ) 121 ; _ phys .
* b 387 * ( 1996 ) 162 .
s. narison , g. shore and g. veneziano , _ nucl .
* b 433 * ( 1995 ) 209 ; _ nucl .
phys . _ * b 546 * ( 1999 ) 235 ; for a recent review , see e.g. s. narison , hep - ph/0601066 .
steele , d. harnett and g. orlandini , hep - ph/0308074 .
h. forkel , hep - ph/0312049 .
steele , d. harnett and g. orlandini , hep - ph/0108232 .
novikov et al .
* b 165 * ( 1980 ) 67 .
p. minkowski and w. ochs , _ eur .
j. _ * c 9 * ( 1999 ) 283 ; _ nucl .
_ * 121 * ( 2003 ) 119 ; _ nucl .
* 121 * ( 2003 ) 121 .
e. witten , _ nucl .
phys . _ * b156 * ( 1979 ) 269 ; + g. veneziano , _ nucl . phys . _
* b159 * ( 1979 ) 213 . a. di giacomo , _ nucl .
* b 23 * ( 1991 ) 191 .
f. gilman and r. kauffman , _ phys .
rev . _ * d 36 * ( 1987 ) 2761 ; l. montanet , _ non - perturbative methods conf .
_ , montpellier , ( wsc 1985 ) .
g. mennessier , s. narison and n. paver , _ phys .
* b 158 * ( 1985 ) 153 ; n. pak , s. narison and n. paver , _ phys .
* b 147 * ( 1984 ) 162 .
pennington , hep - ph/0604212 .
achasov and a.v .
kiselev , hep - ph/0512047 .
a. bramon et al .
, hep - ph/000818 ; yu .
kalashnikova et al .
, nucl - th/0512028 ; t. barnes , _ phys . lett . _
* b 165 * ( 1985 ) 434 ; n. achasov and g.n .
shestakov , _ phys .
_ * d 72 * ( 2005 ) 013006 .
a. deandrea et al . , hep - ph/0012120 .
s. s. gershtein , a. a. likhoded and y. d. prokoshkin , _
z. phys . _ * c24 * ( 1984 ) 305 . h. g. dosch et al . , hep - ph/0203225 ; m. nielsen et al . ,
* 121 * ( 2003 ) 110 .
kisslinger , hep - ph/0103326 .
m. nielsen , f.s .
navarra and i. bediaga , _ nucl .
* 133 * ( 2004 ) 178 .
p. minkowski and w. ochs , hep - ph/0511126 .
glozman , hep - ph/0507206 .
kataev , _ phys . atom .
_ * 68 * ( 2005 ) 567 ; hep - ph/0406305 and references threin .
f. giacosa , th .
gutsche , v.e .
lyubovitskij and a. faessler , hep - ph/0504033 .
c. amsler and f.e .
close , _ phys .
* d 53 * ( 1996 ) 295 ; d. weingarten , _ nucl .
proc . suppl . _
* 73 * ( 1999 ) 283 ; c. amsler , hep - ph/0206104 ; f.e .
close and a. kirk , hep - ph/0103173 .
j. vijande et al . , hep - ph/0508142 .
chanowitz , hep - ph/0506125 .
z.f.zhang , h.y.jin , hep - ph/0511252 . | motivated by several recent data , we test the qcd spectral sum rules ( qssr ) predictions based on different proposals ( @xmath0 , @xmath1 , and gluonium ) for the nature of scalar mesons . in the @xmath2 and 1/2 channels , the unusual _
wrong _ splitting between the @xmath3 and @xmath4 and the @xmath3 width can be understood from qssr within a @xmath5 assignement . however
, none of the @xmath5 and @xmath1 results can explain the large @xmath6 width , which may suggest that it can result from a strong interference with non - resonant backgrounds . in the @xmath7 channel , qssr and some low - energy theorems ( let )
require the existence of a low mass gluonium @xmath8(1 gev ) coupled strongly to goldstone boson pairs which plays in the @xmath9 channel , a similar role than the @xmath10 for the value of the @xmath11 topological charge .
the observed @xmath12 and @xmath13 mesons result from a maximal mixing between the gluonium @xmath8 and @xmath5(1 gev ) mesons , a mixing scheme which passes several experimental tests .
ozi violating @xmath14 , @xmath15 decays and @xmath16 glueball filter processes may indicate that the @xmath17 and @xmath18 have significant gluonium component in their wave functions , while the @xmath19 is mostly @xmath5 .
tests of these results can be provided by the measurements of the pure gluonium @xmath20 and @xmath21 specific @xmath11 decay channels . |
it is generally believed that ferromagnetism should be explained in the framework of a hamiltonian for strongly coupled electrons .
such a hamiltonian on a lattice has been proposed 40 years ago by hubbard . @xcite
it contains a nearest neighbour hopping term @xmath0 and an on - site coulomb repulsion ( of strength @xmath1 ) @xmath2 for the electrons . both terms are `` blind '' to the spin of the electrons and it is therefore far from being obvious that such a model should lead to ferromagnetism .
the limit @xmath3 here called hard - core condition enforces that each site can be occupied by at most one electron . in this limit , the number @xmath4 of electrons which can be accomodated on a lattice of @xmath5 sites is limited @xmath6 . in the charge sector @xmath7
every spin configuration of the electrons is a possible ground state . on a square lattice
this degeneracy is lifted in the sector @xmath8 with one hole .
nagaoka@xcite constructed the ground state and found that the total spin @xmath9 of the electrons is maximal : @xmath10 a clear signature for ferromagnetism .
there has been an intensive search for ferromagnetism in the @xmath11 model with @xmath12 holes .
numerical results on finite clusters @xcite at vanishing exchange coupling ( @xmath13 , see eq.([h_tj ] ) ) show that the ground state does not have maximal spin except for the nagaoka case @xmath8 .
the instability of the nagaoka ferromagnetic state in sectors with 2 and more holes has been investigated with variational ground states of the gutzwiller type.@xcite it was found that there exists a critical hole density @xmath14 such that the ground states can not be ferromagnetic for @xmath15 .
improvements of the variational ansatz led to a successive decrease of @xmath16 for the @xmath17 case @xmath18 on the other hand barbieri , riera and young@xcite have argued that the ferromagnetic ground state emerges in the thermodynamical limit @xmath19 , if the hole density vanishes ( @xmath20 ) .
the existence of a lower bound @xmath15 for the hole density @xmath21 , where ferromagnetism is not possible , is indeed a quite general feature , which has been found also in higher dimensions and for various types of lattices.@xcite an introduction to ferromagnetism in the hubbard model can be found in ref .
[ ] . in this situation one might ask for the reason why hard - core models with fermions ( i.e. electrons ) fail to `` explain '' ferromagnetism on finite clusters ( except for the 1 hole case ) .
we want to demonstrate in this paper that this failure can be traced back to the anticommutation relations for the fermion operators . for this purpose ,
we substitute in the @xmath11 model the fermionic degrees of freedom by bosonic ones . in order to facilitate the comparison of the fermionic and bosonic version , we assume that the bosons carry here as well spin 1/2 and experience the same antiferromagnetic interaction.@xcite therefore the hamiltonians for a hard - core model with bosons ( @xmath22 ) and fermions ( @xmath23 ) are both blind with respect to the spin .
nevertheless , we find pronounced differences ( and similarities ) in the ground state energies @xmath24 , @xmath25 and ground state spins @xmath26 , @xmath25 .
\(a ) for @xmath27 the ground state energies and ground state spins coincide @xmath28 this means in particular that the construction of the nagaoka ferromagnet state holds for fermions and for bosons .
the restriction to the one hole sector @xmath29 , @xmath8 is crucial .
moreover one observes a hole - particle symmetry for @xmath30 , @xmath31 \(b ) for @xmath32 the ground state energies coincide [ ( [ e_bf ] ) for @xmath32 ] , but the ground state spins are different @xmath34 i.e. in the bosonic case the ground state spin is maximal .
\(c ) for @xmath35 ground state energies and ground state spins are different for the fermionic and bosonic version . in the bosonic version
the ground state spin @xmath36 is maximal as shown in appendix [ appendix_a ] .
the ground state energies @xmath37 show the hole particle symmetry ( [ p - h - symm ] ) for these @xmath4-values .
this symmetry is not present in the fermionic version . in the latter case
the ground state spin @xmath38 is `` erratic''.@xcite we think that these properties are interesting enough to justify a detailed study of the ground state properties of the bosonic @xmath11 model and to compare its properties with the corresponding properties in the fermionic version . the outline of the paper is the following : in section[sec2 ] we review the definition of the @xmath11 model and point out the differences between the fermionic and bosonic version .
consequences for the ground state energies on the smallest ( @xmath39 ) cluster are discussed in section[sec3 ] .
the phase diagram of the bosonic @xmath11 model in dimensions @xmath40 and @xmath41 is discussed in section[sec4 ] and [ sec5 ] , respectively .
let us first recall the definition of the @xmath11 hamiltonian with fermionic degrees of freedom : @xmath42 the hopping term @xmath43 and the spin exchange part @xmath44 can be expressed in terms of creation ( @xmath45 ) and annihilation ( @xmath46 ) operators for the electrons , which obey anticommutation relations @xmath47 @xmath48 is the number operator for an electron at site @xmath49 with spin @xmath50 .
owing to the `` hard - core condition '' which is imposed by the projector @xmath51 double occupancy is forbidden i.e. @xmath52 .
the latter can be derived from a hubbard model@xcite with infinite on - site coulomb repulsion ( @xmath53 ) .
the construction of a state with @xmath4 electrons @xmath54 by application of creation operators to the vacuum @xmath55 demands an ordering of all sites on the 2 dimensional square lattice . owing to the anticommutation rules ( [ anticommutation ] ) a different ordering ( e.g. with @xmath56 interchanged ) leads to a state which might differ from the former one in sign . in the following
we will denote the traditional model ( [ h_tj])([h_j ] ) with fermions as `` fermionic @xmath11 model '' .
let us now turn to the `` bosonic @xmath11 model '' , which we simply define by substituting the anticommuting creation and annihilation operators @xmath57 by commuting ones : @xmath58 & = & \big[a_{\sigma}(x),a_{\sigma'}(x')\big]=0 \nonumber\\ \big[a^+_{\sigma}(x),a_{\sigma'}(x')\big ] & = & \delta_{\sigma,\sigma'}\delta_{x , x'}\ , .
% n_{\sigma'}(x')\ , .
\nonumber\\ \label{commutation}\end{aligned}\ ] ] the eigenvalues of the number operator @xmath59 are restricted again to @xmath60 due to the `` hard - core condition '' .
this model has been discussed first in references ( ) .
quite recently a generalized hubbard model with fermionic and bosonic degrees of freedom has been proposed and investigated in order to study a mixture of ultracold bosonic and fermionic atoms in an optical lattice.@xcite the construction of a state with @xmath4 bosons by application of creation operators on the vacuum , however , is symmetric under the permutation of sites due to the commutation relations ( [ commutation ] ) . there does not exist a `` sign problem '' .
this is the only difference between the fermionic and bosonic version of the @xmath11 model [ cf .
one can easily verify , that the action of the spin exchange term ( [ h_j ] ) is the same in the fermionic and bosonic version . on the other hand the hopping term @xmath0 acts indeed in a different way on the fermionic and bosonic states . in the latter case @xmath0
can be expressed in terms of nearest neighbour permutations @xmath61
which interchange a particle and a hole at sites @xmath49 and @xmath62 : @xmath63 here @xmath64 is the number of bosons at site @xmath49 . in appendix [ appendix_a ] the hamiltonian ( [ h_tb ] )
is proven to have a ferromagnetic ground state for all @xmath4 .
moreover it is shown that the ground state energies of the bosonic @xmath11 model at @xmath13 are symmetric under the particle hole transformation @xmath65 as can be seen in fig .
[ fig1 ] . and a @xmath66 lattice for both fermionic and bosonic @xmath11 models at @xmath13.,width=340 ] here we also show the ground state energy for hard - core fermions , where the hole particle symmetry ( [ e_b_ph ] ) is lost ( except for the nagaoka case @xmath8 ) @xmath67 moreover the ground state energies of the hard - core bosons are below those of the hard - core fermions @xmath68 comparing the results from lattices @xmath69 , @xmath70 we find that the finite - size dependence is rather small and smooth in the bosonic case , in contrast to the fermionic case .
the result presented in fig .
[ fig1 ] were obtained with a lanczos algorithm . concerning the fermionic version , they agree with ref .
( ) ( riera , young).@xcite in both versions ( fermionic and bosonic ) the @xmath11 hamiltonian ( [ h_tj ] ) conserves the total charge @xmath71 and total spin @xmath72 such that the eigenvalues of @xmath73 can be used to characterize the eigenstates of the hamiltonian .
@xmath74 have maximal spin @xmath75 .
these states are simultaneous eigenstates of the hopping part @xmath0 and the spin coupling part @xmath76 @xmath77 such that the eigenvalues @xmath78 for the @xmath11 hamiltonian ( [ h_tj ] ) are independent of @xmath79 .
this is a consequence of the fact that the hopping part @xmath0 is blind with respect to the spin of the electrons .
there is no difference in the hopping of spin - up and spin - down particles . on the other hand , the ferromagnetic eigenstates @xmath74
are totally symmetric under any permutation of spin - up and spin - down particles .
therefore , the application of the nearest neighbour spin exchange coupling @xmath80 yields zero : @xmath81
on a plaquette with 4 sites ( @xmath39 ) the @xmath11 hamiltonian can be diagonalized analytically.@xcite it is interesting to compare the ground state energies @xmath82 and total spins @xmath83 in the sectors with @xmath4 spin-1/2 particles for the fermionic and bosonic version .
it turns out they are the same in the sectors @xmath84 * @xmath85 + where the @xmath79 intervals @xmath86 , @xmath87 , @xmath88 are defined by @xmath89 ground state energies with index `` @xmath90 '' correspond to ground states with maximal spin @xmath75 . in the sector with three particles ( @xmath91 ) one finds the same energy in the fermionic and bosonic version of the @xmath11 model for all @xmath92 .
they are different in the sector @xmath32 * \a ) in the fermionic version , @xmath93 the ground state is nondegenerate and has total spin 0 .
+ \b ) in the bosonic version : @xmath94 we find first a ferromagnetic ground state with maximal spin 1 and for larger @xmath79 values degenerate spin 0 ground states .
note that the ground state energy of the fermionic model ( [ e_p_q2_f ] ) is below that of the bosonic model ( [ e_p_q2_b1 ] ) , ( [ e_p_q2_b2 ] ) for @xmath95 .
the two energies coincide at @xmath13 .
it is remarkable to note , that some of these similarities can be found on all square lattices @xmath69 , @xmath96 .
most important is the equality @xmath97 in the 1 hole sector for all @xmath95 .
the @xmath79 dependence is shown in fig .
[ fig2 ] .
-dependence of ground state energies in the 1 hole sector of a @xmath69 lattice for @xmath95 .
bosonic and fermionic @xmath11 ground state energies for @xmath70 coincide.,width=302 ] in the sector with 2 particles ( @xmath32 ) at @xmath13 the ground state energies are the same , but the ground state spin is always @xmath98 in the fermionic , but @xmath99 in the bosonic version , as already stated in ( [ remark_q2 ] ) .
in this paper we are mainly concerned with the question , whether the ground state with @xmath4 ( spin-1/2 ) particles has maximal total spin @xmath75 .
the answer to this question depends on the strength @xmath79 of the spin exchange part ( [ h_tj ] ) which prefers ( for @xmath95 ) antiferromagnetic ordering with total spin @xmath100 .
owing to the property ( [ h_j_f ] ) the expectation value of the spin exchange part @xmath101 can be considered as an order parameter ; it vanishes if the ground state @xmath102 in the sector with @xmath4 particles has maximal spin @xmath75 , but is nonzero in all other cases .
we will demonstrate for the bosonic version of the @xmath11 model on @xmath69 clusters that the ferromagnetic regime : @xmath103 is bounded by a curve @xmath104 depending on the hole density @xmath105 . in our numerical calculations of the ground states @xmath102 , @xmath106 in the bosonic @xmath11 model with periodic boundary conditions
, we only found ferromagnetic states @xmath107 for @xmath79 small enough .
[ fig3 ] shows the order parameter @xmath108 on a @xmath109 cluster in the charge sectors @xmath110 .
the phase boundary @xmath111 can be read off from those @xmath79-values where @xmath108 changes first from 0 to a nonvanishing value .
obviously , @xmath112 is monotonically decreasing with @xmath113 and vanishes for @xmath114 .
its @xmath113-dependence is shown in fig .
[ fig4 ] . for increasing values of @xmath4 , we observe more and more jumps in the order parameter @xmath108 .
e.g. for @xmath115 we find three of these jumps .
each of them signals the transition to a new ground state with increasing values of @xmath79 .
the ground states differ in their total spin @xmath116 which is denoted by the integer ( @xmath4 even ) and halfinteger ( @xmath4 odd ) numbers at the branches between two jumps .
e.g. for @xmath115 : @xmath117 so far we were not able to localize the spin couplings for the ground state with total spin @xmath118 , which we expect to emerge in the interval @xmath119 . for increasing values of @xmath120 , @xmath112 approaches zero and it is more and more difficult to localize the @xmath79 values for those ground states with higher spin @xmath121 .
we performed higher resoluted searches with step width @xmath122 and found e.g. in the @xmath123 sector the ground state with total spin @xmath124 at @xmath125 .
in contrast , a similar search for the @xmath126 ground state in the @xmath127 sector failed . as an alternative we also tried to localize those @xmath79 values : @xmath128 where the total spin @xmath116 of the ground state is first minimal .
the curves @xmath129 , @xmath130 , and @xmath112 are shown in fig .
[ fig4 ] .
all curves are monotonically decreasing with @xmath113 .
we have looked for the finite - size dependence of the curve @xmath112 by comparing results on @xmath69 lattices for @xmath131 on the larger systems with @xmath132 we assumed that the ground state spin @xmath133 is maximal and the curve @xmath112 was extracted from those @xmath79-values , where the spin is lowered first by one or two units . as can be seen from fig .
[ fig5 ] , the finite - size effects can be rather well accounted for by an effective charge density @xmath134 such that @xmath135 scales with the system size @xmath136 . by comparison of results on @xmath69 lattices for @xmath137 .
the finite - size effects has been accounted for by @xmath138 . ,
width=340 ] a similar finite - size analysis for the curves @xmath129 and @xmath139 , which mark the boundaries of the antiferromagnetic phase with minimal ground state spin demands simulations on much larger systems .
they might be accessible with monte carlo techniques if it is possible to determine the ground state spin with high accuracy .
the question of physical interest is , wether the transition region between the ferro- and antiferromagnetic phase in fig .
[ fig4 ] shrinks to zero in the thermodynamical limit .
in one dimension the difference between the bosonic and fermionic version of the @xmath11 model can be absorbed in the boundary conditions .
it turns out that the ground state energies with periodic boundary conditions on a @xmath140 site system coincide for all @xmath79 if @xmath4 is odd @xmath141 for @xmath4 even one observes slight differences for @xmath142 , as can be seen from fig .
[ fig6 ] . ) for a one dimensional periodic chain @xmath140 ( @xmath41 ) for charges @xmath143.,width=340 ] at @xmath13 the equality ( [ eb_ef_al0 ] ) strictly holds for all @xmath4-values and all the ground state energies satisfy the hole particle symmetry ( [ p - h - symm ] ) . concerning the total spin of the ground state @xmath26 @xmath144 at @xmath13 we observe * a maximal value @xmath36 for all @xmath4 * a maximal value @xmath145 only for odd @xmath4 for @xmath4 even the ground states appear to be degenerate with different values for @xmath38 of the total spin .
this leads to the `` erratic '' behaviour observed already in the two dimensional system @xmath40 .
differences between the bosonic @xmath11 model for @xmath41 and @xmath40 become apparent in the variation of the order parameter ( [ eps_j ] ) and the total spin @xmath146 with @xmath79 .
they are shown in fig .
[ fig7 ] for @xmath41 and in fig .
[ fig3 ] for @xmath40 . for a @xmath147 cluster with charge sectors @xmath148 .
small numbers denote the total spin of the bosonic @xmath11 ground states.,width=340 ] the main difference between the two cases @xmath40 and @xmath41 is the following : the ferromagnetic domain @xmath149 where the total spin is maximal shrinks to zero in the 1@xmath150 case .
in particular for @xmath151 , the boundary @xmath112 turns out to be below @xmath152 already on finite chains .
in this paper we have demonstrated that the emergence of ferromagnetism in a hard - core model crucially depends on the statistics of the constituent particles , i.e. whether they are fermions or bosons .
we studied the @xmath11 model in the ( traditional ) fermionic ( @xmath23 ) and in a bosonic ( @xmath22 ) version and compared ground state energies @xmath153 and total spins @xmath154 , @xmath25 , in both models . in particular , we found for @xmath13 that in contrast to hard - core fermions hard - core bosons ( with spin 1/2 ) have a ferromagnetic ground state with maximal spin @xmath36 in all charge sectors q. this means that the corresponding hamiltonian @xmath0 ( [ h_tb ] ) leads intrinsically to a ferromagnetic interaction between spin - up and spin - down particles , as we have demonstrated first on finite square lattices @xmath69 , @xmath96 numerically . a general proof is given in appendix [ appendix_a ] .
the ground state energies for hard - core bosons and fermions behave differently on finite lattices .
they show hole particle symmetry ( [ e_b_ph ] ) in the bosonic version , whereas this symmetry is not present in the fermionic version .
we do not know whether this breaking is a finite - size effect and whether the ground state energies ( fig . [ fig1 ] ) converge to each other in the thermodynamical limit .
if we switch on the spin exchange coupling @xmath79 [ ( [ h_j ] ) , ( [ su3_breaking ] ) ] , the bosonic @xmath11 model in two dimensions ( @xmath40 ) develops a ferromagnetic regime @xmath149 , where @xmath112 is monotonically decreasing with @xmath113 . for @xmath41 ,
@xmath112 shrinks to zero .
ferromagnetism is observed in the fermionic and bosonic @xmath11 model in the sectors @xmath30 ( 1 particle ) and @xmath8 ( 1 hole ) . here ,
the two @xmath11 models are unitary equivalent , which implies that the sign problem for the electrons can be absorbed in a unitary transformation.@xcite in other words `` electrons behave as bosons '' .
this behaviour can be found as well for hard - core electrons in @xmath155 with periodic boundary conditions and @xmath4 odd . on one hand
all ground state energies @xmath156 @xmath157 show the hole particle symmetry ( [ p - h - symm ] ) and are strictly identical with the corresponding quantities @xmath158 in the bosonic version .
the ground state spins @xmath38 are maximal for @xmath4 odd , but this is not the case for @xmath4 even , where the ground state appears to be degenerate with different total spins .
let us finally comment on possible more realistic hard - core boson models for ferromagnetism .
on one hand these bosons have to carry spin , on the other hand the spin statistics theorem demands integer spin for bosons .
it is quite straightforward to extend the hopping hamiltonian ( [ h_tb ] ) to spin 1 bosons . in this case , the number of particles at site @xmath49 : @xmath159 has to be identified with the sum of spin @xmath160 particles .
such a hamiltonian is again invariant under particle hole transformations .
it defines a hard - core boson model with 4 degrees of freedom at each site : 3 for the spin-1 particles , 1 for the holes .
we want to present first a general proof that the ferromagnetic state with all spins up ( @xmath161 ) is indeed a ground state of @xmath0 ( [ h_tb ] ) in the sector with @xmath4 particles . here , the permutation operator @xmath61 for a spin ( @xmath162 ) and a hole can be represented : @xmath163 in terms of pauli matrices at neighbouring sites @xmath49 and @xmath62 which act on particle @xmath164 and hole @xmath165 states as @xmath166 it is easy to verify that ( [ h_tb ] ) in the sector @xmath161 can be mapped on the @xmath167 spin model in two dimensions : @xmath168 the hamiltonian ( [ h_xx ] ) is invariant under the inversion of all `` @xmath169 spins '' ( [ tau_spins ] ) @xmath170 which explains the hole particle symmetry ( [ e_b_ph ] ) .
note , that all the matrix elements in @xmath0 have a negative sign .
therefore we can apply the perron - frobenius theorem,@xcite which states that the ground state @xmath171 is nondegenerate and that in the defining basis all components are greater than zero . since @xmath0 conserves the total spin ( [ s_total ] ) , the application of lowering operators @xmath172 onto the ground state @xmath171 with eigenvalue @xmath78 yields eigenstates of @xmath0 : @xmath173 h_t\psi_f(s_z = q/2-n ) & = & e_f(q)\psi_f(s_z = q/2-n)\nonumber\\\end{aligned}\ ] ] in the sector with @xmath174 with the same eigenvalue @xmath78 .
indeed these states are again ground states in the sector @xmath174 for the following two reasons : * the state @xmath175 has again only positive components in the defining basis in the sector @xmath174 . *
the matrix elements of @xmath0 in the sector @xmath174 are all negative , such that the perron - frobenius theorem can be applied again : the ground state is nondegenerate and has only positive components
. therefore @xmath175 must be the ground state for @xmath174 . finally , we want to stress that all the states @xmath175 are symmetric under all permutations : @xmath176 for the quantum numbers of the particles at sites @xmath49 and @xmath62 , which guarantees that @xmath177 j. hubbard , proc .
soc . * a*276 , 238 ( 1963 ) m. c. gutzwiller , phys .
rev . lett.*10 * , 159 ( 1963 ) j. kanamori , prog .
physics*30 * , 275 ( 1963 ) y. nagaoka , phys . rev . * 147 * , 392 ( 1966 ) j. a. riera , a. p. young , phys . rev . b*40 * , 5285 ( 1989 ) h. fehske , v. waas , h. rder , h. bttner , phys .
b*44 * , 8473 ( 1991 ) , h. fehske : habilitationsschrift [ universitt bayreuth ( 1996 ) , unpublished ] th .
hanisch , e. mller - hartmann , ann .
der physik * 2 * , 381 ( 1993 ) ; p. wurth , g. uhrig , e. mller - hartmann , ann .
der physik * 5 * , 148 ( 1996 ) a. barbieri , j. a. riera , a. p. young , phys .
rev . b*41 * , 11697 ( 1990 ) th .
hanisch , b. kleine , a. ritzl , e. mller - hartmann , ann .
der physik * 4 * , 303 ( 1995 ) ; m. bonisegni , phys .
lett.*87 * , 087201 ( 2001 ) m. bonisegni , phys .
rev . b*65 * , 134403 ( 2002 ) e. altman , w. hofstetter , e. demler , m. d. lukin , new j. phys.*5 * , 113 ( 2003 ) j. smakov , c. d. batista , g. ortiz , phys .
rev . lett.*93 * , 067201 ( 2004 ) m. lewenstein , l. santos , m. a. baranov , h. fehrmann , phys .
92 , 050401 ( 2004 ) a. fledderjohann , a. langari , k .- h .
mtter , j. phys . : condens .
matter*16 * , 8571 ( 2004 ) . except for the incorrect ground state energy for the case @xmath179 , @xmath180 , where the true ground state [ @xmath181 has total spin @xmath182 as previously noted by fehske@xcite in his habilitation thesis . | the problem of ferromagnetism associated with a ground state with maximal total spin is discussed in the framework of a hard - core model , which forbids the occupancy at each site with more than one particle .
it is shown that the emergence of ferromagnetism on finite square lattices crucially depends on the statistics of the particles .
fermions ( electrons ) lead to the well - known instabilities for finite hole densities , whereas for bosons ( with spin ) ferromagnetism appears to be stable for all hole densities . |
clusters of galaxies have been widely used to provide useful constraints on cosmological parameters ( e.g. , henry & arnaud 1991 ; bahcall & fan 1998 ; eke et al .
1998 ; borgani et al . 2001 ) .
this is in part possible because there is a well - developed theoretical framework which allows one to predict the mass function ( mf ) of clusters of galaxies and its evolution as a function of the cosmology . here
, the mf is defined as the number density of clusters as a function of their mass .
one standard method for predicting the mf is to use press - schechter formalism , originally developed by press & schechter ( 1974 , hereafter ps ) , and developed in more detail by bond et al .
( 1991 ) and lacey & cole ( 1993 ) , among others , in combination with the cold dark matter ( cdm ) model . in hierarchical structure formation models like cdm , more massive halos form from the merger of smaller halos .
values of cosmological parameters can be estimated by comparing theoretical models for the mf of clusters with observations .
this technique places the strongest constraints on @xmath0 , the rms mass fluctuations on a scale of @xmath10 mpc where @xmath11 is the hubble constant in units of @xmath12 km / sec / mpc , and on @xmath13 , the ratio of the current mean matter density to the critical density @xmath14 at the current epoch ( e.g. , henry & arnaud 1991 ; kitayama & suto 1996 ; eke et al .
1998 ; borgani et al . 2001 ; ikebe et al . 2002 ) . roughly speaking
, the present - day abundance of clusters determines the relationship between @xmath0 and @xmath15 , while the evolution of clustering with redshift breaks this degeneracy , although reiprich & bhringer ( 2001 ) have been able to constrain @xmath15 using only a local cluster sample by assuming a cdm spectrum of perturbations whose shape parameter depends on @xmath15 . in these determinations ,
the most massive clusters have the greatest leverage .
massive clusters are rare objects , and the abundance of the most massive clusters is very sensitive to the cosmological parameters . unfortunately , the mf of clusters of galaxies can not be directly observed ( save by gravitational lensing in a relatively small number of cases ) .
generally , it is inferred from the x - ray luminosity function ( xlf ) or temperature function ( tf ) , using empirical or semi - empirical relations between the masses of clusters and their temperatures ( @xmath16 ) or x - ray luminosities ( @xmath17 ) .
these scaling relations are usually applied under the assumption that the clusters are dynamically relaxed , although in reality this can not always be true since in the cdm model larger clusters are continually forming via mergers of smaller clusters .
simulations have shown that if two clusters of comparable mass merge to form a larger cluster , there is a temporary increase in the cluster s x - ray luminosity and temperature ( ricker & sarazin 2001 ; ritchie & thomas 2002 ) . if the cluster is observed during this period of boosted luminosity and temperature , the inferred mass will be larger than the actual mass of the cluster .
in fact , such objects will be preferentially detected in x - ray flux - limited samples because they are intrinsically more luminous than equal mass clusters which are not currently experiencing a merger .
this will be particularly true for x - ray selected , high redshift clusters . as a possible examples of this , one of the most distant x - ray selected clusters observed to date , rxj1053.7 + 5735 , has a morphology which may indicate an ongoing merger ( hashimoto et al .
2002 ) , as does the most luminous x - ray selected cluster found to date , rxj1347.5 - 1145 ( allen , schmidt , & fabian 2002 ) .
if this x - ray luminosity - temperature ( hereafter l - t ) boost effect is sufficiently large , and if mergers occur frequently enough , then the observed xlf and tf , and hence the inferred mf , will be different from what they would be if all clusters were dynamically relaxed .
one would expect that merger boosts would affect most strongly the high luminosity and temperature ends of the xlf and tf .
since the massive clusters which have very high values of @xmath17 and @xmath16 are rare , a small contribution of lower mass clusters with l - t boosts could strongly affect the statistics . as we noted above
, the abundance of the hottest and most luminous clusters ( which are usually assumed to be the most massive clusters ) has a strong influence on the inferred values of the cosmological parameters .
in particular , if merger boosts artificially increase the abundance of hot , luminous clusters at high redshift , the real value of @xmath15 will be underestimated . since observations of the numbers of high temperature and luminosity clusters at moderate and high redshifts have been used to infer that @xmath15 is low ( e.g. , bahcall , fan , & cen 1997 ; borgani et al .
2001 ) , this effect could be important .
in this paper we attempt to quantify the effect of mergers on the observed xlf and tf , and to determine how the x - ray l - t boost associated with mergers alters the values of @xmath0 and @xmath15 inferred from observations .
we consider three possible cosmologies : an `` open '' model ( @xmath18 , @xmath19 ) , a `` flat '' model ( @xmath20 ) , and an einstein - de sitter ( eds ) model ( @xmath21 , @xmath19 ) . here , @xmath22 and @xmath23 is the cosmological constant .
we use extended press - schechter ( eps ) theory ( [ sec : basic_ps ] ) and a monte carlo technique to build a collection of merger histories of clusters ( `` merger trees , ''
[ sec : trees ] ) for a variety of cosmologies .
we then build two sets of xlfs and tfs from the mf given by the merger trees for each cosmology by averaging together many merger trees ( [ sec : xlf+tf ] ) .
one set assumes that clusters are always relaxed when transforming from mass to temperature or luminosity ( [ sec : m - t+m - l ] , [ sec : no_boosts ] ) , while the other set includes the l - t boost effect ( [ sec : with_boosts ] ) .
we ignore such non - gravitational effects as preheating since such effects are mainly important for lower mass clusters and groups ( ponman et al .
1996 ) , whereas we choose to focus on massive clusters where the effects of merger boosts are most important .
we also ignore cooling flows at the centers of clusters , which may be disrupted by mergers .
the disruption of cooling flows may increase the effect of mergers on x - ray temperatures , but decrease their effect on x - ray luminosities .
the strength and duration of the l - t boost is estimated for arbitrary merging masses and impact parameters by interpolating and extrapolating from the results of a series of n - body / hydrodynamical simulations of binary cluster mergers ( [ sec : hydro ] and appendix b ) .
the angular momenta or impact parameters for the mergers are drawn from a distribution based on linear theory ( [ sec : impact_param ] ) .
the boosted xlfs and tfs are presented in
[ sec : xlf+tf ] at a variety of redshifts , and are compared to unboosted results .
the boosted and unboosted xlfs and tfs at several redshifts are fitted using a ps mass function and equilibrium relations between the mass and temperature or x - ray luminosity ( [ sec : tree_fit ] ) .
the best - fit parameters , specifically @xmath0 and @xmath15 , are compared to the `` actual '' parameters used to build the trees .
our conclusions are summarized in [
sec : discuss ] .
the quantity which is given most directly by theory is the mf of clusters , @xmath24 , which is defined such that @xmath25 gives the number of clusters per unit comoving volume with masses in the range @xmath26 .
the mf of clusters at various redshifts and for various cosmologies can be derived from numerical simulations of clustering , or from semi - analytic techniques .
since we wish to compare the effects of merger boosts on the properties of an otherwise identical ensemble of clusters in several cosmologies , it is easier to use the semi - analytical techniques .
a simple expression for the mass function of clusters is given by ps formalism .
comparisons to observations of clusters and to numerical simulations show that ps provides a good representation of the statistical properties of clusters , if the ps parameters are carefully selected ( e.g. , lacey & cole 1993 ; bryan & norman 1998 ) . based on large n - body simulations ,
some authors have suggested other forms for the mass function of dark matter halos which differ somewhat from the ps predictions ( e.g. , sheth & tormen 1999 ; jenkins et al .
since we are mainly interested in determining the relative effects of merger boosts on the xlf and tf , the exact form of the distribution function we assume should be unimportant , just so long as the same model is used both when including the effects of merger boosts and when ignoring them .
ps formalism assumes that galaxies and clusters grow by the gravitational instability of initially small amplitude gaussian density fluctuations generated by some process in the early universe .
the fluctuation spectrum is assumed to have larger amplitudes on smaller scales .
thus , galaxies and clusters form hierarchically , with lower mass objects ( galaxies and groups of galaxies ) forming before larger clusters .
these smaller objects then merge to form clusters . according to ps
, the mass function of clusters is given by @xmath27 \
, , \ ] ] where @xmath28 is the current mean density of the universe , @xmath29 is the current rms density fluctuation within a sphere of mean mass @xmath30 , and @xmath31 is the critical linear overdensity for a region to collapse at a redshift @xmath32 . in cdm models ,
the initial spectrum of fluctuations can be calculated for various cosmologies ( bardeen et al .
1986 ) . over the range of scales covered by clusters ,
it is generally sufficient to consider a power - law spectrum of density perturbations , which is consistent with these cdm models : @xmath33 where @xmath34 is the present day rms density fluctuation on a scale of 8 @xmath35 mpc , and @xmath36 is the mass contained in a sphere of radius 8 @xmath35 mpc .
the exponent @xmath37 is given by @xmath38 , where the power spectrum of fluctuations varies with wavenumber @xmath39 as @xmath40 ; following bahcall & fan ( 1998 ) , we assume @xmath41 , which leads to @xmath42 throughout .
the normalization of the power spectrum and overall present - day abundance of clusters is set by @xmath0 .
the observed present - day abundance of clusters leads to @xmath43 ( e.g. , bahcall & fan 1998 ) .
we determine the value of @xmath0 by requiring that our models reproduce the observed local number density of clusters with temperature @xmath44 kev , which we take to be @xmath45 mpc@xmath46 kev@xmath7 ( henry & arnaud 1991 ) , although some more recent observations suggest a slightly smaller value for the local number density of clusters at @xmath44 kev ( ikebe et al .
we assume a value of @xmath47 km / sec / mpc , since the results scale in a simple way with the hubble constant .
we find @xmath48 for the open , flat , and eds models , respectively .
the evolution of the density of clusters is encapsulated in the critical over - density @xmath31 in equation ( [ eq : psdensity ] ) .
the expressions used for @xmath49 for the three cosmologies under consideration are given in appendix [ ap : delta ] .
one can determine the merger histories of clusters either moving forward in time ( starting at high redshift and following the merger history of a set of subclusters ) or moving backward in time ( starting at present with a set of clusters , and de - merging them into subclusters ) .
we follow the latter approach .
this has several advantages .
first of all , the initial populations of clusters is set at the present time , where we have the best observational data on the population of clusters .
secondly , one can start with rich clusters at present , and we do nt need to follow the evolution of low mass systems .
if the merger histories are computed going forward in time , one needs to follow the histories of a large ensemble of small systems , most of which never merge into rich clusters .
since rich clusters are rare objects , this is quite inefficient . to generate our merger trees ,
we follow the eps method outlined by lacey & cole ( 1993 ) .
we begin by writing down the conditional probability that a `` parent '' cluster of mass @xmath50 at a time @xmath51 had a progenitor of mass in the range @xmath52 at some earlier time @xmath53 , with @xmath54 and @xmath55 : @xmath56 dm_1 \ , .\ ] ] here @xmath57 and @xmath58 , with similar definitions for @xmath59 and @xmath60 .
details of the eps theory and derivations of equation ( [ eq : eps_prob ] ) are given elsewhere ( see lacey & cole 1993 ; cohn , bagla , & white 2001 ) .
as noted by lacey & cole ( 1993 ) , the eps expressions are simplified if we replace the mass @xmath30 and time @xmath61 ( or redshift @xmath32 ) with the variables @xmath62 and @xmath63 .
note that @xmath64 decreases as the mass @xmath30 increases , and that @xmath65 decreases with increasing cosmic time @xmath61 .
let @xmath66 be the probability that a cluster had a progenitor with a mass corresponding to a change in @xmath64 of @xmath67 in the range @xmath68 at an earlier time corresponding to @xmath69 .
then , equation ( [ eq : eps_prob ] ) becomes @xmath70 d \delta s \ , .\ ] ] where @xmath69 and @xmath71 .
as the mass of the progenitors is decreased ( @xmath72 ) , the total number of progenitors diverges . on the other hand ,
the total mass associated with these progenitors converges ( equation [ eq : eps_prob ] ) .
thus , it is useful to consider only progenitors with masses exceeding some minimum value @xmath73
. mergers involving smaller masses are considered to be part of a continuous accretion process . in our merger trees , the masses of clusters
increase continuously with cosmic time as a result of this accretion process ( or decrease as we run the merger trees backwards in time ) .
however , we do not follow the individual histories of subclusters with masses @xmath74 .
we chose the value of @xmath75 .
we employ a monte carlo technique to construct merger trees .
each tree starts with a cluster with an initial mass @xmath76 .
we step each cluster back in time , using a small but finite time step corresponding to a positive increase @xmath77 .
we adaptively change the step size of @xmath77 as we run the trees backward in time by satisfying the criterion specified by lacey & cole ( 1993 ) that @xmath78 , where @xmath30 is the mass of the cluster at the current time step and @xmath79 is the smallest subcluster we wish to resolve individually .
specifically , at every step we choose a step size equal to half this maximum value .
we found that using a smaller timestep equal to 1/10 of this maximum value did not significantly alter our results .
the cumulative probability distribution of subcluster masses is given by @xmath80 where @xmath81 is the complementary error function .
the cumulative probability distribution is defined such that @xmath82 as @xmath83 . to draw a @xmath84 from this probability distribution , we select a uniformly - distributed random number @xmath85 between 0 and 1 , and determine the corresponding value of @xmath84 by solving numerically the equation @xmath86 .
the value of @xmath87 of the progenitor is then @xmath88 . the mass of one of the progenitors is given by @xmath89 , where @xmath90 is given by equation ( [ eq : sigma ] ) .
the mass of the other progenitor is @xmath91 .
let @xmath92 and @xmath93 . if @xmath94 , we consider the change in mass to @xmath50 to be due to accretion ( i.e. , to a very small merger or mergers ) .
the transitory effects of such a small merger or mergers are assumed to be small .
we reduce the mass of the cluster to @xmath95 , where @xmath96 becomes the new value of @xmath50 for the next timestep in the merger tree .
we do not follow the smaller mass @xmath97 as a separate subcluster . however , if @xmath98 , we say that a merger has occurred and we follow both of the merging subclusters of masses @xmath99 and @xmath100 separately
. we continue this process until either the branch we are following drops below a mass of @xmath73 or we reach the maximum redshift @xmath101 we wish to study .
we choose to run each tree back to a fixed mass limit , as opposed to a limit expressed as some fraction of the final mass @xmath76 , so that the mass completeness limit of the cluster sample at all redshifts can be readily determined .
an example of a merger tree generated in this way is shown in figure [ fig : merger_tree ] . as pointed out by somerville & kolatt ( 1999 , hereafter sk )
, there are some inherent problems with generating merger trees in this way .
first of all , this technique allows for binary mergers only , while in reality multiple nearly simultaneous mergers are possible .
sk showed that allowing for multiple mergers improves the agreement between monte carlo merger trees and ps theory predictions for the mass function .
however , such multiple mergers mainly affect the relatively low mass halos at rather high redshifts @xmath102
. since we are mainly interested in massive clusters and will only follow the merger trees out to @xmath103 , this is not such a significant problem here . another related problem with the binary merger tree method ,
also pointed out by sk , is that while we draw one merging halo mass @xmath100 from equation ( [ eq : int_prob ] ) , the mass of the other merging halo @xmath99 is simply taken to be the mass of the progenitor @xmath50 minus @xmath100 .
while this approach ensures mass conservation , it distorts the distribution function we are actually drawing halo masses from in a complicated way , such that the effective distribution is no longer described by equation ( [ eq : int_prob ] ) .
the solution is to continue to draw merging halo masses from equation ( [ eq : int_prob ] ) with the restriction that the sum of the halo masses does not exceed the mass of the progenitor ( see sk ) .
this requires that we allow multiple merger events .
the main reason we have not implemented this n - branch tree method developed by sk is that it is unclear how to determine the luminosity and temperature boosts associated with a multiple merger event given the simulation data available ( [ sec : hydro ] ) .
however , it is likely that the effects of multiple mergers with smaller subclusters will be less important than major binary merger events with @xmath104 .
in addition , we find that the predictions of our binary merger tree method agree reasonably well with ps theory over the range of redshifts and halo masses we are interested in ( see [ sec : tests ] ) .
the previous section describes how we determine the merger history of a cluster with a given mass @xmath76 at the present time . in order to describe the statistical properties of mergers
, we construct an ensemble of such merger trees for a set of present day masses , @xmath105 . in principle , the simplest way to select such initial masses would be to draw them from the ps distribution at the present time , @xmath106 ( equation [ eq : psdensity ] ) .
then , the properties of the cluster ensemble at any redshift could be derived simply by averaging over all of the clusters in the set of merger trees . however
, the ps mass function diverges at low masses ( although the mass contained in the clusters converges ) .
this means that most of the merger trees selected in this way would follow the histories of rather small clusters . in order to have enough statistics to sample the larger clusters adequately
, one would require a very large number of merger trees . in order to increase
the efficiency and still have adequate statistics on large clusters , we select the initial ( present day ) cluster masses according to a different distribution function which gives a greater weight to higher mass systems . we draw the values of @xmath105 from the distribution function @xmath107 , which is defined such that @xmath108 is the number of merger trees with initial masses in the range @xmath109 . in practice , we take the initial masses to have logarithmically spaced masses , which corresponds to @xmath110 where @xmath111 is the largest cluster mass considered , and @xmath112 is the total number of trees run .
we construct three separate ensembles of merger trees for the three cosmologies under consideration ( the open , flat , and eds models ) . in each case
we choose @xmath113 , @xmath114 , and a lower limit on the present day mass of clusters considered @xmath115 .
since the method of generating merger trees described above is based on the eps theory , we expect that the mass function produced by the merger trees will agree with the ps function ( equation [ eq : psdensity ] ) for the same cosmological parameters .
of course , the tree - generated mass functions will be underestimated for masses smaller than @xmath116 , the smallest initial mass for a merger tree . the properly - weighted mass function from our ensemble of merger trees
would be given by @xmath117 \
, \frac { d \ , n } { d \ , m^0 } } { \frac { d \ , n } { d \
, m^0 } } \right\ } \ , , \ ] ] where @xmath118 is the present age of the universe , and the progenitor probability distribution @xmath119 $ ] is given in equation ( [ eq : eps_prob ] ) .
the ps cluster mass function is defined per unit comoving volume , and the factor of @xmath120 weights each initial cluster and all of its progenitors by the comoving volume per cluster of mass @xmath76 in the universe .
note that the distribution of initial cluster masses @xmath121 cancels between the numerator and the denominator in the integral .
however , we have a discrete set of initial masses , @xmath105 for @xmath122 . in the discrete limit
, the integral @xmath123 becomes the sum @xmath124 , and the properly weighted mass function is @xmath125 } { \left .
\frac { d \ , n } { d \ ,
m^0 } \right|_{m^0_i } } \right\ } \ , .\ ] ] in the merger trees , the progenitor probability function @xmath126 $ ] is determined from a discrete set of progenitors for each tree .
let @xmath127 for @xmath128 be the masses of the @xmath129 progenitors ( with masses greater than @xmath73 ) at redshift @xmath32 of the present - day cluster with a mass of @xmath105 .
we order the progenitor masses such that @xmath130 we accumulate the mass function in @xmath131 mass bins whose boundary mass values are @xmath132 for @xmath133 , with @xmath134 thus , the @xmath135 mass bin has a width of @xmath136 .
then , the tree - generated , binned mass function at redshift @xmath32 is given by @xmath137 \ , .\ ] ] here , @xmath138 is the number of progenitors at redshift @xmath32 of the tree with a present - day mass of @xmath105 that lie in the @xmath135 mass bin , @xmath139 is the smallest value of @xmath140 such that @xmath141 , and @xmath142 is the largest value of @xmath140 such that @xmath143 .
we compared the merger tree prediction of the mf to the ps expression ( equation [ eq : psdensity ] ) for the ensemble of trees described in [ sec : mass_set ] ; the comparison for the eds model is shown in figure [ fig : ps_compare ] . if equation ( [ eq : psdensity ] ) is fit to the mass function generated by the merger trees allowing @xmath0 and @xmath1 to vary the values used to construct the trees are recovered to better than 2% for @xmath0 and better than 7% for @xmath1 out to @xmath4 . for the eds model , the agreement is much better : less than 2% in @xmath0 and roughly 2% in @xmath1 .
the merger trees give the masses of clusters as functions of time . in order to calculate the tfs and xlfs
, we need to convert cluster masses to temperatures and x - ray luminosities .
we will use the bolometric x - ray luminosity in order to avoid a specific choice of an energy band . in the models including merger l - t boosts ,
the luminosities come from hydrodynamical simulations ( [ sec : hydro ] ) ; however , for comparison to the analytic ps model and to the results of merger tree simulations without merger l - t boosts , we need a set of `` equilibrium '' or non - boosted relations between the cluster mass and its x - ray luminosity and temperature .
we could use empirical relations , but these are generally not available at all redshifts .
we prefer relations which have the correct dependence on redshift and on cosmological parameters , since we consider a range of redshifts and a variety of cosmological models . moreover , we prefer a theoretical dependence on the cluster mass rather than an empirical one , since the empirical relations may already be biased by the effects of merger boosts .
it has been shown that the empirical relationship between cluster mass , luminosity , and temperature is somewhat different from what one would expect from theoretical arguments and numerical simulations where only gravitational heating is considered , even at the relatively high temperatures and luminosities we consider in this paper ( e.g. , arnaud & evrard 1999 ; allen , schmidt , & fabian 2001 ; finoguenov , reiprich , & bhringer 2001 ) . at present , the detailed physics and redshift dependence of these non - gravitational effects are not well understood .
as a result we did not include such effects in our simulations .
while it is true that preheating and other non - gravitational effects may slightly dilute the effect of merger boosts since the fractional increase in x - ray luminosity and temperature due to mergers will be smaller if non - gravitational heating is considered , the effect will be most important for low mass clusters ( ponman et al .
1996 ) which are not of primary interest here
. since we are mainly interested in massive clusters , are not interested in these non - gravitational effects , and are primarily concerned with the relative effects of clusters mergers , it is preferable to use mass - temperature and mass - luminosity relations which are consistent with purely gravitational heating .
we adopt the following mass - temperature and mass - luminosity relations given by bryan & norman ( 1998 ) : @xmath144^{1/3 } \ , { \rm kev } \
, , \ ] ] @xmath145 where we have added our own normalization factors @xmath146 and @xmath147 . in these equations @xmath148 is the baryon density parameter and @xmath149 is defined as @xmath150 , where @xmath151 and @xmath152 is the current radius of curvature of the universe . a fit for @xmath153
is given by bryan & norman ( 1998 ) for the relevant cosmological models : @xmath154 where @xmath155 - 1 $ ] . since our luminosity and temperature boosts
have been modeled from hydrodynamical simulations we choose @xmath146 and @xmath147 so that our relations reproduce the temperature and x - ray luminosity of cluster b in table 1 of ricker & sarazin ( 2001 ) ( assuming cluster b is observed at @xmath8 ) . in applying these relations to theoretical models of clusters , the redshift in equations ( [ eq : m_t ] ) and ( [ eq : m_l ] ) are taken to be the formation redshift of the cluster , @xmath156 .
once a cluster is virialized , its properties ( including the temperature and luminosity ) are assumed to remain constant within these equilibrium relations .
using the method described in [ sec : make_trees ] , we are able to estimate the number of mergers occurring at a particular redshift .
we also know the individual masses of the two merging halos . however , this is all the information we can get from the merger trees themselves ; if we want to know about the individual cluster x - ray luminosities and temperatures , and the effect that a merger has on these quantities , we must turn to other methods . for non - merging , virialized halos we expect a simple relationship between mass and temperature or mass and luminosity . for such halos
we assume the relationships given by equations ( [ eq : m_l ] ) and ( [ eq : m_t ] ) throughout .
however , if a cluster has recently experienced a major merger , it will be in a non - equilibrium state , and we expect its temperature and luminosity to exceed that of a relaxed cluster of equal mass at the same redshift .
since we are unaware of any verified analytic prediction of the l - t boost experienced by two merging clusters of known mass , we turn to hydrodynamical simulations of two merging clusters .
the general method is to simulate a series of binary cluster mergers with varying mass ratios and impact parameters and observe the l - t boost experienced by the merged cluster in each case , and then to fit a function of the merging masses and the impact parameter to the observed l - t boosts .
this fitted function is then used to determine the l - t boost associated with the merger of two clusters of arbitrary mass at any impact parameter .
the details of how this fit is carried out and the results are given below in [ sec : hydro_fit ] the details of the hydrodynamical simulations used to develop a model for the l - t boost are described in detail elsewhere ( ricker & sarazin 2001 ) .
data were available from 8 different simulation runs corresponding to three different mass ratios @xmath157 at three different impact parameters @xmath158 , excluding the one case @xmath159 . here
, @xmath160 is the scale radius of the more massive cluster , which is the parameter in the assumed navarro , frenk , & white ( 1997 , hereafter nfw ) mass profile of the cluster , and is defined in [ sec : hydro_trees ] .
the smaller mass @xmath97 was fixed at @xmath161 for each simulation .
note that while the runs with @xmath162 are not mentioned specifically in ricker & sarazin ( 2001 ) , the method used to generate them is identical to that used for the runs described therein .
figure [ fig : lttt ] shows the behavior of the luminosity and temperature as a function of time for equal mass merger runs at different impact parameters ( see also ricker & sarazin [ 2001 ] , their figures 5 and 8) .
@xmath163 is the emission - weighted x - ray temperature of the merging clusters , while @xmath164 is the same value for the sum of the two clusters prior to the merger .
similarly , @xmath17 is the total x - ray luminosity of the merging clusters , while @xmath165 is the value prior to the merger .
note that we plot the bolometric x - ray luminosity while ricker & sarazin ( 2001 ) consider the x - ray luminosity in the @xmath166 kev band .
the times are the offset from the time of peak luminosity , @xmath167 , and are scaled by the sound - crossing time @xmath168 of the more massive cluster ( @xmath168 is defined in [ sec : hydro_trees ] ) .
several peaks are evident in both the temperature and luminosity which correspond to the dense cores of the merging clusters interacting as they pass close by ( or through ) one another . in applying these results to merger trees ,
one complication is that the merger trees treat the mergers as instantaneous , while the hydrodynamical simulations follow the mergers over time .
moreover , the merger trees are computed with a finite time resolutions ( determined indirectly from @xmath77 ; see [ sec : make_trees ] ) .
thus , we do nt know the precise time or redshift at which each merger occurred in the merger trees . on the other hand ,
we are interested in the statistical effects of mergers ; obviously , numerical hydrodynamical simulations are needed to follow individual mergers in detail .
thus , we assume that each merger occurred at a random time within the time resolution of the merger tree @xmath77 . in all cases ,
the time resolution of the merger trees is fine enough that there are no significant cosmological changes during a time step : for a merger which forms a 10@xmath169 cluster the scale factor changes by less than 4% over one timestep , and for one which forms a 10@xmath170 cluster it changes by less than 0.6% .
we identify the `` time '' of a merger in a merger tree with the time @xmath167 at which the x - ray luminosity boost due to the merger is maximum ( an offset time of zero in figure [ fig : lttt ] ) .
generally , this corresponds to the first time the subcluster cores pass by ( or through ) one another . for the purposes of determining the temperature or luminosity boost assuming the merger occurs at a random time in the merger tree time bin
, it is useful to consider the distribution of cumulative times as a function of the boost . in figure
[ fig : histos]a , we plot the total time @xmath61 that the temperature exceeds some boost @xmath171 $ ] ( scaled by the sound crossing time @xmath168 ) vs. the relative boost @xmath172 .
the corresponding plots for the luminosity boost are given in figure [ fig : histos]b .
of course , we are most interested in the sections of these plots which correspond to strong boosts ( the right side of the plots ) . in appendix [ ap : hydro_fit ] , we fit these cumulative time distributions for the numerical hydrodynamical simulations of mergers to simple analytical forms as a function of the masses of the clusters and the merger impact parameter .
the parameters of these fits are interpolated or extrapolated to approximate the merger boost for mergers with arbitrary masses and impact parameters .
we find that , for boosts greater than a factor of 1.5 , our fits reproduce the boosts given by the hydrodynamical simulations with an average accuracy of better than 2% for the temperature boosts @xmath173 and 7% for the luminosity boosts @xmath174 .
the fits are less accurate for very small boosts , but obviously these have a very small effect on the xlfs and tfs of clusters . to determine the luminosity and temperature of a cluster at some observation time @xmath175 , we consider all the mergers the cluster has undergone prior to @xmath175 and the effect each merger will have on the luminosity and temperature of the cluster at @xmath175 . recall that at each merger @xmath97 and @xmath176 are the masses of the less massive and more massive individual merging subclusters respectively ( see [ sec : make_trees ] ) .
the impact parameter @xmath177 of the merger is drawn from the distribution given by linear theory ( see [ sec : impact_param ] below ) . in the fits of the merger boosts from hydrodynamical simulations ,
the scaled impact parameter @xmath178 is used ( appendix [ ap : hydro_fit ] ) , where @xmath179 and @xmath180 are the core radii of the merging subclusters . following ricker & sarazin ( 2001 ) and appendix [ ap : hydro_fit ] ,
we take the core radius of a cluster to be 1/2 of the scaling radius @xmath160 .
we assume a halo concentration parameter @xmath181 ( ricker & sarazin 2001 ) , where the virial radius @xmath182 is given by ( bryan & norman 1998 ) : @xmath183 the merger boosts depend on the time elapsed since the merger , @xmath184 ( figure [ fig : lttt ] ) .
as noted above , we identify the epoch of the merger in the merger tree with the time @xmath167 , and this time is selected at random within the ( small ) timestep in which the merger takes place .
the times for the merger boosts are scaled by the sound crossing time @xmath168 of the more massive subcluster , @xmath185 where @xmath186 is the sound speed in the cluster , @xmath187 the mean mass per particle in the gas is @xmath188 in units of the mass of a hydrogen atom @xmath189 . when computing the sound speed we make the approximation that the temperature @xmath163 is given by equation ( [ eq : m_t ] ) , although note that ricker & sarazin ( 2001 ) use the central temperature when computing the sound speed .
equation ( [ eq : hyperbola ] ) is then used with @xmath97 , @xmath176 , @xmath190 , and @xmath168 to determine the boost to the x - ray luminosity and temperature in a cluster at @xmath175 due to a merger at @xmath167 .
we apply the maximum boost at @xmath175 from all previous mergers . applying these relations to every cluster in the merger trees at @xmath175
gives a collection of luminosities and temperatures from which the xlf and the tf can be constructed as in [ sec : no_boosts ] .
the boost in x - ray luminosity and temperature during a merger depends on the impact parameter or orbital angular momentum for the merger .
unfortunately , this is not determined by the masses of the individual subclusters , and a range of values are possible for mergers of subclusters with similar masses and similar merger epochs @xmath191 .
the merger tree simulations are based on the assumption of spherical collapse and do not give the impact parameter of the merger .
in hierarchical large scale structure theories , the angular momentum is generally assumed to arise from tidal torques from surrounding material . in linear theory calculations of the growth of perturbations ,
the total angular momentum @xmath192 of a dark matter halo can be characterized by the spin parameter @xmath193 , defined as ( peebles 1971 ) @xmath194 here @xmath195 is its total energy of the dark matter halo and @xmath30 is its mass . in linear theory ,
the median value of @xmath193 is expected to be approximately constant , independent of the mass of the halo , and numerical simulations of large scale structure are in reasonable agreement with this approximation ( e.g. , efstathiou & jones 1979 ; barnes & efstathiou 1987 ; sugerman , summers , & kamionkowski 2000 ) .
the median value of @xmath193 from linear theory and numerical simulations is @xmath196 .
after these calculations were done we became aware of a more recent study of the distribution function of a parameter related to @xmath193 by bullock et al .
they propose an empirical log - normal fit to the distribution of @xmath193 values .
their simulations and their empirical fit are consistent with the distribution we have assumed , as long as the median value is fixed at @xmath197 .
however , their simulations actually give a slightly smaller value of @xmath198 , which would slightly increase the effect of merger boosts .
numerical simulations show that individual dark matter halos of the same mass acquire a range of different values of @xmath193 .
we will assume that the individual components of the angular momenta of dark matter halos have a gaussian distribution with a dispersion @xmath199 , which implies that the total angular momentum @xmath192 or spin parameter @xmath193 has a maxwell - boltzmann - like distribution , @xmath200 where @xmath201 gives the probability that the spin parameter is in the range @xmath202 .
the median value is @xmath203 , which gives @xmath204 .
figure [ fig : lambda ] shows the distribution function predicted by equation ( [ eq : lambda_dist ] ) compared to the distribution measured from simulations by bullock et al .
we will assume that this distribution with @xmath205 applies to the total angular momentum of the merged clusters , and we draw values of @xmath193 at random from this distribution for each merger .
there is a simple kinematic argument which allows one to determine the impact parameter for a merger from the value of @xmath193 ( sarazin 2002 ) .
let @xmath206 be the separation of the subcluster centers at turnaround , and let @xmath207 be their present separation .
then , the impact parameter @xmath177 is given by @xmath208 here , @xmath209 is a function of the masses of the two subclusters which is given by @xmath210^{3/2 } \
, , \ ] ] one can show that @xmath211 for all masses .
if the two subclusters are treated as point masses and one assumes that they dominate the mass in their vicinity , then the turnaround separation is given by ( sarazin 2002 ) @xmath212^{1/3 } \left ( \frac{t_{\rm merge}}{\pi } \right)^{2/3 } \nonumber \\ & \approx & 4.5 \left ( \frac{m _ < + m _ > } { 10^{15 }
\ , m_\odot } \right)^{1/3 } \left ( \frac{t_{\rm merge}}{10^{10 } \ , { \rm yr } } \right)^{2/3 } \ , { \rm mpc } \ , .
\label{eqn : init sep}\end{aligned}\ ] ] here , @xmath191 is the age of the universe at the time of the merger . in the simulations of ricker & sarazin ( 2001 ) , the impact parameter is set at the start of the simulations , when the separation of the two subclusters is @xmath213 where @xmath214 and @xmath215 are the cutoff radii of the two clusters ( ricker & sarazin 2001 ) .
for the purposes of comparison to our merger tree results , both with and without merger boosts , it is useful to have a simple prediction of the xlf and tf based on the ps mass function .
the xlf and tf of virialized , non - merging clusters at an observed redshift @xmath32 can be approximated by combining the ps mass function ( equation [ eq : psdensity ] ) with the mass - luminosity ( equations [ eq : m_l ] ) or mass - temperature ( equations [ eq : m_t ] ) relation .
note that these relations depend on the redshift @xmath156 at which each cluster was formed .
once a cluster is virialized , its temperature and luminosity are assumed to remain constant . for clusters of a given mass observed at some redshift @xmath32 , we need to know the formation redshift @xmath156 . for the merger - tree models
, we can calculate @xmath156 from the merger histories .
however , the formation redshift is not uniquely determined by the mass and observed redshift of a cluster .
a common approximation has been to assume that the formation redshift is approximately the observed redshift , @xmath216 .
however , this turns out to be inadequate to fit our merger tree results without merger boosts .
we have taken the formation redshifts @xmath156 of different clusters of mass @xmath30 observed at redshift @xmath32 to be equal to the mean for each ( @xmath30,@xmath32 ) , as determined by lacey & cole ( 1993 ) using the parameterization @xmath217 the quantities @xmath64 and @xmath65 were defined in [ sec : trees ] ; @xmath64 and @xmath65 are the values at the time of the observation , while @xmath218 and @xmath219 are the values at the epoch of the cluster formation .
cluster formation is assumed to occur when a subcluster has a mass which is greater than or equal to one half the mass of the cluster at the time of observation , so that @xmath220 . so that we will be able to directly compare our analytic luminosity and temperature functions to the merger tree results , we adopt a mean value of @xmath221 from our own merger trees .
we find a mean value @xmath222 , which is in good agreement with the mean of the distribution given by lacey & cole ( 1993 , see their figure 9 ) .
combining equation ( [ eq : psdensity ] ) with equations ( [ eq : m_t ] ) and ( [ eq : m_l ] ) gives us analytic expressions for the xlf and tf : @xmath223 @xmath224 the tf is defined so that @xmath225 gives the number of clusters per unit comoving volume with temperatures in the range @xmath226 ( similarly for the xlf @xmath227 \ , d l$ ] ) .
the mass - temperature relation @xmath228 and mass - luminosity relation @xmath229 are given by equations ( [ eq : m_t ] ) and ( [ eq : m_l ] ) , respectively .
the formation redshift is determined by solving equation ( [ eqn : formation redshift ] ) for @xmath219 with @xmath230 , and then converting @xmath219 to @xmath156 .
the xlf and tf of clusters can be extracted from the merger trees using the method described in [ sec : tests ] . the temperature function in bins @xmath231 is given by equation ( [ eq : mf_ens3 ] ) if we replace @xmath132 with @xmath231 , and if @xmath232 is the number of cluster progenitors with temperatures in the @xmath135 temperature bin .
the luminosity function is given by the same expression , replacing @xmath132 with @xmath233 , the @xmath135 luminosity bin .
we select an observation time or redshift @xmath32 , identify the collection of progenitor masses which exist in the merger trees at that time , transform this collection of masses into a collection of temperatures or luminosities using equation ( [ eq : m_t ] ) or ( [ eq : m_l ] ) , and then apply the appropriate form of equation ( [ eq : mf_ens3 ] ) .
a comparison of the tf given by equation ( [ eq : dndt ] ) to that given by the merger trees is shown in figure [ fig : dndt_compare ] for an eds universe .
the merger tree curves lie slightly above the analytic predictions , even at @xmath8 .
this discrepancy seems to be related to the approximation used to determine the formation redshifts of clusters when determining the ps tf and xlf using equations ( [ eq : m_t ] ) and ( [ eq : m_l ] ) ; we assumed @xmath222 throughout , whereas in reality there is a distribution of formation redshifts for a fixed cluster temperature and luminosity .
in general , this distribution is somewhat skewed ( e.g. kitayama & suto 1996 ) so that convolving it with the ps distribution may produce a different tf from the one obtained if an average formation epoch is assumed .
if , as a test , we set @xmath234 in the merger trees and in equation ( [ eq : dndt ] ) , the agreement between the curves in figure [ fig : dndt_compare ] is improved .
it is interesting to compare the xlfs and tfs given by the merger trees when the effects of merger boosts are included to those given when merger boosts are ignored .
figures [ fig : dndt_tree ] and [ fig : dndl_tree ] plot the boosted and unboosted merger tree tfs and xlfs , respectively , for the three cosmologies under consideration . at low temperatures ( @xmath235 kev ) and luminosities ( @xmath236 erg / sec ) , there is good agreement between the unboosted and boosted curves .
however , at larger temperatures and luminosities , the boosted xlfs and tfs rise above the unboosted curves .
this is due to the fact that majors mergers with large boosts are relatively rare events .
thus , they do not affect the statistics of clusters unless the virialized clusters with the same temperature and luminosity are themselves extremely rare .
there is an exponential drop - off in the expected number of the most massive clusters ( equation [ eq : psdensity ] ) , which have the largest virialized temperatures and luminosities .
thus , relatively rare major mergers of smaller and more common clusters can compete with the numbers of these most massive clusters .
table [ tab : boosts ] gives the fractional increase in the _ cumulative _ luminosity and temperature functions at @xmath237 due to merger boosts for the three cosmologies we consider .
for example , in the flat model at @xmath4 , merger boosts cause the number of clusters with temperatures @xmath5 kev to increase by a factor of 9.49 , and the number of clusters with luminosities @xmath6 erg s@xmath7 is increased by a factor of 8.9 .
the same factors for the eds model at a redshift of @xmath9 are 37.7 and 24.1 for the cumulative temperature and luminosity functions respectively .
we note that the increase in the boosted xlfs and tfs over the unboosted values is largest in the eds model , weakest in the open model , and intermediate in the flat model .
this is to be expected since the evolution of cluster abundance is relatively rapid in the eds model , less rapid in the flat model , and least rapid in the open model .
this means that at low redshifts , the merger rate must be highest in the eds model and lowest in the open model .
in fact , the evolution of the cluster abundance is so rapid in the eds model that the higher temperature and luminosity clusters have essentially disappeared by a redshift of one . for this reason ,
we adapt lower temperature and luminosity limits for the eds model at @xmath4 in table [ tab : boosts ] .
it should be noted that at the highest temperatures and luminosities the merger tree predictions are uncertain for non - zero redshifts since there are fewer and fewer hot luminous clusters at larger and larger redshifts , so that the statistics in the high end bins are poor .
this effect is most noticeable in the eds model where cluster abundance evolution occurs most rapidly so that high mass , high redshift objects are especially rare .
for example , the large discrepancy at @xmath238 kev and @xmath4 between the boosted and unboosted tfs for the eds model seen in figure [ fig : dndt_tree ] is not a real effect but an artifact of the lack of high temperature clusters at this redshift .
note that merger boosts cause @xmath239 kev clusters at @xmath240 in an eds model to be almost as common as in the flat model .
thus , if all of the cosmological parameters other than @xmath1 were known from other measurements , and the abundance of very hot or very x - ray luminous clusters at high redshift were used to determine @xmath1 , the value would be substantially underestimated .
we now treat the binned xlfs and tfs generated from the merger trees as we would observational data , fitting the data with equations ( [ eq : dndt ] ) and ( [ eq : dndl ] ) and treating @xmath1 and @xmath0 as free parameters of the fit .
the models are fit using least squares fitting , where we choose to minimize the square of the difference of the logs of the ps and merger tree tfs and xlfs so that more weight is given to hot , luminous clusters .
if we only consider observations at one redshift , there is a degeneracy between the fitting parameters @xmath1 and @xmath0 , such that @xmath241 ( e.g. , bahcall & fan 1998 ) . one way to break this degeneracy is to simultaneously fit data at two or more redshifts . for each of the three cosmologies under consideration
, we have simultaneously fit the tfs and xlfs at @xmath8 and at either @xmath242 or @xmath243 .
this corresponds to the typical procedure of using observations of a low redshift cluster sample to determine the local tf or xlf ( and possibly infer the mass function ) , and then comparing to a higher redshift sample .
the merger boosted tfs and xlfs for the models are shown in figures [ fig : dndt_fit ] and [ fig : dldt_fit ] along with best - fit ps models .
the best - fit parameters found in fitting equations ( [ eq : dndt ] ) & ( [ eq : dndl ] ) to the boosted and unboosted temperature and luminosity functions of the merger trees are given in tables [ tab : temp_fit ] & [ tab : lum_fit ] .
for comparison , the actual values of the parameters used to build the merger trees are also given in these tables .
if we compare the values given for @xmath0 from the boosted tfs and xlfs to those given by the unboosted functions we see that , in general , merger boosts cause @xmath0 to be overestimated by about 20% ( except for the open model xlf , which only shows a 5% increase ) .
similarly , if we compare the values of @xmath0 from the unboosted tfs to the values actually used to construct the trees , we find that they are about 10% larger .
the same comparison for the xlf results also show that @xmath0 is larger than the original values used to build the merger trees , but there is a lot of scatter in the magnitude of the increase .
the unboosted @xmath0 values are systematically higher than the original values because the merger tree tfs and xlfs lie slightly above the curves given by equations ( [ eq : dndt ] ) & ( [ eq : dndl ] ) with the original values of @xmath0 and @xmath1 . as mentioned in
[ sec : no_boosts ] , this has to do with the fact that the formation times must be estimated when using equation ( [ eq : dndt ] ) , whereas for the merger trees we have a record of the formation time for each cluster .
a similar set of comparisons shows that , in the case of the tfs , merger boosts cause @xmath1 to be underestimated by about 20% , while the unboosted values for @xmath1 are smaller than the actual values used to build the merger trees by about 15% .
the xlf results for @xmath1 show large variations and no systematic trends .
thus , while the results from the tfs suggest that merger boosts can cause @xmath1 to be underestimated by about 20% , the results from the xlfs are less conclusive . it should be noted that the effect of merger boosts on @xmath0 and @xmath1 depends somewhat on the detailed method used to determine the tfs and xlfs and the criteria used to fit them .
thus , the results given here should be viewed as an illustration .
since the merger boost effect is greatest for the highest temperatures and luminosities , the effect on @xmath0 and @xmath1 will be greatest for samples of hot , luminous clusters , or for fitting techniques that weight these clusters more highly . while much of the work done to date on using the xlf and tf of clusters to constrain both @xmath1 and @xmath0 has used the evolution of these functions to break the degeneracy between them , it should be noted that some recent studies simultaneously determine these two parameters using only the shape of the local mass function of clusters ( e.g. , reiprich & bhringer 2001 ) .
this study relies on the predicted variation of the fluctuation spectrum with @xmath1 in cdm models .
since we assume a power - law fluctuation spectrum , we can not compare directly to the results of reiprich & bhringer ( 2001 ) .
however , we did perform a series of fits using only the local @xmath8 data from the merger trees to determine the relative effects of merger boosts on the derived cosmological parameters .
the results are consistent with the results obtained by simultaneously fitting local and higher redshift merger tree data as discussed above .
hydrodynamical simulations of binary cluster mergers have shown that mergers can temporarily boost the x - ray luminosity and temperature of the merged cluster beyond their virial equilibrium values .
these merger boosts can alter the observed tfs and xlfs of clusters , particularly at the high ends .
even a few `` extra '' clusters with high luminosities and temperatures can have a significant impact due to the relative rarity of massive clusters .
the mass function of clusters is commonly inferred from the observed tf or xlf , assuming virial equilibrium .
the inferred mass function is often used to constrain cosmological parameters .
thus , merger boosts can affect the inferred cosmological parameters .
we tested the amplitude of the merger boost effect by using eps theory and a monte carlo technique to numerically reconstruct the merger histories ( trees ) of a population of clusters . at each merger
, we quantified the strength of the merger boost by extrapolating or interpolating from a set of hydrodynamical simulations of cluster mergers .
we then built tfs and xlfs from the merger trees , averaging together our sets of merger histories so that the sample cluster population was representative of the observed present day population .
our results show that merger boosts do indeed affect the apparent number of high mass clusters .
for example , in the flat model at @xmath4 merger boosts cause the number of clusters with temperatures @xmath5 kev to increase by a factor of 9.5 , and the number of clusters with luminosities @xmath6 erg s@xmath7 is increased by a factor of 8.9 .
the effect is strongest for an eds model , since clusters evolve more rapidly in this model than in the open and flat models . in the eds model at a redshift of @xmath9 , the number of clusters with temperatures @xmath5 kev is increased by a factor of 38 , and the number with luminosities @xmath244 erg s@xmath7 by a factor of 24 . at first
this might appear contradictory with the fact that the x - ray luminosity receives a larger boost from mergers than the x - ray temperature ( see figure [ fig : lttt ] ) .
however , the range of x - ray luminosities is larger than the range of x - ray temperatures ; another way of saying this is that the x - ray luminosity - temperature relationship is much steeper than linear . at a result ,
the overall effect of luminosity boosts on the xlf is not as strong as the effect of temperature boosts on the tf . comparing the boosted and unboosted
differential tf for the eds and flat models at @xmath4 shows that merger boosts cause @xmath239 kev clusters to be almost as common in the eds model as in the flat model .
this means that if all cosmological parameters other than @xmath1 were known from other measurements , and the abundance of very hot or very x - ray luminous clusters at high redshift were used to determine @xmath1 , the value would be substantially underestimated .
we fit ps distributions to our simulated tfs and xlfs .
we did this by considering two samples : a low redshift , local sample ( @xmath245 ) , and a moderate or high redshift sample ( @xmath246 or @xmath240 ) .
the two samples are fit simultaneously to determine the best - fit values of @xmath0 and @xmath1 , and these values were compared to the actual values used to construct the merger trees .
merger boosts can cause @xmath0 to be overestimated by about 20% .
merger boosts may also cause @xmath1 to be underestimated by about 20% , although the results from the xlf fits do not show a clear trend .
the effect of merger boosts on the inferred value of @xmath1 is much smaller in these fits than might have been expected from the overall increase in the number of hot clusters at high redshifts produced by merger boosts .
much of the effect of merger boost is `` renormalized '' away by the joint fit of low and high redshift sample .
one way to think of this is that merger boosts increase the numbers of hot clusters , both at low redshift and high redshift .
determining @xmath0 from the low redshift sample or from a joint fit removes much of the boost effect , and the change in @xmath1 is smaller than one might have expected .
studies of the abundance of hot or luminous clusters at high redshift have been used to argue that we live in a low density universe ( e.g. , bahcall & fan 1998 ; borgani et al .
it would appear that merger boosts do not invalidate this conclusion , although the error bars should be increased significantly to include the systematic uncertainties associated with merger boosts .
on the other hand , merger boosts do affect the value of @xmath0 .
as noted before , the number of the hottest and most luminous clusters are affected quite strongly .
thus , any comparison between cluster tfs and xlfs and cosmological parameters derived from other objects ( from cosmic microwave background radiation fluctuations , or from the brightness of distant supernovae ) is likely to be inconsistent and may lead to errors in the deduced cosmological parameters . at the least ,
the systematic uncertainties are likely to be much larger than might be inferred from the statistics of clusters alone .
obviously , the effect of merger boosts could be avoided entirely if the mass function could be determined directly from gravitational lensing observations .
radio detections of the sunyaev - zeldovich ( sz ) effect can also provide a measure of the number of luminous and hot clusters , particularly at high redshift ( e.g. , holder et al .
although mergers should also boost the microwave decrement from clusters , we expect that this effect would be smaller than the effect on the x - ray emission - weighted temperature or the x - ray luminosity , because the sz effect depends on density rather than the square of the density .
we thank y. fujita , t. reiprich , and m. takizawa for useful discussions .
s. w. r. and c. l. s. were supported in part by @xmath247 award number go1 - 2123x , issued by the @xmath247 x - ray observatory center , which is operated by the smithsonian astrophysical observatory for and on behalf of nasa under contract nas8 - 39073 , and by nasa xmm grant nag5 - 10075 .
m. r. is supported by doe under grant b341495 to the asci flash center at the university of chicago .
allen , s. w. , schmidt , r. w. , & fabian , a. c. 2001 , mnras , 328 , l37 allen , s. w. , schmidt , r. w. , & fabian , a. c. 2002 , mnras , in press ( astro - ph/0111368v2 ) arnaud , m. , evrard , a. e. 1999 , mnras , 305 , 631 bahcall , n. a. , fan , x. , & cen , r. 1997 , apj , 485 , 53 bahcall , n. a. , & fan , x. 1998 , apj , 504 , 1 bardeen , j. m. , bond , j. r. , kaiser , n. , & szalay , a. s. 1986 , apj , 304 , 15 barnes , j. , & efstathiou g. 1987 , apj , 319 , 575 bond , j. r. , cole , s. , efstathiou , g. , & kaiser , n. 1991 , apj , 379 , 440 borgani , s. , et al.2001 , apj , 561 , 13 bryan , g. l. , & norman , m. l. 1998 , apj , 495 , 80 bullock , j. s. , dekel , a. , kolatt , t. s. , kravtsov , a. v. , klypin , a. a. , porciani , c. , & primack , j. r. 2001 , apj , 555 , 240 cavaliere , a. , & fusco - femiano , r. 1976 , a&a , 49 , 137 cohn , j. d. , bagla , j. s. , & white , m. 2001 , mnras , 325 , 1053 efstathiou , g. , & jones , b. j. t. 1979 , mnras , 186 , 133 eke , v. r. , cole , s. , freml , c. s. , & henry , j. p. 1998 , mnras , 289 , 1145 finoguenov , a. , reiprich , t. h. , & bhringer , h. 2001 , a&a , 368 , 479 hashimoto , y. , hasinger , g. , arnaud , m. , rosati , p. , & miyaji , t. 2002 , a&a , 381 , 841 henry , j. p. , & arnaud , k. a. 1991 , apj , 372 , 410 holder , g. p. , mohr , j. j. , carlstrom , j. e. , evrard , a. e. , & leitch , e. m. 2000 , apj , 544 , 629 ikebe , y. , reiprich , t. h. , bhringer , h. , tanaka , y. , & kitayama , t. 2002 , a&a , 383 , 773 jenkins , a. , et al . 2001 ,
mnras , 321 , 372 kitayama , t. , & suto , y. 1996 , apj , 469 , 480 lacey , c. , & cole , s. 1993 , mnras , 262 , 627 navarro , j. f. , frenk , c. s. , & white , s. d. m. 1997 , apj , 490 , 493 ( nfw ) peebles p. j. e. 1971 , a&a , 11 , 377 peebles p. j. e. 1980 , the large - scale structure of the universe ( princeton : princeton univ .
press ) ponman , t. j. , bourner , p. d. j. , ebeling , h. , & bhringer h. 1996 , mnras , 283 , 690 press , w. h. , & schechter , p. 1974 , apj , 187 , 425 reiprich , t. h. , & bhringer h. 2001 , apj , 567 , 716 ricker , p. m. , & sarazin , c. l. 2001 , apj , 561 , 621 ritchie , b. w. , & thomas , p. a. 2002 , mnras , 329 , 675 sarazin , c. l. 2002 , in merging processes in clusters of galaxies , ed .
l. feretti , i. m. gioia , & g. giovannini ( dordrecht : kluwer ) , in press ( astro - ph/0105418 ) sheth , r. k. , & tormen , g. 1999 , mnras , 308 , 119 somerville , r. s. , & kolatt , t. s. 1999 , mnras , 305 , 1 sugerman , b. , summers , f. j. , & kamionkowski , m. 2000 , mnras , 311 , 762
the critical overdensity as a function of cosmic time @xmath61 depends on the cosmological parameters @xmath249 and @xmath250 . in this paper
we use the following expressions for @xmath248 : @xmath251 & ( \omega_0 < 1 \ , , \ , \omega_\lambda = 0 ) \\
\frac{3}{2 } d ( t_0 ) \left [ \left ( \frac{t^{\prime}_{\omega}}{t } \right)^{\frac{2}{3 } } - 1 \right ] & ( \omega_0 > 1 \ , , \ , \omega_\lambda = 0 ) \\
\frac{3(12\pi)^\frac{2}{3}}{20 } \left ( \frac{t_0 } { t } \right)^{\frac{2}{3 } } & ( \omega_0 = 1 \ , , \ , \omega_\lambda = 0 ) \\
\frac{d(t_0 ) } { d(t ) } \left(\frac{3(12\pi)^\frac{2}{3}}{20}\right ) \left(1 + 0.0123 \ \log_{10}\omega_{f } \right ) & ( \omega_0 + \omega_{\lambda } = 1 ) \end{array } \right.\ ] ] for the open model ( @xmath252 , @xmath19 ) , @xmath253 represents the epoch at which a nearly constant expansion takes over and no new clustering can occur , and the growth factor can be expressed as @xmath254 where @xmath255 is the standard parameter in the cosmic expansion equations ( peebles 1980 , eqn .
13.10 ) @xmath256 a similar set of relations holds for the closed model ( @xmath257 , @xmath19 ) : @xmath258 @xmath259 in this model @xmath260 represents the lifetime of the universe , that is to say the cosmic time at which the universe formally recollapses to a singularity .
although we do not consider a closed cosmological model when building merger trees it is necessary that our fitting routines be able to consider cosmologies with @xmath257 when fitting cosmological parameters to the merger tree data ( see [ sec : tree_fit ] ) .
the solution for @xmath261 in the einstein - de sitter model can be obtained from the open model solution by the limit @xmath262 ( lacey & cole 1993 ) . to evaluate @xmath261 in the flat model @xmath263 ) ,
we have used an approximation given by kitayama & suto ( 1996 ) . here
@xmath264 is the value of the mass density ratio @xmath265 at the redshift of formation , @xmath266 in this model the growth factor can be written as @xmath267 ( peebles 1980 , eqn .
13.6 ) where @xmath268 and @xmath269 .
as described in [ sec : hydro_fit ] , we wish to fits the portions of the histograms shown in figure [ fig : histos ] which correspond to strong boosts .
we find that these portions of the cumulative time distribution are well described by hyperbolas of the form @xmath270 ^ 2 - 1 \right\ } \left ( \epsilon_t^2 - 1 \right ) } - \ln \left ( \frac{t}{t_{\rm sc } } \right)_t \ , .\ ] ] the parameters in this function are the largest temperature boost during the merger @xmath271_{\rm max}$ ] , the duration of the period of maximum temperature boost @xmath272 , and @xmath273 which describes the eccentricity of the hyperbola . as similar expression fits the time distribution of luminosity boosts , with parameters @xmath274_{\rm max}$ ] , @xmath275 , and @xmath276 .
after fitting the function given by equation ( [ eq : hyperbola ] ) to each of the luminosity and temperature boost histograms shown in figure [ fig : histos ] , we determined the dependence of the boost histogram parameters on the kinematics of the merger .
we characterize the merger kinematics by two dimensionless parameters .
the first is the fractional mass increase in the merger , @xmath277 . following the discussion in above ,
let @xmath97 be the smaller of the two masses of the merging subclusters and @xmath176 be the larger of the two masses . then , @xmath278 .
our second parameter @xmath279 is the impact parameter @xmath177 for the merger , divided by the sum of the core radii of the two merging subclusters , @xmath280 . here ,
@xmath179 and @xmath180 are the core radii of the merging subclusters , which are taken to be 1/2 of the scaling radius @xmath281 for each subcluster .
we find that the variation of the boost histogram parameters with the merger kinematics can be fit with the following functions ( for the temperature boost ) : @xmath282_{\rm max } ( f_m , b ' ) = \frac{a f_m^b } { c + b^{\prime 2 } } + 1 \ , , \ ] ] @xmath283 @xmath284 where @xmath97 and @xmath176 are in units of @xmath285 .
similar expressions are used for the x - ray luminosity boost .
the best - fit parameters in these fits are listed in table [ tab : fits ] .
note that equation ( [ eq : tmaxfit ] ) implies that the temperature and luminosity boosts are largest for equal mass mergers ( @xmath286 ) , and go to zero as the mass of the smaller subcluster goes to zero ( @xmath287 ) .
similarly , equation ( [ eq : efit ] ) implies that the histogram becomes a vertical line as @xmath287 ; that is , all of the time is spent with no boost . as the mass fraction increases
, @xmath288 decreases , and the boost histogram becomes more extended toward large boosts .
the dependence of the maximum boost ( equation [ eq : tmaxfit ] ) and eccentricity ( equation [ eq : efit ] ) on impact parameter is based on the assumption that the gas distributions in clusters are well - fit by the standard beta model ( cavaliere & fusco - femiano 1976 ) , and that the strength of the boost depends on the degree of interactions of the densest portions of the two subclusters .
thus , equations ( [ eq : tmaxfit ] ) & ( [ eq : efit ] ) give the largest boost for a nearly head - on collision , and the boost goes to zero as the impact parameter increases .
the quadratic dependence on @xmath190 reflects the form of the density dependence in the beta model .
the form we adapted for the variation in the time duration of the merger boost ( equation [ timefit ] ) follows from simple scaling for the time scales for mergers . assuming that the two subclusters have fallen from a large distance to a separation @xmath207 , their relative velocity is @xmath289 where we approximate the subclusters as point masses .
the time scale for the merger boost is given by @xmath290 .
for a given formation epoch , the core radii and other length scales associated with clusters scale as @xmath291 .
thus , we assume that the subcluster separation during the time of maximum merger boost scales as @xmath292 .
one might expect that boost time to scale roughly as @xmath293 the virial relation for the temperature of the gas in clusters leads to @xmath294 , which implies that the sound - crossing time @xmath168 is nearly independent of mass .
these arguments determine the mass dependence of equation ( [ timefit ] ) .
if these arguments held exactly , we would expect @xmath295 .
although the best - fit values are close to this ( table [ tab : fits ] ) , the fits are improved if @xmath296 is allowed to vary .
we adapt the same dependence on @xmath190 as in the other fits .
ultimately these fits will be used to determine the luminosity and temperature boosts associated with a pair of merging halos given the time during the merger at which the merging cluster is observed . for a given value of @xmath297 ,
equations ( [ eq : hyperbola ] ) , ( [ eq : tmaxfit ] ) , ( [ eq : efit ] ) , & ( [ timefit ] ) and the parameters in table [ tab : fits ] reproduce the temperature boosts @xmath173 with an average accuracy of better than 2% for all boosts greater than 1.5 .
the accuracy of the fit typically increases with the strength of the boost ; thus the accuracy for the strongest temperature boosts are generally better than 1% .
the fits for the luminosity boosts @xmath174 have an average accuracy better than 7% for boosts greater than 1.5 .
lccccc open&0.0&1.96&2.35&1.97&3.43 + open&0.5&2.24&4.64&1.71&4.97 + open&1.0&2.32&3.30&1.71&3.91 + flat&0.0&2.10&2.24&2.01&4.37 + flat&0.5&2.56&3.54&2.04&8.11 + flat&1.0&3.37&9.49&2.05&8.90 + eds & 0.0&2.73&5.62&4.08&20.7 + eds & 0.5&11.3&37.7&4.51&24.1 + eds & 1.0&3.25&6.72&2.65&21.7 + lcrrrrrrrrr open&0,0.5&0.8274&0.3&0.913&0.26&1.190&0.13 + open&0,1.0&0.8274&0.3&0.912&0.26&1.046&0.21 + flat&0,0.5&0.8339&0.3&0.919&0.25&1.096&0.20 +
flat&0,1.0&0.8339&0.3&0.905&0.26&1.095&0.20 + eds & 0,0.5&0.5138&1.0&0.555&0.86&0.667&0.71 + eds & 0,1.0&0.5138&1.0&0.575&0.77&0.693&0.63 + lcrrrrrrrrr open&0,0.5&0.8274&0.3&1.057&0.49&1.113&0.38 + open&0,1.0&0.8274&0.3&1.085&0.59&1.145&0.46 + flat&0,0.5&0.8339&0.3&0.857&0.19&1.037&0.37 + flat&0,1.0&0.8339&0.3&0.849&0.17&1.000&0.23 + eds & 0,0.5&0.5138&1.0&0.535&0.82&0.655&1.20 + eds & 0,1.0&0.5138&1.0&0.546&0.99&0.638&0.93 + | in the standard cold dark matter model of structure formation , massive clusters form via the merger of smaller clusters .
n - body / hydrodynamical simulations of merging galaxy clusters have shown that mergers can temporarily boost the x - ray luminosity and temperature of the merged cluster above the equilibrium values for the merged system .
the cumulative effect of these `` merger boosts '' will affect the observed x - ray luminosity functions ( xlfs ) and temperature functions ( tfs ) of clusters .
one expects this effect to be most important for the most luminous and hottest clusters .
xlfs and tfs of clusters provide some of the strongest constraints on cosmological and large - scale - structure parameters , such as the mean fluctuation parameter , @xmath0 , and the matter density divided by the critical density , @xmath1 .
merger boosts may bias the values of @xmath0 and @xmath1 inferred from cluster xlfs and tfs if virial equilibrium is assumed .
we use a semi - analytic technique to estimate the effect of merger boosts on the x - ray luminosity and temperature functions .
the boosts from individual mergers are derived from n - body / hydrodynamical simulations of mergers .
the statistics of the merger histories of clusters are determined from extended press - schechter ( ps ) merger trees .
we find that merger boosts can increase the apparent number of hot , luminous clusters .
for example , in a universe with @xmath2 and @xmath3 at a redshift of @xmath4 , the number of clusters with temperatures @xmath5 kev is increased by a factor of 9.5 , and the number of clusters with luminosities @xmath6 erg s@xmath7 is increased by a factor of 8.9 .
we have used our merger - boosted tfs and xlfs to derive the cosmological structure parameters @xmath0 and @xmath1 by fitting press - schechter equilibrium relations to local ( @xmath8 ) and distant ( either @xmath9 or @xmath4 ) cluster samples .
merger boosts cause @xmath0 to be overestimated by about 20% .
the matter density parameter @xmath1 may be underestimated by about 20% , although this result is less clear .
if the parameters of the fluctuation spectrum are derived from the observed tf or xlf ( e.g. , from a low redshift sample ) , then this removes most of the boost effect on @xmath1 .
however , larger merger boost effects may appear when cluster xlfs and tfs are compared to cosmological structure parameters derived by other techniques ( e.g. , cosmic microwave background fluctuations or the brightness of distant supernovae ) . |
we hereafter propose a method that enables us to build a drp scheme while minimizing the error due to the finite difference approximation , by means of an equivalent matrix equation .
+ + consider the transport equation : @xmath0 , \,\,\,t \,\in \,[0,t]\ ] ] with the initial condition @xmath1
. a finite difference scheme for this equation can be written under the form : @xmath2 where : @xmath3
@xmath4 , @xmath5 , @xmath6 , @xmath7 , @xmath8 , @xmath9 denoting respectively the mesh size and time step ( @xmath10 , @xmath11 ) .
+ the courant - friedrichs - lewy number ( @xmath12 ) is defined as @xmath13 .
+ + a numerical scheme is specified by selecting appropriate values of the coefficients @xmath14 , @xmath15 , @xmath16 , @xmath17 , @xmath18 , @xmath19 , @xmath20 , @xmath21 and @xmath22 in equation ( [ scheme ] ) , which , for sake of usefulness , will be written as : @xmath23 where the `` @xmath24 '' denotes a dependance upon the mesh size @xmath8 , while the `` @xmath25 '' denotes a dependance upon the time step @xmath9 .
+ the number of time steps will be denoted @xmath26 , the number of space steps , @xmath27 . in general , @xmath28 .
+ in the following : the only dependance of the coefficients upon the time step @xmath9 existing only in the crank - nicolson scheme , we will restrain our study to the specific case : @xmath29 the paper is organized as follows .
the building of the drp scheme is exposed in section [ drp ]
. the equivalent matrix equation , which enables us to minimize the error due to the finite difference approximation , is presented in section [ sylv ] .
a numerical example is given in section [ ex ] .
the first derivative @xmath30 is approximated at the @xmath31 node of the spatial mesh by : @xmath32 following the method exposed by c. tam and j. webb in @xcite , the coefficients @xmath33 , @xmath34 , and @xmath35 are determined requiring the fourier transform of the finite difference scheme ( [ approx ] ) to be a close approximation of the partial derivative @xmath36 .
+ ( [ approx ] ) is a special case of : @xmath37 where @xmath38 is a continuous variable , and can be recovered setting @xmath39 .
+ denote by @xmath40 the phase . applying the fourier transform , referred to by @xmath41 , to both sides of ( [ approx_cont ] ) , yields : @xmath42 @xmath43 denoting the complex square root of @xmath44 .
+ comparing the two sides of ( [ wavenb ] ) enables us to identify the wavenumber @xmath45 of the finite difference scheme ( [ approx ] ) and the quantity @xmath46 , i. e. : the wavenumber of the finite difference scheme ( [ approx ] ) is thus : @xmath47 to ensure that the fourier transform of the finite difference scheme is a good approximation of the partial derivative @xmath48 over the range of waves with wavelength longer than @xmath49 , the a priori unknowns coefficients @xmath33 , @xmath34 , and @xmath35 must be choosen so as to minimize the integrated error : @xmath50 the conditions that @xmath51 is a minimum are : @xmath52 and provide the following system of linear algebraic equations : @xmath53 which enables us to determine the required values of @xmath33 , @xmath34 , and @xmath35 : @xmath54
the problem ( [ scheme ] ) can be written under the following matricial form : @xmath55 where @xmath56 and @xmath57 are square matrices respectively @xmath58 by @xmath58 , @xmath26 by @xmath26 , given by : @xmath59 the matrix @xmath60 being given by : @xmath61 and where @xmath62 is a linear matricial operator which can be written as : @xmath63 where @xmath64 , @xmath65 , @xmath66 and @xmath67 are given by : @xmath68 @xmath69 the second member matrix @xmath60 bears the initial conditions , given for the specific value @xmath70 , which correspond to the initialization process when computing loops , and the boundary conditions , given for the specific values @xmath71 , @xmath72 .
denote by @xmath73 the exact solution of ( [ transp ] ) .
+ the corresponding matrix @xmath74 will be : @xmath75_{\ , 1\leq i\leq { n_x-1},\ , 1\leq n\leq { n_t}\ , } } \ ] ] where : @xmath76 with @xmath77 , @xmath78 .
we will call _ error matrix _ the matrix defined by : @xmath79 consider the matrix @xmath80 defined by : @xmath81 the _ error matrix _ @xmath82 satisfies : @xmath83 minimizing the error due to the approximation induced by the numerical scheme is equivalent to minimizing the norm of the matrices @xmath82 satisfying ( [ eqmtr ] ) . _ note : _ since the linear matricial operator @xmath84 appears only in the crank - nicolson scheme , we will restrain our study to the case @xmath85 .
the generalization to the case @xmath86 can be easily deduced .
the problem is then the determination of the minimum norm solution of : @xmath87 which is a specific form of the sylvester equation : @xmath88 where @xmath89 and @xmath90 are respectively @xmath91 by @xmath91 and @xmath92 by @xmath92 matrices , @xmath93 and @xmath94 , @xmath91 by @xmath92 matrices .
calculation yields : @xmath95 the singular values of @xmath56 are the singular values of the block matrix @xmath96 , i. e. @xmath97 of order @xmath98 , and @xmath99 of order @xmath98 . + the singular values of @xmath57 are @xmath100 , of order @xmath101 , and @xmath102 , of order @xmath101 .
consider the singular value decomposition of the matrices @xmath56 and @xmath57 : @xmath103 where @xmath104 , @xmath105 , @xmath106 , @xmath107 , are orthogonal matrices .
@xmath108 , @xmath109 are diagonal matrices , the diagonal terms of which are respectively the nonzero eigenvalues of the symmetric matrices @xmath110 , @xmath111 .
+ multiplying respectively [ sylverr ] on the left side by @xmath112 , on the right side by @xmath107 , yields : @xmath113 which can also be taken as : @xmath114 set : @xmath115 @xmath116 we have thus : @xmath117 it yields : @xmath118 one easily deduces : @xmath119 the problem is then the determination of the @xmath120 and @xmath121 satisfying : @xmath122 denote respectively by @xmath123 , @xmath124 the components of the matrices @xmath125 , @xmath126 .
+ the problem [ pb ] uncouples into the independent problems : + minimize @xmath127 under the constraint @xmath128 this latter problem has the solution : @xmath129 the minimum norm solution of [ sylverr ] will then be obtained when the norm of the matrix @xmath130 is minimum .
+ in the following , the euclidean norm will be considered .
due to ( [ ftilde ] ) : @xmath131 @xmath104 and @xmath107 being orthogonal matrices , respectively @xmath58 by @xmath58 , @xmath26 by @xmath26 , we have : @xmath132 also : @xmath133 the norm of @xmath60 is obtained thanks to relation ( [ m0 ] ) .
this results in : @xmath134 @xmath135 can be minimized through the minimization of the second factor of the right - side member of ( [ min ] ) , which is function of the scheme parameters .
+ @xmath136 is a constant .
the quantities @xmath137 , @xmath138 and @xmath139 being strictly positive , minimizing the second factor of the right - side member of ( [ min ] ) can be obtained through the minimization of the following functions : @xmath140 i.e. : @xmath141 setting : @xmath142 one obtains the drp scheme with the minimal error through the minimization of : @xmath143
we denote by @xmath144 the non - dimensional time parameter .
figure [ erreur1 ] displays the @xmath145 norm of the error for an optimized scheme ( in black ) , where @xmath33 , @xmath34 , @xmath35 are given by ( [ opt_values1 ] ) , and a non - optimized one : numerical results perfectly fit the theoretical ones .
+ figure [ erreur2 ] displays the @xmath145 norm of the error for the above optimized scheme ( in black ) , a seven - point stencil drp scheme ( in gray ) , and the fcts scheme ( dashed plot ) . as time increases , the optimized scheme yields , as expected , better results than the fcts one .
also , for @xmath146 , results appear to be better than those of the classical drp scheme . for large values of the time parameter , both latter schemes yield the same results .
+ figure [ erreurl2 ] displays the @xmath147 norm of the error for the above optimized scheme ( in black ) , the seventh - order drp scheme ( in gray ) , and the fcts scheme ( dashed plot ) . as expected , results coincide .
the above results open new ways for the building of drp schemes .
it seems that the research on this problem has not been performed before as far as our knowledge goes . in the near future
, we are going to extend the techniques described herein to nonlinear schemes , in conjunction with other innovative methods as the lie group theory . | finite difference schemes are here solved by means of a linear matrix equation .
the theoretical study of the related algebraic system is exposed , and enables us to minimize the error due to a finite difference approximation , while building a new drp scheme in the same time . * keywords * + drp schemes , sylvester equation |
the demands of fitting models to large data - sets have exploded over the last decade .
increasingly complex data sets are available , which has placed demands on statisticians to develop realistic models to represent these data .
inevitably , for many classes of models , this places a further emphasis on being able to fit such models accurately and in a reasonable time - frame . in this article
, we consider fast bayesian statistical inference for dirichlet process mixture ( dpm ) models @xcite . this particular class of models have proven to be popular in the literature as a tool for both clustering and density estimation and there are a wide variety of elegant mcmc and sequential monte carlo algorithms ; see e.g. @xcite .
such algorithms provide exact inference from dpm models , but can be very computationally demanding when trying to analyze extremely large data - sets and even more , exact statistical inference from mixtures is notoriously difficult ; see @xcite . as mentioned above , this issue often leads to researchers resorting to approximate inference to browse or interrogate the data , so as to refine model specifications for an exact analysis ; one particular important and interesting method in this direction is the sugs algorithm .
the sugs algorithm is a procedure for fast approximate fitting of dpm models .
it relies on an approximation of the exact posterior distribution on the allocation of data to components and the component specific parameters , by sequentially adding data - points to the model .
these data are allocated to a given mixture component and this allocation is frozen and taken as the truth when updating the posterior for new data points .
the method can be sensitive to the ordering of the data , but @xcite provide procedures to select this ordering in a systematic way . in this paper
, we reinterpret the sugs algorithm within a variational bayes framework .
this allows one to derive different approximations of the posterior distribution .
in particular , our interpretation does not mean that one needs to allocate data to a cluster and , instead , provides a probability distribution on these allocations ; this is done at a minor increase in computational cost .
the advantage of this generalization , which we call vsugs , is apparent when one fits the approximation to data whose components are close in some sense . in this scenario , we have consistently found ( and as illustrated in section [ sec : simos ] ) that vsugs outperforms sugs in a variety senses ; this is important as , when the mixture components are highly separated , such initial browsing or interrogation of the data is less important .
moreover , our variational approximation provides a lower - bound on the log - marginal likelihood ; we empirically find that this can be used as a technique for model selection ( e.g. selecting the order in which the data arrive ) which did not seemingly work well in @xcite .
this article is structured as follows .
we begin with a motivating example in section [ sec : motivating ] . in section [ sec : dpm ]
we give a basic summary of dpm models . in section [ sec : sugs ] we discuss sugs and in section [ sec : vsugs ] our generalization vsugs . in section [ sec : simos ] we give numerical examples ; both a simulation study and real data analyses associated to flow cytometry data and snp data via a three - class dirichlet process mixture model . in section [ sec : summ ] we conclude the article , discussing avenues for future work .
a motivating application for us is the problem of genotyping single nucleotide polymorphisms ( snps ) from snp genotyping microarray data @xcite .
figure [ fig : genotyping_example ] illustrates an example dataset .
the statistical problem is to characterise the three genotype classes @xmath2 , @xmath3 and @xmath4 and classify each data point into one of these three classes .
this is a straightforward three - way classification problem that can be approached using hierarchical mixture modelling where , for example , the class - conditional densities are modelled using multivariate normal or student @xmath5-distributions .
although , these models work well in practice , it is clear that the class - conditional densities are not normal ( or student ) .
we can obtain increased accuracy through the use of semi - parametric models for the class - conditional densities using dirichlet process mixtures .
however , the size of the data sets presents a massive challenge for this type of modelling approach . for a single experiment ( individual )
, modern genotyping microarray produces 300,000 - 5,000,000 two - dimensional measurements .
each study may consist of hundreds to thousands of individuals .
the data sizes here prohibit the use of monte carlo inference and motivate approximate approaches that are able to scale to the size of problems encountered .
consider a dirichlet process mixture model of the form , for @xmath6 : @xmath7 where @xmath8 , @xmath9 are observation specific parameters , @xmath10 is a conditional probability which admits a density @xmath11 w.r.t .
a single dominating @xmath12finite measure for each @xmath13 ( which is often lebesgue ) , @xmath14 is an unknown mixing distribution , and @xmath15 indicates that the prior for @xmath14 is a dirichlet process @xcite with precision parameter @xmath16 and base measure @xmath17 . @xmath17 is known and we also consider @xmath18 to be fixed .
we remark that , in connection to subsequent methodology to be presented , @xcite consider a way of handling unknown @xmath18 which can also be used in all the extensions we consider but for simplicity we do not consider this below .
a well known property of the dirichlet process is that a distribution drawn from it will put all its mass on a countable set of points .
following the notation of @xcite we will write @xmath19 for the set of distinct values in the sequence @xmath20 where points in @xmath21 are labelled according to their order of appearance in @xmath22 .
next , let @xmath23 if @xmath24 and write @xmath25 . using the plya urn characterization of the dirichlet process @xcite we can rewrite ( [ dpmmodel ] ) in the form @xmath26 where @xmath27 , @xmath28 is the probability density associated to @xmath17 and @xmath29 is as follows , ( writing @xmath30 , with the convention @xmath31 is the null vector ) , for @xmath32 @xmath33 where @xmath34 , @xmath35 , and @xmath36 is the maximum value in @xmath37 ( i.e. the number of components seen " in the data up to to time @xmath38 ) .
this representation of the model where the unknown measure @xmath14 is integrated out is important for many monte carlo sampling schemes for fitting dp mixtures @xcite .
later we will work with a truncated dirichlet process mixture model , which is often convenient for computations @xcite .
such truncations are based on the stick breaking representation of the dirichlet process @xcite .
suppose we limit the number of distinct values appearing in the sequence @xmath39 to an upper truncation limit @xmath40 . then generalizing the plya urn representation we can consider the truncated dirichlet process with @xmath41 where now @xmath42 and @xmath29 is defined recursively by @xmath34 and similarly to ( [ polyaurn1 ] ) ( see , for example , ( * ? ? ? * section 2.2.2 ) ) for @xmath43 @xmath44 we note that truncations have also been used in the context of variational approximations for dirichlet process mixtures @xcite although these authors consider the truncation point as a variational parameter without truncating the original model .
the algorithm of @xcite , although related to ours , is not a sequential algorithm however - here we are interested in very fast sequential algorithms related to the sugs method of @xcite .
@xcite compare their approach with a variety of other fast computational methodologies for dirichlet process mixtures , and show that their algorithm is competitive with other fast approximation methodologies . in this work we focus only on comparing our new approach with the original sugs algorithm , and refer the reader to @xcite for information about the relative performance of alternative approximations to sugs
sugs is a recursive algorithm that takes at time @xmath38 an estimate @xmath45 of @xmath37 and an approximation of @xmath46 and produces an estimate @xmath47 of @xmath48 and an approximation of @xmath49 . to start the recursion we use @xmath50 and @xmath51 here and in what follows
a term @xmath52 in the conditioning means @xmath53 and similarly for @xmath47 .
suppose at time @xmath38 we have an estimate @xmath45 of @xmath37 .
consider the posterior distribution @xmath54 and approximate this by @xmath55 . that is to say , we initially consider our approximation of @xmath54 , @xmath56 , as @xmath57 using this approximation and integrating out @xmath21 , @xmath58 the sugs algorithm sets : @xmath59 \label{eq : hat_delta}\ ] ] and then one replaces @xmath60 by @xmath61 to form the approximation : @xmath62 in this approximation the components @xmath63 are independent for different @xmath64 , the posterior for @xmath63 for @xmath65 is unchanged from time @xmath38 , and the posterior for @xmath66 is updated by assuming that @xmath53 so that @xmath67 represents an observation from this mixture component .
if the mixture components are from the exponential family and conjugate priors are used , the densities such as @xmath68 can be calculated in closed form and sufficient statistics are updated recursively , leading to a very efficient update .
@xcite consider a number of further innovations in their algorithm .
first , since the fitting algorithm is sequential and there is a dependence of the fit on the ordering of the data , they suggest running their algorithm for different random orderings and then choosing the best according to a pseudo - likelihood criterion . secondly , for model comparison they suggest the approximation @xmath69 and show that this crude approximation can be useful for tasks such as comparison of parametric and nonparametric models .
thirdly , they suggest a way of dealing with unknown @xmath18 in the dirichlet process prior .
here we suggest a simple improvement of the sugs algorithm which we call vsugs ( variational sugs ) . we begin with a brief introduction to variational bayes ( vb ) methods .
suppose we have a parameter @xmath70 and data @xmath71 , @xmath72 is the prior density , @xmath73 the likelihood and @xmath74 denotes the posterior density w.r.t .
lebesgue measure ( which we use for presentational purposes only ) . in vb
@xcite we split @xmath75 into blocks @xmath76 , @xmath77 , with @xmath78 and seek to find a good approximation to @xmath79 of the form @xmath80 where each @xmath81 is a probability density w.r.t . the appropriate dimensional lebesgue measure .
given known probability densities for @xmath81 , @xmath82 , the optimal choice for @xmath83 for minimizing the kullback - leibler divergence @xmath84 is @xmath85 \ } \label{graddescent}\end{aligned}\ ] ] where @xmath86 $ ] denotes expectation w.r.t .
hence there is a gradient descent algorithm for minimizing based on choosing initial values for the factors in @xmath88 and then iteratively updating each term according to . minimizing ( [ kld ] ) is equivalent to maximizing @xmath89 and is a lower bound on the log marginal likelihood @xmath90 where @xmath91 .
@xmath92 is a key quantity in bayesian model selection and the lower bound is tight , @xmath93 , when @xmath94 . generally @xmath95 is often used as an approximation to @xmath90 for model selection in the vb framework . as we have seen in the sugs algorithm recursively
approximates @xmath96 . considering this in a variational framework ,
suppose we have an approximation to the posterior @xmath97 of the form @xmath98 in the above expression , we omit certain conditionings ( as will become apparent below ) to reduce the subsequent notational burdens .
we will suggest a way to update this approximation of @xmath96 , using variational ideas .
the approximation will be of the form @xmath99 .
we start the recursion with @xmath100 , @xmath101 , @xmath102 , @xmath103 .
the idea is to make a particular fixed choice at time @xmath104 for @xmath105 and not to revisit that choice at future times .
that is , the solution to the ( partial ) variational optimization at time @xmath38 is used to initialize the optimization at time @xmath104 .
the original sugs algorithm chooses @xmath106 where @xmath52 is defined in but it is possible to make a better choice than this without sacrificing the attractive computational properties of the original sugs algorithm . in particular , at time @xmath104 , we set @xmath107 , @xmath108 and choose @xmath109 for @xmath110 where @xmath111 is a truncation point for the number of mixture components and @xmath112/t)}{\alpha+i-1 } & \mbox{$j= ( i-1)\wedge t+1$}. \end{array}\right.\ ] ] @xmath113 will be chosen as an approximation to @xmath114 . to provide some intuition for this selection of @xmath113 , we remark that @xmath115 next , note that @xmath116=\mathbb{e}[p(\delta_i|\delta_{1:i-1})|y_{1:i-1}]\ ] ] where the expectation is w.r.t .
this suggests approximating @xmath118 by taking the expectation in the above expression with respect to the variational posterior @xmath119 .
although this approximation is still not easy to work with , if we we condition on @xmath120 in ( [ polyaurn2 ] ) and then take the expectation with respect to @xmath119 , we get @xmath121 .
so @xmath121 is an approximation to @xmath118 in ( [ brule ] ) and the term @xmath122 in simply approximates the integral in by replacing @xmath123 with the variational posterior @xmath124 .
as noted earlier , the original sugs algorithm can be placed in our framework by using @xmath106 , a hard " rather than soft " allocation to clusters which tends to result in greater under - estimation of uncertainty than in our approach . next , using @xmath125 as the prior for @xmath63 at time @xmath104 for processing the data point @xmath67 , the optimal choice for @xmath126 is , via @xmath127 where @xmath128 denotes expectation w.r.t .
\{@xmath129 and @xmath130 .
again , with the choice @xmath106 , exponential family mixture components and conjugate priors this reduces to the sugs update . if the mixture components are normal then the generalized update above can be done in closed form we will give details of this below .
note the attractive form of the above update .
the likelihood contribution from the @xmath131 observation is split among different mixture components @xmath64 according to the weight @xmath113 rather than assuming the most likely allocation as in the original sugs algorithm .
we can also use the variational lower bound ( [ lowerbd ] ) to approximate the marginal likelihood more accurately than with @xmath132 in the sugs algorithm .
details of this are given in the next section for the case of normal mixture components .
@xcite consider the case of dp mixtures of normals in detail . in this case , @xmath133 where @xmath134 is the mean for the @xmath104th component and @xmath135 is the precision and we have @xmath136 .
they consider a normal inverse - gamma prior for @xmath63 , @xmath137 where @xmath138 is a normal density with associated distribution with mean @xmath139 and variance @xmath140 and @xmath139 and @xmath141 are known hyperparameters , and @xmath142 is a gamma density with known parameters @xmath143 and @xmath144 . our vsugs algorithm results in @xmath145 being normal inverse - gamma also , @xmath146 where @xmath147 is the normal density with mean @xmath148 and variance @xmath149 , and @xmath150 is a gamma density with parameters @xmath151 and @xmath152 . the parameters @xmath148 , @xmath153 , @xmath151 and @xmath152 are updated recursively by ( c.f .
@xcite ) @xmath154 to calculate the terms @xmath113 in the vsugs algorithm , we also need to evaluate the integral @xmath155 in the normal case with the priors we have chosen this integral evaluates to a @xmath5 density , @xmath156 @xmath157 where @xmath158 denotes a @xmath5 density for @xmath71 with @xmath159 degrees of freedom , location parameter @xmath160 and scale parameter @xmath161 .
an approximate variational lower bound on @xmath162 can also be computed recursively .
we can think of the posterior at stage @xmath38 as the prior to be updated by the likelihood contribution for the @xmath104th observation : @xmath163 approximating @xmath118 by @xmath121 as we did previously and approximating @xmath46 by @xmath164 and calculating the lower bound ( [ lowerbd ] ) using these priors for the likelihood contribution @xmath165 gives @xmath166 where @xmath167 denotes the digamma function .
_ setup . _ we generated 100 data sets for each setting and all results are averaged over that for each data set . @xmath168 for @xmath169 with @xmath170 .
examples are shown in figure [ fig : mixtures ] .
we compared the relative errors of sugs to vsugs under different values of @xmath18 , i.e. @xmath171 . throughout @xmath172 for vsugs .
we used 50 different ( but random ) orderings of the data and chose the ordering with the maximal variational lower - bound for vsugs in section [ sec : dpm_normal ] and the best ordering for sugs as in @xcite .
we also used a standard collapsed gibbs sampling method @xcite for posterior inference on some of the datasets for comparison . to assess the performance in density estimation we compute the value of @xmath173 where @xmath174 , @xmath175 , are the estimated ( predictive ) and true density for the data , evaluated at data - point @xmath176 and examine the relative errors . _ results . _ figure [ fig : dens_examples ] shows example predictive density estimates from sugs and vsugs .
figure [ fig : relative_error](a ) shows that for large values of @xmath18 and closely spaced clusters @xmath177 , vsugs provides more accurate density estimates than sugs .
however , for @xmath178 and @xmath179 , i.e. well - separated clusters , the density estimates from sugs are relatively more accurate .
the computation time for vsugs is constant for given truncation level @xmath111 as we use a fixed maximum number of mixture components .
in contrast , the computation time required for sugs is variable and depends both on the data set and the order in which the data is processed .
figure [ fig : relative_error](b ) considers the computation burden for the two methods .
in particular , for large values of @xmath18 and more mixture components , sugs can be computationally quite demanding due to the excessive numbers of mixture components that are realised . whilst in practice , one might estimate @xmath18 , this value is not known and hence sugs could both be significantly less accurate and computationally more expensive in many situations .
we compared the sugs and vsugs predictive densities with those obtained from collapsed gibbs sampling , we considered the case @xmath180 and show results in table [ tab : comparison_to_gibbs ] for different data sizes @xmath181 and the truncation parameter @xmath111 .
using collapsed gibbs sampling as a gold standard " , we find that vsugs consistently provides better predictive density estimates .
example computational times for @xmath182 were @xmath183 seconds for sugs , @xmath184 seconds for vsugs ( @xmath185 ) and @xmath186 seconds for collapsed gibbs sampling .
we analyzed the flow cytometry data example , which has been studied thoroughly by @xcite .
flow cytometers detect fluorescent reporter markers that typically correspond to specific cell surface or intracellular proteins on individual cells , and can assay millions of such cells in a fluid stream in minutes .
these data points are associated with one ( or more ) components of a gaussian mixture model ( @xcite ) and are from human peripheral blood cells , with 6 marker measurements each : forward scatter ( measure of cell size ) , side scatter ( measure of cell granularity ) , cd4 ( marker for helper t cells ) , ifng+il-2 ( effector crytokines ) , cd8 ( a marker for cytotoxic t cells ) , cd3 ( marker for all t cells ) ; that is , the observations are 6 dimensional ( the priors are modified to normal - inverse wishart , which leads to a similar derivation of the vsugs algorithm as in section [ sec : dpm_normal ] , in this multivariate scenario ) .
our objective is to compare the performance of vsugs to sugs and collapsed gibbs sampling for clustering and density estimation in this multivariate , large data setting .
the size of the whole data is @xmath187 with @xmath188 dimensions and @xcite state the components of these data are centered closely . in the following simulations
, we adopted a gamma prior for @xmath18 , i.e. @xmath189 for the three approaches . when considering @xmath18 as unknown we use the approach to handling uncertainty in @xmath18 described in @xcite for all algorithms .
the collapsed gibbs sampler was run for a 300 iteration burn - in followed by 1000 iterations .
this low number is adopted due to the size and complexity of the data ; these type of data scenarios are exactly those which motivate the development of sugs and vsugs algorithms . for the vsugs approximation ,
the truncation value @xmath111 is set to be @xmath190 ( we did not find significant differences in our results when @xmath111 is increased or decreased by around 10 ) .
we chose the permutation of the order of the data for vsugs and sugs as in the previous example .
_ results .
_ we first compared the computation time for the three method with @xmath191 data points randomly choose from the whole data set .
this process is repeated for @xmath192 times and we took the average value of the time cost .
the analyses through collapsed gibbs sampling were completed in approximately @xmath193 seconds while approximately @xmath194 seconds and @xmath195 seconds were required for sugs and vsugs respectively .
next , we choose another data sample of @xmath196 data points .
we were interested in the performance of all approaches in clustering and density estimation ( i.e. the predictive density ) .
the predictive density is calculated on the remaining @xmath197 data points ; the collapsed gibbs sampler analysis was repeated @xmath198 times .
the performance of predictive density estimates obtained by the three approaches are shown in table [ tab : flow ] .
the collapsed gibbs sampling method has the greatest predictive ability with vsugs showing greater predictive power than sugs .
this illustrates that the vsugs approximation is performing better than sugs with regard to density estimation and provides an efficient way of detecting and drawing inferences about rare populations in the presence of very large datasets .
figure [ fig : cyto ] shows that sugs has difficulty approximating the data density whilst our vsugs approach better approximates the density estimates by gibbs sampling .
we now turn to our original motivating snp genotyping example and examined the use of vsugs and sugs for a hierarchical bayesian clustering problem .
_ for our experiments , we considered a genotyping dataset that were considered in a recent comparison study @xcite .
the study consists of @xmath188 different individuals , each individual was genotyped three times using the illumina humanhap650 genotyping array which produces approximately 650,000 two - dimensional measurements per sample .
we normalised the data by taking @xmath199 transforms and performed quantile normalisation between the two channels to correct for allele - specific biases .
_ model . _
we clustered the data using a three - class bayesian mixture model : @xmath200 where @xmath201 is the multinomial distribution , we fixed @xmath202 and the class conditional density @xmath203 is given by a dirichlet process mixture of bivariate normal distributions ( one dpm for each genotype ) .
we implemented the model using both the sugs and vsugs approaches to fit the dpms . for comparison , we classified the genotyping data using a standard genotyping tool , genosnp @xcite which models the class - conditional densities using multivariate student-@xmath5 distribution and also performs inference using variational methods .
we used majority vote over the three replicates per sample to obtain the _ true _ genotypes from the genosnp genotype calls .
_ results .
_ over the @xmath204 samples , the average concordance of our vsugs implementation was 99.45% compared to 98.90% for the sugs implementation .
figure [ fig : genotyping_example ] illustrates genotyping performance for one particular sample .
figure [ fig : genotyping_example](c ) indicates that , using genotype calls from genosnp as a reference , vsugs produced the highest concordance with the genosnp results across a range of genosnp call probability thresholds . for the snps with discordant genotype calls between genosnp and sugs / vsugs , we plotted the empirical distribution of the maximum genotype call probabilities for these snps .
figure [ fig : genotyping_example](d ) shows that for vsugs the discordant genotype calls were associated with snps where the maximum genotype classification probability was around 0.5 . with sugs ,
discordant calls have probabilities in excess of 0.5 .
in this paper we have considered vsugs as a generalization of the sugs algorithm for fast inference from dpm models .
we saw that when the components of the mixture appear to be close in some sense , vsugs seems to consistently outperform sugs with regards to density estimation and this improvement is also found by using our variational lower - bound for model selection .
in addition , when @xmath18 grows , we have found vsugs performs significantly better , with less computation time .
we have found that for real data examples , vsugs can detect features of the data which sugs can not . in terms of extensions to our work , we are currently considering the development of vsugs for new models .
in particular , we are developing the ideas for hierarchical mixture models and infinite hidden markov models .
these initial experiments suggests that vsugs can prove to be a very efficient tool for fast , but approximate , inference from a wide class of statistical models .
we thank ioanna manolopoulou for providing codes and data for the flow cytometry example . the second and fourth authors acknowledge support from the moe singapore . | in this article we propose an improvement on the sequential updating and greedy search ( sugs ) algorithm @xcite for fast fitting of dirichlet process mixture models .
the sugs algorithm provides a means for very fast approximate bayesian inference for mixture data which is particularly of use when data sets are so large that many standard markov chain monte carlo ( mcmc ) algorithms can not be applied efficiently , or take a prohibitively long time to converge .
in particular , these ideas are used to initially interrogate the data , and to refine models such that one can potentially apply exact data analysis later on .
sugs relies upon sequentially allocating data to clusters and proceeding with an update of the posterior on the subsequent allocations and parameters which assumes this allocation is correct .
our modification softens this approach , by providing a probability distribution over allocations , with a similar computational cost ; this approach has an interpretation as a variational bayes procedure and hence we term it variational sugs ( vsugs ) .
it is shown in simulated examples that vsugs can out - perform , in terms of density estimation and classification , the original sugs algorithm in many scenarios .
in addition , we present a data analysis for flow cytometry data , and snp data via a three - class dirichlet process mixture model illustrating the apparent improvement over sugs .
+ * key - words * : approximate bayesian inference ; mixture modelling ; variational bayes ; density estimation . *
a sequential algorithm for fast fitting of dirichlet process mixture models * by david j. nott@xmath0 , xiaole zhang@xmath1 , christopher yau@xmath1 & ajay jasra@xmath0 @xmath0department of statistics & applied probability , national university of singapore , singapore , 117546 , sg .
+ e-mail:`[email protected] , [email protected] ` + @xmath1department of mathematics , imperial college london , london , sw7 2az , uk .
+ e-mail:`[email protected] , [email protected] ` |
throughout this paper , a graph @xmath1 always means a simple connected finite graph ( without loops and parallel edges ) .
+ we use the standard notations @xmath10 and @xmath11 for the vertex - set , edge - set and maximum degree of a graph @xmath1 , respectively .
further , the degree of a vertex @xmath12 will be denoted by @xmath13 . +
a matching @xmath14 of @xmath1 is a set of independent edges of @xmath1 and we call @xmath0$]-matching a matching of @xmath1 of size @xmath3 .
we define a graph @xmath1 to be @xmath0$]-coverable if each edge of @xmath1 belongs to an @xmath0$]-matching ( for a complete treatment of matching theory we refer to the classical book of lovasz and plummer @xcite ) . + the main aim of this paper is the study of the excessive @xmath0$]-index of a graph .
the excessive @xmath0$]-index of @xmath1 is defined in @xcite as the minimum number of @xmath0$]-matchings needed to cover the edge - set of @xmath1 and denoted by @xmath2}(g)$ ] .
( set @xmath2}(g)=\infty$ ] if @xmath1 is not @xmath0$]-coverable . )
+ the particular case @xmath15 is largely studied , where we are asking for a cover of the edge - set with perfect matchings ( @xmath16-regular spanning subgraphs ) . in this case the excessive @xmath0$]-index is simply called the excessive index in @xcite and the perfect matching index in @xcite .
the author proved in @xcite and @xcite how two well - known conjectures of berge , fulkerson and seymour ( see @xcite and @xcite ) can be easily stated in terms of the excessive index .
furthermore , some general properties and the exact value of the excessive index for some relevant families of graphs is studied in various recent papers ( see @xcite,@xcite , @xcite and @xcite ) .
+ for what concerns small values of @xmath3 , trivially @xmath4}(g)=|e(g)|$ ] holds and it is quite easy to prove that @xmath6}(g)=\max\{\chi'(g ) , \lceil |e(g)|/2 \rceil \}$ ] holds for each @xmath5$]-coverable graph @xmath1 , where @xmath7 denotes the chromatic index of @xmath1 .
the case @xmath8 is far from being trivial and it is completely solved by cariolaro and fu in @xcite ( see theorem [ 3index ] ) . in this paper
we prove a general formula to compute the excessive @xmath9$]-index of a tree .
this formula gives a complete answer to a problem posed in @xcite .
furthermore , we exhibit a graph ( not a tree ) for which such kind of formula does not work .
finally , we conjecture a general formula for the excessive @xmath0$]-index of any tree ( for all values of @xmath3 ) and for the excessive @xmath9$]-index of a general graph . +
in what follows , we will say that a matching @xmath14 of a graph @xmath1 can be extended to an @xmath0$]-matching if there exists an @xmath0$]-matching of @xmath1 which contains @xmath14 .
a matching is said to be _ maximal _ if it can not be extended to a larger matching . +
a set @xmath17 of @xmath0$]-matchings of a graph @xmath1 such that each edge of @xmath1 belongs to at least one of the @xmath0$]-matchings is an _
@xmath0$]-cover _ of @xmath1 .
an @xmath0$]-cover of @xmath1 is minimum if its size is equal to the excessive @xmath0$]-index of @xmath1 .
+ the concepts of excessive @xmath0$]-index and @xmath0$]-cover are strictly related to chromatic index and edge - colorings of a graph .
indeed the chromatic index can be defined as the minimum number of matchings , without constraints on their sizes , needed to cover the edge - set of a graph .
furthermore , an @xmath0$]-cover @xmath17 of @xmath1 can be viewed as a _ multicoloring _
@xmath18 of the edge - set of @xmath1 .
more precisely , we mean that each @xmath0$]-matching of @xmath17 is a color of @xmath18 and so each edge @xmath19 receives a number of colors equal to the number of @xmath0$]-matchings which contain @xmath20 .
we will say that @xmath18 is the multicoloring associated to the @xmath0$]-cover @xmath17 .
+ it was useful in @xcite and @xcite to consider edge colorings whose color classes have approximately the same size .
the same idea turned up to be useful in our proofs as well .
more precisely , an edge coloring is called _ equalized _ if it has the property that the difference between the sizes of any two color classes is at most @xmath16 .
we shall often use ( without further reference ) the following lemma , due to de werra @xcite and ( independently ) mcdiarmid @xcite . [
equalized ] let @xmath1 be a graph .
then @xmath1 has an equalized edge coloring with @xmath7 colors .
our main aim is the determination of a formula for the excessive @xmath9$]-index of a tree . a particular role in our proofs
is played by caterpillars .
a caterpillar is a tree in which there exists a path that contains every node of degree two or more .
the ( necessarily unique ) path induced by the vertices of degree at least two of a caterpillar is called the _ spine _ of the caterpillar .
we denote a caterpillar by @xmath21 , where the path @xmath22 is the spine of the caterpillar and @xmath23 .
see figure [ caterpillar ] for an example .
, width=207 ]
let @xmath1 be an @xmath0$]-coverable graph .
we call a subgraph @xmath24 of @xmath1 a _ splitting set _ if 1 .
no pair of edges of @xmath24 can be extended to an @xmath0$]-matching of @xmath1 ; 2 .
@xmath24 is not a star @xmath25 for some @xmath26 .
the additional condition @xmath27 is assumed in order to avoid trivial cases .
when the cardinality @xmath3 is clear from the context , we will use the notation @xmath28 in place of @xmath24 .
we will denote by @xmath29 the cardinality of the largest splitting set of @xmath1 .
the concept of a splitting set was introduced in @xcite in order to produce a formula for the computation of the excessive @xmath30$]-index of an arbitrary graph .
more precisely , cariolaro and fu proved the following theorem .
[ 3index ] let @xmath1 be a @xmath30$]-coverable graph .
then , @xmath31}(g)=\max \ { \chi'(g),\lceil |e(g)| / 3 \rceil , s(g)\}.\ ] ]
the same authors introduce in @xcite the notion of an @xmath0$]-compatible graph , that is a graph for which @xmath2}(g)=\max \
{ \chi'(g),\lceil |e(g)| / m \rceil \}$ ] .
in particular , they prove that all trees are @xmath30$]-compatible .
[ 3indextrees ] let @xmath32 be a @xmath30$]-coverable tree .
then , @xmath31}(t)=\max \ { \chi'(t),\lceil |e(t)|/ 3 \rceil \}.\ ] ] furthermore , they construct a tree which is not @xmath33$]-compatible and they leave the determination of @xmath9$]-compatible trees as an open problem . in this paper
, we give an answer to this question in a very strong sense by proving that there are exactly three @xmath9$]-coverable trees which are not @xmath9$]-compatible ( see figure [ no4comp ] ) .
$ ] -compatible trees @xmath34 , @xmath35 and @xmath36.,width=415 ] since our main result concerns trees , we need to establish some general properties of splitting sets of an @xmath0$]-coverable tree .
let @xmath32 be an @xmath0$]-coverable tree .
let @xmath28 be a splitting set of @xmath32 .
then , @xmath37 .
_ suppose @xmath38 holds for a vertex @xmath39 of @xmath28 .
let @xmath40 be a @xmath5$]-matching of @xmath28 such that @xmath41 is incident to @xmath39 ( such a @xmath5$]-matching does exist since @xmath28 is not a star ) .
the tree @xmath32 is @xmath0$]-coverable , so @xmath42 belongs to an @xmath0$]-matching , say @xmath43 , of @xmath32 .
note that @xmath44 for @xmath45 , since @xmath28 is a splitting set .
if @xmath41 is adjacent to an edge @xmath46 , then @xmath47 is an @xmath0$]-matching of @xmath32 containing two edges of @xmath28 , a contradiction . on the other hand , if @xmath41 , @xmath46 are indpendent edges , then @xmath48 is still an @xmath0$]-matching of @xmath32 containing two edges of @xmath28 , a contradiction again .
@xmath49 + let @xmath32 be an @xmath0$]-coverable tree .
let @xmath28 be a splitting set of @xmath32 .
then , @xmath50 .
_ suppose @xmath51 , then @xmath52 .
denote by @xmath53 the vertex set of @xmath28 , and without loss of generality we suppose
@xmath54$],@xmath55 $ ] are two independent edges of @xmath28 .
let @xmath56 be a vertex of @xmath57 adjacent to @xmath58 .
the set @xmath59\ } \cup \{[x_1,x_2],[x_3,x_4]\}$ ] is an @xmath0$]-matching of @xmath32 containing two edges of @xmath28 , a contradiction .
in what follows we will make large use of lemma [ 35covering ] .
this lemma is just a particular case of a more general result proved in @xcite .
we repeat its proof in our particular context for the reader s convenience . [ 35covering ]
let @xmath60 be a set of matchings of @xmath1 such that 1 .
each edge of @xmath1 belongs to an element of @xmath17 , 2 .
@xmath61 .
then , there exists a @xmath9$]-cover of @xmath1 of size @xmath62 .
_ suppose @xmath17 is not a @xmath9$]-cover of @xmath1 , then there exist two matchings @xmath63 and @xmath64 such that @xmath65 .
consider the subgraph @xmath66 , @xmath67 , of @xmath1 : a connected component of @xmath66 is either a path or a cycle of even length ( possibly a single edge belonging to both @xmath63 and @xmath64 ) . in the latter case
the connected component of @xmath66 has the same number of edges as @xmath63 and @xmath64 .
then , by @xmath65 , there exists at least a connected component consisting of a path @xmath68 of odd length starting and finishing with edges of @xmath64 .
the exchange of edges in @xmath68 increases @xmath69 by one and decreases @xmath70 by one .
the iteration of this process furnishes a @xmath9$]-cover of @xmath1 of size @xmath62 .
@xmath49 + [ 4coverabletrees ] let @xmath32 be an @xmath0$]-coverable tree .
if @xmath71 is a tree such that @xmath72 and @xmath73\}$ ] where @xmath74 , then @xmath71 is @xmath0$]-coverable and @xmath2}(t')\leq \chi'_{[m]}(t)+1 $ ] .
_ let @xmath42 be an edge of @xmath32 incident @xmath39 . since @xmath32 is @xmath0$]-coverable @xmath42 belongs to at least an @xmath0$]-matching , namely @xmath75 , of @xmath32 .
we add the @xmath0$]-matching @xmath76,e_2,\ldots , e_m\}$ ] to an @xmath0$]-cover of @xmath32 thus obtaining an @xmath0$]-cover of @xmath71 of size @xmath2}(t)+1 $ ] , as required .
@xmath49 + we prove in theorem [ 4compatibletrees ] that each tree having a @xmath9$]-compatible subtree is itself compatible .
note that this property is false in general for an arbitrary graph , for instance the petersen graph minus an edge is @xmath33$]-compatible but the petersen graph is not . [ 4compatibletrees ] let @xmath32 be a @xmath9$]-compatible tree .
each tree having @xmath32 as a subgraph is @xmath9$]-compatible .
_ let @xmath71 be a tree such that @xmath72 and @xmath73\}$ ] , where @xmath74 .
in order to prove that @xmath71 is @xmath9$]-compatible we distinguish three cases , according to @xmath77 being less than , equal to or greater than @xmath78 , respectively . in
what follows @xmath17 will always denote a minimum @xmath9$]-cover of @xmath32 and @xmath18 the multicoloring of @xmath32 arising from @xmath17 .
consider the case @xmath79 .
since @xmath32 is @xmath9$]-compatible it follows @xmath80}(t)=\delta(t)$ ] . if @xmath81 , then @xmath82 . by lemma [ 4coverabletrees ] , we have @xmath80}(t ' ) \leq \chi'_{[4]}(t)+1 = \delta(t)+1=\delta(t')$ ] , that is @xmath80}(t')=\delta(t')$ ] . if @xmath83 , then @xmath84 .
since @xmath79 , there exists at least an edge @xmath85 of @xmath32 receiving more than one color of @xmath18 .
denote by @xmath86 one of the colors of @xmath85 .
if @xmath39 is an end - vertex of @xmath85 , then we can replace in the @xmath9$]-matching @xmath86 the edge @xmath85 with the edge @xmath87 $ ] , producing a @xmath9$]-cover of @xmath71 of size @xmath88 .
if all the edges incident @xmath39 have just one color of @xmath18 , then , by @xmath89}(t)$ ] , there exists a @xmath9$]-matching @xmath90 not covering @xmath39 . remove the edge @xmath85 from the @xmath9$]-matching @xmath86 and add the edge @xmath87 $ ] to the @xmath9$]-matching @xmath90 . that produces a cover of @xmath71 satisfying the conditions of lemma [ 35covering ] .
hence , there exists a @xmath9$]-cover of @xmath71 of size @xmath88 .
+ consider the case @xmath91 . since @xmath32 is @xmath9$]-compatible it follows @xmath80}(t)=\delta(t)=\lceil \frac{|e(t)|}{4 } \rceil$ ] .
if @xmath81 , then @xmath82 . by lemma
[ 4coverabletrees ] , we have @xmath80}(t ' ) \leq \chi'_{[4]}(t)+1 = \delta(t)+1=\delta(t')$ ] , that is @xmath80}(t')=\delta(t')$ ] . if @xmath83 , then @xmath92 .
hence , by lemma 4 in @xcite , @xmath71 is @xmath9$]-compatible .
+ finally , consider the case @xmath93 . since @xmath32 is @xmath9$]-compatible it follows @xmath80}(t)=\lceil \frac{|e(t)|}{4 } \rceil$ ] .
if @xmath81 , then @xmath82 and @xmath94 . if @xmath95 holds , then @xmath96 . hence , by lemma 4 in @xcite
, @xmath71 is @xmath9$]-compatible . + if @xmath97 , then there exists an edge of @xmath32 having more than one color of @xmath18 and we can use exactly the same argument already used in the case @xmath79 and @xmath83 .
this concludes the proof of the proposition .
@xmath49 + [ diam7 ] let @xmath32 be a @xmath9$]-coverable tree of diameter at least seven .
then @xmath32 is @xmath9$]-compatible .
_ let @xmath32 be a tree of diameter greater than @xmath98 , that is @xmath32 has a path of length @xmath99 as subtree . since a path of length @xmath99 is a @xmath9$]-compatible tree , it follows from theorem [ 4compatibletrees ] that @xmath32 is @xmath9$]-compatible .
from now on , we can assume the diameter of @xmath32 equal to @xmath98 .
+ let @xmath100 be a @xmath98-path @xmath68 in @xmath32 .
if @xmath32 is not a caterpillar then it contains one of the two trees of diameter @xmath98 in figure [ diam7a ] .
since both of them are @xmath9$]-compatible , it follows from theorem [ 4compatibletrees ] that @xmath32 is @xmath9$]-compatible .
+ $ ] -compatible trees of diameter @xmath98.,width=264 ] assume that @xmath32 is a caterpillar @xmath101 with spine @xmath102 .
suppose that exactly one @xmath103 is greater than @xmath104 and we can assume , up to symmetry , @xmath105 .
if @xmath106 ( or @xmath107 ) , then @xmath54 $ ] does not belong to a @xmath9$]-matching of @xmath32 .
if @xmath108 , then @xmath55 $ ] does not belong to a @xmath9$]-matching of @xmath32 . in any case
@xmath32 is not @xmath9$]-coverable .
+ then there are at least two @xmath103 s greater than @xmath104 . if @xmath106 and @xmath109 , for some @xmath110 , then @xmath32 is not @xmath9$]-coverable .
if @xmath107 and @xmath111 , then @xmath54 $ ] does not belong to a @xmath9$]-matching . finally ,
if @xmath108 and @xmath111 , then @xmath55 $ ] does not belong to a @xmath9$]-matching of @xmath32 and , again , @xmath32 is not @xmath9$]-coverable .
that leaves just two cases : @xmath112 and @xmath113 .
then @xmath32 contains one of the two @xmath9$]-compatible caterpillars of figure [ diam7b ] , hence @xmath32 is @xmath9$]-compatible from theorem [ 4compatibletrees ] .
$ ] -compatible caterpillars of diameter @xmath98.,width=302 ] in the same way one can check that if a caterpillar @xmath32 does not contain the two caterpillars in figure [ diam7b ] and it has either three or four @xmath103 s greater than @xmath104 , then @xmath32 is not @xmath9$]-coverable . finally , if five @xmath103 s are greater than @xmath104 , then @xmath32 contains one of the caterpillars in figure [ diam7b ] , hence it is @xmath9$]-compatible from proposition [ 4compatibletrees ] .
@xmath49 + in the following proofs we will use the fact that all trees in figure [ 4comp ] are @xmath9$]-compatible ( for each tree we exhibit a @xmath9$]-cover which uses a number of @xmath9$]-matchings which is equal to the maximum degree of the tree ) .
$ ] -compatible trees.,width=415 ] furthermore , we would like to recall that the chromatic index @xmath114 of a tree is equal to its maximum degree @xmath78 .
[ cat01110 ] let @xmath32 be a tree having the caterpillar @xmath34 as a subgraph . then
, @xmath115}(t)=\max\{\delta(t),\lceil |e(t)|/ 4 \rceil , s(t)\}.\ ] ] _ proof .
_ first of all , we prove that the excessive @xmath9$]-index of the three caterpillars @xmath34 , @xmath35 and @xmath36 is equal to @xmath116 , that is exactly the cardinality of their largest splitting set .
let @xmath117 be the spine of @xmath34 and let @xmath56 be the further vertex adjacent to @xmath58 , for @xmath118 ( see figure [ no4comp ] ) .
it is an easy check that the set of edges @xmath119,[x_2,x_3],[x_3,x_4],[x_4,x_5]\}$ ] is a splitting set of size four for the caterpillar . on the other hand ,
a @xmath9$]-cover of @xmath34 is given by the following four @xmath9$]-matchings @xmath120\ } \cup \bigcup _ { \begin{array}{c } j\neq i \\
j \neq i+1 \end{array } } \ { [ x_j , y_j ] \},\ ] ] for @xmath121 .
the same argument can be easily adapted to obtain the same result for @xmath35 and @xmath36 .
+ let @xmath39 be an arbitrary vertex of @xmath34 .
now we prove that each tree @xmath32 obtained by adding a new vertex @xmath122 and the edge @xmath123 $ ] to @xmath34 is either @xmath9$]-compatible or it is the caterpillar @xmath35 : if @xmath124 ( @xmath125 ) , then @xmath32 has diameter @xmath98 and it is @xmath9$]-compatible from lemma [ diam7 ] , + if @xmath126 ( @xmath127 ) , then @xmath32 contains the graph @xmath128 in figure [ 4comp ] , hence it is @xmath9$]-compatible from theorem [ 4compatibletrees ] , + if @xmath129 , then @xmath32 contains the graph @xmath130 in figure [ 4comp ] , hence it is @xmath9$]-compatible from theorem [ 4compatibletrees ] , + if @xmath131 ( @xmath132 ) , then @xmath32 contains the graph @xmath133 in figure [ 4comp ] , hence it is @xmath9$]-compatible from theorem [ 4compatibletrees ] , + if @xmath134 , then @xmath32 contains the graph @xmath135 in figure [ 4comp ] , hence it is @xmath9$]-compatible from theorem [ 4compatibletrees ] .
+ finally , if @xmath136 ( @xmath137 ) , then @xmath32 is the caterpillar @xmath35 .
we can repeat exactly the same argument starting from @xmath35 and we obtain that each tree which contains @xmath35 is either @xmath9$]-compatible or it is the caterpillar @xmath36 . repeating the same argument again we obtain that each tree containing @xmath36 is @xmath9$]-compatible and the assertion follows .
@xmath49 + [ formula_exc_4_index_trees ] let @xmath32 be a @xmath9$]-coverable tree .
then , @xmath115}(t)=\max\{\delta(t),\lceil |e(t)| / 4 \rceil , s(t)\}.\ ] ] _ proof .
_ we can suppose @xmath138 , otherwise @xmath32 is @xmath9$]-compatible by lemma 4 in @xcite and the relation @xmath80}(t)=\max\{\delta(t),\lceil \frac{|e(t)|}{4 } \rceil \}$ ] holds .
+ suppose the existence of a maximal matching of @xmath32 of cardinality @xmath104 .
hence , all edges of @xmath32 are incident with an edge of this @xmath5$]-matching and so @xmath32 is not @xmath9$]-coverable , a contradiction . on the other hand ,
if all maximal matchings have cardinality at least @xmath116 , then @xmath80}(t)=\delta$ ] since each color class of any equalized @xmath139-coloring of @xmath32 ( see lemma [ equalized ] ) has cardinality at most @xmath116 and all color classes of cardinality less than @xmath116 can be extended to a @xmath9$]-matching of @xmath32 .
so we can assume the existence of a @xmath30$]-matching , say @xmath140,e_2=[x_2,y_2],e_3=[x_3,y_3]\}$ ] , such that each edge of @xmath32 is incident with an edge of @xmath141 .
there are seven possible cases , up to symmetry , to connect the edges of @xmath141 , as illustrated in figure [ cases ] . in the figure
the edges of @xmath141 are the three vertical edges .
+ .,width=226 ] to complete our task we have to prove , for each of the seven cases presented , that either @xmath32 is @xmath9$]-compatible or @xmath32 contains the subgraph @xmath34 .
* : if just one of the vertices @xmath56 has degree greater than one in @xmath32 , then the corresponding edge @xmath142 $ ] does not belong to a @xmath9$]-matching of @xmath32 , a contradiction since @xmath32 is @xmath9$]-coverable .
then , we can assume that at least two of the vertices @xmath56 have degree greater than two in @xmath32 , so we have three subcases : * * @xmath143 , @xmath144 , @xmath145 : + if @xmath146 , then @xmath147 $ ] does not belong to a @xmath9$]-matching of @xmath32 , a contradiction ; if @xmath148 , then @xmath149 $ ] does not belong to a @xmath9$]-matching of @xmath32 , a contradiction again .
hence , we have @xmath150 , @xmath151 which implies @xmath32 has the tree @xmath152 of figure [ 4comp ] as a subgraph . by theorem [ 4compatibletrees ]
, @xmath32 is @xmath9$]-compatible .
* @xmath143 , @xmath153 , @xmath154 : + if @xmath146 ( @xmath155 ) , then @xmath156 $ ] ( @xmath54 $ ] ) does not belong to a @xmath9$]-matching of @xmath32 , a contradiction . hence , @xmath32 contains the tree @xmath34 as a subgraph , by lemma [ cat01110 ] the formula holds .
+ * * @xmath143 , @xmath144 , @xmath154 : + if @xmath157 , then @xmath147 $ ] does not belong to a @xmath9$]-matching of @xmath32 , a contradiction . without loss of generality we can suppose @xmath150 .
hence , @xmath32 contains the graph @xmath130 of figure [ 4comp ] and @xmath32 is @xmath9$]-compatible by theorem [ 4compatibletrees ] . * : if @xmath158 , then @xmath147 $ ] does not belong to a @xmath9$]-matching of @xmath32 , a contradiction .
if @xmath143 and @xmath159 , then @xmath32 has diameter at least @xmath98 , so it is @xmath9$]-compatible by lemma [ diam7 ] . without loss of generality , we can suppose @xmath143 and @xmath160 .
we have @xmath150 and @xmath161 otherwise @xmath162 $ ] does not belong to a @xmath9$]-matching of @xmath32 .
furthermore , @xmath163 otherwise @xmath54 $ ] does not belong to a @xmath9$]-matching of @xmath32 ; hence , @xmath32 has @xmath34 as subgraph and by lemma [ cat01110 ] the formula holds .
* : if exactly one of the vertices @xmath56 has degree greater than one in @xmath32 , then the corresponding edge @xmath142 $ ] does not belong to a @xmath9$]-matching of @xmath32 , a contradiction .
we have three subcases : * * @xmath143 , @xmath144 and @xmath145 : + @xmath32 contains the graph @xmath128 of figure [ 4comp ] and @xmath32 is @xmath9$]-compatible by theorem [ 4compatibletrees ] . + * * @xmath143 , @xmath153 and @xmath154 : + @xmath32 has diameter at least @xmath98 , @xmath32 is @xmath9$]-compatible by lemma [ diam7 ] . + * * @xmath164 , @xmath144 and @xmath154 : + if @xmath155 , then the edge @xmath147 $ ] does not belong to a @xmath9$]-matching of @xmath32 , a contradiction . hence , @xmath165 and @xmath32 contains the graph @xmath130 of figure [ 4comp ] and @xmath32 is @xmath9$]-compatible by theorem [ 4compatibletrees ] . + * : if both vertices @xmath166 and @xmath167 have degree one in @xmath32 , then the edge @xmath147 $ ] does not belong to a @xmath9$]-matching of @xmath32 , a contradiction .
hence , @xmath32 has diameter at least seven and @xmath32 is @xmath9$]-compatible by lemma [ diam7 ] . +
* : if just one of the vertices @xmath56 has degree greater than one in @xmath32 , then the corresponding edge @xmath142 $ ] does not belong to a @xmath9$]-matching of @xmath32 . without loss of generality
, we can suppose @xmath143 .
hence , @xmath32 has diameter at least seven and it is @xmath9$]-compatible by lemma [ diam7 ] . *
: @xmath32 has diameter at least @xmath98 and it is @xmath9$]-compatible by lemma [ diam7 ] . * : if just one of the vertices @xmath56 has degree greater than one in @xmath32 , then the corresponding edge @xmath142 $ ] does not belong to a @xmath9$]-matching of @xmath32 , a contradiction . without loss of generality , we can suppose @xmath143 and @xmath168 .
if @xmath154 holds then @xmath32 contains the @xmath9$]-compatible tree @xmath169 of figure [ 4comp ] and @xmath32 is @xmath9$]-compatible by theorem [ 4compatibletrees ] . if @xmath145 , then at least one between @xmath170 and @xmath171 has degree greater than @xmath104 in @xmath32 , otherwise @xmath32 is not @xmath9$]-coverable . in both cases , @xmath32 contains the @xmath9$]-compatible tree @xmath130 of figure [ 4comp ] and @xmath32 is @xmath9$]-compatible by theorem [ 4compatibletrees ] .
@xmath49 the proof of the previous theorem provides also the complete list of all trees which are not @xmath9$]-compatible .
the caterpillars @xmath34,@xmath35 and @xmath36 are the unique trees which are not @xmath9$]-compatible .
note that theorem [ formula_exc_4_index_trees ] can not be generalized to the entire class of @xmath9$]-coverable graphs .
consider the graph @xmath1 in figure [ 4index_8 ] .
it has @xmath172 , @xmath173 and @xmath174 ( the largest splitting set is a circuit of length @xmath175 ) . on the other hand ,
each @xmath9$]-matching of @xmath1 contains at most two edges in the clique of size six of @xmath1 .
hence , @xmath80}(g ) \geq \lceil 15 / 2 \rceil = 8 $ ] .
it is possible to verify that @xmath1 admits a @xmath9$]-cover of size @xmath99 , that is @xmath80}(g)=8 $ ] .
+ does not hold.,width=151 ] the same idea can be used to prove that an analogous formula does not work to compute the excessive @xmath33$]-index of all trees .
consider the caterpillar @xmath176 .
each @xmath33$]-matching shares at most two edges with the spine of @xmath177 .
since the spine has seven edges , we have @xmath178}(k ) \geq \lceil 7 / 2 \rceil = 4 $ ] , but it is easy to check that @xmath179 .
+ these examples suggest a natural generalization of the concept of a splitting set .
let @xmath1 be an @xmath0$]-coverable graph .
we call a subgraph @xmath180 of @xmath1 a @xmath26-splitting set if @xmath169 no @xmath181$]-matching of @xmath180 can be extended to an @xmath0$]-matching of @xmath1 ; @xmath128 @xmath180 has a @xmath181$]-matching .
we will denote by @xmath182 the cardinality of the largest @xmath26-splitting set of @xmath1 ( for a fixed @xmath3 ) .
+ it is straightforward that a @xmath16-splitting set is exactly a splitting set previously defined and @xmath183 holds . with this new concept in our hands ,
we propose the following conjectures : let @xmath32 be an @xmath0$]-coverable tree .
then , @xmath184}(t)=\max\{\chi'(t),\lceil |e(t)| / m \rceil , s^1(t),\lceil s^2(t ) / 2 \rceil,\ldots,\lceil s^{m-1}(t ) / ( m-1 ) \rceil\}.\ ] ] let @xmath1 be a @xmath9$]-coverable graph .
then , @xmath115}(g)=\max\{\chi'(g),\lceil |e(g)| / 4 \rceil , s^1(g),\lceil s^2(g ) / 2 \rceil , \lceil s^{3}(g ) / 3 \rceil\}.\ ] ] we would like to stress that the natural generalization of the second conjecture to the case @xmath185 is not possible due to the petersen graph .
+ a possible way to attack the first conjecture is the idea that @xmath26-splitting sets of a tree should be not very large ( with respect to @xmath3 ) .
theorem [ 4compatibletrees ] can be easily generalized to arbitrary values of @xmath3 and so the following corollary holds .
let @xmath32 be an @xmath0$]-coverable tree .
if @xmath186 , then @xmath32 is @xmath0$]-compatible . proof .
the path of length @xmath187 is @xmath0$]-compatible .
the author is grateful to one of the anonymous reviewers for his / her very precise comments and advices . | the excessive @xmath0$]-index of a graph @xmath1 , denoted by @xmath2}(g)$ ] , is the minimum number of matchings of size @xmath3 needed to cover the edge - set of @xmath1 .
we set @xmath2}(g)=\infty$ ] if such a cover does not exist and we call a graph @xmath1 @xmath0$]-coverable if its excessive @xmath0$]-index is finite
. obviously @xmath4}(g)=|e(g)|$ ] and it is easy to prove for a @xmath5$]-coverable graph @xmath1 that @xmath6}(g)=\max\{\chi'(g),\lceil |e(g)|/2 \rceil \}$ ] holds , where @xmath7 denotes the chromatic index of @xmath1 .
the case @xmath8 is completely solved by cariolaro and fu in @xcite . in this paper we prove a general formula to compute the excessive @xmath9$]-index of a tree and we conjecture a possible generalization for any value of @xmath3 .
furthermore , we prove that such a formula does not work for the excessive @xmath9$]-index of an arbitrary graph . _
keywords : excessive index , matchings , trees , caterpillars .
+ msc(2010 ) : 05c70 , 05c15 _ |
r. l. and j. l. z. would like to thank professor w. wernsdorfer and professor r. sessoli for providing their paper ( ref . 1 ) .
a. garg , europhys . lett . * 22 * , 205 ( 1993 ) ; h. b. braun and d. loss , europhys . lett . * 31 * , 555 ( 1995 ) ; phys . rev .
b * 53 * , 3237 ( 1996 ) ; e. m. chudnovsky and d. p. divincenzo , phys . rev .
b * 48 * , 10548 ( 1993 ) ; a. garg , phys . rev .
b * 51 * , 15161 ( 1995 ) . fig .
1 the relative tunnel splitting of the first excited level ( @xmath136 ) @xmath137 as a function of @xmath138 for integer @xmath139 and half - integer @xmath140 excess spins at @xmath141 . here
@xmath142 , @xmath143 , and @xmath144 .
2 the relative tunnel splitting of the first excited level ( @xmath136 ) @xmath137 as a function of @xmath138 for @xmath141 , @xmath145 , @xmath146 , @xmath147 , @xmath148 , @xmath149 and @xmath150 , respectively . here
@xmath151 , @xmath142 , @xmath143 , and @xmath144 .
3 the relative tunnel splitting @xmath152 of the ground - state level ( @xmath153 ) and the first excited level ( @xmath136 ) as a function of @xmath138 for integer @xmath154 excess spins at @xmath141 . here
@xmath142 , @xmath155 , and @xmath144 . | the quantum interference effects induced by the topological phase are studied analytically in biaxial antiferromagnets with an external magnetic field at an arbitrarily angle .
this study provides a nontrivial generalization of the kramers degeneracy for equivalent double - well system to coherently spin tunneling at ground states as well as low - lying excited states for antiferromagnetic system with asymmetric twin barriers .
the spin - phase interference effects are found to depend on the orientation of the magnetic field distinctly . *
pacs number(s ) * : + 03.65.vf phase : geometric ; dynamic or topological + 75.45.+j macroscopic quantum phenomena in magnetic systems + 75.50.ee antiferromagnetics + 73.40.gk tunneling one recent experiment on the molecular magnets fe@xmath0 showed a direct evidence of the topological part of the quantum spin phase ( berry phase ) in the spin dynamics.@xcite the importance of the topological berry phase in spin tunneling was elucidated by loss _
et al . , _ and von delft and henly.@xcite they showed that this term can lead to destructive ( for half - integer total spins ) and instructive ( for integer spins ) interference between opposite winding tunneling paths in single - domain ferromagnetic ( fm ) particles . while spin - parity effects are sometimes be related to kramers degeneracy,@xcite they typically go beyond the kramers theorem in a rather unexpected way.@xcite similar spin - parity effects were found in antiferromagnetic ( afm ) particles , where only the integer excess spins can tunnel but not the half - integer ones.@xcite the effects of magnetic field along the hard@xcite and medium@xcite axis were studied in afm particles .
theoretical results@xcite showed that in afm particles the exchange energy is enhanced to the magnetic anisotropy , which leads to tunneling of the nel vector a much stronger effect than tunneling of magnetization in fm particles .
therefore , the afm particle is expected to be a better candidate for observing quantum tunneling than the fm particle with a similar size . up to now theoretical studies on afm tunneling@xcite
have been focused on spin - phase interference between two opposite ground - state tunneling paths .
the spin - phase interference between excited - level tunneling paths are unknown for afm particles .
moreover , the previous works@xcite have been confined to the condition that the magnetic field be applied along the easy , medium , or hard axis , separately .
the purpose of this letter is to study the resonant quantum tunneling and spin - phase interference at excited levels in afm particles placed in a magnetic field at an arbitrary angle @xmath1 .
therefore , _ our study provides a nontrivial generalization of the kramers degeneracy for equivalent double - well system to coherently spin tunneling at ground states as well as low - lying excited states for afm system with asymmetric twin barriers caused by the arbitrarily directed magnetic field . _
the system of interest is a small ( @xmath2 nm in radius ) , single - domain , afm particle at a temperature well below its anisotropy gap . according to the two - sublattice model,@xcite
there is a strong exchange energy @xmath3 between the two sublattices , where @xmath4 and @xmath5 are the magnetization vectors of the two sublattices with large , fixed and unequal magnitudes . in the following , we assume that @xmath6 and @xmath7 .
the system has the biaxial symmetry , with @xmath8 being the easy axis , @xmath9 being the medium axis , and @xmath10 being the hard axis .
the magnetic field is applied in the @xmath11 plane , at an arbitrary angle in the range of @xmath12 . in order to obtain the tunnel splitting for quantum coherence
, we shall calculate the path integral : @xmath13 $ ] , where @xmath14 is the effective euclidean action for afm tunneling , @xmath15 @xmath16 , and @xmath17 is the magnetocrystalline anisotropy energy .
@xmath18 is the volume of the afm particle , and @xmath19 is the gyromagnetic ratio . @xmath20 and @xmath21 are the angular components of @xmath22 , which can also determine the direction of the nel vector .
@xmath23 , and @xmath24 , where @xmath25 is the excess spin of the afm particle due to the non - compensation of two sublattices . for this case ,
the magnetocrystalline anisotropy energy is @xmath26 , where @xmath27 and @xmath28 are the longitudinal and the transverse anisotropy coefficients satisfying @xmath29 .
therefore , the nel vector is forced to lie in the @xmath30 plane , and the fluctuations of @xmath20 about @xmath31 are small .
introducing @xmath32 , @xmath33 , @xmath34 reduces to @xmath35 where @xmath36 , @xmath37 , @xmath38 , @xmath39 , and @xmath40 . in the semiclassical limit
, the dominant contribution to the transition amplitude comes from finite action solution ( instanton ) of the classical equation of motion .
since the configuration space of this problem is a circle , only two types of instantons must be taken into account .
we use @xmath41 to denote the instanton passing through the barrier at @xmath42 , @xmath43 , and @xmath44 as the instanton passing through the barrier at @xmath30 , @xmath45 . correspondingly , there are two kinds of anti - instantons : @xmath46 and @xmath47 .
the small barrier at @xmath48 is @xmath49 , and the large barrier at @xmath45 is @xmath50 .
performing the gaussian integration over @xmath51 , we can map the spin system onto a particle moving problem in one - dimensional potential well . now
the euclidean transition amplitude of this system becomes @xmath52 \left ( \phi _ f-\phi _ i\right ) \right\ } \nonumber \\ & & \times \int d\phi \exp \left\ { -\int d\tau \left [ \frac 12{\cal m}\left ( \frac{d\phi } { d\tau } \right ) ^2+u\left ( \phi \right ) \right ] \right\ } , \eqnum{3}\end{aligned}\ ] ] where @xmath53 is the total spins of two sublattices , and @xmath54 is the sublattice spins . in eq .
( 3 ) we have taken @xmath55 , where @xmath56 is the exchange interaction between two sublattices . the effective mass and effective potential in eq . ( 3 ) are @xmath57 , \eqnum{4a } \\ u\left ( \phi \right ) & = & 2\frac{k_{\vert } v}\hbar \sin ^2\theta _ 0\left ( \sin \phi -\sin \phi _ 0\right ) ^2 .
\eqnum{4b}\end{aligned}\ ] ] the potential @xmath58 has an asymmetric twin barrier .
the degenerate ground states are given by two different types of minima of @xmath59 at @xmath60 and @xmath61 .
@xmath59 can be regarded as a superlattice consisting of two sublattices , and the energy spectrum can be obtained by applying the bloch theorem and the tight - binding approximation .
the translational symmetry is ensured by the possibility of successive @xmath62 extensions .
the periodic instanton configuration @xmath63 satisfies the equation of motion @xmath64 , where @xmath65 can be viewed as the classical energy of the pseudoparticle configuration .
then the periodic instanton @xmath41 solution for @xmath66 is @xmath67 .
sn@xmath68 is the jacobian elliptic sine function of modulus @xmath69 , where @xmath70 , @xmath71 , @xmath72 , @xmath73 , @xmath74 , and @xmath75 .
the associated euclidean action is @xmath76 = { \cal w}% _ a+2e\beta , \eqnum{5a}\ ] ] where @xmath77 .
\eqnum{5b}\ ] ] here @xmath78 , @xmath79 , and @xmath80 are the complete elliptic integral of the first , second , and third kind , respectively .
the similar method can be applied to the periodic instanton @xmath44 , and the result is @xmath81 for @xmath82 , where @xmath77 , \eqnum{6}\ ] ] and @xmath83 . while for @xmath84 , the result is @xmath85 , where @xmath86 , \eqnum{7}\ ] ] with @xmath87 , @xmath88 , @xmath89 , and @xmath90 .
now we turn to the calculation of level splittings of excited states . for a particle moving in a double - well - like potential @xmath91
, the wkb formula gives the tunnel splittings of the @xmath92th degenerate excited levels or the imaginary parts of the @xmath92th metastable levels as @xmath93,@xcite where @xmath94 .
@xmath95 is the period of the real - time oscillation in the potential well , where @xmath96 are the turning points for the particle oscillating inside the potential @xmath97 .
for the present case , we find that the level splittings for instantons @xmath41 and @xmath44 in the domain @xmath66 are @xmath98 , where @xmath99 , and @xmath100 . for @xmath101 ,
the imaginary parts of the metastable energy levels are @xmath102 , where @xmath103 , and @xmath104 .
then we discuss the low energy limit of the level splitting . by using the small oscillator approximation for energy near the bottom of the potential well , @xmath105 , @xmath106 , eqs .
( 5b ) and ( 6 ) are expanded as @xmath107 where `` @xmath108 '' for the instanton @xmath41 , and `` + '' for the instanton @xmath44 .
therefore , the tunnel splittings of @xmath92th excited levels are found to be @xmath109 ^{n+1/2}\exp \left ( -{\cal w}_{a\left ( b\right ) , 0}\right ) .
\eqnum{9}\end{aligned}\ ] ] eq .
( 8b ) shows that at finite magnetic field the wkb exponent for instanton @xmath110 , @xmath111 is smaller than that for instanton @xmath44 , @xmath112 because the barrier through which instanton @xmath44 must tunnel is higher than that for instanton @xmath41 .
it is noted that @xmath113 or @xmath114 is only the level shift induced by tunneling between degenerate excited states through a single barrier .
the periodic potential @xmath115 can be regarded as a one - dimensional superlattice with the sublattices @xmath41 and @xmath44 .
the general translation symmetry results in the energy `` band '' structure which is formally the same as that of a one - dimensional tight - binding model in solid state physics , and the energy spectrum could be determined by the bloch theorem .
the bloch states for sublattices @xmath41 and @xmath44 can be written as @xmath116 , and @xmath117 , where @xmath118 , @xmath119 , @xmath120 , and @xmath121 .
then the total wavefunction @xmath122 is a linear combination of the two bloch states , @xmath123 . including the phase contributions of the topological term
, we derive the secular equation as @xmath124 \left [ \begin{array}{l } a_a\left ( \xi \right ) \\
a_b\left ( \xi \right ) \end{array } \right ] = 0 , \eqnum{10a}\ ] ] where @xmath125 and @xmath105 .
the bloch wave vector @xmath126 in the first brillouin zone .
therefore , the eigenvalues of eq .
( 10a ) are @xmath127 where @xmath128 + 2s\left ( \frac ss\right ) \left ( \frac{k_{\bot } } j\right ) \cos \theta _
0\sin ^2\theta _ 0\right\ } .
\eqnum{11b}\ ] ] the tunnel splitting of @xmath92th excited level is @xmath129 .
we can rederive this result by calculating the transition amplitude , or by the effective hamiltonian method.@xcite at zero magnetic field , @xmath130 , @xmath131 , @xmath132 , the tunnel splitting is suppressed to zero for the half - integer excess spins , which is in good agreement with the kramers theorem .
the presence of a magnetic field perpendicular to the plane of rotation of magnetization yields an additional contribution to the topological phase , resulting constructive and destructive interferences alternatively for both integer and half - integer excess spins .
tunneling is thus periodically suppressed . at finite magnetic field ,
the tunneling spectrum of the degenerate @xmath92th excited levels depends on the parity of excess spins @xmath25 , @xmath133 \right\ } ^{1/2 } , \eqnum{12}\end{aligned}\ ] ] where `` @xmath134 '' for integer @xmath25 , and `` @xmath108 '' for half - integer @xmath25 .
the spin - parity effect and the oscillation of the tunnel splitting with the filed is shown in fig .
1 .
another important observation is that only the @xmath10 component of the magnetic field ( i.e. , along the hard axis ) can lead to the oscillation of the tunnel splitting for the highly anisotropic case . as shown in fig .
2 , even a small misalignment of the field with the @xmath10 axis can completely destroy the oscillation effect , and the oscillation is absent when the field is along the medium axis . for small @xmath135 the tunnel splitting oscillates with the field , whereas no oscillation is shown up for large @xmath1 . in the latter case , a much stronger increase of tunnel splitting with the field
is shown .
this strong dependence on the orientation of the field can be observed for ground - state resonance as well as excited - state resonance . as a result
, we conclude that _ the spin - phase interference effects depend on the orientation of the external magnetic field distinctly .
this distinct angular dependence , together with the oscillation of the tunnel splittings with the field , may provide an independent experimental test for the spin - phase interference effects in afm particles_. in fig .
3 , we plot the tunnel splittings of the ground - state level and the first excited level as a function of the magnetic field for integer excess spins .
it is clearly shown that the splitting is enhanced by quantum tunneling at the excited levels .
detailed calculations of the thermodynamic quantities of the system show that the specific heats oscillate with the magnetic field and are strongly parity dependent of excess spins at sufficiently low temperatures .
due to the topological nature of the berry phase , these spin - parity effects are independent of details such as the magnitude of excess spins , the shape of the soliton and the tunneling potential .
more recently , wernsdorfer and sessoli have measured the tunnel splittings in the molecular fe@xmath0 clusters with the help of an array of micro - squids , and have found a clear oscillation in the tunnel splittings.@xcite _ similar spin - phase interference effects observed in ferromagnetic fe_@xmath0 _ cluster are found theoretically in this letter for afm particles _ , which may bring a new insight to test the manifestation of quantum effects at a macroscopic level and the influence of quantum phases on the tunneling bahaviors of spin systems . |
high computer costs turned out to be a major problem when performing quenched qcd simulations at smaller lattice spacing @xmath5 . improving the action in order to reduce cut - off effects therefore became an important goal . while standard gluonic action has discretization errors of @xmath6 , those for wilson fermions are of @xmath7 .
sheikholeslami and wohlert @xcite proposed a modified fermion action : @xmath8 where @xmath9 is the standard wilson action and @xmath10 if the coefficient @xmath11 of the so - called clover term is chosen appropriately , this action removes all @xmath7 errors from on - shell quantities like hadron masses . a non - perturbative calculation of @xmath11 as a function of @xmath12
was done by the alpha collaboration @xcite . until now
the qcdsf collaboration has presented results for the light hadron spectrum using the improved action for two values of the coupling , @xmath13 and @xmath14 ( see @xcite ) .
these calculations have been carried out on ape quadrics computers on lattice sizes up to @xmath15 . in order to allow a more reliable estimate of the chiral limit for @xmath16 we started calculations with a hopping parameter @xmath17 closer to the critical value , @xmath18 , on a lattice of size @xmath0 . at present
we have evaluated @xmath19 configurations .
we hope for higher statistics , and so the following results should be regarded as preliminary .
we perform the quenched qcd simulations at @xmath20 . to generate a new gauge configuration we use 100 cycles consisting of a single 3-hit metropolis sweep followed by 16 over - relaxation sweeps using the @xmath21 algorithm suggested by creutz @xcite .
we use jacobi smearing @xcite for source and sink .
we chose the number of smearing steps to be @xmath22 and for the smearing hopping parameter we took @xmath23 , for which the radius of the smeared source @xmath24 is about @xmath25 which roughly corresponds to @xmath26 .
although we have calculated the propagator for both smeared and unsmeared sink , we will only use the results for smeared sink here .
the simulations are performed for three different hopping parameters , @xmath27 , @xmath3 , @xmath4 , with clover coefficient @xmath28 chosen according to @xcite . for the matrix inversion we mainly use bicgstab @xcite .
the minimal residue algorithm is used in case bicgstab does not converge . as convergence criterion we chose @xmath29 , where @xmath30 .
we found up to 4 configurations per @xmath17 which show an exceptional pattern ( see fig .
[ fig : except ] ) .
they have been excluded from the evaluation . with
separately plotted exceptional configurations.,height=222 ] [ fig : except ]
until now we have looked at the @xmath31 , @xmath32 and nucleon masses . we find good plateaus when plotting the effective mass @xmath33 $ ] , as shown in fig .
[ fig : effmass_pion ] for the @xmath31 .
[ fig : effmass_pion ] plots of the dimensionless ratio @xmath34 as a function of @xmath35 ( so - called ape plot ) were found to be rather different for improved and wilson fermions at @xmath13 @xcite . as can be seen from fig .
[ fig : ape ] , the results at @xmath16 seem to confirm that the improved results come closer to the physical value than the wilson results . [ fig : ape ] to see how masses scale as @xmath36 is changed , it has been suggested @xcite that @xmath37 should be plotted in units of the square - root of the string tension @xmath38 which has cut - off errors of @xmath6 only . in fig .
[ fig : scale ] the ratio @xmath39 is shown as a function of @xmath40 for fixed physical @xmath31 masses with @xmath41 , @xmath42 and @xmath43 . to obtain the @xmath32 mass for these values of @xmath44 we extrapolated , or interpolated , @xmath37 using the phenomenological ansatz @xcite @xmath45
on the @xmath0 lattice our current program needs @xmath46 per bicgstab iteration step on a t3e with 128 dec alpha 21164 5/375mhz microprocessors .
simulations done on a 256 node quadrics qh2 need @xmath47 for the same operations on a @xmath15 lattice . comparing the peak performance of both machines ( t3e : 96 gflops / qh2 : 12.8 gflops ) one would expect the t3e to do this job about twice as fast as the qh2 ( although the calculations on the t3e are done in double precision ) .
since the communication overhead in the bicgstab routines on the t3e is less then @xmath48 , this indicates a single processor performance problem .
while lattice gauge theory applications on the qh2 typically reach sustained speeds between @xmath49 and @xmath50 of the peak performance , our t3e code currently runs at about @xmath51 .
this might be explained by two disadvantages of the t3e for this kind of problems : the number of registers and the cache size of the dec 21164 microprocessors .
the stream buffers which improve main memory access make the code about @xmath52 faster .
the numerical calculations were performed on the cray t3e lc 512 - 512 at the hlrz jlich and the cray t3e lc 128 - 128 at the konrad - zuse - zentrum berlin .
we wish to thank the operating staff for their support .
11 b. sheikholeslami and r. wohlert , nucl .
b259 ( 1985 ) 572 . m. lscher et al , nucl .
b491 ( 1997 ) 323 , hep - lat/9609035 ; _ ibid _ b491 ( 1997 ) 344 , hep - lat/9611015 . m. gckeler , r. horsley , h. perlt , p. rakow , g. schierholz , a. schiller and p. stephenson , desy preprint desy 97 - 125 ( 1997 ) , hep - lat/9707021 .
p. stephenson et al . , this conference .
m. creutz , phys .
d36 ( 1987 ) 516 . c. r. allton et .
al . , phys .
d47 ( 1993 ) , 5128 .
h. van der vorst , siam j. sc .
stat . comp .
12 ( 1992 ) 631 .
a. frommer , v. hannemann , b. nockel , t. lippert and k. schilling , int .
j. mod .
c5 ( 1994 ) 1073 .
r. sommer , phys .
rep . 275 ( 1996 ) 1 . c. r. allton v. gimenez , l. giusti , and f. rapuano , nucl .
b489 ( 1997 ) 427 .
s. collins , nucl . phys .
b ( proc . suppl . ) 30 ( 1993 ) 393 . | continuing our investigations of quenched qcd with improved fermions we have started simulations for lattice size @xmath0 at @xmath1 .
we present first results for light hadron masses at @xmath2 , @xmath3 , and @xmath4 .
moreover we compare our initial experiences on the t3e with those for ape / quadrics systems . |
10 d. teaney , j. lauret and e. v. shuryak , phys . rev
. lett . * 86 * , 4783 ( 2001 ) ; p. huovinen , p. f. kolb , u. w. heinz , p. v. ruuskanen and s. a. voloshin , phys .
b * 503 * , 58 ( 2001 ) ; t. hirano , u. w. heinz , d. kharzeev , r. lacey and y. nara , phys .
b * 636 * , 299 ( 2006 ) ; p. romatschke and u. romatschke , phys .
lett . * 99 * , 172301 ( 2007 ) .
a. nakamura and s. sakai , phys .
lett . * 94 * , 072305 ( 2005 ) .
d. kharzeev and k. tuchin , jhep * 0809 * , 093 ( 2008 ) ; f. karsch , d. kharzeev and k. tuchin , phys .
b * 663 * , 217 ( 2008 ) .
h. b. meyer , phys .
lett . * 100 * , 162001 ( 2008 ) .
k. paech and s. pratt , phys . rev .
c * 74 * , 014901 ( 2006 ) .
j. w. chen and j. wang , arxiv:0711.4824 [ hep - ph ] .
g. boyd , j. engels , f. karsch , e. laermann , c. legeland , m. lutgemeier and b. petersson , nucl .
b * 469 * , 419 ( 1996 ) .
m. cheng _ et al .
_ , phys .
d * 77 * , 014511 ( 2008 ) .
d. fernandez - fraile and a. g. nicola , arxiv:0809.4663 [ hep - ph ] .
g. policastro , d. t. son and a. o. starinets , phys .
* 87 * , 081601 ( 2001 ) ; p. kovtun , d. t. son and a. o. starinets , phys .
lett . * 94 * , 111601 ( 2005 ) .
s. s. gubser , a. nellore , s. s. pufu and f. d. rocha , arxiv:0804.1950 [ hep - th ] ; s. s. gubser , s. s. pufu and f. d. rocha , arxiv:0806.0407 [ hep - th ] . j. w. chen , m. huang , y. h. li , e. nakano and d. l. yang , phys.lett .
b in press , arxiv:0709.3434 [ hep - ph ] .
l. p. csernai , j. i. kapusta and l. d. mclerran , phys .
lett . * 97 * , 152303 ( 2006 ) . c. ratti , m. a. thaler and w. weise , phys .
d * 73 * , 014019 ( 2006 ) .
s. k. ghosh , t. k. mukherjee , m. g. mustafa and r. ray , phys .
d * 73 * , 114007 ( 2006 ) .
g. torrieri , b. tomasik and i. mishustin , phys .
c * 77 * , 034903 ( 2008 ) ; g. torrieri and i. mishustin , arxiv:0805.0442 [ hep - ph ] .
r. j. fries , b. muller and a. schafer , phys . rev .
c * 78 * , 034913 ( 2008 ) .
z. tang , y. xu , l. ruan , g. van buren , f. wang and z. xu , arxiv:0812.1609 [ nucl - ex ] . | in the hartree approximation of cornwall - jackiw - tomboulis ( cjt ) formalism of the real scalar field theory , we show that for the strongly coupled scalar system near phase transition , the shear viscosity over entropy density is small , however , the bulk viscosity over entropy density is large .
the large bulk viscosity is related to the highly nonconformal equation of state .
it is found that the square of the sound velocity near phase transition is much smaller than the conformal value @xmath0 , and the trace anomaly at phase transition deviates far away from @xmath1 .
these results agree well with the lattice results of the complex qcd system near phase transition .
it has been believed that quark matter created at rhic behaves like a nearly perfect " fluid .
one crucial quantity to identify this property is the small ratio of shear viscosity over entropy density @xmath2 . in order to fit the elliptic flow at rhic , the hydrodynamic
simulation shows that a very small shear viscosity is required @xcite .
lattice qcd calculation confirmed that @xmath2 for the purely gluonic plasma is rather small and in the range of @xmath3 @xcite .
however , recent lattice qcd results showed that another important transport coefficient , the bulk viscosity over entropy density ratio @xmath4 rises dramatically up to the order of @xmath5 near the critical temperature @xmath6 @xcite .
the sharp peak of bulk viscosity at @xmath6 has also been observed in the linear sigma model @xcite , and the increasing tendency of @xmath4 below @xmath6 has been shown in a massless pion gas @xcite . the large bulk viscosity near phase transition is related to the nonconformal equation of state @xcite .
the correlation between large bulk viscosity and the nonconformal euqation of state has been shown in ref .
@xcite .
the bulk viscosity is related to the correlation function of the trace of the energy - momentum tensor @xmath7 : @xmath8\rangle \,.\ ] ] according to the result derived from low energy theorem , in the low frequency region , the bulk viscosity takes the form of @xcite @xmath9 here @xmath10 is the total energy density of the system , @xmath11 the negative vacuum energy density , @xmath12 is the entropy density and @xmath13 is the specific heat .
we have introduced the normalized pressure density @xmath14 and energy density @xmath15 .
the parameter @xmath16 is a scale at which the perturbation theory becomes valid .
there are several quantities to characterize the confomality of the system .
one is the square of the speed of sound @xmath17 , which is related to @xmath18 and has the form of @xmath19 at the critical temperature , the entropy density as well as energy density change most fastly with temperature , thus one expect that @xmath17 should have a minimum at @xmath6 .
another quantity to describe the conformality of the system is the trace anomaly of the energy - momentum tensor @xmath20 , i.e , @xmath21 we define @xmath22 as the `` interaction measure '' , with @xmath23 the degeneracy factor .
the conformal limit has attracted much attention in recent years , since people are trying to understand strongly interacting quark - gluon plasma by using ads / cft techniques . in conformal field theories including free field theory , @xmath24 , @xmath25 , and
the bulk viscosity @xmath26 is always zero .
lattice results show that at asymptotically high temperature , the hot quark - gluon system is close to a conformal and free ideal gas .
however , lattice results show that near deconfinement phase transition , the hot quark - gluon system deviates far away from conformality @xcite .
both @xmath18 and @xmath17 show a minimum around @xmath27 , which is much smaller than @xmath0 . for the @xmath28 pure gluon system ,
the peak value of the trace anomaly @xmath29 reads @xmath30 at @xmath31 and the corresponding `` interaction measure '' is @xmath32 , with the gluon degeneracy factor @xmath33 .
( note that here @xmath34 is the temperature corresponding to the sharp peak of @xmath35 . ) for the two - flavor case , the lattice result of the peak value of the trace anomaly @xmath36 reads @xmath37 , the corresponding interaction measure at @xmath31 is given as @xmath38 , with quark degeneracy factor @xmath39 . due to the complexity of qcd in the regime of strong coupling , results on
hot quark matter from lattice calculation and hydrodynamic simulation are still lack of analytic understanding . in recent years , the anti - de sitter / conformal field theory ( ads / cft ) correspondence has generated enormous interest in using thermal @xmath40 super - yang - mills theory ( sym ) to understand sqgp .
the shear viscosity to entropy density ratio @xmath2 is as small as @xmath41 in the strongly coupled sym plasma @xcite .
however , a conspicuous shortcoming of this approach is the conformality of sym : the square of the speed of sound @xmath17 always equals to @xmath0 and the bulk viscosity is always zero at all temperatures in this theory .
though @xmath4 at @xmath6 is non - zero for a class of black hole solutions resembling the equation of state of qcd , the magnitude is less than @xmath42 @xcite , which is too small comparing with lattice qcd results .
it has been found in ref .
@xcite that in the simplest real scalar model with @xmath43 symmetry breaking in the vacuum , @xmath44 behaves the same way as that in systems of water , helium and nitrogen in first- , second - order phase transitions and crossover @xcite . in ref .
@xcite , we have investigated the equation of state and bulk viscosity in the real scalar model , and compare the result in this simplest relativistic system with that of the complex qcd system .
the lagrangian of the real scalar field theory has the form of @xmath45with @xmath46 the mass square term and @xmath47 the interaction strength .
this theory is invariant under @xmath48 and has a @xmath49 symmetry . in the case of @xmath50 and @xmath51 ,
the vacuum at @xmath52 breaks the @xmath49 symmetry spontaneously .
the @xmath43 symmetry will be restored at finite temperature with a second - order phase transition . at finite temperature , the naive perturbative expansion in powers of the coupling constant breaks down .
a convenient resummation method is provided by the extension of cornwall - jackiw - tomboulis ( cjt ) formalism @xcite to finite temperature .
the cjt formalism is equivalent to the @xmath53-functional approach of luttinger and ward @xcite and baym @xcite . in our calculation
, we only perform the hartree approximation for the effective potential , i.e , only resum tadpole diagrams self - consistently and neglect the exchange diagrams .
the effective potential in the cjt formalism reads @xcite @xmath54 = \frac{1}{2}\int_{k}\left [ \,\ln s^{-1}(k)+s_{0}^{-1}(k)\,s(k)-1\,\right ] + \,\,v_{2}[\bar{\phi},s]+u(\bar{\phi})\,\,,\end{aligned}\]]where @xmath55 is the tree - level potential , and the 2pi potential @xmath56=\frac{3}{4 } b \left(\int_{k}\,s(k,\bar{\phi})\right)^2 $ ] in the hartree approximation .
@xmath57 is the full(tree - level ) propagator and takes the form of @xmath58 with the tree - level mass @xmath59 .
the gap equations for the condensation @xmath60 and scalar mass @xmath61 are determined by the self - consistent one- and two - point green s functions @xmath62 the entropy density is determined by taking the derivative of effective potential with respect to temperature , i.e , @xmath63 . in the symmetry
breaking case , the vacuum effective potential or the vacuum energy density is negative , i.e , @xmath64 .
the normalized energy density @xmath15 and pressure density @xmath14 can be calculated by @xmath65 with @xmath66 and @xmath67 .
as a reference for complex qcd system , we investigate the equation of state and transport properties for the real scalar field theory with @xmath43 symmetry breaking in the vacuum and 2nd order phase transition at finite temperature .
the trace anomaly @xmath35 the specific heat @xmath13 as well as bulk viscosity to entropy density ratio @xmath4 show upward cusp at @xmath6 , and their peak values increase with the increase of coupling strength
. the ratio of pressure density over energy density @xmath18 and the square of the sound velocity @xmath17 show downward cusp at @xmath6 , which is similar to the behavior of @xmath2 found in ref .
@xcite , and the cusp values decrease with the increase of coupling strength .
these cusp behaviors at phase transition resemble lattice qcd results . in fig .
[ intmeas - bulk - fig ] , we only show the trace anomaly @xmath35 and the ratio of bulk viscosity to entropy density ratio @xmath4 as functions of @xmath68 for different coupling strength @xmath47 .
( left ) and the bulk viscosity over entropy density ratio @xmath4 ( right ) as a function of temperature @xmath68 for different coupling strength @xmath47 .
_ , title="fig:",width=287 ] ( left ) and the bulk viscosity over entropy density ratio @xmath4 ( right ) as a function of temperature @xmath68 for different coupling strength @xmath47 .
_ , title="fig:",width=287 ] in the weak coupling case when @xmath69 , the cusp values of @xmath18 and @xmath17 at @xmath70 are close to the conformal value @xmath0 , both the trace anomaly @xmath35 and the bulk viscosity to entropy density ratio @xmath4 at @xmath6 are close to conformal value @xmath1 . however , the shear viscosity over entropy density ratio @xmath2 is around @xmath71 , which is huge comparing with the ads / cft limit @xmath41 .
here we have used the method in ref.@xcite to derive @xmath2 . to our surprise
, we find that when @xmath72 , the strongly coupled scalar system can reproduce all thermodynamic and transport properties of hot quark - gluon system near @xmath6 .
@xmath18 at @xmath70 is close to the lattice qcd result @xmath27 , @xmath73 ( @xmath74 for scalar system ) at @xmath6 is close to the lattice result of the peak value @xmath75 at @xmath31 .
the bulk viscosity to entropy density ratio @xmath4 at @xmath6 is around @xmath76 , which agrees well with the lattice result in ref .
@xcite .
( note , here @xmath77 correspond to @xmath78 , respectively . ) more surprisingly , the shear viscosity over entropy density ratio @xmath2 at @xmath6 is @xmath79 , which also beautifully agrees with lattice result @xmath80 in ref .
@xcite . in table
[ table - all ] , we compare our results of equation of state and transport properties in scalar field theory at @xmath6 , and corresponding results in lattice qcd calculations @xcite , the polyakov - loop nambu jona - lasinio ( pnjl ) model @xcite , and black hole duals @xcite .
[ c]|c|c|c|c|c|c| & @xmath2 & @xmath4 & @xmath22 & @xmath81 & @xmath18 + @xmath69 & @xmath82 & @xmath83 & @xmath84 & @xmath85 & @xmath86 + @xmath72 & @xmath87 & @xmath76 & @xmath88 & @xmath89 & @xmath27 + @xmath90 & @xmath80 & @xmath76 & @xmath91 & @xmath92 & @xmath27 + @xmath93 & @xmath92 & @xmath94 & @xmath95 & @xmath96 & @xmath27 + @xmath97 & @xmath98 & @xmath92 & @xmath99 & @xmath100 & @xmath101 + @xmath102 & @xmath103 & @xmath1 & @xmath1 & @xmath0 & @xmath0 + @xmath104 & @xmath103 & @xmath105 & @xmath92 & @xmath96 & @xmath92 + @xmath106 & @xmath103 & @xmath100 & @xmath92 & @xmath83 & @xmath92 + in summary , in the hartree approximation of cjt formalism , we have investigated the equation of state and transport properties of the real scalar field model with @xmath43 symmetry breaking in the vacuum and 2nd - order phase transition at finite temperature .
we have seen that at phase transition , the system either in weak coupling or strong coupling shows some common properties : 1 ) @xmath18 , the square of the speed of sound @xmath17 as well as @xmath2 exhibit downward cusp behavior at @xmath6 .
2 ) the trace anomaly @xmath35 , the specific heat @xmath13 as well as @xmath4 show upward cusp behavior at @xmath6 .
the cusp behavior is related to the biggest change rate of entropy density at @xmath6 . at weak coupling ,
the scalar system near phase transition is asymptotically conformal .
however , the shear viscosity is huge . at strong coupling ,
the scalar system near phase transition is highly non - conformal , the shear viscosity is small , but the bulk viscosity is large . we can expect that for any nonconfomal field theory , the corresponding system at strong coupling near phase transition exhibits highly nonconformality .
ads / cft method maybe can not help us understand the strongly interacting quark gluon plasma .
unexpectedly , the simplest scalar field model can do the job .
we have found that lattice qcd results on the equation of state and transport properties near phase transition can be amazingly very well described by the simplest real scalar model at strong coupling when @xmath72 .
it is urgent to include the bulk viscosity correction and the nonconformal equation of state in hydrodynamics to investigate hadronization and freeze - out processes of qgp created at heavy ion collisions @xcite , it would be also interesting to investigate how bulk viscosity affects charm radial flow @xcite at rhic . * acknowledgments .
* this work is supported by cas program `` outstanding young scientists abroad brought - in '' , cas key project kjcx3-syw - n2 , nsfc10735040 , nsfc10875134 and nsfc10675077 . |
the light - cone ( lc ) formalism @xcite provides a convenient framework for the relativistic description of hadrons in terms of quark and gluon degrees of freedom and for the application of pqcd to exclusive processes . among them , the @xmath0 mixing is a subject of considerable interest , which has been examined in many investigations , see e.g. experimental ones @xcite and theoretical ones @xcite .
some experiments have been done recently , e.g. the new kloe value of @xmath9 } { \gamma[\phi\to\eta\gamma]}=(4.9\pm0.1_{stat}\pm0.2_{syst } ) \cdot 10^{-3}$ ] @xcite leads to @xmath10 @xcite ; the bes collaboration has announced a new measured value for @xmath11 , i.e. @xmath12 } { \gamma[j/\psi\to\eta\gamma]}=4.94\pm0.40 $ ] @xcite that leads to @xmath13 .
furthermore , the newly measurements of the form factors @xmath1 and @xmath14 at asymptotic region by babar collaboration @xcite , @xmath15 gev and @xmath16 gev , will provide further constraints to the theoretical predictions . the pseudoscalar transition form factors @xmath1 and @xmath14 provide a good platform to study the @xmath6 and @xmath7 mixing effects , which have already been studied in literature by several groups @xcite .
however in these calculations , either only the leading fock - state ( e.g. ref.@xcite ) or only the asymptotic behavior of the form factors ( e.g. ref.@xcite ) have been taken into consideration to determine the mixing angle .
as has been pointed out in refs.@xcite , the mixing angle can not be reliably determined without a proper considering of the contributions from the non - valence quark states due to the fact that even though the higher fock states contributions are power suppressed in large @xmath17 region , they will give sizable contributions to small and intermediate regions .
in fact , it has been pointed out in ref.@xcite that the leading fock state contributes to @xmath18 ( @xmath19 stands for the pseudoscalar mesons ) only half and the remaining half should be come from the higher fock states .
the higher fock states contributions in small @xmath17 region can not be calculated by the perturbative qcd approach due to its nonperturbative feature .
recently , ref.@xcite had constructed a phenomenological expression for the pion - photon transition form factor to estimate the contributions beyond the leading fock state based on its asymptotic behavior at @xmath20 and @xmath4 ( @xmath17 stands for the momentum transfer in the process ) .
the predicted results for @xmath21 there agree well with the experimental data in the whole @xmath17 region . in the present paper
, we will adopt this newly developed method to estimate the higher fock states contributions of the form factors @xmath22 and @xmath14 , and then to derive a possible range for the mixing angle by comparing the predicted results with the experimental data .
as for the @xmath0 mixing , two mixing scheme are adopted in the literature , i.e. the octet - singlet mixing scheme and the quark - flavor mixing scheme .
these two schemes can be related with a proper rotation of an ideal mixing angle ( @xmath23 ) @xcite .
a dramatic simplification can be achieved by adopting the quark - flavor mixing scheme , especially , the decay constants in the quark - flavor basis simply follow the pattern of state mixing due to the ozi - rule @xcite . furthermore , by adopting the quark - flavor mixing scheme and also by carefully dealing with the higher fock states contributions , a naive discussion ( at the end of sec.ii.c ) shows that the value of @xmath24 decreases , while the value of @xmath25 increases , with the increment of the mixing angle @xmath26 , so a possible range for @xmath26 can be derived by comparing with the experimental data on the form factors @xmath22 and @xmath14 .
and then by adopting the relation between the two schemes as shown in refs.@xcite , all the three mixing angles @xmath27 , @xmath28 and @xmath29 involved in the octet - singlet scheme can be determined , where @xmath27 is mixing angle for the states and @xmath30 are mixing angles for the decay constants @xmath31 and @xmath32
. the theoretical and phenomenological considerations performed in refs.@xcite also favor the quark - flavor basis .
therefore , we will adopt the quark - flavor mixing scheme to do our calculation through out the paper .
the paper is organized as follows . in sec.ii
, we outline our techniques for determining the @xmath0 mixing angle , where expressions of the pseudoscalar transition form factors beyond the leading fock state are provided . in sec.iii , we present the numerical results for the @xmath33 and @xmath34 transition form factors , and then derive a possible range for the mixing angle @xmath26 by comparing with the present experimental data on the form factors @xmath1 and @xmath14 . some discussions of the uncertainty sources for @xmath26 determination are provided in sec.iv . the final section is reserved for summary .
in the quark - flavor basis , the two orthogonal basis states are assumed to have the following parton composition in a fock state description : @xmath35 where @xmath36 ( @xmath37 , @xmath38 ) denote the lc wavefunctions of the corresponding parton states , and the dots stand for higher fock states .
the physical meson states are related to the basis ( [ quarkflavor ] ) by an orthogonal transformation @xmath39 where @xmath26 is the mixing angle . under such scheme
, the decay constants in the quark - flavor basis simply follow the pattern of state mixing due to the ozi - rule @xcite , i.e. @xmath40 , \label{decayconstant}\end{aligned}\ ] ] where the two basic decay constants @xmath41 and @xmath42 are defined as @xmath43 with @xmath44 the factorization scale that is of order @xmath45 .
useful constraints to determine the @xmath0 mixing angle can be derived by considering the two - photon decay of @xmath6 and @xmath7 .
the decay amplitudes of @xmath46 and @xmath47 have the similar lorentz structure as that of @xmath48 @xcite : @xmath49 where the fine - structure constant @xmath50 , @xmath51 for the states @xmath52 , and @xmath53 is the corresponding decay constant .
then the decay widths for @xmath54 and @xmath55 can be written as : @xmath56 where the two photon decay widths of @xmath6 and @xmath7 and their masses can be found in pdg @xcite @xmath57 from eq.([etagamma ] ) , we obtain the correlation between @xmath58 and @xmath26 : @xmath59^{-1}\end{aligned}\ ] ] and @xmath60^{-1 } .\label{decayfs}\end{aligned}\ ] ] it shows that if knowing the range of the mixing angle @xmath26 , the ranges of the decay constants @xmath41 and @xmath42 can be determined accordingly ; and vice versa . in order to calculate the @xmath33 and @xmath34 transition form factors , we first give a brief review of the @xmath61 transition form factor @xmath21 .
a comprehensive analysis of @xmath21 has been given in ref.@xcite , in which the transverse - momentum dependence in both the hard scattering amplitude and the lc wavefunction and the contributions beyond the leading fock state have been taken into consideration .
especially , a phenomenological expression to estimate the contributions beyond the leading fock state has been constructed , which is based on the form factor s asymptotic behaviors at @xmath20 and @xmath4 . as has been pointed out in ref.@xcite that the transverse - momentum dependence in both hard - scattering amplitude and the meson wavefunction should be kept to give a consistent analysis of the form factor .
the revised lc harmonic oscillator model as suggested in ref.@xcite was employed for the lc wavefunction @xmath62 , which is constructed based on the brodsky - huang - lepage ( bhl ) prescription @xcite .
more explicitly , the lc wavefunction of @xmath63 can be written as @xmath64,\ ] ] with the normalization constant @xmath65 , the harmonic scale @xmath66 and the light quark mass @xmath67 to be determined . since the contribution from the higher helicity states ( @xmath68 ) has little contribution in comparison to the usual helicity state ( @xmath69 ) ,
so we only write the explicit term for the usual helicity state .
the spin - space wavefunction @xmath70 for the usual helicity state of pion can be written as @xcite , @xmath71 , where @xmath72 .
furthermore , one can derive a relation between @xmath67 and @xmath66 by adopting the constraints from @xmath73 and @xmath48 @xcite @xmath74 there are two basic type of contributions to @xmath21 @xcite , i.e. @xmath75 @xmath76 involves the direct annihilation of @xmath77-pair into two photons , which is the leading fock - state contribution that dominates the large @xmath17 contribution .
@xmath78 involves the case of one photon coupling ` inside ' the lc wavefunction of @xmath79 meson , i.e. strong interactions occur between the photon interactions that is related to the higher fock states contributions . by keeping the transverse - momentum dependence in both the hard scattering amplitude and the lc wavefunction
, the valence quark state transition form factor @xmath76 can be written as @xmath80\int \frac{{\rm d}^2 \mathbf{k}_\perp}{16 \pi^3 } \psi_{\pi}(x,{\bf k}_{\perp } ) t_h(x , x^{\prime},\mathbf{k}_\perp)\ ] ] where @xmath81=dxdx'\delta(1-x - x')$ ] , @xmath82 and the hard - scattering amplitude @xmath83 takes the form @xmath84 @xmath76 can be further simplified as the model wavefunction depends on @xmath85 through @xmath86 only , i.e. @xmath87 , @xmath88 as for the second type of contribution @xmath78 , it is difficult to be calculated in any @xmath17 region due to its non - perturbative nature .
one can construct a phenomenological model for @xmath78 based on the asymptotic behavior at @xmath20 and @xmath4 . as suggested in ref.@xcite ,
we assume it takes the following form : @xmath89 where @xmath90 and @xmath91 are two parameters that are determined by the asymptotic behaviors at @xmath20 , i.e. @xmath92 where the first derivative of @xmath93 over @xmath17 takes the form @xmath94_{q^2\to 0}.\ ] ] from the phenomenological formula ( [ model ] ) , it is easy to find that @xmath78 will be suppressed by @xmath95 to @xmath96 in the limit @xmath4 . as for the lc wavefunctions for the pseudoscalars @xmath97 and @xmath98
, they can be modeled as @xcite @xmath99 , \end{aligned}\ ] ] where @xmath37 , @xmath38 respectively .
they also depends on @xmath85 through @xmath86 only , i.e. @xmath100 . substituting them into the normalization ( [ decayconstant ] ) , we obtain @xmath101 -\mathrm{erf}\left[\sqrt{\frac{m_i^2}{8\beta_{i}^2 x(1-x)}}\right ] \right)dx=\frac{f_{i}}{2\sqrt{3}},\ ] ] where @xmath44 stands for the factorization scale , and following the discussion in ref.@xcite , we take its value to be @xmath102 gev . under such choice , one may safely set @xmath103 to simplify the computation , e.g. @xmath104|_{\mu_0\to\infty}\to 1 $ ] , due to the fact that the contribution from higher @xmath105 region to the wavefunction normalization drops down exponentially for the above model wavefunctions . under the quark - flavor mixing scheme ,
the @xmath33 and @xmath34 transition form factors take the following forms : @xmath106 and @xmath107 where @xmath108 and @xmath109 stand for the @xmath110 and @xmath111 form factors respectively .
similar to the pion - photon transition form factor , the pseudoscalar form factors @xmath112 ( @xmath113 and @xmath114 ) can also be divided into the following two parts , @xmath115 the leading fock - state contribution @xmath116 can be simplified as eq.([simplea ] ) , and we only need to replace @xmath117 and @xmath118 there to the present case of @xmath119 and @xmath120 , where @xmath121 for @xmath122 respectively . and similar to eq.([model ] ) , we assume the following form for the power suppressed non - leading fock - state contribution @xmath123 : @xmath124 where @xmath90 and @xmath91 are two parameters that are determined by : @xmath125 where the first derivative of @xmath123 over @xmath17 takes the form @xmath126_{q^2\to 0}.\ ] ] naively , under strict @xmath127 symmetry , one has @xmath128 , which leads to @xmath129 therefore , if @xmath26 varies within the region of @xmath130 $ ] that is most probably the case , @xmath131 will decrease with the increment of @xmath26 , while @xmath132 will increase with the increment of @xmath26 . in the next section
, we will show that under the case of the broken @xmath127 symmetry , such a fact is still exist .
and then a possible range of @xmath26 can be obtained by comparing with the experimental data on the transition form factors @xmath131 and @xmath132 .
with the help of the constraints from two photon decay amplitudes of @xmath6 and @xmath7 ( e.g. eqs.([decayfq],[decayfs ] ) ) and the experimental data on the @xmath33 and @xmath34 transition form factors @xcite , one can obtain a reasonable region for @xmath26 .
there are several parameters in the wavefunction @xmath120 to be determined . as for the constitute quark masses
, we take the conventional values : @xmath133 mev and @xmath134 mev . by studying the pion - photon transition form factor
, one may observe that the best fit of the experimental data is derived under the case of @xmath135 mev @xcite .
for the transverse parameters @xmath66 , @xmath136 and @xmath137 they are proportional to the inverse of the charged radius of the corresponding valence quark states @xcite .
the difference between them are less than @xmath138 as shown in ref.@xcite . for simplicity ,
we assume @xmath139 throughout this work .
uncertainties from the different choices of @xmath136 , @xmath137 , @xmath140 and @xmath141 will be discussed in sec.iv .
( left ) and @xmath25 ( right ) from the experimental data @xcite . the solid line stands for the average pole - mass fit with @xmath142 mev or @xmath143 mev and the shaded band shows the experimental uncertainty.,title="fig:",scaledwidth=45.0% ] ( left ) and @xmath25 ( right ) from the experimental data @xcite . the solid line stands for the average pole - mass fit with @xmath142 mev or @xmath143 mev and the shaded band shows the experimental uncertainty.,title="fig:",scaledwidth=45.0% ] as for the experimental results of @xmath131 and @xmath132 , we take the pole - mass parameter fit formula that is adopted in those experiments @xcite @xmath144 } { m_p^3}}\frac{q^2}{1+q^2/\lambda_p^2 } .\ ] ] as for the values of @xmath145 and @xmath146 : @xmath147 for cleo collaboration @xcite ; @xmath148 for tpc / two - gamma collaboration @xcite ; @xmath149 for l3 collaboration @xcite ; and @xmath150 for cello collaboration @xcite . averaging the above experimental values
, we obtain the center value for @xmath145 and @xmath146 , i.e. @xmath142 mev and @xmath143 mev .
we draw the pole - mass fit of the form factors @xmath24 and @xmath25 in fig.([etaexp ] ) , where the shaded band is derived by adopting the pole - mass fit formula ( [ polemass ] ) and by varying @xmath145 and @xmath146 within the widest possible range allowed by the above experimental results .
the shaded band ( region ) for @xmath24 and @xmath25 can be regarded as constraints to determine the @xmath6/@xmath151 wavefunctions , i.e. the values of the parameters in the wavefunctions and also the mixing angle @xmath26 should make @xmath24 and @xmath25 within the region of the shaded bands as shown in fig.([etaexp ] ) . and @xmath25 with bhl - like wavefunction . the dash - dot line , the solid line and the dashed line are for @xmath152 , @xmath153 and @xmath154 respectively .
it shows that @xmath24 decreases and @xmath25 increases with the increment of @xmath26 .
the shaded band is the region allowed by the experiments @xcite.,title="fig:",scaledwidth=45.0% ] and @xmath25 with bhl - like wavefunction .
the dash - dot line , the solid line and the dashed line are for @xmath152 , @xmath153 and @xmath154 respectively .
it shows that @xmath24 decreases and @xmath25 increases with the increment of @xmath26 .
the shaded band is the region allowed by the experiments @xcite.,title="fig:",scaledwidth=45.0% ] from eq.([finalp ] ) , we obtain @xmath155 and then with the help of eqs.([decayfq],[decayfs],[ffeta],[ffetap ] ) , it can be found that only the mixing angle @xmath26 is undetermined . as shown in fig.([etaexpbhl ] ) , one may observe that the value of @xmath24 decreases with the increment of @xmath26 , while the value of @xmath25 increases with the increment of @xmath26 , and such a fact can be used to determine a possible range for @xmath26 by comparing with the experimental data .
in fact , it can be found that the lower limit of @xmath26 is determined by @xmath24 and the upper limit of @xmath26 is determined by @xmath25 , i.e. @xmath156 furthermore , we obtain @xmath157 and @xmath158 . by using the correlation between the quark - flavor mixing scheme and the octet - singlet scheme @xcite , we obtain @xmath159 and @xmath160
in this section , we discuss the uncertainties for determining the mixing angle @xmath26 from the above approach .
first , we compare the differences caused by the different model wavefunctions , e.g. the bhl - like one and the cz ( chernyak - zhitnitsky ) -like one @xcite which is much broad . and
then , we restrict ourself to use the bhl - like model wavefunction for a detail analysis on the effects to @xmath26 determination caused by each uncertainty sources separately , where the uncertainty sources mainly include the value of @xmath136 and @xmath137 , the masses of the constitute quarks @xmath161 and @xmath38 , and the possible intrinsic charm components in @xmath6 and @xmath7 . some other even smaller uncertainty sources for the electro - magnetic transition form factors such as the gluon component in @xmath162 will not be discussed .
one typical broad wavefunction is described by the cz - like wavefunction . for convenience and simplicity , we take @xmath163 gev and @xmath164 gev , and @xmath165 . it can be found that @xmath166 gev for the case of cz - like wavefunction @xcite .
the cz - like wavefunctions for @xmath167 and @xmath114 take the form @xmath168 , \end{aligned}\ ] ] where @xmath37 , @xmath38 . following a similar procedure
, it can be found that @xmath169 through numerical calculation , it can be found that the @xmath33 and @xmath34 transition form factors from the cz - like wavefunction raise faster than the case of the bhl - like wavefunction . and the shapes of the @xmath33 and @xmath34 form factors from the bhl - like wavefunction are more close to the pole - mass parameter fits of the experimental data
. although the model is different , one may find that the value of @xmath24 decreases with the increment of @xmath26 , while the value of @xmath25 increases with the increment of @xmath26 .
so the range for @xmath26 can also be estimated , i.e. @xmath170 , which is close to the case of bhl - like wavefunction .
this shows that the @xmath0 mixing angle @xmath26 is almost model - independent in our approach .
recently , babar collaboration has measured the value of @xmath24 and @xmath25 at @xmath171 @xcite : @xmath172 gev and @xmath173 gev , and the ratio of the form factors @xmath174 . under the case of
@xmath175 $ ] , for the bhl - like wavefunction , we have @xmath176 $ ] gev and @xmath177 $ ] gev at @xmath178 , which is close to the experimental values and leads to @xmath179 . while for the cz - like wavefunction , we have @xmath180 $ ] gev and @xmath181 $ ] gev at @xmath178 , which is somewhat bigger than the experimental values and leads to @xmath182 . as a comparison , one may conclude that the asymptotic behavior of the form factors @xmath24 and @xmath25 disfavor the cz - like wavefunction but favor the asymptotic like wavefunction .
it is due to these differences that the form factors of cz - like wavefunction and the bhl - like one are affected differently by the following considered uncertainty sources .
for example , as shown in ref.@xcite , the best fit of @xmath61 form factor to the experimental data is obtained with @xmath183 mev in the case of bhl - like wavefunction , which is shifted to @xmath184 mev for the case of cz - like wavefunction . in the following , we will only take the bhl - like wavefunction as an explicit example to show the uncertainties and the case of cz - like one can be done in the similar way . for clarity , in studying of the uncertainty caused by a certain source ,
the other uncertainty sources are taken to be their center values as adopted above .
we take a wider range for @xmath67 and @xmath141 to study their effects to the mixing angle @xmath26 , e.g. @xmath186 gev and @xmath187 gev . under the present case
we adopt @xmath188 , where the value of @xmath66 varies within the region of @xmath189\;{\rm gev}$ ] according to the value of @xmath67 @xcite . from eq.([finalp ] ) , we obtain the uncertainty from the constituent quark masses @xmath190 where both @xmath191 and @xmath192 increase with the increment of @xmath67 and @xmath141 respectively . and , it can be found numerically that @xmath193 due to su(3)-symmetry breaking , there are differences among @xmath66 , @xmath136 and @xmath137 , which is smaller than @xmath195 through a light - cone quark model analysis @xcite . for clarity , we choose broader ranges @xmath196 gev and @xmath197 gev to make a discussion on how these transverse size parameters affect the mixing angle .
other wavefunction parameters are fixed by setting @xmath163 gev ( or equivalently @xmath198 gev ) and @xmath164 gev . under such condition
, we have the uncertainty from the transverse parameters @xmath136 and @xmath137 , @xmath199 where both @xmath191 and @xmath192 increase with the decrement of @xmath200 and @xmath137 , respectively . and
, it can be found numerically that @xmath201 it has been suggested that a larger intrinsic charm component might be possible to explain the abnormally large production of @xmath7 in the standard model @xcite
. however , some studies in the literature disfavor such a large portion of intrinsic charm component , e.g. @xcite and references therein .
it is has been found @xcite that the mixing between the @xmath203 state with @xmath204-@xmath205 basis is quite small , e.g. less than @xmath206 .
so for simplicity , we do not consider the mixing between @xmath203 and @xmath204-@xmath205 basis .
and then , we have @xmath207 where @xmath208 and @xmath209 corresponds to the contributions from the intrinsic charm component in @xmath6 and @xmath7 respectively , which will be calculated in the following .
the wavefunction of the intrinsic " charm component @xmath210 can be modeled as @xmath211,\end{aligned}\ ] ] where we adopt @xmath212 gev and @xmath213 gev . and @xmath214 within their possible regions ,
the following results will be slightly changed and the present conceptional results are the same . ]
the overall factor @xmath215 is determined by the wavefunction normalization similar to eq.([finalp ] ) , which shows @xmath216 where @xmath217 and @xmath218 are related through @xcite , @xmath219\ ] ] and if taking @xmath220 , we have @xmath221 $ ] .
it is noted that different from the @xmath110 and @xmath111 transition form factors , the helicity - flip amplitude that is proportional to the current quark mass can not be ignored for the present case . for the @xmath222 transition form factor ,
a direct calculation shows @xmath223\int \frac{{\rm d}^2 \mathbf{k}_\perp}{16 \pi^3 } \psi^{c}_{\eta/\eta'}(x,{\bf k}_{\perp } ) t^c_h(x , x^{\prime},\mathbf{k}_\perp).\end{aligned}\ ] ] where the hard - scattering amplitude @xmath224 that includes all the helicity states @xmath225 of @xmath226 takes the form @xcite , @xmath227 } + ( x \leftrightarrow x').\ ] ] the above formula can be further simplified by doing the integration over the azimuth angle : @xmath228 k_\perp d k_\perp , \ ] ] where @xmath229 and @xmath230 . and
@xmath25 for @xmath231 with bhl - like wavefunction .
the shaded band shows the experimental uncertainty @xcite .
the dash - dot line , the solid line , the dashed line and the triangle line are for @xmath232 , @xmath233 mev , @xmath234 mev and @xmath235 mev respectively.,title="fig:",scaledwidth=45.0% ] and @xmath25 for @xmath231 with bhl - like wavefunction .
the shaded band shows the experimental uncertainty @xcite .
the dash - dot line , the solid line , the dashed line and the triangle line are for @xmath232 , @xmath233 mev , @xmath234 mev and @xmath235 mev respectively.,title="fig:",scaledwidth=43.0% ] taking @xmath231 , we draw in fig.([charm ] ) how the value of @xmath236 affects the form factors @xmath24 and @xmath25 .
one may observe that the experimental data disfavors a larger portion of charm component as @xmath237 mev . or
inversely , it can be founded that under the condition of @xmath238 mev , the uncertainty from the possible intrinsic charm components is given by @xmath239 under the quark - flavor mixing scheme , and by carefully dealing with the higher fock states contributions , a possible range for @xmath26 can be derived by comparing with the experimental data on the form factors @xmath1 and @xmath14 due to the fact that the value of @xmath24 decreases , while the value of @xmath25 increases , with the increment of @xmath26 in the whole @xmath17 region .
it has found that the allowable range for the mixing angle @xmath26 is @xmath240 where the first error is from experimental uncertainty @xcite and the second error is from the uncertainties of the wavefunction parameters and the possible intrinsic charm component in @xmath6 and @xmath7 . it should be note that the second uncertainty is lower than the direct sum of the errors caused by each source separately , i.e. @xmath241 , which is due to the fact that these uncertainty sources are correlated to each other .
in the present paper , we have performed a light - cone pqcd analysis of the @xmath33 and @xmath34 transition form factors @xmath1 and @xmath14 involving the transverse momentum corrections , in which the @xmath0 mixing effects and the contributions beyond the leading fock state have been taken into consideration . for such purpose , we have adopted the quark - flavor mixing scheme for the @xmath6 and @xmath7 mixing and have constructed a phenomenological expression to estimate the contributions beyond the leading fock state based on its asymptotic behavior at @xmath20 and @xmath4 .
it has been found that the value of @xmath24 decreases , while the value of @xmath25 increases , with the increment of @xmath26 in the whole @xmath17 region , and then a possible range for @xmath26 can be determined by comparing with the experimental data , which is @xmath242 with the first error coming from experimental uncertainty @xcite and the second error coming from the uncertainties of the wavefunction parameters and the possible intrinsic charm component in @xmath6 and @xmath7 . a more accurate weighted average of the above mentioned value together with the seven adopted experimental values as described in ref.@xcite , and the two new experimental values , @xmath10 @xcite and @xmath243 @xcite , yields @xmath244 .
furthermore , our results show that the @xmath0 mixing angle @xmath26 depends on the different wavefunction models slightly .
however the asymptotic behavior of the form factors @xmath24 and @xmath25 disfavor the cz - like wavefunction but favor the asymptotic like wavefunction .
it has been found that the intrinsic charm component in @xmath6 and @xmath7 can not be too big , e.g. @xmath245 mev .
such a conclusion agrees with other investigations @xcite .
these results are helpful to understand other exclusive processes involving the pseudo - scales @xmath6 and @xmath7 .
brodsky , t. huang and g.p .
lepage , in _ particles and fields-2 _ , proceedings of the banff summer institute , banff , alberta , 1981 , edited by a.z .
capri and a.n .
kamal ( plenum , new york , 1983 ) , p143 ; t. huang , _ in proceedings of xxth international conference on high energy physics _ ,
madison , wisconsin , 1980 , edited by l.durand and l.g .
pondrom , aip conf.proc.no .
69(aip , new york , 1981 ) , p1000 . | the possible range of @xmath0 mixing angle is determined from the transition form factors @xmath1 and @xmath2 with the help of the present experimental data . for such purpose ,
the quark - flavor mixing scheme is adopted and the pseudoscalar transition form factors are calculated under the light - cone pqcd framework , where the transverse momentum corrections and the contributions beyond the leading fock state have been carefully taken into consideration . we construct a phenomenological expression to estimate the contributions to the form factors beyond the leading fock state based on their asymptotic behavior at @xmath3 and @xmath4 . by taking the quark - flavor mixing scheme ,
our results lead to @xmath5 , where the first error coming from experimental uncertainty and the second error coming from the uncertainties of the wavefunction parameters .
the possible intrinsic charm component in @xmath6 and @xmath7 is discussed and our present analysis also disfavors a large portion of intrinsic charm component in @xmath6 and @xmath7 , e.g. @xmath8 .
+ * pacs numbers : * 13.40.gp , 12.38.bx , 14.40.aq |
Subsets and Splits
No saved queries yet
Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.