article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
bose - einstein condensates ( becs ) with active spin degrees of freedom have proven to be a fertile ground for studying cooperative quantum many - body phenomena in condensed matter systems . following the initial realization of becs with spin-1
alkali atoms in an optical trap @xcite and description of the dynamics of the system @xcite , theoretical and experimental investigations of spinor becs have identified such diverse behavior as quantum phase transitions @xcite , domain and topological defect formation @xcite , coherent spin - mixing and amplification @xcite , and , most recently , dipolar effects @xcite . in the experiment described in @xcite , a condensate of @xmath1rb atoms restricted to the ferromagnetic @xmath2 hyperfine manifold was prepared in the @xmath3 state via the quadratic zeeman effect .
when the magnetic field was rapidly quenched , quantum fluctuations triggered the growth of unstable modes , leading to the formation of domains of magnetization with random orientations in the plane perpendicular to the spin - quantization axis .
the instability is characterized by the inverse time constant @xmath4 and corresponding length scale @xmath5 , which are set by the strength of the spin - spin interaction [ cf .
( [ eq : units ] ) ] . @xmath6 coincides with the spin - healing length , with @xmath7 for the conditions of ref
. @xcite .
the exponential growth of the amplitude of a mode is proportional to @xmath8 and is controlled by the dimensionless gain parameter @xmath9 . in a homogeneous system , an unstable mode may be characterized by a wavevector @xmath10 . neglecting the dipolar interaction , one finds @xcite the gain parameter to be @xmath11 , where @xmath12 is measured in units of @xmath13 .
although the nature of the instability at early times ( @xmath14 ) has been successfully analyzed in the case of an infinite condensate @xcite , the effect of the trap geometry on the instability and spatial distribution of the associated magnetization has not been adequately addressed .
this is a relevant question , since the experiment observed significant anisotropy in the magnetization correlation function which was attributed to the spatial anisotropy of the condensate .
it was argued that as a result of the elongated shape of the condensate a small set of discrete modes in the trapped direction , rather than a continuum of states , led to preferential modulation of the magnetization in the confined direction . in this paper
, we account rigorously for the spatial quantization effects in the development of the instability of the polar phase .
concentrating on the case of isotropic interactions within the @xmath2 manifold , we find that the correlation function [ eq . ( [ eq : gfirst ] ) ] displays behavior precisely opposite to the one observed in experiment , i.e. , displays a more prominent magnetization modulation in the unconfined direction
. the inconsistency may be due to distortions introduced by optical aberrations in the imaging device , and new experiments are underway @xcite . to be specific , we consider a condensate whose thickness in the @xmath15-direction is comparable to or smaller than the spin - healing length so that the spin dynamics is effectively constrained to the @xmath16 plane .
moreover , we assume that the condensate is infinite in the @xmath17-direction but short in the @xmath18-direction due to a confining potential .
note that the effect of a trap on single - component condensates has already been analyzed extensively @xcite .
however , the analysis of a condensate with a spinor order parameter is complicated by the fact that , in the case of isotropic interactions within the @xmath2 manifold , the @xmath19 symmetry in spin space must be respected . in particular ,
the simplest approach in which we keep the functional form of @xmath20 for an infinite system but restrict the allowed wavevectors to certain values by applying either hard - wall boundary conditions or periodic boundary conditions ( pbcs ) yields qualitatively incorrect results . in the case of hard - wall boundary conditions
, @xmath21 takes on discrete values which exclude @xmath22 , which results in @xmath23 for all allowed values of @xmath10 and contradicts the existence of a zero mode in the @xmath19-symmetric case @xcite .
meanwhile , although the presence of the zero mode is preserved by choosing pbcs in the @xmath18-direction , this results in the wrong systematics of unstable modes ( only an odd number of such modes are possible , and the maximal gain is identical for all modes with @xmath24 ) .
the errors introduced by either of these two simplifications are insignificant for wide traps ( @xmath25 ) , but grow in importance with decreasing @xmath26 . for the typical trap widths of the order of a few @xmath27 , these errors are substantial enough to affect the qualitative features of the instability development . as an illustration , we plot in fig .
[ fig : gain ] the gain parameter @xmath28 for a system with only a few modes spatially quantized in the @xmath18-direction by ( a ) pbcs and ( b ) a rigorous consideration of a harmonic trap . in the former case
there are 2 ( 3 , accounting for @xmath29 degeneracy ) modes with equal maximum gain and another mode with a nearly - degenerate maximum .
the exact degeneracy of the different maxima for @xmath24 arises because the differences in @xmath21 are compensated by different values of @xmath30 satisfying @xmath31 .
in contrast , the latter is characterized by substantial differences between the maxima of @xmath9 for different modes . in time
, this leads to an increasingly pronounced anisotropy of the magnetization correlation function because a _ single _ mode with the largest gain dominates the growth of the nascent ferromagnetic phase .
since this mode has a well - defined wavelength in the @xmath17-direction , increasingly strong oscillations between regions of positive and negative correlations are expected in the @xmath17-direction .
note that when @xmath32 only a single mode corresponding to @xmath22 is unstable and applying pbcs makes qualitatively correct predictions for the behavior of the magnetization correlation function .
indeed , this justifies our later treatment of the @xmath15-direction in a single - mode approximation , since @xmath33 . in the following section we solve the equations of motion for the amplitude of the confined @xmath34 condensate during the exponential growth stage .
on the basis of these solutions , in sec .
iii we explore the anisotropy of the resulting magnetization correlation function .
in doing so we emphasize the importance of the zero mode guaranteed by @xmath19 symmetry , both to the spectrum of instabilities and to the stability of a spinor bec in the absence of any spin - spin interaction .
( color online ) .
comparison of the gain parameter @xmath28 for a two - dimensional spinor bec in the @xmath16 plane , for different models of confinement in the @xmath18-direction .
( a ) periodic boundary conditions ( the simplest model ) and ( b ) realistic confinement in a harmonic trap potential ; @xmath35 in both cases . in ( a ) the solid and dashed curves correspond to @xmath36 and @xmath37 , respectively , and the two modes with @xmath23 at @xmath38 are each doubly - degenerate .
note that all three modes with @xmath39 have the same maximum value of gain @xmath9 . in ( b ) , curves of the same parity ( even for 1 and 3 , odd for 2 and 4 ) show anti - crossing behavior due to mode - mixing , especially evident in the latter pair .
the position of the arrow indicates the value of @xmath30 corresponding to the maximum gain ; the instability with this gain may become dominant before depletion becomes significant , see sec .
the dashed curve in ( b ) is the @xmath40 curve from ( a ) , for comparison.,width=312 ]
in the absence of magnetic fields , an interacting spin-1 bose gas is described by the hamiltonian @xcite @xmath41}\hat{\psi } + \frac{c_0}{2}\hat{n}^2 + \frac{c_2}{2}{\mathbf{\hat{f}}}^2\label{eq : fullh},\end{aligned}\ ] ] where @xmath42 is a spinor whose components @xmath43 create particles in magnetic sublevel @xmath44 , @xmath45 , @xmath46 with @xmath47 being the spin-1 matrices quantized along the @xmath17-axis , @xmath48 is the trap potential , and @xmath49 is the chemical potential .
the interaction constants are related to the @xmath50-wave scattering lengths @xmath51 by @xmath52 , and we assume that the spin - spin interaction is ferromagnetic ( @xmath53 ) as in @xmath1rb . to find the evolution of the observables of the initially polar ( @xmath3 ) state , we may impose two simplifications based on a zero - depletion approximation whose domain of validity we consider later in the paper : 1 ) keep the @xmath34 components to lowest order , i.e. , to quadratic order in the hamiltonian , and 2 ) treat @xmath54 as a classical real field @xmath55 satisfying the ordinary gross - pitaevskii equation for a scalar condensate : @xmath56}\psi_0 = 0\label{eq : scalargp}.\ ] ] with these assumptions the heisenberg equations of motion yield a closed system for @xmath57 : @xmath58here @xmath59 and , using eq .
( [ eq : scalargp ] ) , @xmath60}\label{eq : h0n}.\end{aligned}\ ] ] note that the latter form for @xmath61 guarantees that @xmath62 is an eigenstate of @xmath61 with zero eigenvalue .
furthermore , since @xmath62 satisfying eq .
( [ eq : scalargp ] ) is nodeless , the ground state energy and all other eigenvalues of @xmath61 are non - negative . as we will see later , this property is essential to respecting spin - conservation in this system in the absence of magnetic fields .
another important consequence is that a spinor bec in the absence of spin - spin interactions is stable for any homogeneous spin state , an expected yet non - trivial fact .
although this is already clear from eqs .
( [ eq : hei1]),([eq : hei2 ] ) with @xmath63 , it is worth seeing this explicitly from the multi - component gross - pitaevskii equation corresponding to eq .
( [ eq : fullh ] ) with @xmath63 : @xmath64}\psi.\ ] ] because of the ( spin- ) rotational symmetry , the solution will be of the form @xmath65 , where @xmath66 is a position - independent spinor .
this allows us to decompose small fluctuations about the mean - field as @xmath67 , where @xmath68 is along @xmath66 and @xmath69 , so that @xmath70}\delta\psi_\perp = h_0\delta\psi_\perp.\ ] ] here @xmath61 is the same hamiltonian from eq .
( [ eq : h0n ] ) .
the fact that @xmath61 is bounded from below by zero therefore guarantees that in the absence of spin interaction , i.e. , @xmath63 , the ground state energy of the bec for any position - independent spin state ( characterized by spinor @xmath66 ) is the same , and coincides with the one for a scalar bec .
returning to the linearized quantum equations of motion with @xmath71 , the density at the center of the condensate , @xmath72 , defines a characteristic inverse time constant for the instability and corresponding length @xmath73 for the spin - spin interaction
. it will be convenient to measure times and distances in terms of @xmath74 and @xmath6 , respectively ; we will use the notation @xmath75 for the dimensionless quantities .
we then arrive at the system of equations @xmath76where @xmath77 : @xmath78 the observable quantity of interest is the transverse magnetization correlation function , which we define using @xmath79 as @xmath80 where the dimensionless correlation function is given by @xmath81 and expectation values are taken with respect to the initial @xmath34 vacuum state . in the following sections we derive the complete solution for @xmath82 from which the homogeneous solution for @xmath83 follows trivially ,
then specialize to the case where the condensate has a finite extent in one direction . the solution of the coupled linear system in eqs .
( [ eq : p1]),([eq : m1 ] ) can be written as @xmath84\label{eq : p1s},\\ \hat{\phi}^\dag_{-1}({{\boldsymbol{\varrho}}},\tau ) & = \int d{{\boldsymbol{\varrho } } } ' \ \bigg [ v({{\boldsymbol{\varrho}}},{{\boldsymbol{\varrho } } } ' , \tau)\hat{\phi}_{+1}({{\boldsymbol{\varrho } } } ' , 0 ) + u^\ast({{\boldsymbol{\varrho}}},{{\boldsymbol{\varrho } } } ' , \tau)\hat{\phi}_{-1}^\dag({{\boldsymbol{\varrho } } } ' , 0)\bigg]\label{eq : m1s},\end{aligned}\ ] ] with the initial conditions @xmath85 , @xmath86 .
we may expand @xmath87 and @xmath88 in an orthonormal basis @xmath89 as @xmath90where @xmath91 , while @xmath92then eqs .
( [ eq : p1s]),([eq : m1s ] ) become @xmath93\label{eq : phip1},\\ \hat{\phi}^\dag_{-1}({{\boldsymbol{\varrho}}},\tau ) & = \sum_{\alpha,\beta}\bigg [ \varphi_\alpha({{\boldsymbol{\varrho}}})v_{\alpha\beta}(\tau)\hat{a}_{+1,\beta } + \varphi^\ast_\alpha({{\boldsymbol{\varrho } } } ) u^\ast_{\alpha\beta}(\tau)\hat{a}^\dag_{-1,\beta } \bigg]\label{eq : phim1}.\end{aligned}\ ] ] let @xmath89 be the eigenstates of @xmath94 in eq .
( [ eq : h0 ] ) with eigenvalues @xmath95 . substituting this expansion into eqs .
( [ eq : p1]),([eq : m1 ] ) and introducing the linear combinations @xmath96then gives @xmath97here @xmath98 are the matrix elements of @xmath99 .
the corresponding initial conditions are @xmath100 this is a particularly convenient formulation of the problem , since the dimensionless magnetization correlation function in eq .
( [ eq : littleg ] ) is easily shown , with the help of eqs .
( [ eq : phip1]),([eq : phim1 ] ) and the relations @xmath101 , to be @xmath102 hence we are primarily interested in computing @xmath103 . now ,
if we navely try to solve eqs .
( [ eq : z]),([eq : w ] ) by taking the time derivative of eq .
( [ eq : w ] ) and substituting eq .
( [ eq : z ] ) , we get @xmath104 where @xmath105 is non - hermitian .
thus it is desirable to apply a transformation that renders the system hermitian .
but we may note now that one mode of the system will be neither oscillating ( positive eigenvalues of @xmath106 ) nor growing ( negative eigenvalues of @xmath106 ) .
this follows from the earlier - noted existence of a mode we label as @xmath107 such that @xmath108 , which implies @xmath109 .
since the first row contains only zeros , the determinant of @xmath106 itself must be zero , which shows that zero is an eigenvalue of @xmath106 .
this is an important consequence of the @xmath19 symmetry of the system @xcite ; specifically , the infinite - wavelength excitation creating an @xmath34 pair is equivalent to a global rotation of the system , and hence does not cost any energy . by performing the similarity transformation @xmath110 for @xmath111 , we find that the solution of eqs .
( [ eq : z]),([eq : w ] ) is governed by the hermitian matrix @xmath112 let @xmath113 be the unitary matrix that diagonalizes @xmath106 , @xmath114_{\lambda\lambda ' } = e_\lambda\delta_{\lambda\lambda'}$ ] .
some of the eigenvalues @xmath115 are negative , representing the unstable modes of the system .
since the oscillatory modes for which @xmath115 is positive are quickly washed out by the exponential growth of the unstable modes , we may restrict all summations over @xmath116 to negative @xmath115 . for these
we define @xmath117 , which we call the `` gain '' in reference to the connection with parametric amplifiers in quantum optics @xcite .
then @xmath118}\label{eq : w3}.\end{aligned}\ ] ] in summary , the transverse magnetization correlation function is determined by eq .
( [ eq : gperp ] ) through the eigenstates @xmath89 of eq .
( [ eq : h0 ] ) together with eqs .
( [ eq : w1])-([eq : w3 ] ) .
note that previous results for the homogeneous case follow trivially from this formalism . if @xmath83 then @xmath119 is the solution to eq .
( [ eq : scalargp ] ) and the eigenvalues and eigenstates of @xmath94 [ eq . ( [ eq : h0 ] ) ]
are @xmath120 and @xmath121 , respectively .
since @xmath122 and @xmath106 are diagonal , @xmath113 is simply the identity matrix and the gain function is given by @xmath123eqs .
( [ eq : w1])-([eq : w3 ] ) and eq .
( [ eq : gperp ] ) then give @xmath124 where @xmath125 which is consistent with @xcite .
the formalism developed above is applicable to any trap geometry . in the case of tight confinement in one of the directions ( the @xmath15-direction , for concreteness , where @xmath126 )
, one may use a single - mode approximation for the @xmath15-dependence of eigenfunctions @xmath89 , as was done in previous works @xcite . in the following sections we take into account the variation of @xmath89 with @xmath18 and @xmath17 only , so that we may constrain the problem to the solution of differential equations in the @xmath16 plane .
to gain an idea of what differences we should expect in the anisotropic case , note that for sufficiently large times we can approximate eq .
( [ eq:2dgperp ] ) as @xmath127carrying out the sum using steepest - descent integration around the maximum gain at @xmath128 gives ( in strictly two dimensions ) @xmath129where @xmath130 is a bessel function .
the dominant contribution to the correlation function is an average over a circle of wavevectors with unit magnitude , and thus manifests as radially decaying `` oscillations '' . as a qualitative first account of finite - size effects
, we may consider quantizing the wavevectors in the @xmath18-direction as @xmath131 while keeping @xmath30 continuous to simulate a highly anisotropic condensate ; see fig .
[ fig : gain](a ) .
we can still carry out steepest - descent integration of eq .
( [ eq : g2d ] ) in the @xmath30 direction to obtain @xmath132we recover eq .
( [ eq : j0 ] ) in the limit @xmath133 by converting the sum over @xmath21 to an integral .
in contrast , for @xmath134 only a few @xmath21 are allowed , and the correlation function develops some anisotropy . as noted in the introduction , however , for any @xmath24 the gain in eq .
( [ eq : omegak ] ) reaches the maximum @xmath135 at some @xmath30 satisfying @xmath136 , which implies that several wavevectors are equally dominant in their contribution to the correlation function .
moreover , it is clear from the denominator in eq .
( [ eq : cos ] ) that the result is highly dependent on the allowed values of @xmath21 , i.e. , the exact value of @xmath137 . as we show in the following sections , both of these predictions are incorrect . to adequately account for the effect of confinement , we now consider more rigorously the spin dynamics in the presence of a harmonic potential . working in dimensionful units for a moment ,
suppose the condensate is infinite in the @xmath17-direction but has a finite extent @xmath138 in the @xmath18-direction .
we model the confining potential as @xmath139 , with the frequency determined by @xmath140 .
then in the thomas - fermi limit of vanishing kinetic contribution ( @xmath141 ) , the solution to eq .
( [ eq : scalargp ] ) is given by @xmath142}^{1/2}\label{eq : psi0}\ ] ] for @xmath143 and zero otherwise .
once @xmath144 , the exact solution vanishes exponentially on the length scale @xmath145 ( ref .
@xcite ) . correspondingly , the effective potential in the hamiltonian @xmath61 of eq .
( [ eq : h0n ] ) grows rapidly for @xmath144 .
this allows us to replace , in the limit @xmath146 , the exact boundary conditions @xmath147 for the eigenfunctions of @xmath61 with @xmath148 and simultaneously use eq .
( [ eq : psi0 ] ) for @xmath62 in eq .
( [ eq : h0n ] ) : @xmath149}.\ ] ] the divergence of the second term in the hamiltonian at @xmath150 is regularized by the boundary condition @xmath151 .
we also note that eq .
( [ eq : scalargp ] ) can be solved directly by numerical integration @xcite to give a solution that is valid for nonzero @xmath152 , but we opt for this simpler solution in order not to obscure the physics . returning to dimensionless units , eq .
( [ eq : h0 ] ) becomes @xmath153where @xmath154 and the solutions obey @xmath155 .
the value of @xmath156 controls the dynamics in the trapped case , which is highlighted by the fact that it may be written as a simple combination of the three energy scales in the system : @xmath157the eigenstates and corresponding eigenenergies of this hamiltonian are @xmath158with @xmath159 satisfying @xmath160}\varphi_n(x ) = 0\label{eq : phin}\]]and @xmath161 . if we make the transformation @xmath162then the eigenvalue problem reduces to finding @xmath163 such that @xmath164}\phi(\eta ) = 0\]]and @xmath165 .
this equation has been solved numerically ; the first 4 eigenstates ( converted back to @xmath166 ) and eigenvalues @xmath167 are shown in fig .
[ fig : phin ] and table [ tab : sech4 ] , respectively .
solutions to the square - well problem ( width @xmath168 ) obtained by neglecting the second term of eq .
( [ eq : phin ] ) are also shown .
the energy of the most relevant low - lying states differ significantly , illustrating the importance of accounting for the trap in a manner that properly preserves the zero mode .
we have already seen the resulting gain function in fig .
[ fig : gain](b ) , and it is worth noting that the presence of off - diagonal elements in @xmath106 , eq .
( [ eq : m ] ) , leads to mode - mixing and anti - crossing behavior among curves of the same parity .
such crossings appear in the nave pbc quantization .
there are remnants of the @xmath29-degeneracy from the homogeneous spectrum , especially near @xmath169 .
but most critical is the degree to which a single mode near @xmath170 is emphasized in the resulting gain function , so that we expect a greater dominance of the corresponding wavelength in oscillations of the magnetization correlation function .
first four eigenstates of eq .
( [ eq : phin ] ) ( solid lines ) , along with eigenstates of the square - well problem ( dashed).,width=307 ] .[tab : sech4]first four eigenvalues for of eq .
( [ eq : phin ] ) compared to the square - well ( sw ) problem obtained by ignoring the second term ; @xmath171 is independent of the trap width .
the bottom of the square well is taken to be at @xmath172 so that the lowest energy is zero as required by @xmath19 symmetry . [
cols="^,^,^,^,^",options="header " , ] based on the gain parameter for the unstable modes [ fig . [ fig : gain](b ) ] , we claimed in the introduction the appearance of strong oscillations with a single wavelength in the @xmath17-direction .
this is most dramatically illustrated by examining cross - sections of the correlation function at @xmath173 [ fig .
[ fig : slices](b ) ] , where it is clear that oscillations of the magnetization correlation function in the @xmath17-direction become stronger in time .
the fact that this indicates the dominance of a _ single _ mode ( contrary to that predicted by pbcs ) can be seen by plotting the positions of the first zero of @xmath174 as shown in the lower curve of fig .
[ fig : nodes ] , which asymptotically approaches the value @xmath175 .
the latter corresponds to the maximum of @xmath28 [ belonging to curve 4 in fig .
[ fig : gain](b ) ] .
we also show the first zero - crossing in the @xmath18-direction for comparison , although there is no meaningful wavelength in this direction and indeed the `` oscillations '' die out over time [ fig .
[ fig : slices](a ) ] .
this is easily explained in the homogeneous picture where the dominant mode occurs at @xmath136 . since @xmath170
, we expect the largest contribution in the @xmath18-direction to be @xmath176 , which has no oscillations .
as we show in the next section , the underlying equations of motion are valid only until @xmath177 .
meanwhile the asymptotic behavior is not realized until @xmath178 , as can be seen in fig .
[ fig : nodes ] .
this can also be predicted on the basis of the gain curves in fig .
[ fig : gain](b ) by noting that the maximum gain of the top curve is separated from the maximum gain of the curve below by approximately 0.2 ; thus @xmath179 represents roughly 3 time constants .
still , we expect even at fairly short times for the anisotropy to be visible in the full correlation function as shown in fig .
[ fig : trap ] , where we plot the correlation function @xmath180 in the @xmath16 plane . note that , since we are interested in the spin correlation , we have factored out @xmath181 .
smaller values of @xmath156 accelerate the development of anisotropy so that it may be conclusively discerned in a real experiment before depletion and nonlinear effects begin to play a significant role .
plot of ( a ) @xmath182 and ( b ) @xmath174 divided by @xmath183 .
arrows indicate the progression of curves with increasing @xmath184 .
note the different scales on the horizontal axes.,width=288 ] ( color online ) .
the position of the first node , defined by @xmath185 for the @xmath17-direction ( squares ) and @xmath186 for the @xmath18-direction ( circles ) , for the curves of fig .
[ fig : slices ] .
the dashed line indicates the asymptotic value , 1.7 , predicted for the @xmath17-direction from the point of maximum gain in fig .
[ fig : gain](b).,width=288 ] ( color online ) .
the density - normalized magnetization correlation function @xmath187 at ( a ) @xmath188 and ( b ) @xmath179 .
dark regions represent alternating negative ( red online ) and positive ( blue online ) correlations , with strongest positive correlation at the center ; the color scale is arbitrary .
large correlations were truncated to white to maximize contrast in the rest of the plot.,width=288 ] in this section we consider for how long the underlying eqs .
( [ eq : p1]),([eq : m1 ] ) , which were based on a zero - depletion approximation , are valid . neglecting thermal effects , the number of @xmath3 particles taken out of the condensate , or equivalently the number of @xmath34 pairs created , is given by @xmath189 where @xmath190rather than evaluate what turns out to be a cumbersome expression , we note from eqs .
( [ eq : lincomb ] ) and ( [ eq : gperp ] ) that @xmath191 .
the expression is well - approximated by evaluating the integrand at @xmath192 , so that the condensate fraction @xmath193 is given by @xmath194where @xmath195 is the dimensionality of the system . in the case of a quasi-2d trap , @xmath196 and @xmath197 .
the condition for the governing equations , and hence the resulting magnetization correlation function , to be a good description of the system is therefore @xmath198 . since the denominator in eq .
( [ eq : kappa ] ) represents the number of particles in a spin correlation volume , physically this condition requires the magnetization per particle in each individual domain to be small .
[ fig : depletion ] plots the depletion as a function of time , showing that the zero - depletion approximation is quite good until @xmath199 ( i.e. , @xmath200 ) for the conditions of @xcite , after which the nonlinear effects not taken into account in the equations of motion are expected to play a significant role in the dynamics .
the difference in gain @xmath201 that can be resolved in the exponential growth of the different modes before depletion renders the underlying equations invalid is indicated by the length of the arrow in fig .
[ fig : gain](b ) . the depletion function @xmath202 .
each unit of time represents approximately 16 ms .
we have chosen @xmath203 , close to experimental conditions @xcite.,width=288 ]
we have analyzed the spin dynamics of a spatially anisotropic spinor bec starting from the polar ( @xmath3 ) state , which is unstable in the absence of a high magnetic field and develops ferromagnetic domains .
we find that anisotropy should develop in the magnetization correlation function in the form of strong oscillations in the unconfined direction .
we have ignored the effects of dipole - dipole interactions , which are inherently anisotropic but are an order - of - magnitude weaker than the spin - spin interactions @xcite .
although dipolar effects can be important under certain circumstances @xcite , preliminary investigation of dipolar forces in the system described here suggests that they will not significantly alter the magnetization correlation anisotropy
. nevertheless , our findings are contrary to what has been observed experimentally @xcite , which poses the question of what is causing the observed preference for the trapped direction .
an investigation of the possible effects of optical aberration in the imaging device used in the experiment is underway , and the new experiments may shed light on the origin of the presently evident inconsistency @xcite .
we are grateful to d. m. stamper - kurn for helpful discussions .
this work was supported by nsf grant no .
dmr-0603369 and doe grant no .
de - fg02 - 08er46482 . | we theoretically investigate the effect of an anisotropic trap on the instability of the polar @xmath0 phase of a spin-1 bose - einstein condensate . by considering rigorously the spatial quantization
, we show that the growth of the nascent ferromagnetic phase at short times becomes anisotropic with stronger oscillations in the magnetization correlation function along the unconfined direction . |
nowadays the mrf prior is quite popular for solving various inverse problems in image processing in that it is a powerful tool for modeling the statistics of natural images .
image models based on mrfs , especially higher - order mrfs , have been extensively studied and applied to image processing tasks such as image denoising @xcite , deconvolution @xcite , inpainting @xcite , super - resolution @xcite , etc . due to its effectiveness
, higher - order filter - based mrf models using the framework of the field of experts ( foe ) @xcite , have gained the most attention .
they are defined by a set of linear filters and the potential function . based on the observation that responses of mean - zero linear filters typically exhibit heavy - tailed distributions @xcite on natural images , three types of potential functions have been investigated , including the student - t distribution ( st ) , generalized laplace distribution ( glp ) and gaussian scale mixtures ( gsms ) function . in recent years
several training approaches have emerged to learn the parameters of the mrf models @xcite .
table [ foelearningresults ] gives a summary of several typical methods and the corresponding average denoising psnr results based on 68 test images from berkeley database with @xmath0 gaussian noise .
@xmath1 foe & st&lap . & contrastive divergence & map , cg & 27.77@xcite + @xmath2 foe & gsms & contrastive divergence & gibbs sampling & 27.95@xcite + @xmath1 foe & gsms & persistent contrastive divergence & gibbs sampling & 28.40@xcite + @xmath1 foe & st & loss - specific(truncated optimization ) & map , gd & 28.24@xcite + @xmath1 foe & st & loss - specific(truncated optimization ) & map , lbfgs @xcite & 28.39@xcite + @xmath1 foe & st & loss - specific(implicit differentiation ) & map ,
cg & 27.86@xcite + [ foelearningresults ] existing training approaches typically fall into two main types : ( 1 ) probabilistic training using ( persistent ) contrastive divergence ( ( p)cd ) ; ( 2 ) loss - specific training .
roth and black @xcite first introduced the concept of foe and proposed an approach to learn the parameters of foe model which uses a sampling strategy and the idea of cd to estimate the expectation value over the model distribution .
et al . _
@xcite improved the performance of their previous foe model @xcite by changing ( 1 ) the potential function to gsms and ( 2 ) the inference method from map estimate to bayesian minimum mean squared error estimate ( mmse ) .
the same authors present their latest results in @xcite , where they achieve significant improvements by employing an improved learning scheme called pcd instead of previous cd .
samuel and tappen @xcite present a novel loss - specific training approach to learn mrf parameters under the framework of bi - level optimization @xcite .
they use a plain gradient - descent technique to optimize the parameters , where the essence of this learning scheme - the gradients , are calculated by using implicit differentiation technique .
domke @xcite and barbu @xcite propose two similar approaches for the training of mrf model parameters also under the framework of bi - level optimization .
their methods are some variants of standard bi - level optimization method @xcite . in the modified setting ,
the mrf model is trained in terms of results after optimization is truncated to a fixed number of iterations , _
i.e. _ , they do not solve the energy minimization problem exactly ; instead , they just run some specific optimization algorithm for a fixed number of steps . in a recent work @xcite ,
the bi - level optimization technique is employed to train a non - parametric image restoration framework based on regression tree fields ( rtf ) , resulting a new state - of - the - art .
this technique is also exploited for learning the so - called analysis sparsity priors @xcite , which is somewhat related to the foe model .
* arguments * : the loss - specific training criterion is formally expressed as the following bi - level optimization problem @xmath3 the goal of this model is to find the optimal parameters @xmath4 to minimize the loss function @xmath5 , which is called the upper - level problem in the bi - level framework .
the mrf model is defined by the energy minimization problem @xmath6 , which is called the lower - level problem .
the essential point for solving this bi - level optimization problem is to calculate the gradient of the loss function @xmath5 with respect to the parameters @xmath4 . as aforementioned
, @xcite employs the implicit differentiation technique to calculate the gradients explicitly ; in contrast , @xcite and @xcite make use of an approximation approach based on truncated optimization .
all of them use the same st - distribution as potential function ; however , the latter two approaches surprisingly obtain much better performance than the former , as can be seen in table [ foelearningresults ] . in principle , samuel and tappen
should achieve better ( at least similar ) results compared to the approximation approaches , because they use a `` full '' fitting training scheme , but actually they fail in practice .
therefore , we argue that there must exist something imperfect in their training scheme , and we believe that we will very likely achieve noticeable improvements by refining this `` full '' fitting training scheme . * contributions * : motivated by the above investigation , we think it is necessary and worthwhile to restudy the loss - specific training scheme and we expect that we can achieve significant improvements . in this paper , we do not make any modifications to the training model used in @xcite - we use exactly the same model capacity , potential function and training images .
the only difference is the training algorithm .
we exploit a refined training algorithm that we solve the lower - level problem in the loss - specific training with very high accuracy and make use of a more efficient quasi - newton s method for model parameters optimization .
we conduct a series of playback experiments and we show that the performance of loss - specific training is indeed underestimated in previous work @xcite .
we argue that the the critical reason is that they have not solved the lower - level problem to sufficient accuracy .
we also demonstrate that solving the lower - level problem with higher accuracy is indeed beneficial .
this argument about the loss - specific training scheme is the major contribution of our paper .
we further show that our trained model can obtain slight improvement by increasing the model size .
it turns out that for image denoising task , our optimized mrf ( opt - mrf ) model of size @xmath7 has achieved the best result among existing mrf - based systems and been on par with state - of - the - art methods . due to the simplicity of our model , it is easy to implement the inference algorithm on parallel computation units , _
e.g. _ , gpu .
numerical results show that our gpu - based implementation can perform image denoising in near real - time with clear state - of - the - art performance .
in this section , we firstly present the loss - specific training model .
then we consider the optimization problem from a more general point of view .
our derivation shows that the implicit differentiation technique employed in previous work @xcite is a special case of our general formulation .
our training model makes use of the bi - level optimization framework , and is conducted based on the image denoising task . for image denoising , the st - distribution based mrf model is expressed as @xmath8 this is the lower - level problem in the bi - level framework .
wherein @xmath9 is the number of filters , @xmath10 is the number of pixels in image @xmath11 , @xmath12 is an @xmath13 highly sparse matrix , which makes the convolution of the filter @xmath14 with a two - dimensional image @xmath11 equivalent to the product result of the matrix @xmath12 with the vectorization form of @xmath11 , _
i.e. _ , @xmath15 . in our training model
, we express the filter @xmath12 as a linear combination of a set of basis filters @xmath16 , _
i.e. _ , @xmath17 . besides , @xmath18 is the parameters of st - distribution for filter @xmath12 , and @xmath19 defines the trade - off between the prior term and data fitting term . @xmath20 denotes the lorentzian potential function @xmath21 , which is derived from st - distribution . the loss function @xmath22 ( upper - level problem )
is defined to measure the difference between the optimal solution of energy function and the ground - truth . in this paper
, we make use of the same loss function as in @xcite , @xmath23 , where @xmath24 is the ground - truth image and @xmath25 is the minimizer of . given the training samples @xmath26 ,
where @xmath27 and @xmath28 are the @xmath29 clean image and the associated noisy version respectively , our aim is to learn an optimal mrf parameter @xmath30 ( we group the coefficients @xmath31 and weights @xmath32 into a single vector @xmath4 ) , to minimize the overall loss function .
therefore , the learning model is formally formulated as the following bi - level optimization problem @xmath33 where @xmath34 .
we eliminate @xmath19 for simplicity , since it can be incorporated into weights @xmath35 . in this paper
, we consider the bi - level optimization problem from a general point of view . in the following derivation
we only consider the case of a single training sample for convenience , and we show how to extend the framework to multiple training samples in the end . according to the optimality condition , the solution of the lower - level problem in is given by @xmath25 , such that @xmath36 .
therefore , we can rewrite problem as following constrained optimization problem @xmath37 where @xmath38 .
now we can introduce lagrange multipliers and study the lagrange function @xmath39 where @xmath40 and @xmath41 are the lagrange multipliers associated to the inequality constraint @xmath42 and the equality constraint in , respectively . here
@xmath43 denotes the standard inner product .
taking into account the inequality constraint @xmath42 , the first order necessary condition for optimality is given by @xmath44 where @xmath45 ( { \langle}k_i^t\rho'(k_ix),p { \rangle})_{n_f \times 1}-\mu\\[1.7ex ] ( { \langle}b_j^t\rho'(k_ix ) + k_i^t{\mathcal{d}}_i b_jx , p { \rangle})_{n \times 1}\\[1.7ex ] { \sum\nolimits_{i=1}^{n_f}}\alpha_i k_i^t\rho'(k_ix ) + x- f\\[1.7ex ] \mu - \max(0,\mu - c\alpha ) \end{pmatrix}.\ ] ] wherein @xmath46 , @xmath47 , in the third formulation @xmath48 .
note that the last formulation is derived from the optimality condition for the inequality constraint @xmath42 , which is expressed as @xmath49 .
it is easy to check that these three conditions are equivalent to @xmath50 with @xmath51 to be any positive scalar and max operates coordinate - wise .
generally , we can continue to calculate the generalized jacobian of g , _ i.e. _ , the hessian matrix of lagrange function , with which we can then employ a newton s method to solve the necessary optimality system .
however , for this problem calculating the jacobian of g is computationally intensive ; thus in this paper we do not consider it and only make use of the first derivatives .
since what we are interested in is the mrf parameters @xmath52 , we can reduce unnecessary variables in . by solving for @xmath53 and @xmath11 in , and substituting them into the second and the third formulation , we arrive at the gradients of loss function with respect to parameters @xmath4 @xmath54 in
, @xmath55 denotes the hessian matrix of @xmath56 , @xmath57 in , we also eliminate the lagrange multiplier @xmath58 associated to the inequality constraint @xmath42 , as we utilize a quasi - newton s method for optimization , which can easily handle this type of box constraints .
we can see that is equivalent to the results presented in previous work @xcite using implicit differentiation
. considering the case of @xmath59 training samples , in fact it turns out that the derivatives of the overall loss function in with respect to the parameters @xmath4 are just the sum of over the training dataset .
as given by , we have collected all the necessary information to compute the required gradients , so we can now employ gradient descent based algorithms for optimization , _
e.g. _ , steepest - descent algorithm . in this paper
, we turn to a more efficient non - linear optimization method the lbfgs quasi - newton s method @xcite . in our experiments
, we will make use of the lbfgs implementation distributed by l. stewart . in our work , the third equation in is completed the l - bfgs algorithm , since this problem is smooth , to which l - bfgs is perfectly applicable .
the training algorithm is terminated when the relative change of the loss is less than a tolerance , _
e.g. _ , @xmath60 or a maximum number of iterations _
e.g. _ , @xmath61 is reached or l - bfgs can not find a feasible step to decrease the loss .
in order to demonstrate that the loss - specific training scheme was undervalued in previous work @xcite , we conducted a playback experiment using ( 1 ) the same 40 images for training and 68 images for testing ; ( 2 ) the same model capacity24 filters of size @xmath1 ; ( 3 ) the same basis ``inverse '' whitened pca @xcite , as in samuel and tappen s experiments .
we randomly sampled four @xmath62 patches from each training image , resulting in a total of 160 training samples .
we then generated the noisy versions by adding gaussian noise with standard deviation @xmath0 .
the major difference between our training experiment and previous one is the training algorithm . in our refined training scheme ,
we employed ( 1 ) our proposed algorithm to solve the lower - level problem with very high accuracy , and ( 2 ) lbfgs to optimize the model parameters , but in contrast , samuel and tappen used non - linear conjugate gradient and plain gradient descent algorithm , respectively . in our refined training algorithm , we used the normalized norm of the gradient , _
i.e. _ , @xmath63 ( @xmath59 is the pixel number of the training patch ) as the stopping criterion for solving the lower - level problem . in our training experiment , we set @xmath64 ( gray - value in range [ 0 255 ] ) , which implies a very accurate solution .
based on this training configuration , we learned 24 filters of size @xmath1 , then we applied them to image denoising task to estimate the inference performance using the same 68 test images .
finally , we got an average psnr value of 28.51db for noise level @xmath0 , which is significantly superior to previous result of 27.86db in @xcite .
we argue that the major reason lies in our refined training algorithm that we solve the lower - level problem with very high accuracy . to make this argument more clear
, we need to eliminate the possibility of training dataset , because we did not exploit exactly the same training dataset as previous work ( unfortunately we do not have their dataset in hand ) .
since the training patches were randomly selected , we could run the training experiment multiple times by using different training dataset .
finally , we found that the deviation of test psnr values based on 68 test images is within 0.02db , which is negligible .
therefore , it is clear that training dataset is not the reason for this improvement , and the only remaining reason is our refined training scheme . * the influence of @xmath65 : * to investigate the influence of the solution accuracy of the lower - level problem
@xmath65 more detailedly , we conducted a series of training and testing experiments by setting @xmath65 to different magnitudes .
based on a fixed training dataset ( 160 patches of size @xmath62 ) and 68 test images , we got the performance curves with respect to the solution accuracy @xmath65 , as shown in figure [ accuracycurve ] ( left ) . from figure [ accuracycurve ] ( left ) , we can clearly see that it is indeed the high solution accuracy that helps us to achieve the above siginificant improvement .
this finding is the main contribution of our paper .
we also make a guess how accurate samuel and tappen solve the lower - level problem according to their result and our performance curve , which is marked by a red triangle in figure [ accuracycurve ] ( left ) . the argument that higher solution accuracy of the lower - level problem is helpful is explicable , the reason is described below .
& the filter size}. it is clear that solving the lower - level problem with higher accuracy is beneficial and larger filter size can normally bring some improvement.,title="fig:",scaledwidth=30.0% ] & the filter size}. it is clear that solving the lower - level problem with higher accuracy is beneficial and larger filter size can normally bring some improvement.,title="fig:",scaledwidth=30.0% ] [ accuracycurve ] as we know , the key aspect of our approach is to calculate the gradients of the loss function with respect to the parameters @xmath4 . according to , there is a precondition to obtain accurate gradients : both the lower - level problem and the inverse matrix of hessian matrix @xmath66 must be solved with high accuracy , _
i.e. _ , we need to calculate a @xmath25 such that @xmath36 and compute @xmath67 explicitly . since the hessian matrix @xmath66 is highly sparse , we can solve the linear system @xmath68 efficiently with very high accuracy ( we use the `` backslash '' operator in matlab ) .
however , for the lower - level problem , in practice we can only solve it to finite accuracy by using certain algorithms , _
i.e. _ , @xmath63 .
if the lower - level problem is not solved to sufficient accuracy , the gradients @xmath69 are certainly inaccurate which will probably affect the training performance .
this has been demonstrated in our experiments .
therefore , for the bi - level training framework , it is necessary to solve the lower - level problem as accurately as possible , _
e.g. _ , in our training we solved it to a very high accuracy with @xmath64 .
* the influence of basis : * in our playback experiments , we used the `` inverse '' whitened pca basis to keep consistent with previous work .
however , we argue that the dct basis is a better choice , because meaningful filters should be mean - zero according to the findings in @xcite , which is guaranteed by dct basis without the constant basis vector . therefore , we will exploit the dct filters excluding the filter with uniform entries from now on . using this modified dct basis ,
we retrained our model and we got a test psnr result of 28.54db . * the influence of training dataset : * to verify whether larger training dataset is beneficial , we retrained our model by using ( 1 ) 200 samples of size @xmath70 and ( 2 ) 200 samples of size @xmath71 , which is about two times and four times larger than our previous dataset , respectively .
finally , we got a test psnr result of 28.56db for both cases .
as shown before , the influence of training dataset is marginal . * the influence of model capacity : * in above experiments
, we concentrated on the model of size @xmath1 to keep consistent with previous work .
we can also train models of different filter sizes , _ e.g. _ , @xmath2 , @xmath7 or @xmath72 , to investigate the influence of model capacity . based on the training dataset of 200 patches of size @xmath70 ,
we retrained our model with different filter size ; the training results and testing performance are summarized in figure [ accuracycurve ] ( right ) .
we can see that normally increasing the filter size can bring some improvement . however , the improvement of filter size @xmath72 is marginal compared to filter size @xmath7 , yet the former is much more time consuming .
the training time for the model with 48 filters of size @xmath7 was approximately 24 hours on a server ( intel x5675 , 3.07ghz , 24 cores ) , but in contrast , the model of size @xmath72 took about 20 days .
more importantly , the inference time of the model of size @xmath72 is certainly longer than the model of size @xmath7 , in that it involves more filters of larger size .
therefore , the model of size @xmath7 offers the best trade - off between speed and quality , and we use it for the following applications .
the learned 48 filters together with their associated weights and norms are presented in figure [ fig : dct7 ] . ) .
the first number in the bracket is the weight @xmath32 and the second one is the norm of the filter.,scaledwidth=98.0% ] [ fig : dct7 ]
an important question for a learned prior model is how well it generalizes . to evaluate this ,
we directly applied the above 48 filters of size @xmath7 trained based on image denoising task to various image restoration problems such as image deconvolution , inpainting and super - resolution , as well as denoising . due to space limitation , here we only present denoising results and the comparison to state - of - the - arts .
the other results will be shown in the final version @xcite .
we applied our opt - mrf model to image denoising problem and compared its performance with leading image denoising methods , including three state - of - the - art methods : ( 1 ) bm3d @xcite ; ( 2 ) lssc @xcite ; ( 3 ) gmm - epll @xcite along with two leading generic methods : ( 4 ) a mrf - based approach , foe @xcite ; and ( 5 ) a synthesis sparse representation based method , ksvd @xcite trained on natural image patches .
all implementations were downloaded from the corresponding authors homepages .
we conducted denoising experiments over 68 test images with various noise levels @xmath73 . to make a fair comparison , we used exactly the same noisy version of each test image for different methods and different test images
were added with distinct noise realizations .
all results were computed per image and then averaged over the test dataset .
we used l - bfgs to solve the map - based mrf model .
when is applied to various noise level @xmath74 , we need to tune the parameter @xmath19 ( empirical choice @xmath75 ) .
ours + 15 & 30.87 & 30.99 & 31.08 & * 31.27 * & 31.19 & 31.18 + 25 & 28.28 & 28.40 & 28.56 & * 28.70 * & * 28.68 * & * 28.66 * + 50 & 25.17 & 25.35 & 25.62 & * 25.72 * & * 25.67 * & * 25.70 * + [ comparison ] table [ comparison ] shows the summary of results .
it is clear that our opt - mrf model outperforms two leading generic methods and has been on par with three state - of - the - art methods for any noise level . comparing the result of our opt - mrf model with results presented in table [ foelearningresults ]
, our model has obviously achieved the best performance among all the mrf - based systems . to the best of our knowledge ,
this is the first time that a mrf model based on generic priors of natural images has achieved such clear state - of - the - art performance .
we provide image denoising examples in the final version @xcite . in additional ,
our opt - mrf model is well - suited to gpu parallel computation in that it only contains the operation of convolution .
our gpu implementation based on nvidia geforce gtx 680 accelerates the inference procedure significantly ; for a denoising task with @xmath0 , typically it takes 0.42s for image size @xmath76 , 0.30s for @xmath77 and 0.15s for @xmath78 . in table
[ runningtime ] , we show the average run time of the considered denoising methods on @xmath77 images . considering the speed and quality of our model , it is a perfect choice of the base methods in the image restoration framework recently proposed in @xcite , which leverages advantages of existing methods .
r|c|c|c|c|c|c & ksvd & foe & bm3d & lssc & epll & ours + t(s ) & 30 & 1600 & 4.3 & 700 & 99 & 12 ( * 0.3 * ) + psnr & 28.28 & 28.40 & 28.56 & 28.70 & 28.68 & 28.66 + [ runningtime ]
in this paper , we revisited the loss - specific training approach proposed by samuel and tappen in @xcite by using a refined training algorithm .
we have shown that the performance of the loss - specific training was indeed undervalued in previous work .
we argued that the major reason lies in the solution accuracy of the lower - level problem in the bi - level framework , and we have demonstrated that solving the lower - level problem with higher accuracy is beneficial .
we have shown that we can further improve the performance of the learned model a little bit by using larger filters . for image denoising task , our learned opt - mrf model of size
@xmath7 presented the best performance among existing mrf - based systems , and has already been on par with state - of - the - art denoising methods .
the performance of our opt - mrf model proves two issues : ( 1 ) the loss - specific training scheme under the framework of bi - level optimization , which is convergence guaranteed , is highly effective for parameters learning ; ( 2 ) map estimate should be still considered as one of the leading approaches in low - level vision . | it is now well known that markov random fields ( mrfs ) are particularly effective for modeling image priors in low - level vision .
recent years have seen the emergence of two main approaches for learning the parameters in mrfs : ( 1 ) probabilistic learning using sampling - based algorithms and ( 2 ) loss - specific training based on map estimate . after investigating existing training approaches
, it turns out that the performance of the loss - specific training has been significantly underestimated in existing work . in this paper , we revisit this approach and use techniques from bi - level optimization to solve it .
we show that we can get a substantial gain in the final performance by solving the lower - level problem in the bi - level framework with high accuracy using our newly proposed algorithm . as a result ,
our trained model is on par with highly specialized image denoising algorithms and clearly outperforms probabilistically trained mrf models .
our findings suggest that for the loss - specific training scheme , solving the lower - level problem with higher accuracy is beneficial .
our trained model comes along with the additional advantage , that inference is extremely efficient .
our gpu - based implementation takes less than 1s to produce state - of - the - art performance . |
fast - slow systems of ordinary differential equations ( odes ) have the general form : @xmath0 where @xmath1 , @xmath2 and @xmath3 represents the ratio of time scales . the functions @xmath4 and @xmath5 are assumed to be sufficiently smooth .
in the singular limit @xmath6 the vector field ( [ eq : fssgen ] ) becomes a differential - algebraic equation .
the algebraic constraint @xmath7 defines the critical manifold @xmath8 .
where @xmath9 is nonsingular , the implicit function theorem implies that there exists a map @xmath10 parametrizing @xmath11 as a graph .
this yields the implicitly defined vector field @xmath12 on @xmath11 called the slow flow .
+ we can change ( [ eq : fssgen ] ) to the fast time scale @xmath13 and let @xmath6 to obtain the second possible singular limit system @xmath14 we call the vector field ( [ eq : fssfss ] ) parametrized by the slow variables @xmath15 the fast subsystem or the layer equations .
the central idea of singular perturbation analysis is to use information about the fast subsystem and the slow flow to understand the full system ( [ eq : fssgen ] ) .
one of the main tools is fenichel s theorem ( see @xcite ) .
it states that for every @xmath16 sufficiently small and @xmath11 normally hyperbolic there exists a family of invariant manifolds @xmath17 for the flow ( [ eq : fssgen ] ) .
the manifolds are at a distance @xmath18 from @xmath11 and the flows on them converge to the slow flow on @xmath11 as @xmath6 .
points @xmath19 where @xmath9 is singular are referred to as fold points onto the @xmath20 coordinates may have more degenerate singularities than fold singularities at some of these points . ] .
+ beyond fenichel s theorem many other techniques have been developed .
more detailed introductions and results can be found in @xcite from a geometric viewpoint .
asymptotic methods are developed in @xcite whereas ideas from nonstandard analysis are introduced in @xcite .
while the theory is well developed for two - dimensional fast - slow systems , higher - dimensional fast - slow systems are an active area of current research . in the following
we shall focus on the fitzhugh - nagumo equation viewed as a three - dimensional fast - slow system .
the fitzhugh - nagumo equation is a simplification of the hodgin - huxley model for an electric potential of a nerve axon @xcite .
the first version was developed by fitzhugh @xcite and is a two - dimensional system of odes : @xmath21 a detailed summary of the bifurcations of ( [ eq : fh ] ) can be found in @xcite .
nagumo et al .
@xcite studied a related equation that adds a diffusion term for the conduction process of action potentials along nerves : @xmath22 where @xmath23 and @xmath24 and @xmath25 are parameters . a good introduction to the derivation and
problems associated with ( [ eq : fhn_original ] ) can be found in @xcite .
suppose we assume a traveling wave solution to ( [ eq : fhn_original ] ) and set @xmath26 and @xmath27 , where @xmath28 represents the wave speed . by the chain rule
we get @xmath29 , @xmath30 and @xmath31 .
set @xmath32 and substitute into ( [ eq : fhn_original ] ) to obtain the system : @xmath33 system is the fitzhugh - nagumo equation studied in this paper .
observe that a homoclinic orbit of ( [ eq : fhn_temp ] ) corresponds to a traveling pulse solution of ( [ eq : fhn_original ] ) .
these solutions are of special importance in neuroscience @xcite and have been analyzed using several different methods .
for example , it has been proved that ( [ eq : fhn_temp ] ) admits homoclinic orbits @xcite for small wave speeds ( `` slow waves '' ) and large wave speeds ( `` fast waves '' ) .
fast waves are stable @xcite and slow waves are unstable @xcite .
it has been shown that double - pulse homoclinic orbits @xcite are possible .
if ( [ eq : fhn_temp ] ) has two equilibrium points and heteroclinic connections exist , bifurcation from a twisted double heteroclinic connection implies the existence of multi - pulse traveling front and back waves @xcite .
these results are based on the assumption of certain parameter ranges for which we refer to the original papers .
geometric singular perturbation theory has been used successfully to analyze ( [ eq : fhn_temp ] ) . in @xcite the fast pulse is constructed using the exchange lemma @xcite .
the exchange lemma has also been used to prove the existence of a codimension two connection between fast and slow waves in ( @xmath34)-parameter space @xcite .
an extension of fenichel s theorem and melnikov s method can be employed to prove the existence of heteroclinic connections for parameter regimes of ( [ eq : fhn_temp ] ) with two fixed points @xcite .
the general theory of relaxation oscillations in fast - slow systems applies to ( [ eq : fhn_temp ] ) ( see e.g. @xcite ) as does - at least partially - the theory of canards ( see e.g. @xcite ) .
+ the equations ( [ eq : fhn_temp ] ) have been analyzed numerically by champneys , kirk , knobloch , oldeman and sneyd @xcite using the numerical bifurcation software auto @xcite .
they considered the following parameter values : @xmath35 we shall fix those values to allow comparison of our results with theirs .
hence we also write @xmath36 . changing from the fast time @xmath37 to the slow time
@xmath38 and relabeling variables @xmath39 , @xmath40 and @xmath41 we get : @xmath42 from now on we refer to ( [ eq : fhn ] ) as `` the '' fitzhugh - nagumo equation . investigating bifurcations in the ( @xmath43 ) parameter space one finds c - shaped curves of homoclinic orbits and a u - shaped curve of hopf bifurcations ; see figure [ fig : cusystem ] .
only part of the bifurcation diagram is shown in figure [ fig : cusystem ] .
there is another curve of homoclinic bifurcations on the right side of the u - shaped hopf curve . since ( [ eq : fhn ] ) has the symmetry @xmath44 we shall examine only the left side of the u - curve
the homoclinic c - curve is difficult to compute numerically by continuation methods using auto @xcite or matcont @xcite .
the computations seem infeasible for small values of @xmath45 . furthermore multi - pulse
homoclinic orbits can exist very close to single pulse ones and distinguishing between them must necessarily encounter problems with numerical precision @xcite .
the hopf curve and the bifurcations of limit cycles shown in figure [ fig : cusystem ] have been computed using matcont .
the curve of homoclinic bifurcations has been computed by a new method to be described in section [ sec : hetfull ] .
hopf bifurcations are shown in green , saddle - node of limit cycles ( snlc ) are shown in blue and gh indicates a generalized hopf ( or bautin ) bifurcation .
the arrows indicate the side on which periodic orbits are generated at the hopf bifurcation .
the red curve shows ( possible ) homoclinic orbits ; in fact , homoclinic orbits only exist to the left of the two black dots ( see section [ sec : hetfull ] ) .
only part of the parameter space is shown because of the symmetry ( [ eq : fhnsym ] ) .
the homoclinic curve has been thickened to indicate that multipulse homoclinic orbits exist very close to single pulse ones ( see @xcite).,scaledwidth=85.0% ] since the bifurcation structure shown in figure [ fig : cusystem ] was also observed for other excitable systems , champneys et al .
@xcite introduced the term cu - system .
bifurcation analysis from the viewpoint of geometric singular perturbation theory has been carried out for examples with one fast and two slow variables @xcite .
since the fitzhugh - nagumo equation has one slow and two fast variables , the situation is quite different and new techniques have to be developed . our main goal is to show that many features of the complicated 2-parameter bifurcation diagram shown in figure [ fig : cusystem ] can be derived with a combination of techniques from singular perturbation theory , bifurcation theory and robust numerical methods .
we accurately locate where the system has canards and determine the orbit structure of the homoclinic and periodic orbits associated to the c - shaped and u - shaped bifurcation curves , without computing the canards themselves .
we demonstrate that the basic cu - structure of the system can be computed with elementary methods that do not use continuation methods based on collocation .
the analysis of the slow and fast subsystems yields a `` singular bifurcation diagram '' to which the basic cu structure in figure [ fig : cusystem ] converges as @xmath6 .
+ _ remark : _ we have also investigated the termination mechanism of the c - shaped homoclinic curve described in @xcite .
champneys et al . observed that the homoclinic curve does not reach the u - shaped hopf curve but turns around and folds back close to itself .
we compute accurate approximations of the homoclinic orbits for smaller values @xmath16 than seems possible with auto in this region .
one aspect of our analysis is a new algorithm for computing invariant slow manifolds of saddle type in the full system .
this work will be described elsewhere .
the first step in our analysis is to investigate the slow and fast subsystems separately .
let @xmath6 in ( [ eq : fhn ] ) ; this yields two algebraic constraints that define the critical manifold : @xmath46 therefore @xmath11 is a cubic curve in the coordinate plane @xmath47 .
the parameter @xmath48 moves the cubic up and down inside this plane .
the critical points of the cubic are solutions of @xmath49 and are given by : @xmath50 the points @xmath51 are fold points with @xmath52 since @xmath11 is a cubic polynomial with distinct critical points .
the fold points divide @xmath11 into three segments @xmath53 we denote the associated slow manifolds by @xmath54 , @xmath55 and @xmath56 .
there are two possibilities to obtain the slow flow .
one way is to solve @xmath57 for @xmath58 and substitute the result into the equation @xmath59 . alternatively differentiating @xmath60 implicitly with respect to @xmath38 yields
@xmath61 and therefore @xmath62 one can view this as a projection of the slow flow , which is constrained to the critical manifold in @xmath63 , onto the @xmath58-axis .
observe that the slow flow is singular at the fold points .
direct computation shows that the fixed point problem @xmath64 has only a single real solution .
this implies that the critical manifold intersects the diagonal @xmath65 only in a single point @xmath66 which is the unique equilibrium of the slow flow ( [ eq : sf ] ) .
observe that @xmath67 is also the unique equilibrium of the full system ( [ eq : fhn ] ) and depends on @xmath48 .
increasing @xmath48 moves the equilibrium from left to right on the critical manifold . the easiest practical way to determine
the direction of the slow flow on @xmath11 is to look at the sign of @xmath68 .
the situation is illustrated in figure [ fig : slowflow ] .
+ , scaledwidth=80.0% ] we are interested in the bifurcations of the slow flow depending on the parameter @xmath48 .
the bifurcations occur when @xmath66 passes through the fold points .
the values of @xmath48 can simply be found by solving the equations @xmath49 and @xmath69 simultaneously .
the result is : @xmath70 where the subscripts indicate the fold point at which each equilibrium is located .
+ the singular time - rescaling @xmath71 of the slow flow yields the desingularized slow flow @xmath72 time is reversed by this rescaling on @xmath73 and @xmath74 since @xmath75 and @xmath76 is negative on these branches . the desingularized slow flow ( [ eq : dssf ] ) is smooth and has no bifurcations as @xmath48 is varied .
the key component of the fast - slow analysis for the fitzhugh - nagumo equation is the two - dimensional fast subsystem @xmath77 where @xmath78 , @xmath79 are parameters and @xmath15 is fixed .
since @xmath15 and @xmath48 have the same effect as bifurcation parameters we set @xmath80 .
we consider several fixed y - values and the effect of varying @xmath48 ( cf .
section [ sec : hetfull ] ) in each case .
there are either one , two or three equilibrium points for ( [ eq : fss ] ) .
equilibrium points satisfy @xmath47 and lie on the critical manifold , i.e. we have to solve @xmath81 for @xmath58 .
we find that there are three equilibria for approximately @xmath82 , two equilibria on the boundary of this @xmath48 interval and one equilibrium otherwise .
the jacobian of ( [ eq : fss ] ) at an equilibrium is @xmath83 direct calculation yields that for @xmath84 $ ] the single equilibrium is a saddle . in the case of three equilibria
, we have a source that lies between two saddles .
note that this also describes the stability of the three branches of the critical manifold @xmath73 , @xmath85 and @xmath74 . for @xmath75
the matrix @xmath86 is singular of rank 1 if and only if @xmath87 which occurs for the fold points @xmath51 .
hence the equilibria of the fast subsystem undergo a fold ( or saddle - node ) bifurcation once they approach the fold points of the critical manifold .
this happens for parameter values @xmath88 and @xmath89 .
note that by symmetry we can reduce to studying a single fold point . in the limit @xmath90 (
corresponding to the case of a `` standing wave '' ) the saddle - node bifurcation point becomes more degenerate with @xmath91 nilpotent .
+ our next goal is to investigate global bifurcations of ( [ eq : fss ] ) ; we start with homoclinic orbits . for @xmath90
it is easy to see that ( [ eq : fss ] ) is a hamiltonian system : @xmath92 with hamiltonian function @xmath93 we will use this hamiltonian formulation later on to describe the geometry of homoclinic orbits for slow wave speeds .
assume that @xmath94 is chosen so that ( [ eq : fss2 ] ) has a homoclinic orbit @xmath95 .
we are interested in perturbations with @xmath96 and note that in this case the divergence of ( [ eq : fss ] ) is @xmath28 .
hence the vector field is area expanding everywhere .
the homoclinic orbit breaks for @xmath75 and no periodic orbits are created .
note that this scenario does not apply to the full three - dimensional system as the equilibrium @xmath97 has a pair of complex conjugate eigenvalues so that a shilnikov scenario can occur .
this illustrates that the singular limit can be used to help locate homoclinic orbits of the full system , but that some characteristics of these orbits change in the singular limit .
+ we are interested next in finding curves in @xmath98-parameter space that represent heteroclinic connections of the fast subsystem . the main motivation is the decomposition of trajectories in the full system into slow and fast segments . concatenating fast heteroclinic segments and
slow flow segments can yield homoclinic orbits of the full system @xcite .
we describe a numerical strategy to detect heteroclinic connections in the fast subsystem and continue them in parameter space .
suppose that @xmath99 so that ( [ eq : fss ] ) has three hyperbolic equilibrium points @xmath100 , @xmath101 and @xmath102
. we denote by @xmath103 the unstable and by @xmath104 the stable manifold of @xmath100 .
the same notation is also used for @xmath102 and tangent spaces to @xmath105 and @xmath106 are denoted by @xmath107 and @xmath108 . recall that @xmath101 is a source and
shall not be of interest to us for now .
define the cross section @xmath109 by @xmath110 we use forward integration of initial conditions in @xmath111 and backward integration of initial conditions in @xmath112 to obtain trajectories @xmath113 and @xmath114 respectively .
we calculate their intersection with @xmath109 and define @xmath115 we compute the functions @xmath116 and @xmath117 for different parameter values of @xmath98 numerically .
heteroclinic connections occur at zeros of the function @xmath118 once we find a parameter pair @xmath119 such that @xmath120 , these parameters can be continued along a curve of heteroclinic connections in @xmath98 parameter space by solving the root - finding problem @xmath121 for either @xmath122 or @xmath123 fixed and small .
we use this method later for different fixed values of @xmath15 to compute heteroclinic connections in the fast subsystem in @xmath124 parameter space .
the results of these computations are illustrated in figure [ fig : het1 ] .
there are two distinct branches in figure [ fig : het1 ] .
the branches are asymptotic to @xmath88 and @xmath89 and approximately form a `` @xmath125 '' . from figure [ fig : het1 ] we conjecture that there exists a double heteroclinic orbit for @xmath126 .
+ ) in parameter space.,scaledwidth=50.0% ] _ remarks _ : if we fix @xmath127 our initial change of variable becomes @xmath128 and our results for heteroclinic connections are for the fitzhugh - nagumo equation without an applied current .
in this situation it has been shown that the heteroclinic connections of the fast subsystem can be used to prove the existence of homoclinic orbits to the unique saddle equilibrium @xmath129 ( cf .
note that the existence of the heteroclinics in the fast subsystem was proved in a special case analytically @xcite but figure [ fig : het1 ] is - to the best of our knowledge - the first explicit computation of where fast subsystem heteroclinics are located .
the paper @xcite develops a method for finding heteroclinic connections by the same basic approach we used , i.e. defining a codimension one hyperplane @xmath130 that separates equilibrium points .
+ figure [ fig : het1 ] suggests that there exists a double heteroclinic connection for @xmath90 .
observe that the hamiltonian in our case is @xmath131 where the function @xmath132 is : @xmath133 the solution curves of ( [ eq : fss2 ] ) are given by @xmath134 .
the structure of the solution curves entails symmetry under reflection about the @xmath58-axis .
suppose @xmath135 $ ] and recall that we denoted the two saddle points of ( [ eq : fss ] ) by @xmath100 and @xmath102 and that their location depends on @xmath94 .
therefore , we conclude that the two saddles @xmath100 and @xmath102 must have a heteroclinic connection if they lie on the same energy level , i.e. they satisfy @xmath136 .
this equation can be solved numerically to very high accuracy .
[ prop : doublehet ] the fast subsystem of the fitzhugh - nagumo equation for @xmath90 has a double heteroclinic connection for @xmath137 . given a particular value @xmath138 there exists a double heteroclinic connection for @xmath139 in the fast subsystem lying in the plane @xmath138 . from continuation of periodic orbits in the full system - to be described in section [ sec : hopf ] - we observe that near the u - shaped curve of hopf bifurcations the @xmath140-coordinate is a faster variable than @xmath58 .
in particular , the small periodic orbits generated in the hopf bifurcation lie almost in the plane @xmath47 . hence to analyze this region we set @xmath141 to transform the fitzhugh - nagumo equation ( [ eq : fhn ] ) into a system with 2 slow and 1 fast variable : @xmath142 note that corresponds to the fitzhugh - nagumo equation in the form ( cf .
): @xmath143 therefore the transformation @xmath141 can be viewed as a rescaling of the diffusion strength by @xmath144 .
we introduce a new independent small parameter @xmath145 and then let @xmath146 .
this assumes that @xmath18 terms do not vanish in this limit , yielding the diffusion free system .
then the slow manifold @xmath147 of ( [ eq : fhnscale ] ) is : @xmath148 [ prop:2slow1fast ] following time rescaling by @xmath28 , the slow flow of system on @xmath147 in the variables @xmath149 is given by @xmath150 in the variables @xmath151 the vector field becomes @xmath152 _ remark : _ the reduction to equations - suggests that is a three time - scale system .
note however that is not given in the three time - scale form @xmath153 for @xmath154 and @xmath155 @xmath156 .
the time - scale separation in - results from the singular @xmath157 dependence of the critical manifold @xmath147 ; see .
( of proposition [ prop:2slow1fast ] ) use the defining equation for the slow manifold ( [ eq : sm2d ] ) and substitute it into @xmath158 . a rescaling of time by @xmath159 under the assumption that @xmath75 yields the result ( [ eq : red1 ] ) . to derive ( [ eq : red2 ] )
differentiate the defining equation of @xmath147 with respect to time : @xmath160 the equations @xmath161 and @xmath162 yield the equations ( [ eq : red2 ] ) . before we start with the analysis of ( [ eq : red1 ] ) we note that detailed bifurcation calculations for exist .
for example , rocsoreanu et al .
@xcite give a detailed overview on the fitzhugh equation and collect many relevant references .
therefore we shall only state the relevant bifurcation results and focus on the fast - slow structure and canards .
equation has a critical manifold given by @xmath163 which coincides with the critical manifold of the full fitzhugh - nagumo system ( [ eq : fhn ] ) .
formally it is located in @xmath164 but we still denote it by @xmath11 . recall that the fold points are located at @xmath50 also recall that the y - nullcline passes through the fold points at : @xmath70 we easily find that supercritical hopf bifurcations are located at the values @xmath165 for the case @xmath166 we get @xmath167 and @xmath168 .
the periodic orbits generated in the hopf bifurcations exist for @xmath169 .
observe also that @xmath170 ; so the hopf bifurcations of ( [ eq : red1 ] ) coincide in the singular limit with the fold bifurcations in the one - dimensional slow flow ( [ eq : sf ] ) .
we are also interested in canards in the system and calculate a first order asymptotic expansion for the location of the maximal canard in ( [ eq : red1 ] ) following @xcite ; recall that trajectories lying in the intersection of attracting and repelling slow manifolds are called maximal canards .
we restrict to canards near the fold point @xmath171 . near the fold point
@xmath171 the maximal canard in @xmath172 parameter space is given by : @xmath173 let @xmath174 and consider the shifts @xmath175 to translate the equilibrium of ( [ eq : red1 ] ) to the origin when @xmath127 .
this gives @xmath176 now apply theorem 3.1 in @xcite to find that the maximal canard of ( [ eq : red1sf ] ) is given by : @xmath177 shifting the parameter @xmath48
back to the original coordinates yields the result .
if we substitute @xmath166 in the previous asymptotic result and neglect terms of order @xmath178 then the maximal canard is predicted to occur for @xmath179 which is right after the first supercritical hopf bifurcation at @xmath180 .
therefore we expect that there exist canard orbits evolving along the middle branch of the critical manifold @xmath181 in the full fitzhugh - nagumo equation .
maximal canards are part of a process generally referred to as canard explosion @xcite . in this situation
the small periodic orbits generated in the hopf bifurcation at @xmath182 undergo a transition to relaxation oscillations within a very small interval in parameter space .
a variational integral determines whether the canards are stable @xcite .
[ prop : stablecanards ] the canard cycles generated near the maximal canard point in parameter space for equation ( [ eq : red1 ] ) are stable .
consider the differential equation ( [ eq : red1 ] ) in its transformed form ( [ eq : red1sf ] ) .
obviously this will not affect the stability analysis of any limit cycles .
let @xmath183 and @xmath184 denote the two smallest @xmath58-coordinates of the intersection between @xmath185 and the line @xmath186 .
geometrically @xmath100 represents a point on the left branch and @xmath101 a point on the middle branch of the critical manifold @xmath187 .
theorem 3.4 in @xcite tells us that the canards are stable cycles if the function @xmath188 is negative for all values @xmath189 $ ] where @xmath190 is the second fold point of @xmath187 besides @xmath191 . in our case
we have @xmath192 with @xmath193 and @xmath194 $ ] .
figure [ fig : canard_stability ] shows a numerical plot of the function @xmath195 for the relevant values of @xmath196 which confirms the required result .
+ for @xmath189 $ ] . ] _ remark : _ we have computed an explicit algebraic expression for @xmath197 with a computer algebra system .
this expression yields @xmath198 for @xmath189 $ ] , confirming that @xmath195 is decreasing .
as long as we stay on the critical manifold @xmath11 of the full system , the analysis of the bifurcations and geometry of ( [ eq : red1 ] ) give good approximations to the dynamics of the fitzhugh - nagumo equation because the rescaling @xmath199 leaves the plane @xmath47 invariant .
next we use the dynamics of the @xmath200-coordinate in system ( [ eq : red2 ] ) to obtain better insight into the dynamics when @xmath201 .
the critical manifold @xmath202 of ( [ eq : red2 ] ) is : @xmath203 we are interested in the geometry of the periodic orbits shown in figure [ fig : subsysgeo ] that emerge from the hopf bifurcation at @xmath204 .
observe that the amplitude of the orbits in the @xmath58 direction is much larger that than in the @xmath140-direction .
therefore we predict only a single small excursion in the @xmath140 direction for @xmath48 slightly larger than @xmath204 as shown in figures [ fig : sub1 ] and [ fig : sub3 ] .
the wave speed changes the amplitude of this @xmath140 excursion with a smaller wave speed implying a larger excursion .
hence equation ( [ eq : red1 ] ) is expected to be a very good approximation for periodic orbits in the fitzhugh - nagumo equation with fast wave speeds .
furthermore the periodic orbits show two @xmath140 excursions in the relaxation regime after the canard explosion ; see figure [ fig : sub2 ] .
the characteristic polynomial of the linearization of the fitzhugh - nagumo equation ( [ eq : fhn ] ) at its unique equilibrium point is @xmath205 denoting @xmath206 , a necessary condition for @xmath207 to have pure imaginary roots is that @xmath208 .
the solutions of this equation can be expressed parametrically as a curve @xmath209 : @xmath210 [ prop : hopf ] in the singular limit @xmath6 the u - shaped bifurcation curves of the fitzhugh - nagumo equation have vertical asymptotes given by the points @xmath211 and @xmath212 and a horizontal asymptote given by @xmath213\quad\text{and}\quad s=0\}$ ] .
note that at @xmath214 the equilibrium point passes through the two fold points .
the expression for @xmath215 in ( [ eq : solhopf ] ) is positive when @xmath216 . for values of @xmath66 between the roots of @xmath217 , @xmath218 in ( [ eq : solhopf ] ) as @xmath219 .
the values of @xmath220 and @xmath221 in the proposition are approximations to the value of @xmath222 in ( [ eq : solhopf ] ) at the roots of @xmath217 . as @xmath219 , solutions of the equation @xmath223 in ( [ eq : solhopf ] ) yield values of @xmath66 that tend to one of the two roots of @xmath224
. the result follows .
the analysis of the slow subsystems ( [ eq : red1 ] ) and ( [ eq : red2 ] ) gives a conjecture about the shape of the periodic orbits in the fitzhugh - nagumo equation .
consider the parameter regime close to a hopf bifurcation point . from ( [ eq : red1 ] ) we expect one part of the small periodic orbits generated in the hopf bifurcation to lie close to the slow manifolds @xmath54 and @xmath55 .
using the results about equation ( [ eq : red2 ] ) we anticipate the second part to consist of an excursion in the @xmath140 direction whose length is governed by the wave speed @xmath28 .
figure [ fig : porbits ] shows a numerical continuation in matcont @xcite of the periodic orbits generated in a hopf bifurcation and confirms the singular limit analysis for small amplitude orbits .
+ furthermore we observe from comparison of the @xmath58 and @xmath140 coordinates of the periodic orbits in figure [ fig : po2 ] that orbits tend to lie close to the plane defined by @xmath47 .
more precisely , the @xmath140 diameter of the periodic orbits is observed to be @xmath18 in this case .
this indicates that the rescaling of section [ sec:2slow1fast ] can help to describe the system close to the u - shaped hopf curve .
note that it is difficult to check whether this observation of an @xmath18-diameter in the @xmath140-coordinate persists for values of @xmath225 since numerical continuation of canard - type periodic orbits is difficult to use for smaller @xmath16 .
+ in contrast to this , it is easily possible to compute the u - shaped hopf curve using numerical continuation for very small values of @xmath16 .
we have used this possibility to track two generalized hopf bifurcation points in three parameters @xmath226 .
the u - shaped hopf curve has been computed by numerical continuation for a mesh of parameter values for @xmath16 between @xmath227 and @xmath228 using matcont @xcite .
the two generalized hopf points @xmath229 on the left half of the u - curve were detected as codimension two points during each continuation run .
the results of this `` three - parameter continuation '' are shown in figure [ fig : gh ] .
+ the two generalized hopf points depend on @xmath16 and we find that their singular limits in @xmath124-parameter space are approximately : @xmath230 we have not found a way to recover these special points from the fast - slow decomposition of the system .
this suggests that codimension two bifurcations are generally diffcult to recover from the singular limit of fast - slow systems .
+ furthermore the hopf bifurcations for the full system on the left half of the u - curve are subcritical between @xmath231 and @xmath232 and supercritical otherwise .
for the transformed system with two slow and one fast variable we observed that in the singular limit for @xmath233 the hopf bifurcation is supercritical . in the case of @xmath166 the periodic orbits for and
exist in overlapping regions for the parameter @xmath48 between the @xmath48-values of @xmath234 and @xmath235 .
this result indicates that can be used to describe periodic orbits that will interact with the homoclinic c - curve .
in the following discussion we refer to `` the '' c - shaped curve of homoclinic bifurcations of system as the parameters yielding a `` single - pulse '' homoclinic orbit .
the literature as described in section [ sec : fhn ] shows that close to single - pulse homoclinic orbits we can expect multi - pulse homoclinic orbits that return close to the equilibrium point multiple times .
since the separation of slow manifolds @xmath236 is exponentially small , homoclinic orbits of different types will always occur in exponentially thin bundles in parameter space .
values of @xmath237 are small enough that the parameter region containing all the homoclinic orbits will be indistinguishable numerically from `` the '' c - curve that we locate .
+ the history of proofs of the existence of homoclinic orbits in the fitzhugh - nagumo equation is quite extensive .
the main step in their construction is the existence of a `` singular '' homoclinic orbit @xmath238 .
we consider the case when the fast subsystem has three equilibrium points which we denote by @xmath239 , @xmath240 and @xmath241 . recall that @xmath100 coincides with the unique equilibrium @xmath67 of the full system for @xmath242 . a singular homoclinic orbit is always constructed by first following the unstable manifold of @xmath100 in the fast subsystem given by @xmath243 .
+ for equation ( [ eq : fss2 ] ) with @xmath243.,scaledwidth=80.0% ] first assume that @xmath90 . in this case
the hamiltonian structure - see section [ sec : hetsub ] and equation ( [ eq : fss2 ] ) - can be used to show the existence of a singular homoclinic orbit .
figure [ fig : homs0 ] shows level curves @xmath244 for various values of @xmath48 .
the double heteroclinic connection can be calculated directly using proposition [ prop : doublehet ] and solving @xmath245 for @xmath48 .
there exists a singular double heteroclinic connection in the fitzhugh - nagumo equation for @xmath90 and @xmath246 .
techniques developed in @xcite show that the singular homoclinic orbits existing for @xmath90 and @xmath247 must persist for perturbations of small positive wave speed and sufficiently small @xmath16 .
these orbits are associated to the lower branch of the c - curve .
the expected geometry of the orbits is indicated by their shape in the singular limit shown in figure [ fig : homs0 ] .
the double heteroclinic connection is the boundary case between the upper and lower half of the c - curve .
it remains to analyze the singular limit for the upper half . in this case , a singular homoclinic orbit is again formed by following the unstable manifold of @xmath100 when it coincides with the equilibrium @xmath67 but now we check whether it forms a heteroclinic orbit with the stable manifold of @xmath102 .
then we follow the slow flow on @xmath74 and return on a heteroclinic connection to @xmath73 for a different y - coordinate with @xmath248 and @xmath249 . from there
we connect back via the slow flow . using the numerical method described in section [ sec : hetsub ] we first set @xmath243 ; note that the location of @xmath97 depends on the value of the parameter @xmath48 .
the task is to check when the system @xmath250 has heteroclinic orbits from @xmath73 to @xmath74 with @xmath251 .
the result of this computation is shown in figure [ fig : het_full ] as the red curve .
we have truncated the result at @xmath252 .
in fact , the curve in figure [ fig : het_full ] can be extended to @xmath253 .
obviously we should view this curve as an approximation to the upper part of the c - curve .
+ heteroclinic connections for equation in parameter space .
the red curve indicates left - to - right connections for @xmath243 and the blue curves indicate right - to - left connections for @xmath254 with @xmath255 ( from top to bottom).,scaledwidth=50.0% ] if the connection from @xmath74 back to @xmath73 occurs with vertical coordinate @xmath256 , it is a trajectory of system ( [ eq : hetateq ] ) with @xmath254 .
figure [ fig : het_full ] shows values of @xmath124 at which these heteroclinic orbits exist for @xmath255 .
an intersection between a red and a blue curve indicates a singular homoclinic orbit .
further computations show that increasing the value of @xmath257 slowly beyond @xmath258 yields intersections everywhere along the red curve in figure [ fig : het_full ] .
thus the values of @xmath257 on the homoclinic orbits are expected to grow as @xmath28 increases along the upper branch of the c - curve . since there can not be any singular homoclinic orbits for @xmath259 we have to find the intersection of the red curve in figure [ fig : het_full ] with the vertical line @xmath260 . using the numerical method to detect heteroclinic connections
gives : the singular homoclinic curve for positive wave speed terminates at @xmath260 and @xmath261 on the right and at @xmath253 and @xmath90 on the left . in @xmath124-parameter space
define the points : @xmath262 in figure [ fig : amazing ] we have computed the homoclinic c - curve for values of @xmath16 between @xmath227 and @xmath263 .
together with the singular limit analysis above , this yields strong numerical evidence for the following conjecture : [ cjt : hom ] the c - shaped homoclinic bifurcation curves converge to the union of the segments @xmath264 and @xmath265 as @xmath266 .
_ remark 1 : _ figure 4 of krupa , sandstede and szmolyan @xcite shows a `` wedge '' that resembles shown in figure [ fig : amazing ] .
the system that they study sets @xmath127 and varies @xmath25 with @xmath267 . for @xmath268 and @xmath127 , the equilibrium point @xmath97
is located at the origin and the fast subsystem with @xmath269 has a double heteroclinic connection at @xmath97 to the saddle equilibrium @xmath270 .
the techniques developed in @xcite use this double heteroclinic connection as a starting point .
generalizations of the results in @xcite might provide a strategy to prove conjecture [ cjt : hom ] rigorously , a possibility that we have not yet considered .
however , we think that 1-homoclinic orbits in the regime we study come in pairs and that the surface of 1-homoclinic orbits in @xmath226 space differs qualitatively from that described by krupa , sandstede and szmolyan .
+ singular limit ( @xmath271 ) of the c - curve is shown in blue and parts of several c - curves for @xmath272 have been computed ( red).,scaledwidth=90.0% ] _ remark 2 : _ we have investigated the termination or turning mechanism of the c - curve at its upper end .
the termination points shown in figure [ fig : cusystem ] have been obtained by a different geometric method .
it relies on the observation that , in addition to the two fast heteroclinic connections , we have to connect near @xmath73 back to the equilibrium point @xmath97 to form a homoclinic orbit ; the two heteroclinic connections might persist as intersections of suitable invariant manifolds but we also have to investigate how the flow near @xmath54 interacts with the stable manifold @xmath273 .
these results will be reported elsewhere , but we note here that @xmath274 .
+ the numerical calculations of the c - curves for @xmath45 are new .
numerical continuation using the boundary value methods implemented in auto @xcite or matcont @xcite becomes very difficult for these small values of @xmath16 @xcite . even computing with values
@xmath275 using boundary value methods is a numerically challenging problem .
the method we have used does not compute the homoclinic orbits themselves while it locates the homoclinic c - curve accurately in parameter space . to motivate our approach consider figure [ fig : splitting ] which shows the unstable manifold @xmath276 for different values of @xmath28 and fixed @xmath48 .
we observe that homoclinic orbits can only exist near two different wave speeds @xmath277 and @xmath278 which define the parameters where @xmath279 or @xmath280 .
figure [ fig : splitting ] displays how @xmath276 changes as @xmath28 varies for the fixed value @xmath281 . if @xmath28 differs from the points @xmath277 and @xmath278 that define the lower and upper branches of the c - curve for the given value of @xmath48 , then @xmath282 increases rapidly on @xmath276 away from @xmath97 .
the changes in sign of @xmath58 on @xmath276 identify values of @xmath28 with homoclinic orbits .
the two splitting points that mark these sign changes are visible in figure [ fig : splitting ] . since trajectories close to the slow manifolds separate exponentially away from them , we are able to assess the sign of @xmath58 unambiguously on trajectories close to the slow manifold and find small intervals @xmath283 and @xmath284 that contain the values of @xmath28 for which there are homoclinic orbits .
+ the geometry of the orbits along the upper branch of the c - curve is obtained by approximating it with two fast singular heteroclinic connections and parts of the slow manifolds @xmath56 and @xmath54 ; this process has been described several times in the literature when different methods were used to prove the existence of `` fast waves '' ( see e.g. @xcite ) .
our results are summarized in the singular bifurcation diagram shown in figure [ fig : singbif ] .
this figure shows information obtained by a combination of fast - slow decompositions , classical dynamical systems techniques and robust numerical algorithms that work for very small values of @xmath16 .
it recovers and extends to smaller values of @xmath16 the cu - structure described in @xcite for the fitzhugh - nagumo equation .
the u - shaped hopf curve was computed with an explicit formula , and the homoclinic c - curve was determined by locating transitions between different dynamical behaviors separated by the homoclinic orbits .
all the results shown as solid lines in figure [ fig : singbif ] have been obtained by considering a singular limit . the lines @xmath264 and @xmath265 as well as the slow flow bifurcation follow from the singular limit @xmath6 yielding the fast and slow subsystems of the fitzhugh - nagumo equation .
the analysis of canards and periodic orbits have been obtained from equations ( [ eq : red1 ] ) and ( [ eq : red2 ] ) where the singular limit @xmath233 was used ( see section [ sec:2slow1fast ] ) .
we have also shown the c- and u - curves in figure [ fig : singbif ] as dotted lines to orient the reader how the results from proposition [ prop : hopf ] and conjecture [ cjt : hom ] fit in .
+ we also observed that several dynamical phenomena are difficult to recover from the singular limit fast - slow decomposition .
in particular , the codimension two generalized hopf bifurcation does not seem to be observable from the singular limit analysis . furthermore the homoclinic orbits can be constructed from the singular limits but it can not be determined directly from the fast and slow subsystems that they are of shilnikov - type . + the type of analysis pursued here seems to be very useful for other multiple time - scale problems involving multi - parameter bifurcation problems . in future work
, we shall give a geometric analysis of the folding / turning mechanism of the homoclinic c - curve , a feature of this system we have not been able to determine directly from our singular limit analysis .
that work relies upon new methods for calculating @xmath54 and @xmath56 which are invariant slow manifolds of `` saddle - type '' with both stable and unstable manifolds .
+ we end with brief historical remarks .
the references cited in this paper discuss mathematical challenges posed by the fitzhugh - nagumo equation , how these challenges have been analyzed and their relationship to general questions about multiple time - scale systems . along the line @xmath264 in figure [
fig : singbif ] we encounter a perturbation problem regarding the persistence of homoclinic orbits that can be solved using fenichel theory @xcite .
the point @xmath86 marks the connection between fast and slow waves in @xmath124-parameter space which has been investigated in @xmath285-parameter space in @xcite .
we view this codimension 2 connectivity as one of the key features of the fitzhugh - nagumo system .
the perturbation problem for homoclinic orbits close to the line @xmath265 was solved using several methods and was put into the context of multiple time - scale systems in @xcite , where the exchange lemma overcame difficulties in tracking @xmath276 when it starts jumping away from @xmath56 .
this theory provides rigorous foundations that support our numerical computations and their interpretation . +
* acknowledgment : * this research was partially supported by the national science foundation and department of energy .
k. bold , c. edwards , j. guckenheimer , s. guharay , k. hoffman , judith hubbard , ricardo oliva , and waren weckesser .
the forced van der pol equation 2 : canards in the reduced system .
, 2(4):570608 , 2003 . | the fitzhugh - nagumo equation has been investigated with a wide array of different methods in the last three decades .
recently a version of the equations with an applied current was analyzed by champneys , kirk , knobloch , oldeman and sneyd @xcite using numerical continuation methods .
they obtained a complicated bifurcation diagram in parameter space featuring a c - shaped curve of homoclinic bifurcations and a u - shaped curve of hopf bifurcations .
we use techniques from multiple time - scale dynamics to understand the structures of this bifurcation diagram based on geometric singular perturbation analysis of the fitzhugh - nagumo equation .
numerical and analytical techniques show that if the ratio of the time - scales in the fitzhugh - nagumo equation tends to zero , then our singular limit analysis correctly represents the observed cu - structure . geometric insight from the analysis can even be used to compute bifurcation curves which are inaccessible via continuation methods .
the results of our analysis are summarized in a singular bifurcation diagram . |
in a recent letter , m. sonnleitner _ et al . _
@xcite studied counterintuitive forces which arise between a black body and neutral atoms or molecules - named black body force ( bbf ) - due to the energy associated with shifts in the absorption lines of their spectra .
these shifts are induced by the so - called dynamical stark effect , in which the incoherent electromagnetic waves generated from the heated black body surface disturbs the electronic transitions in the atomic spectrum . in the above mentioned article ,
the authors have found that attractive forces undoubtedly arise from shifts in the electronic transitions involving the @xmath0 level of the hydrogen atom , and such forces would eventually be stronger than the repulsive force caused by the radiation pressure and the gravitational force due to the black body mass .
however , the energy associated to the spectral shift induced in other atom species could be positive , leading to repulsive forces . in the microscopic level
, this effect can arise from other incoherent electromagnetic emissions , strongly dependent on the shape of the emitting surface .
it is also noticeable in @xcite the emphasis put on the connection between the geometry of the bodies and the bbf , but the authors did not investigate the possible influences of both the topology and spacetime geometry on that force . in other words , those authors limited themselves to flat geometry with trivial topology .
astrophysical and cosmological scenarios as well as some models in condensed matter physics offer possible investigation lines for the bbf .
however , corrections to that force are necessary because even in a weak - field regime , in standard or alternative theories of gravitation , there are non - negligible modifications in the black body radiation law caused by the gravity @xcite , which must be taken into account for a complete understanding of the phenomena .
this is another sense in which geometry is fundamental in understanding of the bbf .
it is worth noticing that the authors of ref.@xcite show that the bbf depends on both the temperature and the solid angle , which are modified by the spacetime geometry .
therefore this work focus on the effects due to both the spacetime geometry and topology on the bbf .
this is considered for static and spherical gravitational sources as well as for infinitely long cylinders , including those ones like cosmic strings .
we compute these force corrections and study their behavior comparing them with the flat case , obtained in @xcite . in the case of the spherical symmetry we will particularize for the schwarzschild solution , making equally manifest the role of the local spacetime geometry in the intensity of the bbf .
we will numerically calculate this corrected force regarding typical neutron stars and the sun . regarding the cylindrical geometry ,
whose bbf is initially calculated for the infinitely long newtonian cylinder , an angular ( deficit ) parameter @xmath1 is appropriately introduced in order to investigate effects of the spacetime topology on that force .
such a parameter multiplies the azimuthal angle entering in the calculation of the solid angle subtended by the atom , and it characterizes the global properties of the cosmic string spacetime , which presents a locally flat geometry , but with global curvature .
this feature generates interesting effects , as self - interaction on particles and dipoles @xcite and gravitational lens @xcite .
in addition , we also consider the bbf caused by a cosmic string rotating stationarily around its symmetry axis , where effects due to the spacetime rotation are expected .
the cosmic string spacetime has a non - trivial topology which is encoded in the aforementioned angular parameter given by @xmath2 @xcite , where @xmath3 is the newtonian constant of gravity and @xmath4 is the linear mass density of the string .
the more recent observations of the cosmic microwave background ( cmb ) provide an upper bound on dimensionless parameter @xmath5 of the order of @xmath6 @xcite .
it is worth emphasize that cosmic strings can have played an important role in the primitive universe , and it is known that in its very early ages the temperature was extremely hot @xcite ; thus it is natural thinking that those objects were in constant interaction with the thermal radiation that bathed the universe .
indeed , the effects of the thermal equilibrium of cosmic strings with radiation were analyzed in @xcite .
although cosmic strings are still subjects of theoretical speculation , a concrete realization of them occurs in some condensed matter systems .
the static cosmic strings are emulated as topological defects termed disclinations , which lie within crystalline structures , including liquid crystals in the nematic phase [ 8,9 ] , while the avatars of the stationary cosmic strings are vortices in a superfluid [ 10,11 ] .
the paper is organized as follows . in section 2
, we analyze the bbf on a neutral atom due to the radiation emitted by spherical sources , studying changes in this force due to the gravitational field . in section 3
, we consider cylindrical sources in the calculation of that force , revealing the role of both the topology and spacetime rotation in the production of it and , finally , in section 4 we discuss the results .
units in which @xmath7 is taken in all calculations .
in this section we examine the influence of spacetime geometry on the bbf . as shown in ref .
@xcite , for the non - relativistic case , the attractive potential associated to this force exerted on the neutral hydrogen atom in its fundamental state , placed at a distance @xmath8 from the center of a spherical black body of radius @xmath9 at temperature @xmath10 is , in natural units , @xmath11 where @xmath12 is the fine structure constant , @xmath13 is the mass of the electron , and @xmath14 is the solid angle subtended by the atom .
about the above expression some points are worth comment . as pointed in ref .
@xcite , for the hydrogen atom in an isotropic thermal bath , the dynamical stark effect will mostly cause a negative energy shift for the ground state .
if the temperature of the thermal bath is @xmath10 , this shift is given by @xmath15 since the atom is in an isotropic thermal bath , the black body field can not have any directional effect on the atom motion , inducing only friction and diffusion .
those authors then argue that if a finite source is considered , this produces a net force .
the point is that in this case , the atom does not see radiation coming from all directions , but only from a solid angle determined by the source .
therefore , only a fraction of the energy shift in the isotropic bath would cause the bbf .
this fraction is given by the solid angle compared to @xmath16 , giving us the final potential energy of eq .
( [ potential ] ) . for the non - relativistic spherically symmetric case
this solid angle is given by @xmath17 from the above expression we can see that the radial dependence of the potential comes only from the solid angle .
however , when we consider curved static spacetimes , the effective temperature will also depend on the coordinates via tolman temperature @xcite . therefore , in order to find corrections to the bbf two aspects must be considered : the first is the curvature correction to the temperature and the second is the modification in the solid angle determined by the spacetime curvature .
this is the general procedure that we will follow throughout this manuscript .
now we must consider the specific case with spherical symmetry .
the cases in which global or topological properties are relevant will be left to the next section , where cylindrical symmetries play an important role .
the static and spherically symmetric exterior solution of einstein s equation of general relativity is considered , which is given by the metric @xmath18 where @xmath19 and @xmath9 is the radius of the source as seem from infinity .
the interior solution is also important to determine the solid angle as seem by the atom . for this
we use the schwarzschild one which is given by @xmath20 as said above , the spacetime modification to the temperature is given by the tolman temperature . at the position of the atom
this is given by @xmath21 where @xmath10 is the temperature of the isotropic thermal bath , as measured by an observer at the infinity .
the fact that the temperature depends on the radial coordinate will contribute to modify the bbf .
now we must consider the modification in the solid angle due to the spacetime curvature which occurs in the system investigated here .
two corrections in expression ( [ solido ] ) must be considered , since the size of the source and the distance of the atom to its center are modified by gravity .
these two quantities are calculated via @xmath22 where @xmath23 is the effective distance of the atom to the source center and the first integral in the rhs is its effective radius @xmath24 .
this latter can be obtained by using eq .
( [ metric < ] ) and is given by @xmath25 and the second integral is @xmath26 thus we obtain that the modification in the solid angle will be given by @xmath27 notice that the above expression reproduces the euclidean solid angle for large values of @xmath8 and at the surface of the source is given by @xmath28 .
plugging eq .
( [ teff ] ) and eq .
( [ solidomodificado ] ) into eq .
( [ potential ] ) we get the final corrected expression for the potential @xmath29 this potential corresponds to a force that also drops with the cube - inverse at large distances and is infinity when @xmath30 , as in the flat spacetime case . ,
gravitationally corrected and flat , both as functions of the radial coordinate.,width=321 ] in fig . 1
, we depict the graph of the relativistically corrected potential @xmath31 over the temperature function @xmath32 , against the same potential energy in the flat case , both depending on the radius @xmath8 , for a compact sphere with @xmath33 .
the difference between both is sensible , and the magnitude of the corrected one is considerably greater than the observed in the other case . at @xmath34 ,
the corrected potential is approximately 14% lower than the flat case . at smaller distances ,
the difference increases even more .
the calculation of the bbf itself is made by performing @xmath35 @xcite .
notice that by considering the sun , this reduction in the temperature - independent part of the bbf potential is approximately 0.001% at @xmath34 . for a quantum gravity correction to this force ,
see @xcite .
in this section we will consider the bbf caused by systems with cylindrical symmetries .
there are three cases to be considered here .
one is the non - relativistic radiating infinity cylinder , which was not considered in the original ref .
the other two cases consider the influence of the spacetime generated by both static and stationary cosmic strings , with and without thickness . from the expression without gravity ( [ potential ] ) , the only change in the potential will be due to the solid angle . supposing that the particle is at a distance @xmath36 of the rectangle center , thus the solid angle of the whole cylinder subtended by the atom is given by @xcite @xmath37 where @xmath38 , @xmath39 ; @xmath40 and @xmath41 are the length and width of the rectangle , respectively .
taking an infinitely long cylinder , @xmath42 , and writing the width in terms of the cylinder radius , @xmath9 , and of the distance of the particle from the cylinder symmetry axis , @xmath8 , which coincides with the @xmath43 axis , we have @xmath44 with @xmath45 .
the argument @xmath46 is just the half - angle measured from the atom and which embraces the cylinder in the plane @xmath47 . in the fig .
1 we depict the neutral atom ( a ) close to the cylindrical black body . according to it ,
the geometrical elements present in our analysis are : @xmath48 , @xmath49 , @xmath50 , @xmath51 , and @xmath52 .
we find that the attractive potential on the hydrogen atom in the ground state due to the cylindrical black body at absolute temperature @xmath10 is given by @xcite @xmath53 where @xmath54 is the thermal shift in the absorption lines of the spectrum of the unexcited hydrogen atom .
thus the magnitude of the radial force exerted on this atom is given by @xmath55 where @xmath12 is now the fine - structure constant .
notice that the bbf is attractive , falling with a square - inverse law at large distances ( or @xmath56 ) . that is different from what happens in the spherical black - bodies where the force drops with the cubic - inverse law at those distances @xcite .
the fact that the cosmic strings spacetime have a locally flat geometry but a non - trivial topology raises interesting points .
first of all , due to the local flatness , the effective temperature is not changed and therefore the only possible correction comes from the solid angle like in the previous case .
the interesting point is that the solid angle now depends on the deficit angular parameter @xmath1 , incorporating then topological features in the bbf .
let us model a static cosmic string , or its realization as a disclination in a liquid crystal , by the cylinder above considered .
first , the conical geometry of the space around of the defect has to be incorporated .
such a geometry is described through the metric @xmath57 this expression describes the space around an ideal cosmic string of zero thickness , or the exterior region of that one endowed with an internal structure @xcite . for this defect , @xmath58 ( angular deficit parameter ) , and @xmath59 . for a disclination , @xmath1 can also be between zero and one , meaning that we have an angular excess parameter .
then all we have to do is rewrite the half - angle from which the atom sees the cylinder , @xmath60 , in terms of the correspondent half central ( azimuthal ) angle , @xmath61 , such that @xmath62 and to insert the factor @xmath63 which multiplies the azimuthal angles , according to the metric ( [ 05 ] ) .
thus the expression ( [ potentialnrc ] ) becomes @xmath64}}}},\ ] ] and calculating the force via @xmath65 , we get @xmath66 it is worth notice that , for the neutral hydrogen atom in the minkowsky space ( _ i.e. _ , when @xmath67 ) we have the expression ( [ 04 ] ) again , and when one takes the limit @xmath68 valid for an ideal cosmic string , the bbf vanishes .
then , an important conclusion is that ideal cosmic strings irradiating do not exert bbfs . for non - zero thickness and @xmath69
, we can approximate eq .
( [ fnu ] ) by @xmath70 the force is attractive , due to the factor @xmath71 to be negative for neutral hydrogen atoms , and it still falls off with the inverse square of the radial distance . such a law is the same one that manages the self - force on neutral ( or charged ) massive particles situated in the spacetime of a gott - hiscock cosmic string , when the particle is far away from the string @xcite .
let us now analyze the bbf on a neutral atom due to a spinning cosmic string , which is the stationary counterpart of the static cosmic string . as seen in the introduction ,
these objects appear in condensed matter physics as vortices in superfluids .
the correspondent spacetime metric is given by @xcite @xmath72 where @xmath73 is the angular momentum per unit length of the string , such that when it vanishes , we recover eq .
( [ 05 ] ) .
from the above expression we can see that the effective angular deficit parameter is given by @xmath74 notice that this provides a consistent definition since it recover the static case for large @xmath8 .
the bbf on the neutral atom , when one modifies @xmath75 in eq .
( [ vnu ] ) is given by @xmath76 where @xmath77 .
we call attention to the fact that the force now is non - zero when the string radius vanishes , yielding @xmath78 while the string is spinning . for a neutral hydrogen atom very distant from the string
, the attractive bbf now falls off with the inverse cube of its radial distance , @xmath79 , and diverges at @xmath80 , _
i.e. _ , at the neighborhood of the boundary that defines the region of the closed time - like curves ( ctc s ) @xcite . in the interior of that region , @xmath81 ,
the bbf is not defined .
strictly speaking , the solid angle given in eq .
( [ cylindersolidangle ] ) is valid only for flat spaces .
however , the modification that must occur in a curved space was effectively implemented in the above systems through the azimuthal angle , since this latter is the only geometric parameter that is changed in fact , as we can see in the metrics ( 23 ) and ( 28 ) , altering solely the global properties of these space - times , which remain locally flat .
we have studied the effects of the black body radiation produced by spherical and cylindrical sources on a nearby neutral hydrogen atom in the ground state , which suffers , via dynamical stark effect , a shift in its spectrum absorption lines producing a attractive force ( bbf ) .
this was made by analyzing geometrical and topological influences of their respective space - times on it .
initially , we have calculated the effects due to a static and spherically symmetric gravitational background , starting by taking the tolman temperature and substituting it in bbf general expression , yielding a first correction to it . in order to arrive at the full correction , we taken also into account modifications in the solid angle subtended by the atom due to the curvature of the spacetime . to do this , we considered the radial proper distance to the center of the source , which was computed from both the exterior and interior schwarzschild solutions . when the results found here are applied to a compact object with this parameter measuring ten times the correspondent schwarzschild radius ,
the bbf potential gravitationally corrected is _ circa _ 14% lower than the flat case at @xmath34 . for the sun
, we verified that the einstein s gravity reduces the bbf potential about 0.001 per cent in relation to this quantity when calculated in the flat spacetime at the same distance , which is therefore a very tiny difference when compared to the previous case .
the analysis continued by taking into account the radiation emitted from an infinitely long cylinder with a diameter equal to @xmath82 , and calculated the attractive potential on the hydrogen atom related to that effect , taking into account just the source s shape . in this case , the bbf is attractive , falling with a square - inverse law at large distances , differently of that happens with spherical black - bodies , in which the force obeys the inverse cube law at those distances as shown in @xcite .
we must point out , as awaited , that the proper body geometry plays a central role in the bbf .
next , we have appropriately introduced the angular parameter @xmath1 in the expression correspondent to the attractive potential due to the cylindric black body radiation .
this parameter multiplies the azimuthal ( or central ) angle which enters in the calculation of the solid angle subtended by the atom , characterizing the conical geometry of the cosmic string ( or of its realization as a disclination in a crystal ) spacetime , and , thus , its topological properties .
the bbf vanishes when we consider an ideal static cosmic string , in which @xmath83 , but it is non - zero when we taken into account an ideal spinning cosmic string ( or vortices in superfluids ) , now falling off at large distances according to an inverse cube law , as in the spherical case .
we can conclude that such a force is weaker than that one presented in the static case because the non - inertial effects caused by the spacetime rotation itself , or , in other words , by the frame dragging associated with stationary cosmic strings @xcite .
we have also shown that this force is not defined in the interior of the ctc s domain , diverging at its boundary .
the authors acknowledge the financial support provided by fundao cearense de apoio ao desenvolvimento cientfico e tecnolgico ( funcap ) , by conselho nacional de desenvolvimento cientfico e tecnolgico ( cnpq ) , and funcap / cnpq / pronex .
m. sonnleitner , m. ritsch - marte and h. ritsch , phys .
lett . * 111 * , 023601 ( 2013 ) . c. r. muniz , v. b. bezerra and m. s. cunha , phys .
d * 88 * , 104035 ( 2013 ) .
bezerra , h. f. mota and c. r. muniz , phys . rev . *
d 89 * , 044015 ( 2014 ) .
m. r. anderson , _ the mathematical theory of cosmic strings : cosmic strings in the wire approximation _ , iop publishing ltd . , ( 2003 ) . c. r. muniz and v. b. bezerra , ann .
phys . * 340 * , 87 ( 2014 ) .
a. vilenkin and e.p.s .
shellard , _ cosmic strings and other topological defects _ , cambrigde university press ( 1994 ) .
p. a. r. ade et al . , arxiv:1303.5085 .
p. j. e. peebles , _ principles of physical cosmology _ , princeton university press , princeton , ( 1993 ) .
p. c. w. davies and v. sahni , class .
quantum grav . * 5 * , 1 ( 1988 ) .
a. l. muoz , _ liquid crystal dynamics : defects , walls and gels _ , cuvillier verlag , goetingen ( 2005 ) . c. stiro and f. moraes , mod .
a * 20 * , 2561 ( 2005 )
. r.l .
davis and e.p.s shellard , phys .
* 63 * , 2021 ( 1989 ) .
g. e. volovik , jetp lett . ,
* 67 * , 881 ( 1998 ) .
a. khadjavi , j. opt .
. am . * a58 * , 1417 ( 1968 ) .
b. allen , b. s. kay and a. c. ottewill , phys.rev . * d53 * , 6829 ( 1996 ) .
n.r . khusnutdinov and v.b .
bezerra , phys.rev .
* d64 * 083506 ( 2001 ) .
v. a. de lorenci , r. d. m. de paola and n. f. svaiter , class .
quantum grav .
* 16 * , 3047 ( 1999 ) . c. r. muniz , v. b. bezerra and m. s. cunha , ann
* 350 * , 105 ( 2014 ) .
b. jensen and h. h. soleng , phys .
rev . * d45 * , 3528 ( 1992 ) .
r. p. feynman , _ statistical mechanics , a set of lectures _ , levant books , kolkata ( 2008 ) .
r.c.tolman , _ relativity , thermodynamics and cosmology _ , dover publications , inc .
, new york ( 1987 ) .
w.rindler , _ relativity : special , general , and cosmological _ , oxford university press , 2nd ed .
( 2006 ) g. alencar , v.b .
bezerra , m.s .
cunha , and c.r .
muniz phys.lett.*b747 * , 536 ( 2015 ) .
h. culetu , j. phys .
68 , 012036 ( 2007 ) . | in this paper we compute the corrections to the black body force due to spacetime geometry and topology .
this recently discovered attractive force on neutral atoms is caused by the thermal radiation emitted from black bodies and here we investigate it in systems with spherical and cylindrical symmetries .
for some astrophysical objects we find that the corrected force is greater than the flat case , showing that this kind of correction can be quite relevant when curved spaces are considered .
then we consider four cases : the schwarzschild spacetime , the non - relativistic infinity cylinder , and both the static and stationary cosmic strings .
for the spherically symmetric case we find that two corrections appear : one due to the gravitational modification of the temperature and the other due to the modification of the solid angle subtended by the atom .
we apply the found results to a typical neutron star and to the sun . for the cylindrical case , which is locally flat , no gravitational correction to the temperature exists . however , we find the curious fact that the black body force depends on the topology of the spacetime through the modification of the azimuthal angle and therefore of the solid angle . for the static cosmic string
we find that the force is null for zero thickness case .
for the stationary one , we find that the force is non - null even for the thin case . |
decays of the @xmath0 provide an excellent source of events with which to study light hadron spectroscopy and search for glueballs , hybrids , and exotic states . since the discovery of the @xmath0 at brookhaven @xcite and slac @xcite in 1974 , more than one hundred exclusive decay modes of the @xmath0 have been reported .
recently , @xmath4 @xmath0 events and @xmath5 @xmath1 events have been obtained with the upgraded beijing spectrometer ( besii ) , and these samples offer a unique opportunity to measure precisely the branching fractions of @xmath0 decays .
the largest @xmath6 decay involving hadronic resonances is @xmath7 .
its branching fraction has been reported by many experimental groups @xcite assuming all @xmath8 final states come from @xmath9 .
the precision of these measurements varies from @xmath10 to 25% . in this paper , we present two independent measurements of this branching fraction using @xmath0 and @xmath1 decays .
the first is an absolute measurement based on @xmath11 directly .
the second , in which many of the systematic errors cancel out , is a relative measurement obtained from a comparison of the rates for @xmath11 and @xmath12 , using @xmath6 events produced via @xmath13 .
the upgraded besii detector operates at the beijing electron - positron collider ( bepc ) ; it is a large solid - angle magnetic spectrometer that is described in detail in ref .
the momentum of charged particles is determined by a 40-layer cylindrical main drift chamber ( mdc ) which has a momentum resolution of @xmath14 , where @xmath15 is in units of gev / c .
particle identification is accomplished using specific ionization ( @xmath16 ) measurements in the drift chamber and time - of - flight ( tof ) information in a barrel - like array of 48 scintillation counters .
the @xmath16 resolution is @xmath17 ; the tof resolution for bhabha events is @xmath18 ps .
radially outside the time - of - flight counters is a 12-radiation - length barrel shower counter ( bsc ) comprised of gas proportional tubes interleaved with lead sheets .
the bsc measures the energies and directions of photons with resolutions of @xmath19 , @xmath20 mrad , and @xmath21 cm .
the iron flux return of the magnet is instrumented with three double layers of counters ( muc ) that are used to identify muons . in the analysis ,
a geant3 based monte carlo program ( simbes ) with detailed consideration of detector performance ( such as dead electronic channels ) is used .
the consistency between data and monte carlo has been checked in many high purity physics channels , and the agreement is reasonable .
each charged track , reconstructed using hits in the mdc , must ( 1 ) have a good helix fit , in order to ensure a correct error matrix in the kinematic fit ; ( 2 ) originate from the interaction region , @xmath22 cm and @xmath23 cm , where @xmath24 , @xmath25 , and @xmath26 are the x , y , and z coordinates of the point of closest approach of the track to the beam axis ; ( 3 ) have a transverse momentum greater than 60 mev / c ; and ( 4 ) have @xmath27 , where @xmath28 is the polar angle of the track . a neutral cluster in the bsc is assumed to be a photon candidate if the following requirements are satisfied : ( 1 ) the energy deposited in the bsc is greater than 0.06 gev ; ( 2 ) the total number of layers with deposited energy is greater than one ; ( 3 ) the angle between the direction of photon emission and the direction of shower development is less than @xmath29 ; and ( 4 ) the angle between the photon and the nearest charged track is greater than @xmath30 .
if the angle between two neutral clusters is less than @xmath31 and their @xmath32 invariant mass is less than 0.05 gev / c@xmath33 , they are combined with the cluster with the largest energy being used for the direction and energy of the combined cluster in the kinematic fit .
events with two oppositely charged tracks and at least two good photons are selected for further analysis . no charged particle identification is required . a 5-constraint ( 5c )
kinematic fit is made under the @xmath34 hypothesis with the invariant mass of the two photons being constrained to the @xmath35 mass .
if the number of the selected photons is larger than two , the fit is repeated using all permutation of the photons . for events with a good fit
, the two photon combination with the minimum fit @xmath36 is selected , and its value is required to be less than 15 . to select a clean sample ,
the following criteria are applied to the remaining events : * to reject the main background events from @xmath37 , a 5c kinematic fit for @xmath37 is performed , and @xmath38 is required .
[ chisq ] shows the scatter plot of @xmath36 versus @xmath39 . +
* background events from @xmath40 conversions ( @xmath41 ) are eliminated by requiring the angle between the two charged tracks , @xmath42 , to be greater than @xmath31 . * radiative events , for example @xmath43
, are removed by the requirement @xmath44 , where @xmath45 is the angle of the @xmath40 in the @xmath35 rest frame .
* contamination from @xmath46 is eliminated by the requirement that the sum of the deposited energies of the two charged pions in the bsc is less than 2 gev .
[ e1e2 ] shows the scatter plot of @xmath47 versus @xmath48 , where @xmath47 and @xmath48 are the deposited energies of the @xmath49 and @xmath50 in the bsc , respectively .
this criteria has almost no effect on @xmath51 events .
+ after the above requirements , 219691 @xmath51 candidates are selected .
remaining backgrounds are evaluated using two different monte carlo simulations . in the first , specific background channels , shown in table [ bkg ] ,
are generated .
the total background from these channels in the selected @xmath51 events is determined to be less than 1% .
the second simulation uses 30 million inclusive @xmath0 mc events generated with the lund model @xcite . after normalizing the selected background events to 58 million @xmath0 events ,
3799 background events are obtained , yielding a contamination of 1.7% . in this paper ,
the latter background estimate is used to correct the branching fraction , giving a correction factor of @xmath52% .
.background contributions from different decay channels . here
@xmath53 is the number of events generated , and @xmath54 is the number of background events selected , normalized by the branching fractions quoted in ref . @xcite . [
cols="<,^,^",options="header " , ] the branching fraction calculated with formula ( [ formula2 ] ) multiplied by the corrrection factor @xmath55 is @xmath56
the absolute branching fraction of @xmath57 has been determined using a sample of 58 million @xmath6 decays , as well as by measuring the relative branching fraction of @xmath58 to @xmath59 in @xmath60 decays with a sample of 14 million @xmath61 events .
the results are in good agreement .
the weighted mean of these two measurements is @xmath62 the only reported branching fraction for @xmath11 is by mark - ii @xcite , whereas many experiments have reported measurements for @xmath63 @xcite , which contributes the dominant part of the @xmath51 final state .
the result obtained here is higher than those of previous measurements and has better precision .
the bes collaboration thanks the staff of bepc for their hard efforts .
this work is supported in part by the national natural science foundation of china under contracts nos .
19991480 , 10225524 , 10225525 , the chinese academy of sciences under contract no .
kj 95t-03 , the 100 talents program of cas under contract nos .
u-11 , u-24 , u-25 , and the knowledge innovation project of cas under contract nos .
u-602 , u-34(ihep ) ; by the national natural science foundation of china under contract no .
10175060 ( ustc ) ; and by the department of energy under contract no .
de - fg03 - 94er40833 ( u hawaii ) . 120 j. j. aubert _ et al .
* 33 * , 1404 ( 1974 ) . j. e. augustin _ et al . _ , phys .
. lett . * 33 * , 1406 ( 1974 ) .
b. jean - marie _ et al .
lett . * 36 * , 291 ( 1976 ) . w. braunschweig _ et al .
_ , phys . lett . *
63b * , 487 ( 1976 ) .
w. bartel _
_ , phys . lett . *
64b * , 483 ( 1976 ) .
g. alexender _
et al . _ ,
72b * , 493 ( 1978 ) .
r. brandelik _ et al .
_ , phys . lett . *
74b * , 292 ( 1978 ) .
m. e. b. franklin _
et al . _ ,
lett . * 51 * , 963 ( 1983 ) .
d. coffman _
et al . _ ,
phys . rev . * d38 * , 2695 ( 1988 )
. j. z. bai _ et al .
_ , phys . rev . *
d54 * , 1221 ( 1996 ) .
j. z. bai _ et al .
methods phys .
* a458 * , 627 ( 2001 ) . k. hanssgen , h. j. mohring and j. ranft , nucl .
88 ( 1984 ) , 551 ; j. ranft and s. ritter , z. phys .
c20 ( 1983 ) , 347 .
et al . _ ,
fluka 92 , proceedings of the workshop on simulating accelerator radiation environments , santa fe , 1993 .
c. zeitnitz and t. a. gabriel , nim a349 ( 1994 ) 106 - 111 .
s. s. fang _
et al . _ , hep&np , 27 ( 4 ) , 277 ( 2003 ) ( in chinese ) . where @xmath66 is the angle between the vector meson and the positron direction .
for @xmath67 , @xmath68 and @xmath69 are the polar and azimuthal angles of the @xmath70 with respect to the helicity direction of the @xmath71 . | using 58 million @xmath0 and 14 million @xmath1 decays obtained by the besii experiment , the branching fraction of @xmath2 is determined .
the result is @xmath3 , which is significantly higher than previous measurements
. 0.2 cm |
these lectures are primarily about light - cone string field theory , which is an ancient subject ( by modern standards ) whose origins lie in the days of ` dual resonance models ' even before string theory was studied as a theory of quantum gravity @xcite .
light - cone string field theory is nothing more than the study of string theory ( especially string interactions ) via hamiltonian quantization in light - cone gauge .
let us immediately illustrate this point .
textbooks on string theory typically begin by considering the polyakov action for a free string , @xmath0 where @xmath1 is the @xmath2-dimensional string worldsheet ( we consider only closed strings in these lectures ) , @xmath3 is the metric on @xmath1 , and @xmath4 is the embedding of the string worldsheet into spacetime .
quantizing this theory is simplest in light - cone gauge , where the ghosts are not needed , and one finds the light - cone hamiltonian @xmath5,\ ] ] where @xmath6 labels the directions transverse to the light - cone . at this stage
most textbooks abandon light - cone gauge in favor of more powerful and mathematically beautiful covariant techniques . however , it is also possible to continue by second - quantizing ( or third - quantizing , depending on how you count ) the hamiltonian ( [ eq : hflat ] ) . to do this
we introduce a multi - string hilbert space , with operators acting to create or annihilate entire strings ( not to be confused with the operators @xmath7 and @xmath8 , which create or annihilate an oscillation of frequency @xmath9 on a given string ) .
all of the details will be presented in lecture 2 , where our ultimate goal will be to write down a relatively simple interaction term which ( at least for the bosonic string ) is able to reproduce ( in principle ) all possible string scattering amplitudes !
light - cone string field theory obscures many important properties of string theory which are manifest in a covariant treatment .
nevertheless , the subject has recently enjoyed a remarkable renaissance following @xcite because it is well - suited for studying string interactions in a maximally symmetric plane wave background , where covariant techniques are more complicated than they are in flat space .
one of the most exciting developments in string theory has been the discovery of the ads / cft correspondence @xcite ( see @xcite for a review ) .
the best understood example of this correspondence relates type iib string theory on @xmath10 to su(@xmath11 ) yang - mills theory with @xmath12 supersymmetry .
although we have learned a tremendous amount about gauge theory , quantum gravity , and the holographic principle from ads / cft , the full promise of the duality has unfortunately not yet been realized .
the reason is simple : string theory on @xmath10 is hard !
in contrast to ( [ eq : flatspace ] ) , the green - schwarz superstring on @xmath10 is nontrivial .
it can be regarded as a coset sigma model on @xmath13 with an additional fermionic wess - zumino term and a fermionic @xmath14-symmetry @xcite . despite some intriguing recent progress @xcite ,
this theory remains intractable and we still do not know the free string spectrum ( i.e. , the analogue of ( [ eq : hflat ] ) ) in @xmath10 .
because of this difficulty , most applications of the ads / cft correspondence rely on the supergravity approximation to the full string theory .
recently , a third maximally symmetric 10-dimensional solution of type iib string theory was found @xcite : the so - called maximally symmetric plane wave background .
this background can be obtained from @xmath10 by taking a penrose limit , and it is essentially flat space plus the first order correction from flat space to the full @xmath10 . in some sense this background sits half - way between flat space and @xmath10 .
it resembles @xmath10 because it is a curved geometry with non - zero five - form flux , and it resembles flat space because remarkably , the free string theory in this background can be solved exactly in light - cone gauge @xcite ( the details will be presented in lecture 1 ) .
furthermore it was realized in @xcite that the penrose limit of @xmath10 has a very simple description in terms of the dual yang - mills theory . in particular ,
bmn related iib string theory on the plane wave background to a sector of the large @xmath11 limit of @xmath15 su(@xmath11 ) gauge theory involving operators of large r - charge @xmath16 .
because string theory in the plane wave is exactly solvable , the bmn correspondence opens up the exciting opportunity to study stringy effects in the holographic dual gauge theory , thereby adding a new dimension to our understanding of gauge theory , gravity , and the holographic principle .
light - cone string field theory , although somewhat esoteric , has emerged from a long hibernation because it naturally connects the string and gauge theory descriptions of the plane wave limit of ads / cft .
light - cone string field theory is a very complicated and technical subject . in order to provide a pedagogical introduction ,
most of the material is developed for the simpler case of the bosonic string , although we do mention the qualitatively new features which arise for the type iib string .
the reader who is interested in a detailed study of the latter , significantly more complicated case is encouraged to refer to the papers @xcite , or to review articles such as @xcite .
our goal here has been to present in a clear manner the relevant background material which is usually absent in the recent literature on plane waves .
we start with a quick review of the remarkable duality between @xmath15 su(@xmath11 ) super - yang mills theory and type iib string theory on @xmath10 , referring the reader to @xcite for an extensive review and for additional references .
this is the best understood and most concrete example of a holographic duality between gauge theory and string theory .
the lagrangian for @xmath12 super - yang mills theory is @xcite @xmath17 ^ 2 + { \rm fermions } \right],\ ] ] where @xmath18 is the gauge field strength and @xmath19 , @xmath20 are six real scalar fields .
all fields transform in the adjoint representation of the gauge group .
the @xmath12 superalgebra in four dimensions is very constraining and essentially determines the field content and the lagrangian ( [ eq : lagrangian ] ) uniquely : the only freedom is the choice of gauge group ( which will always be su(@xmath11 ) in these lectures ) , and the value of the coupling constant @xmath21 .
this theory is conformally invariant , so @xmath21 is a true parameter of the theory .
( in non - conformal theories , couplings become functions of the energy scale , rather than parameters . ) the symmetry of the lagrangian ( [ eq : lagrangian ] ) is so(2,4)@xmath22so(6 ) , where the first factor is the conformal group in four dimensions and the second group is the so(6 ) r - symmetry group which acts on the six scalar fields in the obvious way . on the other side of the duality
we have type iib string theory on the @xmath10 spacetime , whose metric can be written as @xmath23 this is a solution of the iib equations of motion with constant string coupling @xmath24 and five - form field strength @xmath25 the isometry group of the spacetime is so(2,4)@xmath22so(6 ) , where the first factor is the isometry of @xmath26 and the second factor is the isometry of @xmath27 .
according to the ads / cft correspondence , there is a one - to - one correspondence between single - trace operators @xmath28 in the gauge theory and fields @xmath29 in ads .
the holographic dictionary between the two sides of this duality is summarized in table [ table1 ] , where @xmath30 is the t hooft coupling and @xmath31 is the string length .
.the ads / cft correspondence for @xmath10 .
[ table1 ] [ cols="^,^,^ " , ] it will be very convenient to define the quantities @xmath32 which remain finite in the bmn limit . also , we will refer to those operators with @xmath33 and finite @xmath34 as ` bmn operators . ' in table [ table2 ] we have introduced the parameter @xmath35 by rescaling @xmath36 as discussed above .
the question mark in table [ table2 ] indicates that we are still looking for a nice way to characterize this limit of the gauge theory . in other words
, what precisely does it mean to ` truncate ' the theory to a certain class of operators ; or , turned around : is there a simple description of the sector of the gauge theory which is dual to iib string theory on a plane wave ?
the most exciting aspect of the bmn correspondence is that the free iib string on the plane wave background is exactly solvable .
as discussed in the introduction , string theory on @xmath10 is in contrast rather complicated . in light cone gauge ,
the worldsheet theory for iib strings on the plane wave background ( green - schwarz action ) is simply @xcite @xmath37,\ ] ] where @xmath38 , @xmath39 are the bosonic sigma model coordinates , @xmath40 is a complex majorana spinor on the worldsheet and a positive chirality so(8 ) spinor under rotations in the transverse directions , and @xmath41 .
the action ( [ eq : wsact ] ) simply describes eight massive bosons and eight massive fermions , so it is trivially solvable .
let us consider here only bosonic excitations .
then a general state has the form @xmath42 and the hamiltonian be written as @xmath43 we have chosen a basis of fourier modes such that @xmath44 label left movers , @xmath45 label right movers , and @xmath46 is the zero mode .
this convention has become standard in the pp - wave literature and contrasts with the usual convention in flat space , where the left- and right - moving oscillators are denoted by different symbols @xmath47 and @xmath48 .
the alternate convention can be motivated by recalling that in flat space , the worldsheet theory remains a conformal field theory even in light - cone gauge .
therefore the left - moving modes @xmath47 and the right - moving modes @xmath48 decouple from each other
. however , in the plane - wave background , choosing light - cone gauge breaks conformal invariance on the world sheet ( because a mass term appears ) .
therefore all of the modes couple to each other , so there is no advantage to introducing a notation which treats left- and right - movers separately . since this is a theory of closed strings , we should not forget to impose the physical state condition , which says that the total momentum on the string should vanish : @xmath49 where @xmath50 is the occupation number ( the eigenvalue of @xmath51 ) .
we remarked above that the so(8 ) transverse symmetry is broken by the five - form field strength to so(4)@xmath22so(4)@xmath52 .
this manifests itself by the presence of @xmath53 in the worldsheet action ( [ eq : wsact ] ) .
however , the free spectrum of iib string theory on the plane wave is actually fully so(8 ) symmetric @xcite .
one can see this by noting that @xmath53 can be eliminated from ( [ eq : wsact ] ) by first splitting @xmath54 and then making the field redefinition @xmath55 .
one should think of so(8 ) as an accidental symmetry of the free theory . since the background breaks so(8 ) to so(4)@xmath22so(4)@xmath52
, there is no reason to expect that string interactions should preserve the full so(8 ) , and indeed we will see that they do not : the interactions break so(8 ) to so(4)@xmath22so(4)@xmath52 .
before we return to the gauge theory , it will be convenient to rewrite the spectrum ( [ eq : spectrum ] ) in terms of gauge theory parameters using the dictionary from table [ table2 ] : @xmath56 we have seen that iib string theory on the plane wave background has a very simple spectrum ( [ eq : gspec ] ) . in this section we will recover this spectrum by finding the set of operators which have @xmath33 and finite @xmath57 in the @xmath58 limit @xcite . consider first the ground state of the string , @xmath59 . according to ( [ eq : gspec ] ) this
should correspond to an operator with @xmath60 .
the unique such operator is @xmath61 $ ] , where @xmath62 ( recall that we defined @xmath63 to be the u(1 ) generator which acts by rotation in the @xmath64@xmath65 plane ) . the first entry in the
` bmn state - operator correspondence ' is therefore @xmath66.\ ] ] to get the first excited states we can add to the trace operators which have @xmath67 ; for example : @xmath68 ( for @xmath69 ) or @xmath70 ( again for @xmath69 , in the euclidean theory ) .
\cr a_0^{i+4 } d_i z z^j].\end{aligned}\ ] ] to get higher excited states we can add more ` impurities ' to the traces .
for example , a general state with @xmath72 is @xmath73.\ ] ] the sum on the right hand side runs over all possible orderings of the insertions inside the trace .
this sum is necessary to ensure that the operator is bps .
we will always work in a ` dilute gas ' approximation , where the number of impurities is much smaller than @xmath63 , the number of @xmath74 s .
so far we considered only bps operators in the gauge theory .
these have the property that @xmath75 is not corrected by interactions , i.e. @xmath75 does not depend on @xmath21 . according to ( [ eq : gspec ] ) , these
can only correspond to string states with the zero mode ( @xmath46 ) excited . in order to obtain other states
, we can consider summing over the location of an impurity with a phase : @xmath76.\ ] ] but the right hand side is zero ( for @xmath77 ) , because of cyclicity of the trace ! actually this is a good thing , because the string state on the left does not satisfy the physical state condition for @xmath77 . in order to get physical states
, we have to consider ( suppressing the @xmath6 transverse index ) @xmath78.\ ] ] here @xmath79 labels the position , in the string of @xmath74 s , of the @xmath6-th @xmath29 impurity .
cyclicity of the trace now implies that the right - hand side vanishes unless @xmath80 , and this is precisely the physical state condition for the string state on the left - hand side !
the operator on the right - hand side of ( [ eq : genstate ] ) is not bps when the phases are non - zero , so its dimension @xmath75 receives quantum corrections .
one can check that in the bmn limit @xmath81 with @xmath16 , the contribution to @xmath57 from an impurity with phase @xmath9 is @xmath82 precisely in accord with the prediction ( [ eq : gspec ] ) !
this calculation was performed to one loop in @xcite , to two loops in @xcite , and an argument valid to all orders in perturbation theory was presented in @xcite . at this point
we have motivated that the spectrum of iib string theory on the plane wave background can be identified with the set of bmn operators in @xmath15 su(@xmath11 ) yang - mills theory .
one point which we did not consider is the addition of impurities with @xmath83 , for example @xmath84 , which has @xmath85 .
it has been argued that these operators decouple in the bmn limit ( i.e. , their anomalous dimensions go to infinity ) , and can hence be ignored .
we refer the reader to @xcite for details .
next we present the two parameters @xcite which characterize the bmn limit .
something a little miraculous has happened .
the operator ( [ eq : genstate ] ) is not bps ( when the phases are nonzero ) , so its conformal dimension @xmath75 will receive quantum corrections .
generically , the dimension of a non - protected operator blows up in limit of large t hooft coupling .
and we certainly are taking @xmath87 here ( see table [ table2 ] ) !
however , note that ( [ eq : genstate ] ) is bps when all of the phases are zero : @xmath88 . in a sense , then , we might hope that the operator is ` almost ' bps as long as the phases are almost 0 , or in other words @xmath89 for all @xmath6 . by
` almost ' bps we mean that although the dimension does receive quantum corrections , those corrections are finite in the bmn limit despite the fact that @xmath90 .
indeed this is what happens : in the formula ( [ eq : gspec ] ) is it not the
t hooft coupling @xmath91 which appears , but rather a new effective coupling @xmath92 which is finite in the bmn limit .
it is hoped ( and indeed this hope is borne out in all known calculations so far ) that this miracle is quite general in the bmn limit : namely , that many interesting physical quantities related to these bmn operators remain finite despite the fact that the t hooft coupling is going to infinity . in the familiar large @xmath11 limit of su(@xmath11 )
gauge theory , the perturbation theory naturally organizes into a genus expansion , where a gauge theory diagram of genus @xmath94 is effectively weighted by a factor of @xmath95
. in particular , only planar ( @xmath96 ) diagrams contribute to leading order at large @xmath11 .
however , it has been shown @xcite that the effective genus counting parameter for bmn operators is @xmath97 .
the familiar intuition that only planar graphs are relevant at @xmath98 fails because we are not focusing our attention on some fixed gauge theory operators , and then taking @xmath58 .
rather , the bmn operators themselves change with @xmath11 , since we want to scale the r - charge @xmath16 . as @xmath11 becomes large ,
the relevant bmn operators are composed of @xmath99 elementary fields .
because of this large number of fields , there is a huge number @xmath100 of feynman diagrams at genus @xmath94 .
this combines with the @xmath95 factor to give a finite weight @xmath101 in the bmn limit for genus @xmath94 diagrams .
in this lecture we will canonically second - quantize string theory in light - cone gauge and write down its hamiltonian , which will be no more complicated ( qualitatively ) than @xmath102 where @xmath24 is the string coupling .
students of string theory these days are not typically taught that it is possible to write down an explicit formula for the hamiltonian of string theory .
an excellent collection of papers on this subject may be found in @xcite .
the light - cone approach does suffer from a number of problems which will be discussed in detail in the next lecture . however , light - cone string field theory is very well - suited to the study of string interactions in the plane wave background .
for one thing , it is only in the light - cone gauge that we are able to determine the spectrum of the free string .
since other approaches can not yet even give us the free spectrum , they can hardly tell us anything about string interactions .
although we hope this situation will improve , light - cone gauge is still very natural from the point of view of the bmn correspondence .
the dual bmn gauge theory automatically provides us a light - cone quantized version of the string theory , and it is hoped that taking the continuum limit of the ` discretized strings ' in the gauge theory might give us light - cone string field theory , although a large number of obstacles need to be overcome before the precise correspondence is better understood .
many of the fundamental concepts which will be introduced in this and the following lecture apply equally well to all string field theories , and not just the light - cone version .
it is therefore hoped that these lectures may be of benefit even to some who are not particularly interested in plane waves .
this portion of the lecture series is intended to be highly pedagogical .
we will therefore start by studying the simplest possible case in great detail .
here is a partial list of simplifications which we will start with : * during this lecture we will consider only bosons . somewhat surprisingly ,
fermions complicate the story considerably , but we will postpone these important details until the next lecture . the reader should keep in mind that the plane wave background is not a solution of bosonic string theory , so strictly speaking all of the formulas presented in this lecture need to be supplemented by the appropriate fermions .
( the flat space limit @xmath103 does make sense without fermions , if one works in 26 spacetime dimensions rather than 10 . ) * since the bosonic sector of string theory on the plane wave background is so(8 ) invariant , we will completely ignore the transverse index @xmath39 for most of the discussion .
it is trivial to replace , for example , @xmath104 in all of the formulas below . *
finally , we will begin not even with a string in the plane wave background , but simply a particle in the plane wave background !
a string can essentially be thought of as an infinite number of particles , one for each fourier mode on the string worldsheet .
a particle is equivalent to taking just the zero - mode on the worldsheet ( the mode independent of @xmath105 ) . in flat space
, there is a qualitative difference between this zero - mode and the ` stringy ' modes : the former has a continuous spectrum ( just the overall center - of - mass center - of - mass momentum of the string ) while the stringy modes are like harmonic oscillators and have a discrete spectrum .
however in the plane wave background , even the zero - mode lives in a harmonic oscillator potential , so it is not qualitatively different from the non - zero modes .
once we develop all of the formalism appropriate for a particle , it will be straightforward to take infinitely many copies of all of the formulas and apply them to a string .
we consider a particle propagating in the plane wave metric @xmath106 the action for a free ` massless ' field is @xmath107 where we have defined @xmath108.\ ] ] now let us canonically quantize this theory , so we promote @xmath109 from a classical field to an operator on the multi - particle hilbert space . the canonically conjugate field
to @xmath109 is @xmath110 , so the commutation relation is @xmath111 = i \delta(x^- - y^- ) \delta(x - y).\ ] ] let us pass to a fourier basis by introducing @xmath112 note : from now on we * define * @xmath113 in the supersymmetric string theory to be considered below , the supersymmetry algebra will guarantee that @xmath114 for all states .
we will proceed with this assumption , although it is certainly not true in the 26-dimensional bosonic theory .
the commutation relation is now @xmath115 = { 1 \over p^+ } \delta(p^+ + q^+ ) \delta(p+q).\ ] ] since @xmath109 is real ( as a classical scalar field ) , the corresponding operator @xmath109 is hermitian , which means that @xmath116 the hamiltonian can now be written as @xmath117 ( normal ordering will always be understood ) where @xmath118 is the single - particle hamiltonian @xmath119 ( we will always take @xmath120 . ) the subscript `` 2 '' in ( [ eq : hamil ] ) denotes that this is the quadratic ( i.e. , free ) part of the hamiltonian .
later we may add higher order interaction terms .
the single - particle hamiltonian may be diagonalized in the standard way : @xmath121 so that @xmath122 where @xmath123 .
it is important to distinguish two different hilbert spaces .
the single - particle hilbert space @xmath124 is spanned by the vectors @xmath125 the operators @xmath126 , @xmath127 and @xmath118 act on @xmath124 .
the second hilbert space is the multi - particle hilbert space @xmath128 .
let us introduce particle creation / creation operators @xmath129 which act on @xmath128 and satisfy @xmath130 and @xmath131 = e(p^+ ) \delta_{mn } \delta(p^+ + q^+).\ ] ] for @xmath132 , @xmath129 annihilates a particle in the state @xmath133 , while for @xmath134 , @xmath129 creates a particle in the state @xmath135 .
the vacuum of @xmath128 , denoted by @xmath136 , is annihilated by all @xmath129 which have @xmath132 .
normally in scalar field theory we do not introduce this level of complexity because the ` internal ' hamiltonian @xmath118 is so trivial .
we may refer to @xmath124 as the ` worldsheet ' hilbert space and @xmath128 as the ` spacetime ' hilbert space , since this is of course how these should be thought of in the string theory .
we now write the usual expansion for @xmath109 in terms of particle creation and annihilation operators : @xmath137 it is easily checked that the commutation relation ( [ eq : phiccr ] ) follows from ( [ eq : accr ] ) .
note that we have written @xmath109 as simultaneously a state in @xmath124 and an operator in @xmath128 .
this is notationally more convenient than the position space representation , @xmath138 which would leave all of our formulas full of hermite polynomials @xmath139 .
writing the field operator @xmath109 as a state in the single - particle hilbert space has notational advantages other than just being able to do without hermite polynomials .
for example , the equation of motion @xmath140 is just the schrdinger equation on @xmath109 : @xmath141 there is also a simple formula which allows us to take any symmetry generator on the worldsheet ( such as @xmath118 above , and later rotation generators @xmath142 , supercharges @xmath143 , @xmath144 ) and construct a free field realization of the corresponding space - time operator ( @xmath145 , @xmath146 , @xmath147 , @xmath144 ) : @xmath148 we already saw this formula applied to the hamiltonian in ( [ eq : hamil ] ) .
when we further make use of the expansion of @xmath109 into creation operators , we find the expected formula @xmath149 another trivial application is the identity operator it is easily checked that @xmath150 again , the subscript ` 2 ' in ( [ eq : gdef ] ) emphasizes that this gives a free field realization ( i.e. , quadratic in @xmath109 ) .
dynamical symmetry generators ( such as the hamiltonian , in particular ) will pick up additional interaction terms , but kinematical symmetry generators ( such as @xmath146 ) remain quadratic .
now let us consider a cubic interaction , @xmath151 where , for example , we might choose @xmath152 if we insert the expansion of @xmath109 in terms of modes and do the integrals , we end up with an expression of the form @xmath153 the functions @xmath154 are obtained from @xmath155 without too much difficulty : they simply encode the matrix elements of the interaction written in a basis of harmonic oscillator wavefunctions .
we now adopt a convention which is important to keep in mind .
because of the @xmath156 conserving delta function , it will always be the case that two of the @xmath156 s are positive and one is negative , or vice versa .
we will always choose the index ` 3 ' to label the @xmath156 whose sign is opposite that of the other two .
this means that the particle labeled ` 3 ' will always be the initial state of a splitting transition @xmath157 or the final state of a joining transition @xmath158 . in the state - operator correspondence ,
it is convenient to identify the operator cubic @xmath159 with a state in the 3-particle hilbert space @xmath160 ( where @xmath155 stands for ` vertex ' ) with the property that @xmath161 this state can be constructed by taking @xmath162 * exercise .
* compute @xmath163 and @xmath160 for @xmath164 .
it is essentially trivial to promote all of the formulas from the preceding section to the case of a string .
the field @xmath165 is promoted from a function of @xmath166 , the position of the particle in space , to a functional @xmath167 $ ] of the embedding @xmath168 of the string worldsheet in spacetime . in all of the above formulas , integrals over @xmath169
are replaced by functional integrals @xmath170 , and delta functions in @xmath166 are replaced by delta - functionals @xmath171 $ ] . these are defined as a product of delta functions over all of the fourier modes @xmath172 of @xmath173 .
the interacting quantum field theory of strings is described by the action @xmath174 where @xmath175 .
the formula ( [ eq : gdef ] ) is replaced by @xmath176 the worldsheet hamiltonian is now @xmath177 = { 1 \over p^+ } \sum_{n=-\infty}^\infty \omega_n a_n^\dagger a_n,\ ] ] where @xmath178 is the energy of the @xmath9-th mode and we have introduced a suitable basis of raising and lowering operators in order to diagonalize @xmath118 .
note that @xmath179 is identified with the operator @xmath126 corresponding to a particle , while for @xmath44 , @xmath180 are the left - movers and @xmath181 are the right - movers .
the full hilbert space @xmath124 of a single string is obtained by acting on latexmath:[$ ( for all @xmath9 !
note that we do not use any convention like @xmath183 ) .
we therefore label a state by @xmath184 , where the component @xmath185 of the vector @xmath186 gives the occupation number of oscillator @xmath9 .
note that we have to impose the @xmath187 physical state condition @xmath188 the second quantized hilbert space @xmath128 is introduced as before .
it has the vacuum @xmath136 , which is acted on by the operators @xmath189 , which for @xmath134 create a string in the state @xmath190 .
the representation of the hamiltonian at the level of free fields is just @xmath191 our goal now is to construct the state @xmath160 in @xmath192 which encodes the cubic string interactions , in the sense of formulas ( [ eq : eqaa ] ) and ( [ eq : eqbb ] ) .
what is the principle that determines the cubic interaction ?
it is quite simple : the embedding of the string worldsheet into spacetime should be continuous . in a functional representation ,
the cubic interaction is therefore just @xmath193\prod_{r=1}^3 \left ( dp_r^+ d x_r(\sigma)\ \phi [ p_r^+ , x_r(\sigma ) ] \right).\ ] ] there is one very important caveat : the principle of continuity requires the delta functional @xmath194 $ ] , but it does not determine the interaction ( [ eq : split ] ) uniquely because we have the freedom to choose the measure factor @xmath195 arbitrarily .
moreover , in principle the cubic interaction could involve derivatives of @xmath109 , such as @xmath196 ( where @xmath197 is the interaction point ) , whereas the interaction we wrote has only @xmath198 with no derivatives .
we will return to these points later .
our convention about the selection of @xmath199 guarantees that string 3 is always the ` long string ' .
the interaction ( [ eq : split ] ) mediates the string splitting @xmath157 , or its hermitian conjugate , the joining of @xmath158 .
this process is depicted in figure [ fig : fig1 ] .
all we have to do now is fourier transform this delta - functional into the harmonic oscillator number basis !
let us assemble the steps of this calculation .
* step 1 . *
first we recall the definition of a @xmath75-functional as a product of delta - functions for each fourier mode , @xmath200 = \prod_{m=-\infty}^\infty \delta \left ( \int_0^{2 \pi |p_3^+| } d\sigma \ e^{i m \sigma/|p_3^+| } [ x_1 ( \sigma ) + x_2(\sigma ) - x_3(\sigma ) ] \right).\ ] ] let us introduce matrices @xmath201 which express the fourier basis of string @xmath202 in terms of the fourier basis of string 3 ( so that , clearly , @xmath203 )
. then we can write @xmath204 = \prod_{m=-\infty}^\infty \delta\left ( \sum_{r=1}^3 x^{(r)}_{mn } p_{n(r ) } \right).\ ] ] these matrices are obtained by simple fourier transforms , @xmath205 where @xmath206 is the ratio of the width of string 1 to the width of string 3 .
* the expansion of the field @xmath109 in position space is given by @xmath207 = { 1 \over \sqrt{|p^+|}}\sum_{\vec{n } } a_{\vec{n}}(p^+ ) \prod_{n=-\infty}^\infty \psi_{n_n(x_n)},\ ] ] where @xmath208 is a harmonic oscillator wavefunction for the @xmath11-th excited level .
when we plug ( [ eq : sphi ] ) into the cubic action ( [ eq : split ] ) , we find that the coupling between the three strings labeled by @xmath209 , @xmath210 and @xmath211 is simply @xmath212 where the measure is @xmath213.\ ] ] * step 3 . *
the next step is to note that an @xmath166-eigenstate of an oscillator with frequency @xmath214 may be represented as @xmath215|0\rangle.\ ] ] it follows from this that @xmath216|0\rangle.\ ] ] note that the overall constant is irrelevant since we can absorb it into @xmath217
. * step 4 . * now let us assemble the couplings @xmath218 into the state @xmath160 : @xmath219 modes : @xmath220 the delta - functions allow us to replace all of the modes of string 3 in terms of the modes of strings 1 and 2 . then
( [ eq : final ] ) is just a gaussian integral in the infinitely many variables @xmath221 , @xmath222 , @xmath223 ! * step 5 . * the gaussian integral is easily done , and we find @xmath224 |0_{(1)}\rangle |0_{(2)}\rangle |0_{(3)}\rangle,\ ] ] where we have absorbed a constant into @xmath217 ( it was undetermined anyway ) , and we have introduced the matrices @xmath225 and @xmath226 we now give a more straightforward way to arrive at the same final result ( [ eq : vertexone ] ) .
after fourier transforming , the delta - functional can be expressed as local conservation of momentum density on the worldsheet : @xmath227.\ ] ] we are trying to find a state @xmath160 which is an oscillator representation of the position- and momentum - space delta - functionals .
now recall the elementary identity @xmath228 the state @xmath160 must therefore satisfy @xmath229 let us take the fourier transform of these equations with respect to the @xmath230-th fourier mode of string 3 .
then we make use of the same matrices @xmath231 introduced above , and we find the following equations , which must vanish for each @xmath230 : @xmath232 if we make an ansatz for @xmath160 of the form @xmath233 |0_{(1)}\rangle |0_{(2)}\rangle |0_{(3)}\rangle,\ ] ] for some coefficients @xmath234 , and then expand the @xmath166 s and @xmath235 s appearing in ( [ eq : eqns ] ) into creation and annihilation operators , then one obtains some matrix equations whose unique solution is @xmath236 it is clear that @xmath195 remains undetermined by this method .
we have written the hamiltonian of string theory in light cone gauge as a free term plus a cubic interaction .
it turns out ( at least for the bosonic string ) that this is the whole story !
one can use this simple hamiltonian to calculate the string s - matrix , to arbitrary order in string perturbation theory , with no conceptual difficulties .
( the remaining measure factor @xmath217 will be determined in the next lecture . ) since this is a light - cone gauge quantization , the procedure is especially simple .
there are no ` vacuum ' diagrams , so one just uses the simple feynman diagrammatic expansion of the s - matrix : the only interaction vertex is a simple string splitting or joining .
in the last lecture we wrote the cubic string vertex as a squeezed state in the three - string fock space : @xmath237 |0_{(1)}\rangle @xmath238 is a measure factor which we have not yet determined and @xmath239 is the dimensionality of space time .
this enters the formula because one gets one factor of @xmath240 for each dimension transverse to the light cone .
finally we have made the convenient definition @xmath241 the matrix element @xmath242 expresses the coupling between mode @xmath230 on string @xmath202 and mode @xmath9 on string @xmath243 .
these coefficients are called neumann coefficients . although the @xmath244 matrices are independent of @xmath35 , the matrix @xmath245 depends on @xmath35 ( and the three @xmath246 s ) in a highly nontrivial way . in the @xmath103 limit ,
it is rather easy to show that these neumann coefficients reduce correctly to the flat space case , where explicit formulas are known for @xmath242 ( see the papers reproduced in @xcite ) .
a huge technology has been developed towards obtaining explicit formulas for @xmath242 as a function of @xmath35 and @xmath247 .
this material is too technical to present in detail , so we will just summarize the current state of the art @xcite .
recall that the dual bmn gauge theory is believed to be effectively perturbative in the parameter @xmath248 so , in order to make contact with perturbative gauge theory calculations , we are particularly interested in studying string interactions in the large @xmath35 limit . in this limit it can be shown that @xmath249 \cr & & \qquad\qquad+ { \cal{o}}(e^{-2 \pi \mu } , e^{-2 \pi \mu y } , e^{-2 \pi \mu(1-y)}).\end{aligned}\ ] ] the first term encodes all orders in a power series expansion in @xmath86 .
specifically , @xmath250 + { \mbox{nonperturbative}}.\ ] ] it is intriguing that the nonperturbative corrections look like d - branes rather than instantons ( i.e. they are @xmath251 rather than @xmath252 ) .
we have only written the neumann coefficient @xmath253 , but in fact it is easily shown that in the @xmath254 limit , @xmath255 while all other components are zero .
this fact actually has a very nice interpretation in the bmn gauge theory , which we present in table [ table3 ] .
@xmath256 & & matrix elements of @xmath160 at @xmath257 & & & & at @xmath258 & @xmath259 & & @xmath260 & & & & & & & matrix elements of @xmath1 , & & & & splitting - joining operator & & & & & & our vertex ( [ eq : vertex ] ) still has an arbitrary function @xmath217 of the light - cone momenta and @xmath35 , and a factor @xmath261 , which is also a terribly complicated function of the light - cone momenta and @xmath35 . in flat space ( @xmath262 ) , it was shown long ago that lorentz - invariance of the vertex , and in particular , the covariance of s - matrix elements under @xmath263 lorentz transformations , requires @xmath264 and @xmath265 .
this fact is nice for the oscillator representation since these factors then cancel and ( [ eq : vertex ] ) can simply be written as @xmath266 |0_{(1)}\rangle \qquad { \rm for~}\mu=0,\ ] ] with no additional factors ( except perhaps some innocent overall factors like @xmath267 s which we have not carefully kept track of ) .
however , in the functional representation this fact is quite mysterious !
it means that the correct , lorentz - invariant string vertex in flat space , @xmath268\prod_{r=1}^3 \left ( dp_r^+ d x_r(\sigma)\ \phi [ p_r^+ , x_r(\sigma ) ] \right),\ ] ] has a very peculiar measure factor @xmath269 which would have been impossible to guess purely within the functional approach .
moreover , since the function @xmath269 is a highly complicated function of the @xmath270 , if we fourier transform the action ( [ eq : hardact ] ) back to @xmath271 position space , we find that it involves infinitely many @xmath272 derivatives , in a very complicated way .
this tells us that light - cone string field theory is highly non - local in the @xmath271 direction .
the plane wave background with @xmath273 has fewer bosonic symmetries than flat space .
in particular , it does not have the @xmath263 or @xmath274 symmetries .
this means that it is impossible to use mandelstam s method to determine what the corresponding measure factor is when @xmath273 .
our vertex for string interactions in the plane wave background remains ambiguous up to an overall ( possibly very complicated ) function of @xmath275 and @xmath35 .
we determined the form of the vertex by requiring continuity of the string worldsheet , but evidently that is not enough to solve our problem .
in the rest of this lecture we will learn why light - cone string field theory works , and what the physics is that does completely determine the light - cone vertex ( since continuity is not enough ) . to be precise
, we should say that we will discuss the physics which _ in principle _ determines the light - cone vertex uniquely . the actual _ calculation _ of what this overall function is has not yet been performed for @xmath273 , and is likely rather difficult .
( in the supersymmetric theory , it has been speculated that supersymmetry might fix this overall function essentially to 1 , but this has not been proven . )
the first step on this exciting journey into the details of string field theory will be a close look at the four - string scattering amplitude .
we consider a @xmath276 string scattering process at tree level .
this exercise will be useful for showing how to use the formalism of light - cone string field theory to do actual calculations .
without loss of generality we can choose to label the particles so that 1 and 2 are incoming ( positive @xmath156 ) and 3 and 4 are outgoing ( negative @xmath156 ) and furthermore @xmath277 .
the @xmath243-channel amplitude ( see figure [ fig : fig2 ] ) is @xmath278 let us explain each ingredient .
first of all , trivial overall @xmath156-momentum conserving delta functions are always understood but have not been written in order to save space .
the processes @xmath279 and @xmath280 are as indicated , making use of our vertex function @xmath155 . in between these two
we have inserted the light - cone propagator for the intermediate string 5 : @xmath281 by @xmath282 we mean of course the hamiltonian for string 5 : @xmath283 finally , the integral over @xmath284 enforces the physical state condition on the intermediate string by projecting onto those states which satisfy @xmath285 the full amplitude has two additional contributions . in the @xmath286-channel , we have first @xmath287 , and then @xmath288 : @xmath289 finally in the @xmath290-channel , @xmath291 and then @xmath292 : @xmath293 what are these things ?
well , each @xmath294 is just a state in @xmath295 , the fourth power of the string fock space .
if we want to know the amplitude for scattering four particular external states , then we just have to calculate @xmath296 ( summed over channels ) to get the scattering amplitude as a function of @xmath297 and @xmath35 .
it should be emphasized that the harmonic oscillator algebra gives very complicated functions of @xmath298 and @xmath105 which need to be integrated over . actually performing this calculation is far outside the scope of these lectures ( see @xcite ) , but we would like to make one very important point about the general structure of this amplitude .
let us denote @xmath299 \}$ ] , which are the two dimensional regions over which the quantities @xmath300 must be integrated .
there exists a particular map @xmath301 which patches together these three coordinate regions onto a sphere as shown in figure [ fig : fig3 ] .
this much is of course obvious .
let us define @xmath302 to be the image of the three individual @xmath303 , @xmath304 , and @xmath305 on the sphere , patched together via the moduli map .
it turns out that in flat space , precisely in the critical dimension @xmath264 , the function @xmath302 on the sphere is continuous along the boundaries between the images of @xmath306 ( that is , continuous along the dark lines in figure [ fig : fig3 ] ) , which means that the amplitude can be written as @xmath307 the right hand side of ( [ eq : moduli ] ) has a very familiar form .
when we studied string theory , we learned that in order to calculate a four - string amplitude at tree level in closed string theory , one inserts four vertex operators on the sphere .
the positions of three vertex operators can be fixed using the conformal killing vectors , and one is left with some amplitude ( depending on the particular vertex operators inserted ) which must be integrated over @xmath308 , the position of the remaining vertex operator .
the moduli space of a sphere with four marked points ( the positions of vertex operators ) is therefore the sphere itself .
it turns out that the integrand @xmath302 on the right - hand side of ( [ eq : moduli ] ) is precisely the integrand one would derive from the polyakov path integral in the covariant formulation of string theory .
although we have studied only one of the most trivial possible amplitudes , the equation ( [ eq : moduli ] ) indicates a very general feature .
amplitudes calculated in light - cone string field theory , with any number of external states and at arbitrary order in string perturbation theory , are precisely equivalent to those calculated using the covariant polyakov path integral @xcite .
this equivalence relies on two important facts : * property 1 : triangulation of moduli space .
* consider all of the light - cone diagrams which contribute to an amplitude with @xmath94 closed string loops and @xmath9 external particles @xcite .
the diagrams will be labeled by @xmath309 parameters : @xmath94 ` @xmath156-momentum fractions ' , @xmath310 twist angles ( to impose the physical state condition on intermediate string states ) , and @xmath311 interaction times .
the first important fact is that the moduli map provides a one - to - one map between this @xmath312-dimensional parameter space and the moduli space of riemann surfaces of genus @xmath94 with @xmath9 marked points ( the locations of the vertex operators ) . a mathematical way of saying this is that the light - cone vertex provides a triangulation of the moduli space @xmath313 . * property 2 : the measure on moduli space . *
the second important fact is that the integrand of the light - cone vertex , including all of the complicated structure involving the neumann matrices and determinants thereof , maps under the moduli map to precisely the correct integration measure which arises from the polyakov path integral ! the proof of these remarkable facts would take us too far afield , but we can not stress enough the importance of these facts , which are deeply rooted in the underlying beautiful consistency of string theory .
in fact , this equivalence can be used to prove the unitarity of the polyakov path integral @xcite : although the path integral is not manifestly unitary , it is equivalent to the light - cone formalism , which is manifestly unitary ! we are now in a position to answer some questions which may have been bothering some students since the last lecture : why is it sufficient to consider a cubic interaction between the string fields , and why is it sufficient to consider the simplest possible cubic interaction , with only a delta - functional ( and , for example , no derivative terms like @xmath314/\delta x(\sigma_{\rm i})$ ] ) ?
the answer is that the simple cubic interaction is sufficient because ( 1 ) the iterated cubic interaction covers precisely one copy of moduli space and ( 2 ) the vertex we wrote down precisely reproduces the correct integration measure on this moduli space .
we do nt need anything else !
it is sometimes said that the symmetry algebra ( in particular , the supersymmetry algebra , for superstrings ) , uniquely determines the interacting string hamiltonian to all orders in the string coupling .
this is a little bit misleading .
for example , in the supersymmetric theory one could take @xmath315 and then define @xmath316 , and as long as ( anything ) commutes with rotations and translations , one would have a realization of the symmetry algebra ! the symmetry argument , however , provides no motivation for considering only a cubic interaction .
the true criteria are ( 1 ) and ( 2 ) listed above , and fortunately it turns out to be true that ( at least for the bosonic string ) , one can find a purely cubic action with properties ( 1 ) and ( 2 ) .
now , fact number ( 1 ) , that the cubic delta - functional vertex covers moduli space precisely once , is essentially a mathematical theorem about a particular cell decomposition of @xmath313 that holds quite generally @xcite .
however , ( 2 ) can fail in subtle ways in certain circumstances .
in particular , it can happen that one or more of the @xmath300 s has singularities in moduli space .
a typical case might be for example that @xmath317 near @xmath318 , which is not integrable .
this gives rise to divergences in string amplitudes , which need to be corrected by adding new string interactions to the hamiltonian .
however , these interaction terms are always delta - function supported on sets of measure zero ( @xmath318 in this example ) in moduli space , and therefore they do not spoil the beautiful triangulation that the cubic vertex provides . as long as we do nt add any interaction with _ finite _ measure
, the triangulation still works just fine . * definition .
* we define a * contact term * to be any term in the hamiltonian which has support only on a set of measure zero in moduli space
. * corollary . *
all contact terms are divergent .
proof : if they were finite , they would nt give any contribution , there would be no point to include them , since by definition they are integrated over sets of measure zero ! in flat space ,
it is known @xcite that the bosonic string requires no contact terms , while the iib superstring is widely ( though not universally ) believed to require an infinite number of contact terms .
the word ` believe ' can be thought of in the following sense : since the purpose of contact terms is to eliminate divergences ( and indeed we will see how they arise from short - distance singularities on the worldsheet ) , one can think of a contact term as a counterterm in the sense of renormalization .
now , there are infinitely many such counter terms that one can write down for the iib string , and while some of them may have coefficients which are equal to zero , it is widely believed that infinitely many of them will have nonzero coefficients . for example , in open superstring field theory , it was argued in @xcite that a possible contact term in the @xmath243-channel of the @xmath276 amplitude in fact vanishes .
however , it was argued in @xcite that there is a contact term for this process in the @xmath290-channel . for higher amplitudes
the situation is much more complicated and has not been addressed in detail .
a non - zero contact term in the one - loop mass renormalization has been studied in the plane wave background @xcite and will be discussed in the next lecture . for strings in the plane wave background ,
the question of whether property ( 2 ) holds has not been addressed , mostly because we do not have the analogue of the covariant polyakov formalism in which we can actually calculate anything .
first we would need to calculate this overall factor @xmath319 and then see if there are any divergences which give rise to contact terms .
any of the contact terms in iib string theory in flat space will surely give rise to @xmath35-dependent contact terms in the plane wave background . in principle
there could be new contact terms introduced which go to zero in the limit @xmath103 .
certainly we do not know how to disprove such a possibility , but we believe this is unlikely : contact terms may be thought of as coming from short - distance singularities on the string worldsheet , but the addition of a mass parameter @xmath35 on the worldsheet should not affect any of the short - distance behavior . in fact , it is more likely that the opposite is true : that there are infinitely many contact terms in flat space , but all but a finite number vanish in the plane wave background when @xmath35 is large @xcite .
we will have more to say about this in the next lecture .
let s go back to the beginning of lecture 2 , but add fermions to the picture .
we consider now a superparticle on the plane wave solution of iib supergravity .
the physical degrees of freedom of the theory are encoded in a superfield @xmath320 which has an expansion of the form @xcite @xmath321 where @xmath322 is an eight - component so(8 ) spinor ( @xmath323 is short for eight powers of @xmath322 contracted with the fully antisymmetric tensor @xmath324 ) .
initially we allow all the component fields to be complex , but this gives too many components ( 256 bosonic @xmath325 256 fermionic ) so we impose the reality condition @xmath326 which cuts the number of components in half .
note , in particular , that this constraint correctly gives the self - duality condition for the five - form field strength .
when we second quantize , this hermiticity condition implies that the inner product on the string field theory hilbert space @xmath128 is * not * the inner product navely inherited from the single string hilbert space .
instead , @xmath327 where the states @xmath328 and @xmath329 differ by reversing the occupation of all of the fermionic zero modes , i.e. if @xmath330 , then @xmath331 , etc .
the action for the free superparticle is @xmath332 where @xmath41 .
the quantity in brackets is the quadratic casimir of the plane wave superalgebra .
it is straightforward to insert the superfield ( [ eq : superfield ] ) into ( [ eq : superaction ] ) and find the resulting spectrum @xcite .
the action ( [ eq : superaction ] ) of course may also be obtained simply by linearizing the action for iib supergravity around the plane wave background , and the spectrum may be obtained by linearizing the equations of motion around the background and finding the eigenmodes .
this has been worked out in detail in @xcite , but we will use only one fact which emerges from this analysis .
it turns out that there is a unique state with zero energy , which we will call @xmath333 .
the corresponding spacetime field is a linear combination of the trace of the graviton over four of the eight transverse dimensions , @xmath334 , and the components of the four - form gauge potential @xmath335 in the first four directions .
this field lives in the @xmath336 component of the superfield , where we define left and right chirality with respect to @xmath53 ( i.e. , @xmath337 ) .
the only important fact which you might want to keep in mind is that this spacetime field is * odd * under the @xmath338 symmetry which exchanges the two so(4 ) s : @xmath339 the full string theory hamiltonian ( including interactions ) commutes with the @xmath338 operator @xmath74 , and this fact together with ( [ eq : ztwo ] ) can be used to derive useful selection rules for string amplitudes .
when we promote the superfield to string theory , it becomes a functional of the embedding of the string into superspace : @xmath340 $ ] .
the cubic interaction term has a delta - functional for continuity of @xmath168 , and also a delta - functional for the superspace coordinates : @xmath341.\ ] ] one can write this delta - functional in an oscillator representation as a squeezed state involving the fermionic creation operators .
the ` fermionic ' neumann matrices are easily obtained from the bosonic neumann matrices .
from the preceding section one might have the impression that light - cone string field theory for fermionic strings is a relatively trivial modification of the bosonic theory .
unfortunately , this is not true . in the fermionic theory
the cubic string interaction is no longer a simple @xmath198 vertex with derivatives .
instead it is quadratic in string field functional derivatives .
one way to see why this is necessary is to look at the supersymmetry algebra , which constrains the form of the hamiltonian and dynamical supercharges .
the ( relevant part of the ) spacetime supersymmetry algebra is @xmath342 = 0 , \qquad [ \overline{q}{}^- , h ] = 0.\ ] ] at the free level , a realization of this algebra is given by the free hamiltonian @xmath343 we met before , and the free supercharges @xmath344 , which are given by @xmath345 where the worldsheet supercharge is @xmath346.\ ] ] here @xmath91 is the ` fermionic momentum ' conjugate to @xmath322 ( i.e. , it is just @xmath347 ) .
when we turn on an interaction @xmath159 in the hamiltonian , we also need to turn on interactions @xmath348 , @xmath349 in the dynamical supercharges to ensure that the generators @xmath350 provide a ( non - linear ) realization of the supersymmetry algebra ( [ eq : susy ] ) .
now there is a simple argument ( see chapter 11 of @xcite ) which shows that the choice @xmath351 would be incompatible with the supersymmetry algebra .
consider the relation @xmath352 $ ] at first order in the string coupling .
this gives ( via the state - operator correspondence ) @xmath353 now let us consider ( for example ) a matrix element of this relation where we sandwich three on - shell states on the left .
then @xmath354 acts to the left and gives zero , leaving us with only the second term .
now the state @xmath160 indeed is annihilated by the constraints @xmath355 now after looking at ( [ eq : wssc ] ) , the conditions ( [ eq : conser ] ) seem to imply that @xmath356 , and hence that the desired relation ( [ eq : check1 ] ) is true . however , one can check that the operators in ( [ eq : wssc ] ) are actually singular near the interaction point . for example
, we have @xmath357 .
therefore , although @xmath358 vanishes pointwise in @xmath105 ( except at @xmath359 ) , the singular operators nevertheless give a finite contribution when integrated over @xmath105 .
this contribution can be calculated by deforming the @xmath105 contour in an appropriate way and reading off the residue of the pole at @xmath360 . by calculating the residue of this pole
, it can be shown that in order to supersymmetrize the vertex , it is necessary to introduce some operators ( called ` prefactors ' ) @xmath361 , @xmath362 , @xmath363 such that the interacting hamiltonian and supercharges are given by @xmath364 is a second - order polynomial in bosonic mode - creation operators ( the @xmath127 s ) while @xmath365 and @xmath363 are linear in bosonic creation operators .
they also have a very complicated expansion in terms of fermionic modes , and we will not give the complete formula here .
it is essential to note , however , that the last term in ( [ eq : wssc ] ) is non - singular when acting on @xmath160 .
this makes sense , since the parameter @xmath35 introduces a scale in the worldsheet theory , but this should not affect the short distance physics .
therefore the functional form of the prefactor has essentially the same form as in flat space ( there are subtleties in passing from the functional representation to the oscillator representation , though ) .
we have shown that in the functional representation , the cubic interaction between three string super - fields is * not * given simply by the delta - functionals @xmath366 \delta [ \theta_1(\sigma ) + \theta_2(\sigma ) - \theta_3(\sigma ) ] \prod_{r=1}^3 \phi [ x_r(\sigma ) , \theta_r(\sigma)]\ ] ] in addition , there is a complicated combination of functional derivatives acting on the @xmath109 fields , inserted at the point @xmath367 where the strings split .
this interaction point operator is sometimes called the ` prefactor ' .
the presence of this prefactor is associated with the picture changing operator in the covariant formulation .
the prefactor @xmath361 is an operator of weight @xmath368 , which means that at short distances we have @xmath369 .
( in light - cone gauge we do nt have a conformal field theory in the pp - wave , so by ` weight ' we simply mean the strength of the coincident singularity . ) a light - cone string diagram in which two ( or more ) of these prefactors come very close to each other will therefore be divergent .
the simplest example occurs in the two - particle amplitude at one loop ( i.e. , a contribution to the one - loop mass renormalization ) , shown in figure [ fig : fig4 ] .
this amplitude has been studied in the @xmath254 limit of the plane wave background in @xcite .
this amplitude has an integral @xmath370 over the schwinger parameter giving the light - cone time between the splitting and joining ( think of it as coming from the propagator ( [ eq : propagator ] ) for the intermediate state ) , but the integrand is divergent like @xmath371 due to the colliding prefactors .
it is clear that at higher order in the string coupling ( and/or with more external states ) , we can draw diagrams which have arbitrarily many colliding prefactors .
these divergent contributions to string amplitudes must be rendered finite by the introduction of ( divergent ) contact interactions as discussed in the previous lecture .
the belief is that there is a unique set of contact interactions which preserves all the symmetries ( lorentz invariance , supersymmetry ) and which renders all amplitudes finite . but these contact terms are very unwieldy , and almost impossible to calculate explicitly , so they havent really been studied in very much detail
. we will have a little bit more to say about these contact terms below .
in past couple of lectures we have demonstrated how to determine the light - cone hamiltonian of second - quantized iib string theory in the plane wave background .
the two remaining ambiguities are ( 1 ) that we have not determined some overall factor @xmath319 that appears in the cubic coupling , and ( 2 ) that there are ( probably ) infinitely many contact counterterms which need to be added to the action . in principle ( although probably not in practice ) , the light - cone string field theory approach allows one to calculate @xmath40-matrix elements to arbitrary order in the string coupling via a quite straightforward hamiltonian approach : there is a ( large ) hilbert space of states , with a light - cone hamiltonian acting on it , and one can easily apply the rules of quantum mechanical perturbation theory to give the @xmath40-matrix @xmath372 is the transition operator @xmath373 which is usually calculated via the born series @xmath374 where @xmath375 is the ` bare ' propagator . typically , the @xmath40-matrix is the only good ` observable ' of string theory .
local observables are not allowed because string theory is a theory of quantum gravity , and in particular has diffeomorphism invariance .
the question is then , how is the string @xmath40-matrix encoded in the bmn limit of @xmath376 yang - mills theory ?
note that we are not talking about the @xmath40-matrix of the gauge theory ( which does nt exist , since it is a conformal field theory ) .
instead , the question is about how the string theory @xmath40-matrix can be extracted from the bmn limit of the gauge theory .
it has been emphasized by a number of authors that it can be useful to think of gauge theory in the bmn limit as a quantum mechanical system @xcite .
there is a space of states ( the bmn operators ) , an inner product ( the free gauge theory two - point function ) , and a hamiltonian , given by @xmath57 .
perturbation theory in the gauge theory organizes itself into the two parameters @xmath377 which are respectively the effective t hooft coupling and the effective genus counting parameter , respectively .
recall that the spectrum of bmn operators takes the following form @xmath378 where @xmath185 is the number of impurities with phase @xmath9 . to one - loop
, the gauge theory hamiltonian @xmath379 takes the following form @xcite @xmath380 where @xmath381 and @xmath382 respectively increase and decrease the number of traces .
that is , if we act with @xmath381 on a @xmath383-trace operator , then it ` splits ' one of the traces so that we get a @xmath384-trace operator .
similarly @xmath382 ` joins ' two traces .
the first two terms in ( [ eq : gth ] ) are clearly just the first two terms in ( [ eq : spec ] ) , expanded to order @xmath86 , so we have labeled them @xmath385 they constitute the ` free ' hamiltonian .
the third term in ( [ eq : gth ] ) has the structure of a three - string vertex , and incorporates the string interactions , so we have labeled this term @xmath155 .
the next step is to recall that we should keep in mind the basis transformation that appeared in the lectures of h. verlinde in this school . in our first lecture
we identified a precise correspondence between single - trace operators in the gauge theory and states in string theory .
it is natural therefore to identify a double - trace operator in the gauge theory with the corresponding two - particle state in the string theory , etc .
however this identification breaks down at @xmath386 .
one way to see this is to note that in string theory , a @xmath383 particle state and an @xmath387 particle state are necessarily orthogonal ( by construction ) for @xmath388 .
however in the gauge theory , it is not hard to check that the gauge theory overlap ( given by the two - point function in the free theory ) is typically given by @xmath389 it has been conjectured @xcite that one can write an exact formula , valid to all orders in @xmath93 , for the inner product : @xmath390 where @xmath1 is the simple ` splitting - joining ' operator of the bit model ( see table [ table3 ] ) . the inner product ( [ eq : ip ] ) is diagonalized by the basis transformation @xmath391 .
therefore , we propose the following bmn identification at finite @xmath93 @xcite : @xmath392,\cr ( a^i_n a^j_{-n})^\dagger |0 ; p^+\rangle & \longleftrightarrow & s^{-1/2 } \sum_{k=0}^j e^{2\pi i k n / j } { \rm tr } [ \phi^i z^k \phi^j z^{j - k } ] , \cr & { \mbox etc.}&\end{aligned}\ ] ] now the multi - particle states constructed from the operators on the right hand side of this correspondence will have the property that @xmath383-string states are orthogonal to @xmath387-string states for @xmath388 . however , we have lost the identification of ` number of traces ' with ` number of strings ' .
instead , we have something of the form @xmath393\cr & & \qquad\qquad + g_2 \times [ { \mbox{$k{-}1$-trace operator } } + { \mbox{$k{+}1$-trace operator } } ] + \cdots\end{aligned}\ ] ] it is convenient to perform this basis transformation on the operator @xmath145 , to define what we will call the string hamiltonian @xmath394 , given by @xmath395 for some new interaction @xmath396 ( which is easily calculated ) .
now @xmath394 is simply a non - relativistic quantum mechanical hamiltonian , and it is straightforward to derive from it an @xmath40-matrix .
this @xmath40-matrix should be that of iib string theory on the plane wave background .
it is important to recognize that this @xmath40-matrix is not unitarily equivalent to the @xmath40-matrix obtained from the hamiltonian @xmath397 @xcite .
previously we compared contact terms to counterterms , and we saw how certain contact terms in the superstring come about from regulating singularities that arise when two operators collide on the worldsheet .
one could imagine regulating the theory in some way in order to render these divergences finite .
one natural regulator , which is suggested by both the dual gauge theory and the string bit model , is to discretize the worldsheet . discretizing the worldsheet leads to a spacetime hamiltonian which depends on @xmath63 , the total number of bits .
schematically , we might have something like @xmath398 where we suppose that @xmath399 are finite in the @xmath400 limit .
when @xmath63 is finite , the cubic interaction @xmath401 will fail to precisely cover the moduli space , but the additional ` contact terms ' will be finite and will cover regions of moduli space of small but nonzero measure . in the continuum limit @xmath400 , the contact terms become infinite but restricted to regions of measure zero . physical quantities ( such as @xmath40-matrix elements ) should of course be independent of the regulator @xmath63 . the precise way to say
this is that the amplitudes obtained from @xmath402 differ from those obtained from @xmath403 , @xmath404 by something which is brst - exact .
anything brst - exact integrates to zero over moduli space , so this is a sufficient condition for amplitudes to indeed be independent of the cutoff @xmath63 .
the previous two paragraphs have been well - motivated , but not precise : in fact , we do not know how to make precise sense of @xmath405 in string theory .
certainly discretized string theories have been considered in the past ( see for example @xcite ) , but including interactions is frequently problematic .
the bit model @xcite is intended as a step in the direction of constructing @xmath405 .
perhaps , however , the best way to make @xmath405 precise is simply to read it off from the gauge theory !
comparing matrix elements of @xmath57 in the gauge theory to matrix elements of ( [ eq : contact ] ) would let us read off the contact terms perturbatively in the @xmath406 expansion .
in these lectures we have learned that the essence of light - cone string field theory is an interacting string hamiltonian which satisfies properties 1 and 2 from lecture 3 .
furthermore , we learned how to construct a hamiltonian which satsfies these properties , except possibly for measure zero contact interactions for the superstring .
we also learned a little bit about the bmn limit of the @xmath12 su(@xmath11 ) yang - mills theory .
an obvious goal , which has been realized in a number of papers , is to check whether the bmn limit of the gauge theory can reproduce string amplitudes in the plane wave background .
we now summarize the successses of this research program , and end the lectures with a list of open questions .
following closely the original construction of @xcite in flat space , the three - string vertex for type iib superstring field theory in the plane - wave background has been constructed @xcite , including the bosonic and fermionic neumann coefficients , matrix elements of the prefactor , and even explicit formulas for the neumann coefficients to all orders in @xmath86 @xcite .
matrix elements of the cubic interaction in string field theory have been successfully matched to matrix elements of @xmath407 in the dual gauge theory after taking into account the relevant basis transformation discussed above @xcite .
this program is highly developed and further aspects of this approach have been studied in @xcite .
all of the successful checks of this correspondence have been restricted to amplitudes where the number of impurities ( in the gauge theory language ) is conserved . for such amplitudes ,
all calculations so far indicate that the gauge theory can indeed match the string theory prediction , even though we are in a large t hooft coupling limit of the gauge theory .
the successful match works quite generally at first order in @xmath86 and @xmath93 , and at order @xmath408 ( when intermediate impurity number violating processes are omitted ) .
although there is no obstacle in principle to pushing these checks to higher order , the light - cone string field theory becomes prohibitively complicated .
since it is obvious that the repeated splitting and joining ( i.e. , @xmath383-trace operators to @xmath409-trace operators ) in the gauge theory provides , in the large @xmath63 limit , a triangulation of the string theory moduli space ( property 1 ) , these successful tests of the bmn correspondence amount to checking that the gauge theory also knows about the correct measure on moduli space ( property 2 ) , at least in some limits of @xmath86 and @xmath93 parameter space .
we conclude our lectures with a partial list of open problems and interesting directions for further research .
* explore the structure of the string field theory .
* can we say more about light - cone string field theory in the plane wave background ?
in particular , is it possible to determine the measure factor @xmath410 ?
is it possible to fully calculate a 4-particle interaction , or a 1-loop mass renormalization ?
what can be said about the contact terms in the large @xmath35 limit ?
* does an @xmath40-matrix exist in the plane wave ? *
see references @xcite .
obviously , we have assumed in these lectures that the answer is yes .
the existence of an @xmath40-matrix has several very interesting consequences which should be possible to check purely within the gauge theory .
for example , it implies that the one - loop mass renormalization of any @xmath383-particle state should be equal to the sum of the one - loop mass renormalizations of the individual particles .
this is because the existence of an @xmath40-matrix presupposes that the particles can be well - separated from each other ( in the @xmath271 direction ) . for the most trivial case , where one particle has 2 impurities , and the other @xmath411 particles are all in the ground state , @xmath61 $ ] , this has indeed been shown to be true ( it is essentially due to the fact that ` disconnected ' diagrams dominate over ` connected ' ones in the large @xmath63 limit
see @xcite for details ) .
can this proof be generalized to @xmath383 operators , each of which has more than zero ( ideally , arbitrarily many ) impurities ? * calculate the gauge theory inner product .
* the gauge theory inner product is defined as the coefficient of the free ( @xmath412 ) two - point function .
for example , in the simplest case of two vacuum operators it has been shown that @xcite @xmath413 { \rm tr } [ z^j ] \rangle = { \sinh(g_2/2 ) \over g_2/2}.\ ] ] for bmn operators with impurities , the inner product has been calculated only to a couple of orders in @xmath93 .
since this is a free gauge theory calculation , it reduces to a simple gaussian matrix model , with graphs of genus @xmath94 contributing at order @xmath101 . it should be possible to prove ( or disprove ! ) the conjectured formula that the inner product is given in general by @xmath414 , where @xmath1 is the splitting - joining operator of the bit model .
* ` solve ' the quantum mechanics of bmn operators .
* recent studies have uncovered hints of integrable structures in @xmath15 gauge theory .
one interesting question is whether there is any more hidden structure in the quantum mechanical hamiltonian we wrote down .
for example , one can check ( at least in the two - impurity sector ) that the interaction @xmath396 in the string hamiltonian @xmath415 commutes with @xmath1 @xcite ! are there any other hidden symmetries which will allow one to make progress towards solving this quantum mechanics ?
* do higher - point functions play any role in the bmn limit ? * in our discussion we limited our interest purely to two - point functions in the gauge theory ( although we did consider two - point functions of @xmath383-trace operators with @xmath387-trace operators , but that contains only a tiny remnant of the information encoded in the @xmath416-point function ) .
these two - point functions were interpreted ( essentially ) as @xmath40-matrix elements .
it is natural to wonder ( as many papers have ) what ( if any ) role is played by the higher - point functions . in the previous lecture we explained the correspondence between three - point functions , matrix elements of @xmath160 , and matrix elements of @xmath1 at @xmath417 .
we avoided attempting to extend this correspondence to @xmath418 for the following reason . in a conformal field theory ,
three - point functions of operators with well - defined scaling dimension have simple three - point functions .
but bmn operators do * not * have well - defined scaling dimension when @xmath419 ! they are eigenstates of the ` free ' hamiltonian @xmath385 , but they are not eigenstates of @xmath379 .
so in general , the three - point function of bmn operators looks like @xmath420 we do not know how to recover the @xmath421 , @xmath422 , @xmath423 dependence of ( [ eq : three ] ) from the string theory side .
another point to keep in mind is that higher - point functions of bmn operators typically diverge in the large @xmath63 limit @xcite .
an interesting alternative is to consider @xmath9-point functions where 2 operators are bmn , and @xmath424 operators have finite charge ( for example , @xmath425 $ ] ) .
such an @xmath9-point function might be interpreted as an amplitude for propagation from some initial state to some final state , with @xmath424 other operators inserted on the worldsheet which perturb the spacetime away from the pure pp - wave background .
the correspondence between spacetime perturbations and operator insertions can be read off from the familiar dictionary of @xmath10 .
some calculations along these lines have been performed in @xcite .
* define @xmath405 precisely and provide a string theory construction for it .
* equivalently , * can continuum light - cone superstring field theory be honestly discretized ? *
discretized string theories have been studied for a long time ( in particular by thorn @xcite ) , but it seems somewhat problematic to discretize an interacting type iib string . in particular , there are problems with the fermions . there is the usual fermion doubling problem ( see @xcite ) that was mentioned briefly in the lectures of h. verlinde .
there is also the basic question of how to implement the hermiticity constraint ( [ eq : hermiticity ] ) , which in the continuum theory singles out the fermionic zero modes from the non - zero modes .
however , a discretized string does nt really have any fermionic zero modes , there is just one fermionic oscillator on each ` site ' along the string , and the natural ` adjoint ' just takes the adjoint of each fermion at each site , in contrast with ( [ eq : hermiticity ] ) . * which quantities can be calculated perturbatively in the gauge theory ? * it is important to not forget that the bmn limit still involves taking the t hooft coupling to infinity , so the weak coupling expansion is not really valid .
however , it is empirically observed that some quantities , notably the conformal dimensions of bmn operators at @xmath426 : @xmath427 may be calculated at small @xmath91 ( i.e. , in gauge theory perturbation theory ) and finite @xmath63 , and then extrapolated to @xmath428 where magically they agree with the corresponding string calculation !
these quantities are not bps , so we had no right to expect this miracle to occur .
the basic question behind the bmn correspondence is simply this : for which ( if any ! ) other quantities does this miracle occur ?
for example , we commented that there have been successful comparisons of off - shell matrix elements of the light - cone hamiltonian , at one loop ( order @xmath86 ) , but we do nt know if this miracle will continue to hold for more complicated quantities .
this question is essentially the same as : * how is the order of limits problem resolved ? *
the order of limits problem is that in the gauge theory , we want to expand around @xmath429 , which on the string theory corresponds to @xmath430 and seems to be quite a singular limit . in particular , several steps in the derivation of the light - cone string vertex depended on the assertion that at small distances on the worldsheet , the physics is essentially unchanged by the addition of the parameter @xmath35 .
however this does nt really make sense if @xmath35 is strictly infinity . in particular
, the prefactor of the superstring is a local operator insertion at the point where the string splits .
if we view the gauge theory as a discretized version of the string theory ( with @xmath63 ` bits ' ) , how can the prefactor possibly be recovered when all calculations are necessarily done at finite @xmath63 , where there is no notion of ` locality ' on the worldsheet ?
the successful checks of the bmn correspondence in the literature indicate that this problem is somehow resolved , at least to leading order in @xmath86 .
it is not known whether this continues to hold at higher order in @xmath86 .
certainly , nothing guarantees that the bmn correspondence has to work perturbatively in both parameters @xmath86 and @xmath93 .
clearly , what we would like to do is to somehow calculate gauge theory quantities at finite @xmath91 and @xmath63 , and then take the limit @xmath431 holding @xmath92 fixed .
this leads to the question : * can we turn on finite @xmath91 in the bit model ? * one way that it might be possible to make sense of working at finite @xmath91 in the gauge theory is to work at finite @xmath91 in the bit model @xcite , with the hope that the bit model successfully encapsulates all of the relevant gauge theory degrees of freedom .
let @xmath432 , @xmath433 denote the usual operators living on site @xmath166 on the string and satisfying @xmath434 = \delta_{xy}$ ] . at @xmath435 ,
these operators are just raising and lowering operators of the hamiltonian . that is , creating an excitation at some position @xmath166 on the string precisely increases the energy by one unit .
however at @xmath436 it is no longer the case that the hamiltonian is diagonal in this ` local ' basis . instead
, the hamiltonian is diagonal in the fourier basis , and the eigenmodes are not those which are located at some point @xmath166 on the string , but rather are those with well - defined momentum @xmath9 around the string .
one can turn on finite @xmath91 in the bit model by doing a bogolyubov transformation ; the operator @xmath432 will now be expressed in terms of both raising and lowering operators @xmath437 and @xmath438 .
the cubic vertex , which is trivial at @xmath435 because it is just the delta - function overlap matching excitations pointwise in @xmath105 , needs to be reexpressed in terms of the true eigenmodes of the @xmath439 hamiltonian .
* is there another tool ? * rather than attempting to compare gauge theory calculations to the known but complicated @xcite formulation of string field theory , it might also be possible to derive a string field theory directly from the gauge theory i.e .
, to show that the large @xmath63 limit of gauge theory correlation functions gives a correct measure on moduli space .
some work which is more along these lines includes @xcite .
it is a pleasure to thank y. he , i. klebanov , j. pearson , r. roiban , j. schwarz , d. vaman and h. verlinde for collaboration on material presented in these lectures . m.s .
would like to thank the organizers and staff of the trieste spring school on superstring theory and related topics for hosting a very stimulating workshop . this work has been supported in part by the doe under grant de - fg02 - 91er40671 , and by the national science foundation under grant no .
phy99 - 07949 .
any opinions , findings , and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the national science foundation .
s. mandelstam , `` interacting string picture of dual resonance models , '' nucl .
b * 64 * ( 1973 ) 205 .
s. mandelstam , `` interacting string picture of the neveu - schwarz - ramond model , '' nucl .
b * 69 * ( 1974 ) 77 .
e. cremmer and .
l. gervais , `` combining and splitting relativistic strings , '' nucl .
b * 76 * ( 1974 ) 209 .
e. cremmer and .
l. gervais , `` infinite component field theory of interacting relativistic strings and dual theory , '' nucl .
b * 90 * ( 1975 ) 410 .
r. giles and c. b. thorn , `` a lattice approach to string theory , '' phys .
d * 16 * ( 1977 ) 366 .
c. b. thorn , `` a fock space description of the @xmath440 expansion of quantum chromodynamics , '' phys .
d * 20 * ( 1979 ) 1435
. m. b. green and j. h. schwarz , `` extended supergravity in ten - dimensions , '' phys .
b * 122 * ( 1983 ) 143 .
m. b. green and j. h. schwarz , `` superstring interactions , '' nucl .
b * 218 * ( 1983 ) 43 .
m. b. green , j. h. schwarz and l. brink , `` superfield theory of type ii superstrings , '' nucl .
b * 219 * ( 1983 ) 437 .
m. b. green and j. h. schwarz , `` superstring field theory , '' nucl .
b * 243 * ( 1984 ) 475 .
j. h. schwarz ( editor ) , `` superstrings : the first 15 years of superstring theory , volume 2 '' world scientific ( 1985 ) .
s. b. giddings and s. a. wolpert , `` a triangulation of moduli space from light cone string theory , '' commun . math .
* 109 * ( 1987 ) 177 .
j. greensite and f. r. klinkhamer , `` new interactions for superstrings , '' nucl .
b * 281 * ( 1987 ) 269 .
e. dhoker and s. b. giddings , `` unitarity of the closed bosonic polyakov string , '' nucl .
b * 291 * ( 1987 ) 90 .
j. greensite and f. r. klinkhamer , `` contact interactions in closed superstring field theory , '' nucl .
b * 291 * ( 1987 ) 557 .
m. b. green and n. seiberg , `` contact interactions in superstring theory , '' nucl .
b * 299 * ( 1988 ) 559 .
j. greensite and f. r. klinkhamer , `` superstring amplitudes and contact interactions , '' nucl .
b * 304 * ( 1988 ) 108 . c. b. thorn ,
`` supersymmetric quantum mechanics for string - bits , '' phys .
d * 56 * ( 1997 ) 6619 [ arxiv : hep - th/9707048 ] .
j. m. maldacena , `` the large n limit of superconformal field theories and supergravity , '' adv .
* 2 * , 231 ( 1998 ) [ int .
j. theor .
* 38 * , 1113 ( 1999 ) ] [ arxiv : hep - th/9711200 ] .
r. r. metsaev and a. a. tseytlin , `` type iib superstring action in ads@xmath441 s@xmath442 background , '' nucl .
b * 533 * ( 1998 ) 109 [ arxiv : hep - th/9805028 ] .
r. kallosh , j. rahmfeld and a. rajaraman , `` near horizon superspace , '' jhep * 9809 * ( 1998 ) 002 [ arxiv : hep - th/9805217 ] .
o. aharony , s. s. gubser , j. m. maldacena , h. ooguri and y. oz , `` large n field theories , string theory and gravity , '' phys . rept .
* 323 * ( 2000 ) 183 [ arxiv : hep - th/9905111 ] .
r. roiban and w. siegel , `` superstrings on @xmath10 supertwistor space , '' jhep * 0011 * ( 2000 ) 024 [ arxiv : hep - th/0010104 ] .
m. blau , j. figueroa - ofarrill , c. hull and g. papadopoulos , `` a new maximally supersymmetric background of iib superstring theory , '' jhep * 0201 * ( 2002 ) 047 [ arxiv : hep - th/0110242 ] .
r. r. metsaev , `` type iib green - schwarz superstring in plane wave ramond - ramond background , '' nucl .
b * 625 * ( 2002 ) 70 [ arxiv : hep - th/0112044 ] .
m. blau , j. figueroa - ofarrill , c. hull and g. papadopoulos , `` penrose limits and maximal supersymmetry , '' class .
* 19 * ( 2002 ) l87 [ arxiv : hep - th/0201081 ] .
e. dhoker and d. z. freedman , `` supersymmetric gauge theories and the ads / cft correspondence , '' arxiv : hep - th/0201253 .
d. berenstein , j. m. maldacena and h. nastase , `` strings in flat space and pp waves from @xmath12 super yang mills , '' jhep * 0204 * ( 2002 ) 013 [ arxiv : hep - th/0202021 ] .
r. r. metsaev and a. a. tseytlin , `` exactly solvable model of superstring in plane wave ramond - ramond background , '' phys . rev .
d * 65 * ( 2002 ) 126004 [ arxiv : hep - th/0202109 ] .
m. blau , j. figueroa - ofarrill and g. papadopoulos , `` penrose limits , supergravity and brane dynamics , '' class .
* 19 * ( 2002 )
4753 [ arxiv : hep - th/0202111 ] . j. g. russo and a. a. tseytlin , `` on solvable models of type iib superstring in ns - ns and r - r plane wave backgrounds , '' jhep * 0204 * ( 2002 ) 021 [ arxiv : hep - th/0202179 ] .
m. hatsuda , k. kamimura and m. sakaguchi , `` from super-@xmath10 algebra to super - pp - wave algebra , '' nucl .
b * 632 * ( 2002 ) 114 [ arxiv : hep - th/0202190 ] .
j. michelson , `` ( twisted ) toroidal compactification of pp - waves , '' phys .
d * 66 * ( 2002 ) 066002 [ arxiv : hep - th/0203140 ] .
s. r. das , c. gomez and s. j. rey , `` penrose limit , spontaneous symmetry breaking and holography in pp - wave background , '' phys .
d * 66 * ( 2002 ) 046002 [ arxiv : hep - th/0203164 ] .
n. berkovits , `` conformal field theory for the superstring in a ramond - ramond plane wave background , '' jhep * 0204 * ( 2002 ) 037 [ arxiv : hep - th/0203248 ] .
e. kiritsis and b. pioline , `` strings in homogeneous gravitational waves and null holography , '' jhep * 0208 * ( 2002 ) 048 [ arxiv : hep - th/0204004 ] .
r. g. leigh , k. okuyama and m. rozali , `` pp - waves and holography , '' phys .
d * 66 * ( 2002 ) 046004 [ arxiv : hep - th/0204026 ] .
m. spradlin and a. volovich , `` superstring interactions in a pp - wave background , '' phys . rev .
d * 66 * ( 2002 ) 086004 [ arxiv : hep - th/0204146 ] . c. kristjansen , j. plefka , g. w. semenoff and m. staudacher , `` a new double - scaling limit of @xmath12 super yang - mills theory and pp - wave strings , '' nucl .
b * 643 * ( 2002 ) 3 [ arxiv : hep - th/0205033 ] .
d. berenstein and h. nastase , `` on lightcone string field theory from super yang - mills and holography , '' arxiv : hep - th/0205048 .
d. j. gross , a. mikhailov and r. roiban , `` operators with large r charge in n = 4 yang - mills theory , '' annals phys .
* 301 * ( 2002 ) 31 [ arxiv : hep - th/0205066 ] . n. r. constable , d. z. freedman , m. headrick , s. minwalla , l. motl , a. postnikov and w. skiba , `` pp - wave string interactions from perturbative yang - mills theory , '' jhep * 0207 * ( 2002 ) 017 [ arxiv : hep - th/0205089 ] .
r. gopakumar , `` string interactions in pp - waves , '' phys .
* 89 * ( 2002 ) 171601 [ arxiv : hep - th/0205174 ] .
y. kiem , y. kim , s. lee and j. park , `` pp - wave / yang - mills correspondence : an explicit check , '' nucl .
b * 642 * ( 2002 ) 389 [ arxiv : hep - th/0205279 ] .
m. huang , `` three point functions of @xmath12 super yang mills from light cone string field theory in pp - wave , '' phys .
b * 542 * ( 2002 ) 255 [ arxiv : hep - th/0205311 ] . c. s. chu , v. v. khoze and g. travaglini , `` three - point functions in @xmath12 yang - mills theory and pp - waves , '' jhep * 0206 * ( 2002 ) 011 [ arxiv : hep - th/0206005 ] .
h. verlinde , `` bits , matrices and 1/n , '' arxiv : hep - th/0206059 .
p. lee , s. moriyama and j. park , `` cubic interactions in pp - wave light cone string field theory , '' phys .
d * 66 * ( 2002 ) 085021 [ arxiv : hep - th/0206065 ] .
m. spradlin and a. volovich , `` superstring interactions in a pp - wave background .
ii , '' jhep * 0301 * ( 2003 ) 036 [ arxiv : hep - th/0206073 ] .
a. santambrogio and d. zanon , `` exact anomalous dimensions of @xmath12 yang - mills operators with large @xmath443 charge , '' phys .
b * 545 * ( 2002 ) 425 [ arxiv : hep - th/0206079 ] .
c. s. chu , v. v. khoze and g. travaglini , `` pp - wave string interactions from @xmath9-point correlators of bmn operators , '' jhep * 0209 * ( 2002 ) 054 [ arxiv : hep - th/0206167 ] .
i. r. klebanov , m. spradlin and a. volovich , `` new effects in gauge theory from pp - wave superstrings , '' phys .
b * 548 * ( 2002 ) 111 [ arxiv : hep - th/0206221 ] .
m. huang , `` string interactions in pp - wave from @xmath12 super yang mills , '' phys
. rev .
d * 66 * ( 2002 ) 105002 [ arxiv : hep - th/0206248 ] .
a. parnachev and a. v. ryzhov , `` strings in the near plane wave background and ads / cft , '' jhep * 0210 * ( 2002 ) 066 [ arxiv : hep - th/0208010 ] .
u. gursoy , `` vector operators in the bmn correspondence , '' arxiv : hep - th/0208041 .
n. berkovits and j. maldacena , `` n = 2 superconformal description of superstring in ramond - ramond plane wave backgrounds , '' jhep * 0210 * ( 2002 ) 059 [ arxiv : hep - th/0208092 ] . c. s. chu , v. v. khoze , m. petrini , r. russo and a. tanzini , `` a note on string interaction on the pp - wave background , '' arxiv : hep - th/0208148 . n. beisert , c. kristjansen , j. plefka , g. w. semenoff and m. staudacher , `` bmn correlators and operator mixing in n = 4 super yang - mills theory , '' nucl .
b * 650 * ( 2003 ) 125 [ arxiv : hep - th/0208178 ] .
j. h. schwarz , `` comments on superstring interactions in a plane wave background , '' jhep * 0209 * ( 2002 ) 058 [ arxiv : hep - th/0208179 ] .
a. pankiewicz , `` more comments on superstring interactions in the pp - wave background , '' jhep * 0209 * ( 2002 ) 056 [ arxiv : hep - th/0208209 ] .
d. j. gross , a. mikhailov and r. roiban , `` a calculation of the plane wave string hamiltonian from n = 4 super - yang - mills theory , '' arxiv : hep - th/0208231 .
zhou , `` pp - wave string interactions from string bit model , '' phys .
d * 67 * ( 2003 ) 026010 [ arxiv : hep - th/0208232 ] .
n. r. constable , d. z. freedman , m. headrick and s. minwalla , `` operator mixing and the bmn correspondence , '' jhep * 0210 * ( 2002 ) 068 [ arxiv : hep - th/0209002 ] .
p. lee , s. moriyama and j. park , `` a note on cubic interactions in pp - wave light cone string field theory , '' phys .
d * 67 * ( 2003 ) 086001 [ arxiv : hep - th/0209011 ] .
r. de mello koch , a. jevicki and j. p. rodrigues , `` collective string field theory of matrix models in the bmn limit , '' arxiv : hep - th/0209155 .
d. vaman and h. verlinde , `` bit strings from n = 4 gauge theory , '' arxiv : hep - th/0209215 .
b. eynard and c. kristjansen , `` bmn correlators by loop equations , '' jhep * 0210 * ( 2002 ) 027 [ arxiv : hep - th/0209244 ]
. s. dobashi , h. shimada and t. yoneya , `` holographic reformulation of string theory on ads@xmath441 s@xmath442 background in the pp - wave limit , '' nucl .
b * 665 * ( 2003 ) 94 [ arxiv : hep - th/0209251 ] .
r. a. janik , `` bmn operators and string field theory , '' phys .
b * 549 * ( 2002 ) 237 [ arxiv : hep - th/0209263 ] .
j. pearson , m. spradlin , d. vaman , h. verlinde and a. volovich , `` tracing the string : bmn correspondence at finite @xmath444 , '' arxiv : hep - th/0210102 .
j. gomis , s. moriyama and j. park , `` sym description of sft hamiltonian in a pp - wave background , '' arxiv : hep - th/210153 .
a. pankiewicz and b. stefaski , `` pp - wave light - cone superstring field theory , '' nucl .
b * 657 * ( 2003 ) 79 [ arxiv : hep - th/0210246 ] .
n. beisert , `` bmn operators and superconformal symmetry , '' arxiv : hep - th/0211032 .
d. bak and m. m. sheikh - jabbari , `` strong evidence in favor of the existence of s - matrix for strings in plane waves , '' jhep * 0302 * ( 2003 ) 019 [ arxiv : hep - th/0211073 ] . c. s. chu , m. petrini , r. russo and a. tanzini , `` string interactions and discrete symmetries of the pp - wave background , '' class .
grav . * 20 * ( 2003 ) s457 [ arxiv : hep - th/0211188 ] .
y. h. he , j. h. schwarz , m. spradlin and a. volovich , `` explicit formulas for neumann coefficients in the plane wave geometry , '' arxiv : hep - th/0211198 .
y. kiem , y. kim , j. park and c. ryou , `` chiral primary cubic interactions from pp - wave supergravity , '' jhep * 0301 * ( 2003 ) 026 [ arxiv : hep - th/0211217 ] .
r. roiban , m. spradlin and a. volovich , `` on light - cone sft contact terms in a plane wave , '' arxiv : hep - th/0211220 .
u. gursoy , `` predictions for pp - wave string amplitudes from perturbative sym , '' arxiv : hep - th/0212118 . j. a. minahan and k. zarembo , `` the bethe - ansatz for @xmath12 super yang - mills , '' jhep * 0303 * ( 2003 ) 013 [ arxiv : hep - th/0212208 ]
. n. beisert , c. kristjansen , j. plefka and m. staudacher , `` bmn gauge theory as a quantum mechanical system , '' phys .
b * 558 * ( 2003 ) 229 [ arxiv : hep - th/0212269 ] . c. s. chu and v. v. khoze , `` correspondence between the 3-point bmn correlators and the 3-string vertex on the pp - wave , '' jhep * 0304 * ( 2003 ) 014 [ arxiv : hep - th/0301036 ] .
j. gomis , s. moriyama and j. park , `` sym description of pp - wave string interactions : singlet sector and arbitrary impurities , '' nucl .
b * 665 * ( 2003 ) 49 [ arxiv : hep - th/0301250 ] .
s. bellucci and c. sochichiu , `` fermion doubling and bmn correspondence , '' phys .
b * 564 * ( 2003 ) 115 [ arxiv : hep - th/0302104 ] .
k. bobkov , `` graviton - scalar interaction in the pp - wave background , '' mod .
a * 18 * ( 2003 ) 1441 [ arxiv : hep - th/0303007 ] . n. beisert , c. kristjansen and m. staudacher , `` the dilatation operator of @xmath12 super yang - mills theory , '' arxiv : hep - th/0303060 . c. s. chu , v. v. khoze and g. travaglini , `` bmn operators with vector impurities , @xmath445 symmetry and pp - waves , '' jhep * 0306 * ( 2003 ) 050 [ arxiv : hep - th/0303107 ] .
m. spradlin and a. volovich , `` note on plane wave quantum mechanics , '' arxiv : hep - th/0303220 .
di vecchia , j. l. petersen , m. petrini , r. russo and a. tanzini , `` the 3-string vertex and the ads / cft duality in the pp - wave limit , '' arxiv : hep - th/0304025 . b. stefaski jr . , `` open string plane - wave light - cone superstring field theory , '' arxiv : hep - th/0304114 . t. yoneya , `` what is holography in the plane - wave limit of @xmath26/sym@xmath446 correspondence ? , '' arxiv : hep - th/0304183 . c. s. chu and k. kyritsis , `` the string light cone in the pp - wave background , '' phys .
b * 566 * ( 2003 ) 240 [ arxiv : hep - th/0304191 ] .
d. z. freedman and u. gursoy , `` instability and degeneracy in the bmn correspondence , '' jhep * 0308 * ( 2003 ) 027 [ arxiv : hep - th/0305016 ] .
g. dappollonio and e. kiritsis , `` string interactions in gravitational wave backgrounds , '' arxiv : hep - th/0305081 .
i. bena , j. polchinski and r. roiban , `` hidden symmetries of the @xmath10 superstring , '' arxiv : hep - th/0305116 . n. mann and j. polchinski , `` ads holography in the penrose limit , '' arxiv : hep - th/0305230 .
j. maldacena , m. spradlin and a. volovich , unpublished .
j. gomis , s. moriyama and j. park , `` open + closed string field theory from gauge fields , '' arxiv : hep - th/0305264 .
u. danielsson , f. kristiansson , m. lubcke and k. zarembo , `` string bits without doubling , '' arxiv : hep - th/0306147 .
g. georgiou , v. v. khoze and g. travaglini , `` new tests of the pp - wave correspondence , '' arxiv : hep - th/0306234
. n. beisert , `` the complete one - loop dilatation operator of @xmath12 super yang - mills theory , '' arxiv : hep - th/0307015 .
a. pankiewicz , `` strings in plane wave backgrounds , '' arxiv : hep - th/0307027 . c. g. callan , h. k. lee , t. mcloughlin , j. h. schwarz , i. swanson and x. wu , `` quantizing string theory in @xmath10 : beyond the pp - wave , '' arxiv : hep - th/0307032 .
j. c. plefka , `` lectures on the plane - wave string / gauge theory duality , '' arxiv : hep - th/0307101 . s. bellucci and c. sochichiu , `` can string bits be supersymmetric ?
, '' phys .
b * 571 * ( 2003 ) 92 [ arxiv : hep - th/0307253 ] .
a. pankiewicz and b. stefaski , `` on the uniqueness of plane - wave string field theory , '' arxiv : hep - th/0308062 .
l. dolan , c. r. nappi and e. witten , `` a relation between approaches to integrability in superconformal yang - mills theory , '' arxiv : hep - th/0308089 .
y. k. cheung , l. freidel and k. savvidy , `` strings in gravimagnetic fields , '' arxiv : hep - th/0309005 .
r. dijkgraaf and l. motl , `` matrix string theory , contact terms , and superstring field theory , '' arxiv : hep - th/0309238 . | these lecture notes present an elementary introduction to light - cone string field theory , with an emphasis on its application to the study of string interactions in the plane wave limit of ads / cft .
we summarize recent results and conclude with a list of open questions . |
the measure of the star formation rate ( sfr ) in galaxies is based on the analysis of the emission from young stars which escapes the galaxies or is absorbed and re - emitted by the gas or the dust . in this paper , we will focus on uv and ir emissions .
these emissions are closely related by the energetic budget at work in star forming galaxies : the uv light which does not escape the galaxy is absorbed by the interstellar dust and re - emitted in the far - ir .
therefore both emissions originate from the same stellar populations and their comparison is a powerful tracer of the dust attenuation ( buat & xu @xcite , gordon et al .
@xcite ) . they are also closely related to the recent star formation rate over similar timescales ( buat & xu @xcite , kennicutt @xcite ) . in this paper
we will combine the new galex data together with the existing far - ir data from iras .
we have worked on 600 deg@xmath0 observed by galex in nuv ( @xmath1 nm and fuv ( @xmath2 nm ) to build two samples of galaxies . the first one , called the _ nuv selected sample _ includes all the galaxies brighter than m(nuv ) = 16 mag ( ab scale ) , among the 88 selected galaxies ( excluding ellipticals and active galaxies ) only 3 are not detected by iras at 60 @xmath3 .
the second sample , called the _ fir selected sample _ is based on the iras pscz ( saunders et al .
@xcite ) : 118 galaxies from this catalog lie within our galex fields , only 1 is not detected in nuv .
[ [ mean - values - of - the - dust - attenuation ] ] mean values of the dust attenuation + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + for both samples we measure the dust attenuation using the f(dust)/f(nuv ) ratio .
this ratio is a quantitative measure of the dust attenuation at uv wavelengths ( e.g. buat & xu @xcite , gordon et al .
the formulae used here are adapted to the galex bands ( buat et al .
2004 , apjl , in press ) .
the total ( 8 - 1000 @xmath3 ) dust emission is calculated from the fluxes at 60 and 100 @xmath3 using the dale et al .
( @xcite ) recipe .
+ a moderate attenuation is found in the nuv selected sample with @xmath4 mag in nuv and and @xmath5 mag in fuv . as expected the dust attenuation is higher in the fir selected sample with @xmath6 mag in nuv and @xmath7 mag in fuv . + a comparison of the fir and uv luminosity densities from iras ( saunders et al .
@xcite ) and galex ( thilker et al . 2004 , apjl , in press ) leads to a mean dust attenuation at z@xmath80 of 1.1 mag in nuv and 1.6 mag in fuv : the nearby universe is not very obscured .
[ [ the - irx - beta - relation ] ] the `` irx-@xmath9 '' relation + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + from their study of starburst galaxies meurer and coll .
( e.g. meurer et al .
@xcite ) have found a strong correlation between the slope @xmath9 of the uv continuum ( fitted as a power law @xmath10 between 1200 and 2500 ) and f(dust)/f(fuv ) .
they concluded that @xmath9 is a reliable tracer of the dust attenuation in galaxies .
however further works showed that this law is not universal and that ultra - luminous infrared galaxies ( goldader et al .
@xcite ) and normal spiral galaxies ( bell @xcite ) do not follow the relation found for starbursts .
+ the fuv - nuv color from the galex observation is directly linked to @xmath9 ( kong et al .
we have plotted this color versus f(dust)/f(fuv ) for our two samples of galaxies in fig . 1 and compared to the predictions for starburst galaxies .
obviously @xmath9 is not a reliable tracer of the dust attenuation in our nuv and fir selected samples .
moreover , different behaviors are found within both samples : whereas the fir selected galaxies spread over a large area of the diagram most of the nuv selected galaxies lie below the starburst relation .
this may be due to age effects in the stellar populations : assuming a b parameter ( defined as the ratio of the recent sfr divided by the past averaged one ) equal to 0.25 leads to a reasonable fit using the models of kong et al .
uv and total dust emissions can be calibrated in a recent star formation rate assuming a recent ( over @xmath11 years ) star formation history and an initial mass function ( buat & xu @xcite , kennicutt @xcite ) . when using the dust luminosity one must add an additional hypothesis about the absorption of the stellar light by the dust . the classical hypothesis ( also made in this work )
is that all the stellar light from the young stars is absorbed by the dust ( kennicutt @xcite ) . in this paper
we also assume a constant sfr over @xmath12 years , a salpeter imf . using starburst99 synthesis models
we obtain : + @xmath13 and + @xmath14 .
+ in fig 1 are plotted the sfr estimated from the nuv luminosity against the sfr from the dust luminosity . in both samples ,
the _ observed _ nuv luminosity strongly under - estimates the sfr the effect being worse for the fir selected sample .
when the nuv selected sample is corrected for dust attenuation the agreement is very good ( as expected ) between both estimates of the sfr in the fir selected sample since we are dominated in each case by the dust emission .
conversely in the nuv selected sample the sfr estimated from the dust luminosity alone is found to underestimate the sfr as compared to the nuv corrected luminosity .
the discrepancy increases towards the low sfrs to reach a factor 3 for sfr of @xmath8 0.3 m@xmath15 .
therefore , using the dust emission alone to measure the total sfr in all galaxies can be misleading , the best way being to combine uv and ir emissions to estimate reliable sfrs . 1 bell , e. 2002 , _ apj _ 577 , 150 buat , v. , xu , c. 1996 , _ a & a _ , 306 , 61 dale , d. , helou , g. , contursi , a. , silbermann , n. , kolhatkar , s. 2001 , _ apj _ , 549 , 215 goldader , j . , d. , meurer , g. , heckman , t. , m. , seibert , m. , sanders , d. b. , calzetti , d. , steidel , c. , c. 2002 , _ apj _ 568 , 651 gordon , k , clayton , g. , witt , a. , misselt , k 2000 , _ apj _ 533 , 236 kennicutt , r. 1998 , _ araa _ 36 , 189 kong , x. , charlot , s. , brinchmann , j. , fall , m. 2004 , _ mnras _ 349 . 769 meurer , g. r. , heckman , t. m. , & calzetti , d. 1999 , _ apj _ , 521 , 64 saunders , w. et al . 2000 , _ mnras _ 317 , 55 saunders , w. , rowan - robinson , m. , lawrence , a. et al . 1990 , _ mnras _ 242 , 318 | we make use of the on - going all imaging survey of the uv galex satellite cross - correlated with the iras all sky survey to build samples of galaxies truly selected in far - infrared or in ultraviolet .
we discuss the amount of dust attenuation and the star formation rates for these galaxies and compare the properties of the galaxies selected in fir or in uv .
ultraviolet , far - infrared , galaxies , dust attenuation , star formation rates |
an unsolved problem in many - body physics is to augment numerical exact diagonalization , which is only feasible on comparatively small systems , so as to give the maximum information about the thermodynamic limit .
we know from the renormalization group that only a few degrees of freedom , _ viz .
_ those with low energies and long wavelengths , really matter .
hence it should be possible in principle to discard the irrelevant parts of the hilbert space , but no method has been developed in real space for higher - dimensional , interacting lattice models . the density matrix renormalization group ( dmrg)@xcite is the only successful method to date , but it is limited to one - dimensional chains , or two - dimensional systems that can be forced into one dimension as strips.@xcite to get more mileage out of density matrix - based renormalization group methods , surely we must develop a deep understanding of the structure of density matrices of the simplest possible systems , for which analytic results are available to guide us .
therefore , it is appropriate to begin by considering the ground state of a one - dimensional chain of noninteracting spinless fermions described by a nearest - neighbour hopping hamiltonian @xmath3.\ ] ] here we observe that , in one dimension , one has the same fermi sea ground state for a variety of translationally - invariant hamiltonians with hoppings to further neighbors , provided their dispersion relation is monotonic so that the fermi surface occurs at the same wavevector for a given filling .
a ` block ' is then identified within this one - dimensional system by choosing @xmath1 sites that need not be contiguous , following which we can define the many - body density matrix @xmath0 of the block starting from the zero - temperature many - body wavefunction and tracing out all sites outside the chosen block of @xmath1 sites .
the basis of this paper is the simple factorized nature of @xmath0 and its eigenfunctions , which we derived exactly for noninteracting ( spinless or spinfull ) fermions,@xcite by extending a technique introduced by chung and peschel@xcite . in ref .
, it was shown that many - particle density matrix eigenstates are built from a set of single - particle creation operators and eigenvalues , quite analogous to the energy eigenstates of a noninteracting system of fermions .
for a block of @xmath1 sites cut out from a larger overall system , there are @xmath1 such _ pseudo - creation operators _ @xmath4 , and the block s hilbert space is therefore spanned by all @xmath5 products of the @xmath4 s .
the density matrix of the block can be written as @xmath6 which is the exponential of a _ pseudo - hamiltonian _ given by @xmath7 with _ single - particle pseudo - energy spectrum _ @xmath8 .
this paper is devoted entirely to noninteracting fermions , because the analytic results of ref .
permit many calculations that could not be done by numerical brute force in an interacting case , or can be carried out only by quite different methods ( such as monte carlo ) .
furthermore , our hypothetical renormalization or projection algorithm using dm truncation would first be applied to a well - understood system in which the low - energy excitations behave as noninteracting quasiparticles ( as in a fermi liquid or a @xmath9-wave superconducting phase ) .
our imagined numerical method in such a system would extract renormalized pseudo - creation operators @xmath10 which are related to the single - particle operators @xmath11 in the same way that quasiparticle creation operators are related to bare fermion creation operators .
the pseudo - dispersion relation for the renormalized @xmath12 is thus expected to scale in the same fashion as the pseudo - dispersion @xmath8 encoded in of the noninteracting fermions considered in this paper . through the calculations in this paper we aim to understand the analytic structure of @xmath0 , and to begin to understand the quantitative errors due to truncation .
this includes the question : what is the proper measure of error ?
the most familiar measure , the retained fraction of total density matrix weight , does not seem to be the best measure of error , as evidenced by the small errors obtained in section [ sect : cdmts ] for the calculations of the dispersion relation .
another question to be investigated is whether when we are severely truncating the hilbert space , and attempting only to obtain the low - energy excitations the density matrix eigenstates are the optimal basis . as shown by the comparison in section [ sect : cpwts ] , they are certainly superior to the other plausible basis ( plane waves ) .
in section [ sect : dmstructure ] , we summarize first our much improved understanding of the analytic structure of the density matrix for noninteracting fermions ( and by implication for any fermi liquid ) following from and , and elucidate the statistical - mechanical analogy between the density matrix eigenstates and the many - body states of a system of noninteracting spinless fermions . in section [
sect : oldresults ] , we summarize the important results which we obtained in ref . , giving exact relationships between the block density matrix @xmath0 , the green function matrix @xmath13 restricted to the block , as well as their eigenstates and eigenvalues .
then in section [ sect : recipe ] , we develop the main ideas behind a _ operator - based density matrix ( dm ) truncation scheme _ based on the statistical - mechanical analogy described in section [ sect : dmstructure ] .
the relation between the single - particle density matrix eigenstates and single - particle energy eigenstates of a system of noninteracting spinless fermions also suggests how the distribution of single - particle pseudo - energies @xmath8 are expected to scale with the block size @xmath1 .
numerically , a scaling relationship between @xmath8 and @xmath1 was found indeed , for the overall chain at various fillings @xmath2 .
our analytical results from ref .
shed light on this eigenvalue scaling in two ways .
firstly , as in ref . , they allow numerical study of the density matrix for system sizes so large that they would be inaccessible to any other techniques . secondly , the exact connection of the block density matrix @xmath0 to the block green function matrix @xmath13 gives hints about the eigenvalue distribution .
all of these will be discussed further in section [ sect : scaling ] , and the implications of the scaling behaviour of the single - particle pseudo - energies are discussed in section [ sect : implications ] , where we derived the asymptotic behaviour , in the limit of infinite block sizes , of the largest density matrix weight and the truncated weight @xmath14 , which is the sum of weights of the density matrix eigenstates retained in the operator - based dm truncation scheme . compared to the traditional density matrix truncation scheme used in the dmrg , our operator - based density matrix truncation scheme gives for the same number of density matrix eigenstates retained a slightly larger discarded weight @xmath15 ( see fig .
[ fig : neps ] in section [ sect : residue ] ) .
this quantity gives a @xmath16 estimate as an upper bound a worst case scenario for the error incurred when computing the expectation of a most general observable .
as with examples in numerical integration , the performance of an algorithm in integrating some classes of functions may be much better than that expected from the straightforward error analysis . in any case
, we are not really interested in arbitrary observables , but rather , in the dispersion relation of elementary excitations , which we calculate in section [ sect : cdmts ] .
the results are highly encouraging : the dispersion relation calculated in the operator - based dm truncation scheme differ from the true dispersion relation by an amount much smaller than what is suggested by the discarded weight . besides quasiparticle dispersion relations , real space correlation functions are also interesting quantities to calculate , and these invariably depend on the real space structure of the many - body ground state wavefunction . since this wavefunction is to be written in terms of the one - particle density matrix eigenfunctions , it is important to understand the real space structure of these as well .
the one - particle density matrix eigenfunctions kept in our operator - based dm truncation scheme have spatial structures that are very similar to each other . in section
[ sect : wavefunctionscaling ] we look into a representative one - particle density matrix eigenfunction , the pseudo - fermi eigenfunction , for each @xmath1 , and found that they also obey a universal scaling relation .
then in section [ sect : dm2pw ] , we check how well such a truncated basis of one - particle density matrix eigenfunctions can approximate the true single - particle wavefunction at the fermi level , which is a plane wave with wavevector @xmath17 .
we find the approximation to be good even when less than one quarter of the one - particle density matrix eigenfunctions are kept .
this is impressive , considering the fact that in the operator - based dm truncation scheme , the number of many - particle density matrix eigenstates thus represented by the one - particle density matrix eigenstates retained constitutes a miniscule fraction of the total number of density matrix eigenstates .
finally , for systems where we know that the true single - particle wavefunctions are plane waves , it seems _ a priori _ plausible that a plane wave - based operator - based truncation scheme might outperform the operator - based density matrix truncation scheme .
we look into this possibility in section [ sect : cpwts ] , and find that while there are a few aspects in calculating the dispersion relation where the _ operator - based plane wave ( pw ) truncation scheme _ outperforms the operator - based dm truncation scheme , the overall performance of the pw scheme is inferior to the dm scheme .
we then conclude in section [ sect : conclusions ] by summarizing the important results obtained in this paper , and discuss where all these might fit into the numerical analysis of an interacting system .
because the hamiltonian in conserves particle number , the eigenstates of @xmath0 have definite particle number , and may be grouped into various @xmath18-particle sectors , where @xmath19 .
a consequence of our fundamental formulas and is that every eigenstate of @xmath0 has the form @xmath20 each eigenstate is specified by a list of _ pseudo - occupation numbers _
@xmath21 , where @xmath22 for the factors contained in , and is zero otherwise .
furthermore , the corresponding eigenvalue , the _ density matrix weight _ , is simply given by @xmath23,\ ] ] where the quantity @xmath24 appearing in the exponent is the _ total pseudo - energy_. in terms of the single - particle pseudo - energies @xmath25 , the normalization constant of the density matrix in can be written as @xmath26,\ ] ] where the summation is over all @xmath5 combinations of occupations .
it is immediately clear from that the density matrix eigenstate of maximum weight corresponds to the minimum total pseudo - energy .
this is obtained by setting @xmath27 for @xmath28 and @xmath29 for @xmath30 . in complete analogy to the real energy of a noninteracting system of fermions , we simply fill up the _ single - particle pseudo - energy levels ( pels ) _ from the lowest up to a _ pseudo - fermi level_. the maximum density matrix weight always turns out to occur with the block fractional filling that is closest to the bulk filling of the fermi sea ground state .
more generally , the maximum - weight state in the @xmath18-particle sector is obtained by filling the states with the @xmath18 lowest single - particle pseudo - energies .
finally , it is clear that the next - highest weights , or equivalently the next - lowest total pseudo - energies in the @xmath18-particle sector , are obtained by making particle - hole excitations involving the pels near to the last one filled .
the above analogy may be extended to note that is exactly the density matrix that would be obtained ( see for example , ref . ) at temperature @xmath31 if @xmath32 were the hamiltonian .
the reciprocal of the normalization constant @xmath33 of @xmath0 in just corresponds to the grand partition function for the block of @xmath1 sites . among other things
, this implies that @xmath34 , the average particle number in a particular pel , has the functional form of the fermi - dirac distribution .
we will actually apply this idea in a slightly different way , so as to relate the single - particle pseudo - energies @xmath8 for different block sizes .
if we were dealing with an actual hamiltonian , the dispersion relation would imply a density of states which would be multiplied by the system size to obtain the actual distribution of states .
our numerical scaling results in section [ sect : eigscale ] confirm that pseudo - energies behave similarly to real energies . in our earlier work@xcite we obtain an exact formula @xmath35_{ij}{{c_{i}^{\dagger}}}{{c_{j}^{\phantom{\dagger}}}},\ ] ] which , with , relates @xmath0 to the block green function matrix @xmath13 , whose matrix elements are @xmath36 with @xmath37 and @xmath38 restricted to sites within the block . clearly becomes when the pseudo - hamiltonian is diagonalized .
also , tells us that the quadratic form of @xmath32 in and @xmath13 are simultaneously diagonalizable .
if we denote by @xmath39 the single - particle eigenstate of @xmath32 with eigenvalue @xmath8 , then @xmath40 is the eigenvector of @xmath13 with eigenvalue @xmath41 .
the single - particle pseudo - energies are related to the eigenvalues of @xmath13 by @xmath42 or equivalently , @xmath43 i.e. the eigenvalues of @xmath13 are the average pseudo - occupation numbers @xmath34 .
note that we sometimes write @xmath44 to make explicit the dependence on block size @xmath1 .
we will assume that @xmath8 are ordered from the most negative to the most positive values .
another notable result that was derived in ref .
is that @xmath45 along with and , comprises the final ingredients that allow numerical computation of the density matrix even in very large blocks . aside from the possibilities of truncation , through
have completely reduced a @xmath46 diagonalization problem into a @xmath47 problem , a computational shortcut which allows numerical studies of large blocks .
the analytical structure of hints at the proper design of a truncation scheme .
the retained hilbert space of a block _ should not _ be the span of those density matrix eigenstates whose weight exceeds a cutoff . instead , we should implement a ` consistent ' truncation , such that the truncated hilbert space consists of exactly @xmath48 states , built from all combinations of @xmath49 pseudo - creation operators @xmath50 ,
, @xmath51 , acting on a block ` vacuum state ' @xmath52 , and satisfying fermion anticommutation relations . in the traditional density
matrix - based truncation scheme used in dmrg , the recipe for truncation is to sort all density matrix weights in descending order , and then retain only the eigenstates associated with the weights that exceed a certain cut off .
let us refer to this as the _ weight - ranked dm truncation scheme_. in light of our understanding of the structure of the many - body density matrix presented in this paper , we can see that the weight - ranked dm truncation scheme will certainly retain the eigenstate with maximum weight , the pseudo - fermi sea described in section [ sect : dmstructure ] , along with eigenstates that are ` particle excitations ' , ` hole excitations ' and ` particle - hole excitations ' from the pseudo - fermi sea .
if we arrange the entire collection of many - particle density matrix eigenstates into a state graph , then the state graph looks like a @xmath1-dimensional hypercubic lattice near the pseudo - fermi sea .
what the weight - ranked dm truncation does in this state graph picture is to remove nodes , and in effect cut bonds , out from this hypercubic lattice , producing a subgraph that is much less connected and containing tenuous links .
because of this , when the hamiltonian is projected onto the weight - ranked dm truncated basis , spurious interactions are introduced .
we can apply the pseudo - energy analogy in choosing how to truncate , given the form of the density matrix .
it is familiar , in the truncation used in fourier - space - based quantum renormalization groups , to discard all single - particle degrees of freedom except for a shell around the fermi surface . in the same way ,
let us discard all operators @xmath53 as degrees of freedom , except those for which @xmath54 is less than some threshold @xmath55 .
for all other @xmath53 , we ` freeze ' @xmath56 by setting @xmath56 to its ground state value @xmath57 this choice gives the maximum density matrix weight , among the eigenstates having any particular set of @xmath56 for the retained single - particle pseudo - energy levels .
the spirit of this truncation scheme is similar to that used in quantum chemistry@xcite , except that the notion of a fermi surface is more fuzzy in atoms and molecules .
the idea that truncation consists of decreasing the thickness of a shell of wavevectors around the fermi surface appeared in the original renormalization group for a quantum - mechanical solid - state problem ( anderson s poor man s
rg for the kondo problem@xcite ) . this obvious notion
that the action is around the fermi surface necessarily appears in every effective form of truncation intended for a metal ( see for example , ref . and ref .
, among others ) . however , to our knowledge all such schemes used a basis of plane waves or of energy eigenstates .
our variation uses pels in analogy to the use of energy eigenstates in these previous problems . deriving from the density matrix
, it makes sense only in procedures that involving cutting a real - space block out of a larger system . within this _ operator - based density matrix ( dm ) truncation scheme _ , we can define an effective hamiltonian for the truncated hilbert space , just by taking the matrix elements of the true hamiltonian between all the retained states . using an operator - based truncation
, this will have a particularly clean form : first replace each creation operator @xmath58 by the equivalent combination of all @xmath11 ; then replace @xmath59 for each single - particle pseudo - energy @xmath60 ( these are frozen to be always occupied ) , and otherwise remove all terms involving the operators that are truncated .
thus , if the original hamiltonian has at most @xmath49-fermion terms , the same will be true for the truncated hamiltonian .
this prescription shows that such a truncation is possible for general models , once one knows the appropriate density matrix , but in this paper we have applied it only to noninteracting models .
since the many - particle density matrix eigenvalues are built , according to , from the single - particle pseudo - energies , the latter are the focus of our numerical investigations . now
if our entire system consisted of the block in a pure state at @xmath61 , then every eigenvalue @xmath41 of the block green function matrix @xmath13 , being the average pseudo - occupation number @xmath34 of a pel , would either be one or zero . at @xmath62 , @xmath41 follow a fermi - dirac distribution .
we will see later that cutting out a finite block from a @xmath61 system , by tracing over the environment of the block , has a similar effect on the eigenvalues of @xmath13 as would taking @xmath62 when the block is the whole system . in a translationally invariant system with filling @xmath2 ( at @xmath61 ) , a fraction @xmath2 of the eigenvalues of @xmath13 are one , while the rest are zero . cutting out a block of length @xmath1 must smooth out this step ( much as having @xmath62 makes it into a fermi - dirac distribution ) , and we expect the transition from one to zero to occur over a fraction @xmath63 .
this guess was inspired by the analogy of pseudo - energy @xmath8 in to the real energy , which near the fermi level scales linearly with wavevector @xmath63 .
this @xmath64 scaling suggests the conjecture of a scaling form for the single - particle pseudo - energy like @xmath65 , and indeed we find below just such a scaling form .
in this subsection we calculate numerically the eigenvalues @xmath41 of the block green function matrix @xmath13 , and use to compute the single - particle pseudo - energies @xmath8 . for a chain of free spinless fermions in its ground state ,
the matrix elements of the block green function matrix @xmath13 are @xmath66 where @xmath2 is the filling fraction .
[ fig : lambdadiff ] shows how @xmath41 , the eigenvalues of @xmath13 , are distributed for different filling fractions @xmath2 and different block sizes @xmath1 . for different filling fractions and block sizes .
in order to compare @xmath41 for different block sizes , the interval @xmath67 $ ] is rescaled such that @xmath68 . with this rescaling
, @xmath69 always occur at @xmath70 = { \ensuremath{\bar{n}}}$ ] . ] for @xmath71 , our numerical studies suggest a scaling relationship of the form @xmath72 where @xmath73 - \frac{1}{2}$ ] , as shown in fig . [
fig : scalingplot ] .
there are two points on fig .
[ fig : scalingplot ] we would like to note .
first of all , with our choice of the scaling variable @xmath74 , the scaling function @xmath75 always passes through the origin , i.e. @xmath76 secondly , from fig .
[ fig : scalingplot ] , we see that @xmath75 is an odd function , i.e. @xmath77 which is what we would expect from particle - hole symmetry when the overall system is at half - filling , and @xmath75 has a finite positive slope at @xmath78 , i.e. @xmath79 $ ] as a function of the scaling variable @xmath73 - 1/2 $ ] for various block sizes at half - filling , showing a scaling collapse onto the scaling function @xmath75 . for @xmath80 , the largest and smallest pseudo - energies
are severely affected by numerical truncation errors in the diagonalization routines , and thus not shown . ]
similar scaling behaviours of the form @xmath81 are found for all @xmath2 , with the generic scaling variable @xmath82 where @xmath83 plays the role of a fermi wavevector , and a filling fraction - dependent scaling function @xmath84 , as shown in fig .
[ fig : completescaling ] .
the scaling functions continue to satisfy @xmath85 and @xmath86 but @xmath84 is no longer an odd function of @xmath74 for @xmath87 . instead
, the particle - hole symmetry inherent in our model is manifested as @xmath88 , 0.3 , 0.7 and 0.9 , plotted against the scaling variable @xmath89 , for three different block sizes : @xmath90 ( @xmath91 ) , @xmath92 ( @xmath93 ) and @xmath94 ( @xmath95 ) . ] from and , we can write @xmath41 as @xmath96 which tells us that @xmath84 plays the role of a dispersion relation @xmath97 , while @xmath1 plays the role of the inverse temperature @xmath98 .
this confirms our suspicion that the effect of cutting a block out of an overall system in its ground state at @xmath61 is to ascribe to the block an effective temperature . as expected , this effective temperature approaches zero as the block size is increased , since we are keeping more and more information about the overall system , which we know to be at @xmath61 .
having understood that @xmath41 is related to the scaling function @xmath84 as in , we are now ready to investigate the scaling behaviour of the normalization constant @xmath99 , which is related to @xmath41 by .
we do this first at half - filling . as can be seen from fig .
[ fig : lambdadiff ] , at half - filling roughly half of the @xmath41 are approximately one , whereas the other half are approximately zero .
the product @xmath100 is therefore determined primarily by the @xmath101 @xmath41 s that are nearly one . for these eigenvalues , @xmath102 \ll 1 $ ] and thus @xmath103
$ ] ( when it is clear what the filling fraction is , we will drop the argument @xmath2 in @xmath84 to keep the notations neat ) . with this , we find that @xmath104 \\ & \approx \exp\left\{b\int_{-1/2}^0 f(x)\,dx\right\ } \\ & = \exp\left\{-b\int_0^{1/2 } f(x)\,dx\right\ } , \end{split}\ ] ] where we made used the observed odd symmetry of the scaling function in when the overall system is at half - filling , so that the integral within the exponent is positive . from
, we see that @xmath99 decreases exponentially with block size @xmath1 . in general , for @xmath2 not too close to zero , where the argument that those @xmath41 s that are near one makes the dominant contribution holds , we find that @xmath105 where we made use of .
for @xmath2 very close to zero , there are a handful of @xmath41 s of @xmath106 , and the rest are all nearly zero , behaving like @xmath107 $ ] . ignoring these handful of @xmath106 @xmath41 s ,
we find that the contribution to @xmath99 from those @xmath108 is proportional to the product @xmath109 , and so @xmath110.\ ] ] the integral can be evaluated as a cumulant expansion , but we can already see that for large @xmath1 , the integral will not be important , and thus @xmath99 derives most of its value from the few @xmath106 @xmath41 s .
in contrast , when @xmath2 is very close to 1 , then most of the eigenvalues @xmath41 of @xmath13 are close to 1 , and these continue to dominate the product @xmath111 , and the asymptotic formula derived in continues to be valid .
with our understanding of the structure of the many - particle density matrix eigenvalues and eigenstates developed in section [ sect : dmstructure ] , and on the scaling behaviour of the single - particle found in section [ sect : scaling ] , we want to now address the question of how much of the hilbert space we can truncate .
clearly , the answer to this question depends on what measure of error we intend to use as our criteria for judging how well the truncated hilbert space describes the physics associated with the parent model . in this section
we look at the most common measure of error , used in the dmrg@xcite and quantum chemistry calculations:@xcite for a properly normalized density matrix , the density matrix weights @xmath112 satisfy the sum rule @xmath113 if the ordinal numbers @xmath114 are chosen such that @xmath112 is ranked in decreasing order , and a total of @xmath115 density matrix eigenstates are retained , then the truncated weight @xmath116 and the discarded weight @xmath117 are frequently used as figures of merit for the truncation scheme , since for a bounded operator @xmath118 , the truncation error in @xmath119 is @xmath16 .
instead of diving in to look at @xmath14 or @xmath120 , let us consider first a related question : how large is the maximum weight for a block of @xmath1 sites embedded within an overall system of gapless noninteracting spinless fermions ? for our discussions , we will consider the half - filled case ; it will be straightforward to extend the arguments presented below to @xmath87 . for convenience ,
let us take @xmath1 to be even .
odd , apart from the technical annoyance that there are _ two _ many - particle density matrix eigenstates with the largest weight .
] let us denote by @xmath121 , where @xmath122 , the many - particle density matrix eigenstate having the largest weight .
this state is always kept in the operator - based truncation scheme .
recall from section [ sect : consistenttruncation ] that , as shown in fig .
[ fig : fluctuation](b ) , this is the analog among density matrix eigenstates of the fermi sea ground state among energy eigenstates . in this @xmath123-particle state
, the single - particle pseudo - energy is filled up to just before @xmath78 , which we shall call the _ pseudo - fermi level_. , ( b ) @xmath121 and ( c ) @xmath124 , with the largest weights , for a block of @xmath1 ( @xmath1 even ) sites within a overall system that is half - filled . ] the many - particle density matrix eigenstates of the block , of which there are two , having the next largest weights will be called @xmath125 ( fig .
[ fig : fluctuation](a ) ) and @xmath124 ( fig .
[ fig : fluctuation](c ) ) , having respectively one less or one more particle . we can understand the weight ratio @xmath126 or @xmath127 as follows . by
, @xmath121 is the state with @xmath27 for @xmath128 , , @xmath123 and @xmath29 for @xmath129 ,
, @xmath1 .
the state @xmath124 differs only in having @xmath130 , while @xmath131 differs only in having @xmath132 .
near the pseudo - fermi level , the scaling function has a slope of @xmath133 , while the spacing between adjacent pseudo - energies on the rescaled @xmath134 axis is @xmath135 .
thus @xmath136 .
but when the actual filling is @xmath137 , we know by particle - hole symmetry that @xmath138 , so @xmath139 and @xmath140 .
it follows from , that @xmath141 for @xmath142 , would tell us that these ratios are approximately @xmath143 .
this is quite different from what would happen when the ` block ' contains half of the entire system , as considered in the standard dmrg algorithm , or in ref . .
if the fraction ( @xmath144 ) in the block approached one , the state @xmath121 must become the fermi sea ground state of the overall system , and consequently contains all the weight .
if the block is merely a finite fraction of the system , we still expect a much larger ratio than
. it would be interesting to check the behaviour of the ratios in for the case @xmath145 with @xmath146 , but we have not investigated this . for the purpose of understanding the pseudo - energy spectrum of non - interacting systems
better , we also considered the dimerized tight - binding hamiltonian @xmath147\left({{c_{j}^{\dagger}}}{{c_{j+1}^{\phantom{\dagger } } } } + { { c_{j+1}^{\dagger}}}{{c_{j}^{\phantom{\dagger}}}}\right),\ ] ] where the hopping integral @xmath148 is modulated by the @xmath149 term to produce an energy gap .
henceforth we choose the scale of energy to be such that @xmath150 .
this system was solved analytically by gebhard _
et al_,@xcite wherein the hamiltonian can be written as @xmath151 with @xmath152 , where @xmath153 and @xmath154 . in terms of @xmath155 and @xmath97 , we can define an angle @xmath156 such that @xmath157 , and whose sine and cosine we denote as @xmath158 , @xmath159 . in terms of these , the operators @xmath160 and @xmath161 for the upper and lower bands respectively are given by @xmath162 at half - filling , the lower band is completely filled while the upper band is completely empty , and the ground state can written simply as @xmath163 for this ground state , the two - point function is given by @xmath164 using which we can construct the block green function matrix @xmath13 , and hence using the pseudoenergies which correspond to the density matrix eigenvalues . for a fixed block size of @xmath92 , the pseudo - energy spectra for different hoping modulation @xmath165 is shown in fig .
[ fig : dimerphifixedsize ] , compared to that of the gapless case . scaling behaviour of the single - particle pseudo - energies
was found for all @xmath165 , each governed by a different scaling function @xmath166 . the scaling collapse plot for @xmath167
is shown in fig .
[ fig : dimerphifixeddelta ] , compared to the scaling function @xmath168 for the gapless case .
$ ] as a function of the scaling variable @xmath169 , for a block of size @xmath92 , with different hoping modulations @xmath170 .
pseudo - energies for @xmath171 for @xmath172 are not shown because these are severely affected by numerical errors incurred in the numerical integration and diagonalization routines .
the various sets of straight line segments are intended to guide the eye in visualizing the data . ]
$ ] as a function of the scaling variable @xmath169 , for hoping modulation @xmath167 and different block sizes @xmath173 .
the pseudo - energies for @xmath174 are not shown as these are severely affected by numerical errors in the numerical integration and diagonalization routines .
also shown , is an approximate dashed curve for the scaling function @xmath75 for @xmath71 and @xmath175 , obtained from the data for @xmath92 .
the straight line segments are intended to guide the eye in visualizing the data . ]
repeating our analysis for the three density matrix eigenstates with the largest weights , we find again that the ratios of density matrix weights @xmath127 and @xmath176 to be independent of block size @xmath1 when the overall system is at half - filling .
however , these ratios depend strongly on the hopping modulation @xmath165 . as we can see from fig .
[ fig : dimerphifixedsize ] , the slope of the scaling curve at @xmath78 is steeper for larger @xmath165 .
this indicates that everything else being equal for finite @xmath1 a smaller fraction of density matrix states is needed to capture the same total weight if the system is gapped .
we have not investigated the case @xmath177 , as in the standard dmrg algorithm , but we naturally expect the ratio to increase in a gapped system .
thus @xmath121 would be a better approximation to the ground state in a gapped system than in a gapless system , which is known as an empirical fact in the dmrg context .
our approach , if extended to the case @xmath178 , would give an analytic justification for this common observation . for even @xmath1 blocks on a gapless chain of noninteracting spinless fermions
described the hamiltonian , the largest density matrix weight @xmath179 can be numerically computed reliably till @xmath180 , and its dependence on @xmath1 is shown in fig .
[ fig : maxwt ] . also shown in fig .
[ fig : maxwt ] is a fit of the numerical data to @xmath181 where @xmath182 , @xmath183 and @xmath184 are curve - fitting parameters . here
the exponentially decaying term is merely chosen to produce a good curve fit we believe the @xmath1-dependence may be more complex but what is interesting is the fact that @xmath179 tends to a constant , @xmath182 , in the limit of @xmath185 .
we find that we can understand this in terms of the scaling formulas developed so far .
as a function of the block size @xmath1 for @xmath1 even .
the solid curve shown is a fit of the form @xmath186 .
the best fit to this small data set is obtained by neglecting the data points for @xmath187 and @xmath188 , for which we get @xmath189 , @xmath190 and @xmath191 . ] from section [ sect : consistenttruncation ] we saw that the largest many - particle density matrix weight @xmath179 corresponds to the situation for which all pels below the pseudo - fermi level @xmath192 are occupied , and all those above are empty .
this means that @xmath193 using the fact that @xmath33 can be written explicitly as @xmath194 we then find that @xmath195 to evaluate @xmath179 , we evaluate first its logarithm , which is @xmath196 here we make two approximations .
firstly , because of , we know that @xmath197 , and so except for a handful of single - particle pseudo - energies @xmath8 very near @xmath198 , all the exponentials are very small numbers . using the approximation @xmath199 for @xmath200 , we write as @xmath201 secondly , we note that because of , single - particle pseudo - energies far away from @xmath192 will contribute negligibly to the above sum .
for @xmath1 sufficiently large , those single - particle pseudo - energies making significant contribution in will lie within a small interval about @xmath202 where a linear approximation of the form @xmath203 adequately describes the pseudo - dispersion relation .
substituting into , we find then that @xmath204 this is a geometric series which we can readily sum to give @xmath205 i.e. the largest density matrix weight @xmath179 is found to approach a constant value of @xmath206\ ] ] as @xmath185 . from fig .
[ fig : scalingplot ] and fig .
[ fig : completescaling ] , we see that @xmath207 , and so this asymptotic value of @xmath179 is aproximately 0.13 .
this is smaller than the @xmath189 found numerically .
now that we understand more about the scaling behaviour of the largest density matrix weight @xmath179 , let us analyze the discarded weight incurred by the operator - based dm truncation scheme .
we compute numerically the discarded weight incurred by the operator - based dm truncation scheme and that incurred by the weight - ranked dm truncation scheme , and show them in fig .
[ fig : neps ] as a function of the number of many - body states kept as a comparison . as we can see , the discarded weight incurred by the operator - based dm truncation scheme is larger compared to the weight - ranked dm truncation scheme , for the same number of many - particle eigenstates kept .
this is expected , since the weight - ranked dm truncation scheme is by definition the most efficient scheme in exhausting the sum rule given in . in spite of this
seemingly poorer ` convergence ' property , we believe that the operator - based truncation scheme has advantages that can not be reproduced by the weight - ranked truncation scheme , to be argued in detail in section [ sect : cdmts ] . as a function of the number of states kept : weight - ranked ( as done in the dmrg ) , and operator - based . ] writing the total density matrix weight explicitly as @xmath208 where the subscript ` kept ' , ` below ' and ` above ' refer to pels retained , approximated as always occupied and approximated as always empty in the operator - based truncation scheme respectively .
the truncated weight @xmath14 calculated from the operator - based truncation scheme is @xmath209 since @xmath210 , the ratio @xmath211 can be written as @xmath212 this expression has a simple interpretation in terms of the pseudo - occupation numbers @xmath213 . using ,
we find that we can write @xmath14 as @xmath214 i.e. the truncated weight @xmath14 is given by the product of pseudo - occupation numbers @xmath41 of those pels we insist are always occupied , together with the product of the single - hole pseudo - occupation numbers @xmath215 of those pels we insist are always empty . from fig .
[ fig : lambdadiff ] we see that @xmath41 changes fairly rapidly from @xmath216 to @xmath217 , over a small range of pels .
therefore , it appears that there is a fairly large range of @xmath218 s for which @xmath41 is very close to one or very close to zero .
however , this does not mean that we should perform an operator - based truncation scheme keeping only the small number of pels whose @xmath41 s are significantly different from one or zero .
this is because @xmath14 is bounded from above by @xmath219 where @xmath220 is the fraction of pels retained in the operator - based truncation scheme , and @xmath221 because the exponent is @xmath222 , this number can still be very small .
this brings us to the question we posed in the beginning of this section : how much of the hilbert space do we truncate ?
if @xmath14 is the only criterion then we see that a compromise is necessary . for a small block ,
the number of pels with @xmath41 significantly different from one or zero is a sizeable fraction of the total number of pels , but this number is manageable . for a large block ,
the number of pels with @xmath41 significantly different from one or zero is a tiny fraction of the total number of pels , but we still need @xmath220 to be reasonably large for @xmath14 to be appreciable in magnitude .
this of course means that an unmanageably large number of pels has to be retained . to make the above discussions more water - tight , let us make use of the scaling relations obtained thus far to find a formula relating the truncated weight @xmath14 to both the block size @xmath1 and the fraction @xmath220 of pels retained . taking the logarithm of we
find , using the fact that @xmath223 for @xmath218 far below @xmath202 , and @xmath224 for @xmath218 far above @xmath202 , that @xmath225 where @xmath226 is a constant .
if @xmath1 is large and @xmath220 small , then the linear approximation for @xmath8 is valid , in which case the two sums in are equal , and given by @xmath227 with this , we can write @xmath14 as @xmath228,\ ] ] where @xmath229 . we can find @xmath230 by taking the limit @xmath231 , in which case we retain no degree of freedom in the pels . within the operator - based truncation scheme ,
this means that we insist all pels below @xmath198 to be always occupied and all those above @xmath198 to be always empty , i.e. only the density matrix eigenstate with the largest weight is retained , and we should have @xmath232 and so @xmath233.\ ] ] this can be simplified further , using to get @xmath234.\ ] ] in the limit of @xmath235 , we see from the above expression that @xmath14 does not tend to one , but we understand that this is because the linear approximation is only valid for a small range of pels about @xmath198 , i.e. only for small @xmath220 . in this regime , we may further approximate @xmath14 as @xmath236 \\ & \approx w_f \exp\left(\frac{f'(0)}{1 - e^{-f'(0)}}\,l_{\max}\right ) , \end{split}\ ] ] where @xmath237 is the number of pels retained . as we can see , for small @xmath220 , the truncated weight @xmath14 increases exponentially with @xmath49 . also ,
whenever is valid , we get approximately the same truncated weight @xmath14 whether we use @xmath238 and @xmath239 or @xmath240 and @xmath241 .
we will see in section [ sect : cdmts ] that whenever the retained @xmath242 pels lies within the regime where the pseudo - dispersion relation is linear , the truncation errors are essentially determined by @xmath243 .
as noted already in section [ sect : recipe ] , in the many - body eigenstates with largest weights , all the very negative pels will be occupied and all the very positive pels will be empty .
the only pels with significant varying occupancy are those near the pseudo - fermi level . by construction ,
the many - body density matrix eigenstates with large weights constitute the likely configurations of the block .
the difference between the large - weight eigenstates of the @xmath18-particle and @xmath244-particle sectors of the density matrix is in the application of a creation operator @xmath4 such that the pseudo - energy @xmath54 is close to @xmath192 . in real space , it is likeliest that we can add a particle near the ends of the @xmath1-site block , for one can imagine that , in the first configuration , this particle was just past the end in an adjacent block , and we merely hopped it a short distance across the boundary to create the state of @xmath244 particles on the block in question .
it follows that the single - particle eigenfunctions with single - particle pseudo - energies near the pseudo - fermi level have their greatest amplitude near the block s boundaries .
in other words , it is the sites near the end that are most correlated with the environmental information that we discarded . as noted earlier ,
the eigenstates of @xmath0 are all built up from the eigenstates @xmath245 of @xmath13 , which are simultaneously the one - particle eigenstates of @xmath0 .
as such , the effects of basis truncation , particularly in obtaining a truncated expansion of the target state @xmath246 , must be understood in terms of the features of these one - particle eigenstates .
the real - space features of @xmath245 can most easily be understood in terms of the corresponding eigenfunctions @xmath247 , where @xmath248 are sites on the block . at half - filling , the probability densities @xmath249 exhibit particle - hole symmetry , as is shown in fig .
[ fig : chik ] for the example of @xmath250 .
in general , by node counting , we see that the sequence of @xmath1 single - particle eigenfunctions are in one - to - one correspondence with the sequence of @xmath1 plane wave states on the block , where the ordinal number @xmath218 of the single - particle eigenfunctions is related to the wavevector @xmath251 of the plane wave states on the block by @xmath252 ( plotted against the scaling variable @xmath253 ) of @xmath0 on a block of @xmath250 sites at half - filling , showing the particle - hole symmetry of the overall system . the subplots are arranged in order of increasing pseudo - energy . ] for odd @xmath1 , the pseudo - energy @xmath254 sits at the pseudo - fermi level , and we may call the corresponding eigenfunction @xmath255 the _ pseudo - fermi eigenfunction_. the probability density associated with @xmath256 has nodes at every even @xmath38 , as shown in in fig . [
fig : chif ] for the case of @xmath92 . of the pseudo - fermi eigenfunction for @xmath92 , plotted against the scaled variable @xmath257 . ] the most prominent feature of the pseudo - fermi eigenfunction , i.e. the amplitude being strongest near the boundaries of the block , was first observed by white.@xcite this appears to be a generic feature that occurs in both integrable and nonintegrable 1-dimensional systems . using the example of a chain of coupled harmonic oscillators , gaite explained this `` concentration of resolution of quantum states near the boundaries '' as a simple consequence of angular quantization of the density matrix.@xcite to analyze @xmath258 ( where we write the @xmath1 dependence of @xmath256 more explicitly ) more carefully , we first rescale the eigenvectors obtained from @xcite such that @xmath259 for @xmath260 , @xmath261 . for @xmath262 , @xmath263 and the rescaling
can not be carried out as unambiguously as for the @xmath260 series .
this rescaling is harmless , since eigenvectors are only defined up to an arbitrary normalization .
after this trivial rescaling , we find that the pseudo - fermi probability density can be put in a scaling form @xmath264}{\sin^2\pi y},\ ] ] where @xmath265 and @xmath266 is the scaling function shown in fig .
[ fig : envelope ] . for various block sizes
@xmath260 , @xmath261 , compared against @xmath267 and @xmath268 , where @xmath269 is the rescaled coordinate on the block . ] in , @xmath270 is a @xmath1-dependent normalization factor , chosen to ensure that the pseudo - fermi wavefunction given is properly normalized in the limit of @xmath185 , i.e. @xmath271 although we can not compute @xmath270 analytically , we venture a guess to its @xmath1-dependence by noting that the functions @xmath266 and @xmath267 are not very different from the function @xmath268 , and so we expect @xmath272}{\sin^2\pi y } \\ & \sim \sum_{j=1}^b 4 y ( 1 - y ) \frac{\frac{1}{2}[1 - ( -1)^j]}{[4 y ( 1 - y)]^2}\\ & = \frac{1}{4 } \sum_{\text{$j$ odd}}^b \left[\frac{1}{y } + \frac{1}{1 - y}\right ] , \end{split}\ ] ] which we can easily work out to have the form @xmath273}{\sin^2\pi y } \sim b\left(\log_e b + c\right),\ ] ] where @xmath274 is a constant .
numerically , the best fit for @xmath270 in the range of block sizes @xmath275 to @xmath276 is obtained with @xmath277 because of the enhanced amplitude near the edge of the block exhibited in the real - space structure of density matrix eigenfunctions with single - particle pseudo - energies close to the pseudo - fermi level , and conversely , enhanced amplitude near the center of the block exhibited in the real - space structure of density matrix eigenfunctions with single - particle pseudo - energies far away from the pseudo - fermi level , we worry that these eigenfunctions might not be a good basis to use for expanding spatially uniform plane waves , which are the true single - particle energy eigenstates in our model .
we address this concern in section [ sect : dm2pw ] .
in a gapless system , we conjecture that low - lying excitations _ above _
the ground state are built from the same operators as the long - wavelength fluctuations _ within _ the ground state .
this supposition is certainly validated if the system has a continuous symmetry and the long - wavelength modes are goldstone modes . in general
it is justified by the relationships between correlation functions ( for the ground - state fluctuations ) and response functions ( for low - energy excitations ) . despite its poor convergence properties as far as exhausting the sum rule
is concerned , the operator - based truncation would still get the salient features of the physics right .
we check this by projecting the hamiltonian in onto the truncated set of fermion operators @xmath53 , and calculate the dispersion relation therefrom .
there are two physical quantities of interest here : ( a ) for odd number of sites @xmath1 , the middle band crosses the fermi level , and we can ask how the fermi velocity , given by the slope of the dispersion relation at the fermi level , scales with @xmath1 and the fraction @xmath220 of fermion operators kept ; or ( b ) for even @xmath1 , a band gap develops as a result of truncation at the fermi level , and we can ask how the size of this band gap depends on @xmath1 and @xmath220 . for a block of @xmath278 sites , where the effects of truncating 2 , 4 and 6 pels are shown . for this block size , truncating 2 , 4 and 6 pels corresponds to fractions @xmath279 , 0.50 and 0.25 of pels retained . for @xmath279 ,
the energy bands ( dashed curves ) just below and above the fermi level @xmath280 agree with the true dispersion relation ( solid curve ) so well that the difference is not discernible at the scales presented in the figure . ] in fig .
[ fig : ekb8 ] , we show the general features of the dispersion relation @xmath97 calculated within the operator - based truncation scheme , using the example of a block of @xmath278 sites .
apart from the energy gap @xmath281 that opens up at the fermi level @xmath280 , we see that there is a one - to - one correspondence between the pel truncated and the energy band absent from the dispersion relation .
more precisely , if we order the energy bands and the pels from the lowest to the highest as @xmath282 and @xmath283 , then if we truncate pel @xmath8 , the energy band @xmath284 will also be removed from the numerically calculated dispersion relation . for fixed @xmath220 ,
the gap @xmath281 decays exponentially with block size @xmath1 , as is shown in fig .
[ fig : gapdelta ] , i.e. we have @xmath285 where @xmath286 is an attenuation coefficient whose @xmath220-dependence is shown in fig .
[ fig : gapdecay ] . here
we see also that @xmath287 for different @xmath220 appears to converge onto @xmath288 .
of course , there is no physical sense in talking about a block of zero size , but it is nevertheless a useful number to keep in mind when studying the scaling behaviour of @xmath287 as @xmath220 varies .
@xmath289 is approximately 4 , which is the bandwidth of the exact dispersion relation , in all cases . as a function of block size @xmath1 for various constant fractions @xmath290 , @xmath291 , @xmath292 , @xmath293 , @xmath294 , @xmath295 of pels retained .
also shown are fits to the data points for various @xmath220 to @xmath296 . ]
with which the spurious gap @xmath281 decays , as a function of the fraction @xmath220 of pels retained . a cubic spline curve ( dashed curve ) is superimposed on the data points to aid visualization .
also shown ( dotted curve ) is @xmath297 . ] in particular , in the limit of @xmath235 , where all pels are retained , the gap is exactly zero for all nonzero block sizes @xmath1 . in this limit ,
if we start out at a ` gap ' of @xmath289 at a ` block size ' of @xmath298 , then to have @xmath299 at @xmath300 , we need the attenuation coefficient @xmath301 to be infinite , i.e. we expect the limiting behaviour @xmath302 . on the other hand , in the limit of @xmath303 , where we retain none of the pels , it is again physically meaningless to talk of a dispersion relation .
nevertheless , if we pretend that we are able to calculate a ` dispersion relation ' in this limit , then it is reasonable , following the trend observed in fig .
[ fig : gapdelta ] , that the gap never closes , i.e. we expect the limiting behaviour @xmath304 .
these limiting behaviours appear to be borne out in the trend observed in fig .
[ fig : gapdecay ] . another notable feature in fig .
[ fig : gapdecay ] is the fact that @xmath305 for @xmath306 , which is the regime we are most interested to apply the operator - based truncation scheme in . to appreciate the relevance of this observation ,
let us first note from fig .
[ fig : ekb8 ] the general feature that the smaller the gap @xmath281 , the better the truncated dispersion relation matches the true dispersion relation about the fermi level . from
we saw that the truncated weight @xmath14 depends only on the combination @xmath243 in this regime , and as far as @xmath14 is concerned , there is no difference whether we choose to keep 10 out of 100 pels ( @xmath241 ) or 10 out of 200 pels ( @xmath307 ) . here
we see a similar exponential dependence on @xmath49 for the spurious gap @xmath281 that arises due to truncation : if we write @xmath308 in this regime , then @xmath309 .
such an exponential behaviour implies that we have very good control over the numerical accuracy of the dispersion relations in particular near the fermi level - calculated in the operator - based dm truncation scheme .
when the block size @xmath1 is odd , the central energy band crosses the fermi level , and the quantity of interest becomes the fermi velocity @xmath310 .
this can be determined from the truncated dispersion relation by taking the numerical central derivative of the central energy band at @xmath311 . at half - filling , @xmath312
exactly because of particle - hole symmetry .
this feature of the dispersion relation was found to be preserved in the numerical dispersion relations computed within the operator - based dm truncation scheme . on the global scale
, we find numerically that the shifts in the central energy band at the brillouin zone center and brillouin zone edge are such that @xmath313 always .
however , when @xmath1 is large , the numerical diagonalization routine introduces artefacts on the energy scale of @xmath314 , resulting in the locally evaluated @xmath310 coming out to be very slightly less than 2 . as such , we analyze the behaviour of @xmath310 as a function of @xmath1 and @xmath220 only for @xmath315 , as shown in fig .
[ fig : vfdecay ] . calculated from the truncated dispersion relation , as a function of the block size @xmath1 , for various fractions @xmath316 of pels retained
. fits to average exponential decays are also shown . ] as can be seen from fig .
[ fig : vfdecay ] , the difference @xmath317 decays more or less exponentially with @xmath1 for various @xmath220 , i.e. @xmath318 where @xmath319 is the @xmath220-dependent attenuation coefficient for the average exponential decay .
the @xmath220-dependence of @xmath319 is shown in fig .
[ fig : crossover ] . as a function of the fraction @xmath220 of pels retained . also shown ( dashed line )
is the expected behaviour @xmath320 . ]
the eigenfunctions of the untruncated hamiltonian are spatially uniform plane waves , with amplitude @xmath321 on site @xmath38 of the block of @xmath1 sites .
these can be expanded in terms of the density matrix eigenfunctions @xmath322 .
naively , we expect that if we drop those @xmath322 associated with pseudo - energies @xmath323 far from the pseudo - fermi level @xmath198 , as we would in our operator - based truncation scheme which removes these single - particle pseudo - energy levels as degrees of freedom , the remaining terms , all having enhanced amplitudes at the edge of the block , would sum to a function with enhanced amplitude at the edge of the block
. it would therefore seem like we are attempting to approximate a spatially - uniform plane wave with a function with the wrong real - space structure . however , the key insight we gain from our study of block density matrices is that while the system - wide density matrix @xmath324 commutes with @xmath325 in , the block density matrix @xmath0 obtained by tracing down @xmath324 does not commute with @xmath326 , for all @xmath251 .
therefore , after operator - based truncation @xmath327 , we would need to diagonalize @xmath328 to find the truncated dispersion relation @xmath284 .
thus , the function that would approximate the plane wave is not the latter s truncated expansion in terms of the eigenfunction of the one - particle block density matrix , but rather , a particular eigenfunction of @xmath328 , which is an appropriate linear combination of the @xmath322 retained in the operator - based truncation scheme .
we show in fig .
[ fig : dm2pw ] the spatial structure of such a function , for various numbers of density matrix eigenfunctions retained .
as we can see , for a block of @xmath92 sites , keeping 7 density matrix eigenfunctions with pseudo - energies around @xmath198 would produce a decent approximation to the plane wave with @xmath329 . for different numbers of pels retained , for a block of @xmath92 sites at half - filling .
] we speculate that the fact that the operator - based density matrix truncation scheme succeeds so well suggests that appropriate linear combinations of the density matrix eigenfunctions can closely approximate a plane wave with wavevector @xmath330 .
this is only possible by taking the difference of two eigenfunctions so as to cancel the enhancements of the envelope function seen at the block ends ( see section [ sect : wavefunctionscaling ] ) .
indeed , the fact that we get the correct slope @xmath310 of the dispersion near @xmath17 suggests that by taking different weights , a continuously varying effective wavevector can be approximated .
the fact that the goodness of approximation depends only on the number of eigenfunctions kept , means that we approximate the wavefunction about as well in two successive blocks of @xmath1 sites , as we do in one big block of @xmath331 sites .
one could speculate that there might exist some sort of approximate composition formula , analogous to clebsh - gordan formulas for combining angular momenta , that provides the @xmath331-site eigenfunctions in terms of the direct product of the @xmath1-site eigenfunctions .
as we saw in section [ sect : wavefunctionscaling ] , eigenstates of the density matrix @xmath0 are approximately plane waves ( with wavevector @xmath332 determined by the boundary conditions on the block of @xmath1 sites ) modified by some envelope function .
apart from the effects of the envelope functions , the operator - based truncation scheme described above is likened to truncating wavevectors @xmath332 far away from the fermi wavevector @xmath17 .
it is therefore natural to investigate how a operator - based truncation scheme based on plane waves would fare against that based on the density matrix eigenstates .
compared to the _ operator - based dm truncation scheme _ developed above , the most striking feature of the _ operator - based plane wave ( pw ) truncation scheme _ is that it gets the dispersion exactly right at the zone center , as shown for the case of @xmath278 in fig .
[ fig : pwdispersiongapless ] , and for the case of @xmath333 in fig .
[ fig : pwdispersiongapped ] .
we understand this as follows : for a block of @xmath278 sites , where the effect of truncating 2 and 6 plane waves ( pws ) are shown . for @xmath334 ,
the fermi level is located at the zone center , and @xmath97 is always gapless here regardless of the number of plane waves truncated . ] to evaluate the dispersion relation in a blocked description , we start by defining the direct bloch basis states @xmath335 where @xmath336 is the single - particle occupation number basis state at site @xmath337 along the chain . in this basis , the hamiltonian for a chain of @xmath338 non - interacting spinless fermions take on a block diagonal form . diagonalizing the @xmath339 diagonal block @xmath340 for @xmath341
then gives the dispersion relation within the reduced zone scheme .
for the operator - based pw truncation scheme , we need to work with the plane wave states @xmath342 on each block of @xmath1 sites , where the wavevector @xmath332 is determined by periodic boundary condition , i.e. @xmath343 .
these plane wave states are related to the single - particle occupation number basis states by @xmath344 a bloch basis state parallel to can be defined as @xmath345 where @xmath346 . from and , it is easy to see that @xmath347 at the zone center , @xmath348 , and the @xmath339 hamiltonian matrix in the @xmath349 basis that we need to diagonalize becomes @xmath350 it is trivial to check that the eigenstates of this hamiltonian matrix are precisely the plane waves @xmath351 on the block . therefore , in the @xmath352 basis , @xmath326 is diagonal at @xmath348 , and so truncating some plane waves from the hilbert space produces no effect on the dispersion here . to be more precise , in performing truncation ,
a linear subspace of the hilbert space is chosen , and the hamiltonian projected onto this subspace . if @xmath353 is an eigenstate of the full hamiltonian , and if @xmath353 is retained in the truncated hilbert space , then it will continue to be an eigenstate of the truncated hamiltonian , with the same eigenvalue .
for even block sizes with @xmath334 , the fermi level is located at the zone center in the reduced zone scheme , and so there is no energy gap to speak of . on the other hand , for even block sizes with @xmath354 ,
the fermi level is located at the zone boundary . at the zone boundary ,
operator - based pw truncation introduces an energy gap @xmath281 at the fermi level , as shown in fig .
[ fig : pwdispersiongapped ] for @xmath333 . for a block of @xmath333 sites , where the effect of truncating 2 and 6 plane waves ( pws ) are shown . for @xmath354
, the fermi level is located at the zone boundary , and the operator - based plane wave truncation scheme introduces an energy gap @xmath281 here . ] as in the case for the operator - based dm truncation scheme , we investigate the behaviour of the energy gap @xmath281 as a function of the block size @xmath1 for a fixed fraction @xmath355 of block states truncated . however , for the operator - based pw truncation scheme , the number of plane wave states that can be truncated , if @xmath1 is even , is @xmath356 , @xmath357 , 1 , 2 , .
thus the only realizable series of block sizes @xmath1 on which we can perform fixed @xmath355 truncation are of the form @xmath358 , @xmath359 , 3 , .
the fraction @xmath220 of block plane wave states retained is related to the series index @xmath360 by @xmath361 half of these realizable series have block sizes that are multiples of 4 , for which the fermi level is at the brillouin zone center where the dispersion relation we have shown in the previous subsection to be gapless . in this subsection
we are interested in those block sizes for which @xmath360 is an odd integer , since for these block sizes the fermi level is at the brillouin zone boundary , where a gap develops in the dispersion relation as a result of truncation .
the behaviour of @xmath281 as a function of @xmath1 for three series of @xmath360 is shown in fig .
[ fig : pwgap ] . as a function of block size @xmath1 for @xmath362 , 5 and 9 , corresponding to the fractions @xmath363 of block plane wave states retained .
also shown are the fits of the data points to the formula @xmath364 . from the fits
, we have @xmath365 for @xmath366 , @xmath367 for @xmath368 and @xmath369 for @xmath370 . ] as can be seen from fig .
[ fig : pwgap ] , the gap depends on block size as an inverse power law @xmath371 where @xmath372 is a @xmath220-dependent prefactor .
this is in stark contrast to the exponential dependence found for the case of the operator - based density matrix truncation scheme . for odd @xmath1
, we again investigate the behaviour of @xmath310 as a function of @xmath1 for the operator - based plane wave truncation scheme .
the number of block plane waves that can be truncated is @xmath373 , @xmath357 , 1 , 2 , and the series of realizable block sizes are @xmath374 , @xmath362 , 5 , . unlike in the operator - based dm truncation scheme , there appears to be two different systematic behaviours for @xmath375 , one for @xmath376 and another for @xmath377 ( @xmath378 , 2 , ) .
we find that the fermi velocity can be fitted very well to the formula @xmath379 the plots of @xmath380 and @xmath381 as a function of @xmath220 are shown in fig .
[ fig : pwvftilde ] and fig .
[ fig : pwc ] respectively . as a function of @xmath220 in the operator - based pw truncation scheme for both the @xmath382 series and @xmath383 series .
also shown as the dashed curve is @xmath384 $ ] , which appears to fit the data points well near @xmath385 . ] as a function of @xmath220 in the operator - based pw truncation scheme for both the @xmath382 series ( @xmath386 ) and @xmath383 ( @xmath387 ) series . both @xmath386 and @xmath387 appears to be converging towards 2 .
in fact , from the graph we find that @xmath388 . ] as can be seen from fig .
[ fig : pwvftilde ] , the exact value of the fermi velocity is obtained only in the double limit of @xmath185 and @xmath235 .
compared to the operator - based dm truncation scheme , where we manage to achieve the exact fermi velocity for any @xmath220 , this is clearly undesirable .
furthermore , even very close to @xmath385 , the computed fermi velocity approaches the limiting value @xmath380 as @xmath64 .
this is much slower than the exponential convergence of @xmath389 found for the operator - based dm truncation scheme .
to summarize , we have in this paper developed an in - depth understanding of the structure of the eigenvalues and eigenstates of the density matrix @xmath0 of a block of @xmath1 sites embedded within a infinite one - dimensional chain of noninteracting spinless fermions described by the set of fermion operators @xmath390 with dispersion relation @xmath97 . from ref .
we know that the block density matrix @xmath0 can be written in as the exponential of a pseudo - hamiltonian @xmath32 given by acting only within the block , which describes another system of noninteracting spinless fermions with fermion operators @xmath391 and dispersion relation @xmath8 .
we use the prefix _
pseudo- _ when talking about operators @xmath391 and energies @xmath8 to distinguish them from the operators @xmath390 and energies @xmath97 of the system we started out with .
the many - particle eigenstates of @xmath0 can then be interpreted simply as the many - body energy eigenstates of the system of noninteracting spinless fermions described by @xmath32 , and their associated density matrix weights are then their statistical weights in the grand canonical ensemble , _ as though _ the block is at a finite temperature . within this statistical - mechanical picture
, we can apply intuitions learnt from the statistical mechanics of real fermionic systems , and talk about the filling of single - particle pseudo - energy levels ( pels ) as dictated by the fermi - dirac distribution , the pseudo - fermi sea with its pseudo - fermi level @xmath198 and particle - hole excitations of the pseudo - fermi sea . using results obtained in ref .
, we make this statistical - mechanical picture more precise , by identifying the single - particle eigenstates @xmath245 in of @xmath0 as those built up from the eigenvectors @xmath40 , @xmath248 , of the green function matrix @xmath13 restricted to the block .
the eigenvalue @xmath41 of @xmath13 , related to the single - particle pseudo - energies @xmath8 by and , are the average pseudo - occupation numbers of the @xmath218^th^ pel .
the statistical mechanics of real fermionic systems tells us that , at finite temperature , the physically important many - body states are those low - energy particle - hole excitations involving single - particle energy levels within @xmath392 of the fermi level .
single - particle energy levels far away from the fermi level contribute negligibly to the thermodynamic properties of the fermionic system , and are precisely the degrees of freedom to be truncated in a renormalization group analysis .
we capitalize on this insight , and described a recipe for operator - based truncation of the density matrix eigenstates , where out of the @xmath1 pseudo - fermion operators @xmath391 , we truncate those @xmath53 s associated with single - particle pseudo - energies @xmath8 far away from @xmath198 and retain @xmath49 of them with @xmath393 . within this operator - based truncation scheme ,
the effective pseudo - hamiltonian acting on the truncated hilbert space can be made to have the same form as the original pseudo - hamiltonian , in the same spirit of renormalization group transformations in statistical mechanics or quantum field theory .
having laid out the basic principles behind our operator - based dm truncation scheme , we proceeded to look more closely into the distribution of single - particle pseudo - energies @xmath8 , and how these scale with the block size @xmath1 .
there are two related questions that provide the motivation for doing this : ( 1 ) the statistical mechanics of real fermionic systems suggests that single - particle energy levels within @xmath394 of the fermi level are the physically relevant degrees of freedom what then is the effective temperature @xmath395 that we should use as the natural cutoff when performing operator - based truncation on @xmath391 ?
( 2 ) although we have associated the pseudo - dispersion relation @xmath8 with the dispersion relation @xmath97 of a real fermionic system of noninteracting spinless fermions , the wavevector @xmath251 enumerating @xmath97 is an _ intensive _ quantity whereas the ordinal number @xmath218 enumerating @xmath8 is an _ extensive _ quantity what would the intensive analog of @xmath218 that more closely parallels the wavevector @xmath251 , and how would the pseudo - dispersion relation look like in terms of this intensive label ?
the natural answer to the second question would be to write the pseudo - dispersion relation @xmath8 as a function of @xmath134 , totally analogous with how the wavevector @xmath251 is enumerated as @xmath396 for a chain of @xmath338 sites satisfying the born - von karman boundary condition .
in fact , we find strong numerical evidence that suggests that the single - particle pseudo - energies , for various block sizes and fillings , satisfy a scaling relation of the form given in , where the scaling function @xmath84 is the proper analog of the dispersion relation @xmath97 , and the scaling variable @xmath74 given in is the proper analog of the wavevector @xmath251 . from
, we see that the block size @xmath1 plays the role of inverse temperature .
our scaling results in section [ sect : scaling ] indicate that the density matrix eigenstates and eigenvalues behave , as block size @xmath1 is increased , very much as energy eigenstates and eigenvalues do when the system size is increased . in the latter case , we have a dispersion relation and are more or less sampling it at different wavevectors .
it is not quite that simple in the density matrix case , since the scaling function our analog of the energy dispersion relation depends on the filling @xmath2 .
we only note that this analogy still lacks an analytical foundation .
a more penetrating analysis is called for of the relation of @xmath13 to @xmath0 ( or equivalently , the effect on its eigenvalues of restricting @xmath13 to sites on a block ) .
incidentally , we noted that our equation relating @xmath13 to @xmath0 was valid at any temperature , but we assumed zero temperature throughout this paper .
we expect nonzero temperature @xmath397 would become a second scaling variable . since @xmath62 has similar effects on the green function @xmath398 as does the gap introduced in , we expect the scaling also behaves similarly and we did not investigate it . in a real application
, it appears quite unlikely that hamiltonians will be projected directly onto large blocks ( meaning blocks of more than 16 sites ) .
what then is the practical value of extracting scaling forms , if they are unambiguously seen only in blocks of 100 or more sites ?
one answer is that , even though it is an oversimplified cartoon for the non - asymptotic situations in which it usually gets applied , a scaling law is easier to grasp than brute numerical or graphical facts .
the scaling relation is a powerful tool that we can use to derive deeper understanding concerning the structure of the block density matrix , as well as various aspects of truncation .
but in itself , the scaling relation provides only a partial answer to our first question , which is about how much of the hilbert space of many - body states on the block of @xmath1 sites we can truncate . to answer this question more completely , we looked at the three density matrix eigenstates @xmath121 , @xmath125 and @xmath124 with the largest weights . using the scaling relation , we find that the ratios @xmath399 of their weights ( @xmath179 being the largest density matrix weight ) approach a constant , in for a gapless system of noninteracting spinless fermions , as @xmath185 .
the same result was also found for a gapped system of noninteracting spinless fermions , for which we find scaling relations governed by gap - dependent scaling functions .
furthermore , the scaling relation allowed us to conclude that as @xmath185 , the largest density matrix weight @xmath179 approaches a constant given in , and derive approximately the dependence on the number @xmath49 of pels retained and the truncated weight @xmath14 in the operator - based dm truncation scheme .
it is difficult to compare our results with those obtained in the context of the dmrg , because that is an incremental method . rather than obtain the density matrix of a large block all at once
, each iteration of the dmrg takes an approximate density matrix for a block of @xmath1 sites and produces an approximate density matrix for a block of @xmath400 sites .
the fraction of weight kept , which is taken as the figure of merit , refers to the small truncation in each iteration .
the cumulative dmrg truncation might be more appropriate to compare with our results for rather large blocks . nevertheless , let us note that operator - based truncation can be applied independent of whether we use an incremental or one - shot method .
in particular , operator - based truncation could be used in a test run to apply dmrg to a noninteracting fermi chain .
one is given a truncated list of @xmath148 operators @xmath401 where @xmath402 for the original block , and a hamiltonian projected onto it .
one augments this list with the bare creation operators @xmath403 on a new site that will be added , and defines the new hamiltonian by adding the hopping to the new site . in light of the derivation of ref . , we anticipate that the density matrix of the augmented system s ground state must have the same quadratic form , with new operators @xmath404 , which could presumably be obtained merely by diagonalizing the single - particle sector .
one simply deletes the least important member of this list to obtain a new truncated list , no longer than the original one .
this difficulty notwithstanding , we still carried out a naive comparison of the performance of the operator - based dm truncation scheme against the traditional weight - ranked dm truncation scheme used in the dmrg , using the ability to exhaust the sum rule for a given total numer @xmath115 of density matrix eigenstates retained as a criterion .
the conclusion : while the operator - based dm truncation scheme do not exhaust the sum rule as rapidly as the weight - ranked dm truncation scheme , @xmath14 is still of @xmath106 , i.e. the significant parts of the total density matrix weight are ` captured ' by the operator - base dm truncation scheme . however , we believe it is more important to check how well a truncation scheme do by calculating physical quantities , rather than rely solely on the truncated weight @xmath14 as a performance indicator . to this end , we calculated the dispersion relation of elementary excitations within the operator - based dm truncation scheme ( an easy thing to do ) , and found that the error incurred decays exponentially as @xmath49 , the number of pels retained when @xmath405 .
this error is much smaller than @xmath16 , which is expected from a naive analysis based on the discarded weight @xmath15 . here
, there is subtle worry that it may be that truncation works especially well for our chosen hopping hamiltonian is so local .
a hamiltonian with longer range hopping would have the same fermi sea and hence the same density matrix , but the truncation errors might be worse .
it would be desirable to also calculate the dispersion relation within the weight - ranked dm truncation scheme , and compare the results to those obtained within the operator - based dm truncation scheme .
however , in the latter case it is problematic even to define the question , since each retained density matrix eigenstate would be a many - particle state .
the new truncated hamiltonian might be conveniently expressed in terms of the pseudo - creation operators @xmath401 , but many combinations of occupations would not exist
. the situation would be somewhat analogous to taking a simple , noninteracting hopping hamiltonian for spinfull fermions , and imposing a gutzwiller projection ( no doubly occupied sites ) . in effect , the projection made a noninteracting model into an interacting one . similarly a weight - ranked truncation must introduce spurious interactions .
hence , even if a system containing several blocks were to be exactly diagonalized ( using the truncated basis ) we could not immediately identify the elementary excitations .
one would need , for example , to numerically compute a spectral function @xmath406 , where @xmath407 are momentum and energy , and then locate peaks as a function of @xmath408 . on the other hand , a system which is truncated using the operator - based truncation scheme can still be represented by a set of creation and annihilation operators . following this ,
we argued , based on the real - space structure of the one - particle density matrix eigenfunctions shown in section [ sect : wavefunctionscaling ] , that an operator - based truncation scheme can also be defined naturally using the basis formed by single - particle plane wave ( pw ) states on the block of @xmath1 sites .
the dispersion relation was calculated within this operator - based pw truncation scheme , and compared to that calculated within the operator - based dm truncation scheme .
we find that , other than getting the dispersion exactly right at the zone center of the reduced brillouin zone , the operator - based pw truncation scheme is generally inferior to the operator - based dm truncation . instead of decaying exponentially as @xmath49 , the error in the dispersion relation calculated within the operator - based pw truncation scheme decays as a power law @xmath409 , which means that more single - particle basis states need to be retained in the operator - based pw truncation scheme as compared to the operator - based dm truncation scheme .
it is not much harder in principle to analyze the proposed operator - based density matrix truncation scheme for _ noninteracting _ fermions in two dimensions .
however , it will be harder to understand the scaling since we can not simply rank the eigenvalues .
one dimension was special because we know that each successive state has one more node than the previous one .
a further very important difference is that in @xmath411 there are just two fermi points , whereas in @xmath410 there is a fermi surface .
thus , whereas in @xmath411 the density matrix eigenstates near the pseudo - fermi level are related to the energy eigenstates at @xmath330 , in @xmath410 these eigenstates will correspond to mixtures of wavevectors from every piece of the fermi surface .
obviously , noninteracting systems do not require numerical studes , so we must clarify how our results are relevant to the problem of interacting systems .
firstly , many ( gapless ) systems of interest are in a phase fermi liquid , @xmath9-wave superconductor which are noninteracting in the low - energy , long - wavelength limit .
when applied to a fermi liquid system , we expect ( to the extent that the truncation has separated out the low energy modes ) that any iterative renormalization scheme will converge on the noninteracting limit , and it must behave properly in that limit to have even the hope of success .
hence , for a density matrix - based scheme , the first order of business is to study the density matrix for a free fermion ground state ( as in this paper ) or for a bcs state,@xcite i.e. that the density matrix will actually have the simple operator - based form , and hence its many - particle eigenstates of the density matrix really are built from the single - particle eigenstates . for an interacting system
, the truncated basis should of course be constructed using the many - body density matrix for that system ( not the noninteracting system ) .
we expect that the operators generating this many - particle truncated basis will _ not _ just be those that create the 1-particle density matrix eigenstates ( they were in the noninteracting case studied in this paper ) .
more thought will be needed as to discover the best recipe to optimize the truncation rule so as to balance the needs of sectors with different particle numbers in a strongly interacting system .
the simple algebraic structure of the noninteracting density matrix eigenstates , considered in this paper , has motivated investigations ( in progress ) of the relationships among the many - particle density matrix eigenstates for an interacting system .
a separate reason why our results for the noninteracting fermions are relevant to the study of interacting systems is that the scaling behaviour of the noninteracting density matrix should be a good guide to that of interacting systems , although details may differ .
this is in the same sense that mean - field theory is a good guide to the overall pattern of critical phenomena .
however , other interacting models of interest sit at quantum critical points that are not described by quasiparticle interactions , or possess fractionalized excitations . since we do not presently understand the proper way to write their wavefunctions in terms of a spatially blocked basis , nor the proper renormalization step to capture the interblock correlations in the fractionalized systems , we do not know if a plain block density matrix gives the proper basis for truncation of the states . furthermore , in the absence of an analytic construction of the block density matrix , for example , for laughlin s quantum hall wavefunction or the one - dimensional su - schrieffer state , we can not proceed to scaling studies of large blocks like those found in the present paper , but numerical studies of such density matrices might be an illuminating subject for future research . this research is supported by nsf grant dmr-9981744 , and made use of the computing facility of the cornell center for materials research ( ccmr ) with support from the national science foundation materials research science and engineering centers ( mrsec ) program ( dmr-0079992 ) .
r. m. noack , s. r. white , and d. j. scalapino , in _
computer simulation studies in condensed matter physics vii ( proceedings of the seventh workshop , athens , ga , usa , 28 february 4 march 1994 ) _ , edited by d. p. landau , k .- k .
mon , and h .- b .
schttler ( athens , georgia , usa , 1994 ) , pp . 8598 . | in an earlier work [ s. a. cheong and c. l. henley , cond - mat/0206196 ( 2002 ) ] , we derived an exact formula for the many - body density matrix @xmath0 of a block of @xmath1 sites cut out from an infinite chain of noninteracting spinless fermions , and found that the many - particle eigenvalues and eigenstates of @xmath0 can all be constructed out of the one - particle eigenvalues and one - particle eigenstates respectively . in this paper we improved upon this understanding , and developed a statistical - mechanical analogy between the density matrix eigenstates and the many - body states of a system of noninteracting fermions .
each density matrix eigenstate corresponds to a particular set of occupation of single - particle pseudo - energy levels , and the density matrix eigenstate with the largest weight , having the structure of a fermi sea ground state , unambiguously defines pseudo - fermi level .
based on this analogy , we outlined the main ideas behind an operator - based truncation of the density matrix eigenstates , where single - particle pseudo - energy levels far away from the pseudo - fermi level are removed as degrees of freedom .
we report numerical evidence for scaling behaviours in the single - particle pseudo - energy spectrum for different block sizes @xmath1 and different filling fractions @xmath2 . with the aid of these scaling relations , which tells us that the block size @xmath1 plays the role of an inverse temperature in the statistical - mechanical description of the density matrix eigenstates and eigenvalues
, we looked into the performance of our operator - based truncation scheme in minimizing the discarded density matrix weight and the error in calculating the dispersion relation for elementary excitations .
this performance was compared against that of the traditional density matrix - based truncation scheme , as well as against a operator - based plane wave truncation scheme , and found to be very satisfactory . |
let @xmath0 be a dynamical system on a measure space @xmath1 endowed with a @xmath2-algebra @xmath3 , a @xmath2-finite measure @xmath4 and an invertible measure preserving transformation @xmath5 . in 1955
cotlar ( see @xcite ) established the almost everywhere convergence of the ergodic truncated hilbert transform @xmath6 for all @xmath7 with @xmath8 .
the aim of the present paper is to obtain the corresponding result for the set of prime numbers @xmath9 .
let @xmath10 $ ] .
we prove [ thm:1 ] for a given dynamical system @xmath0 the almost everywhere convergence of the ergodic truncated hilbert transform along @xmath9 @xmath11 holds for all @xmath7 with @xmath12 . in view of the transference principle , it is more convenient to work with the set of integers rather than an abstract measure space @xmath1 . in these settings
we consider discrete singular integrals with caldern zygmund kernels .
given @xmath13 satisfying @xmath14 for @xmath15 , together with a cancellation property @xmath16 a singular transform @xmath17 along the set of prime numbers is defined for a finitely supported function @xmath18 as @xmath19 let @xmath20 denote the truncation of @xmath17 , i.e. @xmath21 we show [ thm:2 ] the maximal function @xmath22 is bounded on @xmath23 for any @xmath24 .
moreover , the pointwise limit @xmath25 exists and coincides with the hilbert transform @xmath26 which is also bounded on @xmath23 for any @xmath24 . for @xmath27 ,
the proof of theorem [ thm:2 ] is based on the hardy and littlewood circle method .
these ideas were pioneered by bourgain ( see @xcite ) in the context of pointwise ergodic theorems along integer valued polynomials . for @xmath28 , initially we wanted to follow elegant arguments from @xcite which used very specific features of the set of prime numbers .
however , we identified an issue in @xcite ( see appendix [ apx:1 ] ) which made the proof incomplete .
instead , we propose an approach ( see lemma [ lem:1 ] and [ lem:2 ] ) which rectifies wierdl s proof ( see appendix [ apx:1 ] for details ) as well as simplifies bourgain s arguments .
bourgain s works have inspired many authors to investigate discrete analogues of classical operators with arithmetic features ( see e.g @xcite ) .
nevertheless , not many have been proved for the operators and maximal functions modelled on the set of primes ( see e.g @xcite ) . to the authors
best knowledge , there are no other results dealing with maximal functions corresponding with truncated discrete singular integrals
. it is worth mentioning that theorem [ thm:2 ] extends the result of ionescu and wainger @xcite to the set of prime numbers .
however , our approach is different and provides a stronger result since we study maximal functions corresponding with truncations of discrete singular integral rather than the whole singular integral .
furthermore , we were able to define the singular integral as a pointwise limit of its truncations .
theorem [ thm:2 ] encourages us to study maximal functions associated with truncations of the radon transforms from @xcite . for more details
we refer the reader to the forthcoming article @xcite . throughout the paper ,
unless otherwise stated , @xmath29 stands for a large positive constant whose value may vary from occurrence to occurrence .
we will say that @xmath30 ( @xmath31 ) if there exists an absolute constant @xmath32 such that @xmath33 ( @xmath34 ) .
if @xmath30 and @xmath31 hold simultaneously then we will shortly write that @xmath35
. we will write @xmath36 ( @xmath37 ) to indicate that the constant @xmath32 depends on some @xmath38
. we always assume zero belongs to the natural numbers set @xmath39 .
we start by recalling some basic facts from number theory .
a general reference is @xcite .
given @xmath40 we define @xmath41 to be the set of all @xmath42 $ ] such that @xmath43 . by @xmath4
we denote mbious function , i.e. for @xmath44 where @xmath45 @xmath46 in what follows , significant role will be played by the ramanujan s identity @xmath47 and the mbious inversion formula @xmath48 satisfied by any function @xmath49 .
let @xmath50 be euler s totient function , i.e. for @xmath40 the value @xmath51 is equal to the number of elements in @xmath41 .
then for every @xmath52 there is a constant @xmath53 such that @xmath54 eventually , if we denote by @xmath55 the number of divisors of @xmath56 then for every @xmath52 there is a constant @xmath53 such that @xmath57
the measure space @xmath58 with the counting measure and the bilateral shift operator will be our model dynamical system which permits us to prove theorem [ thm:1 ] .
let us fix @xmath59 $ ] and define a set @xmath60 .
given a kernel @xmath61 satisfying and we consider a sequence @xmath62 where @xmath63,\\ 0 & \text { otherwise . } \end{cases}\ ] ] let @xmath64 denote the fourier transform on @xmath65 defined for any function @xmath66 as @xmath67 if @xmath68 we set @xmath69 then for @xmath70 by integration by parts one can show @xmath71 we define a sequence @xmath72 of multipliers @xmath73 to approximate the multiplier @xmath75 we adopt the argument introduced by bourgain @xcite ( see also wierdl @xcite ) which is based on the hardy littlewood circle method ( see e.g @xcite ) . for any @xmath76 and @xmath77 major arcs are defined by @xmath78 where @xmath79 : { { \lvert { \xi - a / q } \rvert } } \leq \tau^{-j } j^\alpha \big\}.\ ] ] here and subsequently we will treat the interval @xmath80 $ ] as the circle group @xmath81 identifying @xmath82 and @xmath83 .
[ prop:1 ] for @xmath84 @xmath85 the constant @xmath86 depends only on @xmath87 . since for a prime number @xmath88 , @xmath89
if and only if @xmath90 we have @xmath91 let @xmath92 .
if @xmath93 then @xmath94 and consequently @xmath95 using the summation by parts ( see e.g @xcite ) for the inner sum on the right hand side in we obtain @xmath96 where @xmath97 $ ] and for @xmath98 we have set @xmath99 similar reasoning gives @xmath100 by siegel walfisz theorem ( see @xcite ) we know that for every @xmath101 and @xmath98 @xmath102 where the implied constant depends only on @xmath87 . therefore and combined with the estimates and yield @xmath103 eventually , by , @xmath104 next , we can substitute an integral for the sum since for @xmath105 and @xmath106 we have @xmath107 thus @xmath108 repeating all the steps with @xmath88 replaced by @xmath109 we finish the proof .
for @xmath110 we set @xmath111\cap\mathbb{q } : 2^s \leq q < 2^{s+1 } \text { and } ( a , q ) = 1 \big\}.\ ] ] since we treat @xmath80 $ ] as the circle group identifying @xmath82 and @xmath83 we see that @xmath112 .
let us consider @xmath113 where @xmath114 and @xmath115 is a smooth function such that @xmath116 and @xmath117 the value of @xmath118 is chosen to satisfy .
additionally , we may assume ( this will be important in the sequel ) that @xmath119 is a convolution of two smooth functions with compact supports contained in @xmath120 $ ] .
let @xmath121 .
[ prop:2 ] for every @xmath122 @xmath123 the constant @xmath86 depends only on @xmath87 . first of all notice that for a fixed @xmath110 and @xmath124 $ ] the sum consists of the single term .
otherwise , there would be @xmath125 such that @xmath126 and @xmath127 .
therefore , @xmath128 which is not possible whenever @xmath129 , as it was assumed in . *
major arcs estimates : @xmath130 .
* let @xmath131 be such that @xmath132 next , we choose @xmath133 satisfying @xmath134 if @xmath135 then for any @xmath136 , @xmath137 we have @xmath138 therefore , the integration by parts gives @xmath139 combining the last estimate with , we obtain that for some @xmath140 @xmath141 moreover , if @xmath142 then @xmath143 .
by we have @xmath144 .
hence the integration by parts implies @xmath145 in the last estimate it is important that the implied constant does not depend on @xmath131 . since @xmath146 is bounded uniformly with respect to @xmath77 , by and the definition of @xmath133 we have @xmath147 for appropriately chosen @xmath148 . eventually , in view of proposition [ prop:1 ] and definitions of @xmath131 and @xmath133 we conclude @xmath149 * minor arcs estimates : @xmath150 . *
firstly , by the summation by parts , we get @xmath151 where @xmath152 using dirichlet s principle there are @xmath43 , @xmath153 such that @xmath154 thus , by vinogradov s theorem ( see ( * ? ? ?
* theorem 1 , chapter ix ) or ( * ? ? ?
* theorem 8.5 ) ) we get @xmath155 for @xmath156 $ ] . combining @xmath157 with the last bound and
we conclude @xmath158 since @xmath122 . in order to estimate the @xmath159
let us define @xmath133 by setting @xmath160 if @xmath161 for @xmath135 then @xmath162 and @xmath163 then again by the integration by parts we obtain @xmath164 therefore , the first part of the sum may be majorized by @xmath165 as for @xmath166 .
for the second part we proceed as for @xmath167 to get @xmath168 a suitable choice of @xmath169 in both estimates above was possible thanks to .
we start the section by proving two lemmas which will play crucial role .
[ lem:1 ] there is a constant @xmath29 such that for all @xmath110 and @xmath171 @xmath172 @xmath173 we only show for @xmath171 , since the proof of is almost identical .
recall , @xmath174 for @xmath175 smooth functions with supports inside @xmath120 $ ] .
hence , @xmath176 and @xmath177 by cauchy
schwatz s inequality and plancherel s theorem @xmath178 moreover , since @xmath179 we obtain @xmath180 which finishes the proof of . [ lem:2 ] let @xmath181 . for all @xmath182 , @xmath183 and @xmath184 @xmath185
we define a sequence @xmath186 by @xmath187 then @xmath188 where @xmath189 .
since @xmath190 , by minkowski s inequality we obtain @xmath191 where in the last step we have used lemma [ lem:1 ] .
we notice , the constant @xmath29 depends only on @xmath119 .
hence , for all @xmath192 @xmath193 since @xmath194 taking @xmath195 we obtain @xmath196 thus @xmath197 therefore , @xmath198 and using @xmath199 , we get @xmath200 . for the converse inequality
, we use again to conclude @xmath201 [ prop:3 ] for @xmath181 and @xmath202 @xmath203 where @xmath204 . since @xmath205
thus by hlder s inequality we have @xmath206 now we note that @xmath207 and @xmath208 and the implied constants are independent of @xmath118 .
thus we obtain @xmath209 where the last inequality is a consequence of @xcite .
the proof will be completed if we show @xmath210 for this purpose we use from lemma [ lem:2 ] . indeed , @xmath211 this finishes the proof of the proposition .
[ th:3 ] for each @xmath181 there are @xmath212 and @xmath213 such that @xmath214 for all @xmath215 . based on proposition [ prop:3 ]
we may assume @xmath216 .
let @xmath182 be fixed .
firstly , we are going to show that for every @xmath217 we have @xmath218 by mbius inversion formula we see that @xmath219 moreover , for @xmath220 we may write @xmath221 where for @xmath222 we have set @xmath223 therefore , by formula and we have @xmath224 thus in view of it will suffice to prove that @xmath225 where the constant does not depend on @xmath226 . for the proof let us fix @xmath215 and consider a sequence @xmath227 defined by @xmath228 by proposition [ prop:3 ] , we have @xmath229 also for any @xmath230 @xmath231 since @xmath190 , by minkowski s inequality and lemma [ lem:1 ] we obtain that the last expression can be dominated by @xmath232 therefore , by @xmath233 summing up over all @xmath234 we obtain @xmath235 eventually , by lemma [ lem:2 ] we conclude @xmath236 next , we resume the analysis of . using
we get @xmath237 we observe that by the change of variables @xmath238 thus by minkowski s inequality @xmath239 since for @xmath240 @xmath241 we conclude @xmath242 now lemma [ lem:1 ] and lemma [ lem:2 ] imply @xmath243 this completes the proof of .
eventually , by and we obtain that @xmath244 for any @xmath52 and @xmath110 .
if @xmath27 we may refine the estimate ( see also @xcite ) .
let @xmath245 and note that @xmath246 since @xmath190 , and the supports of @xmath247 s are disjoint when @xmath248 varies .
by we have @xmath249 whereas by , we have @xmath250 these two bounds yield @xmath251 where the last estimate follows from cauchy schwartz inequality and the definition of @xmath252 .
eventually , by plancherel s theorem we may write @xmath253 which is majorized by @xmath254 . thus for appropriately chosen @xmath217 we obtain @xmath255 next , for @xmath28 we can use marcinkiewicz interpolation theorem and interpolate between and to conclude the proof . we have gathered necessary tools to illustrate the proof of theorem [ thm:2 ] .
first , we show the boundedness on @xmath23 of the maximal function @xmath256 .
[ th:4 ] the maximal function @xmath256 is bounded on @xmath23 for each @xmath24 .
let us observe that for a non - negative function @xmath257 @xmath258 where @xmath259 is a maximal function corresponding with bourgain
wierdl s averages @xmath260 indeed , suppose @xmath261 for @xmath262 .
then @xmath263 where @xmath264 .
therefore , by , we see @xmath265 since the maximal function @xmath266 is bounded on @xmath23 for any @xmath181 ( see @xcite or appendix [ apx:1 ] ) thus we have reduced the boundedness of @xmath256 to proving @xmath267 let us consider @xmath215 for @xmath268 . by theorem [ th:3 ]
we know that for @xmath269 @xmath270 if @xmath257 is non - negative then @xmath271 thus @xmath272 hence , @xmath273 for @xmath27 we use proposition [ prop:2 ] to get @xmath274 for any @xmath101 big enough .
if @xmath28 we apply marcinkiewicz interpolation theorem to interpolate between and and obtain @xmath275 since @xmath276 by and theorem [ th:3 ] we finish the proof .
next , we demonstrate the pointwise convergence of @xmath277 . if @xmath215 , @xmath24 then for every @xmath278 @xmath279 and @xmath17 is bounded on @xmath23 . if @xmath280 we define an operator @xmath281 by setting @xmath282 for any @xmath215 . by hlder
s inequality we see that for every @xmath278 @xmath283 where @xmath284 stands for the conjugate exponent to @xmath285 , i.e. @xmath286 . the last inequality shows that , on the one hand , @xmath17 is well defined for any @xmath287 , on the other proves .
next , fatou s lemma with boundedness of @xmath256 yield @xmath288 which completes the proof .
let @xmath290 be a strictly increasing sequence of @xmath291 elements .
we set @xmath292 and @xmath293 $ ] . in this section
we consider the kernel @xmath294 .
since each @xmath295 for @xmath77 has mean zero we have @xmath296 let @xmath289 denote the truncated hilbert transform @xmath297 the following argument is based on ( * ? ? ?
* section 7 ) .
[ prop:5 ] there is @xmath29 such that for every @xmath298 and @xmath202 we have @xmath299 let @xmath300
. by plancherel s theorem we have @xmath301 by we have @xmath302 hence , @xmath303 therefore , we obtain @xmath304 similar for @xmath305 , replacing @xmath306 by @xmath307 under the supremum , we can estimate @xmath308 now , using we get @xmath309 thus @xmath310 therefore , we conclude @xmath311 eventually , by proposition [ prop:3 ] @xmath312 which is bounded by @xmath313 .
[ th:5 ] for every @xmath298 there is @xmath314 such that @xmath315 and @xmath316 . by proposition [ prop:2 ] , we have @xmath317 consequently
, it is enough to demonstrate @xmath318 where @xmath316 .
let @xmath319 be defined as @xmath320 .
by theorem [ th:3 ] we have @xmath321 we set @xmath322 by the change of variables , cauchy schwartz inequality and by proposition [ prop:5 ] we get @xmath323 by the definition of @xmath324 we see that @xmath325 thus we achieve @xmath326 which finishes the proof .
let @xmath0 be a dynamical system on a measure space @xmath1 .
let @xmath327 be an invertible measure preserving transformation .
for @xmath328 we set @xmath329 we are going to show theorem [ thm:1 ] .
we start from oscillatory norm .
[ prop:4 ] for each @xmath298 there is @xmath314 such that @xmath330 and @xmath316 .
let @xmath331 .
for a fixed @xmath332 we define a function on @xmath58 by @xmath333 then for @xmath334 @xmath335 hence , @xmath336 therefore , by theorem [ th:5 ] we can estimate @xmath337 since @xmath338 is a measure preserving transformation integration with respect to @xmath332 implies @xmath339 eventually , if we divide both sides by @xmath340 and take @xmath341 we conclude the proof . [ cor:1 ] the maximal function @xmath342 is bounded on @xmath343 for each @xmath24 .
next , we show the pointwise convergence of @xmath344 .
let @xmath345 , @xmath24 . for @xmath4-almost
every @xmath332 @xmath346 and @xmath347 is bounded on @xmath343 .
let @xmath348 , since the maximal function @xmath349 is bounded on @xmath350 we may assume @xmath257 is bounded by @xmath83 .
suppose @xmath351 does not converge @xmath4-almost everywhere .
then there is @xmath52 such that @xmath352 now one can find a strictly increasing sequence of integers @xmath353 such that for each @xmath77 @xmath354 where @xmath292 and @xmath355 .
if @xmath261 then setting @xmath356 $ ] we get @xmath357 by siegel
walfisz theorem we get @xmath358 thus there is @xmath29 such that @xmath359 hence , whenever @xmath360 we have @xmath361 in particular , we conclude @xmath362 for each @xmath363 which contradicts to proposition [ prop:4 ] . indeed , @xmath364 where @xmath365 .
now , the standard density argument implies pointwise convergence for each @xmath7 where @xmath268 , and the proof of the theorem is completed .
in the appendix we discuss why the maximal function @xmath366 is bounded on @xmath23 .
this fact was published by wierdl in @xcite , however , on page 331 in the last equality for * * the factor @xmath56 has the power @xmath83 in place of @xmath88 .
therefore , it is not sufficient to show an estimate ( 24 ) from @xcite to conclude the proof .
in fact , one has to prove the estimate corresponding to from the present paper . for the completeness we provide the sketch of the proof based on the method used in section [ sec:3 ] .
first , we may restrict supremum to dyadic @xmath367 .
we modify the definition of the multiplier @xmath75 by setting @xmath368 hence , it suffices to show that for @xmath181 @xmath369 keeping the definition of the major arcs and setting @xmath370 proposition [ prop:1 ] holds true . for proof we use the well - known result that for @xmath84 ( see e.g ( * ? ? ?
* lemma 8.3 ) ) @xmath371 and then , as in the proof of proposition [ prop:1 ] , we replace the sum by @xmath372 .
also the demonstration of proposition [ prop:2 ] has to be modified .
there , the estimate for @xmath373 is a direct application of vinogradov s theorem . in the proof of proposition [ prop:3 ] in the place of we use @xmath374-boundedness of hardy
littlewood maximal function .
eventually , in the proof of theorem [ th:4 ] we replace the sum @xmath375 with a single term @xmath376 . | the aim of this paper is to prove cotlar s ergodic theorem modeled on the set of primes . |
searches for faint signals in counting experiments are often encountered in particle physics ( e.g. in searches for new resonances ) and astrophysics ( e.g. in searches for new sources of high - energy photons ) .
many practical problems can be reduced to the case of a counting experiment where the events come from the sum of two independent poisson - distributed contributions , usually referred to as `` signal '' and `` background '' .
although an accurate estimate of the background is desired in most applications , the sole parameter of interest is typically the signal intensity . in other words ,
the background intensity can be viewed as a nuisance parameter of the statistical inference problem , whose goal it is to estimate the signal strength given the measured count rate .
sometimes the background is the result of different sources . in such cases , one has a model in which the signal is superimposed to the total background .
if we knew which event comes from which source , we could keep separate lists for the number of observed events for each source of signal or background .
however , in many practical problems , only fewer quantities are directly observable , and sometimes only the total number of observed events can be measured .
an example is the number of events recorded by a geiger - mller counter when measuring the intensity of a weak radioactive sample .
the counts give only information about the total rate of radioactive decays , and not about the isotopic composition of the sample or the environment .
often one has prior knowledge about the individual contributions which can be used to improve the result of the inference . in the example of the weak radioactive source
, this could be any knowledge about the rate of natural radioactivity of the surrounding and the rate of cosmic muons , obtained from a measurement without the source . in this paper
we address the inference problem outlined above by means of objective bayesian methods .
these are based on the likelihood function , on informative background priors , and on an objective signal prior .
this work is based on refs .
@xcite and @xcite , and extends their results . choosing a bayesian approach
forces the scientist to be explicit about all the assumptions made .
some of them may be `` subjective '' , in the sense that they reflect the professional experience of the scientist .
as this is validated by numerous cross checks and by the criticism of the colleagues , a better term might be `` inter - subjective '' but we will keep the standard terminology . on the other hand , `` objective '' means that the posterior depends only on the assumption of the model , although there is always some degree of ( scientific ) subjectivity in the choice of the model in the first place .
this is exactly the meaning attributed to `` objective '' in frequentist approaches .
hence we are concerned here with bayesian results that are by construction as objective ( or subjective ) as frequentist results .
bayesian solutions have the advantage of a very simple interpretation : one obtains a range of admissible values for the parameter of interest ( the signal intensity ) together with the probability that the true value belongs to that range . on the other hand ,
the true value of the parameter is not a random quantity in the frequentist approach and thus one can not speak about the probability that it is inside some interval .
hence , a frequentist solution is a statement about the probability of the data given the true value , in a picture in which the experiment is imagined to be identically repeated a large number of times . in practical situations ,
this is a very idealized picture , as identically repeating the experiment is either unfeasible or too expensive .
furthermore , when feasible it might also appear undesirable .
for example , in high - energy physics data are collected during several `` runs '' , ultimately related to the availability of particle interactions .
colliders inject particles before the run starts and keep them circulating until the collision rate drops enough to justify a new fill .
detectors enable data taking after each fill , and keep running until collisions are detected or some problem forces restarting the run . in practice , although to some extent all runs could be considered as identical repetitions of the same experiment , during data analysis they are joined to form a bigger sample , rather then treating them as repeated experiments . from a computational perspective
, bayesian methods require the calculation of multidimensional integrals while frequentist methods typically adopt maximization procedures .
the latter are typically less cpu - intensive , if those calculations are performed using numerical methods .
however , we are concerned here with a problem that can be treated analytically , hence the computing time is negligible .
both approaches are perfectly legitimate .
when addressing statistical inference , it is important to clearly formulate the question one wants to answer , because this implies the choice of the paradigm . for example , if one wants to speak in terms of the probability of a model ( in our case , a model including contributions from a signal source ) given the experimental result ( e.g. the total count rate ) , then the bayesian approach is the correct paradigm . on the other hand ,
if the question is cast in terms of how ( un)likely the present outcome would be in an ideal infinite sequence of identical repetitions of the same experiment , then the frequentist approach is the choice .
we perform the statistical inference from the bayesian point of view because we believe that this answers the most typical question asked by scientists once an experimental result is made public : what is the probability that this or that model is ( in)correct ? regardless of using a bayesian or frequentist approach , the statistical model used in the inference problem has to be defined carefully .
although the approach chosen in the following guarantees that in the asymptotic limit of a very large number of observed events the posterior will concentrate around the maximum of the likelihood , the most interesting class of problems is on the opposite side , i.e. when collecting a very low ( possibly null ) number of events .
it is hence important that the general solution to this problem does not rely on asymptotic approximations and assumptions , e.g. by using a gaussian model in the regime of small count rates .
considering the contributions of signal and different background sources to the count rate allows not only to infer on the signal - plus - background rate , but also to investigate the individual contributions .
it is possible to model such a situation by either a product of independent poisson distributions one for each signal or background source or by a product of a single poisson distribution and a multinomial distribution . in the latter form
we can encode correlations between the different contributions with the help of an informative dirichlet prior .
the inference problem may be then split into two parts : inference about the total intensity , and inference about the relative contributions of the different sources .
the main result is that an analytic objective bayesian solution can be obtained , which is valid for any number of measured events ( even when it is very low or zero ) .
an example for such a situation is the search for contributions of a new physics process at a hadron collider in a particular part of a spectrum , say , an invariant mass peak on top of a set of standard model background processes .
the expected number of events due to background sources which are estimated based on theoretical cross - sections are partially correlated due to common uncertainties , for example the integrated luminosity of the data set or the scale of the invariant mass .
these correlations can be modeled by suitably choosing a prior for the multinomial part .
this paper is organized as follows .
section [ sec - likelihood ] gives details about the statistical models , while section [ sec - bayes ] describes the bayesian inference .
section [ sec - summary ] concludes the paper .
a statistical model is the specification of the probability to observe any outcome given the values of the parameters .
when considered as a function of the unknown parameters given the observed data , the model is called the likelihood function .
we first address the case of a counting experiment with @xmath0 contributions and for which the counts @xmath1 are independent poisson variables and are observed independently .
we recall the well known result that the likelihood function can either be written as a product of poisson distributions , or as the product of a poisson distribution for the total number of counts , @xmath2 , and a multinomial distribution describing how the counts are partitioned across the different contributions .
this is most often used when considering the entries in different histogram bins .
the number of events in each bin , taken alone , follows a poisson distribution .
the shape of the histogram , normalized to unit area , is described by a multinomial distribution . for this reason ,
it is sometimes useful to imagine that our counts @xmath3 are represented by a histogram with @xmath0 bins .
later , we consider the situation when one additional contribution has a special meaning , that is it represents the signal .
if @xmath4 with @xmath5 are independent poisson random variables , the statistical model of the data is a product of poisson distributions @xmath6 with parameters @xmath7 , where @xmath8 the outcome of the experiment is the @xmath0-tuple @xmath9 of observed counts for each contribution , using a notation in which lowercase letters stand for actual values of the random variables denoted with uppercase letters . the total number of observed events is @xmath10 .
one the other hand , the conditional distribution of the random vector @xmath11 , given the total number of counts @xmath12 , is a multinomial distribution @xmath13 with the conditions @xmath14 and @xmath15 the parameters @xmath16 represent the relative proportions of the poisson rates : @xmath17 the condition implies that there are only @xmath18 independent shape variables in eqn . , and could be used to write the multinomial distribution in a less symmetric form , for example by replacing @xmath19 with @xmath20 and @xmath21 with @xmath22 in eqn . .
the binomial distribution is a special case of the multinomial distribution with @xmath23 , @xmath24 with @xmath25 a well - known result is that the unconditional distribution of @xmath26 can be factored into the product @xmath27 of a single poisson term and a multinomial distribution .
the poisson distribution gives the probability of observing a total number of events @xmath28 , given the sum of all expected contributions , @xmath29 .
the multinomial term describes how the individual counts @xmath30 distribute among the @xmath0 bins .
let us now assume that we have measured a total of @xmath28 counts , and that we want to infer about the model parameter using eqn . as the likelihood function .
an immediate consequence of eqn .
is that @xmath28 carries no information about @xmath31 ( and vice versa ) , but only about @xmath32 .
it is no surprise that the statistical inference about the total rate only requires the total number of counts . on the other hand
, @xmath31 encodes the shape of the distribution of counts across the @xmath0 contributions , but not the normalization : likelihood - based inferences about @xmath31 are the same whether one regards @xmath33 as sampled from @xmath0 independent poisson sources or from a single multinomial process . in other words ,
estimates and tests of any function of @xmath31 will be the same whether we regard the number of counts as a random variable or as a fixed value .
this means that we can separately address the statistical inference problems related to the total count rate and that related to the contributions from different background sources .
the first step is to solve a problem with a single poisson model and @xmath28 total observed events .
the second step is to solve a multinomial problem which focuses on the distribution of counts , given the total number @xmath28 of observed counts .
if a poisson distributed signal source exists on top of @xmath0 background sources , and is independent of the background , one of the @xmath34 contributions to the observed number of counts assumes a special meaning ( the signal ) and the model can be written as @xmath35 where @xmath36 making use of the interpretation of a histogram described earlier , the different counts @xmath37 can be viewed as a histogram with counts from one signal region and @xmath0 auxiliary measurements .
we are now interested in the case that the data show an evidence that , in addition to the known `` background '' source with non - negative bin yields @xmath38 and overall strength @xmath39 representing nuisance ( i.e. not interesting ) parameters , there is a non - null contribution from a `` signal '' source whose non - negative bin yields @xmath40 and overall strength @xmath41 are our parameters of interest .
this is not just the `` discovery '' problem , for which one would typically focus on the total number of observed counts , because here we are also interested in the shape of the signal contribution .
this is likely to be the second step after a discovery phase of a new particle , for example when one wants to conduct a measurement of the mass or width of the new particle , or to distinguish between two competing models which predict different shapes for the distribution of some kinematical quantity . even though we have @xmath42 independent sources , we do not know if any given event has come from a signal or background source .
hence we only have @xmath0 observables , the counts in @xmath0 bins .
formally , we start by considering the signal random variables @xmath43 , which are poisson distributed @xmath44 , and the background variables @xmath45 , which are also poisson distributed @xmath46 .
next , we define the random variables @xmath47 which correspond to the observable counts in the @xmath0 bins . as
the sum of poisson variables is again a poisson variable , we have @xmath48 , i.e. we recover the notation used in section [ sec - k - bins ] above with @xmath49 , @xmath50 and @xmath51 . in terms of the variables @xmath52 and
@xmath37 , if @xmath28 total events have been observed the likelihood function for @xmath42 bins is @xmath53 now we partition the counts into @xmath54 signal events and @xmath55 background events ( see appendix ) .
conditional on @xmath56 ( and on @xmath28 ) , we can now write the previous multinomial as the product @xmath57 because we actually do not know @xmath56 , we need to sum over all possible ways of obtaining @xmath56 signal events out of the total @xmath28 counts , with the help of a binomial distribution @xmath58 whose parameter @xmath59 is the probability of obtaining a signal event when both signal and background are active sources .
the likelihood for @xmath42 bins is then @xmath60 as a cross check , we can count the degrees of freedom : 1 poisson variable , plus 1 binomial variable , plus @xmath18 multinomial variables for the signal , plus @xmath18 multinomial variables for the background . in total , we have @xmath42 variables : @xmath61 together with @xmath18 signal yields ( e.g. @xmath62 ) are the parameters of interest , whereas the @xmath0 background yields @xmath63 are nuisance parameters .
we would have obtained the same result , equation ( [ eq-2k - likelihood ] ) , by considering the likelihood ( [ eq - model - k - bins ] ) as the merged version of separate signal and background sources ( see appendix ) , and defining @xmath64 , such that @xmath65 ( one must not forget the binomial weights due to the splitting of the @xmath28 events into two unknown partitions ) . finally , because we can only observe the sum of signal and background counts in each bin , we have to merge the latter into a set of @xmath0 bins . given the observed @xmath30 counts in bin @xmath66 , there are several ways in which signal and background events can give this sum .
again , we account for this with the help of binomial weights , @xmath67 with @xmath68 .
the full likelihood becomes @xmath69 \right\ } \end{split}\ ] ] that is the product of ( [ eq - model - k - bins ] ) which is the first line in ( [ eq - full - likelihood ] ) with terms accounting for all possible combinations of @xmath0-tuples of signal counts and @xmath0-tuples of background counts which give as a result the observed counts @xmath70 in each bin . in general ,
given the experimental result @xmath70 , the likelihood function ( [ eq - full - likelihood ] ) is the starting point for performing statistical inference about the parameters of interest , which are the signal yields @xmath71 .
however , when @xmath0 is not small this model is computationally very intensive , as the number of possible combinations becomes huge .
in the bayesian approach , the statistical inference relies on the use of bayes theorem , which can be interpreted as the algorithm for updating our knowledge about a model and its parameters ( modeled as a probability distribution ) in view of the outcome of an experiment .
the prior knowledge is encoded into the prior distribution for the parameters of the model , and their joint posterior distribution ( obtained with bayes theorem ) reflects our new state of knowledge about them . in many cases , we are only interested in a subset of the parameters .
one can obtain their joint marginal posterior probability density by integrating the full posterior over all other ( nuisance ) parameters .
we first address the problem of @xmath0 independent poisson sources and @xmath0 observable count rates .
the likelihood function is given by eqn .
with lowercase letters that remind us that the likelihood is a function of the parameters with fixed ( and known ) experimental outcome .
bayes theorem gives the joint posterior probability density for @xmath29 and @xmath31 @xmath72 \ ; [ { \ensuremath{\text{mul}}}(\vec{x } \
, | n , \vec{\eta } ) \ ; p(\vec{\eta } ) ] \end{split}\ ] ] ( where the normalization constant can be found by integrating over the r.h.s . ) in the form of a product of two posterior densities .
the first corresponds to a poisson model with expectation value @xmath29 and prior density @xmath73 , from which @xmath28 events are generated .
the second corresponds to a multinomial model with parameters @xmath31 and prior density @xmath74 , generating the observed vector @xmath75 of counts .
they are two inference problems which can be solved independently , as is shown below . [ [ conjugate - prior ] ] conjugate prior the most convenient functional form for the prior density of the poisson parameter is a gamma function @xcite , i.e. , the conjugate prior for a poisson model , @xmath76 , @xmath77 with shape parameter @xmath78 and rate parameter @xmath79 ( or scale parameter @xmath80 ) .
when the prior is @xmath81 , the posterior is @xmath82 [ [ informative - prior ] ] informative prior in the simple but frequent case in which the prior knowledge about @xmath29 is summarized by its expectation @xmath83 $ ] and variance @xmath84 $ ] , the gamma parameters are determined with the method of moments by imposing @xmath83 = \alpha/\beta$ ] and @xmath84 = \alpha/\beta^2 $ ] .
this gives @xmath85/v[\lambda]$ ] and @xmath86 $ ] .
alternatively , one could start from the prior most probable value ( the gamma mode is at @xmath87 for @xmath88 ) and variance , or from the knowledge of intervals covering given prior probabilities ( e.g. 68.3% or 95% probability ; this requires a numerical treatment to find @xmath89 and @xmath90 ) , or from any set of conditions which is sufficient to determine the shape and rate parameters . in principle , if there is quite detailed information about the prior for @xmath29 _ and _ it is known that the gamma density does not correctly represent its shape , one has two options .
the first one is to adopt a different functional form and solve bayes theorem numerically , and the second is to find a linear combination of gamma densities which approximates the prior well enough . in the latter case , the posterior will be a linear combination of gamma functions .
this simplifies the treatment without biasing the result , provided that enough terms are present in the linear combination of priors , and it is hence preferable to numerical integration . when @xmath28 is large enough , the solution becomes lesser and lesser dependent on the exact shape of the prior , is arbitrarily large the shape of the prior does not matter at all , provided that it is strictly non - null in the region where the true value is . ]
hence one should consider using numerical methods or a linear combinations only if @xmath28 is very small and most importantly when one is very confident that a single gamma density is not good enough to model the prior .
in practical problems , this happens very rarely ( although it is not impossible ) , and a vague knowledge of the prior shape is typically sufficient . in cases where only vague prior information is available ,
it is recommended to assess the dependence of the posterior on the choice of the prior by comparing the results obtained with different priors , all compatible with the available information ( e.g. different functional forms giving the same first two moments ) .
when the posterior does not change appreciably , it means that @xmath28 is large enough for the likelihood to dominate and that the result is robust .
otherwise , one has to conclude that the experiment did not yield enough events for obtaining a solution that does not depend on our prior state of knowledge . in this case , it is best to report the differences of a few key figures of merit ( like the posterior mean and variance , or posterior credible intervals with given probability ) which correspond to the choice of quite different priors .
how different should the priors then be ?
of course , comparing our informative gamma density with a delta - function does not make any sense : there are requirements that admissible priors must satisfy to be considered acceptable .
luckily enough , such requirements are really minimal and typically reasonable , like requiring that the prior is strictly positive over the entire domain and that the posterior be a proper density ( i.e. it integrates to one ) .
in addition , a formal procedure exists which gives us the `` most distant '' prior to our informative prior . we can call it the `` least informative '' prior or the `` objective '' prior , and it is provided by the reference analysis @xcite based on information theory .
this reference prior is the function which maximizes the amount of missing information about the system , hence it is the `` most distant '' choice from our informative prior . in other words ,
the reference prior encodes the minimal set of information about the problem .
indeed , it only depends on the description of the model itself ( in the case under consideration here , on the poisson distribution ) and makes no use of any prior knowledge about the parameter other than the specification of its domain ( which for a poisson problem is the entire positive real line ) .
for this reason , the reference prior shall be used when one wishes to adopt an `` objective prior '' .
[ [ objective - prior ] ] objective prior when assessing the dependence of the result from the prior information , it is recommended to compare against the reference posterior @xcite , i.e. the solution obtained with the reference prior @xmath91 , which for the poisson model coincides with jeffreys prior : @xmath92 .
the reference posterior is then @xmath93 and also represents the best choice when one is required to report an `` objective '' result , or when one claims to have the minimal prior information about the poisson parameter .
compared to the informative prior , which is chosen to best encode all available prior information , the reference prior is the `` most distant '' choice from the information - theory point of view , as it encodes the minimal prior information . [
[ a - simple - example ] ] a simple example in an attempt to measure the amount of background activity in a new lab , one performs two subsequent measurements for the same duration . for the first measurement , no prior knowledge about the expected rate @xmath29 is assumed , hence the non - informative prior is chosen .
the measurement yields an observed number of events of @xmath94 , so that the posterior distribution is @xmath95 with expectation value @xmath83=9.5 $ ] and variance @xmath84=9.5 $ ] .
the posterior of the first measurement is then used as an informative prior for the second measurement .
the number of observed events is @xmath96 and so the posterior distribution is @xmath97 with expectation value @xmath83=10.75 $ ] and variance @xmath84=5.38 $ ] .
one obtains the same result by merging the two data sets and applying the objective prior to this new `` experiment '' . in this case , the posterior is @xmath98 , where the new random variable @xmath99 because merging the two observations is equivalent to observing for a duration twice as long .
it is very simple to show that , after changing variable , one obtains @xmath100 , which means that the posterior for the original parameter is @xmath101 , the same as above .
[ [ conjugate - prior-1 ] ] conjugate prior the conjugate prior of the multinomial model , @xmath102 , is the dirichlet distribution , which means that the posterior is also a dirichlet distribution and its parameters are simple functions of the prior and multinomial parameters .
the dirichlet distribution with concentration parameters @xmath103 ( with @xmath104 and @xmath105 for @xmath5 ) is @xmath106 where the multidimensional beta function has been used to write the normalization constant @xmath107 if the prior for the multinomial problem is @xmath108 , the posterior is a dirichlet distribution with parameters @xmath109 : @xmath110 when @xmath23 , the multinomial becomes a binomial distribution .
the corresponding conjugate prior is the beta density , @xmath111 which is indeed the special case of a dirichlet distribution with @xmath23 , @xmath112 , @xmath113 .
one obtains a beta density also as the marginal distribution of any dirichlet parameter , i.e. the beta density is the result of integrating a dirichlet over @xmath114 parameters .
[ [ informative - prior-1 ] ] informative prior the following properties of the dirichlet distribution can be used to find the values of the concentration parameters which best represent all the available prior information : @xmath115 & = & a_i / a \\
\text{mode}[\eta_i ] & = & \frac{a_i - 1}{a - k } \\
v[\eta_i ] & = & \frac{a_i ( a - a_i)}{a^2 ( a+1 ) } \\ \text{cov}[\eta_i,\eta_j ] & = & \frac{-a_i a_j}{a^2 ( a+1 ) } \end{aligned}\ ] ] often the parameters may be grouped into two classes : a set of parameters that are independent from any other parameter , and a set of parameters that have pairwise correlations . by grouping them correspondingly , one obtains a block - diagonal matrix in which the first block is actually diagonal .
parameters in the first block are then fixed by imposing their prior variance and expectation .
for the others , one has a system of requirements , for example in the form of a collection of variances and covariances .
as there may be more constraints than free variables , the prior concentration parameters may be found e.g. by solving a minimization problem .
alternatively , one may choose to solve the problem multiple times , each with a different choice of concentration parameters which satisfies @xmath0 independent constraints . on the other hand ,
if there are less constraints than the number of parameters , one may impose the `` objective '' choice on the unconstrained variables , i.e. equal prior concentration parameters .
this might be done in a two - step algorithm starting with an objective prior with concentration parameters all set to @xmath116 ( see below ) , and by imagining an intermediate experiment in which only a subset of @xmath117 variables get updated to @xmath118 .
the next step is to find real values @xmath119 such that the first @xmath120 concentration parameters satisfy all available constraints .
the result can be used as the prior for the actual experiment .
[ [ non - informative - prior ] ] non - informative prior the problem of finding an objective prior for all multinomial parameters was addressed by @xcite and @xcite .
being aware that for multidimensional models the reference prior depends on the ordering of the parameters , @xcite argued that , when no priority ranking is desired or allowed ( in other words , when one wants all @xmath121 parameters to be on the same footing ) , some sort of `` overall objective prior '' shall be used which treats all parameters fairly well .
a very reasonable requirement is that the marginal posterior for any given parameter should be as close as possible to the solution obtained with the marginal model and the reference prior for that single parameter .
for the multinomial model , the reference posterior when only @xmath121 is of interest is @xmath122 .
is equivalent to a binomial distribution , this result can also be proved with the latter model .
the reference posterior is the same beta density , as it is shown for example in ref .
] on the other hand , if the prior @xmath123 is used using the same concentration parameter @xmath124 for all multinomial @xmath121 parameters , as they are treated on the same footing , the marginal posterior for @xmath121 is instead @xmath125 .
the goal is to find @xmath124 such that , in an average sense , this result is closest to @xmath126 for each @xmath66 . among different possible approaches to this problem @xcite , @xcite addressed it with a hierarchical approach and showed that the overall objective prior is a dirichlet density with all identical concentration parameters , whose value is very well approximated by the simple choice @xmath127 ( just a bit lower than the intuitive choice @xmath128 ) , which gives an objective dirichlet posterior with parameters @xmath129 .
[ [ a - simple - example - continued ] ] a simple example , continued in the case of two subsequent measurements , one might be interested in the background stability with time .
this is relevant during the preparation of low - background experiments , e.g. if newly produced parts of an experiment are brought into underground laboratories . in our example , we could imagine to have installed new equipment between the two measurements and ask ourselves if the somewhat larger number of counts in the second measurement suggests that the equipment has increased the overall background rate .
one way of checking this is to look at the 2-bins histogram with @xmath130 and @xmath131 counts and quantify the departure from equipartition .
a more formal procedure would be to test against the null hypothesis that the background rate is constant , but it is not necessary in this particular example , which is quite straightforward . to better illustrate the problem , we use first a binomial model with @xmath132 successes out of @xmath133 total trials and
look at the deviation from @xmath134 , where @xmath135 is the ratio between the initial and the total rate .
the reference posterior for @xmath135 is @xcite @xmath136 the posterior mean , mode and standard deviation are @xmath137 = 0.4318 \qquad m[\varepsilon ] = 0.4250 \qquad \sigma[\varepsilon ] = 0.1033\ ] ] if we start with @xmath138 and use the overall objective prior @xmath139 recommended by @xcite , we obtain instead the marginal posterior @xmath140 whose mean , mode and standard deviation are @xmath141 = 0.4312 \qquad m[\varepsilon ] = 0.4242 \qquad \sigma[\varepsilon ] = 0.1037\ ] ] thus the two alternative solutions are hardly distinguishable in practice .
the peaks are @xmath142 and @xmath143 lower than @xmath134 , not a significant difference to conclude that the background level has changed between the two measurements , even without performing a formal test .
hence it is safe to use both measurements to estimate the background level in the cavern .
[ [ an - example - from - high - energy - physics - lepton - universality ] ] an example from high - energy physics : lepton universality assume an experiment built to test lepton universality ( see , e.g. refs .
@xcite for the current experimental status ) in @xmath144-boson production which is equally sensitive to electrons , muons and taus , i.e. assuming equal reconstruction efficiencies , momentum resolutions and so on .
events containing @xmath144-bosons are selected by measuring final states with two opposite - charge leptons and by requiring their invariant mass to be consistent with the mass of a @xmath144-boson .
adding further stringent event - selection requirements , the resulting event sample can be made practically background - free .
the number of observed events in this example are @xmath145 , @xmath146 and @xmath147 .
lepton universality is tested by estimating the branching ratios , e.g. , @xmath148 , where @xmath149 , and comparing them . if they are found to not be equal , then lepton universality is disfavored .
using eqn . with a reference prior for the overall rate @xmath29 and an objective prior for the multinomial part yields @xmath150 \ ; [ { \ensuremath{\text{mul}}}(\vec{x } \
, | n , \eta_{e } , \eta_{\mu},\eta_{\tau } ) \ ; { \ensuremath{\text{dir}}}(\eta_{e } , \eta_{\mu},\eta_{\tau } \,|\ , 0.8/3 , 0.8/3 , 0.8/3 ) ] \ , .
\end{aligned}\ ] ] the resulting two - dimensional marginal posterior , depending on @xmath151 and @xmath152 , is shown in fig.[fig : marginal ] , together with the corresponding one - dimensional projections . the expectation from lepton universality , i.e. , @xmath153 ,
is indicated by the asterisk and consistent with the observation . [ cols="^,^,^ " , ]
in this paper , we have derived a statistical model for counting experiments in which the sources contributing to the resulting count rate are potentially correlated . using bayes theorem and assuming non - informative and informative priors on signal and background contributions respectively , we have shown that the marginal posterior probability for the signal contribution can be expressed in closed form .
simple use cases in the field of experimental physics have been presented . 99 j.o .
berger & j.m .
bernardo , _ ordered group reference priors with application to the multinomial problem _ , biometrika 79 ( 1992 ) 25 - 37 .
berger , j.m .
bernardo , d. sun , _ objective priors for discrete parameter spaces _ , j. amer .
assoc . 107
( 2012 ) 636 .
berger , j.m .
bernardo , d. sun , _ overall objective priors _ , conference on `` recent advances in statistical inference : theory and case studies '' , university of padova , 2123 mar 2013 .
bernardo , `` reference analysis '' , handbook of statistics 25 ( d.k .
dey and c.r .
rao eds . ) .
amsterdam : elsevier ( 2005 ) 1790 .
a. caldwell , d. kollar , k. krninger , _ bat - the bayesian analysis toolkit _ ,
computer physics communications 180 ( 2009 ) 2197 - 2209 , http://arxiv.org/abs/0808.2552[arxiv : 0808.2552 ] .
d. casadei , _ reference analysis of the signal + background model in counting experiments _ , jinst 7 ( 2012 ) p01012 , , http://arxiv.org/abs/1108.4270[arxiv : 1108.4270 ] .
d. casadei , _ estimating the selection efficiency _ , jinst 7 ( 2012 ) p08021 , , http://arxiv.org/abs/0908.0130[arxiv : 0908.0130 ] .
d. casadei , _ reference analysis of the signal + background model in counting experiments ii .
approximate reference prior _ , jinst9 ( 2014 ) t10006 , , http://arxiv.org/abs/1407.5893[arxiv : 1407.5893 ] . k. a. olive _ et al .
_ [ particle data group collaboration ] , _ review of particle physics _ , chin . phys .
c * 38 * ( 2014 ) 090001 , .
l. ventura , s. cabras , w. racugno , _ prior distributions from pseudo - likelihoods in the presence of nuisance paramenters _ , j. amer .
statist . assoc . 104 ( 2009 ) 768 .
we recall here a few properties of the multinomial distribution that are important when considering practical problems , in particular its behavior under re - binning and partitioning .
the first one is about re - binning : merging random variables distributed according to multinomial distributions again gives a multinomial distribution . if @xmath154 with @xmath16 , then @xmath155 where @xmath156 and @xmath157 for @xmath158 .
one may obtain a binomial problem by sequentially merging @xmath114 bins .
the second property is important if we know the total number of counts @xmath159 in a subset of the bins : the conditional distribution given that @xmath160 and @xmath161 is a product of two independent multinomial pieces : @xmath162 for example , this property tells us how to treat two histograms at the same time .
conversely , the third property is that the joint distribution of two independent multinomial random vectors has a multinomial kernal . to see this ,
one starts from @xmath163 and writes their joint distribution as @xmath164 defining @xmath165 such that @xmath166 for @xmath5 and @xmath167 for @xmath168 , one has @xmath169 furthermore , defining @xmath170 one also has @xmath171 such that eqn .
may be rewritten as @xmath172 where we used @xmath173 .
this looks like a multinomial with @xmath174 degrees of freedom .
however , we started from two multinomials with @xmath18 and @xmath175 degrees of freedom , hence must obtain a distribution with @xmath176 free parameters .
eqn . gives the probability distribution conditional to @xmath28 and @xmath159 . if , however , only the total number of counts @xmath177 is known , then one must sum eqn . over all possible pairs of @xmath28 and @xmath159 which result in a total count of @xmath178 .
this reduce by one the number of independent variables . | searches for faint signals in counting experiments are often encountered in particle physics and astrophysics , as well as in other fields .
many problems can be reduced to the case of a model with independent and poisson - distributed signal and background
. often several background contributions are present at the same time , possibly correlated .
we provide the analytic solution of the statistical inference problem of estimating the signal in the presence of multiple backgrounds , in the framework of objective bayes statistics .
the model can be written in the form of a product of a single poisson distribution with a multinomial distribution .
the first is related to the total number of events , whereas the latter describes the fraction of events coming from each individual source .
correlations among different backgrounds can be included in the inference problem by a suitable choice of the priors . counting experiments ; bayesian inference ; objective priors |
infrared observations of emission line galaxies give access to ions not visible in the optical domain . among those
is the high excitation line which has not only been observed in active galactic nuclei but also in several starburst galaxies , although at a much fainter level ( genzel 1998 , lutz 1998 , hereafter lkst98 ) .
hot , massive stars generally emit only few ionizing photons with energies above the edge at 54.42 ev required for the production of .
therefore the origin of this high excitation line in starbursts has been unclear so far .
different excitation mechanisms ( weak agns , super - hot stars , planetary nebulae , and ionizing shocks ) have been discussed by lkst98 . based on simple estimates , photoionization and shock models ,
these authors favor ionizing shocks related to the starburst activity as the most likely explanation for emission _ in general_. according to lkst98 , massive super - hot stars remain , however , an option for the high excitation dwarf galaxies included in their sample .
one of these objects ( ngc 5253 ) was studied in more detail by crowther ( 1999 , hereafter c99 ) , who showed , by computing photoionization models around single stars , that wne - w stars can indeed produce strong and in surrounding regions .
the fact that these lines were not prominent in ngc 5253 led them to exclude the possibility of a significant number of such stars being present in this galaxy .
as will be shown below we do not support their conclusion for a variety of reasons . to shed more light on the origin of the emission in dwarf galaxies
, we use the information on nebular properties and stellar content derived from optical studies to complement the information from ir data .
in addition to ngc 5253 , we also consider ii zw 40 .
these are the two compact low metallicity galaxies which show the highest excitation among the starbursts observed by lkst98 .
both objects are known as so - called wr galaxies ( cf .
conti 1991 , schaerer 1999b ) , where the presence of broad stellar emission testifying to the presence of wr stars provides powerful constraints on the burst age and massive star content ( e.g. schaerer et al .
1999a ) . in this paper
, we present a stellar population model which reproduces the observed stellar features and which , used as an input for photoionization models , explains at the same time the ionization structure of the nebular gas as revealed by the optical and ir fine structure lines .
in the sample of emitting starbursts of lkst98 , ii zw 40 and ngc 5253 show the strongest excitation ( measured by / ) , the largest strength ( quantified by ; cf .
lkst98 , fig .
2 ) and stand out by several properties : * nebular emission indicative of high excitation is present in the region dominating the optical emission ( walsh & roy 1989 , 1993 , hereafter wr89 ,
wr93 , guseva 1998 ) * a significant number of wolf - rayet stars has been detected in these regions ( kunth & schild 1981 , walsh & roy 1987 , vacca & conti 1992 , schaerer 1997 , hereafter sckm97 ) * in both the dwarf ii zw 40 and the amorphous galaxy ngc 5253 one or few star forming regions of a young age clearly dominate the production of ionizing photons ( vanzi 1996 , beck 1996 , calzetti 1997 ) . finding 1 ) confirms the presence of high energy photons ( @xmath0 54 ev ) deduced from the ir observations of and naturally suggests a direct link between the nebular and emission .
furthermore the optical observations allow a more precise localisation of the high excitation regions . 1 ) and
2 ) indicate that ii zw 40 and ngc 5253 are objects where the observed emission is likely due to hot wr stars ( schaerer 1996 , 1997 , 1998 ; de mello 1998 ) .
3 ) justifies , at least to first order , the use of a `` spectral template '' of the brightest starburst region as a representation of the ionizing spectrum of the entire region covered by the iso observations .
wr stars of both wn and wc types have been observed in the two dominant regions of ngc 5253 by sckm97 . in ii zw 40 broad indicative of wr stars was detected by kunth & sargent ( 1981 ) , vacca & conti ( 1992 ) and guseva ( 1998 ) .
the latter also detect broad emission due to wc stars .
the wr and o star content was already analysed by schaerer ( 1996 ) , sckm97 and schaerer ( 1999a ) . in fig .
[ fig_ihb ] the observed intensities of the various wr features are shown and compared to an instantaneous burst model of schaerer & vacca ( 1998 ) with a salpeter imf at the appropriate metallicity ( z/@xmath1 1/5 ) .
the observations of guseva ( 1998 ) refer to the entire `` wr bump '' ( 4643 - 4723 ) and represent therefore an upper limit .
shifts in @xmath2 of the theoretical predictions for the wc lines with respect to the observations are not significant since they correspond to very short timescales .
this and other potential uncertainties affecting such a comparison have been extensively discussed in schaerer ( 1999a ) .
figure [ fig_ihb ] shows that all line strengths are reasonably well reproduced by the model . at the corresponding ages of @xmath1 3 - 5 myr ( cf .
sckm97 ) our synthesis model provides therefore a good description of the massive star content in these regions .
the spectral energy distributions predicted by the synthesis models described above have been used as input to the photoionization code _ photo _ ( same version as in stasiska & leitherer 1996 , hereafter sl96 ) .
the remaining input parameters are the total number of stars , the density distribution and the chemical composition here taken as z/=1/4 for easy comparison with sl96 ( the ionization structure of a nebula is insensitive to a small change in the abundances in the gas ) .
following sl96 we calculate sequences of models for a spherical gaz distribution with a uniform hydrogen density @xmath3 and filling factor @xmath4 , both assumed constant during the evolution of the starburst .
for a given age , models with the same ionization parameter @xmath5 have the same ionization structure .
0 is the total number of photons above 13.6 ev , and @xmath6 a function of the electron temperature ( see sl96 ) .
the densities derived in ii zw 40 and ngc 5253 ( @xmath1 70300 @xmath7 , wr89 , wr93 , c99 ) are low enough , so that collisional deexcitation is negligible for the lines of interest .
we therefore explore the parameter space by simply taking @xmath3=10 @xmath7 and @xmath4=1 , and consider three different initial masses for the starburst : @xmath8 , @xmath9 and @xmath10 .
these three model sequences will be referred to as the sequence with low , intermediate and high @xmath11 . in fig .
[ fig_ss_1 ] we show the temporal evolution of selected line ratios . unlike c99
, we chose to show line ratios that are independent of the abundances of the parent elements , in order to facilitate comparison with observations of different galaxies . the only exceptions are / and / when helium is fully ionized in the region .
also , we limit ourselves to line ratios that involve a dominant ionic stage in the nebula . line ratios like / or / as used by genzel et al .
( 1998 ) are difficult to interpret , as the lines are likely emitted by very different regions . as expected , line ratios from adjacent ionic stages show a progressive decrease of the overall excitation with time for the models we are considering .
helium remains fully ionized up to 4 myr .
notable exceptions to this trend are , and , species with the ionization potential at or above the edge , which appear during a short phase ( at ages @xmath12 3 - 4 myr ) where hot wr stars provide a non - negligible flux above 54 ev ( see schaerer & vacca 1998 ) .
slower temporal changes would of course be obtained for non instantaneous bursts .
it must be noted that those line ratios which are function of the ionization parameter also depend somewhat on the adopted geometry , and this in a non trivial way .
for example , in models with a thin shell geometry , the line ratios shown here differ by factors up to 3 from models for full spheres with the same mean ionization parameter . in the case of the / ratio
, the value predicted during the wr phase can be smaller by a factor of about 10 since , despite the presence of high energy photons , there is no matter emitting at a high ionization parameter , close to the star cluster . in the following
we compare the model predictions with observations of ngc 5253 and ii zw 40 .
observed line ratios are overplotted on the model predictions shown in fig .
[ fig_ss_1 ] .
the optical data is taken from wr89 and wr93 ( region 1 in both objects ) .
ir fluxes for ngc 5253 are taken from genzel ( 1998 ) , lkst98 , and c99 . for / a lower limit
is obtained since different iso apertures are involved in the measurements .
all other limits are `` real '' detection limits . adopting the draine ( 1989 ) extinction curve and @xmath13 mag ( c99 ) increases the / and / ratios by @xmath1 40 % .
other ir line ratios are much less affected .
c99 consider two separate emission regions to be responsible for the high excitation lines ( e.g. ) and lower excitation lines respectively ( e.g. ) .
we see no compelling reason for such a somewhat `` artificial '' separation .
a similar structure is e.g. naturally obtained in an ideal spherical nebula . instead of using line fluxes corrected for such effects we therefore use the original measurements .
ir line ratios for ii zw 40 are from lkst98 .
we have no access to the acquired iso sws spectra , which should , however , become available soon . from the top panels of figure [ fig_ss_1 ]
it is evident that no single model can reproduce at the same time all the observed line ratios .
this finding is not surprising and may be due to several reasons : 1 ) the structure of the galaxies is more complicated than assumed in the models .
2 ) although one or few bursts of similarly young age dominate the ionizing flux ( sect .
[ s_ass ] ) , the ionizing spectrum is likely not fully described by a single burst population .
3 ) atomic data may be inaccurate .
in particular , the computation of collision strenghts for fine structure transitions is a very delicate problem , and the evaluation of the formal uncertainty is difficult .
a comparison between the plasma diagnostics obtained using different ir and optical lines for the planetary nebula led oliva ( 1996 ) to suggest that the collision strengths which enter in the calculation of the intensity of are overestimated by a factor 3 !
similar problems are likely to occur for other fine structure lines . the / ratio , which is one of the best studied from all points of view , indicates that , if the age of the starburst lies between 3 and 4 myr , as indicated by the wolf - rayet features , the models with intermediate @xmath11 are the most adequate to represent the two galaxies under study .
at such an age , helium is still completely ionized , because the radiation field is hard enough .
the discrepancy with the measurement of / in ii zw 40 is likely due to absorption by galactic interstellar sodium intervening at this redshift ( izotov 1999 , private communication ) .
our main result is illustrated in the last three panels of fig .
[ fig_ss_1 ] where we show that during a short phase the stellar population provides enough photons above 54.4 ev to naturally produce the , and lines at levels comparable to the observed ones .
the emission is due to the presence of hot wr stars at ages @xmath12 3 - 4 myr ( cf .
schaerer 1996 , schaerer & vacca 1998 ) . the predicted strength of these lines exceeds even somewhat the observations .
however , this does not invalidate our conclusion .
a more realistic population `` mix '' can easily reconcile the intensity of with the observations .
as for the and lines , they are sensitive to the geometry ( see above ) , which provides ample space for fitting with tailored photoionization models .
this should , however , only be undertaken when the relevant atomic data have been validated by detailed multiwavelength studies of simpler objects and by comparisons with photoionization models .
we propose that 25.9 emission in ngc 5253 and ii zw 40 is due to the presence of hot wr stars observed in both objects .
we draw this conclusion from both empirical and theoretical facts .
first , we note that nebular and emission occur simultaneously in these objects .
furthermore a close link between nebular and wr stars has now been established for the so - called wr galaxies , extra - galactic regions and the few local group regions exhibiting this feature ( schaerer 1996 , 1997 , 1998 ; see schaerer 1999b for a catalogue of these objects ) .
second , quantitative models of the stellar populations using up - to - date non - lte atmospheres including stellar winds and the most recent evolutionary tracks are able to explain the observed massive star population and the optical emission lines ( including nebular ) during the wr rich phase ( schaerer 1996 , 1998 ) , even for the lowest metallicity object , ( de mello 1998 , stasiska & schaerer 1999 ) , where such stars were detected recently .
in addition , as demonstrated here , the emission ( and other ir lines ) is also naturally reproduced by these models .
the two objects considered here are the best available to constrain the origin of emission : they show the highest excitation , and represent the most simple objects in terms of their ionizing population .
ir observations of the 8 local group ( lg ) regions known to exhibit nebular ( cf .
garnett 1991 ) can provide a simple `` consistency '' test : the presence of this line is a necessary condition for showing emission .
the presence or absence of is , however , also influenced by the ionization parameter and the nebular geometry ( cf .
above ) .
observational evidence suggests that such high excitation regions occur preferentially at low metallicities .
this holds both for the lg and extragalactic objects ( cf .
schaerer 1997 , 1998 ) , including ii zw 40 and ngc 5253 .
the same can thus be expected for the contribution of wr stars to .
low metallicity may indeed also justify the neglect of line blanketing in the wr models of schmutz ( 1992 ) included in our synthesis models .
the effects discussed by c99 and crowther ( 1998 ) suppressing the output of photons above the edge in metal - rich wr models and/or high density winds could well be ineffective at low metallicities , as also suggested by the empirical evidence .
our explanation for the stellar photoionization origin of in dwarf - like low metallicity galaxies can not be necessarily generalised to all the objects of lkst98 .
although indeed 6 out of 14 from their list are known wr `` galaxies '' ( schaerer 1999b ) , it is unlikely that the regions where wr stars are detected contribute a significant fraction of the total ionizing flux in these complex objects .
outflows , weak seyfert activity , and other phenomena are known in some of them and provide alternative explanations as discussed by lkst98 .
more complex models will be required to interpret such objects , to provide an theoretical understanding of new empirical ir diagnostic diagrams ( cf .
genzel 1998 ) , and to assess the contribution of stellar sources to high energy photons .
yuri izotov kindly provided us with data prior to publication .
we thank marc sauvage and suzanne madden for useful discussions .
ds acknowledges a grant from the swiss national foundation of scientific research .
crowther , p.a.c .
, 1998 in `` wolf - rayet phenomena in massive stars and starburst galaxies '' , k.a .
van der hucht , g. koenigsberger , & p.r.j .
eenens ( eds . ) , iau symp .
193 , ( san francisco : asp ) , in press ( astro - ph/9812402 ) schaerer , d. , 1998 , in `` wolf - rayet phenomena in massive stars and starburst galaxies '' , k.a .
van der hucht , g. koenigsberger , & p.r.j .
eenens ( eds . ) , iau symp .
193 , ( san francisco : asp ) , in press ( astro - ph/9812357 ) | we propose that the emission of the high excitation line observed with iso in and is due to the presence of hot wolf - rayet ( wr ) stars in these objects .
we construct a consistent evolutionary synthesis and photoionization model which successfully reproduces the constraints on their massive star content and the relevant optical and ir emission lines including .
our explanation for the origin of is supported empirically by : _
1 ) _ the simultaneous presence of nebular and in these objects , and _ 2 ) _ the close relation between nebular and wr stars in extragalactic regions .
photoionization by hot wr stars is mainly expected to be of importance in young low metallicity galaxies .
alternate mechanisms are likely at the origin of emission in other objects . 0 |
in quantum information technology , it is usual to use quantum channel for sending quantum state .
since a quantum channel has noise , it is important to identify quantum channel . in this paper , we consider theoretical optimal performance of quantum channel estimation when we can apply the same unknown channel several times . in order to treat this problem ,
we employ quantum state estimation theory . in our setting
, we can optimize our input state and our measurement@xcite . as is illustrated in fig .
[ f1 ] with @xmath1 , it is assumed to be possible to use entanglement with reference system in the measurement process when the channel @xmath2 with the unknown parameter @xmath3 is applied @xmath4 times .
this setting is mathematically equivalent with the setting given in fig .
[ f2 ] with @xmath1 , which has a single input state in the large input system and a single measurement in the large output system . in this paper , we consider quantum channel estimation with the formulation given by fig .
[ f2 ] . in the state estimation ,
when the number @xmath4 of prepared states goes to infinity , the mean square error ( mse ) behaves as the order @xmath0 as in the estimation of probability distribution .
however , in the estimation of quantum channel , two different analyses were reported concerning asymptotic behavior of mse . as the first case , in the estimations of depolarizing channels and pauli channels
, the optimal mse behaves as @xmath0@xcite . as the second case , in the estimation of unitary , the optimal mse behaves as @xmath5@xcite . in the second case ,
two different types of results were reported : one is based on the cramr - rao approach@xcite .
the other is based on the mini - max approach@xcite .
the cramr - rao approach is based on the notion of locally unbiased estimator , and allows one to give a simple lower bound ( the cramr - rao bound ) to the mean square error ( mse ) at a given point .
the mini - max approach aims to minimize the maximum of the mse over all possible values of the parameter .
the mini - max approach is more meaningful than the cramr - rao approach because the true value of the parameter is unknown .
so , the cramr - rao bound is just a lower bound in general , while it can be asymptotically acheived in the case of quantum state estimation with the @xmath4 copies of the unknown state .
however , the cramr - rao bound has been considered so many times in the literatures@xcite .
the reason seems that computing the mini - max bound is a much harder problem .
indeed , there are only very few examples of calculations of the mini - max bound , and most of them are in the compact group covariant setting , in which , as was shown by holevo @xcite , the mini - max bound coincides with the bayesian average of the mse over the normalized invariant measure .
so , many researchers @xcite applied this approach to quantum channel estimation under the group covariance . as the most simple case for unitary estimation , phase estimation has been treated with the mini - max approach by several papers@xcite , and it has been shown that the minimum mse that behaves as @xmath5 . on the other hand ,
the cramr - rao approach suggests the noon state as the optimal input@xcite in phase estimation .
the later estimation scheme was experimentally demonstrated in the case of @xmath1 @xciteand @xmath6 @xcite .
also , another group @xcite experimentally demonstrated an estimation protocol concerning the group covariant approach proposed by @xcite . in phase estimation with group covariant framework
, the asymptotic minimum mse behaves as @xmath5@xcite .
when we focus on the difference @xmath7 between the true parameter @xmath3 and the estimate @xmath8 , the limiting distribution concerning the random variable @xmath9 can be obtained through fourier transform of a function with a finite domain @xcite .
however , the cramr - rao bound is different from the asymptotic limit of the mini - max bound .
so , these results seem to contradict with each other .
no existing research compares both approaches in a unified viewpoint .
the manuscript consists of two parts . in the first part ,
we discuss the cramr - rao approach for channel estimation , and give a simple formula to compute a bound based on the right logarithmic derivative introduced by holevo@xcite .
this formula is based on the choi - jamiolkowski representation matrix @xcite of the quantum channel , and holds under the condition that , at the given point , the support of the choi - jamiolkowski representation matrix contains the support of its derivative . under this condition
, we prove the additivity of the rld fisher information for quantum channels .
the second part is about the local asymptotic mini - max bound@xcite , in which the maximum of the mse is taken over an interval , the width of which is sent to zero after computing the asymptotic limits .
the obtained results are summarized as follows : ( 1 ) the local asymptotic mini - max bound is strictly larger than the cramr - rao bound in the case of phase estimation .
( 2 ) both bounds coincide with each other when the asymptotic minimum mse behaves as @xmath0 .
( 3 ) the local asymptotic mini - max bound is achievable ( proposition 3 ) , using a variation of the two - step strategy @xcite .
that is , there is an optimal sequence of estimators that achieves the local asymptotic mini - max bound at any point . as consequence
, the maximum of the local asymptotic mini - max bound is shown to be equal to the asymptotic limit of the mini - max bound .
however , the cramr - rao bound has a different type of achievability .
that is , there is an optimal sequence of estimators that achieves the cramr - rao at a given point .
this characteristic is a curious quantum analogue of superefficiency in classical statistics . in estimation of probability distribution , if we assume a weaker condition for our estimator , there exists an estimator that surpasses the cramr - rao bound only in measure zero points .
such an estimator is called a superefficient estimator@xcite . in the estimation of phase action
, we point out that the cramr - rao bound can be attained only by a quantum channel version of a superefficient estimator that works at specific points .
since the cramr - rao approach is based on the asymptotically locally unbiased condition , we can conclude that the asymptotically locally unbiased condition is too weak for deriving the local asymptotic mini - max bound , which can be attained in all points .
indeed , a similar phenomenon happens in quantum state estimation when we use the large deviation criterion@xcite .
this paper is organized as follows .
some of obtained results are based on quantum state estimation with the cramr - rao approach .
section [ s2 ] is devoted to a review of the cramr - rao approach in quantum state estimation . in this section , the symmetric logarithmic derivative ( sld )
fisher information and the right logarithmic derivative ( rld ) fisher information are explained .
however , the cramr - rao bound is obtained only by the locally unbiased condition .
so , it is needed to discuss its relation with the estimator that works globally . in section [ s3 ] ,
we treat sld fisher information and rld fisher information in the quantum channel estimation , and discuss the increasing order of sld fisher information . in section [ s31 ] , we give several examples where the maximum sld fisher information increases with @xmath10 . in section [ s4 ] , we compare the local asymptotic mini - max bound and the cramr - rao bound .
we also show the global attainability of the local asymptotic mini - max bound in the channel estimation .
it is also shown that the cramr - rao bound is closely related to a quantum channel version of superefficiency in the phase estimation .
in quantum state estimation , we estimate the true state through the quantum measurement under the assumption that the true state of the given quantum system @xmath11 belongs to a certain parametric state family @xmath12 . in the following ,
we consider the case when the number @xmath13 of parameters is one .
usually , we assume that @xmath4 quantum systems are prepared in the state @xmath14 .
hence , the total system is described by the tensor product space @xmath15 , and the state of the total system is given by @xmath16 . in this case , when we choose a suitable measurement , the mse decreases in proportion to @xmath17 as in the estimation of probability distribution .
so , we focus on the first order coefficient of the mse concerning @xmath18 . in the most general
setting , any positive operator valued measure ( povm ) @xmath19 on the total system @xmath15 is allowed as an estimator when it takes values in the parameter space @xmath20 .
the mse is given as @xmath21 in the quantum case , there are several quantum extensions of fisher information when the state @xmath14 is differentiable at @xmath3 and @xmath22 , where @xmath23 is the projection to the support of @xmath14 .
the largest one is the right logarithmic derivative ( rld ) fisher information @xmath24 , and the smallest one is symmetric logarithmic derivative ( sld ) fisher information @xmath25 . for these definitions ,
we define the rld @xmath26 and the sld @xmath27 as the operators satisfying @xmath28 then , the rld and sld fisher informations are given by @xcite @xmath29 when the range of @xmath14 contains the range of @xmath30 , the rld fisher information has another expression : @xmath31 when the state family @xmath32 is given by @xmath33 , the condition ( [ 20 - 11-a ] ) does not hold , where @xmath34 is an hermitian matrix . in this case
, the sld fisher information is calculated as follows@xcite .
@xmath35 now , we introduce the unbiased condition by @xmath36 however , this condition is sometimes too restrictive in the asymptotic setting .
so , we consider the taylor expansion at a point @xmath37 and focus on the first order .
then , we obtain the locally unbiased condition at @xmath37 : @xmath38 under the locally unbiased condition at @xmath37 , an application of schwarz inequality similar to the classical case yields the quantum cramr - rao inequalities for both quantum fisher information .
@xmath39 since @xmath24 is greater than @xmath25 , the inequality ( [ 12 - 1 - 1 ] ) is more informative than the inequality ( [ 12 - 1 - 2 ] ) . when the estimator @xmath19 is the spectral decomposition of the operator @xmath40 , the equality in ( [ 12 - 1 - 1 ] ) holds , where @xmath41 is given as @xmath42
. then , we obtain the following inequality @xmath43 this inequality seems to imply that @xmath24 is not as meaningful as @xmath25 in the one - parametric case . however , as will be explained latter , @xmath24 provides a meaningful bound for mse in the case of channel estimation .
in fact , in the asymptotic setting , a suitable estimator usually satisfies the asymptotic locally unbiased condition : @xmath44 for all points @xmath37 . under the above condition , using ( [ 12 - 1 - 1 ] )
, we obtain the inequality @xmath45 further , by using the two - step method , the bound @xmath46 can be universally attained for any true parameter @xmath3 @xcite .
so , defining the cramr - rao bound : @xmath47 we obtain @xmath48 on the other hand , its multi - parameter case is much complicated even in the asymptotic setting @xcite .
so , this paper does not treat the multi - parameter case .
in this section , we apply the cramr - rao approach to estimation of channel . in the quantum system , the channel is given by a trace preserving completely positive ( tp - cp ) map @xmath49 from the set of densities on the input system @xmath50 to the set of densities on the output system @xmath51 . by using @xmath52 linear maps @xmath53 from @xmath54 to @xmath55
, any tp - cp map @xmath49 can be described by @xmath56 .
hence , our task is to estimate the true tp - cp map under the assumption that the true tp - cp map belongs to a certain family of tp - cp maps @xmath57 . in order to characterize a tp - cp map @xmath2
, we formulate the notation concerning states on the tensor product system @xmath58 , where @xmath59 is a system of the same dimensionality as @xmath11 and is called the reference system . using a linear map @xmath60 from @xmath59 to @xmath11 , we define an element @xmath61 of @xmath58 as follows .
@xmath62 where @xmath63 and @xmath64 are complete orthonormal systems ( conss ) of @xmath11 and @xmath59 .
hence , the relation @xmath65 holds . this notation is applied to the cases of @xmath66 and @xmath67 .
now , we focus on the matrix @xmath68:=(\lambda_\theta\otimes \id ) ( |i \rangle \rangle \langle \langle i|)$ ] , which is called the choi - jamiolkowski representation matrix @xcite .
then , when the input state is the maximally entangled state @xmath69 , the output state is @xmath70 = ( \lambda_\theta\otimes \id ) ( | \frac{1}{\sqrt{d}}i \rangle \rangle \langle\langle \frac{1}{\sqrt{d}}i|)$ ] .
when the matrix @xmath71 is a density matrix on @xmath11 , @xmath72 is a pure state on the product system @xmath58 .
thus , the output state is given as @xmath73(i\otimes \overline{a}).\end{aligned}\ ] ] in the one - parameter case , we express the derivative @xmath74 by @xmath75 $ ] . when the input state is the product state @xmath76 , the output state is @xmath77 ( i \otimes \overline{v } \cdot u^{\dagger } ) $ ] . since @xmath78 = u \cdot v^{t } \tr_{{\cal k } } \rho[\lambda_\theta ] \overline{v }
\cdot u^{\dagger } \\
= & \tr_{{\cal k } } ( i \otimes u \cdot v^{t } ) \rho[\lambda_\theta ] ( i \otimes \overline{v } \cdot u^{\dagger } ) = @xmath79 ) | \overline{v}\rangle = 1.\end{aligned}\ ] ] thus , we obtain @xmath80 = i.\label{20 - 4}\end{aligned}\ ] ] taking the derivative in ( [ 20 - 4 ] ) , we obtain @xmath81 = 0.\label{20 - 5}\end{aligned}\ ] ] now , we back to our estimation problem . in this problem ,
our choice is given by a pair of the input state @xmath82 and the quantum measurement @xmath83 .
when we fix the input state , our estimation problem can be reduced to the state estimation with the state family @xmath84 . in the one - parameter case , we focus on the suprema @xmath85:= \sup_{\rho } j^r[\lambda_\theta,\rho ] , \quad j^s[\lambda_\theta]:= \sup_{\rho } j^s[\lambda_\theta,\rho],\end{aligned}\ ] ] where @xmath86 $ ] and @xmath87 $ ] are the sld and rld fisher informations when the input state is @xmath82 . in particular , it is important to calculate the supremum @xmath88 $ ] which is smaller than @xmath89 $ ] .
when @xmath4 applications of the unknown channel @xmath2 are available , the input state @xmath90 and the measurement @xmath19 are given as a state on @xmath91 and a povm on @xmath92 .
for a sequence of estimators @xmath93 , we consider the asymptotic locally unbiased condition : @xmath94 for all points @xmath37 , and denotes the mse of @xmath95 by @xmath96 .
assume that @xmath97 $ ] behaves as @xmath98 when @xmath4 goes to infinity .
when @xmath99 satisfies the asymptotic locally unbiased condition , the inequality ( [ 22 - 1 ] ) yields that @xmath100}.\end{aligned}\ ] ] we define the cramr - rao bound : @xmath101\\ : = & \inf_{\{\rho_n , m^n\ } } \left\{\limsup_{n \to\infty } n^\alpha \mse_\theta(\rho_n , m^n ) \left| \begin{array}{l } \{(\rho_n , m^n)\}\hbox { satisfies the asymptotic}\\ \hbox { locally unbiased condition . }
\end{array } \right .
\right\}.\end{aligned}\ ] ] thus , we obtain @xmath102= \limsup_{n \to \infty } \frac{n^\alpha}{j^s[\lambda_\theta^{\otimes n}]}. \label{3 - 20 - 1}\end{aligned}\ ] ] since @xmath103 \ge j^s[\lambda_\theta^{\otimes n}]+ j^s[\lambda_\theta^{\otimes m } ] , \label{4 - 23 - 2}\end{aligned}\ ] ] the limit @xmath104}{n}$ ] exists .
( for example , see lemma a.1 of @xcite . )
thus , @xmath105 $ ] can be defined by @xmath106}$ ] . in order to treat the above values
, we consider the following condition : ( c ) : : the range of @xmath68 $ ] contains the range of @xmath75 ^ 2 $ ] .
assume that the condition ( c ) does not hold .
when the input state is the maximally entangled state @xmath107 , the rld fisher information diverges .
so , @xmath89 $ ] is infinity .
when the condition ( c ) holds , @xmath85 = \|\tr_{{\cal k } } d[\lambda_\theta ] \rho[\lambda_\theta]^{-1 } d[\lambda_\theta ] \|.\end{aligned}\ ] ] assume that the input state is given by @xmath108 and @xmath60 is an invertible matrix .
then , the range of @xmath109(i\otimes \overline{a})$ ] contains the range of @xmath110(i\otimes \overline{a}))^2 $ ] .
assume that @xmath68 $ ] is invertible .
using the formula ( [ 20 - 11-a ] ) , we obtain @xmath111= \tr((i\otimes a^t)d[\lambda_\theta](i\otimes \overline{a}))^2 ( ( i\otimes a^t)\rho[\lambda_\theta](i\otimes \overline{a}))^{-1 } \\ = & \tr ( i\otimes a^t)d[\lambda_\theta](i\otimes \overline{a } ) ( i\otimes a^t)d[\lambda_\theta](i\otimes \overline{a } ) ( i\otimes \overline{a})^{-1 } \rho[\lambda_\theta]^{-1 } ( i\otimes a^t)^{-1}\\ = & \tr ( i\otimes \overline{a } a^t)d[\lambda_\theta ] \rho[\lambda_\theta]^{-1 } d[\lambda_\theta]\\ = & \tr \overline{a } a^t ( \tr_{{\cal k } } d[\lambda_\theta ] \rho[\lambda_\theta]^{-1 } d[\lambda_\theta]),\end{aligned}\ ] ] where @xmath112(i\otimes \overline{a}))^{-1}$ ] is the inverse of @xmath109(i\otimes \overline{a})$ ] on its range .
so , the supremum of @xmath113 \rho[\lambda_\theta]^{-1 } d[\lambda_\theta])$ ] with the condition @xmath114 equals @xmath115 \rho[\lambda_\theta]^{-1 } d[\lambda_\theta]\|$ ] .
assume that @xmath68 $ ] is non - invertible .
we choose an arbitrary small real number @xmath116 .
similar calculations and the operator monotonicity of @xmath117 yield that @xmath118(i\otimes \overline{a}))^2 ( ( i\otimes a^t)(\rho[\lambda_\theta]+\epsilon i)(i\otimes \overline{a}))^{-1 } \nonumber \\ \le & \tr \overline{a } a^t ( \tr_{{\cal k } } d[\lambda_\theta ] ( \rho[\lambda_\theta]+\epsilon i)^{-1 } d[\lambda_\theta ] ) \nonumber \\ \le & \tr \overline{a } a^t ( \tr_{{\cal k } } d[\lambda_\theta ] \rho[\lambda_\theta]^{-1 } d[\lambda_\theta]),\label{8 - 2 - 1}\end{aligned}\ ] ] which is a bounded value due to condition ( c ) .
now , we focus on two matrixes on the range of @xmath109(i\otimes \overline{a})$ ] , @xmath119(i\otimes \overline{a})$ ] and @xmath110(i\otimes \overline{a}))^2 $ ] . since the right hand side of ( [ 8 - 2 - 1 ] ) is independent of @xmath120 , the range of @xmath109(i\otimes
\overline{a})$ ] contains that of @xmath110(i\otimes \overline{a}))^2 $ ] .
so , taking the limit @xmath121 , we obtain @xmath111 = \tr ( ( i\otimes a^t)d[\lambda_\theta](i\otimes \overline{a}))^2 ( ( i\otimes a^t)\rho[\lambda_\theta](i\otimes \overline{a}))^{-1 } \\ = & \lim_{\epsilon \to + 0 } \tr((i\otimes a^t)d[\lambda_\theta](i\otimes \overline{a}))^2 ( ( i\otimes a^t)(\rho[\lambda_\theta]+\epsilon i)(i\otimes \overline{a}))^{-1 } \\ \le & \tr \overline{a } a^t ( \tr_{{\cal k } } d[\lambda_\theta ] \rho[\lambda_\theta]^{-1 } d[\lambda_\theta]).\end{aligned}\ ] ] the remaining task is to show the inequality @xmath122
\le \|\tr_{{\cal k } } d[\lambda_\theta ] \rho[\lambda_\theta]^{-1 } d[\lambda_\theta]\| \label{3 - 8}\end{aligned}\ ] ] for a non - invertible matrix @xmath60 .
define @xmath123 and @xmath124 as the inverses of @xmath125 and @xmath126 whose domains are the ranges of the matrixes @xmath125 and @xmath127 . letting @xmath23 and @xmath128 be the projections to the ranges of the matrixes @xmath125 and @xmath127 , we have @xmath129 , and @xmath130 .
then , @xmath109(i\otimes \overline{a } ) = p(i\otimes a^t)\rho[\lambda_\theta](i\otimes \overline{a})p = p(i\otimes a^t)p'\rho[\lambda_\theta]p'(i\otimes \overline{a})p$ ] . in the following ,
we consider the case when @xmath68 $ ] is invertible .
the matrix @xmath131 p')^{-1}$ ] is defined as the inverse of @xmath132 p'$ ] whose domain and range are the range of @xmath128 .
thus , @xmath111= \tr p((i\otimes a^t)d[\lambda_\theta](i\otimes \overline{a}))^2 p ( p(i\otimes a^t)\rho[\lambda_\theta](i\otimes \overline{a})p)^{-1 } \nonumber\\ = & \tr ( i\otimes a^t ) d[\lambda_\theta ] p'(i\otimes \overline{a } ) p ( i\otimes \overline{a})^{-1 } ( p ' \rho[\lambda_\theta ] p')^{-1 } ( i\otimes a^t)^{-1}p ( i\otimes a^t ) p ' d[\lambda_\theta ] ( i\otimes \overline{a})\nonumber \\ = & \tr ( i\otimes a^t ) d[\lambda_\theta ] p'(p ' \rho[\lambda_\theta ] p')^{-1}p ' d[\lambda_\theta ] ( i\otimes \overline{a})\nonumber \\ = & \tr ( p ' \rho[\lambda_\theta ] p')^{-1 } p ' d[\lambda_\theta ] ( i\otimes \overline{a } a^t ) d[\lambda_\theta ] p ' \nonumber \\ = & \tr ( p ' \rho[\lambda_\theta ] p')^{-1 } d[\lambda_\theta ] ( i\otimes \overline{a } a^t ) d[\lambda_\theta ] \nonumber \\ \le & \tr \rho[\lambda_\theta]^{-1 } d[\lambda_\theta ] ( i\otimes \overline{a } a^t ) d[\lambda_\theta ] \label{3 - 12}\\ = & \tr \overline{a } a^t ( \tr_{{\cal k } } d[\lambda_\theta ] \rho[\lambda_\theta]^{-1 } d[\lambda_\theta])\nonumber .\end{aligned}\ ] ] therefore , the inequality ( [ 3 - 8 ] ) holds .
next , we consider the case when @xmath68 $ ] is non - invertible .
we choose an arbitrary small real number @xmath133 .
similar calculations and the operator monotonicity of @xmath117 yield that @xmath134(i\otimes \overline{a}))^2 p ( p(i\otimes a^t ) ( \rho[\lambda_\theta ] + \epsilon
i ) ( i\otimes \overline{a})p)^{-1 } \nonumber\\ \le & \tr \overline{a } a^t ( \tr_{{\cal k } } d[\lambda_\theta ] ( \rho[\lambda_\theta]+\epsilon i)^{-1 } d[\lambda_\theta ] ) \nonumber\\ \le & \tr \overline{a } a^t ( \tr_{{\cal k } } d[\lambda_\theta ] \rho[\lambda_\theta]^{-1 } d[\lambda_\theta ] ) , \label{8 - 1 - 1}\end{aligned}\ ] ] which is a bounded value due to condition ( c ) . now , we focus on two matrixes on the range of @xmath23 , @xmath135 ( i\otimes \overline{a})p$ ] and @xmath136(i\otimes \overline{a}))^2 p$ ] .
since the right hand side of ( [ 8 - 1 - 1 ] ) is independent of @xmath120 , the range of @xmath135 ( i\otimes \overline{a})p$ ] contains that of @xmath136(i\otimes \overline{a}))^2 p$ ] .
so , taking the limit @xmath121 , we obtain @xmath111= \tr p((i\otimes a^t)d[\lambda_\theta](i\otimes \overline{a}))^2 p ( p(i\otimes a^t)\rho[\lambda_\theta](i\otimes \overline{a})p)^{-1 } \nonumber\\ = & \lim_{\epsilon \to + 0 } \tr p((i\otimes a^t)d[\lambda_\theta](i\otimes \overline{a}))^2 p ( p(i\otimes a^t ) ( \rho[\lambda_\theta ] + \epsilon i ) ( i\otimes \overline{a})p)^{-1 } \nonumber \\ \le & \tr \overline{a } a^t ( \tr_{{\cal k } } d[\lambda_\theta ] \rho[\lambda_\theta]^{-1 } d[\lambda_\theta])\nonumber .\end{aligned}\ ] ] therefore , the inequality ( [ 3 - 8 ] ) holds . when two channel families @xmath137 and @xmath138 satisfy the condition ( c ) , then the additivity @xmath139 = j^r[\lambda_\theta ] + j^r[\tilde{\lambda}_\theta]\end{aligned}\ ] ] holds .
let @xmath140 and @xmath141 be output systems of the channels @xmath142 and @xmath143 .
then , the relation ( [ 20 - 5 ] ) guarantees @xmath144 \otimes d[\tilde{\lambda}_\theta ] = \tr_{{\cal k } } d[\lambda_\theta ] \tr_{\tilde{{\cal k } } } d[\tilde{\lambda}_\theta ] = 0.\label{4 - 7 - 2}\end{aligned}\ ] ] since @xmath145 = d[\lambda_\theta]\otimes \rho[\tilde{\lambda}_{\theta } ] + \rho[\lambda_\theta]\otimes d[\tilde{\lambda}_{\theta}]$ ] , theorem [ th1 ] and ( [ 4 - 7 - 2 ] ) yield @xmath146\nonumber \\ = & \| \tr_{{\cal k}\otimes \tilde{{\cal k } } } ( d[\lambda_\theta]\otimes \rho[\tilde{\lambda}_{\theta } ] + \rho[\lambda_\theta ] \otimes d[\tilde{\lambda}_{\theta } ] ) ( \rho[\lambda_\theta ] ^{-1}\otimes \rho[\tilde{\lambda}_{\theta}]^{-1 } ) ( d[\lambda_\theta]\otimes \rho[\tilde{\lambda}_{\theta } ] + \rho[\lambda_\theta ] \otimes d[\tilde{\lambda}_{\theta } ] ) \| \nonumber \\ = & \| \tr_{{\cal k}\otimes \tilde{{\cal k } } } ( d[\lambda_\theta ] \rho[\lambda_\theta ] ^{-1 } d[\lambda_\theta ] \otimes \rho[\tilde{\lambda}_{\theta } ] + \rho[\lambda_\theta ] \otimes d[\tilde{\lambda}_\theta ] \rho[\tilde{\lambda}_{\theta}]^{-1 } d[\tilde{\lambda}_\theta ] + 2d[\lambda_\theta ] \otimes d[\tilde{\lambda}_\theta ] ) \| \nonumber \\ = & \| ( \tr_{{\cal k } } d[\lambda_\theta ] \rho[\lambda_\theta ] ^{-1 } d[\lambda_\theta])\otimes i + i \otimes ( \tr_{\tilde{{\cal k } } } d[\tilde{\lambda}_\theta ] \rho[\tilde{\lambda}_{\theta}]^{-1 } d[\tilde{\lambda}_\theta ] ) \| \nonumber \\ = & \| \tr_{{\cal k } } d[\lambda_\theta ] \rho[\lambda_\theta ] ^{-1 } d[\lambda_\theta]\| + \| \tr_{\tilde{{\cal k } } } d[\tilde{\lambda}_\theta]\rho[\tilde{\lambda}_{\theta}]^{-1}d[\tilde{\lambda}_\theta ] \| \label{4 - 7 - 1 } \\ = & j^r[\lambda_\theta ] + j^r[\tilde{\lambda}_\theta],\nonumber \end{aligned}\ ] ] where the equation ( [ 4 - 7 - 1 ] ) follows from the additivity property concerning matrix norm : @xmath147 for any two hermitian matrixes @xmath148 and @xmath149 . when a channel family @xmath137 satisfies the condition ( c ) , then @xmath150= n j^r[\lambda_\theta].\end{aligned}\ ] ] since @xmath151 \le j^s[\lambda_\theta^{\otimes
n } ] \le j^r[\lambda_\theta^{\otimes n}]$ ] , @xmath97 $ ] increases in order @xmath4 under the assumption of theorem [ th1 ] , i.e. , @xmath97=o(n)$ ] .
when the rank of @xmath68 $ ] is the maximum , i.e. , @xmath52 , this condition holds and @xmath97=o(n)$ ] .
however , there is an example that does not satisfy the above condition but satisfies the condition ( c ) as follows .
so , the condition ( c ) is weaker than the condition that @xmath68 $ ] has the maximum rank .
a channel @xmath49 is called a phase damping channel when the output system @xmath140 equals the input system @xmath11 and there exist complex numbers @xmath152 such that @xmath153 where @xmath154 . in this case , the state @xmath155 $ ] is written as the following form @xmath156= \sum_{k , l } d_{k , l}|k\rangle|k\rangle \langle l|\langle l| .\end{aligned}\ ] ] that is , the range of @xmath155 $ ] is included by the space spanned by @xmath157 . when a channel family @xmath57 is given as a one - parameter subfamily of @xmath158 , the condition ( c ) holds
therefore , there exists a channel family that satisfies the condition ( c ) but consists of non - full - rank channels .
further , we have the following observation .
assume that the condition ( c ) holds and there exists a normalized vector @xmath159 in the input system @xmath11 such that @xmath115 \rho[\lambda_\theta]^{-1 } d[\lambda_\theta]\| = \langle u| \tr_{{\cal k } } d[\lambda_\theta ] \rho[\lambda_\theta]^{-1 } d[\lambda_\theta]|u \rangle$ ] , @xmath160|u \rangle , \langle u| d[\lambda_\theta]|u \rangle]=0 $ ] , and @xmath161=0 $ ] .
then , @xmath88 = j^r[\lambda_\theta]$ ] , and this bound can be attained by the input pure state @xmath162 on @xmath11 .
that is , it can be attained without use of the reference system . since @xmath160|u \rangle , \langle u| d[\lambda_\theta]|u \rangle]=0 $ ] , @xmath163 = j^r[\lambda_\theta,|u \rangle\langle u|]$ ] .
so , it is sufficient to show that @xmath164 = \langle u| \tr_{{\cal k } } d[\lambda_\theta ] \rho[\lambda_\theta]^{-1 } d[\lambda_\theta]|u \rangle$ ] .
we consider the case when the input state is @xmath165 .
remember that @xmath166 = j^r[\lambda_\theta,|u \rangle\langle u|]$ ] .
since @xmath161=0 $ ] , the equality in ( [ 3 - 12 ] ) holds for @xmath167 .
then , @xmath168 = \tr \overline{(u \cdot v^t ) } ( u \cdot v^t)^{t } \tr_{{\cal k } } d[\lambda_\theta ] \rho[\lambda_\theta]^{-1}d[\lambda_\theta ] \\
= & \langle u| \tr_{{\cal k } } d[\lambda_\theta ] \rho[\lambda_\theta]^{-1}d[\lambda_\theta]|u \rangle.\end{aligned}\ ] ]
we consider the case where the condition ( c ) does not hold . as the simplest example , we consider the one - parameter unitary case , i.e. , the case when @xmath169 with an hermitian matrix @xmath34 . using ( [ 21 - 1 ] ) , we obtain @xmath170 = ( \lambda_{\max}(h)-\lambda_{\min}(h))^2,\end{aligned}\ ] ] where @xmath171 and @xmath172 are the maximum and minimum of eigenvalues of @xmath34 .
so , we obtain @xmath173 = n^2 ( \lambda_{\max}(h)-\lambda_{\min}(h))^2 . \ ] ] in particular , in the two - dimensional case , when @xmath174 , the optimal input is @xmath175 , which is called the noon state . this unitary estimation is called phase estimation and this estimation with the noon state was experimentally realized with @xmath1@xcite and @xmath6@xcite .
next , we consider the @xmath13-dimensional system @xmath176 spanned by @xmath177 and the unitary matrix @xmath148 defined as @xmath178 mod @xmath13 . using a distribution @xmath179 and a real diagonal matrix @xmath34 with diagonal elements @xmath180
, we define the tp - cp map @xmath2 by @xmath181 this tp - cp map can be regarded as the stochasitc application of the unitary @xmath182 after the application of the unitary @xmath183 .
let @xmath184 and @xmath185 be the maximum and the minimum eigenvalues of @xmath34 . using two - dimensional reference system spanned by @xmath186 and @xmath187 , we choose the following input state : @xmath188 in this case , as the first step , we apply the following measurement @xmath189 : @xmath190 when the outcome of this measurement is @xmath191 , the resulting state is the pure state @xmath192 the sld fisher information of the above family is @xmath193 .
therefore , since the maximum sld fisher information behaves as @xmath10 at most , @xmath194 $ ] behaves as @xmath10 .
in this section , we consider the relation between the discussion in the previous section and estimating protocols in a different viewpoint .
consider the phase estimation with inputing the noon state @xmath175 .
then , the output state is @xmath195 . in this case
, we can not distinguish the parameters @xmath3 and @xmath196 .
for example , when we apply measurement @xmath197 , the probability with the outcome @xmath175 equals @xmath198 , as is shown in fig [ f3 ] , and the fisher information equals @xmath199 . even if the parameter @xmath3 is assumed to belong to @xmath200 $ ]
, we can not distinguish the two parameters @xmath201 and @xmath202 with so high probability because we have only two outcomes . in order to distinguish two parameters @xmath201 and @xmath202 in this measurement
, we need to repeat this measurement with several times , e.g. , @xmath203 .
since the number of application of the unknown unitary is @xmath204 , the error behaves as @xmath205 , which is different from @xmath206 .
so , we can not conclude that the above method attains the order @xmath206 concerning mse .
therefore , we need to discuss what a bound can be attained globally , more carefully .
for this purpose , we focus on an @xmath120-neighborhood @xmath207 of @xmath3 and define the local asymptotic mini - max risk@xcite : @xmath208 & : = \lim_{\epsilon \to 0 } \limsup_{n \to\infty } \sup_{\theta \in u_{\theta_0,\epsilon } } n^\alpha \mse_\theta(\rho_n , m^n)\end{aligned}\ ] ] and the _ local asymptotic mini - max _ bound : @xmath209 & : = \inf_{\{(\rho_n , m^n)\ } } c_\alpha [ \lambda_{\theta_0},\{(\rho_n , m^n)\}].\end{aligned}\ ] ] concerning the local asymptotic mini - max bound , we have the following two propositions .
when @xmath210 $ ] is continuous and the convergence ( [ 3 - 20 - 1 ] ) is compactly uniform , @xmath209 \ge \tilde{c}_\alpha [ \lambda_{\theta_0 } ] .
\label{22 - 9}\end{aligned}\ ] ] for any @xmath211 , we choose an integer @xmath212 and an @xmath120-neighborhood @xmath213 satisfying @xmath209 + \delta \ge n^\alpha \mse_\theta(\rho_n , m^n ) , \quad \forall
n \ge n , \forall \theta \in u_{\theta_0,\epsilon}. \label{3 - 15 - 1}\end{aligned}\ ] ] we introduce two quantities @xmath214 then , we obtain @xmath215 let @xmath216 be the fisher information of the distribution family @xmath217 . this quantity is smaller than @xmath97 $ ] . deforming the classical cramr - rao inequality , we obtain @xmath218
thus , we obtain @xmath219 as is shown later , for any @xmath211 , there exists a sufficiently large integer @xmath212 satisfying the following . for any @xmath220 , there exists @xmath221 such that @xmath222 using ( [ 22 - 7 ] ) , we obtain @xmath223 take the limit @xmath224 .
then , the continuity of @xmath210 $ ] , the compact uniformity of the convergence ( [ 3 - 20 - 1 ] ) , ( [ 3 - 15 - 1 ] ) , and ( [ 4 - 30 - 1 ] ) imply that @xmath209+\delta \ge ( 1-\delta)^2 \tilde{c}_\alpha [ \lambda_{\theta_0}].\end{aligned}\ ] ] taking the limit @xmath225 , we obtain ( [ 22 - 9 ] ) .
finally , we show the existence of the integer @xmath212 satisfying the above condition given by ( [ 3 - 15 - 2 ] ) by using reduction to absurdity .
assume that for any @xmath211 , there exists a subsequence @xmath226 such that @xmath227 for any @xmath228 .
thus , @xmath229 - [ \eta_{n_k}(\theta-\epsilon/2)-(\theta-\epsilon/2 ) ] \\ = & \eta_{n_k}(\theta+\epsilon/2)-\eta_{n_k}(\theta-\epsilon/2 ) -\epsilon < - \epsilon \delta.\end{aligned}\ ] ] then , @xmath230 that is , @xmath231 using ( [ 22 - 7 ] ) , we obtain @xmath232 since @xmath233 behaves as @xmath234 , we obtain contradiction .
assume that @xmath235 .
when the order parameter @xmath236 equals @xmath237 and @xmath210 $ ] is continuous , @xmath238 = \tilde{c}_1 [ \lambda_{\theta_0 } ] = \lim_{n \to \infty } \frac{n}{j^s[\lambda_\theta^{\otimes n}]}.\end{aligned}\ ] ] as is mentioned in section [ s3 ] , ( [ 4 - 23 - 2 ] ) guarantees the convergence of @xmath239}$ ] .
it is enough to show the inequality @xmath240\ge
c_1 [ \lambda_{\theta_0}]$ ] . for an arbitrary real number @xmath211 and an arbitrary integer @xmath241 ,
let @xmath242 be a locally unbiased estimator at @xmath37 such that @xmath243}+\delta > \mse_{\theta_0}(\rho_m , m^m ) .
\label{3 - 15 - 3 - 2}\end{aligned}\ ] ] we define another coordinate @xmath244 by @xmath245 and denote the mse concerning the parameter @xmath246 of an estimator @xmath95 by @xmath247 . for any @xmath211 ,
we choose an integer @xmath241 , a sufficiently small @xmath248-neighborhood @xmath213 such that @xmath249}{m } + \delta & \ge \frac{1}{j^s[\lambda_{\theta_0}^{\otimes m } ] } \label{3 - 15 - 3 } \\
1-\delta \le \frac{\eta(\theta)-\eta(\theta')}{\theta-\theta ' } & \le 1+\delta \label{22 - 5 } \ ] ] for @xmath250 . next ,
let @xmath251 be the estimator given as the average value of @xmath4 times applications of the estimator @xmath252 concerning the original parameter @xmath3 . then
, we can choose a sufficiently large number @xmath4 satisfying the following : when the true parameter is @xmath37 , the estimate of @xmath251 belongs to @xmath213 with the probability higher than @xmath253 .
the second inequality in ( [ 22 - 5 ] ) guarantees that @xmath254 thus , ( [ 3 - 15 - 3 - 2 ] ) , ( [ 3 - 15 - 3 ] ) , and ( [ 22 - 7-a - b ] ) imply that @xmath255}{nm } + \frac{2\delta}{n } + \delta
e ) \ge \mse_{\eta(\theta_0)}(\rho_{nm}',{m^{nm}}').\end{aligned}\ ] ] since @xmath256 is continuous and @xmath257}{nm } + \delta ( e+2 ) ) > ( 1+\delta)^2 ( \frac{\tilde{c}_1 [ \lambda_{\theta_0}]}{nm } + \frac{2 \delta}{n } + \delta e)$ ] , we can choose a sufficiently small number @xmath258 such that @xmath259}{nm } + \delta ( e+2 ) ) \ge \mse_{\eta(\theta)}(\rho_{nm}',{m^{nm}}')\label{4 - 23 - 4}\end{aligned}\ ] ] for @xmath260 .
let @xmath261 be the estimator given as the average value of @xmath262 times applications of the estimator @xmath251 concerning @xmath246 . since the estimator @xmath251 is unbiased concerning the parameter @xmath246 , @xmath263 we choose a sufficiently large number @xmath262 satisfying the following : when the true parameter is @xmath264 , the estimate @xmath246 of @xmath261 belongs to @xmath265 with the probability @xmath266 , where the probability @xmath267 exponentially goes to @xmath268 as @xmath269 .
the first inequality in ( [ 22 - 5 ] ) guarantees that @xmath270 thus , the relations ( [ 4 - 23 - 4 ] ) , ( [ 22 - 6 ] ) , and ( [ 22 - 7-a ] ) imply @xmath271}{nm } + \delta ( e+2 ) ) + e p_l \ge \mse_{\theta}(\rho_{lnm}'',{m^{lnm } } '' ) , \quad \forall \theta \in u_{\theta_0,\epsilon'/2}. \label{22 - 7-e}\end{aligned}\ ] ] taking the limit @xmath272 , we obtain @xmath273 + nm \delta ( e+2 ) ) \ge \lim_{l \to \infty } lnm \mse_{\theta}(\rho_{lnm}'',{m^{lnm } } '' ) , \quad \forall \theta \in u_{\theta_0,\epsilon'/2}.\end{aligned}\ ] ] finally , using the above @xmath4 and @xmath241 , we define a sequence of estimators @xmath274 by the following way . for given @xmath203 ,
we choose maximum @xmath262 such that @xmath275 .
then , the estimator @xmath276 defined as @xmath277 . in this definition , we only use @xmath278 applications , and the remaining @xmath279 applications are discarded .
so , we obtain @xmath273 + nm \delta ( e+2 ) ) \ge \limsup_{k \to \infty } k \mse_{\theta}(\rho_{k}''',{m^{k } } ' '' ) , \quad \forall \theta \in u_{\theta_0,\epsilon'/2},\end{aligned}\ ] ] which implies that @xmath273 + nm \delta ( e+2 ) ) \ge { c}_1 [ \lambda_{\theta_0},\{(\rho_{k}''',{m^{k}}''')\}_{k=1}^{\infty } ] \ge { c}_1 [ \lambda_{\theta_0}].\end{aligned}\ ] ] since @xmath211 is arbitrary , @xmath240\ge c_1 [ \lambda_{\theta_0}]$ ] .
now , remember that the cramr - rao bound can be attained by using the two - step method in the case of state estimation . by using the two - step method@xcite , the local asymptotic mini - max bound @xmath280 $ ]
can be attained at all points @xmath3 as follows .
assume that @xmath235 and @xmath281 $ ] is continuous .
for any @xmath211 , there exists a sequence of estimators @xmath282 such that @xmath283 \le c_\alpha [ \lambda_{\theta}]+\delta \label{7 - 24 - 3}\end{aligned}\ ] ] for all points @xmath3 .
further , when the parameter space @xmath284 is compact , @xmath285 .
\label{7 - 24 - 2}\end{aligned}\ ] ] we use a two - step method slightly different from @xcite . before applying the unknown channel @xmath2 , for any real number @xmath211 , we choose an @xmath286-neighborhood @xmath287 satisfying the following three conditions : ( 1 ) @xmath288 .
( 2 ) for any @xmath289 , there exists a sequence of estimators @xmath290 such that @xmath291+\delta/2 \ge \lim_{\epsilon \to 0 } \limsup_{n \to\infty } n^\alpha \sup_{\theta \in u_{\theta_i,\epsilon } } \mse_\theta(\rho_n(\theta_i ) , m^n(\theta_i ) ) $ ] .
( 3 ) @xmath292 \le c_\alpha [ \lambda_{\theta}]+\delta/2 $ ] .
we divide @xmath4 applications of the unknown channel @xmath2 to two groups : the first group consists of @xmath293 applications and the second group consists of @xmath294 applications . in the first step ,
we apply a povm @xmath83 to the first group .
this povm @xmath83 is a povm on the single system @xmath11 and is required to satisfy that @xmath295 is non - degenerate at all points @xmath3 .
based on @xmath293 obtained data , we estimate which @xmath286-neighborhood @xmath287 contains the true parameter , and obtain the first step estimate @xmath296 .
the error probability @xmath297 of this step goes to @xmath268 exponentially , i.e. , @xmath297 behaves as @xmath298 , where @xmath299 depends on @xmath3 . in the second step ,
we apply the estimator @xmath300 to the second group , and obtain our final estimate from the outcome of the estimator @xmath301 .
we express this estimator by @xmath302 .
its mse is evaluated as @xmath303 since @xmath304 goes to @xmath268 , we obtain @xmath305+\delta/2 \le c_\alpha [ \lambda_{\theta}]+\delta.\end{aligned}\ ] ] thus , we obtain ( [ 7 - 24 - 3 ] ) .
further , the relation ( [ 7 - 24 - 1 ] ) yields that @xmath306 the compactness of @xmath284 guarantees that @xmath307 .
thus , @xmath308+\delta/2 \le \sup_{\theta\in \theta } c_\alpha [ \lambda_{\theta}]+\delta.\end{aligned}\ ] ] since the part @xmath309 of ( [ 7 - 24 - 2 ] ) is trivial , we obtain ( [ 7 - 24 - 2 ] ) .
proposition [ 22 - 1p ] holds even when we replace the mse by a general error function @xmath310 for one - parametric family satisfying the following conditions : ( 1 ) the relation @xmath311 holds with a local coordinate when @xmath8 is close to @xmath3 .
( 2 ) the maximum of @xmath310 exists .
therefore , we can apply proposition [ 22 - 1p ] to the following case : assume that the one - parameter channel family @xmath57 has a compact group covariant structure , that is , its parameter space is given as an interval @xmath312 and there is a unitary representation @xmath313 of @xmath314 such that @xmath315 .
the error is given by @xmath316 instead of the square error @xmath317 .
in this case , due to the group covariance , @xmath318 $ ] does not depend on the true parameter @xmath3 .
application of proposition [ 22 - 1p ] implies that the global min - max error behaves @xmath318\frac{1}{n^{\alpha}}$ ] . in the phase estimation case
, the unknown parameter @xmath3 belongs to @xmath319 , and the minimum of the worst value of the average error @xmath320 behaves as @xmath321 @xcite . that is , the leading decreasing order is @xmath322 and the leading decreasing coefficient is @xmath323 when we apply the optimal estimator .
proposition [ 22 - 1p ] implies that @xmath324=\pi^2 $ ] . since @xmath325 $ ] does not depend on @xmath3 due to the homogenous structure
, we can conclude that @xmath325=\pi^2 $ ] .
so , the equation @xmath97= n^2 $ ] implies the equation @xmath326=1 $ ] .
hence , the cramr - rao bound @xmath326 $ ] can not be attained globally in this model .
however , it can be attained in a specific point in the following sense .
assume that @xmath235 and @xmath281 $ ] is continuous .
for any @xmath211 and any @xmath327 , there exists a sequence of estimators @xmath328 satisfying the asymptotically locally unbiased condition and the relations : @xmath329+\delta , \quad\forall \theta\neq \theta_0 \\
\limsup_{n \to \infty } n^{\alpha}\mse_{\theta_0}(\rho_{n,\theta_0},m^n_{\theta_0 } ) & \le \tilde{c}_\alpha [ \lambda_{\theta_0}]+\delta.\end{aligned}\ ] ] in estimation of probability distribution , there exists a superefficient estimator that has smaller error at a discrete set than the cramr - rao bound@xcite .
since such a superefficient estimator can not be useful , many statisticians think that it is better to impose a condition for our estimators for removing superefficient estimators . in this classical case , if we assume the asymptotic locally unbiased condition , we have no superefficient estimator .
proposition [ pro1 ] means that even if the asymptotic locally unbiased condition is assumed , there exists an estimator that behaves in the similar way to a superefficient estimator in the case of unitary estimation .
so , we call such an estimator a q - channel - superefficient estimator . that is , a sequence of estimators @xmath99 is called _ q - channel - superefficient _ at @xmath3 with the order @xmath330 when @xmath331 $ ] . hence , in order to remove the q - channel - superefficiency problem , it is better to adopt the bound @xmath318 $ ] as the criterion instead of @xmath332 $ ] .
we choose @xmath286-neighborhoods @xmath287 in the same way , and define the neighborhood @xmath333 .
we apply the same first step as proposition [ 22 - 1p ] to neighborhoods @xmath334 , and obtain the first step estimate @xmath296 .
when the first step estimate @xmath296 is not @xmath37 , we apply the same method as proposition [ 22 - 1p ] in the second step .
when the first step estimate @xmath296 is @xmath37 , we apply the asymptotically locally unbiased estimator whose mse behaves as @xmath335+\delta)/n^\alpha$ ] asymptotically .
further , since there exists an asymptotically locally unbiased estimator that surpasses the bound @xmath318 $ ] , the asymptotically locally unbiased condition is too weak for deriving the local asymptotic mini - max bound , which is more meaningful . in order to avoid this problem , it is sufficient to impose the following condition : ( cu ) : : the limit @xmath336 exists for all @xmath3 and this convergence is compactly uniform concerning @xmath3 . under the condition ( cu )
, @xmath336 is continuous concerning @xmath3 , and @xmath337 thus , @xmath338.\end{aligned}\ ] ] therefore , we obtain the following corollary . when a sequence of estimators @xmath93 satisfies the condition ( cu ) , @xmath339.\end{aligned}\ ] ] therefore , the condition ( cu ) is better in estimation of quantum channel than the asymptotically locally unbiased condition .
finally , we consider the relation with the adaptive method proposed by nagaoka@xcite . in this method
, we apply our povm to each single system @xmath11 , and we decide the @xmath203-th povm based on the knowledge of previous @xmath340 outcomes . in this case , fujiwara @xcite analyzed the asymptotic behavior of the mse of this estimator .
now , we consider the case of @xmath341 applications of the unknown channel @xmath2 . in this case , we divide @xmath341 applications into @xmath4 groups consisting of @xmath241 applications .
when we apply the adaptive method mentioned in fujiwara@xcite to these groups , the mse of this estimator behaves as @xmath342}$ ] , which is close to @xmath343}{n m^{\alpha}}$ ] .
so , when @xmath344 , this method can not realize the optimal order @xmath345 .
we have compared the cramr - rao bound @xmath346 $ ] and the local asymptotic mini - max bound @xmath347 $ ] in quantum channel estimation , which contains quantum state estimation . when the model has group covariant structure , the local asymptotic mini - max bound @xmath347 $ ] coincides with the limit of the global mini - max bound .
we have also shown that both bounds @xmath346 $ ] and @xmath347 $ ] coincide in quantum channel estimation when the maximum of sld fisher information @xmath97 $ ] behaves as @xmath348 .
the case of state estimation can be regarded as a special case of this case .
that is , the conventional state estimation has no difference between both bounds .
however , we have shown that the cramr - rao bound @xmath346 $ ] is different from the local asymptotic mini - max bound @xmath347 $ ] in the phase estimation .
so , we can conclude that the local asymptotic mini - max bound @xmath347 $ ] is more meaningful and does not necessarily coincide with the cramr - rao bound @xmath346 $ ] .
in order to clarify the asymptotic leading order of @xmath97 $ ] , we have derived the condition ( c ) as a sufficient condition for @xmath97=o(n)$ ] .
that is , the condition @xmath349 \rho[\lambda_\theta]^{-1 } d[\lambda_\theta]\| = \infty$ ] is a necessary condition for square speedup .
this condition has been derived from the following two facts .
one is the supremum of the rld fisher information satisfies the additive property .
the other is the rld fisher information is an upper bound of the sld fisher information .
this , the supremum of the rld fisher information is the upper bound of the regularized supremum of the sld fisher information , which equals the inverse of the cramr - rao bound .
however , it is an open problem to clarify whether this upper bound can be attained by the regularized sld fisher information .
further , fujiwara and imai @xcite and matsumoto @xcite also obtained another sufficient condition . since the relation with their conditions is not clear , its clarification is an open problem .
our condition ( c ) trivially contains the case when the state @xmath68 $ ] is a full rank state on the tensor product system while it is not so easy to derive the above full rank condition from fujiwara and imai s condition .
further , we have also obtained another example for @xmath97=o(n^2)$ ] under the condition ( c ) .
this example is a larger class than the unitary model .
so , we can expect that @xmath97 $ ] behaves as @xmath10 if the condition ( c ) does not hold .
this is a challenging open problem .
the author was partially supported by a grant - in - aid for scientific research in the priority area ` deepening and expansion of statistical mechanical informatics ( dex - smi ) ' , no .
18079014 and a mext grant - in - aid for young scientists ( a ) no . 20686026 .
the centre for quantum technologies is funded by the singapore ministry of education and the national research foundation as part of the research centres of excellence programme .
the author thanks mr .
wataru kumagai for helpful comments .
he also thanks the referees for helpful comments concerning this manuscript .
in particular , the first referee s report was much help to improve the presentation in section 1 .
[ l3 - 8 ] any strictly positive definite matrix @xmath60 and any projection @xmath23 satisfy the inequality @xmath350 where @xmath351 is the inverse matrix with the domain @xmath23 .
let @xmath352 be the operator norm of the matrix @xmath353 .
l3 - 8 - 2 ] guarantees the inequality @xmath354 for any @xmath133 .
thus , @xmath355 since the function @xmath117 is operator monotone , @xmath356 taking the limit @xmath357 , we obtain ( [ 7 - 23 - 1 ] ) .
[ l3 - 8 - 2 ] let @xmath60 be a positive semi - definite matrix and @xmath23 be a projection .
then , the inequality @xmath358 holds for any @xmath133 , where @xmath352 is the operator norm of the matrix @xmath353 .
nagaoka h. : on the parameter estimation problem for quantum statistical models .
12th symp . on inform . theory and its appl .
_ p 577 ( 1989 ) ; nagaoka h. : _ asymptotic theory of quantum statistical inference _ ed m hayashi ( singapore : world scientific ) p 125 ( 2005 ) ( reprinted ) hayashi m. , matsumoto k. : statistical model with measurement degree of freedom and quantum physics .
_ rims koukyuroku _ no 1055 ( kyoto : kyoto university ) p 96 ( 1998 ) ( in japanese ) ; hayashi m. , matsumoto k. : _ asymptotic theory of quantum statistical inference_. ed m hayashi ( singapore : world scientific ) p 162 ( 2005 ) ( reprinted , english translation ) hayashi m. : quantum estimation and the quantum central limit theorem . in
selected papers on probability and statistics ( american mathematical society translations series 2 ) vol .
277 , pp 95 - 123 ( 2009 ) .
( it was originally published in japanese in bulletin of mathematical society of japan , sugaku , vol .
4 , 368391 ( 2003 ) . )
jones j. a. , karlen s. d. , fitzsimons j. , ardavan a. , benjamin s. c. , briggs g. a. d. , morton j. j. l. : magnetic field sensing beyond the standard quantum limit using 10-spin noon states .
_ science _ * 324 * , 1166 - 1168 ( 2009 ) | in a unified viewpoint in quantum channel estimation , we compare the cramr - rao and the mini - max approaches , which gives the bayesian bound in the group covariant model . for this purpose
, we introduce the _ local asymptotic mini - max _ bound , whose maximum is shown to be equal to the asymptotic limit of the mini - max bound .
it is shown that the local asymptotic mini - max bound is strictly larger than the cramr - rao bound in the phase estimation case while the both bounds coincide when the minimum mean square error decreases with the order @xmath0 .
we also derive a sufficient condition for that the minimum mean square error decreases with the order @xmath0 . |
given a dynamical system on a manifold with first integrals , it is important for a numerical integrator to preserve the manifold structure and the first integrals of the equations of motion .
this has been the focus of much effort in the development of numerical integration schemes @xcite . in this paper
we do not propose any specific numerical integration scheme , but rather propose a new paradigm of integration that can faithfully preserve conserved quantities with existing numerical integration schemes .
the main idea in our paradigm is as follows .
consider a dynamical system on a manifold @xmath0 with first integrals @xmath1 , @xmath2 .
assume that we can embed the manifold @xmath0 into euclidean space @xmath3 and extend the first integrals to a neighborhood @xmath4 of @xmath0 in @xmath3 .
for an arbitrary point @xmath5 , consider the set @xmath6 which is the intersection of @xmath0 with all the level sets of the first integrals containing the point @xmath7 , and is an invariant set of the dynamical system .
we then extend the dynamical system from @xmath0 to @xmath4 and then modify the dynamics outside of @xmath8 such that the set @xmath8 becomes a unique local attractor of the extended , modified system .
since the dynamics have not changed on @xmath8 by the extension and modification to @xmath4 , both the original system on @xmath0 and the extended , modified system on @xmath4 produce the same trajectory for the initial point @xmath9 .
numerically , however , integrating the extended system has the following advantage : if the trajectory deviates from @xmath8 at some numerical integration step , then it will get pushed back toward the attractor @xmath8 in the extended , modified dynamics , thus remaining on the manifold @xmath0 and preserving all the first integrals .
it can be rigorously shown that the discrete - time dynamical system derived from any one - step numerical integrator with uniform step size @xmath10 for the extended , modified continuous - time system indeed has an attractor @xmath11 that contains the set @xmath8 in its interior and converges to @xmath8 as @xmath12 . in this paper
we shall use the word , _ preserve _ , in this sense .
it is noteworthy that the numerical integration of the extended dynamics can be carried out with any ordinary integrator and is done in one global cartesian coordinate system on @xmath3 .
we find conditions for applicability of this method and implement the result on the following three examples : the free rigid body dynamics , the kepler problem , and a perturbed kepler problem with rotational symmetry .
we also carry out simulation studies to show the excellence of our new paradigm of integration for numerical preservation of conserved quantities in comparison with other well - known integration schemes , such as projection and splitting methods and symplectic strmer - verlet integrators .
consider a dynamical system on an open subset @xmath4 of @xmath3 : @xmath13 where @xmath14 is a @xmath15 vector field on @xmath4 .
let us make the following assumptions : * there is a @xmath16 function @xmath17 such that @xmath18 for all @xmath19 , @xmath20 , and @xmath21 for all @xmath22 .
* there is a positive number @xmath23 such that @xmath24)$ ] is a compact subset of @xmath4 . *
the set of all critical points of @xmath25 in @xmath24)$ ] is equal to @xmath26 . adding the negative gradient of @xmath25 to ( [ our : dyn : sys ] ) ,
let us consider the following dynamical system on @xmath4 : @xmath27 since @xmath28 is the minimum value of @xmath25 , @xmath29 for all @xmath30 .
hence , the two vector fields @xmath14 and @xmath31 coincide on @xmath26 . [ theorem : general ] under assumptions a1 a3 , every trajectory of ( [ our : new : dyn : sys ] ) starting from a point in @xmath24)$ ] stays in @xmath24)$ ] for all @xmath32 and asymptotically converges to the set @xmath26 as @xmath33 .
furthermore , @xmath26 is an invariant set of both ( [ our : dyn : sys ] ) and ( [ our : new : dyn : sys ] ) .
let @xmath34 be a trajectory of ( [ our : new : dyn : sys ] ) starting from a point in @xmath24)$ ] .
by a1 @xmath35 for all @xmath36 .
hence , @xmath24)$ ] is a positively invariant set of ( [ our : new : dyn : sys ] ) .
from ( [ v : dot ] ) and a3 , it follows that @xmath37 ) \mid \dot v ( x ) = 0\ } = \{x \in v^{-1}([0,c ] ) \mid \nabla v(x ) = 0\}= v^{-1}(0)$ ] . hence , by lasalle s invariance principle @xcite , @xmath34 converges asymptotically to @xmath26 as @xmath33 , where a2 is used for compactness of @xmath24)$ ] .
the invariance of @xmath26 follows from ( [ vdot : x ] ) and the coincidence of ( [ our : dyn : sys ] ) and ( [ our : new : dyn : sys ] ) on @xmath26 .
let us find a higher - order condition than that in assumption a3 so that a3 can be relaxed .
for the function @xmath25 and the vector field @xmath14 in the statement of assumption a1 , which are now both assumed to be of @xmath38 , let @xmath39 where @xmath40 , and @xmath41 denotes the @xmath42th order directional derivative of @xmath43 along @xmath14 , i.e. , @xmath44 consider the following assumption in place of a3 : * @xmath45 ) \subset v^{-1}(0)$ ] .
the following theorem generalizes theorem [ theorem : general ] : [ theorem : general : more ] under assumptions a1 , a2 and a3@xmath46 , every trajectory of ( [ our : new : dyn : sys ] ) starting in @xmath24)$ ] stays in @xmath24)$ ] for all @xmath32 and asymptotically converges to the set @xmath26 as @xmath33 .
furthermore , @xmath26 is an invariant set of both ( [ our : dyn : sys ] ) and ( [ our : new : dyn : sys ] ) .
consider the dynamics ( [ our : new : dyn : sys ] ) .
it is easy to show that @xmath24)$ ] is a positively invariant set of the dynamics .
let @xmath47 be the largest invariant set in @xmath48)$ ] .
let @xmath34 be an arbitrary trajectory in @xmath47 .
since @xmath49)$ ] as shown in the proof of theorem [ theorem : general ] , the trajectory @xmath34 satisfies @xmath50 , i.e. , @xmath51 for all @xmath52 and @xmath53 . since @xmath50 along @xmath34 , the trajectory @xmath34 satisfies @xmath54 for all @xmath52 . by differentiating ( [ dvdxi ] )
repeatedly in @xmath55 and using ( [ reduced : x ] ) on each differentiation , we can show that the trajectory @xmath34 satisfies @xmath56 for all @xmath57 , @xmath58 and @xmath53 .
thus , the entire trajectory @xmath34 is contained in the set @xmath59 defined in ( [ def : set : s ] ) , implying @xmath60 , from which and a3@xmath46 it follows @xmath61 .
hence , by lasalle s invariance principle , every trajectory starting in @xmath24)$ ] asymptotically converges to @xmath47 and thus to @xmath26 as @xmath33 .
the invariance of @xmath26 follows from ( [ vdot : x ] ) and the coincidence of ( [ our : dyn : sys ] ) and ( [ our : new : dyn : sys ] ) on @xmath26 .
if condition ( [ vdot : x ] ) is replaced by @xmath62 in assumption a1 , then theorems [ theorem : general ] and [ theorem : general : more ] still hold provided that the invariance of @xmath26 is replaced by positive invariance in the statement of the theorems .
2.theorems [ theorem : general ] and [ theorem : general : more ] still hold with the use of the following modified dynamics @xmath63 instead of ( [ our : new : dyn : sys ] ) , where @xmath64 is an @xmath65 matrix - valued function with its symmetric part @xmath66 positive definite at each @xmath67 .
\3 . from the control viewpoint , the added term @xmath68 in ( [ our : new : dyn : sys ] ) can be regarded as a negative feedback control @xmath69 to asymptotically stabilize the set @xmath26 for the control system @xmath70 with control @xmath71 .
suppose that assumptions a1 , a2 and a3 ( or a3@xmath46 instead of a3 ) hold and that we want to integrate the dynamics ( [ our : dyn : sys ] ) for an initial point @xmath72 .
since @xmath26 is positively invariant , the trajectory must remain in @xmath26 for all @xmath32 .
recall that the two dynamics ( [ our : dyn : sys ] ) and ( [ our : new : dyn : sys ] ) coincide on @xmath26 , so we can integrate ( [ our : new : dyn : sys ] ) instead of ( [ our : dyn : sys ] ) for the initial condition . though there is no theoretical difference between the two integrations , integrating ( [ our : new : dyn : sys ] ) has a numerical advantage over integrating ( [ our : dyn : sys ] ) .
suppose that the trajectory numerically deviates from the positively invariant set @xmath26 during integration .
then the dynamics ( [ our : new : dyn : sys ] ) will push the trajectory back toward @xmath26 since @xmath26 is the attractor of ( [ our : new : dyn : sys ] ) in @xmath24)$ ] whereas the dynamics ( [ our : dyn : sys ] ) will leave the trajectory outside of @xmath26 which would not happen in the exact solution . it is noteworthy that this integration strategy is independent of the choice of integration schemes . in the appendix we show that any one - step numerical integrator , as a discrete - time dynamical system , with uniform step size @xmath10 for ( [ our : new : dyn : sys ] ) has an attractor @xmath11 that contains @xmath26 in its interior and converges to @xmath26 as @xmath73 .
let us now apply this integration strategy to numerically integrate dynamics on a manifold while preserving its first integrals and the domain manifold .
consider a manifold @xmath0 and dynamics @xmath74 on @xmath0 that have @xmath75 first integrals @xmath76 , @xmath2 .
suppose that @xmath0 is an embedded manifold in @xmath3 as a level set of a function @xmath77 for some @xmath78 , and that both the dynamics ( [ dyn : on : m ] ) and the functions @xmath79 , @xmath80 extend to an open neighborhood @xmath4 of @xmath0 in @xmath3 .
our goal is to numerically integrate ( [ dyn : on : m ] ) with an initial condition @xmath81 while preserving the manifold @xmath0 and the first integrals .
let @xmath82 and define a function @xmath83 by @xmath84 where @xmath85 is an @xmath86 constant symmetric positive definite matrix .
notice that @xmath87 and that @xmath26 is invariant under the flow of ( [ dyn : on : m ] ) . or , more generally we can define a function @xmath88 as @xmath89 where @xmath90 is a non - negative function that takes the value of @xmath28 only at @xmath91 . if the function @xmath25 satisfies assumptions a1 , a2 and a3 ( or a3@xmath46 instead of a3 ) , then by theorem [ theorem : general ] ( or theorem [ theorem : general : more ] ) , @xmath26 is the local attractor of the modified dynamics @xmath92 that coincide with the original dynamics ( [ dyn : on :
m ] ) on @xmath26 .
the following lemma provides a sufficient condition under which the function @xmath25 defined in ( [ def : v : intro ] ) satisfies assumptions a2 and a3 : [ kuhl : lemma ] consider the functions @xmath93 and @xmath25 defined in ( [ def : f ] ) and ( [ def : v : intro ] ) .
if @xmath26 is compact and the jacobian matrix @xmath94 of @xmath93 has rank @xmath95 for all @xmath96 , then there is a number @xmath97 such that assumptions a2 and a3 hold . by compactness of @xmath26 and the regularity of @xmath98
, there is a bounded open set @xmath14 such that @xmath99 , and @xmath98(x ) has rank @xmath100 for all @xmath101 , where @xmath102 denotes the closure of @xmath14 .
consider now the gradient of @xmath25 .
an easy calculation shows that , @xmath103 now , since for all @xmath104 , @xmath94 is onto as a linear map , @xmath105 is therefore one to one .
it follows that , for @xmath104 , @xmath106 in other words , the set of all critical points of @xmath25 in @xmath14 is equal to @xmath26 . since the boundary @xmath107 of @xmath14 , being closed and bounded ,
is compact and @xmath108 , the minimum value , denoted by @xmath109 , of @xmath25 on @xmath107 is positive .
if necessary , restrict the function @xmath25 to @xmath14 , replacing its original domain @xmath4 with @xmath14 .
then , there is a positive number @xmath23 less than @xmath109 such that @xmath110 ) \subset x$ ] .
therefore , assumption a3 holds for this number @xmath23 . since the closed set @xmath24)$ ] is contained in the bounded set @xmath14 , it is compact , which implies that assumption a2 holds .
[ kuhl : theorem ] for the functions @xmath93 and @xmath25 defined in ( [ def : f ] ) and ( [ def : v : intro ] ) , if @xmath25 satisfies ( [ vdot : x ] ) for all @xmath19 , the set @xmath26 is compact and the jacobian matrix @xmath94 is onto for all @xmath96 , then there is a number @xmath97 such that every trajectory starting in @xmath24)$ ] remains in @xmath24)$ ] for all @xmath32 and asymptotically converges to @xmath26 as @xmath111 . for the functions @xmath93 and
@xmath25 defined in ( [ def : f ] ) and ( [ def : v : intro ] ) , if @xmath25 satisfies ( [ vdot : x ] ) for all @xmath19 , the set @xmath26 is compact and there is an open subset @xmath14 of @xmath4 containing @xmath26 such that the jacobian matrix @xmath94 is onto for all @xmath112 , then there is a number @xmath97 such that every trajectory starting in @xmath24)$ ] remains in @xmath24)$ ] for all @xmath32 and asymptotically converges to @xmath26 as @xmath111 .
modify the proof of lemma [ kuhl : lemma ] appropriately . as discussed above
, we can integrate ( [ dyn : on : m : ext ] ) instead of ( [ dyn : on : m ] ) for the initial condition @xmath113 , which will yield a trajectory that is expected to numerically well remain on the manifold @xmath0 and preserve the values of the first integrals @xmath79 , @xmath2 .
it is noteworthy that the integration is carried out in one cartesian coordinate system on @xmath3 rather than over local charts on the manifold @xmath0 which would take additional computational costs for coordinate changes between local charts . in the following section
, we will apply this strategy to the free rigid body dynamics , the kepler problem and a perturbed kepler problem with rotational symmetry to integrate the dynamics preserving their first integrals and domain manifolds .
consider the free rigid body dynamics : [ rigidbody ] @xmath114 where @xmath115 ; @xmath116 is the moment of inertia matrix ; and @xmath117 for @xmath118 since @xmath119 , from here on we assume that the rigid body dynamics are defined on the euclidean space @xmath120 and that the matrix @xmath121 denotes a @xmath122 matrix , not necessarily in @xmath123 .
this is the _ extension _ of the dynamics step .
define two functions @xmath124 and @xmath125 by @xmath126 where @xmath127 represents the kinetic energy of the free rigid body and @xmath128 the spatial angular momentum vector when @xmath129 .
these quantities are first integrals of ( [ rigidbody ] )
. choose any @xmath130 and let @xmath131 define an open set @xmath4 by @xmath132 and a function @xmath133 by @xmath134 for @xmath135 , where @xmath136 , @xmath137 are constants , and @xmath138 is the 2-norm defined by @xmath139 for a matrix @xmath140 .
observe that we are endowing the space @xmath141 with the standard inner product , and that the trace norm is precisely the norm induced on @xmath142 by this inner product .
we compute all gradients that follow with respect to this inner product .
notice that @xmath143 [ lemma : gradient ] the gradient @xmath144 of the function @xmath25 ( [ def : v : rigid ] ) is given by [ gradequations ] @xmath145 straightforward .
the following lemma shows that the function @xmath25 satisfies assumption a1 stated in
[ section : theory ] .
[ lemma:1:rigid ] the function @xmath25 satisfies @xmath146 one can compute @xmath147 where , in the third equality we use the fact that for @xmath140 symmetric and @xmath148 antisymmetric , @xmath149
. next , we compute , @xmath150 hence , @xmath151 the following lemma shows that the function @xmath25 satisfies assumptions a2 and a3 stated in
[ section : theory ] .
[ lemma:2:rigid ] there is a number @xmath23 satisfying @xmath152 such that @xmath24)$ ] is a compact subset of @xmath4 and the set of all critical points of @xmath25 in @xmath24)$ ] is equal to @xmath26 .
it is obvious that there is a number @xmath23 satisfying ( [ inequality : c ] ) such that @xmath24)$ ] becomes a compact set in @xmath4 .
for such a number @xmath23 , the matrix @xmath121 is invertible for every @xmath153)$ ] .
since @xmath28 is the minimum value of @xmath25 , every point in @xmath26 is a critical point of @xmath25 .
let @xmath154 be a critical point of @xmath25 in @xmath24 ) \backslash v^{-1}(0)$ ] . by lemma [ lemma : gradient ]
it satisfies [ critical ] @xmath155 where @xmath156 post - multiplying ( [ critical:1a ] ) by @xmath157 and pre - multiplying ( [ critical:1b ] ) by @xmath121 yield [ critical2 ] @xmath158 since @xmath159 .
notice that @xmath160 would imply @xmath161 , contradicting @xmath153)$ ] .
hence , @xmath162 .
it follows from ( [ critical ] ) that if any of the three equations @xmath163 holds , then the three of them all hold .
thus @xmath164 since @xmath165 .
since the matrix @xmath166 in ( [ critical:1a ] ) has rank 1 and the matrix @xmath167 is symmetric , there exist a unit vector @xmath168 and a number @xmath169 such that @xmath170 substitution of ( [ rtr_i ] ) into ( [ critical:1a ] ) and ( [ critical:2a ] ) yields @xmath171 which implies @xmath172 where the symbol @xmath173 means ` is parallel to . '
hence , we can express @xmath121 and @xmath128 as @xmath174 for some numbers @xmath175 , @xmath176 and vectors @xmath177 , where @xmath178 and the vectors @xmath179 and @xmath180 can be any vectors such that @xmath181 becomes an orthonormal basis for @xmath182
. substitution of ( [ r_split ] ) into ( [ rtr_i ] ) implies that @xmath183 is an orthonormal basis for @xmath182 .
substitution of ( [ r_split ] ) and ( [ pi_b_pi0 ] ) into ( [ critical:1b ] ) implies @xmath184 , which together with @xmath185 in ( [ parallel_rels ] ) , implies @xmath186 , i.e. , @xmath71 is an eigenvector of @xmath116 .
we can now choose or re - define the unit vectors @xmath179 and @xmath180 such that they become eigenvectors of the symmetric matrix @xmath116 , too .
in the orthonormal basis @xmath181 , we can now write the moment of inertia matrix @xmath116 as @xmath187 where @xmath188 are the eigenvalues of @xmath116 , which are all positive , corresponding to the eigenvectors @xmath189 , respectively . it is then easy to see that equations ( [ critical2 ] ) imply [ abequations ] @xmath190 where we have used @xmath191 .
we consider the following two separate cases : @xmath192 and @xmath193 .
suppose @xmath192 .
if @xmath194 , then @xmath195 by ( [ inequality : c ] ) , which contradicts @xmath153)$ ] . hence , @xmath196 . if @xmath197 , then equation ( [ abequations : a ] ) implies @xmath198 , but equation ( [ abequations : b ] ) implies @xmath199 , implying @xmath200 .
this can not be compatible with @xmath197 .
hence , @xmath197 is ruled out .
similarly , @xmath201 can be ruled out .
hence , @xmath202 , which implies @xmath203 contradicting ( [ contra : eqn ] ) .
thus , when @xmath192 , there are no critical points of @xmath25 in @xmath24)\backslash v^{-1}(0)$ ] .
suppose @xmath204 .
we analyze equations ( [ abequations ] ) using a continuity argument . at @xmath205 , ( [ abequations :
a ] ) implies @xmath206 or @xmath207 , neither of which satisfies ( [ abequations : b ] ) at @xmath205 .
thus , by continuity there exists a number @xmath208 with @xmath209 such that for any @xmath210 with @xmath211 there is no number @xmath212 satisfying both ( [ abequations : a ] ) and ( [ abequations : b ] ) .
hence , @xmath213 .
we now shrink the number @xmath23 such that it not only satisfies ( [ inequality : c ] ) but also @xmath214 .
for such a number @xmath23 , we have @xmath215 which contradicts @xmath153)$ ] .
hence , when @xmath216 , there are no critical points of @xmath25 in @xmath24)\backslash v^{-1}(0)$ ] for some @xmath97
. therefore , there exists a number @xmath97 such that @xmath26 is the set of all critical points of @xmath25 in @xmath24)$ ] .
consider the dynamics [ rigid : modified ] @xmath217 which correspond to ( [ our : new : dyn : sys ] ) . from theorem [ theorem : general ] and lemmas
[ lemma:1:rigid ] and [ lemma:2:rigid ] comes the following theorem : there is a number @xmath97 such that every trajectory of ( [ rigid : modified ] ) starting from a point in @xmath24)$ ] stays in @xmath24)$ ] for all @xmath32 and asymptotically converges to the set @xmath218 as @xmath33 , where the function @xmath25 is defined in ( [ def : v : rigid ] ) .
furthermore , @xmath26 is an invariant set of both ( [ rigidbody ] ) and ( [ rigid : modified ] ) .
the two - body dynamics in the kepler problem are given in the usual barycentric coordinates by [ kepler ] @xmath219 where @xmath220 is the position vector , @xmath221 is the velocity vector and @xmath222 is the gravitational parameter .
define two functions @xmath223 and @xmath224 by @xmath225 where @xmath226 is called the angular momentum vector and @xmath140 is called the laplace - runge - lenz vector .
it is known that both @xmath226 and @xmath140 are first integrals of the two - body dynamics ( [ kepler ] ) and they are orthogonal to each other , i.e. , @xmath227 for all @xmath228 .
the energy function @xmath229 satisfies latexmath:[\[\label{relation } @xmath228 , implying that the energy @xmath127 is also a first integral of the two - body dynamics ( [ kepler ] ) .
it is also known that a non - degenerate elliptic keplerian orbit is uniquely determined by a pair @xmath231 that satisfies @xmath232 , @xmath233 and @xmath234 @xcite . fix a non - degenerate elliptic keplerian orbit , i.e. , a pair of vectors @xmath235 that satisfies @xmath236 define a function @xmath237 by @xmath238 for @xmath228 , where @xmath239 and @xmath240 .
notice that @xmath241 which is the non - degenerate keplerian elliptic orbit whose angular momentum vector and laplace - runge - lenz vector are @xmath242 and @xmath243 , respectively .
[ lemma : gradient : kepler ] the gradient @xmath244 of the function @xmath25 defined in ( [ def : v : kepler ] ) is given by @xmath245 where @xmath246 and @xmath247 .
the following lemma shows that the function @xmath25 defined in ( [ def : v : kepler ] ) satisfies assumptions a1 and a2 stated in
[ section : theory ] .
[ shortlemma : kepler ] 1 .
the function @xmath25 satisfies @xmath248 2 . for any number
@xmath23 satisfying @xmath249 the set @xmath250)$ ] is a compact set in @xmath251 .
the first fact is a straightforward calculation using the previous lemma . for the second
, the essential idea is that the fibers of @xmath25 are homeomorphic to circles , corresponding to the elliptic orbits , and are therefore compact .
for a detailed proof of the second statement , refer to corollary 2.2 in @xcite .
the following lemma shows that the function @xmath25 defined in ( [ def : v : kepler ] ) satisfies assumption a3 stated in
[ section : theory ] .
[ longlemma : kepler ] for any number @xmath23 satisfying ( [ range : c : kepler ] ) the set of all critical points of @xmath25 in @xmath24)$ ] is equal to @xmath26 .
choose an arbitrary number @xmath23 satisfying ( [ range : c : kepler ] ) .
let @xmath252 be an arbitrary critical point of @xmath25 in @xmath24)$ ] .
for notational convenience , let us write @xmath253 suppressing the dependence on @xmath252 .
by lemma [ lemma : gradient : kepler ] , the critical point @xmath252 satisfies [ dv : zero ] @xmath254 if @xmath255 , then @xmath256 , contradicting @xmath257)$ ] .
hence , @xmath233 , which together with ( [ def : l ] ) implies that the three vectors @xmath258 form a basis for @xmath182 .
the dot product of ( [ dvv : zero ] ) with @xmath259 yields @xmath260 so there are numbers @xmath210 and @xmath212 such that @xmath261 substitution of ( [ da : xl ] ) into ( [ dv : zero ] ) gives @xmath262
it follows that there are numbers @xmath109 and @xmath93 such that [ dl : vlx ] @xmath263 from ( [ dl : vlx ] ) , we obtain @xmath264 by linear independence of @xmath265 , @xmath266 substitution of these into ( [ da : xl ] ) and ( [ dl : vlx:2 ] ) gives @xmath267 where we have used the definition of @xmath140 given in ( [ def : a ] ) . hence , @xmath268
from ( [ a0:l0:linear : combo ] ) and the orthogonality @xmath269 and @xmath270 , it follows that @xmath271 since @xmath233 , and recalling that @xmath272 and @xmath273 , we have @xmath274 .
substitution of @xmath274 into ( [ a0:l0:linear : combo ] ) yields @xmath275 which implies @xmath276 .
thus , every critical point of @xmath25 in @xmath24)$ ] is contained in @xmath26 . since 0
is the minimum value of @xmath25 , every point in @xmath26 is a critical point of @xmath25 .
therefore , the set of all critical points of @xmath25 in @xmath24)$ ] is @xmath26 .
choose a non - degenerate keplerian elliptic orbit and let @xmath277 be a point on the orbit .
set @xmath278 to be the angular momentum vector and the laplace - runge - lenz vector of the orbit , respectively .
consider the dynamics : [ kepler : modified ] @xmath279 where @xmath246 and @xmath247 , which correspond to ( [ our : new : dyn : sys ] ) . from theorem [ theorem : general ] and lemmas [ shortlemma : kepler ] and [ longlemma : kepler ]
comes the following theorem : for any @xmath97 satisfying ( [ range : c : kepler ] ) , every trajectory of ( [ kepler : modified ] ) starting from a point in @xmath24)$ ] stays in @xmath24)$ ] for all @xmath32 and asymptotically converges to the set @xmath280 as @xmath33 , where the function @xmath25 is defined in ( [ def : v : kepler ] ) .
furthermore , @xmath26 is an invariant set of both ( [ kepler ] ) and ( [ kepler : modified ] ) .
consider a perturbed kepler problem with rotational symmetry whose equations of motion are given by [ pert : kepler ] @xmath281 where @xmath220 is the position vector , @xmath221 is the velocity vector , and @xmath282 is the potential function that depends only on the radial distance from the origin .
the total energy @xmath283 and the angular momentum vector @xmath284 are defined by @xmath285 and they are conserved quantities of the dynamics ( [ pert : kepler ] ) .
take any point @xmath286 such that @xmath287 let @xmath288 define a function @xmath289 by @xmath290 with @xmath239 and @xmath240 .
then , @xmath291 the gradient @xmath292 of @xmath25 is given by @xmath293 where @xmath294 and @xmath246 .
trivially , @xmath25 satisfies ( [ vdot : x ] ) , i.e. @xmath295 for all @xmath296 .
the modified dynamics , which correspond to ( [ our : new : dyn : sys ] ) , are computed as [ pert : kepler : mod ] @xmath297 [ theorem : main : pert : kep ] suppose that @xmath26 is compact and there is no common solution @xmath298 to the following two equations : latexmath:[\[\begin{aligned } e_0 & = \frac{1}{2}ru^\prime(r ) + u(r ) , \label{e0:cond : pert}\\ assumptions a2 and a3 hold and there is a number @xmath97 such that every trajectory of ( [ pert : kepler : mod ] ) starting in @xmath24)$ ] remains in @xmath24)$ ] for all @xmath32 and asymptotically converges to @xmath26 as @xmath111 .
define a function @xmath300 by @xmath301 then , @xmath302 where the over - hat symbol @xmath303 denotes the hat map defined in ( [ def : hat : map ] ) .
we want to show that the @xmath304 matrix @xmath305 is one - to - one for all @xmath276 .
fix an arbitrary point @xmath276 .
it follows @xmath306 take any point @xmath307 from the kernel of @xmath305 .
then , [ dft : ker ] @xmath308 suppose @xmath309 .
taking the inner product of ( [ dft : ker : a ] ) with @xmath259 and of ( [ dft : ker : b ] ) with @xmath310 , we obtain @xmath311 from which it follows that latexmath:[\[\label{uprime : v2 } with @xmath259 , we get @xmath313 which implies @xmath314 that satisfies both ( [ e0:eqn : up ] ) and ( [ l0:norm : eqn ] ) .
hence , we can not have @xmath309 .
substitute @xmath315 into ( [ dft : ker ] ) .
it follows that @xmath316 is parallel to @xmath317 .
hence , there is a number @xmath212 such that @xmath318 . substituting this in ( [ dft : ker : b ] )
yields @xmath319 .
taking the cross product of this with @xmath259 yields @xmath320 since @xmath321 . since @xmath322 and @xmath323
, we have @xmath274 , so @xmath324 .
it follows that @xmath325 , which implies that @xmath305 is one - to - one for all @xmath276 . in other words
, @xmath326 is onto for all @xmath276 .
hence , the conclusion of the theorem follows from lemma [ kuhl : lemma ] , equation ( [ pert : vx0 ] ) , and theorem [ kuhl : theorem ] .
consider a special case in which the potential function @xmath327 is of the form @xmath328 where @xmath329 and @xmath330 .
then equations ( [ e0:cond : pert ] ) and ( [ l0:cond : pert ] ) become @xmath331 and @xmath242 , it is then easy to check if there is no common solution @xmath298 to ( [ e0:birk ] ) and ( [ l0:birk ] ) .
consider the free rigid body dynamics in [ subsection : rigid : body ] with the moment of inertia matrix @xmath332 and the initial condition @xmath333 the values of the energy @xmath127 and the spatial angular momentum vector @xmath334 corresponding to the initial condition are @xmath335 the period @xmath336 of the trajectory of the body angular velocity vector @xmath337 is computed approximately to be @xmath338 .
we integrate the dynamics over the time interval @xmath339 = [ 0 , 155.7t_\omega]$ ] with step size @xmath340 , using the following four integration methods : a feedback integrator with the euler scheme , a projection method with the euler scheme , a splitting method with three rotations splitting , and the ordinary euler method .
the feedback integrator with the euler scheme denotes the euler method applied to the modified free rigid dynamics ( [ rigid : modified ] ) with the following values of the parameters @xmath341 , @xmath342 , and @xmath343 @xmath344 the projection method is the standard one explained on pp.110111 in @xcite . in order to solve constraint equations for projection at each step of integration in the projection method
, we use the matlab command _ fsolve _ with the parameter _ tolfun _ , which is termination tolerance on the function value , set equal to @xmath345 , which is the same as the integration step size @xmath346 .
the splitting method is the one explained on pp.284285 in @xcite .
the three of the projection method , the splitting method and the ordinary euler method are applied to the original free rigid body dynamics ( [ rigidbody ] ) .
the trajectories of the body angular velocity vector @xmath347 , the energy error @xmath348 , the error @xmath349 in spatial angular momentum , and the deviation @xmath350 of the rotation matrix @xmath351 from @xmath123 are plotted in figures [ figure : rigid : omega ] , [ figure : rigid : energy ] , [ figure : rigid : momentum ] and [ figure : rigid : so3 ] , respectively . in figure
[ figure : rigid : omega ] , it is observed that the trajectories of @xmath347 generated by the feedback integrator and the projection method maintain a periodic shape well whereas those by the splitting method and the euler method drift away significantly from the periodic shape . in figure
[ figure : rigid : energy ] , it is observed that the feedback integrator and the projection method keep the energy error sufficiently small whereas the energy errors by the other two methods increase in time . although the two trajectories of energy error by the splitting method and the euler method seem to coincide in figure [ figure : rigid : energy ] , an examination of the numerical data shows that the energy of the euler method gets larger than that of the splitting method in time .
for example , at @xmath352 , the energy of the euler method is bigger than that of the splitting method by @xmath353 . in figures [ figure : rigid : momentum ] and [ figure : rigid : so3 ] , it is observed that the feedback method preserves the spatial angular momentum vector and the manifold @xmath123 sufficiently well . in terms of computation time ,
the projection method takes much more time than the others , which is due to the steps of solving the constraint equations for projection .
the splitting method is symplectic and of order 2 whereas the other methods are of order 1 .
all of these observations lead us to the conclusion that the feedback integrator overall has produced the best outcome in the simulation of the free rigid body dynamics .
consider the kepler problem in
[ subsection : two : body ] with @xmath354 and the initial condition @xmath355 the corresponding initial values of the angular momentum vector and the laplace - runge - lenz vector are @xmath356 the period @xmath357 and the eccentricity @xmath358 of the kepler orbit containing the initial point are @xmath359 we integrate the kepler dynamics over the time interval @xmath360 $ ] with step size @xmath361 , using the following four integration methods : a feedback integrator with the euler scheme , the standard projection method with the euler scheme , and two strmer - verlet schemes .
the feedback integrator with the euler scheme denotes the euler method applied to ( [ kepler : modified ] ) with @xmath362 and @xmath363 .
the standard projection method is explained on pp.110111 in @xcite . to solve the constraint equations for projection
, we use the matlab command _ fsolve _ with the parameter _ tolfun
_ set equal to @xmath364 , which is the same as the integration step size @xmath346 .
the two strmer - verlet schemes are those in ( 3.4 ) and ( 3.5 ) on pp .
189190 in @xcite , and we call them strmer - verlet - a and strmer - verlet - b , respectively , for convenience .
the strmer - verlet schemes are symplectic methods of order 2 .
the trajectories of the planar orbit @xmath365 , the error of the laplace - runge - lenz vector , @xmath366 , and the error of the angular momentum vector , @xmath367 , are plotted in figures [ figure : kepellipse ] , [ figure : kepa ] and [ figure : kepl ] . in figure
[ figure : kepellipse ] it is observed that the planar trajectories @xmath365 generated by the feedback integrator and the projection method maintain the elliptic shape well whereas those by the strmer - verlet schemes precess .
this can be also verified in figure [ figure : kepa ] , where the feedback integrator and the projection method preserve the laplace - runge - lenz vector well , but the strmer - verlet schemes cause the laplace - runge - lenz vector to noticeably precess . in figure
[ figure : kepl ] , it is observed that the strmer - verlet schemes preserve the angular momentum vector exceptionally well in comparison with the other two methods . in figures
[ figure : kepa ] and [ figure : kepl ] , we can see that the precision of the feedback integrator is comparable with that of the projection method .
however , the feedback integrator takes much less computation time than the projection method .
the feedback integrator and the projection method used here are of order 1 , whereas the strmer - verlet schemes are of order 2 .
all of these observations lead us to conclude that the feedback integrator has produced the best result overall .
consider the perturbed kepler problem in
[ subsection : pert : kepler ] with the potential function @xmath4 given in ( [ pert : kep : pot ] ) with @xmath354 and @xmath368 , which is the one used in example 4.3 on p. 111 in @xcite .
we use the initial conditions @xmath369 with eccentricity @xmath370 as in @xcite .
the corresponding values of the energy and the angular momentum vector are @xmath371 we integrate the perturbed kepler dynamics over the time interval @xmath372 $ ] with step size @xmath373 , just as on p. 111 in @xcite , using the following four integration methods : a feedback integrator with the euler scheme , the standard projection method with the euler scheme , the strmer - verlet scheme in ( 3.4 ) on p. 189 in @xcite , and the matlab command , _
ode45_. the feedback integrator with the euler scheme denotes the euler method applied to ( [ pert : kepler : mod ] ) with @xmath374 and @xmath375 , and it is straightforward to verify that the hypotheses in theorem [ theorem : main : pert : kep ] hold true .
the other three methods are applied to ( [ pert : kepler ] ) . the matlab command _ fsolve _
is used in the projection method with the parameter _ tolfun _ set equal to @xmath376 .
the options of _ reltol _ = _ abstol _
= @xmath377 are used for the matlab integrator , _ ode45 _ , so the result generated by _
ode45 _ can be used as a reference .
the trajectories of the planar orbit @xmath365 , the energy error @xmath378 and the error @xmath379 in angular momentum are plotted in figures [ figure : pertkepellipse ] , [ figure : pertkepenergy ] and [ figure : pertkepl ] . in figure
[ figure : pertkepellipse ] it is observed that the orbits generated by the feedback integrator and the strmer - verlet scheme are similar to that by _
ode45 _ , but the orbit by the projection method precesses too much which is a very poor result .
the projection method excels only at preserving the energy and the angular momentum as expected in view of the nature of the projection method and the small tolerance parameter value , _ tolfun _
= @xmath376 , used for the matlab command , _
fsolve_. in figure [ figure : pertkepenergy ] , it is observed that the feedback integrator is comparable with the strmer - verlet scheme in energy conservation .
the feedback integrator also preserves the angular momentum well in view of the step size @xmath373 , as can be seen in figure [ figure : pertkepl ] .
the feedback integrator and the projection method used here are of order 1 whereas the strmer - verlet scheme is of order 2 . from all of these observations
, we conclude that the feedback integrator has produced the best result overall .
we have developed a theory to produce numerical trajectories of a dynamical system on a manifold that stably remain on the manifold and preserve first integrals of the system .
our theory is not a numerical integration scheme but rather a modification of the original dynamics by feedback .
the actual numerical integration in our framework can be done with any usual integrator such as euler and runge - kutta .
our method is successfully applied to the free rigid body , the kepler problem and a perturbed kepler problem with rotational symmetry , and its excellent performance is demonstrated by simulation studies in comparison with the standard projection method , two strmer - verlet schemes and a splitting method via three rotations splitting . as future work
, we plan to apply our theory to various mechanical systems with symmetry and non - holonomic systems .
we also plan to carry out a quantitative study of the effect of the parameters in the lyapunov function on the performance of our method .
we show , using results in @xcite , that any discrete - time dynamical system derived from a one - step numerical integration scheme with uniform step size @xmath10 for the modified dynamical system ( [ our : new : dyn : sys ] ) has an attractor @xmath11 that contains @xmath26 in its interior and converges to @xmath26 as @xmath73 .
let us first review some definitions from @xcite .
let @xmath140 and @xmath148 be nonempty , compact subsets of @xmath3 and @xmath259 a point in @xmath3 .
the distance between @xmath259 and @xmath140 is defined by @xmath380 the hausdorff separation of @xmath140 from @xmath148 is defined by @xmath381 the hausdorff distance between @xmath140 and @xmath148 is defined by @xmath382 the hausdorff distance is a metric on the space of nonempty compact subsets of @xmath3 . for @xmath298 ,
let @xmath383 denote an @xmath78-neighborhood of @xmath140 .
we say that a nonempty , compact subset @xmath8 of @xmath3 is uniformly stable for an autonomous dynamical system if for each @xmath384 there exists a @xmath385 such that @xmath386 \rightarrow x(t;x_0 ) \in s(\lambda , \epsilon),\ ] ] where @xmath387 is the solution of the given dynamical system with initial condition @xmath388 .
a set @xmath8 is said to be positively invariant for an autonomous dynamical system if @xmath389 for all @xmath9 and @xmath390 .
a nonempty , compact subset @xmath8 of @xmath3 is called uniformly asymptotically stable for an autonomous dynamical system if it is positively invariant and uniformly stable for the dynamical system , and additionally satisfies the following property : there is a @xmath391 and for each @xmath392 a time @xmath393 such that @xmath394 \rightarrow x(t;x_0 ) \in s(\lambda , \epsilon).\ ] ] [ lemma : uas : lambda ] suppose that assumptions a1 , a2 and a3 ( or a3@xmath46 instead of a3 ) stated in [ section : theory ] hold true .
then , the set @xmath26 is uniformly asymptotically stable for the modified dynamical system ( [ our : new : dyn : sys ] ) . since the three assumptions are satisfied , the conclusions of theorem [ theorem : general ] ( or , [ theorem : general : more ] )
hold true . for convenience ,
let @xmath395 , which is invariant under ( [ our : new : dyn : sys ] ) by theorem [ theorem : general ] ( or , [ theorem : general : more ] ) .
let @xmath97 be the number @xmath23 in assumption a2 . using compactness of @xmath24)$ ] and continuity of @xmath25 , it is easy to show that for any @xmath392 there is a @xmath396 such that @xmath397 ) \subset s(\lambda , \epsilon)$ ] .
it is also easy to show that for any @xmath196 there is an @xmath398 such that @xmath399)$ ] .
hence , we can use the family of sets @xmath400 ) , b>0\}$ ] instead of the family of open sets @xmath401 to show uniform stability and uniform asymptotic stability of @xmath8 for ( [ our : new : dyn : sys ] ) .
let us first show uniform stability of @xmath8 for ( [ our : new : dyn : sys ] ) . given any @xmath392 , take any @xmath208 such that @xmath402 .
then , for any @xmath403)$ ] , @xmath404 ) \subset v^{-1}([0,\epsilon])$ ] for all @xmath390 since @xmath25 is decreasing along the trajectory of @xmath387 of ( [ our : new : dyn : sys ] ) . hence , @xmath8 is uniformly stable for ( [ our : new : dyn : sys ] ) .
let us now show uniform asymptotic stability of @xmath8 for ( [ our : new : dyn : sys ] ) .
take any @xmath405 such that @xmath406 . by continuous dependence of @xmath387 on initial point @xmath7 , compactness of @xmath407)$ ] ,
continuity of the function @xmath25 , and the property that @xmath408 decreases to 0 as @xmath111 for any @xmath409)$ ] , it is easy to show that for any @xmath392 there is a time @xmath393 such that for any @xmath410)$ ] we have @xmath411)$ ] for all @xmath412 .
hence , @xmath8 is uniformly asymptotically stable for ( [ our : new : dyn : sys ] ) .
suppose the vector field @xmath14 is @xmath413 and the function @xmath25 is @xmath414 in the modified dynamical system ( [ our : new : dyn : sys ] ) .
consider a discrete analogue of ( [ our : new : dyn : sys ] ) described by any one - step numerical method of @xmath415th order @xmath416 with uniform step size @xmath417 , where @xmath418 for each @xmath10 .
suppose that the vector field @xmath14 is @xmath413 and the function @xmath25 is @xmath414 , and that assumptions a1 , a2 and a3 ( or a3@xmath46 instead of a3 ) are satisfied .
then there is a number @xmath419 such that for each @xmath420 the discrete - time dynamical system ( [ discrete : system ] ) has a compact , uniformly asymptotically stable set @xmath11 which contains @xmath26 in its interior and converges to @xmath26 with respect to the hausdorff metric as @xmath73 .
moreover , there is a bounded , open set @xmath421 , which is independent of @xmath10 and contains @xmath11 , and a time @xmath422 where @xmath140 and @xmath148 are constants depending on the stability characteristic of @xmath26 , such that the iterates of ( [ discrete : system ] ) satisfy @xmath423 for all @xmath424 , @xmath425 and @xmath426 .
we have only to show that the hypotheses in theorem 1.1 of @xcite hold . since @xmath14 is @xmath413 and @xmath25 is @xmath414 , the vector field @xmath427 of ( [ our : new : dyn : sys ] ) and its derivatives of order up to @xmath415 are all continuous and bounded on the compact set @xmath24)$ ] .
the set @xmath26 is uniformly asymptotically stable for ( [ our : new : dyn : sys ] ) by lemma [ lemma : uas : lambda ] in the above .
therefore , the conclusions of this theorem follow from theorem 1.1 and lemma 3.3 of @xcite .
refer to @xcite to see how to obtain the set @xmath421 and values of the parameters @xmath428 , @xmath140 and @xmath148 that appear in the statement of the above theorem .
the above theorem extends to multi - step numerical integrators ; refer to @xcite for detail .
this research was supported in part by dgist research and development program ( cps global center ) funded by the ministry of science , ict & future planning , global research laboratory program ( 2013k1a1a2a02078326 ) through nrf , and institute for information & communications technology promotion ( iitp ) grant funded by the korean government ( msip ) ( no .
b0101 - 15 - 0557 , resilient cyber - physical systems research ) . | a new method is proposed to numerically integrate a dynamical system on a manifold such that the trajectory stably remains on the manifold and preserves first integrals of the system .
the idea is that given an initial point in the manifold we extend the dynamics from the manifold to its ambient euclidean space and then modify the dynamics outside the intersection of the manifold and the level sets of the first integrals containing the initial point such that the intersection becomes a unique local attractor of the resultant dynamics .
while the modified dynamics theoretically produces the same trajectory as the original dynamics , it yields a numerical trajectory that stably remains on the manifold and preserves the first integrals .
the big merit of our method is that the modified dynamics can be integrated with any ordinary numerical integrator such as euler or runge - kutta .
we illustrate this method by applying it to three famous problems : the free rigid body , the kepler problem and a perturbed kepler problem with rotational symmetry .
we also carry out simulation studies to demonstrate the excellence of our method and make comparisons with the standard projection method , a splitting method and strmer - verlet schemes . |
here we briefly discuss the lifetime of the rydberg dressed ground state atoms @xcite .
lifetimes in the rydberg states @xmath110 and @xmath111 are 252 @xmath112s and 410 @xmath112s @xcite , correspondingly .
the lifetime in the dressed state @xmath0 is amplified by a factor of @xmath113 due to the weak mixing with the rydberg states . with the parameters in the caption of fig . 1 in the main text , we find the effective lifetime is about @xmath114 second , which is far longer than other time scales in the hamiltonian .
in eq.(4 ) in the main text we studied the transition between a gas of @xmath54-mers and @xmath55-mers assuming the effective filling fraction @xmath115 for both gases . here we want to justify analytically that assumption .
as it stands eq.(4 ) does not constrain the effective filling fractions of the two gases , which can in principle be different .
the crucial constraint comes from the requirement that at the transition point both the filling fractions of the gas of @xmath54- and @xmath55-mers become unstable . for the gas of @xmath54-mers ,
the boundaries of the stability region of a certain filling fraction can be determined from the energetics : we call @xmath116 the points at which a gas with one more / less @xmath54-mer becomes energetically favorable .
an analytic expression for @xmath117 was found in @xcite and it reads @xmath118 \label{eq : v1- } \\
w^{+ } & = & -\frac{1}{r-1 } \left [ e^{\rm b , int}_r + \sum_{l=1}^\infty ( r_l+1 ) e^{\rm int}_r(r_l ) - r_l e^{\rm int}_r(r_l+1 ) \right ] , \label{eq : v1+}\end{aligned}\ ] ] where @xmath119 is the repulsive interaction energy between the constituents of a single @xmath54-mer and @xmath120 is the distance between first spins of @xmath121-th nearest @xmath54-mers .
the expressions in eqs.([eq : v1-],[eq : v1 + ] ) are derived under the assumption that the interaction between the @xmath54-mers is repulsive , i.e. in our case it breaks down when the configuration of the gas can allow two @xmath54-mers sitting next to each other , which happens for high filling fractions .
we need then to restrict to the case @xmath122 as for @xmath123 the distances @xmath124 in eq.([eq : v1- ] ) would correspond to adjacent @xmath54-mers .
we plot the devil s staircase for different @xmath54-mers in fig .
[ fig : rmers ] . -mers .
the black , red , green and blue staircases correspond to the gas of dimers , trimers , pentamers and octamers respectively .
the orange crosses indicate the transition points between different polymers ( e.g. the transition point between dimers and trimers is represented by the orange cross at filling fraction 0.5 ) .
as @xmath54 is increasing the transition value @xmath37 quickly approaches the asymptotic value @xmath125 , which is indicated by the orange dashed line .
the parts of the staircase for higher filling fractions ( after the large plateaus as @xmath37 is decreased ) are not shown as they are not captured by our simple theory ( see text for details ) . ] with the stability region at hand one can find the transition points between different kinds of @xmath54-mers as follows : firstly we find the regions in @xmath37 over which two different @xmath54-mers are stable , and extract the effective filling fractions of the two gases .
second we plug the extracted filling fractions in eq.(4 ) to find the transition point .
this point corresponds to a real transition only if it falls in the overlap of the region of stability of the two gases . to illustrate this with an example ,
let us consider e.g. a filling fraction @xmath126 which admits both dimers ( see fig .
[ fig : rmers ] , black line , @xmath127 ) and trimers ( red line , @xmath128 ) .
the condition eq.(4 ) then yields @xmath129 , which is inconsistent with the given stability region @xmath130 $ ] of trimers . in this way
one can show that as @xmath37 is decreased , the chain is filled with dimers only down to @xmath131 , where the first transition to trimers occurs ( orange cross at @xmath126 in fig .
[ fig : rmers ] ) . following this logic
it is then straightforward to show , that all the remaining transitions between @xmath54-mers and @xmath55-mers occur at @xmath127 and quickly approach the asymptotic value @xmath132 .
the values of @xmath37 for some of the transitions are shown in table [ tab : trans points ] and plotted in fig .
2c in the main manuscript .
.values of @xmath37 where a transition between @xmath133-mers occurs , determined from eq.(4 ) with @xmath134 . [ cols="^,^",options="header " , ]
lobe . the left panel shows the transverse magnetization @xmath136 in the vicinity of @xmath135 ( see the caption of fig . 3 in the main text for further details ) .
the right panel shows the idmrg calculations of the density - density correlation function @xmath95 vs. @xmath96 for three different values of @xmath137 at @xmath138 corresponding to the points a , b , c in the left panel . ] in the main text , we have focused on the quantum melting of the @xmath50 lobe , which is the largest lobe between @xmath139 and @xmath61 .
however , the repulsive van der waals tail produces an intricate staircase - like structure of the crystalline ground states ( see fig.1 of the main text ) . to understand the effects of the repulsive tail , as a first approximation
, we could keep only the nearest - neighbour attraction and the next - nearest - neighbour repulsion , and as such the model reduces to the well - known annni ( axial next - nearest - neighbour ising ) model @xcite , for which only three possible ground states exist at @xmath45 , i.e. , the @xmath139 , @xmath50 and @xmath61 phases .
the transition from the phase of @xmath139 to that of @xmath50 happens at @xmath60 as explained in the main text , and the transition from the @xmath50 phase to that of @xmath61 happens at @xmath140 @xcite . now if one introduces the seemingly small third nearest - neighbour repulsion @xmath141 , which is equal to @xmath142 due to the @xmath108 law , two additional new phases appear at filling fractions @xmath135 and @xmath143 .
the filling fraction @xmath135 is stabilized for @xmath144 $ ] which is roughly half of the width of the stability region of @xmath50 ( @xmath145 $ ] ) .
importantly , the filling fraction @xmath135 is stabilized due to the interaction terms going beyond next - to - nearest neighbour . for interaction with the van der waals tails ,
the physics in the sector with filling fraction @xmath146 is dominated by pure dimers and the effective interaction between these dimers is given by the effective potential @xmath147 rather than by the van der walls potential itself . in fig .
[ fig : sf2 ] we show the melting process of the @xmath135 lobe .
the qualitative features are analogous to those for @xmath50 studied in the main text . inside the @xmath135 lobe , the crystalline structure is well preserved as indicated by the periodic behaviour of the @xmath95 correlation function ( point a in fig.[fig : sf2 ] ) .
the ground state wave function at small @xmath17 is given by @xmath148 , from which one can calculate @xmath149 which matches perfectly the numerical results .
experimentally , one can distinguish the different crystalline orders at small @xmath17 by measuring the different periodic behaviours of @xmath95 . at intermediate @xmath17 ,
one observes the quasi long - range order indicated by the algebraic decay of the correlations ( point b in fig.[fig : sf2 ] ) .
eventually the crystal melts into a paramagnetic phase with exponentially decaying correlations ( point c in fig.[fig : sf2 ] ) . | the devil s staircase is a fractal structure that characterizes the ground state of one - dimensional classical lattice gases with long - range repulsive convex interactions .
its plateaus mark regions of stability for specific filling fractions which are controlled by a chemical potential .
typically such staircase has an explicit particle - hole symmetry , i.e. , the staircase at more than half - filling can be trivially extracted from the one at less than half filling by exchanging the roles of holes and particles . here
we introduce a quantum spin chain with competing short - range attractive and long - range repulsive interactions , i.e. a non - convex potential . in the classical limit
the ground state features generalized wigner crystals that depending on the filling fraction are either composed of dimer particles or dimer holes which results in an emergent complete devil s staircase without explicit particle - hole symmetry of the underlying microscopic model . in our system
the particle - hole symmetry is lifted due to the fact that the staircase is controlled through a two - body interaction rather than a one - body chemical potential .
the introduction of quantum fluctuations through a transverse field melts the staircase and ultimately makes the system enter a paramagnetic phase . for intermediate transverse field strengths , however , we identify a region , where the density - density correlations suggest the emergence of quasi long - range order .
we discuss how this physics can be explored with rydberg - dressed atoms held in a lattice .
_ introduction._systems with long - range interactions can feature intriguing physics that is not necessarily present in their short - range counterparts . for example , classical particles in a one - dimensional ( 1d ) lattice interacting via repulsive infinite - range convex potentials lower their interaction energy by assuming a distribution in space as uniform as possible . remarkably this property makes the ground state configuration independent of the actual details of the interactions , e.g. the specific power - law of the interaction potential , and leads to the formation of a so - called generalized wigner crystal @xcite .
furthermore , the permitted filling fractions of the ground state configuration versus the chemical potential form a fractal curve known as the complete devil s staircase @xcite .
recent years have seen a great success in emulating many - body systems with long - range interactions using ultracold gases . here , crystalline structures which were originally studied in the context of solid state physics have shown to be present also in ensembles of cold trapped ions @xcite , polar molecules @xcite and gases of rydberg atoms @xcite . in particular rydberg gases have witnessed a recent experimental breakthrough in which the adiabatic preparation of a crystalline state and the onset of a staircase structure were shown @xcite . a case currently much less studied
is when long - range interactions feature competing attraction and repulsion .
systems with such interactions are known to exhibit intricate behavior and in fact appear in a number of contexts , e.g. , langmuir monolayers @xcite , classical fluids @xcite , mixtures of colloids and polymers @xcite and ferrofluids @xcite . and
@xmath0 are coupled coherently through a raman process . the @xmath0 state is weakly dressed to two rydberg states @xmath1 and @xmath2 by a blue- and red - detuned lasers . (
b ) interaction potential between two atoms in the rydberg dressed @xmath0 state .
varying the laser parameters and the lattice spacing @xmath3 , we achieve a competing interaction where the nn interaction is attractive and the longer range interactions are repulsive .
the blue - detuned laser induces the two - atom resonant excitation at a distance @xmath4 ( dashed line in the figure ) .
( c ) tuning the competing interaction by changing the laser detuning @xmath5 . in the example , two rydberg states @xmath6 and @xmath7 of rubidium atoms are considered .
the dispersion constants of the respective rydberg states are @xmath8 and @xmath9 .
laser parameters are given in the following . in ( b ) , we assume @xmath10 mhz . other parameters used in plotting ( b ) and ( c ) are @xmath11 mhz , @xmath12 mhz , @xmath13 mhz and @xmath14 m ( see text for details ) . ] in this work we explore such scenario in a cold atomic lattice gas in which interatomic interactions are controlled via a so - called double rydberg dressing scheme @xcite .
this permits the realization of competing nearest - neighbor ( nn ) attractive and long - range repulsive interactions which can stabilize generalized wigner crystals and lead to a new mechanism for the formation of a complete devil s staircase . here
the filling fraction is controlled via the nn interaction strength and the emergent devil s staircase is a union of two sub - staircases a dimer particle staircase and a dimer hole staircase despite of the fact that the microscopic description of the system has no explicit particle - hole symmetry .
going beyond the classical limit we eventually perform a qualitative study of the melting of the crystalline states through quantum fluctuations introduced by a transverse field . in our analysis of the density - density correlations
we find signatures of a regime featuring quasi long - range order before a paramagnetic phase is reached . _
the system._ as shown in fig .
[ fig : potential](a ) we consider atoms in two electronic ground states @xmath15 and @xmath0 ( typically zeeman or hyperfine states ) held in a 1d optical lattice .
they form a pseudo - spin @xmath16 particle whose states are ( raman ) coupled with strength @xmath17 .
one blue - detuned and one red - detuned laser are applied simultaneously to weakly dress the @xmath0 state with two rydberg @xmath18-states @xmath1 and @xmath2 , which induces long - range interactions between atoms in the rydberg dressed @xmath0 state @xcite .
the double rydberg dressing in this setting is motivated by recent experimental achievements on rydberg excitations in lattices or microtraps @xcite and laser dressing of ground state atoms to a single rydberg state @xcite . in the following
, we will elaborate on how to build the competing attractive and repulsive interactions in this double dressing scheme .
the van der waals ( vdw ) interaction of the rydberg state @xmath19 is @xmath20 with @xmath21 the corresponding dispersion constant ( @xmath22 ) , while the inter - state rydberg interaction can be neglected when the two states are far separated energetically @xcite .
the laser ( rabi frequency @xmath23 and detuning @xmath24 ) induced interaction is given by @xmath25 with @xmath26 and @xmath27 . here
the @xmath28 and @xmath29 sign correspond to the red- and blue - detunted laser , respectively . for the dressing by the blue - detuned laser
, @xmath30 determines the distance of the two - atom resonant excitation when @xmath31 @xcite [ see fig . [
fig : potential](b ) ] .
therefore the resulting interaction @xmath32 is attractive for @xmath33 and repulsive when @xmath34 .
the dressing of the red - detuned laser generates a barely repulsive soft - core interaction @xmath35 .
the overall interaction potential is given by @xmath36 .
as depicted in fig .
[ fig : potential](b ) , it is divided into an attractive ( negative ) branch and a long - range repulsive vdw part .
moreover the relative value @xmath37 at the attractive branch can be tuned by controlling laser parameters , which is illustrated in fig .
[ fig : potential](c ) . in a 1d lattice ,
the number of lattice sites ( red circles ) on the attractive branch depends on the length @xmath4 . in this work we focus on the case where only the nn interaction is attractive .
we parameterize the nn interaction through @xmath37 and the long - range tail approximately through @xmath38 for each @xmath39 . here
@xmath40 and @xmath3 is the lattice spacing .
the above system is described by the hamiltonian @xmath41 where @xmath42 and @xmath43 .
note , that we set @xmath3 as the length unit and @xmath44 as the energy unit without loss of generality .
_ emergent devil s staircase._let us now determine the ground state configuration of the model hamiltonian ( [ hamiltonian ] ) in the classical limit ( @xmath45 ) .
one notable feature is that the crystal formed in the ground state depends on the sign of the nn interaction @xmath37 . in the case of positive @xmath37
the introduction of an excitation is energetically unfavorable and the ground state of eq .
( [ hamiltonian ] ) is the empty state ( @xmath46 ) . on the other hand in the limit of infinite negative @xmath37 the ground state
is the ferromagnetic state ( @xmath47 ) .
we are interested in how the ground state interpolates between these two cases . versus nn interaction strength @xmath37 . the particle part ( @xmath48 ) and the hole part of the staircase ( @xmath49 ) are indicated by the brown and blue shaded areas and merge at filling fraction @xmath50 .
( b ) magnified view of the dimer particle staircase .
the brown curve is obtained with the effective dimer particle theory of eq .
( [ eq : dimerparticle ] ) .
the black curve ( in all panels ) is the numerically calculated staircase taking into account any rational filling fractions @xmath51 with @xmath52 where the range of the interaction has been cut off at @xmath53 lattice sites .
( c ) at large filling fraction ( @xmath49 ) the main plateaus ( black ) correspond to arrangements of a single kind of @xmath54-mers .
the small intermediate plateaus ( green ) contain mixtures of @xmath54-mers and @xmath55-mers .
the red stars correspond to the predicted transition points given by eq .
( [ eq : trans cond ] ) .
( d ) a unified picture for the staircase structure at large fillings is obtained by employing an effective dimer hole theory of eq .
( [ eq : dh_theory ] ) which agrees excellently with the numerical data . ]
we first perform a numerical analysis of the infinite system . to this end
we consider configurations which are periodic with period @xmath56 , i.e. considering rational filling fractions @xmath51 , with @xmath57 , and we set the maximal @xmath58 .
for each value of the interaction @xmath37 we identify the ground state numerically as the configuration with minimum energy density , i.e. , energy of a single period divided by its length . using this method
, we obtain a staircase pattern connecting the empty state ( @xmath59 at @xmath60 with the completely filled state ( @xmath61 ) at @xmath62 , depicted in fig .
[ f2](a ) .
the data show that we never observe isolated excitations , i.e. an atom in state @xmath0 surrounded by atoms in state @xmath15 , but rather polymers of neighboring excitations .
this is because aggregating at least two excitations is energetically favorable for negative @xmath37 .
we find that the excitations arrange in dimers for filling fractions @xmath48 , which is in agreement with the results of ref .
@xcite .
once the threshold of @xmath50 is reached the ground state configuration becomes @xmath63 . in the region of @xmath49
larger polymers form and the numerical results suggest that the ground state exhibits a rather intricate crystalline structure . in the following
, we will aim at understanding the staircase analytically in both the regimes of @xmath48 and @xmath49 . in the regime of @xmath48
inspired by the numerical observations that two excitations are always on neighbouring sites we formulate an effective dimer particle theory to describe the resulting crystalline structure .
we consider the presence of a dimer particle ( dp ) on sites @xmath64 as an effective single particle with associated dimer particle number operator @xmath65 .
the effective hamiltonian of the dimers then reads @xmath66 where @xmath37 assumes the role of a chemical potential for the dimers . here
@xmath67 is the interaction potential between two dimers separated by @xmath68 sites , which is convex .
note , that for @xmath48 the convexity of the potential ensures that no neighbouring dimers are allowed .
we can now employ a method used by hubbard @xcite to identify stability regions of different densities of dimers with respect to @xmath37 .
the result is plotted in fig .
[ f2](b ) .
one observes a remarkable agreement between the dimer model and the numerical result by solving the full hamiltonian . in the regime of @xmath49
unlike the traditional staircase where the structure of the staircase in this hole sector is trivial , a detailed study must be carried out since there is no explicit particle - hole symmetry in our system ( [ hamiltonian ] ) . from the numerical data we find that in this regime excitations form polymers constituted of @xmath54 neighboring excitations ( so - called @xmath54-mers ) .
we find regions in which only one kind of @xmath54-mer is stable , and regions in which different kinds of @xmath54-mers coexist .
the former correspond to the wider plateaus in the staircase [ black data points in fig .
[ f2](c ) ] .
the latter `` mixed '' cases correspond to the narrow regions interpolating between two wider plateaus [ green data points in fig .
[ f2](c ) ] . as a first approximation
, we can construct an @xmath54-mer theory to describe the staircase as the dominant plateaus correspond to different kinds of @xmath54-mers ( without mixtures ) .
our aim is to determine when the crystal changes from @xmath54-mer to @xmath55-mer as we change @xmath37 , which is achieved by examining the stability of @xmath54-mers for a given @xmath37 . considering an @xmath54-mer as a single particle
, we define the effective filling fraction of @xmath54-mers as @xmath69 , which is related to the real filling fraction @xmath51 through @xmath70 $ ] . in
the following we focus on filling fractions with @xmath71 .
the energy of an @xmath54-mer in the thermodynamic limit reads @xmath72 where @xmath73 is the bond energy of the @xmath54-mer , which is associated to the mutual interactions of its internal excitations .
$ ] is the interaction energy between two @xmath54-mers separated by @xmath75 ( @xmath76 ) , where @xmath77 is the length of the unit cell .
when the crystal changes from @xmath54-mer to @xmath55-mer , the corresponding unit cell length changes from @xmath78 ( for @xmath54-mer ) to @xmath79 ( for @xmath55-mer ) .
the infinite chain can be divided in equal periods on which either @xmath80 @xmath54-mers or @xmath77 @xmath55-mers are disposed , so that we can compute the transition point by solving @xmath81 we observe numerically that in the regions in which only one kind of @xmath54-mer is stable the @xmath54-mers are homogeneously distributed and separated by two holes , giving @xmath82 and therefore @xmath83 .
this numerical observation is confirmed analytically in @xcite , where we show that any other filling fraction makes the resulting @xmath55-mer arrangement unstable .
the comparison of @xmath37 obtained from the simple theoretical model eq .
( [ eq : trans cond ] ) with the numerical data is shown in fig .
[ f2](c ) .
one can see that the transition values of @xmath37 are in excellent agreement .
the presence of two neighboring holes in each unit cell stimulates us to search for an effective theory of dimers of holes ( dh ) . in analogy to the dimers of particles eq .
( [ eq : dimerparticle ] ) , the effective hamiltonian reads @xmath84 where @xmath85 is the interaction energy between dimer holes , @xmath86 is the dimer hole number operator and @xmath87 can be regarded as a chemical potential .
we can again study the stability regions with a method used by hubbard @xcite as the interaction is convex . in fig .
[ f2](d ) we report the filling fraction of excitations found in this way , and we compare them with the numerical results of the full hamiltonian , finding a perfect agreement in the region @xmath49 . note by setting the chemical potential @xmath88 , no holes are present , and a single polymer extends over the whole system corresponding to the aforementioned ferromagnetic state . we can then identify the transition value of @xmath37 to the ferromagnetic state exactly , i.e. , @xmath89 . note that the emergent staircase is the union of two sub - staircases and is complete due to the fact that the long - range interactions @xmath90 and @xmath91 in the respective effective models are convex @xcite . .
left panel shows the transverse magnetization @xmath92 as function of @xmath37 and @xmath17 , where the simulation was done for @xmath93 lattice sites with open boundary conditions and the range of the interactions has been truncated to 10 lattice sites .
the four largest stability regions @xmath94 are marked by dashed lines , see also fig .
[ f2 ] .
right panel shows the idmrg calculations of the density - density correlation function @xmath95 vs. @xmath96 for three different values of @xmath97 at @xmath98 corresponding to the points a , b , c in the left panel ( see text for details ) . ] _ melting of the staircase._ in the case of a long - range repulsive interaction , it was shown in @xcite that the introduction of a quantum term leads to a two - stage melting process when the excitations are dilute . for weak transverse fields the crystalline structure melts into a floating solid which will eventually melt into a paramagnet for higher values of @xmath17 . here
we explore the qualitative features of the melting process in the context of competing attractive and repulsive interactions for a relatively dense crystal .
we perform numerical studies by means of matrix product states implementations of the density matrix renormalization group ( dmrg ) @xcite and its infinite version idmrg @xcite .
the results are presented in fig .
[ f3 ] .
the left panel shows the transverse magnetization @xmath92 as function of @xmath37 and @xmath17 , where @xmath99 is the length of the chain . the transverse magnetization increases with increasing @xmath17 .
typically crystals in the middle of a plateau are more robust against the quantum fluctuation , which leads to a series of lobes similar to the one found in @xcite . here
we study the melting of the @xmath50 lobe which has large stability region and is thus naturally easier to access experimentally ( see @xcite for a detailed discussion of the melting process for lobes with different filling fraction ) .
we use the density - density correlation function @xmath100 ( which is independent of the initial site @xmath101 for an infinite system ) to characterize the melting process along the black line in fig .
[ f3 ] . performing a numerical simulation for vanishingly
small values of @xmath17 ( close to the classical limit ) is challenging due to the presence of quasi translational invariance which leads to a highly degenerate ground state .
we consider @xmath102 ( see point a in fig .
[ f3 ] ) , which allows the degeneracy to be slightly lifted while still inside the lobe .
we find a periodic correlation function @xmath95 characteristic for a crystalline state . in our case ( @xmath50 )
the ground state is quasi - four - fold degenerate and for small values of @xmath17 has the structure @xmath103 that allows us to derive the approximate expression @xmath104 which reproduces remarkably well the results shown for point a in fig .
[ f3 ] .
when @xmath17 is sufficiently large , the ground state of the system is in a paramagnetic phase and the envelope of @xmath95 shows an exponential decay ( point c in fig . [ f3 ] ) . by considering intermediate values of @xmath17 we study the melting of the crystalline phase and find that in this regime the correlations are significantly longer ranged . in the right panel of fig .
[ f3 ] , we show a typical result with bond dimensions of 100 at @xmath105 ( point b in fig .
[ f3 ] ) .
the correlations in the vicinity of the point b show similar behavior suggesting that there is a region with quasi long - range order extending from @xmath106 to @xmath107 .
_ summary and outlook._in conclusion , we have shown that a long - range interaction with competing attractive and repulsive parts can result in a new mechanism to form a complete devil s staircase .
in particular we have considered the case of attractive nn interactions which lead to a dimer staircase without manifest particle - hole symmetry .
this suggests an interesting way to control the structure of the hole sector of the staircase by tailoring the repulsive tail .
for example , one will obtain a trimer - hole staircase through replacing the longer ranged interaction part from the vdw shape ( @xmath108 ) to a dipolar interaction potential ( @xmath109 ) .
moreover , since in the outlined double rydberg dressing scheme the range of the attractive branch of the potential can be tuned freely , and therefore be extended over more lattice spacings , we expect that more exotic polymer staircases can be explored . _
acknowledgements. _ the research leading to these results has received funding from the european research council under the european union s seventh framework programme ( fp/2007 - 2013 ) / erc grant agreement no .
335266 ( escquma ) , the eu - fet grant no .
512862 ( hairs ) , the h2020-fetproact-2014 grant no .
640378 ( rysq ) , and epsrc grant no .
ep / m014266/1 .
wl is supported through the nottingham research fellowship by the university of nottingham .
38ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty link:\doibase 10.1103/physrevb.17.494 [ * * , ( ) ] http://stacks.iop.org/0022-3719/11/i=16/a=022 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.49.249 [ * * , ( ) ] http://stacks.iop.org/1367-2630/12/i=11/a=113037 [ * * , ( ) ] link:\doibase 10.1103/physrevb.80.174519 [ * * , ( ) ] link:\doibase 10.1103/physrevb.86.064501 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.108.255302 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.105.230403 [ * * , ( ) ] http://stacks.iop.org/1367-2630/14/i=9/a=095009 [ * * , ( ) ] e. sela , m. punk , and m. garst , phys .
rev .
b * 84 * , 085434 ( 2011 ) .
link:\doibase 10.1088/1367 - 2630/17/1/013008 [ * * , ( ) ] link:\doibase 10.1126/science.1258351 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.71.779 [ * * , ( ) ] http://stacks.iop.org/0953-8984/19/i=7/a=073101 [ * * , ( ) ] link:\doibase 10.1103/physreve.76.031501 [ * * , ( ) ] http://stacks.iop.org/0953-8984/16/i=42/a=009 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.94.208301 [ * * , ( ) ] link:\doibase 10.1103/physreve.64.041506 [ * * , ( ) ] http://stacks.iop.org/1367-2630/10/i=9/a=093009 [ * * , ( ) ] http://stacks.iop.org/1367-2630/13/i=7/a=073044 [ * * , ( ) ] http://stacks.iop.org/1367-2630/16/i=12/a=123036 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.104.195302 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.105.160404 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.105.135301 [ * * , ( ) ] link:\doibase 10.1103/physreva.85.053615 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.110.263201 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.107.060402 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.107.263001 [ * * , ( ) ] link:\doibase 10.1038/nature11596 [ * * , ( ) ] link:\doibase 10.1038/nature12227 [ * * , ( ) ] @noop ( ) , link:\doibase 10.1103/physreva.84.041607 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.109.233003 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.110.213005 [ * * , ( ) ] http://stacks.iop.org/0295-5075/50/i=3/a=307 [ * * , ( ) ] @noop , , , , * * , ( ) .
, , , ( ) , . , , , , ( ) , .
s. suzuki , j .-
i inoue , and b. k. chakrabarti , _ quantum ising phases and transitions in transverse ising models _ , lecture notes in physics vol .
862 ( springer , 2013 ) .
\doibase http://dx.doi.org/10.1016/j.aop.2010.09.012 [ * * , ( ) ] @noop ( ) , http://stacks.iop.org/1367-2630/12/i=10/a=103044 [ * * , ( ) ] |
here i solve directly in minkowsky space the mass gap equation for spontaneous chiral symmetry breaking ( s@xmath0sb ) .
s@xmath0sb was introduced by the original work of nambu and jona - lasinio @xcite , and it is presently accepted to occur in hadronic physics , where it generates a constituent mass for the quarks .
s@xmath0sb also implies the partially conserved axial current theorems @xcite , including a small mass for the pion . in the literature the mass gap equation
is usually solved either in equal time @xcite or in euclidean space @xcite , in order to avoid the poles and complex quantities which are expected in full minkowsky calculations .
recently the scientific community has been exploring different approaches to the minkowsky space @xcite . an exact solution in minkowsky space
will test the quality of the aproximate solutions .
moreover a solution in minkowsky space opens wider applications .
for instance the bethe salpeter amplitudes can be computed on the mass shell momentum @xmath1 , and at the same token it is possible to boost the hadrons to any convenient frame . in this paper
an analytical approach is applied to finite and analytic kernel . for simplicity
the infrared or ultraviolet divergences are not addressed here .
although a complete solution can only be achieved numerically , an interesting insight is nevertheless presented by the analytical approach . in the schwinger - dyson formalism truncated at the rainbow level ,
the mass gap equation is , @xmath2 for the quark - quark interaction i consider the class of kernels that can be decomposed in a class of yukawa potentials , which is finite both in the ultraviolet and infrared limits , @xmath3 to ensure that the kernels are finite in the ultraviolet limit of large @xmath4 i assume that @xmath5 , since this implies that the kernel vanishes at least like @xmath6^{-2}$ ] . in the infrared limit
the kernels are finite when the @xmath7 are finite , or when infrared cancellations occur . the class of kernels defined in eq .
( [ kernel ] ) is quite general , for instance it includes the coulomb interaction regularized by a pauli - villars term .
it also includes the fourier transform of the confining linear potential , @xmath8 and of the linear potential with a negative infinite shift @xmath9 which are both obtained in the limit of vanishing @xmath10 . using finite differences , the derivatives in @xmath10 can be decomposed in the class of potentials of eq .
( [ kernel ] ) .
the class of kernels addressed in this paper is not only general , it is also covariant , analytical and forward propagating , and this is convenient to study the mass gap equation in minkowsky space .
chiral symmetry breaking occurs when a non - vanishing mass is dynamically generated . for clarity
let us also assume that @xmath11 in eq .
( [ propagator ] ) .
this approximation is qualitatively acceptable when the kernel is finite @xcite .
computing @xmath12 is not difficult , but it obscures the result of the paper . then the mass gap equation is a single non - linear and integral equation for @xmath13 , @xmath14 where all dirac and colour algebraic factors are absorbed in the coupling constants @xmath15 of the potential @xmath16 . the chiral limit of vanishing current quark mass ( the mass in the free quark propagator ) @xmath17 , is also assumed .
this is particularly interesting because it applies to the physics of the quarks @xmath18 and @xmath19 .
it is clear that eq .
( [ mass equation ] ) then has a trivial solution @xmath20 .
the problem that this paper addresses is the other possible solutions of eq .
( [ mass equation ] ) , with the kernel of eq .
( [ kernel ] ) .
the technical difficulties reside in the multiple integral with poles and complex quantities and in the non - linearity of the self - consistent equation .
this paper is organised in sections . in section [ standard ]
standard approximations are applied to the the rainbow schwinger - dyson , and the need to perform a calculation in minkowsky space is motivated . in section
[ one loop ] the integral in the mass gap equation is computed analytically in the case where the quark mass is assumed to be constant .
the solution of the minkowsky non - linear integral mass gap equation is addressed in section [ non - linear ] . finally in section
[ conclusion ] the results and conclusion are presented .
in the literature the mass gap equation is usually solved either in equal time @xcite or in euclidean space @xcite . in the equal time approximation , the lorentz invariance is lost .
the space and time components of physical constants , say @xmath21 or @xmath22 may differ @xcite .
moreover it is not clear how to boost the hadrons outside the centre of mass frame . in the euclidean approximation
it is not clear if a simple wick rotation @xmath23 is exact because there may exist poles in the path ot the @xmath24 axis .
moreover it is very hard to rotate back to minkowsky with the inverse rotation @xmath25 when only a numerical expression of the dynamical mass is known .
this is connected to the topological problem of discretising a curved surface .
although these two methods are approximate , they are used in the literature because they are fully consistent with dynamical symmetry breaking , providing the same approximation is also used in the bound state equations @xcite . at the same token these methods avoid the technical problem of addressing poles and complex quantities , which are expected in minkowsky space . by providing a solution of the mass gap equation , both lorentz invariant and in minkowsky space , both the equal time and the euclidean time approximations
may be better understood . here
the approximate methods for solving the schwinger - dyson equation are reviewed . in the instantaneous or equal time approximation , it is assumed that the dependence in @xmath24 and in @xmath26 is irrelevant in the kernel of the mass gap equation .
the angular integral of the coulomb potential with neglected time component is , @xmath27 \ , \label{instantaneous kernel}\end{aligned}\ ] ] where the notation for quadri - vectors and for tri - vectors is @xmath28 . in what concerns the mass ,
a solution exists where the mass is independent of @xmath26 , and the integral in @xmath26 is trivial , @xmath29 } = { \pi \over \sqrt { { \bf q}^2 + m^2 ( { \bf q}^2 ) } } \end{aligned}\ ] ] and the mass gap equation in the instantaneous approximation is , @xmath30 \nonumber \\ & & { m ( { \bf q}^2 ) \over \sqrt { { \bf q}^2 + m^2 ( { \bf q}^2 ) } } \ .
\label{mge instantaneous}\end{aligned}\ ] ] this is a one - dimensional non - linear integral equation .
it has no singularities and it is solvable numerically @xcite . in the euclidean approximation , it is assumed that the time component of the momentum can be replaced @xmath31 or @xmath32 .
a convenient angular description of the variables is , @xmath33 where the euclidean momentum is , @xmath34 . in this case
the boson exchange potential of eq .
( [ kernel ] ) is finite for all momenta .
covariance allows the choice of the external momentum parallel to the fourth axis , and three angular integrals can be performed , @xmath35 ^ 2 \label{euclidean kernel}\end{aligned}\ ] ] which is a function of the real and positive variable @xmath36 .
again the mass gap equation is reduced to a one - dimensional non - linear integral equation , @xmath37 { m(q^2 ) \over -q^2 + m^2(q^2 ) } \label{mge euclidean}\end{aligned}\ ] ] which has no singularities and is solvable numerically @xcite . a third perspective , differing from the equal - time approximation because it includes the retardation in the kernel and differing from the euclidean approximation because it addresses time - like momenta , is provided by the coupled channel approach
. here the approximation consists in only considering the positive energy poles of the propagators .
for simplicity i consider here a single yukawa term , @xmath38 then the integral in @xmath26 produces , @xmath39 the @xmath40 , @xmath41 , and the fermion mass @xmath42 can be included in the boson creation vertex , @xmath43 then the mass gap equation is equivalent to a coupled channel hamiltonian equation , where the one fermion channel is coupled to the fermion plus boson channel by the boson creation and annihilation vertices , @xmath44 \left ( \begin{array}{c } \phi_f \\ \phi_{f , b } \end{array } \right ) = 0 \ .\ ] ] reducing the equation by substitution of the fermion plus boson wave - function , the secular equation is , @xmath45 \phi(p ) = 0 \ , \label{mge secular}\end{aligned}\ ] ] where it is clear that when @xmath46 , the threshold for a boson production is open .
this produces a cut in the function @xmath47 , with an imaginary component for the mass .
actually more cuts appear when the energy is further increased , and the system couples to to 1 fermion and 2 bosons , 1 fermion and 3 bosons ... therefore expect cusps and imaginary components are expected to appear when @xmath48 , @xmath49 ... when the mass gap equation is solved in the minkowsky space , then all the features of the three different approaches of eqs .
( [ mge instantaneous ] ) , ( [ mge euclidean ] ) and ( [ mge secular ] ) are expected to appear .
non - trivial solutions of the mass gap equation , and imaginary masses above thresholds are expected .
moreover it will be interesting to study the effect of the negative energy components on eigenvalues .
as a first approach to the minkowsky integral , i assume a real constant mass @xmath42 and compute analytically the integral of eq.([mass equation ] ) . this can be regarded a one loop approximation to the mass gap equation , in the sense that the correct solution can be obtained iteratively , in an infinite loop calculation .
the analytical result may also be eventually used to remove poles form the numerical iterative program . in this sense
i choose to perform the multiple integrals in the same ordering that may be used in a numerical integration . a first attempt to start by an analytical integral of the three angular integrals
( both trigonometric and hyperbolic ) was abandoned because the result was indeterminate .
therefore the two angular integrations of the three dimensional space are first performed , in order to simplify the kernel . finally the double integration in the temporal @xmath26 and in the spatial @xmath50 are performed .
the main aim of this section is the covariant study of the positive @xmath51 case which was not accessible to the approximate approaches .
covariance allows one to consider the @xmath52 case .
analyticity can be used later , to continue this function to a @xmath51 negative case .
the two dimensional angular integrals are trivial .
the integral in @xmath26 is also directly computed , using the residue theorem , @xmath53 the integral in @xmath54 can also be performed analytically .
it is interesting to remark that the threshold for an imaginary contribution appears when @xmath55 .
indeed the pole appears at the root @xmath56 of , @xmath57 where @xmath56 coincides with the mass shell momentum above threshold .
a simple algebraic simplification transforms the integral into , @xmath58 where the pole @xmath59 defined in eq .
( [ pole ] ) is real when the threshold opens , at @xmath55 .
@xmath56 is also real when @xmath60 , below the pseudo - thershold , @xmath61 \left[p_0 ^ 2 -(m - \lambda_i)^2 \right ] \over 4 p_0 ^ 2 } } + i \epsilon \ \ , \ ] ] however in this case the residue vanishes , therefore @xmath62 is only expected to possess an imaginary component above the threshold for the fermion - boson production . after cumbersome calculations , where the ultraviolet divergent terms cancel because they are proportional to @xmath63 , the integral @xmath64 can finally be reduced to the exact form , @xmath65 - \log \left [ i \ , \left(1 + { \sqrt{\rho^2 + \lambda_i^2}\over \rho } \right ) \right ] \biggr\ } \biggr ) \nonumber \\ & & + ( m^2 + p_0 ^ 2 - \lambda_i^2 ) { \rho \over \sqrt{\rho^2 + m^2 } } \biggl\ { \log \biggl [ i \ , \left ( 1 - { \sqrt{\rho^2 + m^2 } \over \rho } \right ) \biggr ] - \log \left [ i \ , \left ( 1 + { \sqrt{\rho^2 + m^2 } \over \rho } \right ) \right ] \biggr\ } \biggr ) \biggr ] \ , \label{analytical}\end{aligned}\ ] ] where @xmath56 is defined in eq . ( [ pole ] ) .
the result of eq .
( [ analytical ] ) is not only correct for time - like momentum , it also applies to negative @xmath66 , where the integrals of eq .
( [ first integral ] ) remain correct .
the integral @xmath67 is depicted in fig .
[ real and imaginary ] .
there are three particular cases of eq .
( [ first integral ] ) that can be tested independently .
the imaginary part can be directly computed with the residue theorem applied to eq.([pole separated ] ) . when @xmath55 the imaginary part is ,
@xmath68=\sum_i -\alpha_i \ , \pi^3{\sqrt{\left(p_0 ^ 2 -m^2 - \lambda_i^2\right)^2 -4 m^2 \lambda_i^2 } \over p_0 ^ 2 } \ . \label{imaginary}\ ] ] ( 250,170)(0,0 ) ( 0,0 ) ( 140,160)@xmath69 $ ] ( 180,15)@xmath70 $ ] ( 230,95)@xmath71 another important particular case is the matching point , between the time - like momentum and the space - like momentum , of @xmath72 .
the integral of eq .
( [ first integral ] ) is then easy to compute , @xmath73 finally in the space - like case @xmath74 can be considered .
this implies that the poles are on the correct quadrants to enable a trivial wick rotation .
i can also use the angular integrals already performed in eq .
( [ euclidean kernel ] ) .
the resulting integral can be computed analytically , @xmath75\biggr\ } \ .
\label{space - like}\end{aligned}\ ] ] these three particular cases comply with eq .
( [ analytical ] ) .
here the mass gap equation is solved in an analytical one loop approximation . the standard method to solve the mass gap equation ( [ mass equation ] )
is the iterative method , where one starts by an initial educated guess for the mass @xmath76 .
i consider a constant @xmath77 , which allows the analytical computation of the integral in eq .
( [ mass equation ] ) , with the techniques used in section [ one loop ] .
this produces the next mass in the iterative series , @xmath78 where the integral @xmath67 is defined in eq .
( [ analytical ] ) . to find an exact solution to the mass gap equation , one
would then need to continue this iterative process , computing again the integral with the function @xmath79 and so on until the method converges .
however these further iterative steps would probably need a numerical computation , since the function @xmath80 is already a complicated one .
therefore , in this analytical approach , i choose to stop at the second step of the iteration .
to minimise the error of this one loop computation , i demand that the mass @xmath80 coincides with the mass @xmath77 at the mass shell momentum @xmath81 , @xmath82 this is equivalent to assume that the mass dependence of the integral in eq .
( [ mass equation ] ) is dominated by the pole neighbourhood .
the quark mass , dynamically generated in the mass gap equation , is expected to coincide with the constituent quark mass .
the constituent quark mass is estimated in the quark model , where it is a crucial parameter to produce the hadronic spectrum , and where it is of the order of 0.3 gev .
therefore the coupling constants @xmath15 of the quark interaction ( [ kernel ] ) are adjusted to reproduce @xmath83 gev .
an example of this is illustrated in fig .
[ double solution ] .
i now discuss the different scenarios for the parameters @xmath7 of the potential .
i start with the case with _
three yukawa terms_. in the particular case of in eq .
( [ irfinite ] ) , this corresponds to the linear potential , which vanishes in the infrared limit .
it occurs that in this case the generated quark mass @xmath80 is quite small close to the origin @xmath84 .
it is then very difficult to arrive at a quark mass of 0.3 gev .
this may be related to the infrared cancellation , @xmath85 which occurs when the @xmath7 are equally spaced , @xmath86 .
moreover the structure of the potential is quite complicated above threshold , with large cusps
. therefore this class of models is abandoned .
( 250,170)(0,0 ) ( 0,0 ) ( 230,65)@xmath87 ( 100,50)re@xmath88 ( 100,35)im@xmath88 ( 100,20)@xmath77 ( 100,05)re@xmath89 next the simpler case of _ two yukawa terms _ is studied .
this is related to the first derivative of eq .
( [ irdivergent ] ) . in this case
it is easy to adjust the strength parameters @xmath90 to produce a quark mass of 0.3 gev at the pole position of @xmath91 gev @xmath92 .
this seems to agree with the infrared finite result , @xmath93 nevertheless , in this simple case of two yukawa terms , there are two different scenarios . depending on the steepness of the function @xmath94
, it intercepts the function @xmath51 either at a single point or in two points . from a numerical exploration one
concludes that the two scenarios are separated by the line @xmath95 gev@xmath92 .
this line is depicted in fig .
[ one and two solutions ] , in the parameter space @xmath96 .
there is a single solution when @xmath97 for instance this includes the case of a large @xmath98 , which acts as an ultraviolet pauli - villars cutoff , say @xmath99 .
when @xmath100 and @xmath98 are quite large , the solution @xmath101 of the mass gap equation is very smooth , and the instantaneous or the euclidean approaches constitute very good approximations to the actual minkowsky solution .
it is also clear that in the case of a pauli - villars regularisation @xmath11 is not acceptable .
@xmath12 would need to be computed , nevertheless its computation could be performed with the techniques presented in this paper .
( 250,170)(0,0 ) ( 0,0 ) ( 35,150)@xmath102 ( 230,20)@xmath103 ( 150,80)*one solution * ( 40,110)*two solutions * one also concludes that there is a double solution when @xmath104 this includes the case of small and similar @xmath7 .
this case is related to the first derivative of eq .
( [ irdivergent]).in the limit of vanishing parameters @xmath7 , eq .
( [ irdivergent ] ) corresponds to a linear potential with a negative and infinite constant shift , @xmath105 comparing with the potential defined in eq .
( [ kernel ] ) the string tension is @xmath106 . assuming a string constant of the order of 0.14 gev@xmath92 estimated from the quark spectrum , one arrives at the following parameters , @xmath107 and this implies that the potential also includes a negative constant shift of the order of -1.7 gev .
an example of the generated masses is depicted in fig .
[ double solution ] . to proceed iterating the equation ( [ mass equation ] ) , say at two loop order or more , or to compute the quark condensate @xmath108 , the analytical integral becomes quite complicated .
a numerical study is probably necessary and this is not addressed in this paper .
nevertheless the qualitative changes to the one loop computation can be anticipated . in what concerns the positive @xmath51 ,
it is expected that new channels will open whenever @xmath109 , where @xmath110 and @xmath111 are positive integers .
this will affect the imaginary part of the mass . in what concerns the behaviour of the mass @xmath101 for very large positive or negative @xmath51 , an inspection of eq .
( [ mass equation ] ) shows that @xmath112 and therefore the mass should vanish proportionally to @xmath113 . in the one loop approximation the mass only vanishes like @xmath114 , see fig [ real and imaginary ] .
therefore the large momentum behaviour of the generated quark mass is expected to improve in the next iterations .
i address the technically challenging problem of solving the non - linear integral schwinger - dyson equation in full minkowsky space
. an analytical approach is followed , and approximate but analytical expressions for the quark mass are obtained .
i find that the quark mass exhibits a branch cut above the threshold for boson creation , including an imaginary component . in the case of a linear potential
the mass gap equation is expected to have at least two solutions , and this agrees with references @xcite .
the analytical continuation of a the numerical euclidean space solution ( with negative @xmath51 ) into the full minkowsky has been studied in the literature @xcite . here
i verify that this analytical continuation is not uniquely defined , even when a dense set of points is known in the the space - like @xmath115 sector .
in particular the equation ( [ mass equation ] ) includes at least three vanishing imaginary numbers @xmath116 , summed respectively to the masses @xmath117 and @xmath42 . for external space - like momenta @xmath115 ,
these @xmath118 are irrelevant .
however for time - like momenta @xmath119 there are possible branch cuts both above threshold @xmath120 and below the pseudo - threshold @xmath121 , and the integral depends on the sign of each of the three vanishing @xmath122 . only one integral , with all the three @xmath123 is causally correct .
for instance in fig .
[ real and imaginary ] the correct imaginary part of the mass is negative , but a continuation with a positive imaginary would also be analytically possible . as another example , a naive extension of the euclidean eq .
( [ space - like ] ) to @xmath119 fails to coincides with the full minkowsky solution eq .
( [ analytical ] ) for some time - like momenta .
the next step of this study will consist in applying numerical methods to proceed with the study of the mass gap equation .
a numerical integration seems to be necessary to compute the quark condensate @xmath108 .
moreover the exact numerical solution of the mass gap equation ( [ mass equation ] ) may also be attempted . with the analytical method of this paper ,
it may be possible to subtract the poles from the integrand and to compute their contribution analytically .
another interesting method may be the fully numerical monte - carlo method @xcite .
these numerical methods will be applied elsewhere .
y. nambu and g. jona - lasinio , phys .
rev . * 122 * , 345 ( 1961 ) ; phys .
rev . * 124 * , 246 ( 1961 ) .
r. delbourgo and m. d. scadron , j. phys .
g * 5 * , 1621 ( 1979 ) ; p. bicudo , phys .
c * 67 * , 035201 ( 2003 ) [ arxiv : hep - ph/0311277 ] ; f. llanes - estrada and p.bicudo , phys .
d 68 , 094014 ( 2003 ) , a. holl , a. krassnigg and c. d. roberts , arxiv : nucl - th/0311033 ; a. krassnigg and c. d. roberts , arxiv : nucl - th/0308039 .
a. le yaouanc , l.oliver , o. pene and j - c .
raynal , phys rev d * 29 * , 1233 ( 1984 ) ; a. le yaouanc , l.oliver , s. ono , o. pene and j - c .
raynal , _ ibid _ * 31 * , 137 ( 1985 ) ; s. adler and a. c. davis , nucl .
b * 224 * , 469 ( 1984 ) ; p. bicudo and j. e. ribeiro , phys .
d * 42 * , 1611 ( 1990 ) ; p. bicudo , j. e. ribeiro and j. rodrigues , phys .
c * 52 * , 2144 ( 1995 ) ; f. j. llanes - estrada and s. r. cotanch , phys .
lett . * 84 * , 1102 ( 2000 ) [ arxiv : hep - ph/9906359 ] ; a. p. szczepaniak and e. s. swanson , phys .
lett . * 87 * , 072001 ( 2001 ) [ arxiv : hep - ph/0006306 ] ; a. p. szczepaniak and e. s. swanson , phys .
d * 65 * , 025012 ( 2002 ) [ arxiv : hep - ph/0107078 ] .
j. praschifka , r. t. cahill and c. d. roberts , int . j. mod
. phys . a * 4 * , 4929 ( 1989 ) ; c. d. roberts and b. h. j. mckellar , phys .
d * 41 * , 672 ( 1990 ) ; y. b. dai , c. s. huang and d. s. liu , phys .
d * 43 * , 1717 ( 1991 ) ; l. von smekal , p. a. amundsen and r. alkofer , nucl . phys .
a * 529 * , 633 ( 1991 ) .
f. t. hawes , k. kusaka and a. g. williams , arxiv : hep - ph/9411238 ; k. m. maung , c. a. hill , m. t. hill and g. derise , arxiv : hep - ph/0302228 ; v. sauli , jhep * 0302 * , 001 ( 2003 ) [ arxiv : hep - ph/0209046 ] ; v. sauli and j. adam , nucl .
a * 689 * , 467 ( 2001 ) [ arxiv : hep - ph/0110298 ] ; r. alkofer , w. detmold , c. s. fischer and p. maris , arxiv : hep - ph/0309077 ; r. alkofer , w. detmold , c. s. fischer and p. maris , arxiv : hep - ph/0309078 .
a. le yaouanc , l. oliver , s. ono , o. pene and j. c. raynal , phys .
d * 31 * , 137 ( 1985 ) ; p.
bicudo , arxiv : hep - ph/9905345 .
p. bicudo , j. e. ribeiro and a. v. nefediev , phys .
d * 65 * , 085026 ( 2002 ) [ arxiv : hep - ph/0201173 ] ; a. v. nefediev and j. e. f. ribeiro , phys .
d * 67 * , 034028 ( 2003 ) [ arxiv : hep - ph/0212371 ] ; p. bicudo and a. v. nefediev , phys .
d * 68 * , 065021 ( 2003 ) [ arxiv : hep - ph/0307302 ] ; p. j. a. bicudo and a. v. nefediev , phys .
b * 573 * , 131 ( 2003 ) [ arxiv : hep - ph/0308273 ] ; a. a. osipov and b. hiller , arxiv : hep - th/0307035 ; b. hiller and a. a. osipov , arxiv : hep - ph/0301025 .
d. e. soper , phys .
lett . * 81 * , 2638 ( 1998 ) [ arxiv : hep - ph/9804454 ] . | the mass gap equation for spontaneous chiral symmetry breaking is studied directly in minkowsky space . in hadronic physics ,
spontaneous chiral symmetry breaking is crucial to generate a constituent mass for the quarks , and to produce the partially conserved axial current theorems , including a small mass for the pion . here
a class of finite kernels is used , expanded in yukawa interactions .
the schwinger - dyson equation is solved with an analytical approach .
this improves the state of the art of solving the mass gap equation , which is usually solved with the equal - time approximation or with the euclidean approximation .
the mapping from the euclidean space to the minkowsky space is also illustrated . |
as electronic devices become smaller , they leave the regime of classical physics and enter the realm of quantum physics . many classical quantities such as resistance
must be reinterpreted for systems on a mesoscopic scale .
one such classical concept is that of the refrigerator : a device that uses an external source of work to cool a gas . in this paper
we consider whether this classical concept can be applied to an electron gas so that one could cool such a gas by applying a voltage to a device .
there are many ways to cool electrons in a condensed - matter system .
for example , thermoelectric coolers based on the peltier effect @xcite are available commercially .
a different kind of electron - cooling mechanism in semiconductor devices is based on a quasistatic expansion of a two - dimensional electron gas.@xcite still other possibilities include taking advantage of many - body effects that can lead to liquid / gas phase transitions in the electron population in a semiconductor quantum well.@xcite in this paper we investigate electron cooling in a mesoscopic solid - state device using evaporative emission .
this method entails removal of ( `` filtering '' ) electrons from a high - energy subband of a many - electron system , followed by relaxation of the remaining electrons to a temperature lower than that of the initial system .
evaporative cooling is widely used in bosonic systems.@xcite but this method is harder to implement for fermionic systems , as we shall discuss below . in section [ sec : theory ]
we describe the theory of a two - dimensional device to cool electrons in quantum wires .
we use the landauer formula@xcite to analyze the cooling properties of these devices .
this formula was originally developed to explain the transport properties of electrons in a quantum device .
it relates these properties to the quantum mechanical scattering amplitudes for electrons that pass through the device . to calculate these amplitudes
, we use an extension of r - matrix theory@xcite that we summarize in the appendix . in section
[ sec : results ] we use this theory to calculate cooling properties of several two - dimensional devices .
we begin with a simple t - junction device and show that by optimizing its design we can achieve electron cooling .
we improve upon this result by switching to a `` plus - junction '' design , which can give up to 15% cooling . in section [ sec : experiment ] we discuss applications and realistic parameters for a device to cool electrons , and in [ sec : conclusions ] we summarize our key results and describe future research .
our theoretical approach is analogous to the working principle of the classical hilsch vortex tube , @xcite which uses a t - shaped assembly of pipes to separate high - pressure air into a high - temperature system and a low - temperature system .
this separation does not violate the second law of thermodynamics , because the system is driven by an external force .
we use a similar idea to cool electrons in a quantum mechanical system .
the simplest such device uses a t - shaped assembly of quantum wires to remove higher - energy electrons from an electron gas at fixed temperature .
figure [ fig : tregions ] is a schematic of such a configuration that defines the regions of the device .
we assume that our device is formed from a quantum well whose thickness is sufficiently small that the device can be considered two - dimensional .
that is , we assume that the device confines electrons to a layer of thickness @xmath0 such that the confinement energy associated with motion in the @xmath1-direction is much larger than any other energy in the problem .
( this confinement energy is @xmath2 , where @xmath3 is the electron effective mass . ) ) of the junction .
the `` adiabatic region '' is within the electron - phonon scattering length ( @xmath4 ) of the junction .
it is in the adiabatic region that we achieve cooling . at larger distances , in the `` equilibrium region , ''
the electrons have returned to the temperature of the lattice .
[ fig : tregions],width=3 ] there are three leads in the t - junction : the input lead , output lead , and sidearm , as shown in fig.[fig : tregions ] .
we shall label physical quantities by subscripts `` i , '' `` o , '' and `` s '' respectively ; for example , we denote the widths of the leads by @xmath5 , @xmath6 , and @xmath7 .
electrons are injected into this device through the input lead , and in the input region are in thermal equilibrium at an initial temperature @xmath8 and chemical potential @xmath9 .
filtering of higher - energy electrons from the initial electron gas occurs in the _ scattering region_. the rate of scattering into the sidearm depends upon the electron energy : if the subband energy in lead of width @xmath10 is given by @xmath11 , then the `` force '' electrons exert on the sides of the lead , @xmath12 , is larger for larger @xmath13 .
thus when an electron encounters the sidearm , the higher - subband states are more likely to squirt down the sidearm .
alternatively , one can see that the higher - subband electrons scatter preferentially to the sidearm because they are in states with wave functions that are linear combinations of plane waves with larger transverse momenta .
electrons that scatter forward into the output lead proceed into the _ ballistic region _
, where the electron population is determined entirely by the product of the initial electron distribution and the scattering probability . next the electrons enter the _ adiabatic region _ , where they exchange energy among themselves and so relax to a temperature @xmath14 and a chemical potential @xmath15 . in principle , the values of these properties can be calculated from conservation laws for energy and particle number . in section [ sec : results ] we report such calculations and show that for some device geometries the temperature @xmath14 is less than the initial temperature : _ this temperature decrease is the desired cooling effect_. finally , at large distances ( as determined by the electron - phonon scattering rate ) the electrons will return to equilibrium with the lattice at the initial temperature @xmath16 ; this re - equilibration occurs in the _ equilibrium region_. to see cooling , one must measure the electron temperature before they get to this region below we define our notation and describe how we calculate the electron distributions in the input and output leads .
we also define and describe the calculation of a quantitative measure of electron cooling in the device .
provided the electrons in each lead are in thermodynamic equilibrium , we can treat them as an ideal fermi gas . for lead @xmath17 , therefore , the subband electron density ( per unit volume ) in an open ( energetically accessible ) subband @xmath13 is @xmath18 for @xmath13 such that the subband energy obeys @xmath19 . in this equation ,
the density of states is @xmath20^{-1/2}.\ ] ] the quantity @xmath21 is the dimensionless electron energy measured in units of the lowest subband energy of the input channel , @xmath22 .
we choose the zero of energy at the energy of the ground transverse state in the input lead .
we measure all other subband thresholds relative to this energy and in units of @xmath23 , so that @xmath24 the occupation probability for lead @xmath17 is given by the fermi - dirac distribution function for total electron energy @xmath25 , @xmath26 where @xmath27 is the electrochemical potential in lead @xmath17 , @xmath28 is the _ longitudinal kinetic energy _ of the electron , and @xmath29 is boltzmann s constant .
the corresponding subband population in lead @xmath17 is the integral of the subband density ( eq . ( [ eq : subbanddensity ] ) ) over all allowed total electron energies @xmath21 : @xmath30 the total population in lead @xmath17 is the sum of the subband populations for that lead , @xmath31 immediately upon leaving the scattering region the transmitted electrons are in a highly nonequilibrium distribution and can not be characterized by a temperature or a chemical potential . by the time
these electrons have traveled a distance along the lead comparable to several times their relaxation length , they have come to equilibrium at @xmath32 and @xmath33 , and it is meaningful to describe them by a fermi - dirac distribution function @xmath34 .
these output - lead properties refer to the electron population in the adiabatic region , where the electrons are in thermodynamic equilibrium .
given @xmath16 and @xmath35 for the _ input _ lead , our goal is to determine the values of @xmath32 and @xmath33 for the _ output _ lead that yields the lowest @xmath36 ; i.e. , which maximizes cooling of transmitted electrons . for a given @xmath16 and @xmath35
, we can determine the temperature in the output lead @xmath32 by requiring that the number of electrons in the output lead at equilibrium equals the number of electrons transmitted into this lead ( conservation of electrons in the output lead ) . to set up equations to implement this strategy
, we must define subband and lead populations in terms of electrons transmitted from the input lead into the output lead .
the ( state - to - state ) density of electrons transmitted from an open subband @xmath13 of the input lead @xmath37 into an open subband @xmath38 of a lead @xmath39 is @xmath40 for @xmath41 , where @xmath42 is the transmission coefficient from subband @xmath13 in lead @xmath37 to subband @xmath38 in lead @xmath39 .
the restriction that the total electron energy @xmath21 be greater than or equal to @xmath43 ensures that subbands @xmath13 and @xmath38 are both open ; were this restriction violated , then the transmission coefficient @xmath44 would be undefined .
the population of electrons transmitted into subband @xmath38 of the output lead , the transmitted subband density , is @xmath45 hence the total population of electrons transmitted into the output lead , the transmitted density , is @xmath46 for a given @xmath16 and @xmath35 , we can determine the equilibrium temperature in the output lead @xmath32 by solving simultaneously the equation for conservation of the number of electrons , @xmath47 and the equation for conservation of electron energy in this lead : @xmath48 in eq .
( [ eq : conserveenergy ] ) the energy of the electrons at equilibrium in the output lead is @xmath49 where the subband density in the output lead , @xmath50 is defined by eq .
( [ eq : subbanddensity ] ) .
the energy of the electrons _ transmitted into the output lead _ is @xmath51 where the transmitted - electron density @xmath52 is given by eq .
( [ eq : transdensity ] ) . using eqs .
( [ eq : transmittedleadpop ] ) , and ( [ eq : transleadenergy ] ) we calculate the number of transmitted electrons and their energy .
we then calculate the equilibrium temperature and chemical potential that would give the same total number and energy .
this calculation produces the parameters @xmath32 and @xmath33 for the output lead . as a measure of the effectiveness of a given device for cooling electrons
, we define the cooling parameter @xmath53 if @xmath54 , the device _ heats _ electrons . our goal , therefore , is to determine the device geometry and initial electron properties @xmath16 and @xmath35 that minimize @xmath55 .
to achieve optimum cooling the higher - subband electrons should scatter into the sidearm , and the lower - subband electrons should scatter into the output lead . in a real device not all higher - subband electrons will scatter into the sidearm , and not all the lower - subband electrons will scatter into the output lead .
the probabilities for electron scattering , as quantified in transmission coefficients , depend on system properties such as the geometry , and on scattering potentials .
we consider `` perfect devices '' that have no impurities , so electrons are scattered only by the boundaries of the device .
we further assume that the potential energy of the electrons in the leads is zero ; this simplifying assumption is _ not _ essential to either the cooling effect or to our formalism .
we can alter the scattering of electrons by changing the ratio of the width of the sidearm to that of the input lead ( fig .
[ fig : tregions ] ) or by changing the geometry altogether . to calculate the cooling parameter @xmath56 we need to know the population of electrons in each subband . to determine this quantity we must know transmission coefficients from a state in the input lead to states in the output lead .
to calculate these transmission coefficients we use a generalization of r - matrix theory that we summarize in the appendix . to determine the results reported here , we calculated cooling parameters @xmath56 using transmission coefficients in a t - junction device for various ratios of the width @xmath7 of the sidearm to the width @xmath5 of the input lead , keeping the width @xmath6 of the output lead equal to @xmath5 . .
the curves correspond to different subbands of the incident electrons : @xmath58 ( solid curve ) , 2 ( dashed ) , 3 ( short dash ) , 4 ( dash - dot ) , 5 ( dotted ) . the horizontal axis is the energy of the incoming electron measured in terms of the first - subband energy of the input lead , @xmath59 , from a zero of energy at @xmath23 .
lower panel : the cooling parameter @xmath56 for the coefficients in ( a ) for initial chemical potential @xmath60 .
the `` reduced initial temperature '' ( horizontal axis ) is the dimensionless quantity @xmath61 .
[ fig : trans10],title="fig : " ] + .
the curves correspond to different subbands of the incident electrons : @xmath58 ( solid curve ) , 2 ( dashed ) , 3 ( short dash ) , 4 ( dash - dot ) , 5 ( dotted ) .
the horizontal axis is the energy of the incoming electron measured in terms of the first - subband energy of the input lead , @xmath59 , from a zero of energy at @xmath23 .
lower panel : the cooling parameter @xmath56 for the coefficients in ( a ) for initial chemical potential @xmath60 .
the `` reduced initial temperature '' ( horizontal axis ) is the dimensionless quantity @xmath61 .
[ fig : trans10],title="fig : " ] we first consider a t - junction device in which all leads have the same width : @xmath62 . figure [ fig : trans10](a ) shows `` state - to - lead '' transmission coefficients for scattering into the output lead of electrons in different subbands of the input lead .
these coefficients are sums over all energetically accessible ( `` open '' ) subbands @xmath63 of the output lead of state - to - state transmission coefficients @xmath64 [ see eq .
[ eq : transdensity ] ] from a given state @xmath65 of the input lead .
this figure illustrates the loss of some higher - subband electrons from the initial electron distribution .
the cooling parameter @xmath56 for this case is shown in fig .
[ fig : trans10](b ) for different values of initial temperature @xmath16 with the initial chemical potential @xmath60 .
( note that we measure the energy in terms of @xmath23 , and all the energies are measured from @xmath23 .
so @xmath60 means that the external potential of the system is such that the fermi energy is @xmath66 . ) for this geometry @xmath54 for all initial temperatures the cooling parameter .
that is , this device _
heats _ electrons the opposite of the desired effect .
this case is important because it demonstrates that even if high - energy electrons are lost due to scattering , a compensatory loss of low - energy electrons may produce an overall heating effect .
loss of low - energy electrons opens gaps in the electron distribution at low energies .
higher - energy electrons can then relax into these newly accessible low - energy states , with the resulting energy difference liberated as thermal energy . if this happens , then the loss of low - energy electrons will _ heat _ the system .
even a small dip in the scattered - electron distribution at low energies will significantly affect the final temperature . to _
electrons , therefore , it is not sufficient to merely scatter higher - energy electrons .
we must scatter thermally excited electrons _ but not significantly scatter electrons in lower - energy subbands_. as a function of the reduced initial temperature with initial chemical potential @xmath60 for different values of @xmath67 .
the cooling parameter decreases with increasing magnitude of @xmath68 ( solid line ) , 0.9 ( long dash ) , 0.75 ( medium dash ) , 0.6 ( short dash ) , 0.5 ( dash - dot ) , and 0.4 ( dotted ) .
the reduced initial temperature is defined in the caption to fig .
[ fig : trans10 ] .
[ fig : coolingfordifferentt ] ] to determine whether a t - junction device can cool electrons , we now consider several widths @xmath7 of the sidearm in fig .
[ fig : tregions ] . for each geometry
we determine the initial temperature @xmath16 and chemical potential @xmath35 that _ minimize _ the cooling parameter @xmath56 .
table [ tbl : cooltab ] shows these data and the corresponding final temperature @xmath32 and chemical potential @xmath33 for maximum cooling .
cddddd w_s & _ i & t_i & t_o & _ o & + 1.0 & 0.0 & 4.65 & 5.35 & -6.24 & 1.15 + & 3.0 & 5.85 & 7.06 & -5.63 & 1.21 + & 6.0 & 6.77 & 8.69 & -5.06 & 1.28 + 0.9 & 0.0 & 5.68 & 6.34 & -6.27 & 1.11 + & 3.0 & 6.52 & 7.58 & -4.97 & 1.16 + & 6.0 & 7.63 & 9.33 & -4.31 & 1.22 + 0.75 & 0.0 & 4.18 & 4.463 & -3.68 & 1.07 + & 3.0 & 5.36 & 6.01 & -2.58 & 1.22 + & 6.0 & 6.59 & 7.84 & -1.80 & 1.19 + 0.6 & 0.0 & 1.79 & 1.84 & -1.09 & 1.02 + & 3.0 & 6.00 & 6.53 & -1.84 & 1.09 + & 6.0 & 7.39 & 8.40 & -0.77 & 1.14 + 0.5 & 0.0 & 1.85 & 1.79 & -0.75 & 0.96 + & 3.0 & 3.62 & 3.86 & 0.38 & 1.06 + & 6.0 & 6.52 & 7.35 & 0.58 & 1.13 + 0.4 & 0.0 & 2.89 & 2.73 & -0.71 & 0.95 + & 3.0 & 3.28 & 3.25 & 1.54 & 0.99 + & 6.0 & 5.47 & 5.74 & 2.75 & 1.05 + to illustrate these data and the effect changing the geometry in this way , we show in fig [ fig : coolingfordifferentt ] the variation of @xmath56 with initial temperature @xmath16 ( for initial chemical potential @xmath60 ) for different sidearm widths .
these results show that we achieve cooling ( @xmath55 ) for some geometries and heating ( @xmath69 ) for others .
only the transmission coefficients depend on the device geometry , so it is through these quantum - mechanical scattering probabilities that we can control the extent to which a device can cool electrons . and @xmath70 .
electron are scattered into the output lead from subbands of the input lead @xmath71 ( solid curve ) , 2 ( long dash ) , 3 ( medium dash ) 4 ( short dash ) , and 5 ( dotted ) . the horizontal axis is the energy of the incoming electron measured in terms of the first - subband energy of the input lead , @xmath59 , from a zero of energy at @xmath23 .
see also the data in tbl .
[ tbl : cooltab ] , which analyzes cooling for this case .
[ fig:0point4trans ] ] we shall now consider in detail a device with @xmath72 , which gives @xmath73 cooling .
figure [ fig:0point4trans ] shows state - to - lead transmission coefficients for such a device .
different curves correspond to different initial subbands . at low energies
the transmission probability is nearly unity , so for this geometry no low - energy electrons are lost from the initial distribution .
were we to adjust the fermi energy of this device so only the lowest two subbands were occupied , we would see cooling . as a function of the reduced initial temperature for a t - junction device with @xmath74 and @xmath75 for three initial chemical potentials , @xmath76 ( solid curve ) , @xmath77 ( dashed ) , and @xmath78 ( dotted ) .
these data are based on the transmission coefficients shown in fig .
[ fig:0point4trans ] .
maximum cooling is obtained for @xmath60 .
the reduced initial temperature is defined in the caption to fig .
[ fig : trans10 ] .
[ fig : cooling0point4 ] ] in fig .
[ fig : cooling0point4 ] we illustrate the dependence of the cooling effect on the initial chemical potential @xmath35 .
this figure shows the cooling parameter @xmath56 for @xmath72 as a function of the initial temperature @xmath16 for @xmath76 , @xmath77 , and @xmath78 .
electron cooling is maximized for @xmath60 , the edge of the lowest subband . at this chemical potential
all electrons scattered into the sidearm are in the thermally active region of the fermi distribution .
cooling is also obtained for @xmath79 , the edge of the second subband .
but at larger values of @xmath35 the device heats electrons .
this example shows that a t - junction _ can _ cool electrons .
we would prefer , however , a device that produces more cooling than @xmath80 . investigation of other t - junction geometries showed that such a device can not produce significantly more cooling for any value of @xmath67 .
so we next investigate the addition of a second sidearm .
this change produces the `` plus junction '' illustrated in fig .
[ fig : plusfig ] .
since we achieved cooling in a t - junction device with @xmath81 , we shall consider a plus junction with the same lead ratios : @xmath74 and @xmath82 .
( note that the widths of the two sidearms in fig .
[ fig : plusfig ] are the same . )
state - to - lead transmission coefficients for this device are shown in fig .
[ fig : coolplus](a ) , and the resulting cooling parameter @xmath56 as a function of initial temperature @xmath16 in fig .
[ fig : coolplus](b ) .
the latter figure shows that this geometry yields appreciably more electron cooling
greater than @xmath83than did the t - junction devices of tbl .
[ tbl : cooltab ] .
a single unit of a plus - junction device with this geometry would cool room - temperature electrons by about @xmath84 .
cooling could be further enhanced by combining several such units in sequence .
comparison of this plus - junction to the t - junctions discussion previously , as illustrated in fig .
[ fig : cooltogether ] , show how greatly adding a sidearm improves the effectiveness with which this simple device cools electrons .
one could try to further optimize the geometry of this device by , for example , increasing the width of the sidearms with increasing distance from the junction .
alternatively , one could round the sharp corners at each junction into smooth curves .
however , our initial explorations of such alterations ( for a t - junction ) did not produce substantially more cooling than we obtained with the far simpler plus junction in fig .
[ fig : plusfig ] . the essential features that makes the plus junction more effective than any t junction
is the presence of more than one sidearm .
it is also essential the threshold for scattering into these sidearms lie above the threshold for the second subband of the input lead . )
, the width @xmath7 of the two sidearms may be smaller or larger than @xmath5 .
this geometry gives substantially more cooling than any device we have explored that has only a single sidearm .
[ fig : plusfig ] ] .
the curves correspond to different subbands of the incident electrons : @xmath58 ( solid curve ) , 2 ( dashed ) , 3 ( short dash ) , 4 ( dash - dot ) , 5 ( dotted ) .
the horizontal axis is the energy of the incoming electron measured in terms of the first - subband energy of the input lead , @xmath59 , from a zero of energy at @xmath23 .
lower panel : cooling parameter @xmath56 based on the coefficients shown in the upper panel .
the reduced initial temperature is defined in the caption to fig . [
fig : trans10 ] .
[ fig : coolplus],title="fig : " ] + .
the curves correspond to different subbands of the incident electrons : @xmath58 ( solid curve ) , 2 ( dashed ) , 3 ( short dash ) , 4 ( dash - dot ) , 5 ( dotted ) .
the horizontal axis is the energy of the incoming electron measured in terms of the first - subband energy of the input lead , @xmath59 , from a zero of energy at @xmath23 .
lower panel : cooling parameter @xmath56 based on the coefficients shown in the upper panel .
the reduced initial temperature is defined in the caption to fig .
[ fig : trans10 ] .
[ fig : coolplus],title="fig : " ] ( solid line ) , a t - junction with @xmath74 and @xmath85 ( dashed ) , and a plus - junction with @xmath74 and @xmath86 ( short dash ) . in all cases
the initial chemical potential is @xmath60 .
the reduced initial temperature is defined in the caption to fig .
[ fig : trans10 ] .
[ fig : cooltogether ] ]
we now consider a plus - junction device using realistic experimental parameters for insb and gaas . while in sec .
[ sec : results ] we reported results in dimensionless units , we here use dimensional units .
all material properties depend on the fermi energy @xmath87 of the system .
this quantity is inversely proportional to the effective mass @xmath3 of the electrons in the material and is determined by the electron density in the reservoir .
for our device we chose for the initial chemical potential ( which is approximately equal to the fermi energy ) @xmath60 . since we have scaled the energy by @xmath23 and chosen @xmath23 as the zero of energy , setting the initial chemical potential equals zero means that @xmath88 where @xmath5 is the width of the input lead of the device in fig .
[ fig : plusfig ] .
the condition ( [ determintthewidth ] ) allows us to determine the width of the quantum wire as a function of electron density .
since the fermi energy and the subband energies depend on the effective mass in the same fashion , the width of the quantum wire is independent of the material . for a sample with electron density @xmath13 , we have @xmath89 which gives for the width of our quantum wire @xmath90 if , for example , @xmath91 , we obtain @xmath92 , quite a small value .
we can increase this value by decreasing the electron density . for a plus - junction
, we were able to maximize cooling by setting the initial temperature to @xmath93 in dimensionless units .
in dimensional units , this optimum temperature is @xmath94 using eq .
( [ determintthewidth ] ) , we obtain @xmath95 for a sample with @xmath96 , the initial temperature for maximum cooling is @xmath97 for gaas and @xmath98 for insb .
room temperature ( 300k ) corresponds to for insb .
_ one can therefore obtain substantial cooling due to quantum effects in a room - temperature device . _
a cooling parameter of @xmath99 implies that the electron population is cooled by @xmath100 .
the resulting cooled electrons could be used for photo - detection of optical frequencies corresponding to thermal energies near room temperature .
the devices we have considered have only one cooling stage .
one could increase cooling by connecting multiple plus junctions in series .
the spacing between junctions , however , must be large enough that resonances in scattering between sidearms are negligible . if not , one would have to treat the device as a single large quantum mechanical scattering target .
while the presence of such resonances would not preclude cooling , it would make calculations for a chain of junction devices more difficult and sensitive to details of phase breaking .
many photo - detection applications require a cold detector . we have presented results for a prototype device that demonstrates electron cooling in a single - particle picture .
we have shown that , while a naive t - junction can produce modest cooling and may produce heating , adding an additional sidearm yields a device that can produce appreciable cooling at least @xmath83 .
the abrupt discontinuities in the confining potentials in these models are not essential to cooling ; what is essential is that higher - subband states , which consist of states with larger transverse momenta , scatter appreciable into the sidearms .
we therefore expect electron cooling in such devices to be insensitive to details of the potential so long as the potential does not eliminate the states of the lowest subband .
this project was supported in part by the us national science foundation under grant , and , and .
one of the authors ( tj ) would like to acknowledge the university of oklahoma , graduate college for travel and research grants .
we consider the two - dimensional system in fig .
[ fig:2dfig ] .
this system has a central region @xmath101 connected to @xmath102 external regions or `` leads . ''
the leads and the interior region meet at a set of boundary surfaces we denote by @xmath103 , @xmath104 , , @xmath105 .
we treat the boundaries between the shaded and unshaded regions as `` hard walls '' ( infinite potential ) so electron wave functions are non - zero only in the shaded regions .
since there may be more than three leads , we depart from the notation used in the body of this paper ( in which the input , output and sidearm leads were denoted by subscripts @xmath37 , @xmath106 , and @xmath107 ) and denote the input lead by a zero subscript and all other leads by positive integer subscripts .
we measure all distances in units of @xmath108 and energies in terms of @xmath109 .
we seek an analytic solution for the amplitudes of outgoing states in the leads when only one incoming state is occupied .
the time - independent schroedinger equation for the scattering function is @xmath111 where @xmath112 represents the state of an electron with kinetic energy @xmath21 incident in input - lead subband @xmath113 .
note that @xmath114 is well - defined in all leads . in a _
finite _ region , the hamiltonian @xmath115 is not hermitian .
we can produce a hermitian operator by adding to @xmath115 the so - called bloch operator @xmath116.@xcite we denote the eigenfunctions of the sum of these operators in the interior region by @xmath117 and write the so - called bloch eigenvalue equation as @xmath118 inserting the bloch operator into the schroedinger equation we get @xmath119 we now expand the scattering wave function @xmath120 in the set of orthonormal bloch eigenfunctions @xmath121 inserting this expansion into the schroedinger equation and using the properties of the bloch eigenfunctions yields @xmath122 where @xmath123 is the eigenvalue corresponds to the bloch eigenfunction @xmath124 .
this expansion is valid throughout the interior region @xmath101 and on its surface ( see fig .
[ fig:2dfig ] ) . to derive an equation for the r matrix
, we now apply this expansion of the scattering state on each boundary @xmath125 . at each such boundary
we can expand the scattering function in either lead eigenfunctions or bloch eigenfunctions in the interior region . to be specific , we introduce a local cartesian coordinate system for each lead : @xmath126 and @xmath127 are the longitudinal and transverse coordinates of the @xmath128 lead , respectively .
we choose @xmath129 on each boundary .
( one can easily choose any orthonormal coordinate system , _ mutatis mutandis _ ) .
each lead eigenfunction is then a product of a plane wave in the @xmath126 direction and a transverse bound - state eigenfunction @xmath130 .
the scattering wave function in the @xmath128 lead therefore becomes @xmath131 where @xmath132 and @xmath133 are the wave vector and transmission _ amplitude _ for the channel with quantum number @xmath134 in channel @xmath135 . also , @xmath136 is the @xmath137transverse eigenfunction of lead @xmath135 .
finally , @xmath138 is the kroniker delta - function , which ensures that each wave function is defined only in one lead .
if we measure energy in units of @xmath139 then we can express energy conservation in lead @xmath135 as @xmath140 , where @xmath141 is the width of the @xmath128 lead ( in units of @xmath108 ) .
we use this equation to determine the wave vector @xmath132 .
after some algebra we get a set of linear algebraic equations that we can solve for the transmission amplitudes : @xmath142 in writing these equations we have defined @xmath143 finally , the r - matrix is given by @xmath144 this equation is general in that we can easily adapt it to any number of leads and to different choices of input lead .
_ thermodynamics and an introduction to thermostatics _ , herbert .
h. callen , ( john wiley and sons , 1985 ) .
l. g. c. rego , and g. kirczenow , appl .
lett . , 75 , 2262 ( 1999 ) .
thushari jayasekera , k. mullen and michael a. morrison , _
unpublished_. junren shi and x. c. xie , _ phys .
lett . _ * 88 * , 86401 ( 2002 ) .
w. ketterle and n.j .
van druten , in _ advances in atomic molecular and optical physics _ , * 37 * , b. bederson and h. walther eds .
( academic press , san diego , 1996 ) , p181 .
r. landauer , _ phil
. mag . _ * 21 * , 863 , ( 1970 ) .
m. bttiker , _ phys .
lett . _ * 57 * 1761 ( 1986 ) .
a. douglas stone and a. szafer , _
ibm j. res .
_ , * 32 * , 384 , ( 1988 ) .
e. p. wigner and i. eisenbud , _ phys .
* 72 * , 29 ( 1947 ) . c. bloch , _ nucl .
_ * 4 * , 503 ( 1957 ) .
u. wulf , j. kucra , p. n. racec and e. sigmund , _ phys .
b _ * 58 * 16209 , ( 1998 ) .
thushari jayasekera , k. mullen and michael a. morrison , `` r - matrix theory in confined geometries , '' university of oklahoma ph.d thesis .
r. hilsch , _ rev .
* 18 * , 108 ( 1947 ) . _
and yet it moves : strange systems and subtle questions in physics _ , m. p. silverman , ( cambridge university press , 1993 ) . | we discuss the theory of cooling electrons in solid - state devices via `` evaporative emission . ''
our model is based on filtering electron subbands in a quantum - wire device .
when incident electrons in a higher - energy subband scatter out of the initial electron distribution , the system equilibrates to a different chemical potential and temperature than those of the incident electron distribution .
we show that this re - equilibration can cause considerable cooling of the system .
we discuss how the device geometry affects the final electron temperatures , and consider factors relevant to possible experiments .
we demonstrate that one can therefore substantial electron cooling due to quantum effects in a room - temperature device .
the resulting cooled electron population could be used for photo - detection of optical frequencies corresponding to thermal energies near room temperature . |
in the last few years , a new class of cosmological models has been proposed .
the more usual cosmological constant @xmath1-term is replaced with a dynamical , time dependent component , the quintessence or dark energy , added together with baryons , cold dark matter ( cdm ) , photons and neutrinos .
the equation of state of the dark energy is given by @xmath2 , @xmath3 and @xmath4 being , respectively , the pressure and energy density ; since @xmath5 , it still implies a negative contribution to the total pressure of the cosmic fluid . when @xmath6 , we recover a constant @xmath1-term .
one of the more interesting physical realizations of the quintessence is a cosmic scalar field minimally coupled with the usual matter action @xcite .
such a field induces the repulsive gravitational force dynamically , allowing to explain the accelerated expansion of our universe , as discovered with observations of supernovae ( sne ) of type ia @xcite .
together with an accelerated expansion , another experimental fact is the strong evidence of a spatially flat universe @xcite : quintessence could be responsible for the missing energy in a flat universe with a subcritical matter density .
since quintessence drives the cosmological expansion at late times , it also influences the growth of structures arisen from gravitational instability .
quintessence clusters gravitationally at very large scales ( @xmath7 mpc ) , leaving an imprint on the microwave backgroud anisotropy @xcite ; at small scales , fluctuations in the dark energy are damped and do not enter in the evolution equation for the perturbations in the pressureless matter @xcite . at the scale we are considering in the following ,
quintessence behaves as a smooth component : it does not partecipate directly in cluster formation , but it only alters the background cosmic evolution . in this work ,
we study the evolution of density perturbations for a general exact solution for quintessence model , found by one of us together with p. scudellaro ( hereafter the rs model ) @xcite . in section 2
, we present the basic equations of the quintessence models examined in the paper . in section 3
, the linear perturbations equation is solved for the rs potential and the properties of the peculiar velocity field are discussed ; in section 4 , the 2dfgrs data are considered .
section 5 is devoted to some final considerations .
we investigate spatially flat , homogeneous and isotropic cosmologies filled with two non - interacting components : pressureless matter ( dust ) and a scalar field @xmath8 , minimally coupled with gravity .
we consider the rs potential , @xmath9 where @xmath10 .
we refer to rubano & scudellaro @xcite and pavlov et al .
@xcite for details and a discussion of the general exact solution of the field equations and we present here only what is needed for the present work .
it is @xmath11 and @xmath12 where @xmath13 is a dimensionless time ; @xmath14 is a time scale of the order of the age of the universe ; @xmath15 is the time dependent hubble parameter ; @xmath16 is the pressureless matter density ( @xmath17 ) in units of the critical density @xmath18 .
the relation between the dimensionless time @xmath19 and the redshift @xmath20 is given by @xmath21 where @xmath22 is the present value of @xmath19 .
this very simple cosmological model has two free parameters , @xmath14 and @xmath22 , or , equivalently , @xmath23 , the today value of the hubble parameter , and @xmath24 , the present value of the pressureless matter density . as can be seen from eqs .
( [ scal3],[scal4 ] ) , @xmath24 depends only on @xmath22 we shall compare the results for the above universe with a flat model with constant equation of state @xmath25 ( as well known , this includes the @xmath1cdm cosmologies for @xmath26 ) . in this case , the redshift dependent hubble parameter is @xmath27 this class of models has three free parameters , @xmath24 , @xmath23 and @xmath25 .
for perturbations inside the horizon , the equation describing the evolution of the cdm component density contrast , @xmath28 ( for unclustered quintessence ) is @xcite @xmath29 where the dot means derivative with respect to comoving time . in eq .
( [ grow1 ] ) , the relative contribution of dark energy to the energy budget enters into the expansion rate @xmath15 .
we shall consider eq .
( [ grow1 ] ) in the matter dominated era , when the radiation contribution is really negligible . with a rs potential , eq .
( [ grow1 ] ) reduces to @xmath30 equation ( [ grow2 ] ) has two linearly independent solutions , the growing mode @xmath31 and the decreasing mode @xmath32 , which can be expressed in terms of the hypergeometric function of second type @xmath33 .
we get @xmath34,\ ] ] and @xmath35.\ ] ] by comparing eqs .
( [ grow3],[grow4 ] ) with eq .
( [ scal4 ] ) , we observe that the density contrast , as a function of the redshift , depends on only one parameter , @xmath22 , i.e. @xmath24 . for @xmath36 , eqs .
( [ grow3],[grow4 ] ) reduce to the standard results @xmath37 and @xmath38 for the evolution of the density contrast for quintessence with constant equation of state , we refer to silveira and waga @xcite .
the linear theory relates the peculiar velocity field @xmath39 and the density contrast by @xcite @xmath40 where the growth index @xmath41 is defined as @xmath42 @xmath43 is the scale factor , @xmath44 , and @xmath45 is its present value . for a rs potential ,
we get @xmath46 the growth index is usually approximated by @xmath47 . for @xmath1cdm models ,
@xmath48 ; if @xmath49 , @xmath50 ; if @xmath51 , @xmath52 @xcite . in fig . ( [ ln - f - ln - omega ] )
, we show the logarithm of @xmath41 as a function of the logarithm of @xmath16 for the rs potential .
the value @xmath53 provides a good approximation to the model . in order to compare the various models
, we have to stress that the right " value for @xmath24 depends on the model .
it is the result of a best fit prodecure on experimental data , i.e. from sn ia observations .
the well known value @xmath54 refers only to the @xmath1cdm model . with the same procedure applied to sne ia data
, the rs potential gives @xmath55 @xcite .
due to the large errors , the two results are still marginally compatible , so the value @xmath54 holds for the rs model too . in the case of a constant @xmath56 ,
the situation is more complicate , due to the degeneration between @xmath24 and @xmath25 @xcite . in this preliminary study
, we decided to consider only two situations : _
i ) _ a @xmath1cdm , a @xmath56 and a rs model with @xmath57 ; _ ii ) _ the same three models with @xmath58 . as can be seen from fig .
( [ f - z ] ) , a large variation in @xmath24 entails a remarkable difference in the growth index also at low redshifts .
even more remarkable is the difference at large redshifts in the case of the same @xmath24 .
this suggest that with the deep surveys planned in the future , this approach could provide a constraint independent of sne ia @xcite . in fig .
( [ deltaf - z ] ) , we show the relative variations of rs models from @xmath1cdm cosmologies . for models with the same @xmath24 ,
the relative variations are quite small and decrease with increasing pressureless matter density . for @xmath59 ,
the relative variation is @xmath60 ; for @xmath61 , it is @xmath62 .
low redshifts have been investigated by the 2dfgrs collaboration @xcite .
peculiar velocity distorts the correlation function of galaxies observed in redshift space @xcite . in linear theory
, the characteristic quadrupole distortion enables to measure the growth index from redshift galaxy catalogs .
the 2dfgrs has obtained redshifts for more than @xmath63 galaxies with an effective depth of @xmath64 . from a precise measurement of the clustering
, a value of @xmath65 has been determined @xcite , where @xmath66 is the bias parameter , connecting the relative density fluctuations of the galaxies and of the total mass .
the bias parameter has been measured by computing the bispectrum of the 2dfgrs ; it is @xmath67 @xcite . combining the measurements of @xmath68 and @xmath66 , the growth index at @xmath69
can be estimated : it is @xmath70 @xcite . by using the rs potential , from this value of @xmath41
we can obtain an estimate of the today pressureless matter density parameter .
we get @xmath71 this value is significantly compatible with the estimate from the sn ia data .
the present state of accuracy in observations does not allow to discriminate among the illustrated alternatives . in our opinion
, the above discussion adds another argument in favour of the use of exponential potentials in quintessence cosmology .
the strong dependence of the growth index on the model , at high redshifts , suggests that more stringent constraints can be given in the future .
this work has been in part financially sustained by the m.u.r.s.t . grant prin2000 sin.te.si . " . | we investigate the evolution of matter density perturbations and some properties of the peculiar velocity field for a special class of exponential potentials in a scalar field model for quintessence , for which a general exact solution is known .
the data from the 2-degree field galaxy redshift survey ( 2dfgrs ) suggest a value of the today pressureless matter density @xmath0 .
keywords : cosmology : theory - cosmology : quintessence - large - scale structure of universe |
the period - luminosity ( p - l , also known as leavitt law ) relation for cepheid variables is important in modern astrophysics , as it is the first rung in the distance scale ladder .
the p - l relations can also be used to constrain the theoretical pl relations based on stellar pulsation and evolution models .
one important issue in the application of p - l relation in distance scale work is its universality is the slope of p - l relation independent of metallicity ? in this work ,
the multi - band p - l relations were derived for the large magellanic cloud ( lmc ) and small magellanic cloud ( smc ) cepheids , based on the latest release of ogle - iii ( the third phase of optical gravitational lensing experiment ) catalogs .
large numbers ( @xmath2 ) of cepheids in the lmc and smc permit the determination of accurate p - l slopes , hence testing the universality of the p - l relation in low metallicity galaxies .
periods and intensity mean magnitudes in @xmath0 bands were available from ogle - iii catalogs for @xmath3 lmc cepheids @xcite and @xmath4 smc cepheids @xcite .
all of these cepheids are fundamental mode cepheids .
smc cepheids with @xmath5 , however , were removed from the sample , as they followed different p - l relation @xcite .
these cepheids were cross - matched to 2mass point source catalog .
the random - phase @xmath6 photometry were converted to mean magnitudes using the prescription given in @xcite .
the mid - infrared photometry were available from _ spitzer _ archival data
sage @xcite and sage - smc .
zaritsky s extinction maps for lmc @xcite and smc @xcite were used for extinction corrections .
in addition , the extinction - free p - l relation the wesenheit function in the form of @xmath7 , was also derived .
outliers presented in the p - l plane were removed using an iterative sigma clipping algorithm @xcite .
additional period cuts need to be applied in certain bands , as the faint end ( hence shorter period ) of these p - l relations may be affected by incompleteness bias .
llll band & lmc p - l slopes & smc p - l slopes & slope difference + @xmath8 & @xmath9 & @xmath10 & @xmath11 + @xmath12 & @xmath13 & @xmath14 & @xmath15 + @xmath16 & @xmath17 & @xmath18 & @xmath19 + @xmath20 & @xmath21 & @xmath22 & @xmath23 + @xmath24 & @xmath25 & @xmath26 & @xmath27 + @xmath28 & @xmath29 & @xmath30 & @xmath31 + @xmath32 & @xmath33 & @xmath34 & @xmath35 + @xmath36 & @xmath37 & @xmath38 & @xmath39 + @xmath40 & @xmath41 & @xmath42 & @xmath43 + @xmath44 & @xmath45 & @xmath46 & @xmath47 + comparison of the multi - band p - l slopes for these two metal poor galaxies are presented in table [ tab ] .
note that the lmc p - l slopes have been published in @xcite . as can be seen from this table ,
the p - l slopes are within @xmath1 in all bands between the lmc and smc cepheids .
the @xmath44 band p - l slopes are almost identical , suggesting the extinction - free @xmath44 band p - l relation is a good choice for distance scale application .
ccn thanks the funding from national science council ( of taiwan ) under the contract nsc 98 - 2112-m-008 - 013-my3 .
8 bauer , f. , et al . 1999 , , 348 , 175 meixner , m. , et al . 2006 , , 132 , 2268 ngeow , c. , et al . 2009 , , 693 , 691 soszynski , i. , et al . 2005 , , 117 , 823 soszynski , i. , et al . 2008 , acta astron .
, 58 , 163 soszynski , i. , et al .
2010 , acta astron . , 60 , 17 zaritsky , d. , et al .
2002 , , 123 , 855 zaritsky , d. , et al .
2004 , , 128 , 1606 | the period - luminosity ( p - l ) relation for cepheid variables is important in modern astrophysics . in this work
, we present the multi - band p - l relations derived from the large magellanic cloud ( lmc ) and small magellanic cloud ( smc ) cepheids , based on the latest release of ogle - iii catalogs .
in addition to the @xmath0 band mean magnitudes adopted from ogle - iii catalogs , we also cross - matched the lmc and smc cepheids to the 2mass point source catalogs and publicly available _ spitzer _ catalogs from sage program .
mean magnitudes for these cepheids were corrected for extinction using available extinction maps .
when comparing the p - l slopes , we found that the p - l slopes in these two galaxies are consistent with each others within @xmath1 level . |
in four space - time dimensions , there is no scale invariant interacting quantum field theory which contains massive particles .
even if scale invariance is preserved in massless field theory at classical level , it would be broken by renormalization effect which is known as the trace anomaly . nevertheless , it is possible that at a much higher scale there exists a scale invariant sector with a nontrivial infrared fixed point .
recently , it has been argued that one kind of fields , under the name banks - zaks ( @xmath13 ) fields @xcite , might appear at tev scale . at low energy ,
these fields manifest themselves by matching onto a new sector called `` unparticle '' ( @xmath14 ) with a non - integral number of scale dimension @xmath6 @xcite .
although the underlying structure of unparticles is still unclear , there indeed exist respectable interesting phenomena for testing unparticles experimentally . in many processes , @xmath15 @xcite , @xmath16 and @xmath17 @xcite , the productions of these stuff
might be detected by measuring the missing energies and the momentum distributions .
the phenomenological studies on the unparticle effect in charged higgs decays , anomalous magnetic moments , @xmath18 and @xmath19 mixing and hadronic flavor changing neutral current ( fcnc ) in @xmath20 decays are carried out in ref . @xcite . in ref .
@xcite , the lepton flavor violation interaction is introduced to explore the phenomenology in @xmath21 . in this work
, we will investigate the lepton flavor changing processes @xmath22 , @xmath23 and some other related processes .
there are a number of experimental upper bounds on these processes which give stringent constraints to the effective couplings of unparticles .
this can probably shed light on the internal structure of unparticles .
at very high energy , the theory contains the standard model ( sm ) fields and the fields with a nontrivial infrared fixed point , called @xmath13 fields @xcite .
these two sectors interact with each other by exchanging particles with a very large mass @xmath24 . below this mass scale ,
heavy particles are integrated out and thus the sm particles and the @xmath13 sector interacts through non - renormalizable operators : @xmath25 where @xmath26 and @xmath27 are local operators made up of the sm and @xmath13 fields , respectively .
the renormalization effects in the scale invariant sector induce the dimensional transmutation at the scale @xmath28 . below the scale @xmath29
, the @xmath13 operators match onto unparticle operators while the non - renormalizable operators in eq .
( [ eq : nonren ] ) match onto an effective interaction operator : @xmath30 where @xmath31 and @xmath6 are the scaling dimensions of the @xmath32 and unparticle @xmath33 operators respectively . the scaling dimension of unparticle field has been taken as @xmath34 in the literature . to be more specific ,
unparticles have some different characters from ordinary particles in phase space , virtual propagator , and the effective interaction with the sm particles .
it was demonstrated in @xcite that scale invariance can be used to fix the two - point functions of the unparticle operators and the propagators .
the propagator of the scalar unparticle field can be written by @xcite : @xmath35|0\rangle = i\frac{a_{d_{\cal u}}}{2}\frac{1}{\mbox{sin}(d_{\cal u}\pi)}(-p^2-i\epsilon)^{d_{\cal u}-2}.\end{aligned}\ ] ] if the vector unparticle is assumed to be transverse , its propagator has the form @xmath36|0\rangle = i\frac{a_{d_{\cal u}}}{2}\frac{-g^{\mu\nu}+p^\mu p^\nu / p^2}{\mbox{sin}(d_{\cal u}\pi)}(-p^2-i\epsilon)^{d_{\cal u}-2},\end{aligned}\ ] ] where the coefficient @xmath37 is given by @xmath38 there are other possible lorentz structures : a spinor field @xcite or even a tensor field @xmath39 .
the only difference is the spin structure which has been comprehensively discussed in ref .
@xcite .
the effective interactions that satisfy the standard model gauge symmetry for the scalar and vector unparticle operators with standard model fields are given , respectively , by @xmath40 where @xmath41 stands for a standard model fermion and @xmath42 are dimensionless effective couplings @xmath43 with the index @xmath44 labeling the scalar and vector unparticle operators , respectively . in principle , the coupling constants @xmath45 can be different for different flavors and are then distinguished by additional indices .
for example , the hadronic fcnc via scalar unparticle , taking @xmath46 as an example , can proceed through the effective interaction term : @xmath47 where we have used the subscript @xmath48 to denote the coupling with @xmath49 and @xmath50 quark .
similarly , for the vector unparticle , the effective interaction is considered as , @xmath51
in this section , we will consider the lepton flavor violation in neutral meson decays , including @xmath53 , @xmath54 , @xmath55 , @xmath56 and @xmath57 . to be more specific
, we will consider the scalar coupling in eq.([eq : scalar ] ) and the vector coupling in eq.([eq : vector ] ) .
as discussed above , the vector unparticle is assumed to be transverse .
thus this kind of unparticle gives zero contribution , when contraction with @xmath58 from the matrix element @xmath59 . in the following ,
we will first focus on the decay channel @xmath60 induced by scalar unparticle .
the decay amplitude for @xmath61 reads @xmath62v(p_\mu ) \nonumber\\ & & \times \frac{ia_{d_{\cal u}}}{2}\times \frac{1}{\mbox{sin}(d_{\cal u}\pi ) } \times ( -m_b^2-i\epsilon)^{d_{\cal u}-2 } \times i\frac{\lambda_{\mu\tau}}{\lambda_{\cal u}^{d_{\cal u}}}(-p^\mu_b)(-if_b { p_{b}}_{\mu})\nonumber\\ & = & -m_{\tau}\frac{a_{d_{\cal u}}}{2\mbox{sin}(d_{\cal u}\pi)}\frac{\lambda_{db}\lambda_{\mu\tau}}{\lambda_{\cal u}^{2d_{\cal u } } } f_b m_b^2(-m_b^2-i\epsilon)^{d_{\cal u}-2 } \times\bar u(p_{\tau})(1+\gamma_5)v(p_\mu),\end{aligned}\ ] ] where the mass of lighter lepton @xmath63 is neglected .
thus , the decay width for this decay channel can be written as @xmath64 this decay width is proportional to the lepton mass square , due to the @xmath65 current for the interaction type , which indicates this kind of process is helicity - suppressed . as an illustration
, we use the following inputs @xmath66 the experimental upper bound on the branching fraction : @xmath67 leads to a constraint on the coupling constant @xmath68 .
if we take the scaling dimension @xmath6 larger , the constraint becomes less stringent , for example , the constraint for @xmath69 is obtained by taking @xmath11 . as mentioned above , the @xmath65 coupling of unparticles with the standard model fermions results in the famous helicity suppression . if other interactions are introduced such as @xmath70 , the helicity rule is invalid , which can constrain @xmath71 more strictly .
the analysis can be easily generalized to other processes , such as @xmath72 .
these processes can give different constraints .
for example , the constraint on @xmath73 is obtained from @xmath74 . in table [ tab : mtoll ] , we collect the experimental upper bounds @xcite for these channels and their constraints on the leptonic flavor violating processes in @xmath53 , @xmath54 , @xmath55 , @xmath75 , @xmath57 decays . from this table
, we can see the results dramatically depend on the scaling dimension of the unparticle field . since the transversely polarized vector unparticle gives zero contribution to the neutral meson decays
, there is no constraint from experiments to their effective couplings .
.experimental upper bounds on @xmath76 at @xmath77 confidence level @xcite and their constraints on the effective coupling constants @xmath71 performed at @xmath78 tev and @xmath79 , @xmath80 .
[ cols="^,^,^,^",options="header " , ] the constraint on the axial - vector coupling constants can also be similarly analyzed .
since the scalar unparticle coupling does not contribute to the @xmath81 process in the zero lepton mass limit , there is no constraint for their effective coupling from these experiments .
in a short summary , we have explored the phenomenology of unparticle physics with the help of lepton flavor changing processes @xmath82 , @xmath83 and other related processes . in the zero lepton mass limit
, the vector unparticle coupling does not contribute to the neutral meson leptonic decays ; while the scalar unparticle coupling does not contribute to the @xmath81 scattering processes .
therefore , the experimental upper bounds of @xmath82 decays will only constrain the scalar unparticle coupling @xmath71 , which is sensitive to the scaling dimension of the unparticle @xmath6 . for the scattering process
@xmath84 , there is only constraint to the vector unparticle coupling from current experiments .
we also study the @xmath8 dependence of the cross section @xmath9 of process @xmath84 with different values of @xmath6 .
if @xmath11 , the cross section is independent on the invariant mass . for @xmath12 , the cross section increases with @xmath8 .
this work is partly supported by national science foundation of china under grant no.10475085 and 10625525 .
the authors would like to thank ying li , yi liao , yue - long shen and shun zhou for helpful discussions .
m. luo and g. zhu , arxiv:0704.3532 [ hep - ph ] . c. h. chen and c. q. geng , arxiv:0705.0689 [ hep - ph ] .
g. j. ding and m. l. yan , arxiv:0705.0794 [ hep - ph ] .
y. liao , arxiv:0705.0837 [ hep - ph ] .
x. q. li and z. t. wei , arxiv:0705.1821 [ hep - ph ] . | we study the virtual effects of unparticle physics in the lepton flavor violating processes @xmath0 and @xmath1 scattering , where @xmath2 denotes the pseudoscalar mesons : @xmath3 and @xmath4 denote two different lepton flavors . for the decay of @xmath5 ,
there is no constraint from the current experimental upper bounds on the vector unparticle coupling with leptons .
the constraint on the coupling constant between scalar unparticle field and leptons is sensitive to the scaling dimension of the unparticle @xmath6 . for the scattering process
@xmath7 , there is only constraint from experiments on the vector unparticle couplings with leptons but no constraint on the scalar unparticle .
we study the @xmath8 dependence of the cross section @xmath9 of @xmath10 with different values of @xmath6 .
if @xmath11 , the cross section is independent on the center mass energy . for @xmath12 , the cross section increases with @xmath8 . |
the importance of understanding proteins@xcite has attracted much attention to the statistical mechanics of heterogeneous polymers .
a particular type of heteropolymers built with a random mixture of positively and negatively charged groups along their backbone are called polyampholytes ( pas ) .
the presence of long range electrostatic interactions causes a rather unique behavior in such polymers : the behavior of a single pa with unscreened electrostatic interactions at a low temperature @xmath8 is extremely sensitive to its total ( excess ) charge @xmath9 .
geometrical properties of polymers can be conveniently described by their radius of gyration ( root mean squared size ) @xmath10@xcite . at high @xmath8 ,
the effect of electrostatic interactions is small and @xmath10 is approximately equal to that of an uncharged polymer .
however , upon lowering of @xmath8 the pa attempts to take advantage of the presence of two types of charges along its backbone by assuming spatial conformations in which every charge is predominantly surrounded by charges of an opposite sign .
this behavior can be approximately described using a debye
type theory@xcite , which leads to the conclusion that at low @xmath8 the polymer should collapse into an dense state with condensation energy @xmath11 , where @xmath1 is the number monomers , @xmath12 is the typical charge of a monomer , and @xmath13 is a microscopic distance such as diameter of the monomer . in such a collapsed state ,
@xmath14 . on the other hand ,
renormalization group inspired scaling arguments showed@xcite that at low @xmath8 one should expect a strongly stretched state with @xmath15 .
this apparent contradiction was resolved by noting@xcite that the low@xmath8 behavior is extremely sensitive to the overall charge @xmath9 : it has been observed@xcite that randomly charged pas with _ vanishing _
indeed collapse at low @xmath8 , while @xmath10 , which is averaged over _ unrestricted _ quenches , grows with decreasing @xmath8 .
such sensitivity is consistent with experimental observations of pas@xcite . from a detailed study of the @xmath9dependence of @xmath10 ,
the following picture began emerging@xcite : consider a dense ( globular , approximately spherical ) low@xmath8 state of the pa .
its energy can be roughly separated into three terms , as @xmath16 ( in this description we omit the dimensionless prefactors of order unity . )
the first term in this equation represents the debye
hckel type condensation energy ,
the second term is the surface energy ( where the surface tension @xmath17 , and the surface area @xmath18 ) , while the last term is the electrostatic energy of the globule of radius @xmath19 . for vanishing @xmath9 ,
the globule remains approximately spherical .
however , when @xmath20 , the electrostatic term exceeds the surface tension term , the spherical shape becomes unstable and the polymer starts to stretch in order to minimize the electrostatic energy .
since the threshold charge @xmath21 increases with @xmath1 exactly as the standard deviation of the total charge @xmath9 in a random sequence of charges , for any @xmath1 there will be a finite portion of chains with @xmath9 exceeding @xmath21 .
( note that this property is specific to three
dimensional electrostatic interactions . for the @xmath1dependence of @xmath21 in general space dimensions , see ref .
while the above arguments suggest that a typical pa should stretch out at low @xmath8 , such stretching may lead to a loss of the condensation energy .
a reasonable compromise between stretching ( which minimizes the electrostatic energy ) and remaining compact ( which gains in condensation energy ) is for the pa to form a _ necklace _ of weakly charged blobs connected with highly charged necks " , by taking advantage of the charge fluctuations along the chain .
the results of the monte carlo@xcite and exact enumeration@xcite studies qualitatively support such a picture .
an example of such a low energy configuration is shown in fig .
[ paconf ] .
while the exact treatment of electrostatic interactions is not possible , we can pose a simplified problem which , we hope , captures some essential features of this necklace model .
for example , we may ask what the typical size of the largest neutral ( or weakly charged ) segment in a random sequence of @xmath1 charges will be . in order to answer this question , we investigated the size distribution of the largest @xmath22segments ( segments with a total charge @xmath0 ) in such @xmath23mers .
this problem can be mapped to a one - dimensional random walk ( rw ) : the sequence of charges @xmath24 ( @xmath25 ) , is mapped into a sequence of unit steps in the positive or negative directions along an axis .
the sequence of charges with vanishing total charge @xmath9 now corresponds to a rw which returns to the origin after @xmath1 steps , while a neutral segment inside the sequence of charges corresponds to a loop inside the rw .
similarly , a segment with charge @xmath0 corresponds to a segment ( in the corresponding rw ) whose end is displaced by @xmath0 units from its beginning .
the primary objective of this work is to investigate the probability @xmath26 that the _ largest _ @xmath0segment in an @xmath1step rw has length @xmath2 .
there is an apparent simplicity of the formulation of the problem , i.e. it is similar ( and related ) to the classical rw problems@xcite , such as the problem of first passage times or the problem of last return to the starting point , for which probability distributions can be computed exactly by using the method of reflections@xcite .
however , the search for the _ longest _ segment of the rw , among all possible starting points , creates a more complicated problem .
some of the results presented in this paper have been briefly reported before@xcite . in this work we present a complete exposition of those results , as well as many new results related to this problem and its generalized version . in section [ secdefin ]
we define the problem accurately and argue that in the large @xmath1 limit it can be described in terms of a probability density @xmath27 , where @xmath28 and @xmath29 are the reduced length and charge , respectively .
this probability density is investigated using monte carlo ( mc ) and exact enumeration methods , as well as by analytical arguments .
in particular , we show that the function @xmath30 has an essential singularity in the @xmath31 limit , and diverges as @xmath32 in the limit @xmath33 . these properties can be easily understood from qualitative arguments presented in section [ secapprox ] .
in section [ analytic ] we construct an exact integral expression which enables an analytic investigation of certain properties of @xmath27 . in section [ related ]
we show that our problem is related to a different problem of two random walkers ( which we call the `` staircase problem '' ) .
this relation enables us to use the latter problem to investigate the behavior of @xmath30 in the limit @xmath33 .
while some of the properties of @xmath27 can be deduced analytically , we had to complement our results by mc and exact enumeration studies , which appear in almost every section of the paper along with analytical arguments on the subject .
an additional insight into the problem can be gained by considering its generalization to @xmath7dimensional rws .
( this generalization is _ not _ related to the original problem of pas or to their embedding dimension . ) in this generalization , which is described in sec .
[ highd ] , @xmath34 is treated as a @xmath7dimensional vector rather than a scalar .
as in the one dimensional case , @xmath35 corresponds to a loop in the rw .
since the generalized problem investigates the presence of large loops , it is somewhat related to the problem of self avoiding walks@xcite , whose behavior is also controlled by self intersections ( i.e. loops ) . in particular , the probability distribution of the @xmath36segments becomes trivial for @xmath37 , when large loops are virtually absent .
in this section , we present an exact definition of the problem of extremal segments of a one dimensional sequence and review the qualitative features of the resulting probability distributions .
consider the set @xmath38 which contains all @xmath1element sequences @xmath24 ( @xmath39 ) .
here , @xmath40 physically corresponds to the charge ( positive or negative ) on the @xmath41th monomer of the @xmath1-mer .
alternatively , it can be thought of as the direction of the @xmath41th step of an @xmath1step one dimensional rw . a randomly charged polymer ( or , alternatively , rw )
can then be represented as a random sequence ( rs ) @xmath42 picked with equal probability @xmath43 .
[ randomwalk ] depicts an example of such a sequence and the corresponding path , where the position @xmath44 of the path at index @xmath41 gives the accumulated charge from the beginning of the polymer till the @xmath41th monomer .
( @xmath45 . ) in the language of the rws , @xmath46 is simply the displacement of the walk from the origin after @xmath41 steps .
every segment of the sequence between , say , steps @xmath41 and @xmath47 , has a certain charge @xmath48 .
a segment for which @xmath49 will be called a @xmath0segment . given a randomly chosen sequence @xmath42 and a charge @xmath0 ,
let @xmath26 denote the probability that the _ largest _ @xmath0segment in @xmath50 has length @xmath2 .
it should be stressed that the definition refers to the largest @xmath0segment among many possible @xmath0segments with different starting points which may exist in @xmath50 .
for example , the dotted lines in fig .
[ randomwalk ] indicate the longest 0segments ( @xmath51 ) and the dot dashed lines show the longest 4segments ( @xmath52 ) in a sequence with @xmath53@xcite .
clearly , the longest @xmath0segment does not have to be unique .
if there is at least one @xmath0segment in the sequence then its length @xmath2 satisfies @xmath54 . from the definitions is clear that the 0segment is always present and therefore @xmath55 .
however , the set of @xmath0segments in a given sequence may be empty for @xmath56 : for example , the sequence shown in fig .
[ randomwalk ] has no 8segments .
thus , @xmath57 for @xmath56 .
most properties of rss have simple continuum limits .
we demonstrate this in sections [ secapprox ] and [ related ] by discussing rw problems that are exactly solvable , and relating them to the behavior of @xmath26 in certain limits .
thus , we also expect @xmath26 to approach a similar scaling form when @xmath58 , while the reduced length @xmath28 and the reduced charge @xmath29 are kept constant . in this continuum limit
, it is more convenient to work with the _ probability density _ @xmath59\ .\ ] ] of course , for small @xmath1 , this definition of @xmath27 will still depend on @xmath1 .
we expect it to become a function only of the reduced variables in the @xmath58 limit
. note that at least one term the the square brackets of eq .
( [ eqdefp ] ) vanishes since @xmath60 for odd @xmath61 . to prevent even odd oscillations ,
we included two terms in the definition of @xmath62 , as in the definitions which are used in continuum limits of discrete rws .
we have initially examined the behavior of @xmath26 using numerical ( exact enumeration and monte carlo ) methods , details of which are given in the appendix .
monte carlo results obtained for a variety of large @xmath1s up to @xmath63 were virtually indistinguishable from each other when plotted in the properly scaled variables .
the results for @xmath64 are depicted as a solid curve in each one of the graphs in fig .
[ pl0split ] . for that particular value of @xmath1
we evaluated the probability density from @xmath65 randomly selected sequences . for short chains ( up to @xmath66 )
it was possible to perform a complete enumeration and get the exact results for @xmath26 .
when these exact results are plotted in the scaled form , as presented in fig .
[ pl0split ] , we can see that even for such modest values of @xmath1 , there is an extremely fast convergence to the continuum distribution @xmath30 , depicted by the solid curve ( especially for @xmath67 ) .
the probability density @xmath30 shown in fig .
[ pl0split ] has several remarkable properties : \(a ) mc results show that @xmath62 at @xmath68 is very close to unity ( @xmath69 ) . at that point
the slope of the curve changes by an order of magnitude .
while it is impossible to ascertain from the numerical results that there is actually a discontinuity in the first derivative of @xmath30 with respect to @xmath70 , both the mc results and analytical arguments indicate that @xmath71 is a very special point of the curve .
\(b ) for @xmath31 , the function exhibits an essential singularity of the form @xmath72 , where @xmath73 and @xmath74 2 .
the estimates of the coefficient @xmath75 and of the exponent @xmath76 have been obtained from the mc data . however , in the @xmath31 limit we are dealing with almost vanishing probabilities , and therefore the statistical accuracy is small .
thus the estimates depend on the precise range of @xmath70s for which the fit is performed .
nevertheless , the existence of the singularity can be easily understood from the fact that for small @xmath70 the absence of large loops in the entire chain can be though of as requirement that such loops are absent in many separate and independent segments of the sequence . in the section [ secapprox ]
this argument will be discussed in detail .
\(c ) for @xmath33 , @xmath30 diverges as @xmath77 , with @xmath78 .
this estimate of the constant @xmath79 has been obtained from mc results for @xmath64 sequence . in section [ analytic ]
we prove the existence of the square
root singularity from an integral relation which is derived for @xmath27 .
the proof , however , does not provide a value for the prefactor @xmath79 , and we are limited to mc estimates , as well as results extracted from exact enumeration studies which will be presented in sec .
[ analytic ] .
( the accumulated evidence of mc and exact enumeration shows that @xmath79 is definitely larger than 1 . )
some more intuitive , although less rigorous , results regarding the @xmath31 and @xmath33 limits are presented in section [ secapprox ] .
the exact enumeration results depicted in fig . [ pl0split ] are not suitable for extraction of asymptotic behavior , since the @xmath1s are too small . in the section [ analytic ]
we show that it is possible to exactly calculate @xmath80 for small @xmath81 ( i.e. @xmath82 ) and _ arbitrary _ sequence length @xmath1 . in principle , the correct behavior of @xmath30 in the @xmath33 limit can be deduced from the exact values of @xmath83 only if the limit @xmath84 ( while keeping @xmath85 constant ) is taken before the @xmath86 limit .
somewhat surprisingly , if we attempt to match the asymptotic form of @xmath30 near @xmath6 with @xmath87 for @xmath88 , we find @xmath89 , i.e. we reproduce almost the exact value of the prefactor .
thus , the discrete distribution approaches its asymptotic ( continuum ) form within a few steps of the extreme @xmath90 .
consider next the full probability density @xmath27 , which is depicted in fig .
introduction of an additional variable @xmath91 significantly increased the cpu time needed to analyze a single rs .
the mc data in this figure represent only @xmath92 sequences of length @xmath93 , i.e. its accuracy is smaller than the mc results depicted by the solid line in fig .
[ pl0split ] .
[ plq ] demonstrates further peculiarities of @xmath27 : for fixed @xmath70 , the @xmath91-dependence of @xmath62 is qualitatively different for @xmath94 and @xmath95 : when @xmath94 , the distribution has a single peak at @xmath96 , which approaches a gaussian shape as @xmath70 increases , while for @xmath97 we see a minimum at @xmath96 and two peaks symmetrically located around the minimum .
while qualitatively such behavior can be easily understood ( e.g. for small @xmath70 the 0segments are very unprobable , since they are typically large , and consequently the maximum must be reached for non zero value of @xmath91 ) the transition between the @xmath97 and @xmath98 regions is rather sharp : we analyzed the @xmath91dependence of the graphs representing the fixed@xmath70 sections of fig .
[ plq ] and concluded that the transition from single maximum to a minimum surrounded by two maxima can not be obtained by a variation of parameters in a simple function ( the way it is done in the mean field description of a phase transition near the critical temperature ) .
the numerical data creates an impression of two different functions glued along @xmath68 . the areas @xmath99 under fixed-@xmath70 sections are shown in fig .
[ intplq ] .
for @xmath98 it will be proven in sec .
[ analytic ] that @xmath100 ; fig .
[ intplq]b demonstrates the numerical validity of this relation @xmath101 remains approximately constant in the range of validity .
the accuracy of small @xmath70 regime is rather low ; we only note that @xmath102 is approximately linear in @xmath70 for @xmath103 , as can be seen from fig .
[ intplq]a .
in this section we present approximate derivations of several features of @xmath27 . despite the approximate nature of the arguments , they are rather intuitive , and will be useful when we generalize the problem to @xmath7dimensional rws .
most properties of rws have simple continuum limits . as an example , let us consider the special case @xmath90 of our probability distribution : the probability @xmath104 that the largest @xmath0segment has length @xmath1 is simply equal to the probability that the overall charge @xmath9 of the rs is equal to @xmath0 .
this probability ( for even @xmath105 ) is given by @xmath106![(n+q_o)/2 ] ! } \\ & \displaystyle{=}\atop{n\to\infty } & \quad \sqrt{\frac{2}{\pi n}}\exp(-q_o^2/2n ) .
\nonumber\end{aligned}\ ] ] consider a restricted subset of all rss in @xmath38 which consists only of sequences with _ total _
charge @xmath9 .
the _ conditional _
probability for the largest @xmath0segment in a sequence selected from this subset to have length @xmath2 will be denoted as @xmath107 .
this probability is related to @xmath26 by the relation @xmath108 in the case of @xmath109 , i.e. for 0segments , we note that from the definition it follows that the conditional probability is normalized , i.e. @xmath110 .
we further note that as a function of @xmath2 , the conditional probability is expected to be peaked at value which depends on @xmath9 .
let us assume for simplicity that the peak is very narrow , i.e. the length of the largest 0segment is uniquely determined by @xmath9 and can be described by a function @xmath111 .
indeed , when @xmath112 , the longest 0segment typically has @xmath113 , while for very large @xmath9 , the longest 0segment must be short .
thus @xmath111 is a monotonically decreasing function .
this approximation is especially reasonable for the extremes @xmath31 or 1 . in that case ,
@xmath114 , and thus @xmath115 standard scaling arguments suggest that for @xmath116 we can relate @xmath117 , where @xmath13 is of order unity .
this gives @xmath118 , and finally leads to @xmath119 on the other hand , for @xmath120 , the length of the longest 0segment will be of order of a scale at which the random excursion of the rw becomes comparable to the drift produced by @xmath9 , i.e. , when @xmath121 , where @xmath75 is a constant of order unity .
thus , @xmath122 and @xmath123 thus , this simple scaling argument correctly reproduces the square - root divergence for @xmath33 , and the @xmath124 singularity for @xmath31 .
it is useful to consider an alternative derivation of the behavior in @xmath31 limit , since such derivation involves a somewhat different view of the same properties .
a rs with an extremely short 0segment must have a strong imbalance between the charges ( large @xmath9 ) , i.e. resemble a biased random walk .
consider the probability @xmath125 that the largest 0segment in an @xmath1step sequence _ does not exceed _
length @xmath2 .
if @xmath126 , this quantity can be used to estimate @xmath127 for a sequence twice as long : two halves of the sequence of length @xmath128 must be biased walks with the same direction of bias to prevent creation of long loops , which start in one half of sequence and end in the other half .
in addition , loops longer than @xmath2 must be absent from each half of the sequence .
thus , @xmath129 .
this relation is only approximate since it disregards the correlation between the two halves of the sequence close to its middle .
( loops longer than @xmath2 can begin in one half of the sequence and end at the other half ; correction for this effect may introduce an @xmath2dependent prefactor . ) if the continuum limit is well defined , we can express this relation in the form @xmath130 this relation is satisfied by @xmath131 .
the approximation casts serious doubts on the exact value of the power of @xmath70 in the prefactor to the exponential .
note that two different derivations of the behavior of @xmath30 in the @xmath31 limit produced different preexponential powers . as was mentioned in the previous section ,
our mc results are not accurate enough to distinguish between these predictions .
the method of reflections is a standard tool in calculating the behavior of random walkers near reflecting or absorbing walls ( see ref .
it can be used to calculate various seemingly nontrivial probabilities in terms of probabilities that are easily evaluated .
one such result , which is important for the following discussion , is that the probability for an @xmath1step rw to never return to its starting point is equal to the probability that it reaches its starting point _ exactly _ at the @xmath1th step@xcite , i.e. , @xmath132 where @xmath133 was defined in eq .
( [ allcharge ] ) .
this relation permits , for instance , an exact solution to a simplified version of our problem . in the modified problem ,
the largest @xmath0segments are selected among those that _ start from the beginning _ of the rs , rather than all possible starting positions .
this modified probability @xmath134 is given by the probability that the path @xmath50 reaches position @xmath0 at the @xmath2th step , and that it never again passes through position @xmath0 until the @xmath1th step .
using eqs.([allcharge ] ) and ( [ identity ] ) , we obtain the result ( for @xmath1 , @xmath2 and @xmath0 all even or all odd ) @xmath135![(l+q)/2 ] ! } \frac{(n - l)!}{[(n - l)/2]![(n - l)/2]!}\nonumber \\ & \displaystyle{=}\atop{n\to\infty } & \quad \frac{2}{\pi\sqrt{l(n - l)}}\exp(-q^2/2l).\end{aligned}\ ] ] unfortunately , the search for the longest @xmath0segment in the rs among all possible starting points creates a more complicated problem .
however , we similarly expect @xmath134 to approach a scaling form when @xmath58 , while the reduced length @xmath28 and the reduced charge @xmath29 are kept constant . in this continuum limit ,
it the _ probability density _ is defined analogously with @xmath62 : @xmath136 .
$ ] in this limit eq .
( [ pprime ] ) reduces to @xmath137 we intuitively expect @xmath62 and @xmath138 to behave similarly , at least in the @xmath33 limit , and indeed in that limit @xmath138 resembles @xmath62 ( see eq .
( [ plimit ] ) )
the probabilities @xmath26 for different values of @xmath1 , @xmath2 and @xmath0 satisfy an interesting relation , which in the continuum limit becomes an integral expression that relates @xmath27 at arbitrary values of @xmath98 and @xmath91 to the values of @xmath139 .
while such a relation is insufficient to completely determine the function @xmath27 , it suffices to determine some of its important features . in this section
, we derive this relation and explore its consequences .
we first consider the following sets of random sequences , for @xmath140 and arbitrary @xmath0 : @xmath141 @xmath142 is the set of all ( @xmath143)step sequences with _ total _ displacement ( charge ) @xmath0 .
this set has @xmath144 elements , where the function @xmath145 has been defined in eq .
( [ allcharge ] ) .
the set @xmath146 contains all ( @xmath147)step sequences whose largest @xmath0segments are exactly half as long as the whole sequence . by definition , there are @xmath148 such sequences .
finally , @xmath149 is our `` target set '' which consists of all @xmath1step sequences whose largest @xmath0segment has length @xmath2 .
this set contains @xmath150 sequences .
we shall use the sequences from the @xmath79 and @xmath75type sets to construct the sequences of the `` target set '' : it is possible to construct a one - to - one onto mapping @xmath151 i.e. , each sequence in @xmath149 can be uniquely associated with a pair of sequences from @xmath152 and @xmath153 for some value of @xmath154 , and vice versa .
the mapping @xmath155 is schematically shown in fig . [ mapping ] .
basically , the sequence from @xmath153 is inserted into the sequence from @xmath152 _ at its midpoint _ to create a sequence in @xmath149 .
after such an insertion we obtain a sequence of length @xmath156 , which contains a segment of charge @xmath157 of length @xmath158 .
thus we created an @xmath1step sequence with a @xmath0segment of size @xmath2 . from the process of construction it is clear , that this is the _ largest _
@xmath0segment in the sequence : if a larger @xmath0segment had existed in the resulting chain , we could have reversed the process by removing a segment of length @xmath143 from the center of the chain .
this would have yielded a @xmath159 step chain whose largest @xmath160segment was longer than half of its entire length , contradicting the initial assumption regarding the chain from the set @xmath161 .
the `` reversibility '' of the process also proves the one to
one correspondence between the sets .
it should be stressed , however , that this process requires that the midpoint of the resulting @xmath1step sequence is necessarily included in the largest @xmath0segment .
thus , the proof is valid only for @xmath162 .
since @xmath163 and @xmath164 are disjoint when @xmath165 , equating the number of elements in the domain and range of @xmath155 gives the identity @xmath166 taking the continuum limit of the above equation , we replace the probabilities @xmath167 by the probability density @xmath62 , and the discrete probability @xmath145 by its continuum ( gaussian ) form which follows from eq .
( [ allcharge ] ) and obtain @xmath168 where @xmath169 . since the equation is linear in the function @xmath62 , it can not be used to determine proportionality constants .
( since the equation is valid only for @xmath170 , the normalization condition of @xmath62 can not be used either . ) eq .
( [ eintrel ] ) expresses an unknown function in an interval of @xmath70s via the values of the same unknown function at a particular point @xmath68 . despite these limitations , eq .
( [ eintrel ] ) can be utilized to explain some properties of @xmath27 and to extract information using alternative methods , as will be explained below . before proceeding ,
we note , that in the @xmath171 limit the gaussian term in the integrand of eq .
( [ eintrel ] ) ( the exponential term with the prefactor @xmath172 ) becomes @xmath173 , and the integral relation reduces to identity . by integrating both sides of eq.([eintrel ] ) over @xmath91 , we find a relation between the areas @xmath102 , for @xmath174 : @xmath175 which confirms the observation from the mc data that for @xmath98 , @xmath102 is proportional to @xmath176 .
the relation ( [ eqal ] ) provides a method for measuring the otherwise unknown proportionality constant by detailed calculation of probability density at @xmath68 , i.e. measurement of @xmath177 . in the @xmath33 limit
, the variable @xmath178 disappears from the exponent in eq.([eintrel ] ) , and the relation reduces to @xmath179 this relation both confirms our contention that @xmath30 has a square - root divergence @xmath77 with @xmath180 , and demonstrates that the fixed@xmath70 sections of the surface in fig .
[ plq ] approach a pure gaussian shape when @xmath33 . the proportionality coefficient of the square
root divergence @xmath79 is simply related sum over @xmath0 of the probabilities for the largest @xmath0segment to be exactly half of the length of the rs . by complete enumeration
we calculated the probabilities @xmath181 for all @xmath0 and @xmath182 , and formed the sums @xmath183 .
( only even sequence lengths @xmath81 were used . )
the sums @xmath184 converge to @xmath79 in the @xmath185 limit .
[ am ] depicts the sequence of the estimates @xmath184 plotted versus @xmath186 .
the extrapolation to @xmath187 provides an estimate @xmath188 .
this result is consistent with the mc estimates of @xmath79 , and has smaller error bars .
it is interesting to note , that despite the fact that @xmath79 is almost unity , it is definitely larger than 1 .
finally , we note that the discrete relation in eq .
( [ recrel ] ) can be used to produce exact analytical forms for @xmath189 .
consider cases when @xmath190 and @xmath81 is a small number .
( [ recrel ] ) can be rewritten in the terms of @xmath81 as follows : @xmath191 consider a case of , say , @xmath192 .
the function @xmath193 is nonzero only for @xmath194 , and can be easily found for those cases by examining all random sequences of length 4 .
the function @xmath195 is known exactly for _ arbitrary _ values of @xmath1 and @xmath196 .
the sum over @xmath154 is finite it contains only 5 terms , and therefore can be performed . as a result we can find an exact expression for @xmath197 for an arbitrary value of @xmath1 .
a similar procedure can be performed for @xmath198 .
thus , for _ arbitrarily large _ ( even ) @xmath1 we get @xmath199 ^ 2}\ , \\ p_n(n-4,0)&=&2^{1-n } { ( n-8)!\over\left[\left({n-8\over2}\right)!\right]^2 } \cdot { 91n^2 - 1186n+3576\over(n-4)(n-6)}\ .\end{aligned}\ ] ] unfortunately , the expressions become increasingly complex with increasing @xmath81 , and it is not possible to use this method to determine the continuum limit of @xmath27
. we did not find analogous integral relations for @xmath97 . here
, the situation is complicated by the fact that , in a given sequence , there may be several longest @xmath0-segments that are disjoint .
in this section , we define a new problem in the theory of random walks , related to two simultaneous walkers , and analyze it detail .
we derive the relation between this problem , and the problem of extremal segments , and use this relation to investigate the properties of @xmath30 in the @xmath33 limit .
consider a random sequence ( walk ) @xmath200
. it can be graphically represented by a plot of @xmath46 versus @xmath41 , where @xmath46 represents the total displacement of the @xmath41th step from the origin of the walk .
let us define the following variables : @xmath201 the variables @xmath202 and @xmath203 represent the maximal and minimal coordinates achieved by the random walker up to ( and including ) @xmath41th step . in fig .
[ definitions]a , the dot dashed and dotted lines depict @xmath202 and @xmath203 , respectively , corresponding to a rs @xmath50 shown above the graph .
( the corresponding @xmath46 is depicted by the solid line . )
the variable @xmath202 ( @xmath203 ) is a monotonic non decreasing ( non increasing ) function of @xmath41 which graphically looks like an ascending ( descending ) staircase .
one can also view @xmath202 and @xmath203 as two walls that contain the entire rw .
initially the walls are located at @xmath204 , and they gradually separate from each other : whenever the random walker inside reaches a wall and performs an additional step in the direction of the wall , it pushes the wall to a new position thus increasing the distance between the walls .
consider two rss , @xmath205 and @xmath206 , selected from @xmath38 .
we are interested in the probability @xmath207 that the path @xmath206 remains _ above _ the maximum point of @xmath205 that far , for the first @xmath2 steps .
the dotted line in fig . [ 3walks ] depicts the rs @xmath205 , which generates the staircase ( solid line ) that the rs @xmath206 is supposed to remain entirely above of .
we denote the determination of @xmath208 as the staircase problem . "
the dot dashed line in fig .
[ 3walks]a depicts a permitted @xmath206 , while dot dashed lines in figs . [ 3walks]b and [ 3walks]c show examples of forbidden cases .
( analogously , one can define of problem of rw staying below @xmath203 , and a problem of rw staying either above @xmath202 or below @xmath203 , i.e. staying outside the walls pushed by the rs @xmath205 . )
every step of the staircase begins when the rs @xmath205 arrives to that particular maximal value of @xmath209 for the first time .
the step ends when the sequence exceeds that value for the first time .
the sizes of the these steps are independent of each other , and their distribution is given by the first arrival time to index 1 , i.e. , prob(size of a step@xmath210 .
( for a general expression of first arrival times see ref .
this probability is normalizeable , but the _ mean _ step size is divergent .
the probability @xmath208 of @xmath206 staying above the staircase after @xmath2 steps decreases with increasing @xmath2 .
it is easy to put loose upper and lower bounds to @xmath208 : ( i ) @xmath206 needs to remain above the origin up to the @xmath2th step , since @xmath211 .
thus , @xmath208 decays faster than @xmath212 , which is the asymptotic behavior of the probability of never returning to the origin given by eq .
( [ identity ] ) .
( ii ) the condition is satisfied if @xmath205 remains completely below the origin and @xmath206 remains above the origin up to step @xmath2 . therefore , @xmath208 decays slower than @xmath213 .
given these bounds , it is reasonable to expect an asymptotic power law for @xmath208 : @xmath214 where @xmath215 .
we will later argue that @xmath216 .
we performed a mc investigation of the staircase problem for @xmath2 ranging from 10 to 40960 and sample sizes of about @xmath217 ( yielding approximately @xmath218 survival events for each @xmath2 ) , and confirmed this particular value of @xmath219 to within one percent .
[ cphi ] shows @xmath220 as a function of @xmath221 . the fact that this combination remains independent of @xmath2 when @xmath222 demonstrates the assumed power law .
the points on the graph provide successive estimates of the prefactor @xmath223 ; the error bars indicate statistical uncertainties ( one standard deviation ) for each @xmath2 .
we estimate the asymptotic value of the coefficient as @xmath224 . a very closely related probability distribution is @xmath225 i.e.
, this time the two paths are allowed to meet at positions where @xmath205 has reached a new maximum .
[ 3walks]a and [ 3walks]c both correspond to the permitted events in the definition of @xmath226 .
now let @xmath227 denote the probability of such a meeting occurring for the first time at step l. meeting at the @xmath2th step represents an extremely simple event , i.e. , despite the fact that we are considering the behavior of _ two _ random walkers , it is easy to construct all possible cases for short @xmath2 . in fig .
[ fterms ] , the solid and dot dashed lines represent @xmath205 and @xmath206 , respectively , for @xmath2281 , 3 , 4 and 5 .
we can see that there is a single possibility for @xmath229 and five possibilities for @xmath230 .
( the diagram in the bottom right represents 4 different cases ; the dashed lines indicate the alternative segments in both @xmath205 and @xmath206 . )
@xmath231 is simply equal to @xmath232 ( probability of a single diagram ) multiplied by the number of distinct such diagrams .
since @xmath233 is a rapidly converging series , we can easily evaluate the infinite sum @xmath234 to a high accuracy by summing the first few terms .
( the convergence of the infinite series @xmath234 can be easily seen from the fact that it is bounded from above by the probability that @xmath205 is at a maximum when the two rws meet for the first time . )
we can use the probabilities @xmath235 to relate @xmath226 to @xmath208 via the following relation : @xmath236 fast decay of @xmath231 with increasing @xmath2 , allows the replacement of @xmath237 , @xmath238 in eq.([phitophi ] ) by @xmath208 in the @xmath222 limit , leading to @xmath239 the coefficient @xmath240 can be calculated to high accuracy by summing up the series @xmath233 .
we have obtained the value @xmath241 by extrapolating from finite sums of @xmath235 , which we have obtained exactly for @xmath2 up to 18 , and up to @xmath242 using a monte carlo method .
the results are shown in fig .
[ cf ] . finally , we are in a position to discuss the connection of the staircase problem to the problem of our main interest
. for simplicity , let us only consider @xmath87 in the @xmath243 limit and examine all rss with @xmath244 whose largest neutral segments are @xmath2 steps long .
to construct such a sequence , we can start with a neutral segment @xmath245 of size @xmath2 , depicted by a solid line in fig .
[ twowalkers ] .
this segment is completed into the @xmath1step rs by adding pieces to its two ends ( thick dashed and dotted lines in fig . [ twowalkers ] ) , in such a way _ that a larger neutral segment is not created_. in order to avoid overcounting when there is more than one largest neutral segment , we can for example require that the initially selected segment is the _ leftmost _ of all largest segments .
let @xmath246 be the size of the piece @xmath247 added to the rhs of @xmath245 .
( the lhs piece @xmath248 will then have length @xmath249 . ) to avoid creating a larger neutral segment which begins somewhere inside @xmath245 and ends somewhere inside @xmath247 , the sequence @xmath247 must remain above the staircase generated by the successive maxima of @xmath245 , i.e. if the sequence @xmath247 is translated to the beginning of the sequence @xmath245 ( as depicted by the thin dashed line in fig .
[ twowalkers ] ) they must satisfy conditions defined in the staircase problem .
similar restrictions apply to the segment @xmath248 ; however , this time both @xmath245 and @xmath248 should be viewed `` backwards '' ( thin dotted line in fig .
[ twowalkers ] ) .
( formally , for any sequence @xmath50 it is convenient to define a _
conjugate sequence _ @xmath250 , which consists of the elements @xmath40 of @xmath50 written in reverse order , as illustrated in the fig .
[ definitions]b .
the conjugate of a given path can be obtained by rotating the original path by @xmath251 around the axis normal to the plane .
thus , the staircase conditions have to be satisfied between the sequences @xmath252 and @xmath253 . )
since @xmath245 is a neutral segment , its elements are not completely independent , while our original definition of the staircase problem required the presence of two completely random sequences . however , when @xmath254 , the two ends of @xmath245 can be treated approximately as independent rss , and they become completely independent in the @xmath255 ( i.e. , @xmath33 ) limit .
finally , we notice that the above requirements were somewhat over restrictive : we _ are allowed _ to create neutral segments exactly of length @xmath2 between @xmath245 and @xmath247 , and therefore the probability will be described by @xmath226 rather than by @xmath208 .
the segment @xmath248 , however , must satisfy probabilities described by @xmath208 because we initially required that the neutral segment created by @xmath245 is the _ leftmost _ segment in the rs .
this yields @xmath256 the factor 2 coming from rws with @xmath257 .
finally , taking the sum over @xmath246 in the large @xmath1 limit , we obtain ( for even l ) @xmath258^{1-\alpha } } \nonumber \\ & = & \frac{\gamma(\alpha)\gamma(\alpha)}{\gamma(2\alpha ) } \frac{2c_\phi^2c_f}{(n - l)^{1/2 - 2\alpha } } \sqrt\frac{2}{\pi l(n - l)}.\end{aligned}\ ] ] in the above , @xmath259 is the gamma ( factorial ) function .
this result has several remarkable consequences : first of all , this result suggests that @xmath30 has a well - behaved continuum limit only if @xmath260 .
this implies that @xmath261 , a result we have not yet found in the literature .
knowledge of @xmath223 and @xmath240 now enables an independent calculation of the proportionality coefficient @xmath79 through the relation @xmath262 ^ 2 /\gamma(1/2)=1.025\pm0.015 $ ] .
although it is slightly larger and less accurate , this result is consistent with other estimates of @xmath79 .
the fact that @xmath30 has a singularity at @xmath6 is a consequence of the fact that a rw in one dimension returns to its starting position very often .
thus , it is clear that the behavior of @xmath30 depends strongly on the dimensionality of the rw . in order to investigate this ,
we have generalized the original problem to rws on a @xmath7dimensional hypercubic lattice . now
the `` elementary charge '' ( scalar ) of the one dimensional problem is replaced by an elementary step ( vector ) between neighboring sites on that lattice along one of @xmath263 possible directions , and there are @xmath264 possible @xmath1step walks .
( we can not use the analogy with the sequence of charges , anymore . )
the probability distribution @xmath26 can be easily generalized : @xmath265 where @xmath266 is now the @xmath7-dimensional displacement of a segment in the rw , and @xmath267 .
most of the arguments used to explore the features of one dimensional rws can be applied with minor changes to the @xmath7dimensional walks . as an example , let us consider the qualitative derivation of the asymptotic properties of @xmath268 in the @xmath33 limit as derived for the @xmath269 case in sec .
[ secapprox ] : as in the one dimensional case we may assume that the length of the longest loop can be approximately thought of as a function of the _ overall displacement _ @xmath270 ( end to end vector ) of the entire walk . under such assumption
we expect @xmath271 , which is analogous to the one dimensional case , except for the overall charge @xmath9 that is replaced by the _ modulus _ ( length ) of the vector @xmath270 .
the generalization of eq.([eqapproxp ] ) to @xmath7dimensions is @xmath272 where the one
dimensional @xmath273 of eq .
( [ eqapproxp ] ) has been replaced by @xmath274 , which is the probability that the _ length _ of a @xmath7dimensional end to end vector of an @xmath1step rw is @xmath275 . near @xmath276
this probability is proportional to @xmath277 .
substituting , this expression to eq.([eqapproxpd ] ) and using the relation between @xmath2 and @xmath275 we find @xmath278 .
thus , we expect the probability density to approach a constant in the @xmath33 limit in @xmath279 , and to decay to zero as @xmath280 in @xmath281 .
the relations which have been demonstrated from an approximate argument above can be proven exactly by generalizing eq.([eintrel ] ) to @xmath7 dimensions .
the generalization is straightforward and leads to the form @xmath282 thus , for the @xmath86 limit we obtain @xmath283 fig .
[ alldimens ] depicts @xmath284 for @xmath285 and 3 , obtained from mc simulations @xmath64 with samples of @xmath65 , @xmath286 and @xmath286 rws , respectively .
the peak of the distribution shifts towards @xmath4 as the dimensionality is increased .
[ alldimens ] also demonstrates the verification of the form @xmath287 for these dimensions .
the asymptotic relation described above assumes that @xmath288 does not vanish .
note that @xmath289 for each sequence @xmath50 , there are at most @xmath1 nonzero terms in the summation over * q * , and @xmath290 since it is a probability .
thus , @xmath291 .
this implies that in dimensions higher than 4 , @xmath292 for @xmath293 .
it is easy to understand why @xmath294 is a special dimension in this problem : it is known from the study of the self avoiding random walks@xcite that large loops are absent in space dimensions @xmath37 .
thus we expect that in terms of the reduced variable @xmath70 , all loops will have reduced `` length '' @xmath4 , i.e. , @xmath295 in this regime .
while we expect an asymptotic probability density @xmath296 for @xmath37 , it should be noted that for _ finite _ @xmath1 the probability @xmath297 is a monotonically increasing function of @xmath2 for small values of @xmath2 . therefore , the probability density @xmath268 measured for finite @xmath1 will have a peak at finite ( small ) value of @xmath70 . as @xmath1 increases the entire distribution should drift towards @xmath4 .
[ dimhigh ] depicts such a trend for @xmath298 . a convenient measure of such behavior is calculation of value of @xmath2 such that most of the statistical weight corresponds to loops shorter than the threshold value .
we verified the approach of the distribution to a @xmath299function by examining the finite size scaling of the 90% threshold @xmath300 , defined through @xmath301 .
this means that 10% of the time , there is a loop larger than @xmath300 in a @xmath7dimensional rw of size @xmath1 .
we examined the cases @xmath298 , 6 , and 7 using the mc method described in the appendix , for @xmath1 ranging from 14 to 896 and sample sizes of @xmath286 .
the threshold lengths are also shown in fig .
[ dimhigh ] .
we find that @xmath302 , where @xmath303 and @xmath304 .
since the exponents @xmath305 are positive , the threshold in the terms of the reduced variable @xmath306 vanishes with increasing @xmath1 .
the above arguments do not provide a definite answer for the borderline dimension of @xmath294 .
( the reader is reminded that the self avoiding walk problem at the critical dimension @xmath294 slightly differs from the @xmath37 cases : e.g. , regular power laws are modified by logarithmic corrections . ) through mc calculations with up to @xmath92 rw samples and values of @xmath1 up to 5000 , we find that in @xmath294 the _ entire _ distribution @xmath307 can be fitted very well to the form @xmath308 , for finite values of @xmath1 .
[ dim4 ] depicts such curve for @xmath64 .
( the sample size is @xmath92 . )
the peak position @xmath309 approaches 0 either logarithmically @xmath310 , or with a very small power of @xmath1 , i.e. @xmath311 where @xmath312 .
thus , the distribution still converges to a delta function in the continuum limit .
although the qualitative behavior of @xmath307 is easily understood , it would be interesting to obtain a quantitative understanding of the distribution , especially at the borderline dimension of four .
the problem of extremal segments originated from the desire to consider a simplified description of the ground states of randomly charged polymers .
we used mc , exact enumeration and analytical techniques to analyze the problem , and our results provide convenient tools for a semi quantitative analysis of the the ground states of pas .
in particular , we show that a `` typical '' rs contains very large neutral segments , i.e. it is possible to construct a ground state from a single very large blob with relatively short ends of the chain dangling outside the blob . besides the original motivation , the problem of extremal segments is interesting in its own right .
it looks like one of the classical problems of random walks and , nevertheless , is highly non trivial , and the results indicate a solution with very rich and unexpected structure .
the problem can be related to other interesting problems of the rws , such as the `` staircase problem . ''
while several features of the problem have been established analytically , we did not find a complete analytical solution of the problem .
we think that such a solution is possible and further attempts of finding it are worthwhile .
generalization of the problem to arbitrary space dimension @xmath7 is not related to the original problem of charged polymers , nevertheless interesting in its own right . the numerical `` proof '' of the continuum limit in our work
was limited to a particular class of rws , in which a unit displacement appears at each step . within that class we presented evidence of a continuum limit where the properly scaled functions become independent of @xmath1
. it would be interesting to perform a numerical test of the `` universality '' of the solution for a broader class of rws .
it may be possible to prove the universality of the continuum limit by attempting to perform a renormalization group like treatment of the problem , i.e. attempting to define the problem in the limit where the rw becomes a true gaussian walk ( walk of idealized brownian particle ) .
this limit , however , is far from being trivial .
in particular the definition of what is called a loop ( i.e. how close two different points of the walk should be located so that the segment will be called a closed loop ) presents a non trivial problem in the continuum limit .
such short distance scale can undergo a non trivial scaling , similarly to the excluded volume parameter in the treatment of self avoiding walks .
a different approach to the question of universality may begin from an expansion of the solution near the dimension @xmath294 , as in the treatment of self avoiding walks .
we would like to thank m. kardar and a. yu . grossberg for helpful discussions .
this work was supported by the us
israel bsf grant no .
9200026 , and by the nsf through grant nos .
dmr9400334 ( at mit s cmse ) and dmr9115491 ( at harvard ) .
in this appendix , we describe the numerical methods used in our study .
all of the algorithms were implemented on a silicon graphics r4000 workstation .
we use two approaches to attack the problem numerically : the first approach is to compute the exact distribution @xmath26 for small values of @xmath1 by considering all possible @xmath1-step walks .
since the computational time increases exponentially with @xmath1 , this method practical only up to @xmath313 , and we have analyzed rws with up to 36 steps this way .
thus , in order to determine the scaling form @xmath27 , it is necessary to use a random sampling of the set @xmath38 for large values of @xmath1 .
using such a monte carlo ( mc ) procedure , we have investigated rws of up to 1024 steps .
since the @xmath96 case is especially interesting , we have used more efficient algorithms to determine @xmath314 to a higher accuracy . for both the exact enumeration and mc calculations ,
our algorithms require @xmath315 operations to process one sample from @xmath38 for @xmath314 , and @xmath316 operations to process the full probability distribution @xmath317 .
further details on the individual algorithms , as well as the algorithm used to determine @xmath307 are given below .
the only difference between exact enumeration and mc algorithms involve the number of rws analyzed : in exact enumeration , the number of analyzed rws increases exponentially with @xmath1 , whereas the samples are chosen at random in the mc routines , and the sample size is usually set to a constant .
standard random number generators are used to generate the rws in the mc algorithm . for each rw ,
the size of the largest loop is determined and this is recorded in a histogram ( with sizes from 0 to @xmath2 ) that eventually represents the probability distribution we are looking for .
the determination of the largest neutral segment in a given sequence is identical in both enumeration and mc algorithms , and is described below . given a rw @xmath50 , an array @xmath318 stores the step number @xmath41 when @xmath319 for the first time .
initially , @xmath320 for all @xmath0 . at each step of the rw ( including step 0 ) , the current step number @xmath41 is recorded in @xmath321 if the site is visited for the first time , i.e. if @xmath322 .
if the site was visited earlier , the maximum loop size is replaced by the maximum of itself and the difference @xmath323 . since @xmath324 stores the first time a site is visited
, the largest loop in the walk must correspond to one of such differences .
a finite number of operations are needed for each step , therefore this part of the algorithm involves @xmath315 operations .
the selection of rws ( enumeration or mc ) and the creation of the histogram are also straightforward for this more general problem .
the main task is to find an efficient algorithm that produces the sizes of largest @xmath0segments ( for all @xmath0 ) in a given sequence @xmath50 .
a straightforward generalization of the algorithm for @xmath30 would have required @xmath325 operations per sequence .
however , our algorithm takes advantage of the fact that the same positions are visited many times , and it requires only @xmath316 operations instead . as usual , the algorithm traces the sequence one by one .
there are two main arrays . at a given step @xmath41 ,
one of them keeps track of the sizes of largest @xmath0segments encountered that far .
the second array is actually a dynamically allocated list of pairs of integers .
each pair in the list stores a charge @xmath91 and size of the largest @xmath91segment that ends at the current step @xmath41 .
the size of this array grows as @xmath326 on the average . at each increment in step size , all pairs in the list are updated by adding the next element in the sequence to @xmath91 and incrementing the corresponding lengths by one .
these lengths are then compared with the corresponding values in the first array , which is updated if the new length is larger .
a new element is added to the list of pairs whenever the walk reaches a position for the first time , a condition that is checked for separately .
all the operations in an update can be accomplished by a single pass through the list of pairs , thus the whole algorithm requires only @xmath316 operations to complete , as mentioned earlier . for the mc determination of @xmath307 at higher dimensions , the @xmath315 algorithm described in sec .
[ alg1 ] requires @xmath327 storage elements for the array @xmath328 , which quickly becomes prohibitive with increasing @xmath7 .
the storage requirement can be reduced to @xmath329 by storing the time series of the position @xmath330 of the rw instead .
however , the simplest algorithms would require @xmath325 operations to find the largest * 0*segment given such a data structure .
note that the typical rw in dimensions @xmath331 does not revisit the same site more than a few times , and therefore the total number of 0segments in a rw should be only of @xmath315 .
we have taken advantage of this fact in order to devise an algorithm that requires only @xmath332 operations to do the job .
the algorithm is as follows : after the position array @xmath333 is formed , its contents [ which are the position vectors @xmath334 are indexed in lexicographical order .
this operation requires only @xmath332 operations , when an efficient sorting algorithm like heapsort@xcite is used .
all 0segments in the sequence start and end at the same position by definition , therefore the two endpoints will be adjacent in the lexicographical index .
going through the index sequentially , it is then possible to determine the largest of the 0segments in only @xmath315 operations .
the extraordinary speedup of this algorithm makes is possible to go up to sample sizes of @xmath286 for 1000step rws in 7 dimensions .
see , e.g. , t. e. creighton , _ proteins : their structure and molecular properties _ , freeman , san francisco ( 1984 ) .
p. g. de gennes , _ scaling concepts in polymer physics _ , cornell univ .
press , ithaca ( 1979 ) .
p.g . higgs and j
joanny , j. chem . phys .
* 94 * , 1543 ( 1991 ) .
y. kantor and m. kardar , europhys .
* 14 * , 421 ( 1991 ) .
y. kantor , h. li and m. kardar , phys .
rev . lett . *
69 * , 61 ( 1992 ) .
y. kantor , m. kardar and h. li , phys .
rev . * e49 * , 1383 ( 1994 ) . x .-
yu , a. tanaka , k. tanaka , and t. tanaka , j. chem .
phys . * 97 * , 7805 ( 1992 ) ; x .- h .
yu , ph.d .
thesis , mit ( 1993 ) ; m. annaka and t. tanaka , nature * 355 * , 430 ( 1992 ) ; m. scouri , j. p. munch , s. f. candau , s. neyret , and f. candau , macromol .
* 27 * , 69 ( 1994 ) .
y. kantor and m. kardar , europhys .
* 27 * , 643 ( 1994 ) , and phys . rev . *
e51 * , 1299 ( 1995 ) .
y. kantor and m. kardar , phys . rev .
* e52 * , in press ( 1995 ) .
s. chandrasekhar , rev .
phys . * 15 * , 1 ( 1943 ) .
w. feller , _ an introduction to probability theory and its applications _ , vol . 1 , 3rd ed . ,
john wiley and sons , new york ( 1968 ) .
y. kantor and d. erta , j. phys .
* 27 * , l907 ( 1994 ) . in the language of rws
, the term `` 0segment '' corresponds to a loop . throughout this paper
we will use the `` language of charged @xmath1-mers '' and the `` language of rws '' interchangeably . ref .
@xcite , p. 76 .
w. h. press _ et al .
_ , _ numerical recipes in c : the art of scientific computing _ , 2nd ed .
, cambridge univ . press , new york ( 1992 ) , p. 336 . | motivated by an investigation of ground state properties of randomly charged polymers , we discuss the size distribution of the largest @xmath0segments ( segments with total charge @xmath0 ) in such @xmath1mers . upon mapping the charge sequence to one dimensional random walks ( rws ) , this corresponds to finding the probability for the largest segment with total displacement @xmath0 in an @xmath1step rw to have length @xmath2 . using analytical , exact enumeration , and monte carlo methods ,
we reveal the complex structure of the probability distribution in the large @xmath1 limit .
in particular , the size of the longest neutral segment has a distribution with a square - root singularity at @xmath3 , an essential singularity at @xmath4 , and a discontinuous derivative at @xmath5 .
the behavior near @xmath6 is related to a another interesting rw problem which we call the staircase problem " .
we also discuss the generalized problem for @xmath7dimensional rws . |
in recent years it has been possible to study in detail the underlying chaotic dynamics of many - particle systems , thus finding interesting connections between their microscopic dynamics and the observed macroscopic behavior .
for example , in the case of static properties , the largest lyapunov exponent ( lle ) , which measures the dynamical instability of phase - space trajectories to infinitesimal perturbations in the initial conditions , is related to the kosterlitz - thoules transition temperature in a system of coupled rotators with nearest - neighbor interactions @xcite , as well as to first @xcite and second - order phase transitions @xcite in particle systems with long - range interactions . for dynamical properties ,
the sudden change in the gradient of the lle against energy , which corresponds to the transition from weak to strong chaos , can be detected in the macroscopic behavior of a brownian particle coupled to the system @xcite .
it has also been related to transport coefficients of fluid systems with continuous potentials in non - equilibrium situations @xcite . in equilibrium
, a relation has also been proposed @xcite , although for this case arguments both against @xcite and in favor @xcite have been advanced .
nevertheless , these last results entail interesting possibilities , since they suggest that dynamical instability is at the origin of macroscopic transport phenomena .
two different methodologies have related the lyapunov spectrum ( ls ) , defined as the sorted set of lyapunov exponents ( les ) which give the exponential rate of expansion or contraction of nearby trajectories along each independent component of the phase space , to the transport coefficients of simple fluids @xcite .
however , they involve setting up non equilibrium simulations or locating special phase space trajectories .
furthermore , neither of these approaches considers the perturbations in phase space underlying the les , which have attracted a lot of attention in recent years .
therefore it would be desirable first to attain a more thorough understanding of the phase - space perturbations associated to the les of the most general type of dynamical models employed to study fluid systems before attempting the construction of a general theory that could relate , with enough confidence , the dynamical instability with the macroscopic behavior as described by the transport coefficients . for some particular many - particle systems , e.g. hard - sphere fluids , the theory of les
is highly developed @xcite .
nevertheless , their most interesting features have been discovered by means of molecular dynamics simulations , which revealed that the slowly - growing and decaying perturbations associated with the non - vanishing les closest to zero are related to , and in some cases may even be represented as , almost exact periodic vector fields coherently spread out over the physical space with well defined wave vectors , together with an almost precise stepwise structure of the considered les @xcite . from their similarity to the classical modes of fluctuating hydrodynamics , these perturbations have been named _ lyapunov modes _ ( lms ) . since their discovery @xcite , much work has been done to understand their origin and possible relevance .
some analytical approaches that have been advanced to understand these small les include random - matrix dynamics @xcite , periodic orbit models @xcite , and kinetic theory @xcite .
all these approaches have met with only a very limited success , since none of them has been able to satisfactorily describe the mode dynamics in all the known situations accessible to simulations .
a more serious drawback is that neither has been capable to extract the transport coefficients from the lms , or to relate them , in a simple way , to hydrodynamic fluctuations . until recently the existence of these lms
could only be verified in one , two and three - dimensional hard - sphere fluids @xcite ; their existence in the case of atomic fluids with soft interactions ( both attractive and repulsive ) remained controversial @xcite , since certain features , such as the step structure in the ls , dissappear in soft - potential systems .
hence , to extend the concept of lms beyond the realm of hard - core systems , a more generic definition has to be adopted , which involves both the appearance of a sharp peak of low wave number in the spatial fourier spectrum of lvs corresponding to the les closest to zero and of a minimum in the average spectral entropy , defined in sec .
[ sec : tiiq ] of the present work , in that same le region . employing this more generic definition convincing evidence of the existence of the
so defined lms was obtained for some soft - potential systems , namely in fluids with a one - dimensional ( 1d ) lennard - jones ( lj ) @xcite , where the aforementioned definition of lms and the spectral analysis techniques needed to detect them were developed , and two - dimensional ( 2d ) weeks - chandler - anderson ( wca ) interaction potentials @xcite .
more recently , the existence of lms has been also verified in lattices of coupled hamiltonian and dissipative maps @xcite , as well as in fermi - pasta - ulam anharmonic oscillator lattices @xcite .
thus all information so far available seems to indicate that the hamiltonian structure , conservation laws , and translational invariance are not necessary conditions for the existence of lms .
nevertheless , since there is no theoretical scheme that can predict the existence of lms , the discovery and characterization of lms has to be done in a case - by - case basis .
for example , the introduction of damping in a system of coupled hamiltonian maps does not destroy the lms , whereas the addition of that same damping to a system of coupled circle maps wipes off the lms @xcite .
therefore , although lms are present in the aforementioned systems , their existence is in no way guaranteed in other situations not studied so far , such as in three - dimensional hamiltonian fluid systems with both attractive and repulsive interactions . in this paper
we perform a study of the dynamical instabilities of a three - dimensional ( 3d ) atomic fluid interacting with the full lj potential under the simulation conditions most frequently encountered in molecular dynamics studies @xcite .
we assess the validity of various dynamical indicators previously proposed in the literature for our particular system .
the main result of our work is that , although the spectral analysis methods proposed in ref.@xcite can be readily applied to our system , the dimensionality greatly enhances the mixing among perturbations that was already present in the case of the 1d lj gas , thus rendering both the detection and characterization of the collective perturbations even more problematic .
this paper is organized as follows : in sec .
[ sec : the - model ] we present a survey of the model , as well as the computational details needed to obtain our results in the phase space of our system as well as a short account of the relevant theory necessary to study the perturbations of the phase space .
[ sec : lyapunov - spectrum ] describes the results for the complete ls , its dependence on the particle density , and provides evidence of the mixing among perturbations . in sec .
[ sec : spatial - structure - of - perturbations ] we show the results of the spectral analysis methods applied to the proposed dynamical indicators . section [ sec : discussion ] is devoted to discuss and analyze the results reported in the previous sections . in the appendix we report some preliminary results for the temporal correlations of the phase space perturbations .
we present our conclusions in sec .
[ sec : conclusions ] .
the lj potential @xmath0 between particles @xmath1 and @xmath2 is given by @xmath3,\label{eq : potlj}\ ] ] where @xmath4 is the distance between particles @xmath1 and @xmath2 ; here @xmath5 and @xmath6 are the lj atom diameter and the strength of the interparticle interaction , respectively . the actual interaction potential employed in our simulations is the spherically truncated and shifted ( sts ) potential , which can be written as @xmath7 where @xmath8 is the cut - off radius at which the potential is truncated in order to save computer time .
it is important to notice that the force derived from the above potential , as well as its derivative , are not continuous at the truncation point @xmath9 .
therefore the following potential , first proposed in ref .
@xcite , can be employed @xmath10 + c_2\left(\frac{r}{r_c}\right)^2 + u_{c } & r\le r_{c}\\ 0 & r > r_{c},\end{array}\right.\label{eq : sf}\ ] ] with @xmath11 $ ] and @xmath12 $ ] .
this potential has a continuous derivative in the truncation point @xmath13 .
nevertheless , for the 1d lj model it has been shown , by adopting the above expression for the potential , that the qualitative behavior of the lms is not greatly affected by the discontinuity of the sts potential at @xmath13 @xcite . in the next section
it will be shown explicitly that this same phenomenology holds for our 3d lj fluid .
the complete hamiltonian of our system is then@xmath14 with @xmath15 being the momenta of the atoms and @xmath16 their corresponding masses . in all our simulations we took the masses of the atoms equal to unity , i.e. @xmath17 , with @xmath18 , and @xmath19 as well .
the hamiltonian is then written in terms of the reduced variables @xmath20 @xmath21 , and @xmath22 .
the initial configuration for all simulations was set up from a fcc lattice on a square cell of sides @xmath23 .
the initial momenta were drawn from a maxwell - boltzmann distribution .
then the @xmath24 equations of motion were numerically integrated by means of the verlet leap - frog algorithm with periodic boundary conditions and the minimum image convention applied in all directions .
a time - step of @xmath25 was used in all simulations , with an equilibration period of @xmath26 steps at constant temperature obtained by a uniform rescaling of the velocities at each time - step .
after the equilibration period the system was allowed to evolve at constant total energy for a period of @xmath27 time - steps . the relative energy drift for this number of time - steps is @xmath28@xmath29 , which is an acceptable compromise between accuracy and speed , since there is no systematic drift and thus the energy fluctuations are stable for the chosen time - step .
all the reported results were obtained for systems of @xmath30 at a supercritical reduced temperature of @xmath31 , where @xmath32 is the boltzmann constant .
the reduced densities @xmath33 were taken within the range @xmath34 $ ] .
the critical temperature and density for the lj fluid with sts potential are @xmath35 and @xmath36 @xcite ; thus the employed values for these variables ensure that the system state is far away from the two - phase region in the @xmath37 vs @xmath38 phase diagram , being a homogeneous fluid . finally ,
since in the rest of the work reduced variables will be exclusively employed , from now on we will drop the asterisk from all symbols without risk of confusion .
the phase space trajectory is represented by the variable @xmath39 , where @xmath40 . to study the local dynamical stability of our system we introduce the _ lyapunov vector _ ( lv ) as @xmath41 , where @xmath42 and @xmath43 representing the _ _
i__th particle contribution to the infinitesimal perturbations of the trajectory @xmath44 along all possible directions ( position and momentum axes ) of the phase space , thus defining the so called _ tangent space_. according to oseledec s multiplicative ergodic theorem @xcite the remote past limit symmetric operator @xmath45^{1/[2(t - t_0)]}$ ] exists for almost every initial condition @xmath46 , where @xmath47 is the fundamental matrix governing the time evolution of the perturbations @xmath48 in tangent space as @xmath49 @xcite .
the set of instantaneous les is defined as @xmath50 , where @xmath51 are the eigenvalues of @xmath52 . the herein employed standard procedure to compute the les consists in periodically reorthonormalizing a set of offset vectors that are time evolved by means of the matrix @xmath47 @xcite .
the time - averaged values of the logarithms of the renormalization factors , i.e. @xmath53 are the les @xmath54 and the set of offset vectors right after the reorthonormalization are the eigenvectors of @xmath52 , which are called _ backward _ lvs .
the equivalence of the les computed by means of these two methods can be proved rigorously @xcite , but the relation between the oseledec eigenvectors and the lvs obtained via the standard method is more subtle .
it is known that the backward lvs converge at an exponential rate to the oseledec eigenvectors for the inverse - time dynamics of the original system @xcite .
the latter are obtained as the eigenvectors of the far future limit operator @xmath55^{1/[2(t_0-t)]}$ ] and are called _ forward _ lvs .
it is to be noted that recently it has been possible to obtain from the intersection of the embedded subspaces spanned by the eigenvectors of @xmath56 ( backward lvs ) and @xmath57 ( forward lvs ) the so called _ characteristic _ lvs , which are independent of the norm and do not form an orthogonal basis @xcite .
a somewhat similar algorithm has also been proposed in ref .
the aforementioned schemes are based on ideas already discussed long ago @xcite , but have only been recently proposed because their implementation is by no means a simple task from a computational point of view , which is why they only have been tested in simple 1d systems .
thus it is not at all clear that their implementation could be feasible in the near future for the case of our system .
furthermore , since the purpose of the present work is to investigate the possible existence of lms in a 3d lj fluid , it is important to recall that the discovery of lms in all studied systems so far has been made employing the backward lvs .
so it seems that the use of the clvs is not essential for the detection of lms .
finally , meaningful results are still obtained by means of the backward lvs obtained from the standard procedure , such as in the recent characterization of lms in a diatomic system @xcite and the discovery of lms in the xy rotator model @xcite ; in both studies the backward lvs were employed .
therefore , from now on we will refer to the numerically computed vectors ( backward lvs ) as the lvs without confusion . in the hamiltonian case ( which we treat here ) , and also in some special homogeneous non - equilibrium situations
@xcite , the les and the corresponding lvs have a symmetry property which makes unnecessary to calculate the whole spectrum . in these cases
the ls thus computed is symmetrical around zero , which means that each le has a partner that is exactly its negative .
this is the so called _ conjugate - pairing rule _ @xcite .
therefore , only @xmath58 linearized equations for the lvs were simultaneously integrated , along with the @xmath24 nonlinear equations for the reference trajectory @xmath59 . the initial lvs @xmath60 in all our simulations consisted in a set of @xmath58 orthogonal vectors with randomly selected components .
as mentioned in the previous section , both the force derived from the sts potential , as well as its derivative , are not continuous at @xmath13 . in order to assess the relevance of this particular feature in the computation of the properties in the tangent space we perform additional simulations employing the potential given by eq .
( [ eq : sf ] ) . in fig .
[ cap : ls - sts - sf ] we present the ls computed from this last potential , as well as that computed by means of the sts potential , for @xmath61 , @xmath62 , and @xmath30 . as can be readily appreciated
there are no significant differences , a result which supports the use of the simpler sts potential .
the most relevant feature of this figure is that , as in the case of the 1d lj gas , in the smallest positive le region there is no evidence of the stepwise structure that signals the appearance of the lms in the case of hard - core systems .
thus it is plausible that the same mechanism that accounts for the absence of the stepwise structure in the ls of the 1d lj gas is also at work in the present model . in fig .
[ cap : probability - density - function ] we present the probability density function for the instantaneous les @xmath63and @xmath64 for @xmath30 and @xmath65 .
it can be observed that the fluctuations around their mean values increase as the lyapunov index @xmath66 increases , i.e. when going from a high to a low value of the le .
it is to be observed that the fluctuations in the @xmath64 value are so great that the tails in the probability density function overlap with those corresponding to @xmath63 .
these large fluctuations are certainly a reason why no stepwise structure can be found in the low-@xmath66 region of the ls .
other dynamical indicators that we will present later on are also affected by this behavior of the ls at large @xmath66 values . in the same figure results for the potential given by eq .
( [ eq : sf ] ) are also presented ; these are quite similar to those obtained with the sts potential .
thus it is confirmed that the aforementioned fluctuations are not a result of the discontinuity of the sts potential at @xmath13 .
therefore this potential will be used in the rest of this work .
other dynamical indicators that we will present later on are also affected by this behavior of the ls at large @xmath66 values . in fig .
[ cap : normalized - ls ] we show the normalized ls for @xmath30 and @xmath61 computed for the employed range of reduced densities .
the values of the scalar length of on edge of the cubic simulation cell , as well as the values of the lle corresponding to each value of the reduced density , are reported in table [ tab : table1 ] . as can be readily appreciated in fig .
[ cap : normalized - ls ] , the ls @xmath54 corresponding to the lowest density @xmath65 can be separated into two regions . in both of them
the ls is a decreasing function of the lyapunov index @xmath66 , but in the former the decrease is more pronounced than in the latter .
this bending of the ls has been observed in a quasi 1d hard - disk gas @xcite as well as in the 1d lj gas @xcite , and has been related to the separation of two time scales . to properly explain this point it is important to remember that each le indicates a time scale given by its inverse , so the ls can be considered as a spectrum of time - scales .
the smallest positive le region of the spectrum is dominated by macroscopic time and length scale behavior . on the other hand , the opposite region of the ls is dominated by short time scale behavior , such as local collision events . as the density increases , the collisions increase both in frequency and in the number of particles involved ; the correlation among them increases , and becomes increasingly difficult to distinguish them individually . thus it is no longer possible to make a separation of time scales , and so the ls does not present the aforementioned bending for @xmath67 . although this explanation is plausible for the ls of our 3d lj fluid , we also point out that the bending depicted in fig .
[ cap : normalized - ls ] for @xmath65 ( the lowest reduced density value employed ) is less pronounced than the corresponding lowest density instance of the ls of the 1d lj gas @xcite .
we interpret this fact as a signature that , even for the lowest density case , the effect of the dimensionality in the microscopic events is to enhance the correlation among them , and thus reduce the separation of time scales that is more evident in the 1d case .
this explanation will be further supported in the next sections .
.[tab : table1 ] length @xmath68 of the cubic simulation cell along with the values of the lle for each reduced density value . [ cols="<,<,<",options="header " , ]
as discussed in sec . [ sec : lyapunov - spectrum ] , the strong fluctuations of the smaller les make the perturbations in tangent space extremely unstable , rendering any coherent structure that may exist in configuration space difficult to detect . to investigate the possibility that these structures ( lms ) exist in the case of the 3d sts lj potential we have to establish a measure of the contribution of a given lv @xmath69 at each point @xmath70 of the configuration space ,
regardless of which particle makes the contribution to the magnitude of the chosen lv .
we have to remember that the _ _ i__th particle contribution to the infinitesimal perturbation @xmath69 consists of spatial and momentum components .
since the lms can be in general considered as goldstone modes resulting from translational invariance in coordinate space @xcite , we will consider only the spatial part of the full perturbation component @xmath71 .
thus , in analogy with the definition of microscopic density fluctuations @xcite , we define the _ spatial lv density _ as @xmath72 which was first introduced in ref .
@xcite and is the function to be studied afterwards . in fig .
[ cap : snapshot - lv - density - components ] we present a snapshot of a single component @xmath73 of the aforementioned spatial density along the @xmath74 axis of our system . since the latter is isotropic , the results are completely analogous to those corresponding to the @xmath75 and @xmath76 components projected along their respective coordinate axes .
we can readily appreciate that the spatial density corresponding to the lle is more localized than that corresponding to the le with @xmath77 , which is the lowest index value corresponding to a lv not related to the space and time translational invariance symmetries of the system and the associated conserved quantities , the total energy and the total momentum @xcite .
a point to be noticed is that the localization of the lle is not so clearly defined and strong as in the 1d lj system @xcite or in hard - disk systems @xcite .
this is another indication that the dimensionality of the system has indeed a strong influence on the tangent - space dynamics .
next , we proceed to consider the spatial fourier transform of @xmath78 , which can be written as @xmath79.\end{aligned}\ ] ] to proceed further we invoke the _
static lv density correlation function _ of ref .
@xcite , defined as @xmath80 which in the case of our 3d system is a second rank tensor . for our isotropic
fluid the cartesian components @xmath81 of @xmath82 can be written in terms of longitudinal @xmath83 and transverse @xmath84 static correlation functions as @xmath85 with @xmath86 . from this last expression
it is immediate to obtain the explicit form of the longitudinal and transverse static correlation functions as @xmath87 to simplify the analysis we will consider a coordinate system such that the wave vector @xmath88 is parallel to the _ z _ axis . then @xmath89.\ ] ] for homogeneous systems with translational invariance the fourier transformed components of a quantity such as the spatial lv density @xmath90 , where @xmath91 , are uncorrelated @xcite .
thus , our simplification in no way destroys essential information .
the spatial fourier spectrum @xmath92 corresponding to each component of the spatial density of @xmath93 can be readily computed by an algorithm for unequally - spaced data points @xcite which has been previously applied to the 1d lj @xcite and the 2d wca systems @xcite .
furthermore , we observe that the diagonal components of the static lv density correlation function correspond to the spatial fourier spectrum , i.e. @xmath94 . thus the static longitudinal and transverse correlation functions can be obtained from the aforementioned power spectra as @xmath95 finally , since averaging over several spatially equivalent directions will improve the statistics , we take successively the @xmath88 vector along the remaining coordinate axes _ x _ and _ y _ to obtain two more sets of longitudinal and transverse correlation functions and then average over all sets .
the result are longitudinal @xmath96 and transverse @xmath97 correlation functions independent of the employed coordinate system . at variance with the hard - core systems in which the patterns resembling transverse modes do not survive time averaging @xcite , the fourier spectral techniques so far presented have been quite successful in the case of the 1d lj system , and so there is a reasonable possibility of success in the case of our 3d lj fluid . due to their mutual interaction , the lms in all soft potential systems are only of finite life - time . to investigate their time stability we present in fig .
[ cap : time - evolution - of - peak wavenumber ] the time evolution of two quantities associated with the longitudinal static correlation function @xmath96 for the lv @xmath77 : the peak wave number @xmath98 , which indicates the position of the highest peak in the spatial fourier spectrum @xmath99 , and the _ spectral entropy _
@xcite , which measures the @xmath101 distribution properties of the aforementioned spectrum .
this last quantity was first employed in ref .
@xcite in the study of the 1d lj gas and is defined as @xmath102 for the 1d lj system these quantities show an intermittent behavior , i.e. large intervals of nearly constant low values ( _ off state _ ) are interrupted by short periods of bursts ( _ on state _ ) where they experience large values @xcite . as can be appreciated in fig .
[ cap : time - evolution - of - peak wavenumber](a ) for the time evolution of the peak wave number @xmath98 , this behavior is somewhat different in our case ; although there is an alternation between the on and off states , the mixing among them is so great that there are no long - lived time intervals for either state , in sharp contrast with the 1d case .
the instant value of the spectral entropy , presented in fig .
[ cap : time - evolution - of - peak wavenumber](b ) , has a similar behavior as that of @xmath98 in the sense that it is not easy to identify a correspondence between its temporal evolution and that of the on and off states .
a virtually identical behavior is obtained for @xmath98 and @xmath103 , but now defined in terms of @xmath97 ( not shown ) , and for all values of the reduced density .
this intermittency in the time evolution of the spatial fourier spectrum of lvs is a typical feature of soft - potential systems .
however , this behavior is greatly increased in comparison to the 1d case .
there are two possible causes for this difference with the 1d results : first , the reduced temperature at which we are working is much higher ( an order of magnitude ) than that employed in ref .
@xcite ; second , the dimensionality of our system is also higher , thus the mixing is easier by this enlargement in the phase space , as already explained .
now , in order to average out temporal fluctuations , and thus extract useful information about the collective modes , from now on we will study the properties of the average spectra @xmath104 and @xmath105 , where @xmath106 means temporal average .
further insight into the temporal dynamics of the lms can be obtained by studying the dynamic lv density correlation function first introduced in ref .
@xcite for the 1d lj gas . in the appendix
some preliminary results of the aforementioned dynamical correlation function applied to our 3d lj fluid will be presented .
figures [ cap : low - density - lt - correlation - functions](a ) and [ cap : low - density - lt - correlation - functions](b ) display , for @xmath65 , @xmath107 and @xmath108 respectively . in both cases
we can readily appreciate that the contribution of the high wave number components is not small .
the instantaneous power spectra are not dominated by a single peak ; rather , several small peaks , which are related to intermediate length scales , are present , which in turn make significant contributions to the overall shape of the time averaged correlation functions for large @xmath101 values .
this feature can be explained by the less pronounced time - scale separation than in the 1d case , as mentioned in sec . [
sec : lyapunov - spectrum ] in relation to the ls .
nevertheless , the highest value of the time averaged correlation functions is always dominated by certain low - wave - number components ; for the longitudinal correlation function we observe that the sharp - valued peak in the spectrum corresponds to diminishing @xmath98 values as the lyapunov index @xmath109 , i.e. as we go from the region of high les to the region of low les . at this point
it is important to notice that this correspondence is not monotonic , since the @xmath98 value of the highest peak is attained at @xmath110 , and the height of the corresponding peak slightly diminishes , although remains well defined , for higher @xmath66 values . for the transverse correlation function depicted in fig .
[ cap : low - density - lt - correlation - functions](b ) the results are similar , except for a very important feature : the highest value of the spectra is again attained at @xmath110 , a value far away of the region corresponding to the lowest les , but then vanishes as @xmath109 .
this feature is not consistent with the existence of well - defined transverse lms , since it would be expected that the highest peak in the transverse correlation function should correspond to small values of @xmath98 in the @xmath111 regime . in the inset
the static structure factor @xmath112 @xcite for the corresponding thermodynamic state is plotted .
the highest peak of this function is located at @xmath113 .
now , the position of the the peak in @xmath107 is located at @xmath114 .
this value suggest that there is no obvious direct connection between the short range order of the atoms and the lowest wavevector peak of the longitudinal lv correlation function .
the longitudinal and transverse correlation functions corresponding to the reduced density @xmath115 are presented in figs .
[ cap : high - density - lt - correlation - functions](a ) and [ cap : high - density - lt - correlation - functions](b ) , respectively .
a difference with respect to the low density results is that the highest peak in each power spectra is broader than in the corresponding low - density case .
this can be interpreted as an indication that the influence of higher wave numbers is stronger than in the low - density case .
we further observe that the highest peak is located at a higher wave number value than the corresponding peaks in figs .
[ cap : low - density - lt - correlation - functions](a ) and ( b ) .
now , although at this density there is a higher degree of spatial order in the atoms of the system , the lowest wavevector peak of @xmath112 , which is plotted in the inset , has no relation whatsoever with the @xmath98 value of the time averaged transverse lv correlation function .
next we notice that the highest peak of the transverse correlation function is reached at a value of the lyapunov index of @xmath110 , with a monotonic decrease for higher @xmath66 values .
this same result was obtained for the transverse correlation function in the low - density regime , although in this case the peak remains well defined , in contrast with the result displayed in fig . [ cap : low - density - lt - correlation - functions](b ) , which is almost a flat spectrum .
so far our results are consistent with the existence of longitudinal lms . to quantify the properties of these modes in fig .
[ cap : low - density - l - kmaxcmaxh](a ) we plot @xmath98 versus the lyapunov index @xmath66 .
it is clear from this figure that , as @xmath109 , @xmath98 diminishes .
a feature that stands out is that , despite the time averaging over a large number of data points ( @xmath27 time steps ) , the obtained values remain somewhat noisy .
nevertheless it is clear from the figure that the decrease in the @xmath98 value as @xmath66 increases is approximately monotonic , in clear contrast to the 1d case @xcite in which a sudden change of @xmath98 from a finite value to zero was interpreted as a signature of the separation of time scales mentioned in sec .
[ sec : lyapunov - spectrum ] .
next , in fig .
[ cap : low - density - l - kmaxcmaxh](b ) the height @xmath116 of the highest peak in the time averaged longitudinal correlation function is also plotted as function of the lyapunov index @xmath66 . on average
we can observe a monotonic increase of the peak value as the lyapunov index @xmath66 goes from small to large values , although a small decrease can be noticed in the @xmath111 region .
finally , from the definition given in eq .
( [ eq : sentro ] ) , it is immediate to obtain the average spectral entropy @xmath117 which is presented , again as function of @xmath66 , in fig . [ cap : low - density - l - kmaxcmaxh](c ) .
its value decreases as the lyapunov index increases .
this in turn means that lvs corresponding to smaller positive les are more localized in fourier space , i.e. they have more wave - like character than those corresponding to larger les .
however , the decrease in the value of @xmath117 as the lyapunov index increases is not monotonic .
we notice that already for @xmath110 the average spectral entropy has reached its minimum value , but presents a slight increase as @xmath109 , which certainly accounts for the decrease in the height of the peak in the average spectra of fig .
[ cap : low - density - lt - correlation - functions](a ) .
our tentative conclusion at this point is that the longitudinal lms are more vaguely defined than in the 1d case for this density value .
the results corresponding to the transverse correlation function @xmath108 are presented in fig .
[ cap : low - density - t - kmaxcmaxh ] .
we observe that the @xmath66 range for which @xmath98 has a small value is broader compared to the results in fig [ cap : low - density - l - kmaxcmaxh](a ) .
however , @xmath98 has a slight increase in value as @xmath111 , a result which is not consistent with the existence of a transverse lyapunov mode for this @xmath66 range .
this conjecture is supported by the behavior of @xmath118 displayed in fig .
[ cap : low - density - t - kmaxcmaxh](b ) .
this last quantity has its maximum at @xmath119 , with a corresponding minimum of @xmath120 at the same @xmath66 value , as seen in fig .
[ cap : low - density - t - kmaxcmaxh](c ) .
this feature of the spectral entropy is inconsistent with localization in fourier space ; that is , the wave - like character of the lvs is largely diminished in the low @xmath66 region . taken together these results
make us difficult to unambiguously ascertain the existence of transverse lms in the 3d lj system . for the case of high reduced density @xmath115 fig .
[ cap : high - density - l - kmaxcmaxh](a ) presents the results for @xmath98 versus the lyapunov index @xmath66 corresponding to the longitudinal correlation function @xmath107 .
the behavior displayed is very different to that shown in fig .
[ cap : low - density - l - kmaxcmaxh](a ) .
we first notice that the fluctuations of all quantities are much smaller than those corresponding to the low - density value . in the present case
the value of @xmath98 stays close to 1 for @xmath121 .
then , as @xmath66 further increases and after a small transient interval , drops rather sharply to a value slightly lower than @xmath122 , which is consistent with that obtained from fig .
[ cap : high - density - lt - correlation - functions](a ) .
next , in fig .
[ cap : high - density - l - kmaxcmaxh](b ) we observe that the value of @xmath116 decreases smoothly from its maximum at @xmath111 down to its minimum at @xmath123 . finally , in fig .
[ cap : high - density - l - kmaxcmaxh](c ) we observe that , at variance with the low - density result , the value of the average spectral entropy @xmath124 decreases monotonically in the whole range of @xmath66 values , with a minimum for @xmath111 where the les are the smallest possible ones . from the definition of the spectral entropy
we conclude that the corresponding spectra for these lvs are most significantly dominated by a few @xmath101 values .
thus the lms are more sharply defined than in the low density case : a very intriguing state of affairs , since no relation was found between the spatial order of the atoms as described by @xmath112 and the highest peak in the lv power spectrum at any density value .
the results for the transverse correlation function @xmath105 for the highest density studied are presented in fig .
[ cap : high - density - t - kmaxcmaxh ] , which display a much reduced fluctuation level , but have nevertheless a very similar behavior , compared to those in fig .
[ cap : low - density - t - kmaxcmaxh ] corresponding to the low - density case .
that is , the maximum value of @xmath125 , which is coincident with the minimum value of @xmath126 , is present in a range of @xmath66 values that is far from the region in which @xmath111 , a result not entirely consistent with the existence of transverse lms .
the overall picture that emerges from our results so far indicates a tangent space dynamics much more complicated than that of hard - core systems or the 1d lj fluid .
the strong fluctuations in the les mentioned in sec .
[ sec : lyapunov - spectrum ] produce a strong mixture among modes , so it is unreasonable to expect that results of the spectral analysis could convey information concerning pure modes .
indeed , the maximum wave number @xmath98 for both the longitudinal and transverse correlation functions is highly unstable . in the time interval depicted in fig .
[ cap : time - evolution - of - peak wavenumber ] for the longitudinal case the @xmath98 value very rarely stays close to zero .
rather , it seems to wander randomly between zero and one .
the instantaneous value of the spectral entropy , eq .
( [ eq : sentro ] ) , also seems to have a seemingly random time evolution .
finally , there seems to be no correlation between the time evolution of @xmath98 and @xmath100 , which was discovered in the case of the 1d lj system and which greatly contributed to a clear - cut definition of the so called on and off states @xcite .
this is a first indication that the detection of lms becomes more difficult than in the corresponding 1d case .
it turns out that the time averaged longitudinal and transverse correlation functions are the relevant variables from which meaningful information about the lms can be obtained .
this fact can be understood in terms of the zwanzig - mori formalism , in which the fourier components of the fluctuation of a conserved density vary slowly for a small wave number @xcite ; from these `` slow '' variables a meaningful description is then extracted .
for the tangent - space dynamics the time averaged longitudinal and transverse correlation functions can be considered as the `` slow '' variables .
the results presented in figs . [ cap : low - density - lt - correlation - functions ] and [ cap : high - density - lt - correlation - functions ] show that this is indeed the case ; the spectra are dominated by low wave number values , i.e. @xmath114 for @xmath65 and @xmath127 for @xmath115 , both for longitudinal and transverse correlation functions , with a corresponding broader peak in the latter case . in both cases
@xmath128 , which is the smallest nonzero wave number allowed by the periodic boundary conditions used ( smaller @xmath101 values are due to the oversampling inherent to the method , see ref .
another point to be noted is that , for our particular system , the longitudinal correlation function reaches its maximum value at @xmath119 , and then its height diminishes slightly , but with the peak position @xmath98 unchanged , as @xmath111 .
these results can be attributed to a complicate mixing of pure modes in the low @xmath101 regime , which produces the observed degeneracy of the @xmath101 value with respect to the @xmath66 index .
thus the obtained lms lose their hydrodynamic character and no dispersion relation @xmath98 vs @xmath129 as those observed in the 1d lj fluid @xcite and hard - core systems @xcite could be detected .
our results even stand in contrast with those of 2d coupled map lattices , for which a dispersion relation @xmath130 indeed exist @xcite , and even more sharply with those of the 2d lj fluid , for which a corresponding dispersion relation has been claimed to hold , although for this last model the evidence seems so far to be inconclusive @xcite .
the most convincing evidence of the existence of longitudinal lms for the low density state comes from the results of fig .
[ cap : low - density - l - kmaxcmaxh ] .
first we observe that the average spectral entropy attains its minimum at @xmath119 , a result not entirely inconsistent with those of random matrix theory @xcite .
next , as @xmath131 , the value of this quantity remains close to this minimum .
taken together with the already obtained position of the highest peak in the power spectrum in the region @xmath111 depicted in fig .
[ cap : low - density - lt - correlation - functions](a ) , these results are compelling evidence of the existence of longitudinal lms .
the fuzziness of the obtained values of the reported quantities also led us to suppose that the mixing between modes is strong . up to this point
we can affirm that longitudinal lms do indeed exist , but are more vague than in the 1d lj gas and with no hydrodynamical character at all . on the other hand , our results on the transverse correlation function displayed in fig .
[ cap : low - density - t - kmaxcmaxh ] , despite their fluctuations , show a tendency that does not allow us to unambiguously classify them as transverse lms .
the most important result of this paper was presented in fig .
[ cap : high - density - l - kmaxcmaxh ] for @xmath115 . besides a much reduced fluctuation in the studied variables , a very defined jump in @xmath98 at @xmath123 is observed .
the importance of this fact is that , for the 1d lj gas , a similar jump was observed , but for a low - density value @xcite . in that system
this behavior was interpreted in terms as of a separation of time scales signaled by a bending in the ls . on the contrary
, we obtain a sharply defined jump in @xmath98 at a @xmath132 value in which the corresponding ls shows no bending , as can be observed in fig .
[ cap : normalized - ls ] . in order to try to understand this seemingly puzzling result
, we have to remember that , from the suspected importance of hyperbolicity for the appearance of the lms @xcite , the main results for the 1d lj fluid were obtained in a relatively diluted regime @xcite , which made this system somewhat similar to hard - core systems previously studied @xcite .
however , for @xmath115 our system is far from complete hyperbolicity , since the combination of attractive and repulsive interactions induces strong correlations between the collision events that makes difficult to separate them from each other .
at the same time , the aforementioned density value is not high enough to make the effective interaction among atoms similar to that of a lattice of anharmonic oscillators in which hydrodynamic lms have been detected @xcite .
thus there is no obvious mechanism that could account for the sharp jump in @xmath98 when there is no bending in the ls .
therefore the role of the alleged separation of time scales , inferred from the ls bending for the 1d lj gas , seems to bear no relation to the appearance and phenomenological description of the lms in our 3d lj fluid .
in this paper we have performed a study of the tangent space dynamics of the 3d lj fluid in order to investigate the possible existence of the lms which are a distinctive feature of the hard - core systems .
our results indicate that longitudinal modes indeed exist for low and high reduced density values , and no conclusive evidence of transverse modes for either density studied .
the lack of a dispersion relation between the les and the maximum wave number make both types of modes markedly different from those already encountered in other systems .
the longitudinal lms turn out to be much better defined for high values of the reduced density .
previously only in hyperbolic systems could lms be detected at high density values @xcite .
it is highly plausible that by changing the thermodynamic state of the system the lms could present a different behavior than the one reported in the present communication .
one of the authors ( m.r.b . )
wishes to acknowledge karen haneman and d. castaeda - valle for their comments and suggestions . financial support from consejo nacional de ciencia y tecnologa ( conacyt )
is also acknowledged .
from the fourier transform of the spatial lv density @xmath78 , eq .
( [ eq : dens ] ) , we can define the _ intermediate two - time correlation function _ as @xmath133 which in the 3d case is a second rank tensor with components @xmath134 . since the results of sec .
[ sec : taps ] point clearly to the existence of longitudinal lms , we will concentrate on the longitudinal component of this function . by following the same methodology of sec .
[ sec : slvd ] we are led to an expression for the longitudinal component of the form @xmath135 . taking * k * parallel to the other coordinate directions and averaging we obtain the final form @xmath136 .
for @xmath137 we recover the time average of the static lv correlation function , i.e. @xmath138 . the dynamical lv correlation function encodes structural as well as temporal correlations , and thus provides more detailed information of the system . by fourier transformation with respect to time
we obtain the _ dynamical lv density correlation function _ ( dlvdcf ) as @xmath139 . in fig .
[ cap : low - density - dlvdcf ] we present the result for @xmath65 and @xmath77 for the longitudinal dlvcf at various integer multiples of the wavevector @xmath140 .
it can be observed that , besides the central peak , smaller peaks are present as the frequency @xmath141 increases .
this result implies that the tangent space dynamics is described by a set of characteristic frequencies , which may have their origin in the lack of timescale separation already advanced in sec .
[ sec : lyapunov - spectrum ] .
however , restricting our attention to the lowest frequencies , it is clear that the position @xmath142 of the first peak can be unambiguously defined for the lowest @xmath101 value ; a similar identification can be made for other low @xmath101 values .
thus a dispersion relation @xmath143 for different @xmath66 values can be extracted .
the result is presented in the inset of the same figure .
it is observed that an approximately linear dispersion relation is obtained and thus , besides the characteristic wavevector @xmath144 for each lm already determined in sec .
[ sec : taps ] , each mode can also be characterized by a frequency @xmath145 .
the non - vanishing value of @xmath146 seems to imply propagating wave - like excitations .
the result for the longitudinal dlvdcf at @xmath115 is presented in fig .
[ cap : high - density - dlvdcf ] . in sharp contrast with the low density state
no structure can be observed whatsoever at any wavevector number .
no characteristic frequency can be easily identified , which implies that no propagation of any wave - like structures occurs at any wavevector number .
thus the tangent space structure described in fig .
[ cap : high - density - l - kmaxcmaxh ] remains unaltered within the time scales studied .
more details of the temporal dynamics in tangent space will be presented in a forthcoming study .
10 butera p and caravati g 1987 phys . rev .
a * 36 * 962 torcini a and antoni m 1999 phys .
e * 59 * 2746 firpo m -c 1998 phys .
e * 57 * 6599 romero - bastida m and braun e 2002 phys .
e * 65 * 036228 ; romero - bastida m 2004 _ ibid_. * 69 * 056204 evans d j , cohen e
g d , and morriss g p 1990 phys .
a * 42 * 5990 barnett d m , tajima t , nishihara k , ueshima y and furukawa h 1996 phys .
* 76 * 1812 ; ueshima y , nishihara k , barnett d m , tajima t and furukawa h 1997 phys . rev .
e * 55 * 3439 ( 1997 ) ; barnett d m and tajima t 1996 _ ibid_. * 54 * 6084 torcini a , dellago ch and posch h a 1999 phys .
. lett . * 83 * 2676 barnett d m , tajima t and ueshima y 1999 phys .
* 83 * 2677 evans d j and morriss g p 1990 _ statistical mechanics of non - equilibrium liquids _ ( academic : new york ) gaspard p 1999 _ chaos , scattering , and statistical mechanics _
( cambridge : cambridge university press ) dorfman j r 1999 _ an introduction to chaos in nonequilibrium statistical mechanics _
( cambridge : cambridge university press ) milanovi jj , posch h a and hoover wm h 1998 mol . phys . * 95 * 281 posch h a and hirschl r 2000 _ hard ball systems and the lorentz gas _ ed d. szsz ( springer : berlin ) p. 279 .
eckmann j
-p and gat o 2000 j. stat . phys . * 98 * 775 taniguchi t and morriss g p 2002 phys .
e * 65 * 056202 taniguchi t , dettmann c p and morriss g p 2002 j. stat .
* 109 * 747 mcnamara s and mareschal m 2001 phys .
e * 64 * 051103 ; mareschal m and mcnamara s 2004 physica d * 187 * 311 de wijn a s and van beijeren h 2004 phys .
e * 70 * 016207 hoover wm g , posch h a , forster ch , dellago c and zhou m 2002 j. stat . phys .
* 109 * 765 yang h l and radons g 2005 phys .
e * 71 * 036211 forster ch and posch h a 2005 new j. phys .
* 7 * 32 yang h l and radons g 2006 phys .
lett . * 96 * 074101 yang h l and radons g 2006 phys .
e * 73 * 016202 yang
h l and radons g 2006 phys .
e * 73 * 016208 yang h l and radons g 2006 phys .
e * 73 * 066201 allen m p and tildesley d j 1987 _ computer simulations of liquids _
( oxford : oxford university press ) stoddard s d and ford j 1973 phys .
a * 8 * 1504 smit b 1992 j. chem . phys . * 96 * 8639 oseledec v i 1968 trans .
* 19 * 197 benettin g , galgani l and strelcyn j m 1976 phys .
rev . a * 14 * 2338 shimada i and nagashima t 1979 prog .
. phys . * 61 * 1605 johnson r a , palmer k j and sell g r 1987 siam j. math . anal . *
18 * 1 goldhirsch i , sulem p l and orszag s a 1987 physica d * 27 * 311 ershov s v and potapov a b 1998 physica d * 118 * 167 szendro i g , paz d , rodrguez m a and lpez j m 2007 phys . rev .
e * 76 * 025202 ginelli f _ et .
_ 2007 phys .
lett . * 99 * 130601 eckmann j -p and ruelle d 1985 rev .
* 57 * 617 yang h l and radons g 2007 phys .
96 * 164101 yang h l and radons g 2008 phys .
e * 77 * 016203 dettmann c p and morriss g p 1996 phys .
e * 53 * r5545 ruelle d 1999 j. stat . phys . * 95 * 393 taniguchi t and morriss g p 2003 phys .
e * 68 * 046203 boon j p and yip s 1991 _ molecular hydrodynamics _
( new york : dover ) radons g and yang h l _ preprint _ nlin.cd/0404028 .
berne b j and pecora r 2000 _ dynamic light scattering _
( new york : dover ) press w h , teukolsky s a , vetterling w t and flannery b p 1992 _ numerical recipes in fortran 77 _ ( cambridge : cambridge university press ) livi r , pettini m , ruffo s , sparpaglione m and vulpiani a 1985 phys .
rev . a * 31 *
1039 forster c , hirschl r , posch h a and hoover wm h 2004 physica d * 187 * 294 | recent studies on the phase - space dynamics of a one - dimensional lennard - jones fluid reveal the existence of regular collective perturbations associated with the smallest positive lyapunov exponents of the system , called hydrodynamic lyapunov modes , which previously could only be identified in hard - core fluids . in this work we present a systematic study of the lyapunov exponents and lyapunov vectors , i.e. perturbations along each direction of phase space , of a three - dimensional lennard - jones fluid . by performing the fourier transform of the spatial density of the coordinate part of the lyapunov vector components and then time - averaging this result we find convincing signatures of longitudinal modes , with inconclusive evidence of transverse modes for all studied densities . furthermore , the longitudinal modes can be more clearly identified for the higher density values .
thus , according to our results , the mixing of modes induced both by the dynamics and the dimensionality induce a hitherto unknown type of order in the tangent space of the model herein studied at high density values . |
a microscopic description of freezing / melting phenomena presents a formidable challenge to present - day condensed matter theory @xcite . among the simple liquids ,
liquid metals have received limited attention in this field , stemming from the relative complexity of their interatomic potentials and thermodynamic properties @xcite .
previous related studies include those of stroud and ashcroft @xcite , who addressed the melting of simple metals using variational principles ; igli _ et al . _
@xcite , who studied the freezing of liquid aluminum and magnesium by means of density - functional ( df ) methods ; and moriarty and coworkers @xcite , who calculated the structural stabilities of third - period simple metals using first - principles techniques . in this contribution , we report on a new study of the freezing phenomena of simple metals that exploits recent advances in thermodynamic perturbation theory for solids . in line with ref .
@xcite , and most modern theories of freezing , our approach incorporates classical df theory @xcite in calculating the helmholtz free energy .
although formally rigorous , df theory is limited in practice by insufficient knowledge of the free energy functional for most except hard - body systems . to circumvent this limitation ,
we apply weeks - chandler - andersen ( wca ) perturbation theory @xcite , which splits the effective interatomic pair potential into reference and perturbative parts and maps the reference system onto an effective hard - sphere ( hs ) system . for the inhomogeneous solid phase
, df theory provides an adequate approximation for the hs free energy , while an accurate model of the hs radial distribution function yields directly the first - order perturbation free energy .
the homogeneous liquid phase is treated within the _ same _ perturbative framework , using essentially exact expressions for the hs free energy and pair distribution function .
this treatment of the liquid and solid by the same theoretical approach constitutes an important requirement for a consistent description of phase coexistence .
the next section outlines the theoretical methods used to calculate effective pair potentials and free energies of simple metals . in sec .
3 we present and discuss predictions of the theory for freezing transitions and relative stabilities of competing crystal structures . finally , in sec .
4 we summarize and conclude .
restricting attention to the simple metals , we construct effective density ( @xmath0)-dependent interatomic pair potentials , @xmath1 , and a volume - dependent , but structure insensitive , contribution to the binding energy , @xmath2 , via pseudpotential theory , assuming linear response of the conduction electrons to the ions . in practice
we use empty - core pseudopotentials and the ichimaru - utsumi electron dielectric function , which are known to give reliable results for bulk liquid properties @xcite . taking @xmath1 and @xmath2 as the basic microscopic input , we proceed to calculate the thermodynamics of the liquid and solid phases by means of wca perturbation theory .
thus , we separate @xmath3 at its first minimum into a short - range , purely repulsive , reference part , @xmath4 , and a long - range , weakly oscillating , perturbative part , @xmath5 , and map the reference system onto an effective hs system with a hs diameter prescribed by the well - known wca criterion .
this separation allows the ionic contribution to the helmholtz free energy to be calculated within the coupling - constant formalism @xcite .
to first order in @xmath5 , the resulting general ( inhomogeneous ) expression for the total helmholtz free energy reads @xmath6 = f_{\rm hs}[\rho({\bf r } ) ] + \frac{2\pi n^2}{v}\int_0^{\infty}{\rm d}r'~r'^2~ g_{\rm hs}(r';[\rho({\bf r})])~\phi_{\rm p}(r';\rho ) + e_v(\rho ) , \label{pert2}\ ] ] where @xmath7 is the number of particles , @xmath8 the volume , and @xmath9 $ ] and @xmath10)$ ] the free energy and radial distribution function ( rdf ) , respectively , of the hs reference system , both functionals of the equilibrium one - particle number density @xmath11 .
the rdf of the solid is defined as an orientational and translational average of the two - particle density @xcite . for the liquid we use the homogeneous version of eq .
( [ pert2 ] ) , taking for the hs free energy and rdf the carnahan - starling and verlet - weis parametrizations of computer simulation data @xcite . to calculate the free energy of the hs solid we invoke classical df theory @xcite .
the df approach is based on the existence of a functional @xmath12 $ ] of the density @xmath11 that satisfies a variational principle , according to which @xmath12 $ ] is minimized for a given average density and external potential by the equilibrium density , its minimum value equaling the helmholtz free energy @xmath13 . in practice ,
@xmath12 $ ] is split into an ( exactly - known ) ideal - gas contribution @xmath14 $ ] and an excess contribution @xmath15 $ ] , the latter depending entirely upon internal interactions . here
we approximate @xmath15 $ ] by the modified weighted - density approximation ( mwda ) @xcite .
this approximation maps the excess free energy per particle of the solid onto that of a corresponding uniform fluid of effective density @xmath16 , according to @xmath17 = n
f_{\rm hs}(\hat \rho ) , \label{mwda}\ ] ] where the effective density , defined as @xmath18 is a self - consistently determined weighted average of @xmath11 .
the weight function , @xmath19 , is specified by normalization and by the requirement that @xmath20 $ ] generates the exact two - particle direct correlation function in the uniform limit ( see ref .
@xcite for details ) . practical calculation of @xmath21 $ ] and @xmath22)$ ] requires specifying the solid density , _
i.e. _ , the coordinates of the lattice sites and the shape of the density distribution about these sites . here
we consider the _ hcp _ , _ fcc _ , and _ bcc _ crystals with the density distribution modeled by the usual gaussian ansatz , introducing a parameter @xmath23 determining the width of the distribution .
this parametrization allows the ideal contribution to the free energy functional to be accurately approximated by @xmath24 , where @xmath25 is the thermal de broglie wavelength , and yields the hs free energy as the minimum with respect to @xmath23 of the approximate functional @xmath21={\cal f}_{\rm id}[\rho({\bf r})]+ { \cal f}_{\rm ex}^{\rm mwda}[\rho({\bf r})]$ ] . for the rdf appearing in the perturbation contribution to the free energy [ eq .
( [ pert2 ] ) ] we adopt the approach of rascn _ et al .
_ @xcite . expressing @xmath10)$ ] as a sum over coordination shells ,
this scheme approximates the distributions of the second and higher coordination shells in simple mean - field fashion , but parametrizes the first - shell distribution in such a way as to incorporate nearest - neighbour correlations .
the first - peak parameters are determined by imposing sum rules , _
e.g. _ , the virial theorem and coordination number , so as to specify the contact ( @xmath26 ) value and shape ( area and first moment ) of the first peak ( for details see ref .
for a given solid structure , the lattice distances , the coordination number , the @xmath23 that minimizes @xmath21 $ ] , and the hs pressure , derived directly from @xmath9 $ ] , then combine to determine @xmath10)$ ] and hence the perturbation free energy . following the above procedure , we obtain the _ ionic _ free energy , @xmath27 , which is the only _
structure_-dependent contribution to the total free energy , and which suffices for assessing relative stabilities of different solid structures . to fully describe freezing / melting , however , it is essential to consider the _ total _ free energy , @xmath28 , where @xmath29 is an additional contribution from the conduction electrons . the electronic free energy ( or volume energy ) , @xmath29 , although independent of structure , depends on the average density and thus influences the densities of coexisting phases . here
we take for @xmath29 the usual expression following from second - order perturbation theory @xcite and consistent with the effective pair potential .
figure 1 illustrates the density dependence of the effective hs diameter , @xmath30 , for three different crystal structures of al near its melting point . according to the wca prescription
, @xmath30 depends both on the reference pair potential and on the first peak of the hs pair distribution function .
since the first peak is identical for _ hcp _ and _ fcc _ crystals , @xmath30 is the same for these two close - packed structures .
differences in the first peak for the _ bcc _ crystal result in a smaller diameter for that more open structure . because of the sensitivity of @xmath31 to the hs packing fraction ,
these structural variations in @xmath30 , though only of the order of 1 % , can amount to significant variations in the free energy . in passing ,
we note that the widely - used barker - henderson prescription for @xmath30 @xcite neglects such structural dependencies and that only knowledge of the hs rdf permits consistent application of the more accurate wca prescription @xcite
. predicted structural free energy differences for na , mg , and al are displayed in fig .
2 . evidently the theory correctly predicts the observed equilibrium structures for these three metals , namely _
bcc_-na , _ hcp_-mg , and _ fcc_-al , although the density at crossover from _ fcc_- to _ bcc_-na is somewhat overestimated .
compared with predictions of first - principles approaches , our results agree qualitatively well . for mg and al
the stability trends of fig .
2 match those from generalized pseudopotential theory ( gpt ) and the linear muffin - tin orbitals ( lmto ) method @xcite . as well , for na we predict the same trend as gpt , while lmto fails for this case .
compared with available experimental data , our structural free energy differences are in somewhat better quantitative agreement than those of either gpt or lmto .
for example , for mg and al we predict differences between _ hcp _ and _ fcc _ free energies of @xmath32 and @xmath33 mry , respectively , compared with @xmath34 and @xmath35 mry from gpt , @xmath36 and @xmath37 mry from lmto , and @xmath38 and @xmath39 mry from experiment ( extrapolations of thermochemical alloy data , as quoted in ref .
. it should be emphasized , however , that while the first - principles methods calculate ground - state ( @xmath40 ) energies , thermodynamic perturbation theory yields _ free _ energies of finite - temperature systems .
we note further that application of perturbation theory is restricted to those structures for which the hs system is at least metastable , _
i.e. _ , for which @xmath21 $ ] has a local minimum .
this excludes , _
e.g. _ , the diamond structure of si , the hs diamond crystal being unstable . finally , we investigate the freezing / melting transition by comparing predicted free energies for the liquid and solid phases .
figure 3 exemplifies the comparison for al near its observed melting temperature .
crossing of the ionic free energy curves for the liquid and _ fcc _ phases implies a distinct transition near the observed equilibrium solid density .
similar comparisons for na and mg also demonstrate liquid - solid phase transitions in qualitative agreement with observation . in this respect ,
the effective - pair - potential approach reasonably describes the many - body interacting system .
note that the liquid - solid crossover density serves as a useful upper ( lower ) bound on the liquid ( solid ) density at coexistence .
adding the volume - energy contribution
identical for the two phases while not affecting the crossover density , does alter the shape of the curves . as a result , the densities of coexisting liquid and solid phases , as determined by a maxwell common tangent construction , depend on @xmath29 .
an accurate volume energy is known to be essential for calculating realistic pressures and bulk moduli of metals from pseudopotential theory at the level of effective pair interactions @xcite .
similarly , we find the liquid - solid coexistence analysis to be sensitive to the form of @xmath29 , particularly its curvature with respect to @xmath0 .
in fact , the sensitivity is such as to preclude here a reliable determination of coexistence densities .
given the reasonable predictions for ionic free energies , this apparent limitation of the theory may reflect a need to go beyond the linear response approximation to include higher - order response effects in the volume energy ( see also @xcite , p. 42ff ) . the relative significance of such effects , as well as distinctions @xcite between the real - space representation of @xmath29 used here and the reciprocal - space representation of ref .
@xcite , are interesting issues deserving of further attention .
summarizing , working within an effective - pair - potential framework , we have implemented a consistent form of thermodynamic perturbation theory , one that is of comparable accuracy for both the liquid _ and _ solid phases , and used it to study structural stabilities and freezing behaviour of third - period simple metals .
predictions for structural free energy differences are at least as accurate as those from two first - principles methods and are in generally good qualitative agreement with experiment . on the basis of ionic free energies , freezing / melting transitions are predicted near the observed densities . as the densities of coexisting liquid and solid phases
are found to depend sensitively on the electronic free energy , a complete coexistence analysis has not been attempted pending more accurate knowledge of this part of the free energy .
the computationally practical approach demonstrated here should prove useful in future studies of the freezing of liquid metal mixtures , as well as in guiding more sophisticated _ ab initio _ simulations .
ard acknowledges the forschungszentrum jlich for use of its computing facilities .
gk and jh acknowledge the support of the sterreichische forschungsfonds under proj .
p11194 and p13062 and of the oesterreichische nationalbank under proj . | freezing of simple liquid metals and the relative stabilities of competing crystalline solids are investigated using thermodynamic perturbation theory , the interactions between ions being modeled by effective pair potentials derived from pseudopotential theory .
the ionic free energy of the solid phase is calculated , to first order in the perturbation potential , using classical density - functional theory and an accurate approximation to the hard - sphere radial distribution function .
free energy calculations for na , mg , and al yield well - defined freezing transitions and structural free energy differences for _ bcc , fcc , _ and _ hcp _ crystals in qualitative agreement with experiment
. = -2.0truecm , and |
quantum information processing ( qip ) has become one of the most promising applications of quantum mechanics @xcite . the first criterion to be fulfilled for actual implementation of qip
@xcite is the successful manipulation of a qubit , the basic unit for encoding quantum information @xcite . in this context
, manipulation means coherent control of the quantum dynamics of the chosen qubit .
the density matrix formalism is the perfect tool @xcite to explore those quantum dynamics , particularly when dealing with multipartite systems @xcite . for open quantum systems
, being in contact with a reservoir , this formalism provides a theoretical environment to describe the action of decoherence @xcite .
since the general state of one qubit is defined as a quantum superposition on a two - dimensional basis , it is natural to describe it as a @xmath0 spin system @xcite .
coupling such a system with electromagnetic radiation is one of the paths to implement a coherent manipulation of the qubit , and the theoretical models to describe this interaction must take into the account the classical or quantum nature of the radiation @xcite .
when considering quantum radiation , the jaynes - cummings model @xcite is one of the most successful theoretical descriptions of the spin - boson interaction and subsequent dynamics @xcite .
thinking about physical implementations , semiconductor nanostructures have become potential candidates for applications in qip and quantum computing @xcite .
specifically , relevant contributions arise from the study of physical properties of quantum dots ( qds ) , often recognized as artificial atoms , in front of the discrete character of the energy spectrum due to the confinement of carriers @xcite .
a qubit can be encoded inside a qd by using the charge @xcite , the spin @xcite , as well as excitonic states @xcite of the confined particle .
one of the advantages of the qds is the versatility on the experimental manipulation of ( valence - conduction ) band gap , with the subsequent customization of its optical properties .
the unique characteristics of quantum states in qd are behind the advantages of coupling a qd qubit with quantum light @xcite , particularly involving nanocavities @xcite .
once the qds are created using semiconductor materials , it is possible to confine carriers in order to maximize the dipole - dipole interaction , coupling the nanostructure strongly with a chosen mode of electromagnetic field of the cavity @xcite .
the crescent interest on such an experimental setup lies on its potential for the miniaturization of the cavity quantum eletrodynamics ( cqed ) , as found in atomic physics context @xcite .
once a typical setup is smaller than a micrometer @xcite , one can think on the implementation of an _ on - chip _ cqed @xcite using this kind of arrangement .
the rich optical response of quantum dots inside nanocavity includes nonlinearities @xcite and non - trivial emission spectra @xcite .
several groups concentrate efforts on developing applications of qd - cavity setup as quantum light emitters@xcite , exploring quantum dynamics in a similar way of successful procedures for production of quantum states of light in atomic cqed @xcite .
other approach , directly related with qip , is the use of quantum light for the coherent manipulation of the qd qubit .
typical phenomena of coherent dynamics as rabi oscillations @xcite , entanglement between a spin qd - qubit and photons @xcite and exciton - photon entanglement @xcite have been observed on qd - cavity systems . in this work ,
we propose encoding a qubit using an excitonic state , interacting with a coherent state of light prepared in a nanocavity .
the jaynes - cummings model is used to study the interaction between the quantum dot and the nanocavity and we consider photons losses in our treatment .
we calculate the density matrix operator dynamics , once we are dealing with a multipartite open quantum system . instead of a resonant condition between quantum dot and the cavity @xcite ,
we assume a non - resonant , self - trapped dynamics . to create and maintain a coherent state inside the cavity we use a continuous laser field applied in resonance with the cavity mode . to control the qubit rotations , the coherent manipulation of the system
is done by applying external laser pulses . to find the best set of parameters for the coherent manipulation
, we calculate the average occupation of the exciton state as a function of two detunings : laser - cavity and exciton - cavity .
the effect of the pulse area is also evaluated and the residual effect of the pulses over the cavity state is surveyed by checking the average number of photons and photon distribution .
we explore the behavior of the qd qubit by studying the population inversion , the dynamics over the poincar sphere and the purity of qubit , the last one using von neumann entropy .
our system is composed of a qd , treated here as a two - level system , coupled to a single mode nanocavity . to explore this physical setup
, we use the jaynes - cummings model @xcite under the rotating wave approximation ( rwa ) @xcite . to model the external lasers we use the dipole and rwa approximations @xcite and
assume that the pulsed laser only interact with the qd while the continuous laser only with the cavity .
the hamiltonian can be written as ( @xmath1 ) @xmath2 where @xmath3 are the pseudospin operators for the qd exciton qubit , @xmath4 and @xmath5 are the creation and annihilation operators for photons inside the cavity , @xmath6 , @xmath7 , @xmath8 and @xmath9 are the frequencies of exciton , cavity mode , pulsed laser and continuous laser field , respectively , @xmath10 is the rabi frequency that describes the exciton - laser interaction , with @xmath11 being the electric dipole strength of the exciton transition and @xmath12 the amplitude of the electric field of the laser , which can be constant or have different shapes in the pulsed excitation , and @xmath13 contains information about the laser field amplitude and the cavity transmission coefficient .
we encode a quantum bit using the qd exciton state , being @xmath14 the state with no exciton and @xmath15 the exciton state , and the cavity is described by the usual fock basis , @xmath16 .
the hamiltonian basis is depicted as @xmath17 with @xmath18 or @xmath19 , indicating the state of the qd qubit and @xmath20 being the number of photons in the cavity . to obtain the dynamics of our physical system
, we numerically solved the time dependent density matrix in the lindblad form @xmath21+\kappa\mathcal{d}[a]+\gamma\mathcal{d}[\sigma_-]+\phi\mathcal{d}[\sigma_z]\end{aligned}\ ] ] where @xmath22 is the full hamiltonian [ eq . ] and @xmath23=l\rho l^\dagger-\frac{1}{2}(l^\dagger l \rho + \rho l^\dagger l ) $ ] is the lindblad superoperator , which contains the incoherent terms of the density matrix and assumes a markovian approximation . here
@xmath24 is the photon loss rate of the cavity , @xmath25 is the decay rate of the qd and @xmath26 is the pure dephasing rate of the qd . in order to get a deeper understanding on the qd qubit dynamics we use the poincar - sphere representation , an analogous of the bloch - sphere representation for mixed states @xcite .
the use of this representation is common in the context of cavity quantum electrodynamics @xcite .
the qd qubit treated here is not in a pure state : when losses are not considered ( @xmath24 and @xmath25 null ) , the qd qubit is one of the parts of a bipartite system , and has some degree of entanglement with the cavity . for the contrary ,
when considering losses , one can say the qd qubit exchanges just a portion of its information with the cavity , once the bipartite system is now open . to obtain the components @xmath27 , @xmath28 and @xmath29 of the bloch vector
we first compute the reduced density operator for the qd qubit by doing the partial trace over the cavity variables @xmath30_{\mathrm{cav}}. \ ] ] then @xmath31 on poincar sphere , the @xmath29 component of the mixed , block - like , vector coincides with the population inversion .
the value of azimuthal angle is known as relative phase which corresponds to @xmath26 in the pure qubit represented as @xmath32 , with @xmath33 being related to the population of the state @xmath34 .
a bloch vector with a null azimuthal angle lies on the @xmath35 plane.@xcite to quantify the purity of the qd qubit , we use the von neumann entropy @xmath36 , which is defined as @xcite @xmath37 .
\label{eq : entropy}\end{aligned}\ ] ] from the definition , @xmath38 indicates that the qd qubit is a pure system , described by a state separated from the cavity and the reservoir . if losses are not considered on the description of the system , maximal degree of entanglement between the qubit with the cavity corresponds to @xmath39
to analyze the dynamics of qd qubit we parameterize all frequencies in units of @xmath40 , so it is easy to convert the values obtained here to real experimental situations .
for example , in photonic crystal nanocavity @xmath41 is of the order of @xmath42 mev @xcite .
this parametrization also allows our results to be applied to other cavity systems , which we will not discuss here . to solve the time - dependent master equation we carefully choose the size of photon basis to describe accurately the cavity state according to its mean occupation , taking into account the effects of interaction between qd , cavity and external lasers .
in most of the cases studied , a fock basis of @xmath43 was sufficient , so we set this value for all simulations presented here .
to get some insight to what we can do , lets first ignore incoherent effect ( @xmath44 ) and consider the initial state of our system as a direct product of a coherent state @xmath45 in the cavity and qd in the ground state @xmath14 , thus , @xmath46 .
the coherent state can be represented in the basis of fock states as @xmath47 where @xmath48 is mean number of photons .
a coherent state in the cavity can be created by sending a laser in resonance with the cavity mode for a short period of time ( @xmath49 and @xmath50 in our hamiltonian ) .
figure [ fig : resonant ] illustrates several aspects of the qd qubit dynamics when interacting with the cavity without the action of any external laser field ( @xmath51 ) and neglecting incoherent effect ( @xmath44 ) .
we consider that the exciton state is resonant with the cavity mode , @xmath52 , and an initial cavity state with @xmath53 .
figure [ fig : resonant](a ) shows the behavior of the population inversion , @xmath54 , with the apparition of collapses and revivals around an average value @xmath55 , in agreement with the predictions of the jaynes - cummings model .
notice that the average occupation of the cavity stays almost constant at a value around @xmath56 .
the dynamics of the qd qubit on poincar sphere is shown in the upper panel of fig .
[ fig : resonant ] , illustrated by constructing a vector which components are defined by eqs . .
in this figure , a scale of colors and arrows helps to visualize the temporal evolution .
first , the qd qubit starts from the pure state @xmath14 , indicated by a purple curves at the north pole of the poincar sphere .
then , at short times , the qd qubit performs a rotation around the @xmath27 axis , which is restricted to the @xmath35 plane ( @xmath57 ) , showing that the relative phase is null under the evolution .
it is also observed that the norm shrinks as time increases , which becomes evident at long times , as shown by green / yellow / red arrows . and in the absence of lasers fields .
( a ) population inversion and cavity occupation ( dashed blue line ) as a function of @xmath58 .
( b ) von neumann entropy as a function of @xmath58 .
upper panel shows the evolution of qd qubit over the poincar sphere . ]
the qd qubit dynamics restricted to a small region inside the poincar sphere , is connected with a high degree of entanglement with the cavity .
this behavior is better understood by checking the evolution of the von neumann entropy @xmath59 , eq . , shown in fig .
[ fig : resonant](b ) .
the initially pure qd qubit ( @xmath38 at @xmath60 ) performs oscillations between high and low entangled states , with an almost complete purification at @xmath61 associated with the collapse on population inversion .
this behavior repeats subsequent collapses ( not shown ) . as time
further increases , the degree of entanglement increases and approaches the maximal entangled states indicated by @xmath62 . and in the absence of lasers fields .
( a ) population inversion and cavity occupation ( dashed blue ) as a function of @xmath58 .
( b ) von neumann entropy as a function of @xmath58 .
upper panel shows the evolution of qd qubit over the poincar sphere . ]
it is also important to explore the behavior of the system under a non - resonant condition between qd qubit and cavity .
figure [ fig : nonresonant ] shows the same theoretical tools used on the description of the qd qubit resonant dynamics but considering @xmath63 .
figure [ fig : nonresonant](a ) shows that the population inversion does not perform complete oscillations between the qd qubit states , although collapses and revivals around the average value @xmath64 are still present .
this phenomena is called _ self - trapping _ or population trapping and is a well - know aspect of non - resonant dynamics concerning two - level systems .
the average occupation of the cavity stay constant at the value of @xmath53 over the evolution as seen by the blue line linked to the right axis .
the self - trapping is more evident in the upper panel of fig .
[ fig : nonresonant ] , once the mixed state vector is confined in a restricted region on the north hemisphere of poincar sphere .
it is worth noting that the mixed vector performs rotations around @xmath29 axis .
thinking in terms of a general qubit state written as @xmath65 , changes in the relative phase @xmath26 are connected with changes with the value of azimuthal angle of the qubit bloch vector .
thus , the non - resonant dynamics brings a gain of relative phase of the mixed vector , which oscillated between @xmath66 and @xmath67 .
figure [ fig : nonresonant](b ) shows the von neumann entropy , where differently from fig . [
fig : resonant ] the entanglement of the qd qubit suffers a stabilization , having asymptotic value around @xmath68 .
notice also that an increase of the purification appears at the revival of oscillations in the population inversion , indicating that the system try to recover its initial state . in general ,
nonresonant condition preserves the purity of qd qubit , once it prevents the qubit to interact in an efficient way with the cavity .
after the system being prepared in the state @xmath46 for a constant laser with @xmath69 as functions of detunings @xmath70 and @xmath71 with @xmath72 .
dark regions correspond to energy configurations where the qubit exciton state @xmath15 is being populated .
dashed red line illustrates the condition @xmath73 , where individual average occupations are shown in the upper panel .
dotted gray line in the upper panel correspond to @xmath63 , being the best condition for population inversion . ] after checking the basic aspects on qd qubit - cavity dynamics , we are ready to understand the action of the external lasers and incoherent effects . the main problem here is to find the right parameters to send the laser pulses as the coupling between qd and cavity modifies its interaction with the laser . a practical and fast way for survey the effect of this new ingredient
can be done through calculations of the average occupation @xmath74 of basis states @xmath17 as a function of some physical parameters neglecting incoherent process and assuming a constant laser excitation ( @xmath75 constant ) , which allow us to use the time independent schrdinger equation to compute the time evolution .
as we are interested in the qubit dynamics and wants to investigate where the population inversion occurs , we write the average occupation of the @xmath76 qd qubit state as @xmath77 . here
we choose to seek for best exciton - cavity @xmath70 , and laser - cavity @xmath71 detunings and the reason for this choice is that the cavity frequency @xmath7 is usually fixed by construction , making our procedure very similar to what one would do in an experimental setup .
figure [ fig : averpop ] ( lower panel ) shows the average occupation of the exciton state as functions of @xmath78 and @xmath79 for our system under a constant laser excitation with @xmath69 interacting with the qd , keeping @xmath80 .
the initial state is again @xmath46 with @xmath53 .
bright regions indicate low population of the exciton state in opposition to dark areas , therefore , dark areas are good candidates for laser induced population inversion . in the upper panel of fig .
[ fig : averpop ] we show a cut in the false color plot for @xmath73 ( red dashed line in the lower panel ) , showing the average occupation of the two qd qubit components .
gray dotted lines at @xmath63 indicates the maximum value of @xmath81 , being a good candidate to present a high degree of population inversion under pulsed excitation . as a function of the pulse area @xmath82 after the application of a single gaussian pulse at @xmath83 with duration of @xmath84 and detunings @xmath63 and @xmath73 .
( b ) average of the population inversion @xmath85 as a function of @xmath79 for a pulse with same parameters as in ( a ) with an area @xmath86 . ] in order to gain a fine control of the dynamics in realistic situations , we check the effect of the pulse parameters more closely .
we proceed the simulations using a single gaussian pulse of duration of @xmath84 , which in typical strong - coupled qd - cavity system would be a pulse with duration of the order of 4 to 5 ps , centered at @xmath87 , with optimal parameters obtained from fig .
[ fig : averpop ] ( @xmath63 and @xmath73 ) .
we observed that the average of the population inversion @xmath85 after the pulse is strongly dependent of the pulse area , defined as @xmath88 , as we can see in fig .
[ fig : central_av](a ) .
the maximum change in @xmath85 is obtained for @xmath89 and @xmath90 .
the dependence of @xmath85 with @xmath82 shows that the self - trapped dynamics , initially fixed at some polar angle over poincar sphere , can be set up on demand by the action of a pre - arranged laser pulse .
it is interesting to note that high values of pulse area bring nonlinear effects , linked with the increase of the pulse intensity and resulting in a lack of control of the dynamics .
this is indicated by nontrivial values of @xmath85 for @xmath91 .
we now simulate the behavior of @xmath85 as a function of the laser detuning to the cavity frequency considering a laser pulse with area @xmath86 , corresponding to the first minimum in fig .
[ fig : central_av](a ) . as we can see in fig .
[ fig : central_av](b ) , @xmath85 changes drastically for values around @xmath73 , in accord with the predictions for continuous laser as shown in fig .
[ fig : averpop ] .
it is also interesting to notice that there is a dip of @xmath85 around @xmath92 , in a range of @xmath93 , meaning the effects of the pulse on dynamics permits some flexibility over the exact value of @xmath79 . for practical applications ,
once it is established an adequate value of @xmath79 ( using the survey of populations ) , a specific pulse can be set up on demand in order to perform controlled self - trapped dynamics of the qd qubit .
this set of parameter , @xmath63 , @xmath92 and @xmath89 , will be kept constant for the rest of the paper .
to corroborate our approach , we proceed to explore the dynamics of population inversion , qd qubit on poincar sphere and von neumann entropy .
we discuss the action of a sequences of three pulses .
we consider again the qd qubit initialized at the ground state @xmath14 and the cavity in the coherent state @xmath45 with @xmath53 and the conditions described before . , @xmath92 , neglecting losses and under the effect of three gaussian , the first and second with the same pulse area @xmath86 , and the last with @xmath94 , applied at @xmath95 , @xmath96 and @xmath97 , all with duration of @xmath84 . ( a ) sequence of pulses used in the simulation ,
( b ) population inversion and cavity occupation ( dashed blue ) as a function of @xmath58 and ( c ) von neumann entropy as a function of @xmath58 .
upper panel shows the evolution of qd qubit over the poincar sphere , with the color code and arrows indicating the time sequence . ]
figure [ inv2 ] shows the qd qubit dynamics under the action of a sequence of three pulse , all with duration of @xmath84 for the best detuning found before .
the format of the pulse is shown in fig .
[ inv2](a ) . the population inversion , fig .
[ inv2](b ) , shows that the qd qubit initially ( for @xmath98 ) performs self - trapped oscillations around @xmath99 .
this can be better visualized in the top panel of fig .
[ inv2 ] , where the dynamics is restricted to the north hemisphere of the poincar sphere . after the application of the first pulse at @xmath95 with pulse area @xmath100 , the population inversion starts an oscillation with a new structure of collapses and revivals , oscillating around @xmath101 , still in a self - trapped dynamics as can also be seen in the poincar sphere , moving the oscillation from north to the south hemisphere .
after the application of the second pulse at @xmath96 with same pulse area @xmath102 , the self - trapping oscillations changes to an average value around @xmath103 , moving back to the north hemisphere of the poincar sphere . further control can be obtained with the last pulse , applied at @xmath97 with a pulse area half of the previous case ( @xmath104 ) , creating a superposition between exciton and ground state , showing that we can control the qd qubit state reasonable well with this choice of pulse sequence . concerning the entanglement dynamics ,
again explored using the von neumann entropy plotted in fig .
[ inv2](b ) , it is worthy noting that the two first pulses changes slightly the entanglement degree of qd qubit , keeping the value of the entropy below @xmath105 .
the third pulse , on the other hand , creates a situation similar to the full resonant case , fig .
[ fig : resonant ] , where the entropy goes to maximum , and the inversion exhibits oscillations around the average value of @xmath55 .
one advantage of using non - resonant condition is that the qd lifetime increases from a few hundred of picoseconds to a few nanoseconds due to the reduction of fluctuations in the vacuum of the electromagnetic field @xcite .
taking the lower limit , lets assume that lifetime of the qd is about @xmath106 ns , which gives a decay rate of the order of @xmath107 . since we are proposing the manipulation the qd qubit in a timescale of a few picoseconds , the effects of this decay rate can be neglected from our analysis .
we also consider a pure dephasing rate of the same order of magnitude @xmath108 and as we shall see later , pure dephasing plays an important role in the preparation of the initial stated used in dynamics of the coupled system . , but now considering decoherence .
( a ) sequence of pulses used in the simulation .
( b ) population inversion and cavity occupation ( dashed blue line ) as a function of @xmath58 for @xmath109 , @xmath107 and @xmath110 .
( c ) same as ( b ) for @xmath111 . ]
cavity loss is another term that can not be neglected since the average life span of a photon within a cavity may be as small as a few picoseconds in a bad cavity . to understand the effects of the cavity loss , in fig .
[ inv2ab ] we show the dynamics of the system using same parameters and sequence of pulses as in fig . [ inv2 ] , but now including incoherent effects . in fig .
[ inv2ab](b ) we use @xmath107 , @xmath110 and @xmath109 , while in fig .
[ inv2ab](c ) we use @xmath107 , @xmath110 and @xmath111 .
as we can see in this figure , even a small loss of the cavity is enough to break the qd qubit manipulation .
notice that the average occupation of the cavity photons decease exponentially and we have @xmath112 , @xmath113 and @xmath114 when the first , second and third pulses are applied , respectively .
pure dephasing and exciton spontaneous decay plays no role in this particular case as the time scale is small . notice that we have used @xmath115 as initial state , which is a pure state . for the case of @xmath111 ,
[ inv2ab](c ) , the occupation of the cavity goes to zero very quickly , being zero before the second pulse , and the coherent manipulation of the qd qubit is completely destroyed .
thus , even thought the cavity is out of resonance , the loss of the cavity plays an important role on the qd qubit manipulation .
it is important to mention that the coherent state under cavity loss will still be a coherent state , but with a lower average of photons .
the main problem here is that the resonance condition ( best parameter we found in figs .
[ fig : averpop ] and [ fig : central_av ] ) changes over time as the photon population decreases and a complete control of the qd state will require an previous evaluation of the best parameter for each pulse of the pulse sequence to control its states . experimentally , this is challenging , as it require sending pulses with different frequency and intensities in a short time scale . , @xmath107 , @xmath110 and @xmath116 assuming @xmath117 as initial state .
right panel shows the photon distribution for the quantum state at @xmath118 . ]
to solve this problem , instead of using the laser described by the coupling @xmath13 to only prepare the initial coherent state in the cavity , lets keep it as a constant pump to maintain a steady coherent state in the cavity . in fig .
[ coh1 ] we plot the evolution of the photon distribution in the fock basis considering @xmath63 , @xmath92 , ( same parameters used before ) , @xmath50 , @xmath116 , @xmath111 , @xmath107 and @xmath108 as a function of @xmath58 , assuming @xmath117 as initial state . here
we choose @xmath119 because that produces an steady coherent state with @xmath120 , close to the condition used in fig .
the right panel is a view of the photon distribution for @xmath118 , showing a coherent photon distribution .
, now considering @xmath111 , @xmath107 , @xmath110 and @xmath116 and the initial state the state at @xmath118 , whose photon distribution is shown in the right panel of fig .
( a ) sequence of pulses used in the simulation , ( b ) population inversion and cavity occupation ( dashed blue line ) as a function of @xmath58 .
( c ) von neumann entropy as a function of @xmath58 .
upper panel shows the evolution of qd qubit over the poincar sphere , with the color - code indicating the time sequence . ] using the state at @xmath118 obtained with @xmath116 as initial state of the system , in fig .
[ inv2d ] we shows the dynamics of the system under the same pulse sequence and parameters as in fig . [ inv2ab](c ) , with additional parameters @xmath121 , @xmath111 , @xmath107 and @xmath110 . as we can see in fig .
[ inv2d](b ) , we have now a better control the inversion of population with our pulse sequence as the cavity occupation is kept almost constant over the evolution despite of the application of the lasers pulses interacting with the qd .
the value is very close to our ideal situation ( with no losses ) , with an average of photons in the cavity around @xmath56 as we initially planed .
notice , however that the range of population inversion is decreased , this is due to the pure dephasing rate acting in the preparation of the state of cavity ( the evolution during the pulse sequence is too short for the pure dephasing and decay of the exciton to play a role ) .
the effect of the pure dephasing rate in the initial state can also be seen is in fig .
[ inv2d](c ) , where the initial value of the von neumann entropy is about @xmath122 . despite the initial difference ,
the evolution of the von neumann entropy is very similar to the case analyzed previously in fig .
[ inv2](c ) . neglecting pure dephasing completely and keeping all parameters as in fig .
[ inv2d ] we obtain the fig . [ inv2e ] .
this two figures have similar general characteristics .
the main differences are in range of the population inversion and the initial value of the von neumann entropy , which is a result of different initial state as we mention before . in the case with pure dephasing , fig .
[ inv2d](b ) , the population inversion is restricted to smaller absolute values , resulting in self trapping close to the central region of the poincar sphere , indicating that we have a mixed state .
this results in a large value of the von neumann entropy , indicating that the initial state prepared ( fig .
[ coh1 ] for @xmath118 ) is not pure .
here it is important to mention that pure dephasing in solid state systems can not be neglected , being always present due to the nature of the system and the presence of phonons , in this way , the preparation of the initial state in solid state system will always face this problem . to minimize its effect
, the exciton - cavity detuning can be increased . in systems where pure dephasing can be safely neglected ,
the result presented in fig .
[ inv2d ] might be useful .
, now neglecting pure dephasing and considering the initial state as the state at @xmath118 for @xmath123 . ( a ) sequence of pulses used in the simulation , ( b ) population inversion and cavity occupation ( dashed blue line ) as a function of @xmath58 .
( c ) von neumann entropy as a function of @xmath58 .
upper panel shows the evolution of qd qubit over the poincar sphere , with the color - code indicating the time sequence . ]
in this work , we discuss the dynamics of an exciton in a quantum dot which interacts with a coherent state , supported by a cavity , under the action of a continuous laser , for controlling the cavity losses , and external gaussian laser pulses , to control exciton - cavity dynamics . our study is a step towards the implementation of an _ on - chip _ cavity quantum eletrodynamics .
we define a qubit using two levels on the quantum dot so @xmath14 is the vacuum state ( no exciton ) and @xmath15 is the exciton state ( qd qubit ) .
we use the population inversion , @xmath54 , a mixed vector definition of quantum dot qubit on poincar sphere to study the dynamics of the qubit and the von neumann entropy , to analyze its entanglement degree with the cavity mode ( closed system ) as well as its purity ( open system ) .
the treatment without considering losses provides important information : by the average population is used to define an efficient condition of qubit - cavity and pulse - cavity detunings for populating the dressed states @xmath17 .
the dynamics shows self - trapping on population inversion , among with oscillations of the relative phase of the qubit . by including the pulses
, we check the effect of the pulse area on the dynamics showing that the central value of population inversion changes significantly after the pulse .
the sequence of pulses can be also used to increase the entanglement degree between the qubit and the cavity in non - resonant ideal ( no losses ) condition .
we also discuss the effect of losses on our approach . because the qd qubit lifetime is long enough ( and it can be even longer in a non - resonant condition with the cavity ) , we avoid the effects of spontaneous emission by choosing a short time scale defined by @xmath124 . the qd qubit dynamical control , on - demand ,
is attained by using a sequence of short pulses . to protect the manipulation against cavity looses , we explore the use of an additional continuous laser , which maintains a steady coherent state inside the cavity .
this mechanism already sustains the qd qubit dynamics assuring the success of our proposal .
we also analyzed the effects of pure dephasing of the qd qubit in the dynamics , showing that it produces no effects in the time evolution during our short pulse sequence , but it might affect the preparation of the initial state . as future works , we intend to continued studying the entanglement between the qd qubit and cavity by quantifying the existence of quantum correlations when considering losses .
the goal is to explore the potential of this system as an entanglement resource .
a second issue is to engage a study about production quantum light but focusing on other than single photons .
we would thank the referees for the helpful critics and questions .
their careful and detailed review is behind the enrichment of our work .
we acknowledge financial support of capes , cnpq , fapemig , disse and inct .
54ifxundefined [ 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 * * , ( ) link:\doibase 10.1103/physreva.51.2738 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.29.74 [ * * , ( ) ] @noop _ _ ( , ) link:\doibase 10.1103/revmodphys.76.1267 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.75.715 [ * * , ( ) ] @noop * * , ( ) @noop _ _ ( , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevb.69.113301 [ * * , ( ) ] link:\doibase 10.1103/physrevb.59.2070 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevlett.83.4204 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.87.067401 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevb.69.125342 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1038/nature02063 [ * * , ( ) ] \doibase http://dx.doi.org/10.1038/nature05586 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/revmodphys.87.347 [ * * , ( ) ] link:\doibase 10.1103/physrevb.91.241305 [ * * , ( ) ] link:\doibase 10.1103/physrevb.85.195301 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.103.087405 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.114.233601 [ * * , ( ) ] @noop * * ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physreva.69.032305 [ * * , ( ) ] @noop * * , ( ) @noop * * ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevlett.110.167401 [ * * , ( ) ] @noop * * ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevlett.114.220501 [ * * , ( ) ] | in this work we demonstrate theoretically how to use external laser field to control the population inversion of a single quantum dot exciton qubit in a nanocavity .
we consider the jaynes - cummings model to describe the system , and the incoherent losses were take into account by using lindblad operators .
we have demonstrated how to prepare the initial state in a superposition of the exciton in the ground state and the cavity in a coherent state .
the effects of exciton - cavity detuning , the laser - cavity detunings , the pulse area and losses over the qubit dynamics are analyzed .
we also show how to use a continuous laser pumping in resonance with the cavity mode to sustain a coherent state inside the cavity , providing some protection to the qubit against cavity loss . |
the modern trends in miniaturization of electronic devices eventually led to fabrication of single molecular junctions and molecular transistors ( see e.g. review @xcite ) .
the electric properties of single molecular transistors ( smts ) in many cases are similar to the analogous characteristics of single electron transistors ( sets ) fabricated in two - dimensional electron gas .
smts demonstrate such effects as coulomb blockade and coulomb blockade oscillations on gate voltage .
the significant difference between smt and semiconducting set is that the former can function even at room temperatures that makes them to be very promising basic elements for nanoelectronics .
another specific feature of molecular transistors is the interaction between electronic and vibrational degrees of freedom .
the electron in the process of tunneling through the molecule can excite ( and absorb at finite temperatures ) molecular vibrational quanta ( vibrons ) - the phenomenon known as `` phonon - assisted tunneling '' @xcite . the opening of inelastic channels results in appearance of additional peaks ( side - band peaks ) in differential conductance . for weak electron - vibron interaction
the magnitudes of inelastic peaks are much smaller then the value of the elastic resonance peak .
the situation is changed in the regime of strong electron - vibron interaction when nonperturbative ( and , in particular , polaronic ) effects determine electron transport through a vibrating molecule ( see e.g.@xcite ) .
polaronic effects are most pronounced in the case when the molecule ( quantum dot ) is well - separated from the leads and the width , @xmath0 , of conducting molecular states is small compared to other energy scales ( temperature @xmath1 , driving voltage @xmath2 ) . in this case
the mechanism of electron transport through the vibrating molecule is ( inelastic ) sequential tunneling via the real polaronic intermediate state .
the amplitude of this tunneling is exponentially suppressed since the wave functions of free electron in the leads and polaronic state ( holstein polaron ) in the dot are almost orthogonal .
this effect ( named as frank - condon @xcite,@xcite or polaronic @xcite blockade ) strongly suppresses elastic channel of electron transport and changes the temperature behavior of conductance .
recently , frank - condon blockade was observed in experiment of electron tunneling through a suspended single - wall carbon nanotube @xcite .
a one more novel phenomenon appears for vibrating quantum dot when the matrix element of electron tunneling to the left and to the right lead differently depends on the position of the dot center of mass .
this is always the case when the dot ( molecule ) vibrates in the direction of electron tunneling and then the effect of electron shuttling takes place at finite voltages @xcite-@xcite . in papers
@xcite,@xcite the problem of electron shuttling was considered for a model of single level ( @xmath3 ) vibrating quantum dot weakly coupled to the leads of noninteracting electrons .
it was shown that in the regime of weak electromechanical coupling the shuttle instability occurs at bias voltages @xmath4 ( @xmath5 is the vibron energy ) and the increment of instability is @xmath6 . here
@xmath7 is the level width , @xmath8 is the dimensionless electron - vibron interaction constant and @xmath9 ( @xmath10 is the amplitude of zero - point fluctuations of quantum dot , @xmath11 is the electron tunneling length ) .
both coupling constants @xmath8 , @xmath12 were assumed to be small in refs.@xcite,@xcite . in the problem of electron shuttling the electron - vibron interaction strength @xmath13 linearly depends on the driving voltage .
therefore , at sufficiently high voltages the regime of strong electron - vibron interaction ( @xmath14 ) is realized . in this regime
polaronic effects could play significant role in electron shuttling . in the present paper
we reconsider the problem of shuttle instability for the regime of strong electromechanical coupling .
denotes the `` shifted '' ( due to polaronic shift ) bias voltage - dependent fermionic level ; @xmath5 is the energy of vibrational mode ; @xmath15 are the coordinate - dependent tunneling amplitudes and @xmath16 are the chemical potentials of the leads @xmath17 .
characteristic distances in the qd : @xmath18 is the distance ( `` gap '' ) between the leads ; @xmath11 is the tunneling length of the electron in the qd ; @xmath19 is average coordinate of the shuttle.,width=8,height=7 ] we derived the equation of motion for the shuttle average coordinate @xmath19 assuming only the weak character of dot - lead interaction .
it was shown that the increment @xmath20 of shuttle instability is a nonmonotonic function of bias voltage with a maximum at @xmath21 , where @xmath22 is the distance between the leads .
the maximum value of the increment is sensitive to value of coupling constant @xmath12 .
the interplay of two effects - the increase of @xmath23 caused by the increase of inelastic channels which contribute to the increment when rising the applied voltage and the decrease of @xmath23 with the increase of bias voltage caused by polaronic blockade - results in oscillation of @xmath23 on small energy scale @xmath5 .
our results show that polaronic effects determine the physics of electron shuttling in the case of moderate or strong mechanical damping when the transition to a shuttle - like regime of electron transport is possible only at sufficiently high bias voltages .
our starting point is the model of vibrating quantum dot weakly coupled to the leads of noninteracting electrons .
this model was repeatedly considered in the literature for the problem of electron transport in molecular transistors ( see e.g.@xcite and referencies therein ) .
we expand this model to the problem of electron shuttling @xcite by taking into account the explicit dependence of tunneling amplitude on the center of mass coordinate of quantum dot . for simplicity we will study the case of a single level ( with the energy @xmath3 ) quantum dot coupled to a single vibronic mode ( with the energy @xmath24 ) .
the total hamiltonian of our system is @xmath25 where @xmath26 is the standard hamiltonian of noninteracting electrons ( @xmath27 ) in the left ( @xmath28 ) and right ( @xmath29 ) leads , @xmath30 is the corresponding chemical potential : @xmath31 ( @xmath32 is the driving voltage ) ; @xmath33 ( @xmath34 ) are the creation(destruction ) operators .
the hamiltonian of vibrating quantum dot takes the form ( see e.g.@xcite ) @xmath35 where @xmath36 is the characteristic energy of electron - vibron interaction ( see below ) , @xmath37(@xmath38 ) and @xmath39(@xmath40 ) are fermionic and bosonic creation ( destruction ) operators with commutation relations @xmath41 ; @xmath42=1 $ ] .
the tunneling hamiltonian is @xmath43\,.\ ] ] in hamiltonian eq.([4 ] ) we take into account the dependence of tunneling amplitude @xmath44 on the center of mass coordinate @xmath45 of quantum dot . in quantum description
the coordinate @xmath45 becomes an operator @xmath46 where @xmath47 , ( @xmath48 is the mass of qd ) is the amplitude of zero - point oscillations of quantum dot . in eq.([5 ] ) we defined also the dimensionless momentum operator @xmath49 with the canonical commutation relation @xmath50=i$ ] . in
what follows the tunneling amplitude is model @xcite by the exponential function @xmath51 here @xmath52 and @xmath11 is the tunneling length .
notice that electron - vibron interaction term in eq.([3 ] ) in our model originates from the electrostatic interaction of charge density on the dot with the electrostatic potential produced by the leads @xcite .
it is convinient to characterize this interaction by the dimensionless bias voltage - dependent coupling constant @xmath53 ( @xmath32 is the bias voltage , @xmath18 is the distance between the leads ) .
we use the notations usually accepted in the literature on molecular transistors .
notice that our notations for the coupling constants differ from refs.@xcite .
the problem of electron shuttling in the model eqs.([1])- ( [ 3 ] ) was studied in refs.@xcite for the case of weak electromechanical coupling ( in our notations : @xmath54 , @xmath55 ) . in molecular transistors
the electron - vibron interaction can be strong @xmath14 ( see e.g. @xcite ) .
here we reconsider the problem of shuttle instability @xcite in the regime of strong coupling . to study electron transport in the presence of polaronic effects ( @xmath14 ) it is convenient to use unitary transformation ( see e.g.@xcite ) which eliminates electron - vibron interaction term in the dot hamiltonian eq.([3 ] ) .
shuttle instability results in appearance of classical time - dependent coordinate @xmath19 of quantum dot .
it means that bosonic operators @xmath40 and @xmath39 acquire @xmath56-number part @xmath57 : @xmath58 here @xmath59(@xmath60 ) are the bosonic creation(destruction ) operators , which describe quantized vibron modes @xmath61 ( @xmath62 denotes the thermal average ) .
we transform total hamiltonian ( 1 ) using the unitary operator ( see e.g.@xcite ) @xmath63 the unitary transformed hamiltonian takes the form @xmath64 where @xmath65\hbar\omega_{0}$ ] , @xmath66 . the time - independent shift ( @xmath67 ) of the dot level
is called polaronic shift @xcite .
the transformed tunneling hamiltonian in eq.([10 ] ) is @xmath68\nonumber\\ = \bar{t}_{0j}\sum_{k}\left[\hat{t}_{j}(\hat{x}_{t})\hat{a}_{kj}^{+}\hat{v}_{j}^{+}\hat{c}+h.c.\right]\,.\end{aligned}\ ] ] here @xmath69 is the transformed coordinate operator @xmath70 and @xmath71
the heisenberg equations of motion for dimensionless operators of coordinate @xmath72 and momentum @xmath49 @xmath73 ( where @xmath66 , @xmath74 ) can be represented in the form of hamilton equations @xmath75 with the hamiltonian @xmath76 given by the following expression @xmath77 here we denote by @xmath78 the tunneling hamiltonian defined by eq.([11 ] ) . in our model equations ( [ 14 ] ) take the form @xmath79 where @xmath80 are the current operators @xmath81\,.\ ] ] these operators satisfy ( as it should be ) the continuity equation @xmath82 with the help of eq.([18 ] ) we can rewrite the first expression of eq.([16 ] ) in the following form @xmath83 and the second equation in ( [ 16 ] ) is transformed to @xmath84 where @xmath85 . to make the operator equations ( eq.([16 ] ) or eqs.([19 ] ) ) , ( [ 20 ] ) complete we have to write down the equations of motion for fermionic operators @xmath34(@xmath33 ) and @xmath38(@xmath37 ) .
these equations @xmath86 @xmath87 are linear and they can be readily solved ( the equations of motion for creation operators are obtained from eqs.([21]),([22 ] ) simply by taking hermitian conjugation ) .
we will follow refs.@xcite and find the solution of eqs.([21]),([22 ] ) in the so called wide band approximation``(''wba")(see e.g. @xcite ) .
the only difference of our system of equations ( [ 21]),([22 ] ) from the corresponding one in ref.@xcite is the presence in our case bosonic operator factors @xmath88 and @xmath89 .
formally , these factors make the level width @xmath90 ( see the definition below in eq.([24 ] ) , which appears in eq.([22 ] ) after substituting in this equation the solution of eq.([21 ] ) , to be an operator as well .
the dynamical equation for the dot level operator @xmath38 in wba takes a simple form @xmath91\hat{c}(t)\nonumber\\ -i\sum_{k , j
= l , r}\bar{t}_{0j}\hat{v}_{j}(t)\hat{a}_{kj}(0)e^{-i(\varepsilon_{k}-\mu_{j})t}\,,\end{aligned}\ ] ] where @xmath92 here @xmath93 and @xmath94 is the density of states in the @xmath95 electrode ( @xmath94 is assumed to be energy independent in the wide band approximation ) . to proceed further we have to make additional simplifications .
our purpose in this section is to derive the equation of motion for the average coordinate in the presence of fluctuations .
notice that in the hamiltonian eq.([10 ] ) electron - vibron interaction appears only in the tunneling term which for a weak tunneling can be treated perturbatively . in perturbation theory on the bare level widths @xmath96 vibrons are decoupled from fermions and when evaluating averages of the products of fermion ( @xmath97 ) and boson ( @xmath98 ) operators we can use the decoupling procedure @xmath99 the averages of fermion operators are taken with `` fermionic '' part of the total transformed hamiltonian ( including the tunneling part ) and averages of boson operators are calculated with the quadratic hamiltonian of nonointeracting vibrons @xmath100 .
this decoupling procedure is the starting point of calculations in many papers dealing with the polaronic effects in electron transport in molecular junctions ( see e.g. review @xcite and referencies therein ) . the linear equation ( [ 23 ] ) can be formally solved ( using time - ordering procedure ) for the operator level width @xmath101 .
however the closed equation for the average coordinate @xmath19 can be obtained only in perturbation theory on @xmath96 . in ref.@xcite
such equation was obtained in the limit of weak electromechanical coupling @xmath54 , @xmath55 . here
we derive the equation for classical coordinate @xmath19 valid also for the regime of strong electron - vibron coupling @xmath102 . in perturbation theory on the bare level width
@xmath96 we can neglect the time dynamics of the level width and replace @xmath101 by constant @xmath103 .
it is convinient in what follows to represent `` vertex '' operator @xmath88 as a product of classical @xmath104 and quantum @xmath105 parts @xmath106 @xmath107\,,\,\hat{q}_{j}(\hat{x},\hat{p})=\exp(j\lambda_{t}\hat{x}+i\lambda\hat{p})\,,\nonumber\\ \,j=(l , r)\equiv(-,+)\ , \nonumber\\\end{aligned}\ ] ] in the discussed approximation the solution of eq.([23 ] ) takes the form @xmath108\right\}\ , .
\nonumber\\ \end{aligned}\ ] ] the equation of motion for the classical coordinate @xmath19 in perturbation theory on @xmath96 can be readily obtained from the exact operator equation ( [ 20 ] ) by taking the average and using the discussed above `` fermion - boson '' factorization procedure @xmath109+\omega_{0}^{2}\bar{x}(t)=-\frac{\omega_{0}\lambda_{t}}{\hbar}\sum_{j=_{+/r}^{-/l}}jh_{j}\left\{\bar{x}(t)\right\}\,,\ ] ] where @xmath110 it is useful to rewrite eq.([27 ] ) in terms of a new variable @xmath111 . notice that according to eqs.([11]),([26 ] ) both quantities in eq.([27 ] ) - averaged tunneling hamiltonian @xmath112 and the level occupation number @xmath113 - are proportional to the bare level width @xmath114 .
since eq.([27 ] ) is derived in the born approximation ( up to the second order in the tunneling amplitude ) we can replace @xmath19 by @xmath115 in the averaged quantities . then eq.([27 ] ) for the variable @xmath115 ( shifted coordinate ) takes the form of the newton s equation derived in ref.@xcite @xmath116 where @xmath117 this equation in the limit @xmath54 , @xmath55 , when we can omit operator factors ( @xmath105 ) in the vertex function @xmath88 , exactly coincides with the corresponding equation for the coordinate of classical " shuttle ( see refs.@xcite )
. one can analyze shuttle instability by using either eq.([27 ] ) or eq.([29 ] ) .
the only difference is the starting ( equilibrium ) position of the shuttle .
this is @xmath118 for eq.([27 ] ) and @xmath119 for eq.([29 ] ) ( see ref.@xcite ) .
we will use eq.([27 ] ) .
the conditions for a shuttle instability can be found by analyzing linearized equation of motion .
we will follow ref.@xcite where these conditions were obtained for the regime of weak electromechanical coupling .
the linear integral - differential equation for the shuttle coordinate @xmath19 in the dimensionless units ( @xmath120 , energy scale is @xmath121 ) reads ( see appendix ) @xmath122\left|\lambda g_{lj}^{(+)}(\varepsilon)-j\lambda_{t}\right|^{2}\right\}\ , \nonumber\\\end{aligned}\ ] ] in eq.([31 ] ) the following notations are introduced : @xmath123 @xmath124(t-\tau)\right\}\,,\end{aligned}\ ] ] and @xmath125 where @xmath126 is the polaronic shift .
vibrational degrees of freedom result in franck - condon factors @xmath127 \left|\frac{\lambda+j\lambda_{t}}{\lambda - j\lambda_{t}}\right|^{l}\nonumber\\ \cdot i_{l}(2|\lambda^{2}-\lambda_{t}^{2}|\sqrt{n_{b}(1+n_{b})})e^{-\beta l/2}\,\end{aligned}\ ] ] here @xmath128 is the modified bessel function of the second kind and @xmath129 , ( @xmath130 ) , @xmath131 $ ] are bose - eistein and fermi - dirac distribution functions . remind that all energies in above expressions are dimensionless ( in the units of @xmath121 ) . in the limit @xmath132 the vibron - induced correlation factor eq.([34 ] ) coincides with the well - known in the literature expression ( see e.g. refs.(@xcite ) .
the solution of eq.([31 ] ) with the initial condition @xmath133 is @xmath134 where @xmath135 is an arbitrary constant and @xmath136 is the dimensionless increment ( when @xmath137 ) of shuttle instability . our purpose here is to find conditions ( driving voltage ) for the realization of shuttle motion in the presence of strong electron - vibron interaction . for a weak electromechanical coupling @xmath54 , @xmath55 we can neglect quantum and thermodynamical fluctuations of vibrons ( they are of higher orders on coupling constants ) and omit all terms in the right - hand side of eq.([36 ] ) but @xmath138 . in this limit @xmath139 , @xmath140 , @xmath141 and eq.([36 ] )
is transformed to the corresponding equation of refs.@xcite .
the increment eq.([36 ] ) in this case coincides with the one found earlier ref.@xcite @xmath142 ( there is a misprint in eq.(24 ) of ref.@xcite - the sign - changing factor @xmath143 is missing in the coefficients in front of distrubution functions ) .
it is qualitatively clear that shuttle instability is most pronounced at low temperatures @xmath144 when there are no thermally activated vibrons . at low temperatures shuttle instability results in a sharp ( step - like ) features in the @xmath145 curve of single electron transistor @xcite .
finite temperature effects broaden the transition region and at @xmath146 the shuttle instability is less pronounced ( nevertheless as far as @xmath147 the instability results in a strong increase of electric current ) . in what follows
we will consider only low-@xmath1 effects and ( for simplicity ) the case of a symmetric junction ( @xmath148 , we set fermi energy of the leads @xmath149 ) . at low temperatures the @xmath150 terms in eq.([36 ] )
are exponentially suppressed in comparison with the negative @xmath151 ( formally , due to the factor @xmath152 since @xmath153 for the integer @xmath151 ) .
physically , positive @xmath151 corresponds to vibron absorption - an energetically forbidden process at @xmath154 .
negative @xmath151 describes emission of vibrons and the summation over @xmath151 at finite bias voltage is limited by a certain @xmath155 ( see below ) . by using the well - known asymptotics of the bessel function at small arguments @xmath156 and replacing fermi distribution functions in eq.([36 ] ) by the heaviside theta - functions we obtain the desired formula for the increment ( @xmath147 ) of shuttle instability @xmath157 where @xmath158 is the incomplete gamma function ( see e.g. @xcite ) and @xmath159\ ] ] ( @xmath160 $ ] denotes the integer part of @xmath161 ) .
we find from eqs.([38]),([lm ] ) that in the regime of weak electromechanical coupling the instability occurs at @xmath162 @xcite,@xcite and the increment @xmath163 ( see ref.@xcite ) is a linear function of bias voltage .
remind that we consider the case when the uncertainty in the quantum dot initial position due to quantum fluctuations ( @xmath164 ) is small compared to the geometrical size @xmath18 of the junction ( that is our parameter @xmath165 , eq.([7 ] ) ) . in this case at the threshold voltage @xmath166 the electron - vibron coupling is small @xmath167 ( we always can put @xmath168 at the resonance condition ) and polaronic effects are not pronounced .
nevertheless there is a small negative correction to the threshold voltage due to polaronic shift @xmath169 and the multiplicative renormalization of the bare level width by quantum fluctuations @xmath170 , as one can see from eqs.([38])([lm ] ) . the increment @xmath171 is described by eq.([38 ] ) at bias voltages in the interval @xmath172 . at higher voltages the electron distribution function of the right electrode ( biased by @xmath173 ) in eq.([36 ] ) @xmath174 and the processes of inelastic electron tunneling to the right bank start to contribute to @xmath171 .
their contributions according to eq.([36 ] ) at low temperatures are negative and they could only diminish the increment .
we show now that due to polaronic ( franck - condon ) blockade both the `` right '' and `` left '' contributions at voltages @xmath175 are exponentially suppressed and shuttle instability takes place in the finite interval of bias volages ( we restore here the dimensions ) @xmath176 the finite series on @xmath151 in eq.([38 ] ) can be approximated as follows @xmath177 with the help of asymptotics eq.([s_l ] ) we obtain the following formula for @xmath171 at @xmath178 ( the corresponding electron - vibron coupling constant " @xmath179 ) @xmath180 for a very high biases @xmath181 it is easy to show from eqs.([38]),([s_l ] ) ( using sterling s formula to estimate asymptotics of @xmath182 ) that @xmath183\rightarrow 0 $ ] at @xmath184 .
so we see that the dependence of increment @xmath171 on the bias voltage is strongly nonmonotonic with the maximum at @xmath185 .
it is interesting to notice here that at the excitation energy @xmath186 ( the corresponding number of excited vibrons @xmath187 ) the characteristic width , @xmath188 , of the wave function of harmonic oscillator which represents quantum dot in our model ( eq.([3 ] ) ) is of the order of gap , @xmath18 , between the leads ( @xmath189 ) .
the approximation eq.([s_l ] ) does not reveal the fine structure ( on the @xmath121 scale ) of the dependence @xmath190 .
on this small energy scale one could expect the appearance of steps each time the additional vibration channel contributes to the increment ( see eq.([lm ] ) ) . at low voltages
the steps are slightly modified by the franck - condon factors . with the increase of voltage in the regime of strong coupling @xmath191 each
additional channel modifies @xmath171 by the value of the order ( few times smaller ) of @xmath171 ( sharp big steps )
. however between two sequential steps @xmath171 is diminished ( due to the @xmath192 factor in eq.([38 ] ) ) approximately by the same amount .
that is on the scale of vibron energy there are oscillations of @xmath171 . these oscillations are smeared out at temperatures @xmath193 . ) as a function of bias voltage for @xmath194 ; @xmath195 ( solid line ) and @xmath196 .
the dotted line represents the result of ref @xcite extended to the region of strong electromechanical coupling.,width=8,height=6 ] .,width=8,height=5 ] the dependence @xmath190 is plotted in figs.2,3 for the different values of parameters of our model ( for numerical calculations we used the basic eq.([38 ] ) and put the inverse temperature @xmath197 ) . for both figures we set @xmath198 , since the further decrease of @xmath199 yields the results which practically coincide with ref.@xcite in the considering region of applied voltages .
the dimensionless increment @xmath23 of the shuttle instability is measured in the units of dimensionless @xmath200 ( @xmath2 is in the units of @xmath201 ) . in fig.2
we set @xmath202 , that corresponds to the case when @xmath203 . in fig.3
we put @xmath204 , that corresponds to the case when @xmath205 .
the straight line on both figures shows the increment for the `` classical '' shuttle of ref.@xcite for the same values of all parameters .
we see from the figures , that , the dependence of increment on bias voltage is strongly nonmonotonic .
the increment @xmath23 grows until @xmath206 and decreases when @xmath207 .
the magnitudes of @xmath23 in fig.3 ( @xmath208 ) are much greater than the ones in fig.2 ( @xmath209 ) .
this means , that shuttle instability is very sensitive to the value of tunneling lengths @xmath12 .
so , when @xmath210 ( @xmath211 ) , the greater is @xmath12 the stronger is shuttle instability ( i.e. it develops on a shorter time scale @xmath212 ) . in the presence of strong mechanical friction , characterized by phenomenological damping term , @xmath213 , introduced in eq.([27 ] ) , and for @xmath214 ( i.e. @xmath215 ) , when @xmath216 , the increment at all voltages could be less then the friction coefficient @xmath217 . in this case shuttle regime of electron transport is not realized .
in this paper we have studied the influence of quantum and thermodynamical fluctuations on shuttle instability .
these fluctuations are significant in the case of strong electromechanical coupling that can be realized in electron transport through vibrating quantum dot at high bias voltages .
it was shown that the increment of shuttle instability is a nonmonotonic function of bias voltage @xmath32 with a maximum at @xmath21 , which corresponds to the region of strong electron - vibron coupling @xmath218 . at higher voltages
polaronic blockade suppresses shuttle instability .
the maximum value of increment is sensitive to the electromechanical dimensionless coupling constant @xmath9 ( @xmath11 is the electron tunneling length ) . in the presence of mechanical friction ( characterized by phenomenological friction coefficient @xmath219 ) and when @xmath216 the increment @xmath20 at all voltages could be less then friction @xmath220 and shuttle regime of electron transport is not realized . in the regime of strong electromechanical coupling @xmath221 , @xmath222 the pumping of energy in mechanical shuttle motion could overcome even strong damping .
we showed that in this regime the increment of shuttle instability strongly oscillates on the energy scale of vibron energy @xmath223 . if the friction coefficient is comparable with the amplitude of increment oscillations , the small change of bias voltage ( @xmath224 ) drastically changes the regime of electron transport through a vibrating quantum dot ( from a phonon - assisted tunneling to a shuttle regime of electron transfer and vice - a - versa ) .
one can speculate that reentrant transitions to a shuttle - like regime of electron transport will result in unusual current - voltage characteristics with pronounced negative differential conductance .
the authors thank s.i.kulinich for valuable discussions . this work was supported in parts by the swedish vr and ssf , by the faculty of science at the university of gothenburg through the `` nanoparticle '' research platform and by the grant `` effects of electronic , magnetic and elastic properties in strongly inhomogeneous nanostructures '' provided by the national academy of sciences of ukraine .
ivk thanks the department of physics at the university of gothenburg for hospitality .
to study the shuttle instability we linearize the eq.([29 ] ) ( @xmath225 , @xmath226 ) . in the linear approximation
we have @xmath227 , @xmath228 and @xmath229 , where @xmath230(t - t_{1})\right\}\ , \nonumber\\\end{aligned}\ ] ] and @xmath231\nonumber\\ \cdot \hat{q}_{j}(t_{1})\exp\left\{i\left[\varepsilon_{k}-\mu_{j}-\delta+\frac{i}{2}\gamma_{0}\right](t - t_{1})\right\}\ , \nonumber\\\end{aligned}\ ] ] here @xmath232 is the polaronic shift . then , the linearized force @xmath233 in the right - hand side of the eq.([29 ] ) is represented as the sum of two contributions : @xmath234 , where @xmath235 and @xmath236\ , \nonumber\\\end{aligned}\ ] ] to proceed further , we calculate the averages in eqs.([a6]),([a7 ] ) . in perturbation theory on the level width
@xmath7 the averages of boson and fermion operators factorize @xmath237 where @xmath238
. for noninteracting equilibrated electrons in the leads @xmath239^{-1}$ ] is the fermi - dirac distribution function .
it is easy to calculate boson correlation function in eq.([a8 ] ) , assuming vibrons to be at equilibrium at temperature @xmath1 ( this is a plausible assumption for a weak tunneling regime we are dealing with ) .
the standard calculations ( see ref.@xcite ) results in @xmath240 where @xmath241 is defined in the main text ( see eq.([34 ] ) ) .
now , by substituting eq.([a11 ] ) into the eqs.([a6]),([a7 ] ) and taking all time integrals in the limit @xmath242 , we obtain the right - hand side of eq.([29 ] ) ( i.e.@xmath243 ) as a linear functional of @xmath19 . in this case
the `` force '' term @xmath244 does not depend on time @xmath245 it determines the initial position of qd .
the linear in @xmath19 term @xmath246 takes the form of the right - hand side of eq.([31 ] ) . | shuttle - like mechanism of electron transport through a single level vibrating quantum dot is considered in the regime of strong electromechanical coupling .
it is shown that the increment of shuttle instability is a nonmonotonic function of the driving voltage .
the interplay of two oppositively acting effects - vibron - assisted electron tunneling and polaronic blockade - results in oscillations of the increment on the energy scale of vibron energy . |
pion - nucleon interaction plays a fundamental role in determining the nuclear dynamics involving pions . despite very extensive investigations in the past two decades ,
several outstanding problems remain to be solved .
for example , an accurate description of pion absorption by nuclei @xcite is still not available and hence the very extensive data for pion - nucleus reactions and pion productions from relativistic heavy - ion collisions have not been understood satisfactorily . to make progress , it is necessary to improve our theoretical description of the @xmath5 off - shell amplitude which is the basic input to most of the existing nuclear calculations at intermediate energies .
the importance of the @xmath5 off - shell t - matrix in a dynamical description of pion photoproduction has also been demonstrated @xcite in recent years .
quantum chromodynamics ( qcd ) is now commonly accepted as the fundamental theory of strong interactions . however , due to the mathematical complexities , it is not yet possible to predict @xmath5 interactions directly from qcd . on the other hand ,
models based on meson - exchange picture @xcite have been very successful in describing the @xmath11 scattering .
it is therefore reasonable to expect that the @xmath5 dynamics at low and intermediate energies can also be described by the same approach .
most of the recent attempts @xcite in this direction were obtained by applying various three - dimensional reductions of the bethe - salpeter equation for @xmath5 scattering . as is well known @xcite , the derivation of a three dimensional formulation from the bethe - salpeter equation is not unique .
it is natural to ask whether the resulting off - shell dynamics in the relevant kinematic regions depends strongly on the choice of the reduction scheme .
this question concerning the @xmath11 models was investigated @xcite quite extensively in 1970 s .
no similar investigation for the @xmath5 interactions has been made so far . in this paper
we report the progress we have made on this question . in section
ii , we specify the approximations that are used to derive a class of three - dimensional @xmath5 scattering equations from the bethe - salpeter formulation . in section iii , we define the dynamical content of the resulting meson - exchange models
. the phenomenological aspects of the models are described in section iv .
the results and discussions are presented in section v.
to illustrate the derivations of three - dimensional equations for @xmath5 scattering from the bethe - salpeter formulation , it is sufficient to consider a simple @xmath6 interaction lagrangian density @xmath12 where @xmath13 and @xmath14 denote respectively the nucleon and pion fields and @xmath15 is a bare @xmath6 vertex , such as @xmath16 in the familiar pseudo - scalar coupling . by using the standard method @xcite , it is straightforward to derive from eq .
( 1 ) the bethe - salpeter equation for @xmath5 scattering and the one - nucleon propagator . in momentum space
, the resulting bethe - salpeter equation can be written as @xmath17 where @xmath18 and @xmath19 are respectively the relative and total momenta defined by the nucleon momentum @xmath20 and pion momentum @xmath21 @xmath22 here @xmath23 and @xmath24 can be any function of a chosen parameter @xmath25 with the condition @xmath26 obviously we have from the above definitions that @xmath27 in analogy to the nonrelativistic form , it is often to choose @xmath28 and @xmath29 .
the choice of the @xmath30s is irrelevant to the derivation presented below in this section provided that eq .
( 3 ) is satisfied .
note that @xmath31 in eq .
( 2 ) is the `` amputated '' invariant amplitude and is related to the @xmath5 s - matrix by @xmath32 with @xmath33 denoting the nucleon spinor .
the driving term @xmath34 in eq .
( [ bseq ] ) is the sum of all two - particle irreducible amplitudes , and @xmath35 is the product of the pion propagator @xmath36 and the nucleon propagator @xmath37 . in the low energy region
, we neglect the dressing of pion propagator and simply set @xmath38 where @xmath39 is the physical pion mass .
the nucleon propagator can be written as @xmath40 where @xmath41 is the bare nucleon mass and the nucleon self energy operator @xmath42 is defined by @xmath43 the dressed vertex function @xmath44 on the right hand side of eq .
( [ nucl1 ] ) depends on the @xmath5 bethe - salpeter amplitude @xmath45 it is only possible in practice to consider the leading term of @xmath34 of eq .
( 2 ) . for the lagrangian eq . ( 1 ) the leading term consists of the direct and crossed @xmath1 diagrams , as illustrated in figs .
1(a ) and 1(b ) @xmath46 where @xmath47 with _
@xmath48p+k+k'$ ] . _ equations ( 2)-(11 )
form a closed set of coupled equations for determining the dressed nucleon propagator of eq .
( 6 ) and the @xmath5 bethe - salpeter amplitude of eq .
it is important to note here that this is a drastic simplification of the original field theory problem defined by the lagrangian eq .
( 1 ) . however , it is still very difficult to solve this highly nonlinear problem exactly . for practical applications , it is common to introduce further approximations .
the first step is to define the physical nucleon mass by imposing the condition that the dressed nucleon propagator should have the limit _
@xmath49 _ as @xmath50 with @xmath51 being the physical nucleon mass .
this means that the self - energy in the nucleon propagator eq .
( 6 ) is constrained by the condition @xmath52 the next step is to assume that the @xmath20-dependence of the nucleon self - energy is weak and we can use the condition eq .
( 13 ) to set @xmath53 .
this approximation greatly simplifies the nonlinearity of the problem , since the full @xmath5 propagator @xmath35 in eqs .
( 2 ) , ( 7 ) and ( 8) then takes the following simple form @xmath54 to be consistent , the driving terms eqs . ( 10 ) and ( 11 ) are also evaluated by using the simple nucleon propagator of the form of eq .
( 12 ) .
the next commonly used approximation is to reduce the dimensionality of the above integral equations from four to three .
in addition to simplifying the numerical task , this is also motivated by the consideration that the above covariant formulation is not consistent with most of the existing nuclear calculations based on the three - dimensional schroedinger formulation .
the procedure for reducing the dimensionality of the above equations is to replace the propagator @xmath35 of eq .
( 14 ) , by a propagator @xmath55 which contains a @xmath56-function constraint on the momentum variables . in the low energy region , this new propagator must be chosen such that the resulting scattering amplitude has a correct @xmath5 elastic cut from @xmath57 to @xmath58 in the complex @xmath59-plane , as required by the unitarity condition .
it is well known ( for example , see ref .
@xcite ) that the choice of such a @xmath55 is rather arbitrary . in this work ,
we focus on a class of three dimensional equations which can be obtained by choosing the following form @xmath60 \nonumber \\ & \times & \delta^{(+)}([\eta_n(s')p ' + k]^2 - m_{n}^2 ) \delta^{(+ ) } ( [ \eta_\pi ( s')p'- k]^2 - m_{\pi}^2 ) .
\label{2gh}\end{aligned}\ ] ] in the above equation , @xmath61 is the invariant mass of the @xmath5 system , and @xmath62 defines the `` offshellness '' of the intermediate states .
the superscript ( + ) associated with @xmath56-functions means that only the positive energy part is kept in defining the nucleon propagator .
the relative momentum @xmath18 in the @xmath56-functions is defined by setting @xmath63 in @xmath64 , i.e. , @xmath65 . to have a correct @xmath5 elastic cut , the arbitrary functions @xmath66 and @xmath67 must satisfy the conditions @xmath68 it is easy to verify that for @xmath69 , eqs .
( 15)-(17 ) give the correct discontinuity of the propagator @xmath55 @xmath70 & = & \frac{-i}{(2\pi)^2 } ( \eta_n(s ) p \hspace{-0.07cm}\slash / + k \slash /+m_n ) \delta^{(+)}([\eta_n(s ) p + k]^2-m_n^2 ) \nonumber \\ & \times & \delta^{(+)}([\eta_\pi ( s ) p -k]^2 - m_\pi^2).\end{aligned}\ ] ] several three dimensional formulations developed in the literature can be derived from using eqs .
( 15)-(17 ) .
these are given by blankenbecler and sugar ( @xmath71 ) @xcite , kadyshevsky ( @xmath72 ) @xcite , thompson ( @xmath73 ) @xcite , and cooper and jennings ( @xmath74 ) @xcite . in table 1
, we list their choices of the functions _
@xmath66 _ and _
@xmath67_. all schemes set @xmath75 and @xmath76 where @xmath77 and @xmath78 are the center of mass ( cm ) energies of nucleon and pion , respectively . in the rest of the paper
, we will present the formulation in the cm frame . in this frame
, we have _
@xmath79 _ for the total momentum , @xmath80 and @xmath81 .
the integral over @xmath82 in eq .
( [ 2gh ] ) can then be carried out to yield @xmath83 where @xmath84 and @xmath85 are the nucleon and pion energies , and we have defined @xmath86 .
\nonumber\end{aligned}\ ] ] replacing @xmath35 by @xmath55 in eq .
( 2 ) and performing the integration over the time component @xmath87 , we then obtain a three - dimensional scattering equation of the following form @xmath88 where @xmath89 with @xmath90 and @xmath91 .
+ substituting the @xmath92 and @xmath93 listed in table 1 into ( 23 ) , we find @xcite that the propagator of the three - dimensional scattering equation eq .
( 20 ) for each reduction scheme is 1 .
_ cooper - jennings propagator _ @xmath94.\end{aligned}\ ] ] 2 .
_ blankenbecler - sugar propagator _ @xmath95 .
\end{aligned}\ ] ] 3 .
_ thompson propagator _ @xmath96 .
\end{aligned}\ ] ] 4 .
_ kadyshevsky propagator _ @xmath97 .
\end{aligned}\ ] ] if we consistently replace @xmath35 by @xmath55 in evaluating eqs .
( [ nucl1]-[nucl2 ] ) , we then also obtain a numerically much simpler three - dimensional form @xmath98 for the nucleon self energy @xmath99 and @xmath100 for the dressed vertex function @xmath101 .
the resulting equations in the cm frame are @xmath102 accordingly , the nucleon pole condition eq .
( 13 ) becomes @xmath103 this completes the derivations of the three - dimensional formulations considered in this work .
to define the @xmath5 potential by using eq .
( 22 ) , we assume that the driving term @xmath104 is the sum of all tree diagrams calculated from the following interaction lagrangian @xmath105 \vec{t}_{\delta
n}n \cdot \partial_{\nu } \vec \pi + h.c.\ } , \label{largrang}\end{aligned}\ ] ] where @xmath106 is the rarita - schwinger field operator for the @xmath2 , @xmath107 is the isospin transition operator between the nucleon and the @xmath2 .
the notations of bjorken - drell @xcite are used in eq .
( [ largrang ] ) to describe the field operators for the nucleon @xmath1 , the pion @xmath108 , the rho meson @xmath109 , and a fictitious scalar meson @xmath4 . for @xmath110 coupling , a mixture of the scalar and vector couplings
is introduced to simulate the broad width of the s - wave correlated two - pion exchange mechanism @xcite . as illustrated in fig .
1 , the resulting driving term consists of the direct and crossed @xmath1 and @xmath2 terms , and the t - channel @xmath4- and @xmath3-exchange terms . to write down the resulting matrix elements of the @xmath5 potential , defined by eq .
( 22 ) , we introduce the following notations : @xmath111 is the four - momentum for the pion and @xmath112 for the nucleon .
the nucleon helicity is denoted as @xmath113 .
we then have ( isospin factors are suppressed here ) @xmath114 the diagrams ( a)-(b ) of fig .
1 give @xmath115 @xmath116 the @xmath4-exchange diagram fig .
1(c ) has a component from the scalar coupling and a component from the vector coupling @xmath117 while the @xmath118exchange diagram of fig .
1(d ) gives @xmath119 with @xmath120 ( p+p ' ) \cdot q , \nonumber \\ & & b_4 = \frac{\kappa_v^\rho}{2m_n}\frac{\delta-1}{4m_\rho^2}[(p - p ' ) \cdot q ] [ ( p+p ' ) \cdot q'].\end{aligned}\ ] ] the contributions from the @xmath2 excitations are depicted in diagrams of figs .
1(e ) and 1(f ) @xmath121 \nonumber \\ & & \times \frac{2m_\delta q'_{\mu'}\lambda^{\mu\nu}(p+q)q_{\nu'}}{(p+q)^2-m_\delta^2 } [ g^{\nu'\nu}-(z+\frac{1}{2 } ) \gamma^\nu \gamma^{\nu'}],\end{aligned}\ ] ] @xmath122 \nonumber
\\ & & \times \frac{2m_\delta q_{\mu'}\lambda_{\mu\nu}(p - q')q'_{\nu'}}{(p - q')^2-m_\delta^2 } [ g^{\nu'\nu}-(z+\frac{1}{2 } ) \gamma^\nu \gamma^{\nu'}],\end{aligned}\ ] ] where @xmath123 is @xmath124.\\\ ] ] the partial - wave decomposition of these potential matrix elements was discussed in detail in ref .
because of the appearance of one - particle intermediate state in of fig .
1(a ) , the @xmath5 scattering amplitude , defined by eq . ( 20 ) , in @xmath125 channel can be decomposed into a sum of pole and non - pole ( background ) terms . in the operator form
, the @xmath125 amplitude can be written as @xmath126 where @xmath127 is the bare one - nucleon state and @xmath128 , \label{gamma}\\ \sigma_n(e ) & = & < n_0|\gamma_0 g(e)\gamma^\dagger(e^*)|n_0>.\end{aligned}\ ] ] in the above equations , @xmath129 and @xmath130 denotes the bare @xmath131 vertex in fig .
@xmath132 is due to the background potential @xmath133 which is the sum of contributions ( b ) , ( c ) , ( d ) , and ( f ) of fig . 1 .
@xmath100 is the dressed @xmath6 vertex .
we follow the procedure of afnan and his collaborators @xcite to constrain the fit of @xmath125 phase shifts by imposing the nucleon pole condition eq .
this also leads to a condition which relates the bare coupling constant @xmath134 to the empirical @xmath6 coupling constant . as @xmath135
, the self - energy @xmath136 can be expressed as @xmath137 where latexmath:[\[\sigma_1(m_n)= \left .
\frac{\partial \sigma_n(e)}{\partial e } \right the above relations lead to a renormalization of the @xmath6 coupling constant .
the renormalized coupling constant @xmath139 is related to the bare coupling constant @xmath134 by @xmath140z_2^{1/2}.\end{aligned}\ ] ] where the nucleon wave function renormalization constant is given by @xmath141 the renormalized coupling constant is identified with the empirical value @xmath142 .
equations for @xmath143 channel can also be written in the form of eqs .
( 39)-(43 ) with @xmath1 replaced by @xmath2 .
to complete the model we need to introduce form factors to regularize the potential matrix elements defined by eqs .
( 28)-(38 ) . in this work
we follow pearce and jennings @xcite and associate each external leg of the potential matrix elements with a form factor of the form @xmath144^{n } , \label{ff}\ ] ] where @xmath145 with @xmath146 defined by the on - shell momentum @xmath147 of the incident energy .
it is interesting to note that as @xmath148 , @xmath149 approaches to a gaussian form .
the parameters which are allowed to vary in fitting the empirical phase shifts are : @xmath150 , and @xmath56 for the t - channel @xmath4 and @xmath3 exchanges , @xmath151 for the @xmath2 mechanisms , and the cut - off parameters @xmath152 s of the form factors of eq .
( [ ff ] ) . in the crossed @xmath1 diagram ,
the physical @xmath6 coupling constant is used . for the crossed @xmath2 diagram ,
the situation is not so clear since the determination of the `` physical '' @xmath153 coupling constant depends on the nonresonant contribution in the @xmath143 channel . in principle , it can be determined by carrying out a renormalization procedure similar to that used for the nucleon .
however , it is a much more difficult numerical task .
the complication is due to the fact that the @xmath2 pole is complex . as in ref .
@xcite , such a renormalization for the @xmath2 is not carried out in this work and we simply allow the coupling constant used in the crossed @xmath2 diagram to also vary in the fit to the data . this coupling constant is denoted as @xmath154 .
we first consider the models using rank @xmath155 form factor defined by eq .
( [ ff ] ) .
the constructed models are called @xmath156 , @xmath157 , @xmath158 , and @xmath159 for the cooper - jennings , blankenbecler - sugar , thompson , and kadyshevsky reduction schemes , respectively . for each model , we adjust the parameters described in the previous section to fit the data of @xmath5 scattering phase shifts @xcite . the results for the @xmath159 model is shown in fig .
we see that the data can be described very well .
the results of other three models are very similar in all channels except in the @xmath125 channel .
this is illustrated in fig .
the difficulty in getting the same fit to this channel is mainly due to the nucleon renormalization conditions eqs .
( 26 ) and ( 44 ) .
this difficulty is well known in the literature .
our results for the k2 , b2 and t2 are acceptable . we , however , are not able to improve the result for c2 unless we ignore the fit to other channels . the resulting parameters of the constructed four models are listed in table 2 .
we first notice that the bare @xmath6 coupling constant @xmath160 is considerably smaller than the physical value @xmath161 in all models .
this large vertex renormalization is closely related to an about 150 mev mass shift between the bare mass @xmath162 and @xmath51 , as seen in the first two rows of table 2 .
the determined physical coupling constant @xmath163 for the crossed @xmath2 term , fig .
1(f ) , is also significantly larger than the bare coupling constant @xmath164 .
the large difference between the bare mass @xmath165 mev and the resonance position @xmath166 mev seems to be a common feature of the constructed models .
the parameters associated with the @xmath3-exchange are comparable to that of other meson - exchange @xmath5 models .
the @xmath167exchange turns out to be not important in the fit .
if we set the coupling constant @xmath168 of all models to zero , the resulting phase shifts are not changed much .
this is consistent with ref .
@xcite in which the fit was achieved without including a @xmath4-exchange mechanism .
it is also interesting to note that the fit to the data seems to favor a soft @xmath6 form factor with @xmath169 mev for the models @xmath156 , @xmath157 , and @xmath158 .
the value @xmath170 mev for the model k2 is also not too hard compared with the range used in defining nucleon - nucleon potential and consistent with previous findings @xcite .
an essential phenomenology in constructing the meson - exchange models is the use of form factors to regularize the potential . to develop theoretical interpretations of the determined parameters listed in table 2 ,
it is important to investigate how the models depend on the parameterization of the form factors .
for this we also consider models with very high rank form factors defined by eq .
( [ ff ] ) with @xmath171 . as
discussed in ref .
@xcite , this very high rank form is close to the gaussian form .
we find that a fit comparable to that shown in figs . 2 and 3 can also be obtained with this parameterization of form factors .
there are some significant , though not very large , changes in the resulting parameters .
this is illustrated in table 3 in which the parameters from using @xmath155 ( t2 and k2 ) and @xmath171 ( t10 and k10 ) form factors are compared .
the constructed four models can be considered approximately phase - shift equivalent .
we therefore can examine how the resulting @xmath5 off - shell dynamics depends on the chosen three - dimensional reduction .
the @xmath5 off - shell amplitudes are needed to study nuclear dynamics involving pions . to be specific ,
let us first discuss how the constructed models can be used to investigate the near threshold pion production from nucleon - nucleon collisions .
the most important leading mechanism of this reaction is that a pion is emitted by one of the nucleons and then scattered from the second nucleon .
the matrix element of this rescattering mechanism can be predicted by using the dressed @xmath6 form factor and the half - off - shell t - matrix .
the predicted @xmath6 form factors for the near threshold kinematics , @xmath172 , are compared in fig .
4 . in fig . 5 , we compare the half - off - shell t - matrix elements in the most relevant @xmath173 and @xmath174 channels at at pion lab .
energy @xmath175 above threshold .
we see that there are rather significant differences between the considered reduction schemes at @xmath176 mev / c which is close to the momentum of the exchanged pion at the production threshold .
consequently , a study of near threshold pion production from nn collisions could distinguish the considered four different reduction schemes .
we next discuss the reactions at the @xmath2 excitation region . in figs . 6 and 7 , we show the predicted dressed @xmath177 form factor @xmath178 , defined analogously to @xmath179 of eq .
( [ gamma ] ) , and the half - off - shell t - matrix at the @xmath2 resonance energy .
these quantities are the input to the investigations of the @xmath2 excitation in pion photoproduction @xcite .
the results shown in figs . 6 and 7 suggest that the considered reduction schemes can also be distinguished by investigating the pion photoproduction reactions .
this however requires a consistent derivation of the photoproduction formulation for each reduction scheme , and is beyond the scope of this work .
the differences shown in figs . 6 and 7
can also have important consequences in determining pion - nucleus reactions in the @xmath2 region .
for instance , the constructed four models will give rather different predictions of pion double - charge reactions which are dominated by two sequential off - shell @xmath5 single - charge exchange scattering .
they can also be distinguished by investigating pion absorption which is induced by the dressed @xmath177 vertex , fig .
6 , followed by a @xmath180 transition . in summary
, we have shown that the @xmath5 scattering data up to 400 mev can be equally well described by four reduction schemes of bethe - salpeter equation .
the resulting meson - exchange models yield rather different off - shell dynamics . with the high quality data obtained in recent years
, they can be best distinguished by investigating pion productions from nn collisions and pion photoproductions .
their differences in describing pion - nucleus reactions are also expected to be significant .
our effort in these directions will be published elsewhere .
+ we thank drs .
b. pearce and c. schtz for useful communications concerning their works . c.t.h . also wishes to thank mr .
guan - yeu chen for checking part of the program .
this work is supported in part by the national science council of roc under grant no .
nsc82 - 0208-m002 - 17 and the u.s .
department of energy , nuclear physics division , under contract no.w-31 - 109-eng-38 , and also in part by the u.s .- taiwan national science council cooperative science program grant no .
int-9021617 .
h. lee and d.o .
riska , phys .
lett . * 70 * , 2237 ( 1993 ) ; c.j .
horowitz , h.o .
meyer , and d.k .
griegel , phys .
c * 49 * , 1337 ( 1994 ) ; e. hernandez and e. oset , phys .
350b * , 158 ( 1995 ) ; c. hanhart , j. haidenbauer , a. reuber , c. schutz , and j. speth , phys . lett . * 358 * , 21 ( 1995 ) .
park , f. myhrer , t. meissner , j.r .
morones , and k. kubodera , phys .
c * 53 * , 1519 ( 1996 ) ; t.d .
cohen , j.l .
friar , g.a .
miller , and u. van klock , phys .
c * 53 * , 2661 ( 1996 ) ; t. sato , t .- s .
h. lee , f. myhrer , and k. kubodera , phys .
c * 56 * , 1246 ( 1997 ) .
ccccc parameter & c2 & b2 & t2 & k2 + @xmath51 & 939 & 939 & 939 & 939 + @xmath181 & 1090 & 1072 & 1071 & 1116 + @xmath39 & 137 & 137 & 137 & 137 + @xmath182 & 1232 & 1232 & 1232 & 1232 .
+ @xmath183 & 1415 & 1412 & 1410 & 1461 + @xmath184 & 770 & 770 & 770 & 770 + @xmath185 & 654 & 662 & 654 & 654 .
+ @xmath186 & 14.3 & 14.3 & 14.3 & 14.3 + @xmath187 & 3.82 & 6.28 & 5.49 & 6.08 + @xmath188 & -0.49 & -0.37 & -0.50 & -0.39 + @xmath189 & 33.20 & -1.53 & -1.40 & -1.40 + @xmath190 & 2.54 & 2.87 & 2.87 & 2.90 + @xmath56 & 1.02 & 1.05 & 1.06 & 1.10 + @xmath191 & 0.41 & 0.31 & 0.29 & 0.34 + @xmath192 & 0.14 & 0.17 & 0.17 & 0.18 + @xmath193 & -0.14 & -0.036 & -0.075 & -0.029 + @xmath194 & 1.00 & 1.00 & 1.19 & 1.55 + + @xmath195 & 1227 & 1383 & 1321 & 1239 + @xmath196 & 417 & 704 & 681 & 648 + @xmath197 & 1521 & 1700 & 1637 & 1548 + @xmath198 & 1026 & 1555 & 1542 & 1429 + @xmath199 & 674 & 690 & 666 & 859 + ccccc parameter & t10 & t2 & k10 & k2 + @xmath51 & 939 & 939 & 939 & 939 + @xmath181 & 1065 & 1071 & 1073 & 1116 + @xmath39 & 137 & 137 & 137 & 137 + @xmath182 & 1232 & 1232 & 1232 & 1232 + @xmath183 & 1407 & 1410 & 1420 & 1461 + @xmath184 & 770 & 770 & 770 & 770 + @xmath185 & 654 & 654 & 654 & 654 + @xmath186 & 14.3 & 14.3 & 14.3 & 14.3 + @xmath187 & 5.77 & 5.49 & 6.82 & 6.08 + @xmath188 & -0.49 & -0.50 & -0.39 & -0.39 + @xmath189 & -1.52 & -1.40 & -1.43 & -1.40 + @xmath190 & 3.05 & 2.87 & 2.68 & 2.90 + @xmath56 & 0.65 & 1.06 & 1.26 & 1.10 + @xmath191 & 0.29 & 0.29 & 0.33 & 0.34 + @xmath192 & 0.17 & 0.17 & 0.18 & 0.18 + @xmath193 & -0.13 & -0.075&-0.065 & -0.029 + @xmath194 & 1.45 & 1.19 & 1.41 & 1.55 + + @xmath195 & 1300 & 1321 & 1373 & 1239 + @xmath196 & 653 & 681 & 400 & 648 + @xmath197 & 1431 & 1637 & 2272 & 1548 + @xmath198 & 1522 & 1542 & 1507 & 1429 + @xmath199 & 682 & 666 & 767 & 859 + + + | the pion - nucleon scattering is investigated by using several three - dimensional reduction schemes of the bethe - salpeter equation for a model lagrangian involving @xmath0 , @xmath1 , @xmath2 , @xmath3 , and @xmath4 fields .
it is found that all of the resulting meson - exchange models can give similar good descriptions of the @xmath5 scattering data up to 400 mev .
however they have significant differences in describing the @xmath6 and @xmath7 form factors and the @xmath5 off - shell t - matrix elements .
we point out that these differences can be best distinguished by investigating the near threshold pion production from nucleon - nucleon collisions and pion photoproduction on the nucleon .
the consequences of using these models to investigate various pion - nucleus reactions are also discussed .
-0.0 cm 15.5 cm / * meson - exchange @xmath5 models in three - dimensional bethe - salpeter formulation * + cheng - tsung hung@xmath8 , shin nan yang@xmath9 and t .- s.h .
lee@xmath10@xmath8chung - hua institute of technology , taipei , taiwan 11522 , roc + @xmath9department of physics , national taiwan university , taipei , taiwan 10617 , roc@xmath10physics division , argonne national laboratory , argonne , illinois 60439 , u.s.a . |
the celebrated volume conjecture of kashaev @xcite and murakami and murakami @xcite claims that the colored jones polynomial determines the volume of a hyperbolic knot complement .
this conjecture is wide open .
recall that the colored jones polynomial of a knot @xmath1 is a sequence of one variable laurent polynomials @xmath2 indexed by a positive integer @xmath3 , the color . for @xmath4 , the normalized version of @xmath2 , @xmath5 ,
it is the classical jones polynomial . by using results of lackenby , agol and thurston @xcite we showed in @xcite that there is a volume - ish theorem for the jones polynomial of alternating knots : the volume of a hyperbolic alternating knot is linearity bounded by the sum of the absolute values of the second leading and the second lowest coefficient of the jones polynomial @xmath6 . here
we will extend our results to the colored jones polynomial of alternating knots .
we will show that in this case the colored jones polynomial has a well defined head and tail : the leading three and the lowest three coefficients are independent of the color , for @xmath7 .
it turns out that these coefficients have a particular nice form and can easily be computed .
it is interesting to note , though , that the third coefficient of the colored jones polynomial @xmath8 is in general not determined by the jones polynomial @xmath9 .
moreover , the coefficients that gave rise to the volume - ish theorem for the jones polynomial persists in all the colored jones polynomials : the second leading coefficient and the second lowest coefficient are independent of the color . as a corollary
we will get a volume - ish theorem for the colored jones polynomial , independent of the color @xmath3 .
thus there is indeed a deeper connection between the colored jones polynomial and the volume of hyperbolic knots , at least for alternating knots .
* acknowledgment : * we are more than indebted to dror bar - natan for making his wonderful mathematica package knottheory publicly available on his web page @xcite . the colored jones polynomial routine allowed us to experimentally observe some of the phenomena that we were able to show here .
let @xmath10 be a knot diagram . for each crossing of @xmath10 , we may apply two kinds of splicings ( see figure [ ab - splicing ] ) as in the kauffman skein relation .
one splicing is called @xmath11-splicing and the other @xmath12-splicing .
a state @xmath13 of @xmath10 is a choice of @xmath11-splicing or @xmath12-splicing for each crossing of @xmath10 . after applying a state @xmath13 to splice each crossing of @xmath10
, we change @xmath10 to a diagram of a collection of disjoint simple closed curves ( circles ) in the plane .
let @xmath14 be the number of the resulting circles . also let @xmath15 ( @xmath16 , resp . )
the number of @xmath11-splicings ( @xmath12-splicings , resp . ) in the state @xmath13 . with these notations ,
the kauffman bracket of @xmath10 is @xmath17 where the sum is over all states @xmath13 of @xmath10 .
the @xmath11-graph @xmath18 of a knot diagram @xmath10 has its set of vertices equal to the set of circles in the all @xmath11-splicing state , and there is an edge joining two vertices of @xmath18 for each crossing in @xmath10 between the corresponding circles .
similarly , we define the @xmath12-graph @xmath19 of @xmath10 using the all @xmath12-splicing state . to the best of our knowledge ,
the notion of @xmath11-graph and @xmath12-graph of a knot diagram @xmath10 was first introduced by a. stoimenow ( @xcite , compare with @xcite ) .
a knot diagram @xmath10 is called @xmath11-adequate ( @xmath12-adequate , resp . ) if there are no loops , i.e. edges that start and terminate at the same vertex , in the @xmath11-graph @xmath18 ( the @xmath12-graph @xmath19 , resp . ) .
a knot diagram @xmath10 is called adequate if it is both @xmath11-adequate and @xmath12-adequate .
it is easy to see that a reduced alternating knot diagram is adequate .
in fact , in this case the @xmath11- and the @xmath12-graph are the two checkerboard graphs of the knot diagram .
a diagram @xmath10 of a knot is @xmath11-adequate if and only if the corresponding diagram @xmath20 of its mirror image is @xmath12-adequate .
furthermore , the kauffman bracket of the mirror image @xmath20 equals the kauffman bracket of @xmath10 after substituting @xmath21 for the variable @xmath11 .
therefore , all theorems on @xmath11-adequate knots have a corresponding formulation for @xmath12-adequate knots . in the sequel
we will frequently omit this formulation to make the theorems more readable .
the reduced @xmath11-graph @xmath22 of a knot diagram is obtained from the @xmath11-graph @xmath18 by keeping the same set of vertices but reduce all multiple edges to one for each pair of vertices .
so a knot diagram @xmath10 is @xmath11-adequate if and only if @xmath22 is a simple graph , i.e. does not have loops or multiple edges . similarly , we have the reduced @xmath12-graph @xmath23 and @xmath10 is @xmath12-adequate if and only if @xmath23 is a simple graph .
let @xmath10 is a knot diagram of @xmath24 crossings .
the number of vertices of @xmath22 , which equals the number of vertices in the unreduced graph @xmath18 , is denoted by @xmath25 and the number of edges of @xmath22 by @xmath26 .
the following theorem is a generalization of our corresponding result in @xcite that we showed for alternating knots : ( stoimenow)[stoim ] suppose @xmath10 is @xmath11-adequate .
then we have @xmath27 recall that @xmath26 is the number of edges of the reduced graph @xmath22 .
let @xmath28 be the multiplicity of the @xmath29-th edge of @xmath22 in @xmath18 , for @xmath30 . for a state @xmath13 of @xmath10 where there are at least two @xmath12-splicings appearing at edges of @xmath18 which reduce to different edges in @xmath22 ,
its contribution to @xmath31 is @xmath32 where @xmath33 .
now @xmath34 so we only need to consider states whose @xmath12-splicings all appear on the same @xmath28 multiple edges of @xmath18 .
the contribution of such states in @xmath31 is @xmath35 this proves the theorem .
note that when the graph @xmath22 is connected , we have @xmath36 this will be important to us . [ cab - b-1 ]
let @xmath10 be an @xmath11-adequate ( resp .
@xmath12-adequate ) knot diagram and @xmath37 be its blackboard @xmath3 cabling . then @xmath37 is @xmath11-adequate ( resp .
@xmath12-adequate ) .
furthermore , we have @xmath38 the first claim was shown by lickorish ( see e.g. @xcite ) .
the following picture depicts the @xmath11-splicing of the blackboard cabling of a knot diagram @xmath10 : if @xmath10 is @xmath11-adequate , then @xmath22 is a simple graph . for a vertex @xmath39 of @xmath22 ,
let the edges coming out of @xmath39 be @xmath40 . then in @xmath41
, we have @xmath3 vertices associated with @xmath39 : @xmath42 , and @xmath43 edges @xmath44 .
the edges @xmath40 of @xmath22 are still in @xmath41 as edges coming out of @xmath45 .
thus , @xmath41 is equal @xmath22 with a chain @xmath46 attached to each vertex @xmath47 .
so @xmath41 is still a simple graph and @xmath37 is @xmath11-adequate .
the second claim follows easily from the structure of @xmath41 described above .
we will use the chebyshev basis of the kauffman bracket skein module of @xmath48 $ ] to express the colored jones polynomial .
see e.g. @xcite .
let @xmath49 , @xmath50 , be polynomials of @xmath51 specified by the following recurrence relation and initial values : @xmath52 inductively , we have @xmath53 suppose that @xmath10 is a knot digram of a knot @xmath1 .
let @xmath2 be the colored jones polynomial of @xmath1 corresponding to the @xmath3 dimensional irreducible @xmath54 module . here
, @xmath55 is the classical jones polynomial multiplied by @xmath56 .
we have @xmath57^{-w(d)}\,(-1)^{n-1}\,[2]\ , \langle s_n(d)\rangle,\ ] ] where @xmath58=a^2+a^{-2}$ ] , @xmath59 is the writhe of @xmath10 , and @xmath60 is a linear combination of blackboard cablings of @xmath10 obtained using the chebyshev polynomial @xmath49 . by equation ( [ cheby ] )
, we have @xmath61 for the unknot diagram @xmath62 , we have @xmath63.\ ] ] the normalized colored jones polynomial is @xmath64\ ] ] and @xmath6 is the classical jones polynomial .
we are interested in the leading coefficients of @xmath65 .
we write @xmath66 thus @xmath67\\ & = ( aa^k+ba^{k-4}+ca^{k-8}+\cdots)(a^{2n-2}+a^{2n-6}+\cdots)\\ & = aa^{k+2n-2}+(a+b)a^{k+2n-6}+(a+b+c)a^{k+2n-10}+\cdots \end{aligned}\ ] ] suppose now that @xmath10 is an @xmath11-adequate knot diagram . since all blackboard cablings of @xmath10 are adequate , using theorem [ stoim ] , lemma [ cab - b-1 ] , and equation ( [ cheby2 ] ) , we have @xmath68 compare the second highest degree of @xmath69 and the highest degree of @xmath70 : @xmath71 so , @xmath72 finally , we have @xmath73\,\langle s_n(d)\rangle&=(a^2+a^{-2})\langle s_n(d)\rangle\\ & = ( -1)^{nv-1}a^{n^2c+2nv}+((-1)^{nv-1}+(-1)^{nv-2}\beta_1(a(d)'))a^{n^2c+2nv-4}\\ & \quad+ \text{lower degree
terms}. \end{aligned}\ ] ] let @xmath10 be an @xmath11-adequate knot diagram .
write @xmath74 then , we have @xmath75 this follows directly from the previous calculations .
* remark : * note that the difference between highest degrees of @xmath69 and @xmath70 , respectively , is @xmath76 .
so when @xmath3 is getting large , any fixed leading portion of coefficients of @xmath5 will only depend on the coefficients of @xmath77 .
for the first and second leading coefficients of @xmath69 , we have shown that they depend only on @xmath3 by a sign when @xmath1 has an @xmath11-adequate diagram .
we consider the third leading coefficient of @xmath69 in the next section .
the difficulties in computing the third leading coefficient are that it does not depend on the topological type of the @xmath11-graph and its reduced graph anymore .
informations on the embedding of the @xmath11-graph are needed .
we first will extend theorem [ stoim ] .
suppose that a knot diagram @xmath10 is @xmath11-adequate .
the vertices of @xmath18 are represented by disjoint circles in the plane .
let @xmath78 and @xmath79 be two different edges in @xmath22 .
we say that they are _ disjoint _ if @xmath80 .
if @xmath78 and @xmath79 share a common vertex @xmath39 , then there are two possibilities . in one case
, the circle representing @xmath39 can be divided into two semi - circles such that vertices of multiple edges @xmath78 and @xmath79 lie on different semi - circles respectively .
we say that @xmath78 and @xmath79 are _ separated _ at @xmath39 in this case .
otherwise , we say that @xmath78 and @xmath79 are _ mixed _ at @xmath39 .
see figure 3 , in which the pairs @xmath81 are @xmath82 are mixed , and the pairs @xmath83 and @xmath84 are separated .
suppose @xmath78 and @xmath79 are two edges of @xmath22 mixed at a vertex @xmath39 . then in @xmath18 , the multiple edge @xmath78 ( @xmath79 , resp . )
is partitioned into sets of @xmath85 ( @xmath86 , resp . )
parallel multiple edges , so that the endpoints of these edges on the circle @xmath39 are placed alternatively as @xmath87 .
we must have @xmath88 .
see figure 4 .
if @xmath10 is a reduced alternating knot diagram , we do not have such pairs of edges @xmath78 and @xmath79 in @xmath22 .
[ mixededges ] [ theorem41 ] suppose @xmath10 is @xmath11-adequate .
then we have @xmath89 where @xmath90 is the number of edges in @xmath22 whose multiplicity in @xmath18 is larger than 1 , @xmath91 is the number of pairs of edges in @xmath22 mixed at a vertex , and @xmath92 is the number of triangles in @xmath22 . when @xmath10 is a reduced alternating knot diagram , then @xmath93 .
so this theorem generalized the formula of @xmath31 in @xcite for reduced alternating knot diagrams .
let @xmath13 be a state of @xmath10 with at least three @xmath12-splicings such that ( 1 ) they appear at edges of @xmath18 which reduce to three different edges in @xmath22 , and ( 2 ) these three edges of @xmath22 do not form a triangle in @xmath22 .
then the contribution of @xmath13 to @xmath31 is @xmath94 where @xmath95 .
now @xmath96 so we only need to consider the following four cases .
* case 1 : * states whose @xmath12-splicings all appear on the same @xmath28 multiple edges of @xmath18 .
the total contribution of such states in @xmath31 is @xmath97 the last equation follows from : @xmath98 * case 2 : * states whose @xmath12-splicings appear in a pair of distinct multiple edges , which are either disjoint or separated .
the total contribution of such states in @xmath31 is @xmath99 * case 3 : * states whose @xmath12-splicings appear in a pair of distinct multiple edges @xmath100 which are mixed at a vertex .
this is the most delicate of the four cases .
we claim that there is no contribution to the three leading coefficients that involve @xmath12-splicings at both edges .
suppose the @xmath11-diagram is locally similar to figure [ mixededges ] , i.e. @xmath101 with @xmath102 and @xmath103 for @xmath104 .
we will deal with the following subcases : 1 .
[ subcase1 ] one of the @xmath105 or one of the @xmath106 is greater than @xmath107 .
assume that @xmath108 .
in this case one can see that the three leading coefficients of @xmath109 equal the three leading coefficients of @xmath110 .
therefore , to show the claim we can assume that @xmath111 .
[ subcase2 ] if @xmath112 , i.e. our diagram is @xmath113 , a straightforward computation shows that the contribution of the type of @xmath12-splicings we consider here vanishes .
we are left with the case of @xmath114 .
we compare the leading terms of @xmath115 and @xmath116 + yet another direct , straightforward computation shows that the three leading coefficients coincide .
thus we can reduce this case to ( [ subcase1 ] ) and ( [ subcase2 ] ) .
thus we showed that the sum of all states with @xmath12-splicings at a pair of mixed edges does not contribute to the leading three coefficients .
recall that in our notation @xmath91 is the number of pairs of edges in @xmath22 that are mixed . to summarize case 2 and case 3 : the total contribution of states whose @xmath12-splicings appear in a pair of distinct multiple edges is @xmath117 * case 4 : * states with three @xmath12-splicings appearing on three different edges in @xmath22 that form a triangle .
note , that for reasons of planarity the edges can not be mixed at any of the three vertices .
the total contribution of such states is @xmath118 this proves the theorem .
similar to the discussion in 3 , we need to compare first the third highest degree of @xmath69 and the highest degree of @xmath70 : @xmath119 since we have @xmath120 and @xmath121 for non - trivial cases .
thus , we just need to consider the third leading coefficient of @xmath69 .
denote by @xmath122 the number of vertices of @xmath41 , @xmath123 the number of edges in @xmath41 , @xmath124 the number of edges in @xmath125 whose multiplicity in @xmath125 is larger than 1 , @xmath126 the number of pairs of edges in @xmath41 mixed at a vertex , and @xmath127 the number of triangles in @xmath41 .
suppose the @xmath11-adequate diagram @xmath10 contains no kinks , or @xmath18 has no vertices of valence 1 .
we have : @xmath128 , @xmath129 , @xmath130 when @xmath131 , and @xmath132 . furthermore , if @xmath10 is a reduced alternating knot diagram and @xmath131 , then @xmath133 .
this is clear from the structure of @xmath41 .
the relevant terms in theorem ( [ theorem41 ] ) for the third coefficient are now simplified with the following lemma : if @xmath10 is a reduced alternating knot diagram and @xmath131 , we have @xmath134 thus , we have finally arrived at our main theorem .
[ main theorem ] let @xmath1 be an alternating knot .
write @xmath135 with positive @xmath136 and @xmath137 .
let @xmath18 and @xmath19 be the @xmath11- and @xmath12-graphs of a reduced alternating diagram @xmath10 of @xmath1 with crossing number @xmath138 .
the reduced graphs @xmath22 and @xmath23 have @xmath139 and @xmath140 edges and @xmath141 and @xmath142 vertices .
note , that @xmath143 .
furthermore , there are @xmath144 and @xmath145 triangles in @xmath22 and @xmath23
. then 1 .
[ span ] the span @xmath146 of @xmath5 , i.e. the difference of the highest and lowest exponent , in the variable @xmath147 is : @xmath148 2 .
@xmath149 3 .
@xmath150 and @xmath151 .
4 . @xmath152 and @xmath153 for @xmath154 .
the coefficients @xmath155 and @xmath156 were identified in @xcite : @xmath157 where @xmath158 ( or @xmath159 ) is the number of edges in @xmath160 ( or @xmath161 ) that correspond to edges of multiplicity at least @xmath162 in the unreduced graph @xmath18 ( or @xmath19 ) .
in particular , @xmath163 and @xmath164 are independent of @xmath3 when @xmath131 and @xmath165 and @xmath166 are independent of @xmath3 when @xmath167 .
item ( [ span ] ) was essentially shown by kurpita and murasugi in @xcite and le in @xcite .
the highest degree of @xmath168 in @xmath11 is @xmath169 and the lowest degree is @xmath170 .
thus for the span of @xmath168 in the variable @xmath11 we get @xmath171 this implies that the span of @xmath172 is @xmath173 and thus the span of @xmath174 in the variable @xmath11 is @xmath175 changing to the variable @xmath176 this shows claim ( [ span ] ) . for the remaining claims : recall that we showed @xmath177 for some number @xmath178 and @xmath179 .
the remaining claims now follows from a comparison of the coefficients of @xmath168 and @xmath180 and the fact that the @xmath11-graph of the knot is the @xmath12-graph of the mirror image of the knot and vice versa .
as an example we pick the alternating knots @xmath181 and @xmath182 in the knotscape census @xcite .
the two knots are given in figure [ figure 12_217 ] . according to knotscape both knots share the same jones polynomial @xmath6 .
[ figure 12_217 ] we have for the knot @xmath181 : the @xmath11-graph is the checkerboard graph corresponding to the color not containing the outer face .
its reduced graph has @xmath183 edges and @xmath184 vertices .
furthermore , it contains @xmath185 triangles .
thus the three leading coefficients are , up to a common sign change : @xmath186 and @xmath187 similarly : @xmath188 and @xmath189 .
thus the last three coefficients are : @xmath190 , up to a common sign .
for the knot @xmath182 the data are : @xmath191 and we get for the leading coefficients @xmath192 and for the last three coefficients : @xmath190 . again
, these two lists are up to sign changes .
since the jones polynomials of the two knots coincide but already the third coefficient of the colored jones polynomials for color @xmath167 does not , we see : for @xmath167 the leading third coefficient of the colored jones polynomial @xmath5 is in general not determined by the jones polynomial @xmath6 .
as a corollary to the main theorem [ main theorem ] and the results in @xcite we also get the volume - ish theorem for the colored jones polynomial : | the colored jones polynomial is a function @xmath0 $ ] associated with a knot @xmath1 in 3-space .
we will show that for an alternating knot @xmath1 the absolute values of the first and the last three leading coefficients of @xmath2 are independent of @xmath3 when @xmath3 is sufficiently large .
computation of sample knots indicates that this should be true for any fixed leading coefficient of the colored jones polynomial for alternating knots . as a corollary
we get a volume - ish theorem for the colored jones polynomial . |
there is a lot of interest in the study of non - perturbative systems , which can not be solved by exploiting the existence of a small expansion parameter . in this paper
we discuss two formalisms that have been proposed to address non - perturbative problems : @xmath0-particle irreducible ( @xmath0pi ) effective theories , and the exact renormalization group ( rg ) .
the @xmath0pi formalism has been used to study finite temperature systems and non - equilibrium dynamics ( see @xcite and references therein ) , and transport coefficients @xcite .
the exact rg has been applied to a variety of problems @xcite .
we show that the 4pi effective action produces two bethe - salpeter ( bs ) equations that can be used to truncate the rg flow equations at the level of the second equation , and that the resulting flow equations for the 2- and 4-point functions are total derivatives whose integrals give the 4pi eom s .
this result is surprising ( since the full hierarchy of rg flow equations are obtained using a single bi - local source term ) , and suggests that a bs truncation at arbitrary orders produces equations whose integrals give the @xmath0pi eom s .
we work with a scalar field theory with quartic coupling and consider only the symmetric case where the expectation value of the field is zero .
we use a compactified notation in which the space - time coordinates are represented by a single numerical subscript .
we also use an einstein convention in which a repeated index implies an integration over space - time variables .
the @xmath0pi effective action is obtained by taking the @xmath0th legendre transform of the generating functional which is constructed by coupling the field to @xmath0 source terms of the form @xmath1 .
the effective action has the form @xmath2 + \frac{i}{2 } { \rm tr } \,{\rm ln}g^{-1 } + \frac{i}{2 } { \rm tr}\left [ \left(g^0\right)^{-1 } g\right ] -i\phi_2 \,,\nonumber\end{aligned}\ ] ] where @xmath3 contains all contributions which have two or more loops .
for example , @xmath3 for the 4-loop 4pi effective action in the symmetric theory @xcite is shown in fig . [
fig : phiaandb-2 ] . for the 4-loop 4pi
effective action .
[ fig : phiaandb-2],width=264 ] the self consistent propagator and vertex are obtained through the variational principle by solving the equations produced by taking the functional derivative of the effective action and setting the result to zero .
we define kernels : @xmath4 the factors @xmath5 and @xmath6 indicate the number of @xmath7 s and @xmath8 s with respect to which the functional derivative is taken .
the inverse propagators truncate the legs that are left behind by the functional derivatives with respect to @xmath8 , and they are moved through the derivative in ( [ lam0nl ] ) only so that the equation is easier to write - functional derivatives do not act on legs by definition .
the definition of the kernel @xmath9 also excludes a disconnected piece that contains only inverse propagators .
it is well known that the 2pi effective action gives a bs equation for a diagonal 4-point function .
writing the effective action as a functional of the form @xmath10 $ ] we obtain : @xmath11 calculating the derivative gives @xmath12 substituting ( [ gderr2 ] ) into ( [ side1 ] ) and using the kernel definition ( [ lam0nl ] ) produces : @xmath13 higher order bs equations are calculated the same way @xcite . using the 4pi effective action ( [ side1 ] ) becomes : @xmath14 the method used to obtain ( [ gderr2 ] ) gives : @xmath15\,,\nonumber\end{aligned}\ ] ] and combining pieces we obtain the result in the first line of fig .
[ fig : bs - combo ] . , width=529 ] similarly we obtain a bs equation for a 6-point function from @xmath16 : @xmath17 substituting the derivatives and kernels gives , after a long calculation , the second line in fig .
[ fig : bs - combo ] @xcite .
the rg is constructed by building a family of theories indexed by a continuous parameter @xmath18 , with the dimension of a momentum , such that fluctuations are smoothly taken into account as @xmath18 is lowered from the microscopic scale @xmath19 ( at which the couplings are defined ) down to zero .
we add to the original action a non - local term which is quadratic in the fields @xmath20 : @xmath21 the function @xmath22 approaches zero for @xmath23 , so that modes @xmath24 are unaffected , and @xmath25 for @xmath26 , which suppresses modes @xmath27 by giving them a mass @xmath28 . generating functionals are defined in the usual way but with the action @xmath29 replaced by @xmath30 .
we obtain ( @xmath31 , @xmath32 ) : @xmath33 functionally differentiating ( [ gkappa - def ] ) with respect to @xmath34 produces the exact rg flow equations which form an infinite coupled hierarchy .
the first two flow equations are ( see fig .
[ fig : renormgroup ] ) : @xmath35 the factor ( 6 ) in eq .
( [ flow2 ] ) and fig .
[ fig : renormgroup ] is a short - hand notation which means that there are 6 permutations of external legs .
we use numerical factors in brackets in equations ( figures ) to represent additional terms that correspond to permutations of external indices that are not written ( drawn ) . ) and ( [ flow2 ] ) .
[ fig : renormgroup ] , width=377 ] in order to do calculations , the infinite hierarchy of rg flow equations must be truncated .
this is a common feature of non - perturbative methods , and often leads to difficulties ( see for example @xcite ) .
the hierarchy of rg flow equations can be truncated at the level of the first equation using the bethe - salpeter ( bs ) equation derived from the 2pi effective action @xcite .
we can perform the truncation at the 4pi level using the bs equations in fig .
[ fig : bs - combo ] .
we extend these equations to the deformed theory by using kernels obtained from functional derivatives of the 4pi effective action which are subsequently evaluated at @xmath36 .
diagrammatically , we substitute the bs equations in fig . [
fig : bs - combo ] into the tadpole diagrams in fig .
[ fig : renormgroup ] . after a long calculation
this produces the diagrams in figs .
[ fig:22truncation-4pi ] and [ fig : bs24truncation - all ] @xcite .
we represent the sum of terms @xmath37 by a small grey dot ( red on - line ) .
is substituted into the tadpole diagram in the first part of fig .
[ fig : renormgroup ] .
[ fig:22truncation-4pi],width=226 ] is substituted into the tadpole diagrams in the second part fig .
[ fig : renormgroup ] .
[ fig : bs24truncation - all],width=302 ] we compare our results for @xmath38 and @xmath39 with the derivatives @xmath40 and @xmath41 using the 4pi equations of motion extended to the deformed theory .
@xmath42 equation ( [ sigma-4pi - b ] ) is precisely the result that is shown in fig .
[ fig:22truncation-4pi ] if we identify @xmath43 , which means we have @xmath44 for the 4-point function a similar calculation gives @xcite .
@xmath45 equations ( [ 2result ] ) and ( [ 4result ] ) show that the rg equations for the 2- and 4-point functions are total derivatives whose integrals can be written as the 4pi equations of motion .
we have shown that the bs truncation produces flow equations that are total derivatives with respect to the flow parameter , and that the integrals of the flow equations give the equations of motion of the 4pi effective theory .
this establishes a direct connection between two non - perturbative methods . for the @xmath0pi formalism
, there could be a practical advantage in reformulating the integral equations as flow equations , because initial value problems are usually easier to solve than non linear integral equations .
furthermore , regarding the vertices from the @xmath0pi effective theory as flow equations gives new insight into the problem of how to renormalize the @xmath0pi effective theory for @xmath46 .
j. p. blaizot , e. iancu , and a. rebhan , phys .
lett . * 83 * , 2906 ( 1999 ) ; phys .
d 63 * , 065003 ( 2001 ) ; j. berges , sz .
borsnyi , u. reinosa , and j. serreau , phys .
rev.*d 71 * , 105004 ( 2005 ) .
j. berges and j. cox , phys .
* b 517 * , 369 ( 2001 ) ; j. berges , nucl .
phys . * a 699 * , 847 ( 2002 ) ; g. aarts and j. berges , phys .
lett . * 88 * , 041603 ( 2002 ) ; g. aarts , d. ahrensmeier , r. baier , j. berges , and j. serreau , phys . rev . *
d 66 * , 045008 ( 2002 ) .
g. aarts and j. m. martnez resco , jhep * 02 * , 061 ( 2004 ) .
carrington and e. kovalchuk , phys.rev .
* d77 * , 025015 ( 2008 ) ; ibid , phys .
rev . * d80 * , 085013 ( 2009 ) ; ibid , phys
* d81 * , 065017 ( 2010 ) . c. wetterich , phys .
lett . , * b301 * , 90 ( 1993 ) ; u. ellwanger , z. phys . * c58 * , 619 ( 1993 ) ; n. tetradis and c. wetterich , nucl . phys . *
b422 * , 541 ( 1994 ) ; t.r .
morris , phys .
b329 * , 241 ( 1994 ) ; j. berges , n. tetradis and c. wetterich , phys . rept . * 363 * , 223 ( 2002 ) ; j. m. pawlowski , annals phys . * 322 * 2831 , ( 2007 ) . m.e .
carrington , eur .
j. * c35 * , 83 ( 2004 ) ; j. berges , phys . rev . *
d70 * , 105010 ( 2004 ) .
carrington , phys . rev . *
d87 * , 045011 ( 2013 ) m.e .
carrington , weijie fu , t. fugleberg , d. pickering , i. russell , phys . rev . *
d88 * , 085024 ( 2013 ) .
a. arrizabalaga and j. smit , phys . rev .
* d66 * , 065014 ( 2002 ) ; m.e .
carrington , g. kunstatter and h. zaraket , eur .
j. * c 42 * , 253 ( 2005 ) ; d. binosi and j. papavassiliou , physics reports , * 479 * , 1 ( 2009 ) .
blaizot , phil .
r. soc . * a 369 * , 2735 ( 2011 ) ; j.p .
blaizot , j.m .
pawlowski , u. reinosa , phys . lett .
* b 696 * , 523 ( 2011 ) . | the 4pi effective action provides a a hierarchy of integral equations which have the form of bethe - salpeter equations .
the vertex functions obtained from these equations can be used to truncate the exact renormalization group flow equations .
this truncation has the property that the flow is a total derivative with respect to the flow parameter and is equivalent to solving the @xmath0pi equations of motion .
this result establishes a direct connection between two non - perturbative methods . |
light is known as fastest carrier of energy and information and this property makes light indispensable for communications , warfare and fundamental research . while light can be relatively easy created and guided , its manipulation is somewhat difficult .
manipulation of the light requires dynamical control of the refractive index of the host medium . while the materials for active control of the light are still under development , the theoretical studies are already started .
recently , the light manipulation was investigated in work @xcite where the weak scattering by the finite object was studied by using the first born approximation . in practice , the light scattering by photonic cluster may be not weak and it should be studied with other methods . the local perturbation method ( lpm ) is suitable tool for such analysis .
the lpm correctly describes scattering by particles with arbitrary large refractive index which are small compared to the incident wavelength and the scattering can be strong ( see for example work @xcite-@xcite and references wherein ) . in works
@xcite-@xcite the lpm was used to study the wave propagation in the photonic cluster . in this paper
we optimize the light scattering from the cluster made of small particles by using the lpm and special optimization technic . by using our method we modify the cluster in such a way that the scattering from the cluster is significantly minimized or maximized at one point .
we present several examples which will show that the scattering by the cluster can be increased and decreased by several orders of magnitude for selected wavelength and direction .
the formalism we use is presented in many works ( see for example @xcite-@xcite ) and we only briefly present it here for convenience and consistency . consider the photonic cluster made of particles which characteristic sizes are small compared to the incident wavelength @xmath0 . the electric field @xmath1 propagating in the host medium filled with @xmath2 particles is described by the following equation @xcite @xmath3 where @xmath4 here @xmath5 is laplacian and @xmath6 nabla operators , @xmath7 defines tensor product , @xmath8 is a wave number in the host medium ( @xmath9 is the angular frequency and @xmath10 is the light velocity in vacuum ) , @xmath11 and @xmath12 are the permittivity of the @xmath13-th particle and the medium respectively , @xmath14 is the function describing the shape of the @xmath13-th scatterer , and @xmath15 is the field source .
the characteristic size of @xmath13-th scatterer we denote as @xmath16 .
note , that the equation ( [ inv11 ] ) is an approximate one and this is true only when the small scatterers ( @xmath17 ) are considered .
this equation is easily solvable with respect to the fields @xmath18 when the positions @xmath19 are known ( it will invoke solution of @xmath20 linear equations for each frequency @xmath21 .
the solution of the equation ( [ inv11 ] ) can be written in the following form @xmath22 where @xmath23 and @xmath24 here @xmath25 is the @xmath26 unitary tensor in polarization space and @xmath19 is the radius vector of the n - th particle .
the field @xmath27 is the field inside the @xmath13-th particle , @xmath28 is the fourier transform of the function @xmath14 , and @xmath29 defines scalar product .
the incident field @xmath30 is created by the source @xmath15 in the host medium and it is not important for our consideration .
the formula ( [ inv12 ] ) is rather general one and it describes the field in the medium with photonic cluster of arbitrary form made of small particles of arbitrary form .
the fields @xmath31 should be found by solving the system of @xmath32 linear equations obtained by substituting @xmath33 into eq .
( [ inv12 ] ) .
the formula for the scattered field ( [ inv13 ] ) can be simplified even further when the distance between the observer and an @xmath13-th scatterer ( @xmath34 ) is large , i.e. when @xmath35 . in this case
the integral @xmath36 can be calculated approximately .
we note also that the integral @xmath36 can be calculated exactly at least for the spherical particles .
when @xmath37 integration in ( [ inv13 ] ) gives @xmath38 where @xmath39 here @xmath34 is the distance between the observation point and @xmath13-th scatterer , @xmath40 is the volume of the @xmath13-th scatterer , and @xmath41 brackets denote absolute value . in practice ,
the distance between the cluster and the observer is much more larger than the size of the cluster and the inequality @xmath42 is fulfilled . in this case
the field ( [ inv15 g ] ) can be simplified and rewritten in the form @xmath43 where @xmath44
the idea of our approach is the following .
we assume that scatterers in the cluster can be repositioned in a such a way that the field scattered by the cluster will be tuned to required value at the observation point . at first
, we calculate the intensity of the scattered field from the given ( not modified ) cluster for selected wavelength @xmath45 and observation point @xmath46 . after this
we select one particle in the cluster and reposition it a bit in @xmath47 , @xmath48 , and @xmath49 directions ( usually for tenth of the particle size ) recalculating the scattered field in @xmath46 each time we move the particle . from the array of the calculated intensities we select the value nearest to the required one and we place the particle at the point corresponding to this intensity value .
we perform this procedure for all particles in the cluster and repeat it several times .
after several iterations , the particles in the cluster will be rearranged in such a way that the intensity of the scattered light will correspond the required value at the selected observation point @xmath46 and for the wavelength @xmath50 . versus @xmath51 for the not optimized @xmath52 cluster .
the particles in the cluster are arranged into cubic lattice .
the permittivity of the host medium is @xmath53 , the characteristic size of the particles is @xmath54 nm , the period of the cluster is @xmath55 , and the total number of the particles is @xmath56 .
the source is positioned at @xmath57 and the observation points @xmath58 are located at the circle with radius @xmath59 .
the angle @xmath60 is the angle between vector @xmath46 and @xmath61 in @xmath62 plane.,width=340 ] for example , super reflective cluster can be created when the intensity of the scattered field @xmath63 will be maximized at some point and minimized at all other points ( directions ) .
another important example closely related to optical cloaking is the cluster with significantly reduced reflectivity for selected wavelength at chosen direction . when the reflectivity of the cluster is lower than the sensitivity level of the receiver , this cluster is actually invisible for observer .
moreover , when the intensity of the field scattered by the cluster is changed in some direction , this can be used in directional optical switch .
this kind of device can be used for light houses , for example . when the spectrum of the field scattered by the cluster is modified in time when the cluster can be used as a filter .
below we present several examples demonstrating the possibility to optimize the cluster reflectivity in chosen direction for selected wavelength .
we note that similar approach was successfully used in work @xcite where the phase of incident field was tuned to maximize the field scattered from the complex object ( eggshell and tio powder ) in chosen direction . versus @xmath51 for the optimized spherical cluster made of @xmath52 cubes .
the permittivity of the host medium is @xmath64 , the characteristic size of the cubes is @xmath54 nm and the total number of the particles is @xmath56 .
the source is positioned at @xmath65 and the observation points @xmath46 are located at the circle with radius @xmath59 .
the angle @xmath66 is the angle between vector @xmath46 and @xmath67 in @xmath62 plane .
the cluster reflectivity was minimized for @xmath68 at @xmath69 .
the red line shows extremely low deep in cluster reflectivity after the optimization.,width=340 ] consider the spherical photonic cluster consisting of @xmath52 cubes .
the characteristic size of the cubes is @xmath70 and they are organized into simple cubic lattice with period @xmath71 . the center of the cluster is positioned at the beginning of the coordinates .
the incident field is generated by the point source positioned at the point @xmath57 and it is linearly polarized in @xmath48 direction
. the scattering by the cluster will be optimized ( to minimum or maximum ) for the chosen wavelength @xmath72 for the observer positioned at the point @xmath46 .
we assume that the permittivity of the scatterers is the same as the permittivity of the bulk @xmath52 and the actual values of permittivity were taken from @xcite .
note that the source and the observer are far from the cluster such that @xmath73 and @xmath74 . versus @xmath51 for the optimized spherical cluster made of @xmath52 cubes .
the permittivity of the host medium is @xmath64 , the characteristic size of the cubes is @xmath54 nm and the total number of the particles is @xmath56 .
the source is positioned at @xmath65 and the observation points @xmath46 are located at the circle with radius @xmath59 .
the angle @xmath66 is the angle between vector @xmath46 and @xmath67 in @xmath62 plane .
the cluster reflectivity was maximized for @xmath68 at @xmath69 .
the red line shows extremely high peak in cluster reflectivity after the optimization.,width=340 ] the examples of the optimization are presented on figs .
[ fig1]-[fig5 ] .
the figures show the normalized intensity of the light scattered by the clusters versus @xmath51 .
[ fig1 ] shows the light intensity scattered from the not optimized cluster . figs .
[ fig2 ] and [ fig3 ] show the results for the clusters which scattering was optimized at the point @xmath75 and the figs .
[ fig4 ] and [ fig5 ] show the results for the clusters which scattering was optimized at the point @xmath76 . versus @xmath51 for the optimized spherical cluster made of @xmath52 cubes .
the permittivity of the host medium is @xmath64 , the characteristic size of the cubes is @xmath54 nm and the total number of the particles is @xmath56 .
the source is positioned at @xmath65 and the observation points @xmath46 are located at the circle with radius @xmath59 .
the angle @xmath66 is the angle between vector @xmath46 and @xmath67 in @xmath62 plane .
the cluster reflectivity was minimized for @xmath68 at @xmath77 .
the red line shows extremely low deep in cluster reflectivity after the optimization.,width=340 ] the positions of the particles in the clusters were changed in order to maximize or minimize the scattering for the wavelength @xmath78 nm ( @xmath68 ) in chosen directions .
the results are presented for different angles of observation @xmath79 in @xmath62 plane . the fig .
[ fig2 ] shows normalized ( with respect to the intensity of the incident field @xmath80 ) intensity of the light @xmath81 scattered by the optimized photonic cluster .
the cluster was optimized to minimal intensity of the scattered field in the direction of the source ( @xmath82 ) . comparing fig .
[ fig1 ] and fig .
[ fig2 ] one can see huge deep ( about five orders of magnitude ) in the intensity of the scattered field for the selected wavelength @xmath45 and the selected direction @xmath82 .
the fig . [ fig2 ] shows also that unintendently , the optimization significantly increased intensity of the scattered field for other wavelengths ( near @xmath83 ) in almost all directions .
[ fig3 ] shows normalized intensity of the light scattered by the cluster optimized to maximal intensity of the scattered field in the direction @xmath82 . comparing fig .
[ fig1 ] and fig .
[ fig3 ] one can see relatively large increase ( about two times ) in the intensity of the scattered field for the selected wavelength @xmath45 and the selected direction @xmath82 .
[ fig3 ] shows also that the optimization significantly increased intensity of the scattered field for @xmath84 and significantly decreased the intensity for @xmath85 in directions with @xmath86 and @xmath87 respectively .
[ fig4 ] shows normalized intensity of the light scattered by the cluster optimized to minimal intensity of the scattered field in the direction perpendicular to the source - cluster direction ( @xmath86 ) . comparing fig .
[ fig1 ] and fig .
[ fig4 ] one can see significant decrease ( about two orders of magnitude ) in the intensity of the scattered field for the selected wavelength @xmath45 in the selected direction @xmath86 .
in distinction to previous examples , the intensity of the scattered field in other directions is almost not affected in this case . versus @xmath51 for the optimized spherical cluster made of @xmath52 cubes .
the permittivity of the host medium is @xmath64 , the characteristic size of the cubes is @xmath54 nm and the total number of the particles is @xmath56 .
the source is positioned at @xmath65 and the observation points @xmath46 are located at the circle with radius @xmath59 .
the angle @xmath66 is the angle between vector @xmath46 and @xmath67 in @xmath62 plane .
the cluster reflectivity was maximized for @xmath68 at @xmath77 .
the black line shows extremely high peak in cluster reflectivity after the optimization.,width=340 ] the fig .
[ fig5 ] shows normalized intensity of the light scattered by the cluster optimized to maximal intensity of the scattered field in the direction @xmath86 . comparing fig .
[ fig1 ] and fig .
[ fig5 ] one can see huge increase ( about three orders of magnitude ) in the intensity of the scattered field for the selected wavelength @xmath45 in the selected direction @xmath86 .
we note that the intensity of the scattered field increased also in direction @xmath87 ( for @xmath72 ) and for all other directions the intensity peak shifted to @xmath84 .
examination of the figs .
[ fig1]-[fig5 ] reveals several important things .
the first one is that the optimization works because the intensity of the light scattered by the optimized cluster has very clear minima or maxima for the chosen wavelengths and positions .
the second one is that the intensity of the scattered field has deep or peak not only for designed wavelength but for other wavelengths and directions .
it should be emphasized that the scattering was optimized for one direction and for one wavelength only and that is why the optimization is effective for narrow spectrum of wavelengths and scattering directions .
the reflectivity of the spherical photonic cluster made of small particles was optimized to maximal and minimal values by using the local perturbation method and the special optimization algorithm .
the possibility to design the photonic cluster with required scattering characteristics was demonstrated for selected wavelength and direction . | the optimization of the light scattered by photonic cluster made of small particles is studied with the help of the local perturbation method and special optimization algorithm .
it was shown that photonic cluster can be optimized in a such a way that its reflectivity will be increased or decreased by several orders of magnitude for selected wavelength and direction . |
apart from their technological importance , emulsions have served as model systems accessible to rigorous theoretical modeling .
the study of emulsions consisting of droplets of a liquid dispersed in another liquid has thus contributed substantially to our understanding of the rheology of complex fluids . however , although major theoretical achievements date back to the beginning of the @xmath0 century , further progress turned out to be difficult .
the macroscopic rheological properties of emulsions are determined by the reaction of the individual drops to the flow field , which in turn is modified by the presence of other drops .
the mutual hydrodynamic interactions of drops complicates substantially the mathematical description .
moreover , depending on system parameters and flow type , droplets may break under steady flow conditions if a certain critical strain rate is exceeded .
rigorous calculations of the constitutive equation have therefore concentrated on very dilute emulsions and on conditions where drops are only weakly deformed .
sometimes , however , it is desirable to have an approximate expression , which though not rigorous can serve for practical purposes as a quantitative description in the parameter region beyond the ideal limits . as far as the dependence of the viscoelastic properties of an emulsion on the volume fraction @xmath1 of the dispersed phase is concerned , such an approximation has been given by @xcite .
it is not rigorous beyond first order in @xmath1 but serves well some practical purposes even at rather high volume fractions .
it seems not reasonable to look for a comparably simple approximation for the dependence of shear viscosity @xmath2 on shear rate @xmath3 that covers the whole range of parameters , where all kinds of difficult break up scenarios are known to occur . in the next section ,
we propose instead a less predictive expression which contains an average drop size @xmath4 ( that may change with shear rate ) as a phenomenological parameter .
the latter has to be determined independently either from theory or experiment .
it turns out , however , that for a substantial range of viscosity ratios and shear rates , the expression for @xmath5 is to a large extent independent of morphology . for conditions , where drops do not break outside this region , we point out a similarity relation between this expression and the frequency dependent viscoelastic moduli @xmath6 , @xmath7 , similar to the _
rule in polymer physics .
more precisely , we show that in the limit of small drop deformation , the constitutive equation of an emulsion composed of newtonian constituents of equal density can be obtained from the frequency dependent linear response to leading order in capillary number @xmath8 and ( reciprocal ) viscosity ratio @xmath9 .
and we argue that the identification is likely to represent a good approximation beyond this limit in a larger part of the @xmath10 parameter plane .
experimentally , this similarity relation can be tested directly , without interference of theoretical modeling , by comparing two independent sets of rheological data . in summary ,
our theoretical discussion provokes two major empirical questions .
( 1 ) if drops break : does the expression for the viscosity derived in eq.([eq : cm_miracle ] ) describe the data with @xmath4 the average drop size at a given shear rate ? ( 2 ) if drops do not break below a certain characteristic capillary number @xmath11 : does the proposed similarity rule hold ? to what extent does it generalize to non
newtonian constituents ? in section
[ sec : experiment ] we address mainly the second question by experiments with a quasi
static droplet phase of a mixture of moderately viscoelastic polymer solutions .
a common way to characterize the rheological properties of complex fluids such as emulsions , suspensions , and polymer solutions , is by means of a _ constitutive equation _ or an equation of state that relates the components @xmath12 of the _ stress tensor _ to the _ rate of strain tensor _ @xmath13 .
this relation can account for all the internal heterogeneity and the complexity and interactions of the constituents if only the system may be represented as a homogeneous fluid on macroscopic scales .
the form of possible constitutive equations is restricted by general _
symmetry arguments _ , which provide guidelines for the construction of phenomenological expressions @xcite . on the other hand , for special model systems
the constitutive equations may be calculated directly at least for some restricted range of parameters .
an early example for a direct computation of the constitutive equation of a complex fluid is einstein s formula @xmath14 for the shear viscosity @xmath15 of a dilute suspension ( particle volume fraction @xmath16 ) .
it is obtained by solving stokes equation for an infinite homogeneous fluid of viscosity @xmath17 containing a single solid sphere . for a sufficiently dilute suspension
, the contributions of different particles to the overall viscosity @xmath2 can be added independently , giving an effect proportional to @xmath1 . in close analogy
@xcite calculated @xmath2 for a steadily sheared dilute suspension of droplets of an incompressible liquid of viscosity @xmath18 in another incompressible liquid of viscosity @xmath17 . for weakly deformed drops he obtained @xmath19 which we abbreviate by @xmath20 in the following .
this expression includes einstein s result as the limiting case of a highly viscous droplet , @xmath21 . as in einstein
s calculation , interactions of the drops are neglected .
the result is independent of surface tension @xmath22 , shear rate @xmath3 , and drop radius @xmath4 ; i.e. , it is a mere consequence of the presence of a certain amount @xmath1 of dispersed drops , regardless of drop size and deformation ( as long as the latter is small ) .
moreover , the dynamic ( frequency dependent ) linear response of an emulsion has been calculated by @xcite .
his results are quoted in section [ sec : cox - merz ] below . under steady flow conditions ,
drop deformation itself is proportional to the magnitude of the rate of strain tensor @xmath13 .
more precisely , for simple shear flow with constant shear rate @xmath23 , the characteristic measure of drop deformation for given @xmath24 is the _ capillary number _ @xmath25 also introduced by @xcite .
it appears as dimensionless expansion parameter in a perturbation series of the drop shape under shear .
to derive taylor s eq.([eq : taylor ] ) it is sufficient to represent the drops by their spherical equilibrium shape .
aiming to improve the constitutive equation , @xcite took into account deformations of drops to first order in @xmath26 .
the refined analysis did not affect the off
diagonal elements of the constitutive equation , i.e. taylor s eq.([eq : taylor ] ) for the viscosity , but it gave the ( unequal ) normal stresses to order @xmath27 .
another limit , where exact results can be obtained , is the limit of large viscosity ratios @xmath21 @xcite . to clarify the physical significance of the different limits we want to give a brief qualitative description of the behavior of a suspended drop under shear , based on work by @xcite and @xcite . in a quiescent matrix fluid of viscosity @xmath17 ,
a single weakly deformed drop relaxes exponentially into its spherical equilibrium shape ; i.e. , defining dimensionless deformation by @xmath28 with @xmath29 and @xmath30 the major and minor axis of the elongated drop , one has for a small initial deformation @xmath31 , @xmath32 the characteristic _ relaxation time _
@xcite @xmath33 also characterizes the macroscopic _ stress relaxation _ in an unstrained region of a dilute emulsion . at @xmath34
one observes the characteristic relaxation mode in the frequency dependent moduli .
the relaxation time diverges for @xmath35 since it takes longer for a weak surface tension to drive a viscous drop back to equilibrium . what happens if the matrix is steadily sheared at shear rate @xmath36 ?
for @xmath37 , the flow induced in the drop by the external driving is weak compared to the internal relaxation dynamics and the equilibrium state is only slightly disturbed , i.e. , the drop is only weakly deformed .
similarly , for large viscosity ratio @xmath24 , the elongation of the drop becomes very slow compared to vorticity , and hence again very small in the steady state , even if @xmath38 is not small .
technically , this is due to the asymptotic proportionality to @xmath9 of the shear rate within the drop . in both limits of weak deformation , the time @xmath39 also controls the orientation of the major axis of the drop with respect to the flow according to @xmath40 eq.([eq : d_relax ] ) and eq.([eq : angle ] ) both can be used to determine the surface tension @xmath22 from observations of single drops under a microscope . in passing ,
we note that the classical method based on the result obtained by @xcite for the steady state deformation , can only be used if @xmath24 is not too large , whereas eq.([eq : d_relax ] ) and eq.([eq : angle ] ) are more general .
the exact calculations mentioned so far became feasible because ( and are applicable if ) deviations of the drop from its spherical equilibrium shape are small . on the other hand , if neither the capillary number ( or @xmath38 ) is small nor the viscosity ratio is large , i.e. , @xmath41 drops can be strongly deformed by the symmetric part of the flow field .
experiments with single drops by @xcite and others have shown that this eventually leads to drop break up if @xmath24 is smaller than some critical viscosity ratio . no general rigorous result for the viscosity of an emulsion is known in this regime , where arbitrary drop deformations and break up may occur .
below we will also be interested in such cases , where the conditions required for the rigorous calculations are not fulfilled . for the following discussion
we introduce some additional notation .
the rate of strain tensor @xmath13 and the vorticity tensor @xmath42 are defined as symmetric and antisymmetric parts of the velocity gradient @xmath43 . in particular , for a steady simple shear flow @xmath44 , and @xmath45 the components @xmath46 of the stress tensor in the shear plane ( @xmath47 ) as obtained for finite @xmath24 by @xcite read @xmath48 as usual , the material derivative has been defined by @xmath49 where summation over repeated indices is implied , and the first two terms vanish for steady shear flow . note
that since @xmath50 is diagonal , eq.([eq : schowalter ] ) implies @xmath51 , and hence eq.([eq : taylor ] ) remains valid to first order in @xmath8 as we mentioned already . ) , ( [ eq : cm_miracle ] ) for the viscosity @xmath2 of an emulsion give rigorous results . in the shaded region
they predict @xmath2 to be practically independent of capillary number .
the frame of the box is meant to comprise the whole @xmath26-@xmath24 parameter plane from zero to infinity ( small reynolds number understood ) .
the curved solid line is a sketch of the break up curve for steadily sheared isolated newtonian drops according to @xcite .
dashed lines indicate schematically our viscosity measurements ( see section [ sec : experiment ] ) . ]
can we extrapolate the exact second order result eq.([eq : schowalter ] ) for the stress tensor to arbitrary @xmath8 and @xmath24 by using the constraints provided by general invariance arguments ?
for example , since the shear stress has to change sign if the direction of the shear strain is inverted whereas the normal stresses do not , the shear stress and the normal stresses have to be odd / even functions of @xmath3 , respectively . from this observation
we could have foreseen that eq.([eq : taylor ] ) can not be improved by calculating the next order in @xmath3 , i.e. , by considering droplet deformation to lowest order .
more important are galilean invariance and invariance under transformations to rotating coordinate frames , which give rise to the material derivative introduced above . applying the operator @xmath52 to eq.([eq : schowalter ] )
adds to the right hand side of the equation a term @xmath53 plus a term of order @xmath54 , so that one obtains ( in the shear plane ) @xmath55 as another short hand notation we have introduced a second characteristic time @xmath56 , which to the present level of accuracy in @xmath1 is given by @xmath57 it sets the time scale for _ strain relaxation _ in an unstressed region and was named _ retardation _
time by @xcite .
@xcite realized that up to the partly unknown terms of order @xmath58 on the right hand side , eq.([eq : acrivos ] ) belongs to a class of possible viscoelastic equations of state already discussed by @xcite .
hence , setting @xmath59 on the right hand side of eq.([eq : acrivos ] ) , we can define a ( minimal ) model viscoelastic fluid that behaves identical to the emulsion described by eq.([eq : schowalter ] ) for small shear rates .
in contrast to eq.([eq : acrivos ] ) , the truncated formula for the viscosity @xmath60\ ] ] thus obtained has a manifestly non perturbative form .
however , no phenomenological parameters had to be introduced .
note that eq.([eq : miracle ] ) comprises both exactly known limits : @xmath61 for fixed @xmath24 , and @xmath21 for arbitrary @xmath8 .
obviously , @xcite have forgotten a term @xmath62 in their eq.(3.6 ) for @xmath46 in the limit @xmath21 for fixed @xmath26 . if the latter is included ,
eq.([eq : miracle ] ) is also in accord with their @xmath63analysis .
moreover , eq.([eq : miracle ] ) has the proper limiting behavior for @xmath64 , @xmath65 , i.e. @xmath66 , which is an extreme case of eq.([eq : restriction ] ) .
since we assume equal densities for the two phases , the two phase fluid actually reduces to a one phase fluid in this degenerate case , and the viscosity is simply @xmath17 , independent of morphology . for illustration , the rigorous limits of eq.([eq : miracle ] ) in the @xmath10 plane
are depicted graphically in fig .
[ fig : limits ] . following @xcite , a qualitative break up curve for single drops under steady shear is also sketched . in summary
, eq.([eq : miracle ] ) is correct for arbitrary @xmath24 if @xmath67 , and for arbitrary @xmath8 if @xmath21 , and for small and large @xmath8 if @xmath64 .
therefore , one can expect that eq.([eq : miracle ] ) works reasonably well within a large parameter range ( small reynolds number understood ) .
this is further supported by the observation that the error made in going from eq.([eq : schowalter ] ) to eq.([eq : miracle ] ) rather concerns the _ shape _ of the droplet than its _ extension _ ( it consists in truncating a perturbation series in shape parametrisation ) .
the final result , though sensitive to the latter , is probably less sensitive to the former .
nevertheless , one would not be surprised to see deviations from eq.([eq : miracle ] ) when drops become extremely elongated .
finally , due to changes in morphology by break up and coalescence , the avarage drop size @xmath4 may change .
observe , however , that for most viscosity ratios ( @xmath24 not close to unity ) , eq.([eq : miracle ] ) is practically independent of capillary number ( and thus of @xmath4 ) when break up might be expected according @xcite and others . as an analytic function that is physically known to be bounded from above and from below ( the latter at least by the viscosity of a stratified two phase fluid depicted in fig . [ fig : mixing ] ) , @xmath68 has to have vanishing slope in the @xmath69direction for large @xmath8 .
according to eq.([eq : miracle ] ) , @xmath68 is almost independent of @xmath8 for @xmath70 where @xmath71 is the turning point in the dilute limit , determined by @xmath72 . for finite volume fractions @xmath73 from eq.([eq : tau_1_oldroyd ] ) replaces @xmath39 .
hence , for @xmath74 , eq.([eq : miracle ] ) and its extension to higher volume fractions derived below in eq.([eq : cm_miracle ] ) are practically independent of drop deformation and morphology . for finite volume fractions a rough interpolation for @xmath71 derived from eq.([eq : cm_miracle ] ) is given by @xmath75 . in a large part of parameter space we thus expect eqs.([eq : miracle ] ) , ( [ eq : cm_miracle ] ) to be applicable to monotonic shear histories with @xmath4 given by the average initial radius of the droplets . for non
monotonic shear histories , there can of course be hysteresis effects in @xmath76 that result from morphological changes for @xmath74 .
these can only be avoided by substituting for @xmath4 the radius corresponding to the actual average drop size at the applied shear rate @xmath36 . of the continuous phase / the viscosity @xmath2 of the whole emulsion , within / outside the `` free volume sphere '' of radius @xmath77 .
] a general limitation of the equations discussed so far , is the restriction to small volume fractions . above
, we have implicitly assumed that second order effects from drop interactions are small compared to second order effects from drop deformation .
any direct ( coalescence ) and indirect ( hydrodynamic ) interactions of droplets have been neglected in the derivation .
hydrodynamic interactions can approximately be taken into account by various types of cell models .
recently @xcite proposed a self consistent method analogous to the clausius mossotti or lorentz sphere method of electrostatics .
for the case of a disordered spatial distribution of drops his results reduce to those already obtained by @xcite .
oldroyd artificially divides the volume around a droplet of viscosity @xmath78 into an interior `` free volume '' with a viscosity @xmath17 of the bare continuous phase and an exterior part with the viscosity @xmath2 of the whole emulsion ( see fig .
[ fig : mf - trick ] ) . according to this scheme , an improved version of eq.([eq : taylor ] ) should be @xcite @xmath79 this equation predicts a larger viscosity than its truncation to first order in @xmath1 , eq.([eq : taylor ] ) . both are shown as dot dashed lines in fig .
[ fig : mixing ] .
eq.([eq : cm_taylor ] ) is _ qualitatively _ superior to eq.([eq : taylor ] ) .
we note , however , that the limit @xmath80 deviates in second order in @xmath1 from the result obtained for suspensions by @xcite .
eq.([eq : cm_taylor ] ) and likewise all of the following equations containing quantities @xmath81 , @xmath73 , @xmath82 are only rigorous to first order in @xmath1 .
the same reasoning as to the viscosity applies to the characteristic times @xmath39 and @xmath56 which now read @xcite @xmath83[2\lambda+3 - 2\phi(\lambda-1 ) ] } { 40(\lambda+1)-8\phi(5\lambda+2 ) } \;,\ ] ] @xmath84[2\lambda+3 + 3\phi(\lambda-1 ) ] } { 40(\lambda+1)+12\phi(5\lambda+2 ) } \;.\ ] ] their ratio @xmath85 is still given by eq.([eq : tau_2 ] ) .
finally , eq.([eq : miracle ] ) becomes @xmath86 which to our knowledge has not been given before , and is one of our main results .
( for a graphical representation see fig .
[ fig : graph ] . ) in the limit @xmath87 it reduces to eq.([eq : cm_taylor ] ) , whereas for @xmath88 only the second term in parentheses contributes and the _ curved _ dashed lines in fig .
[ fig : mixing ] are obtained . from the foregoing discussion
one should expect eq.([eq : cm_miracle ] ) to be applicable within a large range of shear rates , viscosities , and volume fractions .
finally , we note that more cumbersome expressions for @xmath89 , @xmath90 and @xmath82 have been derived within a different cell model by @xcite . here
we only quote their expression for @xmath89 , @xmath91 } { 4(\lambda+1)-5(5\lambda+2)\phi+42\lambda\phi^{5/3}-5(5\lambda-2)\phi^{7/3 } + 4(\lambda-1)\phi^{10/3}}\ ] ] which is also represented graphically by the dotted lines in fig .
[ fig : mixing ] . since our data favor eq.([eq : cm_taylor ] ) over eq.([eq : choi ] ) , and similar observations have been made by others before ( see section [ sec : experiment ] ) , we will not pursue this alternative approach further in the present contribution .
( chosen arbitrarily ) .
the dot dashed and dashed _ straight _ lines pertain to dilute emulsions described by the extrapolation formula eq.([eq : miracle ] ) , which reduces to taylor s formula eq.([eq : taylor ] ) for small capillary numbers .
the corresponding _ curved _ lines are obtained from eq.([eq : cm_miracle ] ) where interactions of the droplets are taken into account in a mean field approximation .
the curved _ dotted _ lines are the predictions of the cell model by @xcite for small shear rates , eq.([eq : choi ] ) .
the curved _ solid _ line represents the viscosity @xmath92 $ ] of a two phase stratified fluid and is a lower bound for any viscosity mixing rule . ]
it is interesting to observe that if morphology is conserved ( drop size @xmath4 independent of shear rate ) for @xmath93 , our eq.([eq : cm_miracle ] ) for the nonlinear shear viscosity is closely related to the expressions for the frequency dependent complex viscosity @xmath94 of an emulsion of two incompressible newtonian liquids as derived by @xcite , @xmath95 for convenience , we also give the corresponding viscoelastic shear modulus @xmath96 , @xmath97 obviously , the shear rate dependent viscosity of eq.([eq : cm_miracle ] ) is obtained from the real part of the complex frequency dependent viscosity @xmath98 by substituting @xmath36 for @xmath99 , @xmath100 in the same way , the first normal stress difference @xmath101 from eq.([eq : schowalter ] ) is obtained to leading order in @xmath1 and @xmath36 from @xmath102 , i.e. , @xmath103 reverting the line of reasoning pursued so far , we can conclude that to leading order in @xmath8 and/or @xmath9 the weak deformation limit of the constitutive equation of emulsions is obtained from the linear viscoelastic spectra @xmath104 , @xmath105 .
further , this identification can possibly be extended ( at least approximately ) into regions of the @xmath10 plane where the critical capillary number for drop breakup is somewhat larger than @xmath71 of eq.([eq : region ] ) . in hindsight , it is not surprising that in the case of weakly deformed drops the frequency dependent viscosity and the steady shear viscosity are related .
note that under steady shear , drops undergo oscillatory deformations at a frequencey @xmath106 if observed from a co rotating frame turning with vorticity at a frequency @xmath107 .
if we take eq.([eq : first_normal_stress ] ) seriously beyond the rigorously known limit , we obtain an interesting prediction for the first normal stress difference .
in contrast to eq.([eq : fnsd_schowalter ] ) , eq.([eq : first_normal_stress ] ) implies that the first normal stress difference saturates at a finite value @xmath108 ^ 2 $ ] for high shear rates .
thus , although the initial slope of the first normal stress difference with @xmath3 increases with @xmath24 , its limit for large @xmath3 decreases with @xmath24 .
finally , we remark that based on qualitative theoretical arguments , the similarity relation contained in eq.([eq : modified_cox - merz ] ) and eq.([eq : first_normal_stress ] ) has recently been proposed also for polymer melts @xcite .
usually , in polymer physics a slightly different relation is considered ; namely a similarity between @xmath109 and @xmath110 , also known as _
cox merz rule _ @xcite . in our case , since @xmath111 , we can write @xmath112 under the conditions mentioned at the beginning of this section , the usual cox
merz rule is fulfilled to first order in @xmath1 for an emulsion .
eqs.([eq : modified_cox - merz ] ) , ( [ eq : first_normal_stress ] ) are interesting from the theoretical point of view , because they suggest a similarity of two _ a priory _ rather different quantities .
the results of this section also can be of practical use , since they suggest that two different methods may be applied to measure a quantity of interest . ) normalized to @xmath17 as a function of viscosity ratio @xmath24 and capillary number @xmath8 .
the volume fraction of the dispersed phase is chosen to be @xmath113 . ]
generalization of the above theoretical discussion to the case of non newtonian constituents is not straightforward . indeed , as @xcite already knew , his linear response results quoted in eq.([eq : eta_w_oldroyd ] ) and eq.([eq : g_w_oldroyd ] ) are readily generalized to viscoelastic constituents by replacing the viscosities @xmath114 in the expression for @xmath115 ( or @xmath116 ) by complex viscosities @xmath117 @xcite . as a consequence , the decompositions of @xmath115 and @xmath116 in real and imaginary parts
are no longer those of eqs.([eq : eta_w_oldroyd ] ) and ( [ eq : g_w_oldroyd ] ) , and @xmath118 , @xmath119 , @xmath120 , @xmath121 are given by more cumbersome expressions . for the steady state viscosity , on the other hand , one has to deal with a non homogeneous viscosity even within homogeneous regions of the emulsion , since the strain rate itself is non homogeneous and the viscosities are strain rate dependent .
we do not attempt to solve this problem here , nor do we try to account for elasticity in the nonlinear case .
yet , it is an intriguing question , whether the similarity rule eq.([eq : modified_cox - merz ] ) can be generalized to the case of non - newtonian constituents if the constituents themselves obey the cox
merz rule ( what many polymer melts and solutions do ) . if both constituents have similar phase angles @xmath122 the generalized viscosity ratio @xmath123 that enters the expressions for @xmath115 and @xmath116 , transforms approximately to @xmath124 by applying the cox
merz rule .
therefore , in this particular example , eq.([eq : cm_miracle ] ) supports the expectation that the generalization may work at least approximately .
if , on the other hand , the phase angles of the constituents behave very differently , the answer is less obvious .
this problem has been investigated experimentally and is further discussed in section [ sec : experiment ] . in any case , the generalization can only work if the representation of the emulsion by a simple shear rate dependent viscosity ratio @xmath125 , with @xmath36 the external shear rate , is justified . in the remainder of this section we construct an argument that allows us to estimate the effective shear rate dependent viscosity ratio that should replace @xmath24 in eq.([eq : cm_miracle ] ) .
we take into account the deviation of the strain rate from the externally imposed flow only within the drops , because outside the drops the discrepancy is always small . inside a drop
, the strain rate can be small even for high external shear rates if the viscosity ratio @xmath126 is large .
since we are looking for an effective viscosity @xmath127 for the whole drop to replace the viscosity @xmath128 at small shear rate , we replace the non newtonian drop of non homogeneous viscosity by an effective pseudo newtonian drop of homogeneous but shear rate dependent viscosity .
a possible ansatz for @xmath129 is obtained by requiring that the total energy dissipated within the drop remains constant upon this substitution .
hence , we have @xmath130 where @xmath131 and @xmath132 denote the rate of strain fields in the real drop and in the corresponding model drop of effective viscosity @xmath129 , respectively .
they both depend on the position within the drop , whereas the average strain rate @xmath133 that enters the right hand side of eq.([eq : dissipation ] ) does not .
since the strain field @xmath132 within a drop of homogeneous viscosity is unique for a given system in a given flow field , @xmath129 itself can be expressed as a function @xmath134 of the average strain rate . here , we approximate this functional dependence by the strain rate dependence @xmath135 of the viscosity of the dispersed fluid .
further , neglecting drop deformation we calculate @xmath136 from the velocity field within a spherical drop @xcite and obtain @xmath137 a different prefactor ( @xmath138 in place of @xmath139 ) in the expression for the effective strain rate @xmath140 was obtained by @xcite using instead of the average in eq.([eq : avg_strain ] ) the maximum norm of @xmath132 . to obtain the correction to eq.([eq : cm_miracle ] ) due to eq.([eq : eta_eff ] ) in the case of non newtonian constituents , the implicit equation for @xmath141 has to be solved for given functions @xmath142 and @xmath135 . for shear thinning constituents , eq.([eq : eta_eff ] ) implies a tendency of eq.([eq : cm_miracle ] ) to overestimate @xmath143 if @xmath36 and @xmath24 are large . in the actual case of interest , for the constituents that were used in the experiments discussed in section [ sec : experiment ] , the viscosity ratio @xmath24 ( @xmath128 and @xmath17 taken at the external shear rate @xmath3 ) varies almost by a factor of 10 .
however , the corrections discussed in this section only become important for high shear rates , where the constituents are shear thinning .
in this regime , the viscosity ratio ( viscosities taken at the external shear rate ) only varies between @xmath144 and @xmath145 , and hence the corrections expected from eq.([eq : eta_eff ] ) are at best marginally significant at the level of accuracy of both eq.([eq : eta_eff ] ) and the present measurements .
therefore , a representation of the drops by pseudo newtonian drops of homogeneous but shear rate dependent viscosity is most probably not a problem for the measurements presented in the following section .
the question as to a generalization of the similarity rule eq.([eq : modified_cox - merz ] ) to non
newtonian constituents seems well defined .
the experimental investigation deals with a phase separated aqueous solution containing a polysaccharide ( alginate ) and a protein ( caseinate ) . this type of solutions are currently used in the food industry .
the methods for characterizing the individual polymers in solution are in general known , especially when dealing with non gelling solutions where composition and temperature are the only relevant parameters
. the polymers are water soluble .
when the two biopolymers in solution are mixed , a miscibility region appears in the low concentrations range and phase separation at higher concentrations .
the binodal and the tie lines of the phase diagram can then be established by measuring the composition of each phase at a fixed temperature . in general , the rheological behavior of phase separated systems is difficult to investigate , and a suitable procedure is not fully established . in some cases , two phase solutions
macroscopically separate by gravity within a short period of time , but in some other cases ( such as ours ) they remain stable for hours or days without appearance of any visible interface .
these `` emulsion type '' solutions have no added surfactant . following approaches developed for immiscible blends
, one may try to characterize the partially separated solution as an effective emulsion if the coarsening is slow enough . in order to establish a comparison between phase separated solutions and emulsions ,
it is necessary to know * the volume fraction of the phases , * their shear
rate dependent viscosities ( flow curves ) , * their viscoelastic spectra , * the interfacial tension @xmath22 between the phases , * the average radius of @xmath4 the drops a difficulty when working with phase separated solutions , as opposed to immiscible polymer melts , arises from the fact that each phase is itself a mixture ( and not a pure liquid ) and therefore the rheology of the phase depends on its particular composition .
if one wishes to minimize the number of parameters , it is important to keep the composition of the phases constant upon changing the volume fractions .
this can be achieved by working along a tie line of the phase diagram . and
this is precisely the procedure that we followed .
the polymers were first dissolved , then a large quantity of the ternary mixture was prepared ( 350 ml ) and was centrifuged .
the two phases were then collected separately .
both pure phases were found to be viscoelastic and to exhibit shear thinning behavior , which is especially pronounced for the alginate rich phase with @xmath147 , while the caseinate rich phase is almost newtonian below @xmath148 s@xmath149 .
the viscosity of the alginate rich phase is higher than that of the caseinate rich phase for shear rates below @xmath150 s@xmath149 and lower for higher shear rates .
we checked that both phases obey the usual cox
merz rule in the whole range of applied shear rates . by mixing various amounts of each phase
, the volume fraction of the dispersed phase was varied between @xmath151 and @xmath152 while the composition of each phase was kept constant .
in particular , the temperature was kept constant and equal to the centrifugation temperature in order to avoid redissolution of the constituents . to prepare the emulsion ,
the required quantities of each phase were mixed in a vial and gently shaken .
then the mixture was poured on the plate of the rheometer ( ar 1000 from ta instruments fitted with a cone and plate geometry 6 cm/@xmath153 ) and a constant shear rate was applied . the apparent viscosity for a particular shear rate
was then recorded versus time until it reached a stable value . by shearing at a fixed shear rate
, one may expect to create a steady size distribution of droplets , with a shear rate dependent average size .
after each shear experiment a complete dynamic spectrum was performed . in this way , shear rates ranging between @xmath154 s@xmath149 and @xmath155 s@xmath149 were applied .
the analysis of each spectrum according to @xcite allowed us to derive by curve fitting the average drop radius @xmath4 at the corresponding shear rate .
more technical details about the experimental investigation along with more experimental results will be presented elsewhere . here , we concentrate on the analysis of those aspects of the rheological measurements pertinent to the theoretical discussion in section [ sec : theory ] . in this section
we present our experimental observations and address the questions posed at the end of the introduction .
before we present our own data we want to comment briefly on related data recently obtained for polymer melts by @xcite .
these authors measured the shear rate dependent viscosity of binary polymer melts and compared them to the low volume fraction limit of eq.([eq : cm_miracle ] ) , i.e. eq.([eq : miracle ] ) , and to ( a truncated form of ) results of @xcite .
they reported much better agreement with eq.([eq : miracle ] ) than with the truncated series from @xcite .
comparison with the full expressions of @xcite would have made the disagreement even worse ( cf .
fig [ fig : mixing ] ) .
the average radius @xmath4 that enters the equation , was determined independently for each shear rate applied .
the constituents where moderately non
newtonian polymer melts , the viscosity ratio varying between @xmath156 over the range of shear rates applied .
hence , these experiments , are located in the interesting parameter range , where eqs.([eq : cm_miracle ] ) and eq.([eq : miracle ] ) for @xmath157 are expected to be sensitive to drop deformation and break up .
surprisingly , the results show that they describe the data very well over the whole range of shear rates although one would not necessarily expect average drop deformation to be very small .
unfortunately , drop sizes have not been reported by the authors , so conclusions concerning the location in the @xmath10 parameter plane and the validity of the similarity rule eq.([eq : modified_cox - merz ] ) can not be drawn . also the question , whether eq.([eq : cm_miracle ] )
holds for small viscosity ratios @xmath158 , can not be answered .
( opaque squares ) and the real part @xmath159 of the dynamic viscosity @xmath160 ( lines ) of a droplet phase of a mixture of weakly viscoelastic polymer solutions ( alginate / caseinate ) . also shown
is eq.([eq : cm_miracle ] ) for the viscosity of an emulsion of newtonian constituents evaluated for the actually non
newtonian viscosities of the constituting phases with the drop size obtained from the spectra ( open triangles ) . due to the scatter in the dynamic viscosity at low frequencies
there is an uncertainty in the average drop size , resulting in corresponding error bars ( multiple points ) for eq.([eq : cm_miracle ] ) . ]
our own measurements were located in about the same @xmath161range . as we noted in the preceding section ,
only the ratio @xmath146 enters rheological equations and knowledge of either @xmath4 or @xmath22 allows the other quantity to be inferred from rheological measurements .
@xcite determined the interfacial tension of the alginate / caseinate system used in our experiments by observing drop relaxation under a microscope and analyzing the data according to section [ sec : theory ] .
they found @xmath162 n / m . using this
, we obtained an average drop size @xmath163 m from the measured spectra @xmath104 , @xmath164 according to @xcite for the experiments reported in fig .
[ fig : cox - merz ] . by the method based on @xcite
, we could not detect a decrease in drop size with shear rate as expected from the phenomenological phase diagram for single newtonian drops under shear as established by @xcite and others .
thus , the limit of high capillary numbers and moderate viscosity ratios ( the region above the break up curve ) in fig .
[ fig : limits ] has been accessed experimentally .
corresponding locations have been indicated qualitatively in the figure by dashed lines . with the drop size being constant , one can try to test the proposed similarity rule eq.([eq : modified_cox - merz ] ) . by identifying the axis for frequency @xmath99 and shear rate @xmath36 , data for the real part @xmath159 of the frequency dependent dynamic viscosity @xmath165
are compared to data for the shear rate dependent viscosity @xmath157 in fig .
[ fig : cox - merz ] .
the emulsion containing @xmath152 of the alginate rich phase and @xmath166 of the caseinate rich phase has been prepared at room temperature as described in the preceding section .
steady shear rates ranging between @xmath154 s@xmath149 and @xmath155 s@xmath149 corresponding to capillary numbers @xmath167 have been applied . the shear viscosity ( opaque squares )
is reported in the figure for each of these individual measurements .
the multiple data sets for @xmath118 ( lines ) taken each between two successive steady shear measurements , superimpose fairly well ; i.e. the spectra appear to be remarkably independent of the preceding steady shear rate .
the good coincidence of @xmath168 and @xmath157 in fig .
[ fig : cox - merz ] show that the data obey the proposed similarity rule eq.([eq : modified_cox - merz ] ) over a large range of shear rates .
the agreement near @xmath169 s@xmath149 is a consequence of the proximity to the trivial limit @xmath64 , @xmath170 .
nevertheless , the data provide strong evidence that eq.([eq : modified_cox - merz ] ) is an excellent approximation for a large range of viscosity ratios and capillary numbers .
similar results ( not shown ) have been obtained for other volume fractions .
comparison with eq.([eq : cm_miracle ] ) represented by the open triangles in fig .
[ fig : cox - merz ] , on the other hand , is less successful at large shear rates , although it is still not too far off for a theoretical curve without any adjustable parameter .
a discrepancy had to be expected as a consequence of the non newtonian character of the constituents at large shear rates , which is definitely not taken into account in eq.([eq : cm_miracle ] ) .
for the plot of eq.([eq : cm_miracle ] ) in fig .
[ fig : cox - merz ] we merely substituted @xmath171 taken at the external shear rate @xmath36 for @xmath24 .
the average drop size @xmath4 was obtained from fitting the viscoelastic moduli .
the scatter in the dynamic viscosity data gives rise to an uncertainty in @xmath4 , which is reflected by the multiple open triangles at low shear rates .
it seems that the similarity rule eq.([eq : modified_cox - merz ] ) is more general than eq.([eq : cm_miracle ] ) , i.e. , it still holds for rather viscoelastic constituents ( that obey the usual cox merz rule ) , where the latter fails .
this relation certainly deserves further investigation with different materials and methods .
in summary , we have succeeded in establishing an analogy first between a partially phase separated polymer solution and an emulsion , and further between the viscoelastic spectrum of the system and its nonlinear shear viscosity even in the case of ( moderately ) non newtonian constituents .
this work was supported by the european community under contract no fair / ct97 - 3022 .
we thank p. ding and a. w. pacek ( university of birmingham ) for measuring the surface tension and s. costeux and g. haagh for helpful discussions and suggestions .
1998 the morphology - dependent rheological behavior of an immiscible model polymer blend . in _ proceedings of the @xmath173 european rheology conference _
i. emri & r. cvelbar ) , _ progress and trends in rheology _ , vol . 5 , p. 80 .
darmstadt : steinkopff . | combining direct computations with invariance arguments , taylor s constitutive equation for an emulsion can be extrapolated to high shear rates .
we show that the resulting expression is consistent with the rigorous limits of small drop deformation and that it bears a strong similarity to an _ a priori _ unrelated rheological quantity , namely the dynamic ( frequency dependent ) linear shear response .
more precisely , within a large parameter region the nonlinear steady state shear viscosity is obtained from the real part of the complex dynamic viscosity , while the first normal stress difference is obtained from its imaginary part .
our experiments with a droplet phase of a binary polymer solution ( alginate / caseinate ) can be interpreted by an emulsion analogy .
they indicate that the predicted similarity rule generalizes to the case of moderately viscoelastic constituents that obey the cox
merz rule . |
lateral semiconductor superlattices are realized by modern lithographic techniques to impose a periodic pattern onto a two - dimensional electron system ( 2des ) at a semiconductor heterostructure .
this pattern can be a uni- or bi - directional periodic potential modulation of different strength allowing for a variety of artificial periodic structures.@xcite magnetotransport measurements in such structures have revealed novel oscillations in the conductivity , the weiss oscillations which are superimposed on top of the well - known shubnikov - de haas oscillations.@xcite besides weiss oscillations ( a commensurability effect),novel magneto - resistance oscillations with @xmath0periodicity have been observed in short - period superlattices due to magnetic miniband energy spectrum.@xcite the above developments have renewed the interest in the problem of bloch electrons in a magnetic field , which was extensively studied in the past by renowned theorists@xcite but it is experimentally accessible now with these artificial periodic structures.@xcite the first theoretical approach to the problem is due to peierls@xcite and it is based on the formulation of an effective single - band hamiltonian , arising from a tight - binding dispersion relation @xmath1 , through the substitution @xmath2 , where @xmath3 the vector potential . when combined to a semiclassical picture@xcite , the effective hamiltonian theory leads to the idea of quantized magnetic orbits , which can be commensurate with the lattice period . within the framework of effective single - band hamiltonian theory , further investigations lead to butterfly - like patterns for the energy spectrum known as hofstadter spectrum.@xcite
this spectrum reflects , as a consequence of harper@xmath4s equation@xcite , the splitting of the single bloch - band into magnetic subbands , according to the number of magnetic flux quanta piercing the unit cell of the lattice .
some recent experimental results@xcite supports the existence of the hofstadter@xmath4s spectrum in high magnetic field regime .
furthermore , an experimental realization of the hofstadter s spectrum was achieved considering the transmission of microwaves@xcite and acoustic waves@xcite through an array of scatterers . in recent years
special attention has been paid to investigate the energy miniband structure@xciteas well as the transport properties of two - dimensional short - period lateral supperlattices in the presence of a perpendicular magnetic field @xmath5 in the case of weak modulation.@xcite some calculations of superlattice electronic structure under the above conditions have used perturbation theory@xcite or a tight - binding scheme@xcite and look for solutions of the one - electron hamiltonian that are linear combination of the unperturbed extended landau wavefunctions .
another approach to the problem is based on ferrari s wavefunctions@xcite , that is , wavefunctions constructed out of coherent states associated to a von neumann - type lattice.@xcite besides the effect of electric modulation , the effect of magnetic modulation on the superlattice energy spectrum has been studied@xcite and a formal connection between these two has been investigated.@xcite in this paper we present calculations on the electronic band structure of a lateral superlattice subject to a perpendicular magnetic field by employing a projection operator technique in the context of magnetic translation group ( m.t.g ) theory .
we use a new set of magnetic - consistent wannier functions given in terms of schrauben functions introduced by jannussis@xcite in order to construct bloch - like eigenfunctions that reflect the crystal symmetry of the superlattice .
schrauben functions describe free electrons in a uniform magnetic field in the symmetric gauge and have been used in the past for magnetoconductance calculations.@xcite the paper has the following structure : in section 2 the model is formulated and the projection operator of the m.t.g corresponding to the problem is introduced . in section 3
we construct a basis of appropriately symmetric bloch - like wavefunctions as linear combination of schrauben functions . in section 4
, the electronic spectrum of the 2des , under weak periodic modulation , is obtained . in section 5
, we present specific results for periodic modulations with rectangular or triangular symmetry .
the density of states is calculated as well .
we conclude with a summarizing section 6 .
to describe the electrons in the conduction band of a lateral superlattice at the algaas - gaas interface in a constant perpendicular magnetic field @xmath6 , we employ a model of strictly two - dimensional electron gas ( 2deg ) with a 2d periodic potential modulation @xmath7 , @xmath8 .
following the common practice , we adopt an effective mass approximation , to take into account the effect of the crystalline atomic structure over the charge carriers .
the system is described by the well - known single - particle hamiltonian of bloch electrons in a magnetic field : [ eq.2.1]= ( + e ) ^2 + v ( ) where we neglect the zeeman splitting and spin - orbit interactions which are very small for gaas systems .
the hamiltonian ( [ eq.2.1 ] ) does not commute with the usual translations operators [ eq.2.2 ] ( ) = where @xmath9 is any translation vector , but commutes with unitary operators of the form [ eq.2.3 ] _
m()= ] introduced by brown@xcite and zak@xcite , if the vector potential @xmath3 fulfills the condition [ eq.2.4 ] + = 0 j , k=1,2,3 this relation holds , for the symmetric gauge @xmath10 , which was used by both authors .
it is worth noting that introducing local gauge one can consider any vector potential.@xcite the above operators ( [ eq.2.3 ] ) form the magnetic translation group ( m.t.g ) .
this is in fact a ray group@xcite , because the product of two elements of the m.t.g yields another element , multiplied by a constant phase [ eq.2.5 ] _
m(_2 ) = _ m(_1+_2 ) where @xmath11 and @xmath12 are translation vectors . as a consequence of ( [ eq.2.5 ] )
, we can verify the commutation relation [ eq.2.6 ] = 2 i _
m(_1+_2 ) where the symmetric gauge for the vector potential has been used .
therefore , the magnetic translation operators commute with the hamiltonian , but in general not with each other . only in the case
that the phase constant in ( [ eq.2.6 ] ) is a multiple of @xmath13 , [ eq.2.7 ] ( _ 1_2 ) = q , q n we get @xmath14=0 $ ] and the operators form an abelian group . the condition ( [ eq.2.7 ] ) is compatible with the demand that the eigenfunctions of the system satisfy the born - von karman periodic boundary conditions , e.g. are periodic under magnetic translations corresponding to the full lattice size . the term @xmath15 in ( [ eq.2.7 ] ) is the magnetic flux through the cell spanned by lattice vectors @xmath16 and @xmath17 and can be written as an integer multiple ( @xmath18 ) of the flux per unit cell @xmath19 , where @xmath20 is the magnetic flux quantum .
consequently , the condition([eq.2.7 ] ) is fulfilled each time the number of magnetic flux quanta through the unit cell is a rational number : [ eq.2.8 ] = , q , p thus , imposing a commensurability relation between the magnetic field intensity and the lattice size .
after a brief preamble on the m.t.g , we follow brown@xmath4s formulation@xcite to introduce the projection operators of m.t.g .
we consider the irreducible representations of the m.t.g labelled by the magnetic crystal momentum vector @xmath21 , restricted to a single magnetic brillouin zone ( mbz ) , or to @xmath22 part of it if @xmath23 .
the @xmath24 irreducible representation is given by a familiar relation [ eq.2.9 ] d^()= ) d^0 ( ) where @xmath25 , with @xmath26 , @xmath27 the primitive lattice vectors and @xmath28 with @xmath29 , @xmath30 are the primitive reciprocal - lattice vectors .
the vectors @xmath31 and @xmath32 are related in the usual fashion @xmath33 and @xmath34 , @xmath35 are integer indices .
the @xmath36 matrix @xmath37 for any translation @xmath38 is generated from those corresponding to the primitive lattice vectors [ eq.2.10 ] d^0_kk(_1)=_kk , d^0_kk(_2)=_k , k-1^mod p and the ray - group multiplication law ( [ eq.2.5 ] ) yields to [ eq.2.11 ] d^0_kk()=_k , k-n_2^mod p , k , k=0,1,2,p-1 where the modulo kronecker delta , @xmath39 equals 1 if @xmath40 and @xmath41 are equal or differ by a multiple of @xmath18 and vanishes otherwise . using ( [ eq.2.9 ] ) and ( [ eq.2.11 ] )
we obtain the following form for the irreducible representations @xmath42 } \exp \bigg [ i\pi \frac{q}{p}n_1 \big ( n_2 + 2(k-1 ) \big ) \bigg ] \ ] ] we can see that increasing any of @xmath43 , @xmath44 by @xmath45 produces an equivalent representation .
thus the values of @xmath21 have to be restricted to a single magnetic brillouin zone : @xmath46 .
the rank of matrices implies the existence of @xmath18-partner functions transforming according to different rows of the same representation @xmath47 we introduce the projection operator of m.t.g . of bloch electrons in a uniform magnetic field@xcite [ eq.2.13 ] p^ _ kk()=_^_m ( ) which projects out the component belonging to @xmath48-th row of the representation @xmath21 and then generates the partner belonging to @xmath49-th row . using the expression for the magnetic translation operator ( [ eq.2.3 ] ) , eq . ( [ eq.2.13 ] )
leads to [ eq.2.14 ] p^_kk()= _ c _ where [ eq.2.15 ] c_=_k , k-n_2 ^mod p we finally remark that acting the projection operator @xmath50 to an arbitrary function we can find a partner function of the @xmath21-th representation of m.t.g . in the next section ,
we find a basis of bloch electrons in a magnetic field by acting the found projection operator ( [ eq.2.14 ] ) onto a basis of free electrons in a magnetic field .
landau functions , eigenfunctions of angular momentum and localized functions based on coherent states@xcite are usually used to describe free electrons moving in a plane under the influence of a magnetic field .
the coherent states@xcite on a von neumann lattice are the most localized eigenfunctions for an electron in magnetic field .
for such a lattice , with a single flux quantum per unit cell , the area of the unit cell equals @xmath51 , where @xmath52 is the magnetic length .
jannussis@xcite working in the symmetric gauge , introduced a different basis of wave functions for free electrons in uniform magnetic field , who called them schrauben functions : [ eq.3.1 ] _
n,()=c_n ( -_y - i_x ) ^n where @xmath53 is the landau index , @xmath54 the electron wavevector and @xmath55 , @xmath56,the normalization coefficient is given by [ eq.3.2 ] c_n= the above functions have several important properties , some of which are : i)they form a basis in the hilbert space of free electrons in a uniform magnetic field .
[ eq.3.3 ] _ n , _
n,^ ( ) _ n,()=(- ) [ eq.3.4 ] ( _ n , ( ) , _
n , ( ) ) = _ nn ii)they are fully localized around the centers [ eq.3.5 ] _
m(x_m , y_m ) : x_m = l^2k_y , y_m =- l^2k_x and in the two - dimensional plane take the symmetric form : @xmath57^n \nonumber \\ & \cdot \exp{\bigg [ -\frac{1}{4l^2}\big [ ( x - x_m)^2+(y - y_m)^2+\frac{i}{2l^2}(x_my - xy_m ) \big ] \bigg ] } \ ; \end{aligned}\ ] ] also , the probability density @xmath58 has a rotational symmetry about the centers @xmath59 and takes the form of the poisson distribution [ eq.3.7 ] ( r_m)= where @xmath60 \iii ) they obey harper@xmath4s symmetric condition@xcite , i.e. for a lattice translation @xmath61 [ eq.3.8 ] _
n,(+)=_n , ( ) , = - ( ) it has been shown@xcite that a wannier function @xmath62 for electrons in a magnetic field can be obtained directly form the schrauben function if the wave - vector @xmath63 is substituted by the vector potential @xmath64 in what follows , we construct a basis for two - dimensional bloch electrons in a uniform magnetic field applying the projection operator method described in section 2 . in order to find the principal partner function we choose the projection operator @xmath65 ( see eq .
( [ eq.2.14 ] ) ) , projecting onto the first row of @xmath21-th representation . acting the projection operator
( [ eq.2.14 ] ) onto the functions @xmath66 of the basis ( [ eq.3.1 ] ) we find the following partner functions of the @xmath21-th representation of the m.t.g .
@xmath67 where @xmath68 are the magnetic wannier functions of free electrons in a uniform magnetic field@xcite , @xmath69 \bigg \}^n \nonumber \\ & \cdot \exp \bigg \ { -\frac{1}{4l^2}\big [ ( x - n_1a_1)^2+(y - n_2a_2)^2 \big ] + \frac{i}{2l^2}(yn_1a_1-xn_2a_2 ) \bigg \}\end{aligned}\ ] ] where @xmath70 , @xmath71 are the 2d - lattice periods along @xmath72 and @xmath73 directions respectively and the normalization constant @xmath74 is given by eq.([eq.3.2 ] )
. therefore , eq . ( [ eq.3.9 ] ) represents modified bloch functions which are formally expanded in the corresponding magnetic wannier functions , that is , functions strongly localized on the lattice sites .
note that , for commensurate magnetic fields ( [ eq.2.7 ] ) one can recover bloch@xmath4s theorem [ eq.3.10 ] ( ) _ n,()=_n , ( ) with the values of the magnetic crystal momentum @xmath75 restricted to the first magnetic brillouin zone .
also , @xmath76@xmath77 with different values of @xmath21 are orthogonal since they belong to different irreducible representations of m.t.g . finally , after some straightforward steps , we arrive at the following expression for the eigenfunctions eq .
( [ eq.3.9 ] ) @xmath78 } \bigg \ { \frac{1}{2l^2}[(x - n_1a_1)+i(y - n_2a_2 ) ]
\bigg \}^n \nonumber \\ & \cdot \exp \bigg \ { -\frac{1}{4l^2}\big [ ( x - n_1a_1)^2+(y - n_2a_2)^2 \big ] + \frac{i}{2l^2}(yn_1a_1-xn_2a_2 ) \bigg \ } \;\end{aligned}\ ] ] where @xmath79 with @xmath80 integer multiple of @xmath18 .
the values of crystal momentum @xmath75 are determined from the periodic boundary conditions for @xmath81 with respect to magnetic translations with periods @xmath70 and @xmath71 .
figure 1 shows the contour plots of the modulus @xmath82 for the case of a square periodic lattice @xmath83 in the presence of a perpendicular magnetic field that corresponds to two flux quanta per unit cell , for different values of the crystal momentum @xmath21 in the first brillouin zone .
it is seen that the wavefunctions possess the full symmetry of the lattice and have regularly distributed zeros as it is expected from the translational invariance of the system .
now we can find the energy spectrum of bloch electrons in a constant magnetic field and weak periodic modulation potential @xmath84 using the basis @xmath85 introduced in the preceding section 3 .
it is well - known@xcite that the matrix elements of the first part of the hamiltonian ( [ eq.2.1 ] ) which describes free electrons in a uniform magnetic field are given by [ eq.4.1 ] = ( n+ ) _
c_nn where @xmath86 is the cyclotron frequency and @xmath87 labels the landau levels . in order to find the corresponding energy spectrum of bloch electrons we first evaluate the matrix elements of the periodic potential . since @xmath88 is periodic with respect to @xmath38 it can be expressed in terms of a fourier series [ eq.4.2 ] v()= _ v _ with the fourier coefficients given by [ eq.4.3 ] v_=_a d^2r v ( ) where @xmath89 stands for the area of the unit cell in the two - dimensional lattice and @xmath90 is a reciprocal lattice vector . in the two dimensional lattice @xmath91 the potential of eq .
( [ eq.4.2 ] ) is written as [ eq.4.4 ] v(x , y)=_g_1g_2v_g_1,g_2 with @xmath92 , @xmath93 integers and @xmath70 , @xmath71 the modulation periods in @xmath72 and @xmath73 direction , respectively . using the basis ( [ eq.3.11 ] )
, the matrix elements of the potential @xmath94 can be derived for each crystal momentum @xmath95 : @xmath96\nonumber \\ \cdot & \exp \bigg [ -\frac{a_2 ^ 2}{l^2 } ( n_2-n_2')(n_2-n_2'+2g_1\frac{\phi_0}{\phi } ) -\frac{a_1 ^ 2}{l^2 } ( n_1-n_1 ' ) ( n_1-n_1'-2g_2 \frac{\phi_0}{\phi } ) \bigg ] \nonumber \\ \cdot & \exp \bigg [ i\pi\big [ ( n_1'n_2-n_2'n_1)\frac{\phi}{\phi_0}+(n_1+n_1')g_1+(n_2+n_2')g_2\big]\bigg ] i_{nn'}(g_1,g_2 ) \;\end{aligned}\ ] ] where @xmath97 and @xmath98 with @xmath99 and @xmath100 the associate laguerre polynomials .
furthermore , from the expression ( [ eq.4.5 ] ) one sees that for high magnetic fields or if the lattice points are far apart each other , the matrix elements @xmath101 approach zero unless the following conditions are valid : @xmath102 under the above conditions , which mean that the magnetic flux through a unit cell of the periodic lattice should be a rational number , the matrix elements ( [ eq.4.5 ] ) take the form [ eq.4.8 ] v_nn=_g_1g_2v_g_1g_2 i_nn(g_1,g_2 ) finally , the first order energy perturbation can be found by calculating the diagonal matrix elements from eqs .
( [ eq.4.8 ] ) and ( [ eq.4.6a ] ) : [ eq.4.9 ] v_nn(q_x , q_y)=_g_1g_2v_g_1g_2 l_n(u ) thus , the complete energy spectrum to first order approximation can be written as [ eq.4.10 ] e_n(q_x , q_y)=(n+ ) _
c+v_nn(q_x , q_y )
in this section , we present analytical results concerning the energy spectrum of superlattices with perfect rectangular or triangular symmetry applying the formalism of the preceding section 4 .
we consider smooth periodic potentials described by a few lowest fourier components with weak modulation strengths .
it is well - known that the numerical solution of the corresponding tight - binding equation@xcite for 2d - lattices in perpendicular magnetic field , shows that each landau level splits into several sub - bands when a rational number of flux quanta @xmath103 pierces the unit cell and that the corresponding gaps are exponentially small . in what follows ,
we assume that the gaps are closed due to disorder in samples of not exceptionally high mobilities , that is , we assume that the landau levels are bands neglecting the above fine structure .
this assumption is justified since the reported experiments@xcitedid not indicate that this fine structure of the energy spectrum was resolved .
we consider a two - dimensional periodic modulation with rectangular symmetry , i.e : [ eq.5.1 ] v(x , y)=v_x + v_y where @xmath104 , @xmath105 are the modulation strengths and @xmath70 @xmath71 the periods in the corresponding directions .
we assume that the modulation is weak compared to the cyclotron energy @xmath106 so that eq .
( [ eq.4.10 ] ) is valid and mixing of different landau levels can be neglected . in terms of the plane wave expansion eq .
( [ eq.4.4 ] ) , the periodic potential ( [ eq.5.1 ] ) can be described by four fourier coefficients @xmath107 applying eq .
( [ eq.4.10 ] ) we find the energy spectrum @xmath108 where f(u)=l_n(u ) , u_x=(_0/ ) u_y=(_0/ ) thus , the unperturbed landau levels broaden into bands , with width equal to @xmath109 , that oscillates with inverse magnetic flux @xmath110 . using the asymptotic form of laguerre polynomials in the large n - limit
, it can be shown that the bandwidth tends to zero when @xmath111 and reaches maximum when @xmath112 with @xmath113 an integer .
therefore , the energy bandwidths in eq .
( [ eq.5.2 ] ) oscillate with the landau index
. one important consequence of the non - zero bandwidth is finite electron - velocities leading to a finite band conductivity@xcite which is absent when the modulation is not present .
in fig.2 we plot the energy spectrum for the first two landau levels , @xmath114 , modulation strengths @xmath115 mev and for ( a ) an integer @xmath116 and ( b ) a rational @xmath117 number of flux quanta .
the square superlattice period has been taken @xmath118 nm .
it is seen that the energy spectrum reflects the symmetry of modulated square lattice .
the interesting characteristics in the energy spectrum of the weak modulated superlattice are also reflected in the density of states ( dos ) per unit area defined by [ eq.5.3 ] g(e)=_n=0^ _ q_xq_y where @xmath119 is the electro - spin degeneracy .
for the spectrum given by eq .
( [ eq.5.2 ] ) the dos per unit area takes the following form : @xmath120 ^ 2-[e-(n+1/2\hspace{3pt } ) \hbar\hspace{1pt}\omega_c - v_yf_n(u_y)\cos{\theta}]^2 \}^{-1/2}\end{aligned}\ ] ] where @xmath121 .
figure 3 plots the dos per unit area for the same values of parameters as in fig.2 .
as can be seen , the dos does not exhibit van hove singularities at the edges of each landau band reflecting the 2d nature of the electron motion in this band with finite mean velocities . finally , we have to note that similar results for the band structure of a two dimensional surface superlattice have been obtained previously by direct application of perturbation theory to a different unperturbed basis of extended landau wavefunctions.@xcite we now consider a laterally modulated 2deg with triangular symmetry , i.e : [ eq.5.5 ] v(x , y)=v_x + v_y where @xmath122 and @xmath123 .
this periodic modulation can be realized in a gaas / algaas heterostructure by means of a diblock copolymer nanolithography technique.@xcite the above periodic potential can be described by six fourier coefficients @xmath124 where we assume again the condition of weak modulation @xmath125 .
applying eq .
( [ eq.4.10 ] ) we find the energy spectrum @xmath126 where the parameters @xmath127 , @xmath128 and the function @xmath129 are defined in the preceding subsection 5.1 . as in the rectangular case ,
the landau levels have broadened into bands with a bandwidth equal to @xmath130 , that oscillates with inverse magnetic flux @xmath131 and landau band index n. in fig.4 the energy bands for @xmath132 and @xmath133 , are plotted for the same values of parameters as in the rectangular case . the energy spectrum reflects the symmetry of the modulated triangular lattice .
furthermore , the dos per unit area of the superlattice with triangular symmetry can be calculated according to eq .
( [ eq.5.3 ] ) and is given by @xmath134 ^ 2\nonumber\\ & -[e-(n+1/2\hspace{3pt } ) \hbar\hspace{1pt}\omega_c - v_yf_n(4u_y)\cos{2\theta}]^2 \}^{-1/2}\end{aligned}\ ] ] where @xmath121 .
the dos for this case has similar characteristics to the one shown in fig.3 and is not plotted .
the approximate analytic energy spectrum of the preceding subsections @xmath135 and @xmath136 , provides a good insight into the problem and can facilitate the study of other properties , e.g. , the thermodynamic and the transport properties of the superlattice . on the other hand ,
the exact energy spectrum can be obtained by diagonalizing numerically the hamiltonian matrix produced by means of eq.([eq.4.8 ] ) .
the diagonalization is carried out by truncating the matrix dimensions while including a sufficient number of landau levels .
we were careful to choose the rank of the matrix being diagonalized large enough to ensure the convergence of our result . in fig.5
the exact energy spectrum of the superlattice with square symmetry , given by eq.([eq.5.2 ] ) and the approximate one are plotted as a function of the crystal momentum @xmath137 for the first ten landau levels , where the modulation strengths are @xmath115 mev and @xmath138 .
we see that the approximate energy spectrum ( solid curve ) agrees well in the whole range of crystal momentum with the exact one ( dotted line ) .
similar agreement is found for other values of magnetic flux @xmath139 provided that the condition of weak modulation is satisfied .
in summary , we have studied the energy spectrum of electrons in lateral semiconductor superlattices subject to a perpendicular magnetic field .
such systems , in which the simultaneous influence of a bi - directional periodic potential and a magnetic field is studied , are realized nowadays in layered semiconductor structures with modern technologies of growth and lithography .
we introduced a projection operator method applied to irreducible representations of the magnetotranslation group in order to construct an appropriately symmetrized basis of bloch - type wavefunctions in terms of a new set of magnetic - consistent wannier functions .
the well - localized wannier functions in this linear combination , were produced by eigenfunctions of free electrons in a magnetic field obtained in the symmetric gauge . using the above bloch - type wavefunctions we have calculated the magnetic energy spectrum of bloch - electrons in a two dimensional periodic potential obtaining similar results to previous pertubative or numerical approaches .
the spectrum is expressed as a function of crystal - momenta , defined in a magnetic brillouin zone , which are good quantum numbers of the system . especially , we have presented explicit results concerning the band structure of short - period lateral superlattices with both square and triangular symmetry .
the obtained oscillatory behavior of energy bandwidths , in the weak modulation regime , manifest themselves in the magnetoresistance measurements where they give rise to the so called weiss oscillations in the dependance of the conductance on the magnetic field strength .
finally , we note that our approach provides a symmetry - adapted basis - functions based on group theory and handles the cases of integer and rational magnetic flux in a uniform way . for rational magnetic fluxes
it is assumed that the fine structure of landau bands can be neglected in samples of not exceptionally high mobilities , since the mini - gaps in a landau band are closed due to disorder . from the presented results
, it seems that our basis is appropriate in the intermediate regime of magnetic fields where both the lattice period and the magnetic length are comparable in magnitude .
we also expect that this basis is convenient for further both analytic and numerical calculations .
f. m. peeters and j. d. boeck , _ handbook of nanostructured materials and nanotechnology _ vol . 3 , ed .
h. s. nalwa ( new york , academic press , 2000 ) ch . 7 .
d. weiss , _ ad .
solid state phys .
_ * 31 * 341 ( 1991 ) ; r. w. winkler , j. p. kotthaus and k. ploog _ phys .
_ * 62 * 1177 ( 1989 ) .
r. a. deutschmann et al.,_phys .
lett . _ * 86 * 1857 ( 2001 ) . c. albrecht et al.,_physica b _ * 249 - 251 * 914 ( 1998 ) .
p. peierls , _ z. phys .
_ * 80 * 763 ( 1933 ) ; j. m. luttinger , _ phys .
rev . _ * 133 a * 1030 ( 1956 ) ; g. h. wannier,_rev .
phys . _ * 34 * 645 ( 1962 ) ; e. brown , _ phys .
_ * a166 * 626 ( 1968 ) ; j. zak , _ phys .
_ * 177 * 1151 ( 1969 ) . c. albrecht et al.,_phys
. rev .
lett . _ * 83 * 2234 ( 1999 ) .
l. onsager , _ philos .
mag . _ * 43 * 1006 ( 1952 ) ; a. b. pippard , _ proc . r. soc
. london _ * sera 256 * 317 ( 1964 ) ; w. g. chambers , _ phys . rev . _ * 140 * a135 ( 1965 ) ; f. h. fofstadter , _ phys . rev . _
* b 14 * 2239 ( 1976 ) .
r. a. claro and g. h. wannier _ phys .
rev . _ * b 19 * 6068 ( 1979 ) . p.g .
harper , _ proc .
london _ * a68 * 874 ( 1955 ) . c. albrecht et al . ,
lett . _ * 86 * 147 ( 2001 ) ; c. albrecht et al .
, _ physica _ * e 20 * 143 ( 2003 ) .
u. kuhl and h. stockmann , _ phys .
* 80 * 3232 ( 1998 ) .
d. richoux and v. pagneux , _ europhys .
lett . _ * 59 * 34 ( 2002 ) .
d. e. grant , a. r. long and j. h. davies _ phys .
* b 61 * 13127 ( 2000 ) .
u. rossler and m. suhrke , _ ad .
solid state phys .
* 40 * 35 ( 2000 ) .
m. langenbuch , m. suhrke and u. rossler _ europhys .
lett . _ * 61 * 520 ( 2003 ) .
n. usov,_zh .
* 94 * 305 ( 1988 ) [ _ sov . phys .
jetp _ * 67 * 2564 ( 1988 ) ] f. m. peeters and p. vasilopoulos _ phys .
* b42 * 5899 ( 1990 ) ; _ phys .
rev . _ * b46 * 4667 ( 1992 )
. i. labbe _ phys .
rev . _ * b35 * 1373 ( 1987 ) ; h. silberbauer , _ j. phys .
: contens .
matter _ * 4 * 7355 ( 1992 ) ; v. gudmundsson and r. r. gerhards , _ phys .
_ b52,16744 - 16752 ( 1995 ) .
r. ferrari , _ phys .
_ * b42 * 4598 ( 1990 ) ; r. ferrari , _ int .
j. mod .
* 9 * 3333 ( 1995 ) .
q. w. shi and k. y. szelto , _ phys .
_ * b56 * 9251 ( 1997 ) ; g. y. oh , _ phys .
_ * b60 * 1939 ( 1999 ) ; m. c. chang and m. f. yang , _ phys . rev .
_ * b69 * 115108 ( 2004 ) .
m. c. chang and q. niu , _ phys .
* b50 * 10843 ( 1994 ) .
a. jannussis , _ phys .
sol . _ * 6 * 217 ( 1964 ) .
k. chung and b. mrowka , _
z. physik _ * 259 * ( 1973 ) .
g. s. kliros , e. skuras and a. jannussis , _
nuovo cimento _ * d 12 * 879 ( 1990 ) . e. brown , _ phys . rev . _ * 102 * 1038 ( 1964 ) .
j. zak , _ phys .
* 134 a * 1602 ( 1964 ) .
p. wiegmann and a. zabrodin , _ phys .
lett . _ * 72 * 1890 ( 1994 ) .
w. florek , _ j.phys : condens matter _ * 11 * 2523 ( 1999 ) ; e. brown , _ aspects of group theory in electronic dynamics _ , _ solid state physics _ , vol . 22 , ( academic press , n. york and london , 1969 ) .
m. hamermesh,_group theory and its applications to physical problems _ , (
dover , new york , 1989 ) chap .
12 s. l. altmann,_induced representations in crystals and molecules _ , ( london , academic press , 1977 ) .
h. j. fischbeck , _ phys .
sol . _ * 38 * 11 ( 1970 ) .
d. a. hutchinson , j. c. inkson and j. m. rowe , _
solid state commun .
_ * 97 * 515 ( 1996 ) .
a. feldman and a. h. kahn , _ phys .
rev . _ * b1 * 4584 ( 1970 ) ; i. a. malkin and v. i. manko , _ sov
. phys .
jetp _ * 28 * 527 ( 1969 )
. a. jannussis,_z .
_ * 21a * 1577 ( 1961 ) ; g. gountaroulis , _ phys .
* a40 * 132 ( 1972 ) x. f. wang , p. vasilopoulos and f. m. peeters , _ phys .
rev . _ * b69 * 035331 ( 2004 ) . s. chowdhury , a. long , j. davies , k. lister and e. skuras,_proceedings of ep2ds-14 _ , p.583 ( 2001 ) ; s. chowdhury , a. long , e. skuras , j. davies , k. lister , g. panneli and c. stanley , _ phys . rev . _
* b69 * 035330 ( 2004 ) .
p. vasilopoulos and f. m. peeters , _ phys
. rev .
lett . _ * 63 * 2120 ( 1989 ) .
s. melinte et al . , _ phys .
* 92 * 036802 ( 2004 ) . for a square lattice with @xmath118 nm and @xmath138 , for different values of the crystal momentum @xmath21 in the first brillouin zone
: ( a ) @xmath140 , ( b ) @xmath141 , ( c ) @xmath142 , ( d ) @xmath143,title="fig:",scaledwidth=100.0% ] + | in this paper we present calculations on the electronic band structure of a two - dimensional lateral superlattice subject to a perpendicular magnetic field by employing a projection operator technique based on the ray - group of magnetotranslation operators .
we construct a new basis of appropriately symmetrized bloch - like wavefunctions as linear combination of well - localized magnetic - wannier functions . the magnetic field was consistently included in the wannier functions defined in terms of free - electron eigenfunctions in the presence of external magnetic field in the symmetric gauge . using the above basis , we calculate the magnetic energy spectrum of electrons in a lateral superlattice with bi - directional weak electrostatic modulation .
both a square lattice and a triangular one are considered as special cases .
our approach based on group theory handles the cases of integer and rational magnetic fluxes in a uniform way and the provided basis could be convenient for further both analytic and numerical calculations . |
early observations of the sub - mm sky , mostly carried out using the submillimeter common - user bolometer array ( scuba ) on the james clerk maxwell telescope ( jcmt ) , revealed a population of very luminous , l@xmath6@xmath710@xmath8l@xmath3 corresponding to associated star - formation rates ( sfr ) @xmath9500m@xmath3yr@xmath4 , high redshift , @xmath1@xmath71 , galaxies which are extremely faint at optical wavelengths ( e.g. , @xcite ) , the so - called sub - mm galaxies ( smgs ) .
these sources , as a population , are responsible for the release of a significant fraction of the energy generated by all galaxies over the history of the universe @xcite .
their very high ir to sub - mm fluxes , in comparison to the optical and uv outputs , indicate that a significant fraction of the luminosity in these galaxies is heavily obscured @xcite .
intriguingly , smgs contribute about 20% @xcite to the integrated star formation rate density ( @xmath10sfr ) , thus indicating that a large fraction of the star formation in the universe is hidden by gas and dust ( e.g. , * ? ? ?
* ) . mostly due to their optical faintness ,
measuring redshifts and assigning accurate multi - wavelength associations have been difficult and expensive in terms of telescope time .
there is strong evidence that the majority of the smgs are at redshifts greater than unity , with a median redshift for the population of @xmath1@xmath112 - 3 @xcite .
accurate spectroscopic redshifts and well - determined multi wavelength properties have now being derived for a few hundred smgs ( see * ? ? ? * for a review ) .
the observed gas and stellar surface densities of these smgs , when coupled to plausible spectral energy distributions ( seds ) , are significantly higher than those of low - redshift iras galaxies @xcite .
furthermore , their luminosities and masses demand that the smg phase be short - lived as compared with the age of the universe ( e.g. , * ? ? ?
the typical depletion timescale in smgs measured from molecular gas co observations is @xmath2100 - 200 myr , in contrast to the much longer , @xmath21 gyr , timescales derived for normal galaxies ( e.g , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . still , the timescales for high redshift smgs are longer than those seen in local ulirgs , primarily due to the elevated gas fractions in high - redshift galaxies @xcite .
smgs are mostly found at @xmath1@xmath71 , and luminous ( @xmath12@xmath710@xmath13@xmath14 ) and ultra - luminous ( @xmath12@xmath710@xmath8@xmath14 ) ir galaxies , ( u)lirgs @xcite are considered their analogues in the local universe . however , while ( u)lirgs in the local universe are extremely rare , with space densities @xmath210@xmath15mpc@xmath16 ( comparable to those of quasars ; @xcite ) , smgs at @[email protected] have space densities @xmath210@xmath17 higher at similar ir luminosities ( e.g. , * ? ? ?
* ; * ? ? ?
this clearly points to a rapid evolution in the most extreme ir luminous galaxies , which has been parameterized as ( 1+@xmath1)@xmath18 with @[email protected] for the total energy output attributed to these systems out to z@xmath21 @xcite .
hence , while they are relatively unimportant in the local universe , ( u)lirgs can account for up to @xmath250% of the total star formation density by @xmath1@xmath22 ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ?
the nature of the large energy output in smgs is now commonly assumed to be associated with their vigorous star formation activity .
nuclear activity is also expected to provide a contribution to the energy emission in a fraction of the systems .
early studies ( e.g. , * ? ? ?
* ) claimed that a very high fraction , @xmath275% of the radio - selected spectroscopically - identified smgs , can be associated with active galactic nuclei ( agn ) .
in contrast , a lower agn fraction , @xmath215 - 30% , was reported for the general smg population @xcite , the vast majority of them subject to heavy obscuration @xcite .
a strong connection between agn activity and strong starburst episodes has been proposed based both on observations @xcite and theory @xcite , in which these events are triggered by the major merger of gas rich galaxies .
spatially - resolved observations of smgs using the _ hubble space telescope _ have not been conclusive in this regard @xcite , finding that smgs at @xmath1=0.7 - 3.4 are in general compact with spheroid / elliptical light profiles .
similarly , recent alma observations ( e.g. , * ? ? ?
* ) have also found that brighter smgs are compact , @xmath20.2@xmath5 , in the sub - mm as well .
integral field unit ( ifu ) spectrographs , in particular those assisted by adaptive optics ( ao ) , have been critical in studying the morphologies and kinematics of smgs at high spatial resolutions .
the first comparison of the dynamical properties from sinfoni ( spectrograph for integral field observations in the near infrared ) installed at the vlt ( very large telescope ) by @xcite showed that bright smgs have larger velocity widths and are much more compact than the uv and optically selected @xmath1@xmath22 star forming galaxies ; they also have lower angular momenta and higher matter densities .
this indicates that dissipative major mergers may dominate the smg population .
early studies in the near - ir by @xcite of 6 high redshift smgs show that they consist of more than one sub - component with significant velocity offsets .
similar results were found by @xcite for eight smgs , who interpreted them as evidence that smgs are systems in the early stages of major mergers .
more recently , @xcite found evidence for the presence of a compact region characterized by broad h@xmath0 emission which can be associated with an agn together with clumps of narrow h@xmath0 emission due to star formation processes . at the same time
, they do not see evidence for ordered motions , hence consistent with a connection between smgs and galaxy mergers .
@xcite observed eight ulirgs that host agn activity and smgs at high redshift with ifu , concluding that the agn activity could be the dominant power source for driving all of the observed outflows , although star formation may also play a significant role in some of the sources .
unfortunately , given the relatively low spatial densities of smgs and the required exposure times using 8 m telescopes and state - of - the - art instrumentation , only relatively small samples of galaxies have been studied so far . in this paper , we present observations of eight smgs between @xmath1=1.4 and 2.5 using the sinfoni near - ir ifu at the european southern observatory ( eso ) very large telescope ( vlt ) .
the main goal of this work is to study the morphology and kinematics of the star forming regions in high redshift smgs using the h@xmath0 line as a tracer . in section [ sec : obs ]
we present the properties of the sample studied , the observations , data reduction and analysis .
the results of these observations are presented in section [ sec : results ] , while the discussion and conclusions are presented in section [ sec : discussion ] and [ sec : conclusions ] respectively .
we assume a @xmath19cdm cosmology with @xmath20=0.7 , @xmath21=0.27 and @xmath22=0.73 @xcite .
as described in the previous section , only relatively small samples of targeted smgs at high - z have been studied so far using ifu spectroscopy .
this paper focuses on follow - up ifu observations for eight @xmath1@xmath71 ulirgs selected based on two different and complementary criteria .
hence , it will be possible to study how these selections are related to the properties of the observed sources .
the basic properties of the sources studied in this work are presented in table [ tab_properties ] .
we categorize our sources in two groups , group a focuses on warm - dust ulirgs , while in group b we include only smgs with a suitable reference star for the ao corrections to have high spatial resolution for kinematic measurements . ccccccccccccc j033129.874 - 275722.40 & a & 03:31:29.87 & -27:57:22.40 & 1.482 & 11/2011 & h & 0.80 & 8@xmath5@xmath178@xmath5 & 2.0 & 1.2@xmath23 & 45@xmath244 + j033212.866 - 274640.89 & a & 03:32:12.86 & -27:46:40.89 & 1.930 & 11/2011 & k & 0.64 & 8@xmath5@xmath178@xmath5 & 2.5 & 2.2@xmath25 & 42@xmath249 + j033246.329 - 275327.01 & a & 03:32:46.32 & -27:53:27.01 & 1.382 & 10/2011 & h & 0.65 & 8@xmath5@xmath178@xmath5 & 2.0 & 0.4@xmath26 & 53@xmath2413 + j033249.352 - 275845.07 & a & 03:32:49.35 & -27:58:45.07 & 2.326 & 10/2011 & k & 0.86 & 8@xmath5@xmath178@xmath5 & 2.0 & 8.1@xmath27 & 56@xmath245 + l50879 & a & 03:33:37.67 & -27:46:35.26 & 2.509 & 11/2011 & k & 0.64 & 8@xmath5@xmath178@xmath5 & 2.5 & 0.8 & 46 + rgj030258 + 001016 & b & 03:02:58.94 & + 00:10:16.3 & 2.236 & 10/2012 & k & 0.34 & 3@xmath5@xmath173@xmath5 & 3.75 & 0.77@xmath28 & @xmath746 + smmj04431 + 0210 ( n4 ) & b & 04:43:07.25 & + 02:10:23.3 & 2.509 & 10/2012 & k & 0.35 & 3@xmath5@xmath173@xmath5 & 3.75 & 0.3 & 40 + smmj2135 - 0102 & b & 21:35:11.60 & -01:02:52.0 & 2.326 & 10/2012 & k & 0.34 & 3@xmath5@xmath173@xmath5 & 3.75 & 0.23@xmath29 & 30 - 60 + group a consists of several warm - dust star - forming ulirgs at @xmath1@xmath21 - 2 . while smgs are proposed to contribute to as much as half of the star formation density at early epochs @xcite , smg observations at 850 @xmath31 m naturally select colder - dust ulirgs with t@xmath32@xmath3345k and
thus potentially miss a whole subpopulation of high - z ulirgs by virtue of their warmer dust temperatures .
recent work by @xcite show that there is a population of ulirgs that are fainter at 850 @xmath31 m and have higher dust temperatures , t@xmath32@xmath1150 k ( e.g. , * ? ? ?
* ; * ? ? ?
these so - called warm - dust ulirgs may contribute significantly to the cosmic star formation rate density at its peak , similar to the smg contribution .
however , the warmer - dust temperatures implied by the absence of 850 @xmath31m - flux suggests that they might have inherently different evolutionary origins compared to the well - studied cold - dust smgs .
it is possible that these higher temperatures are due to the effects of agn heating .
we have assembled a sample of warm - dust star - forming ulirgs from spectroscopic samples of 250 @xmath31m - selected blast galaxies from the observations described by @xcite .
all of them have existing redshift measurements from long - slit spectroscopy . in four of the five galaxies (
j033246 , j033249 , j033129 and j033212 ) the redshift was determined based on the detection of the h@xmath0 emission line , while for l50879 the ly@xmath0 line was detected and identified .
j033246 was first identified in spectroscopic surveys by @xcite and @xcite .
it was later classified as a star - forming radio galaxy based on a significant sub - mm detection by @xcite .
the galaxy has a redshift of @xmath1=1.382 , as indicated by the detection of an emission line using vlt / isaac identified as h@xmath0 by @xcite .
the measured rest - frame [ nii]/h@xmath0 line flux ratio of 0.13 suggests that the emission in this galaxy is dominated by star formation activity .
j033249 is a very similar source , also classified as a vigorous star forming galaxy at @xmath1=2.326 @xcite .
two systems ( j033212 and j033129 ) in our sample were not detected by our vlt / sinfoni observations and thus are not further discussed .
sources in this group were selected roughly independently of the target physical properties . the main goal in this case is to achieve the highest possible spatial resolution for our kinematic measurements .
hence , the only smg selection criteria were the availability of a suitable reference star for the ao corrections ( which for vlt / sinfoni was a r@xmath3314 star closer than 30@xmath5 ) , a measured spectroscopic redshift of @[email protected] or @[email protected] , so that h@xmath0 would fall in either the h or k near - ir bands , and that the source is observable at high enough elevation by the vlt .
this resulted in the selection of three sources for this group : smmj2135 - 0102 , rgj0302 + 0010 and smmj04431 + 0210 ( n4 ) .
the mean redshift for these sources is @xmath1=2.189 .
smmj04431 + 0210 ( n4 ) is one of the 15 original sub - mm sources found in the scuba lens survey ( s@xmath34=7.2 mjy ; @xcite ) .
smmj2135 - 0102 was identified by @xcite in a apex / laboca 870 @xmath31 m observation of the galaxy cluster macsj2135 - 010217 . the galaxy rgj0302
+ 0010 was selected from the catalogs of smgs of @xcite .
[ [ rgj030258001016 ] ] rgj030258 + 001016 + + + + + + + + + + + + + + + + was identified as a ulirg by @xcite .
rest - frame uv spectroscopy shows strong [ civ ] emission , which together with a high [ nii]/h@xmath0 suggests the presence of agn activity and a redshift of [email protected] @xcite .
the [ o iii]@xmath354959,5007 emission - line doublet shows a narrow and redshifted broad component .
this broad component is offset from the center by about 8 kpc and can be explained by an outflow with its near - side obscured by dust @xcite .
[ [ smmj044310210-n4 ] ] smmj04431 + 0210 ( n4 ) + + + + + + + + + + + + + + + + + + + is an extremely red galaxy identified as the near - ir counterpart of the faint sub - mm source reported by @xcite , which was undetected in deep optical images @xcite .
it is located behind the @xmath1=0.18 cluster ms0440 + 02 , with an amplification factor for the background galaxy of @xmath31=4.4 @xcite .
a spectroscopic redshift of [email protected] was measured by @xcite based on the detection of an h@xmath0 , [ nii]@xmath36@xmath366583,6548 , and [ oiii]@xmath365007 emission lines using the nirspec spectrograph at keck .
the rest - frame [ nii]/h@xmath0 line flux ratio of [email protected] suggests that this emission can be explained by a narrow - line agn / liner nucleus surrounded by a resolved starburst . [
[ smmj2135 - 0102 ] ] smmj2135 - 0102 + + + + + + + + + + + + + ( also known as the `` cosmic eyelash '' ) is a gas - rich starburst galaxy .
the galaxy was identified in the field of the massive lensing cluster macsj2135 - 010217 ( @xmath1@xmath37=0.325 ) via ground - based 870@xmath38@xmath31 m imaging @xcite .
they unambiguously identified the redshift of the source as @[email protected] thanks to the detection of carbon monoxide ( co ) j=1@xmath380 emission at 34.64 ghz and derive an amplification factor for the background galaxy of @[email protected] .
later , @xcite used the iram plateau de bure interferometer and the evla to obtain maps of the co(6 - 5 ) and co(1 - 0 ) emission , finding that the molecular gas kinematics are well described by a rotationally - supported disk with an inclination - corrected rotation speed v@xmath39=320@xmath2425 km s@xmath4 , a ratio of rotational to dispersion - support of v/@[email protected] and a dynamical mass of ( [email protected])@xmath1710@xmath41 m@xmath42 within a radius of 2.5 kpc .
follow - up sub - mm studies of both molecular and atomic transitions in this galaxy reported by @xcite and @xcite found evidence for a two - phase medium and physical properties such as ism density and far - uv radiation fields similar to those observed in nearby ulirgs and central regions of starburst galaxies .
near - ir ifu observations of the sample described above were performed using the sinfoni instrument @xcite , installed at the eso vlt .
these data were obtained as part of program 088.a-0452 in october and november 2011 for sample a and program 090.a-0464 observed from october 2011 to november 2012 for sample b. for the observations of the group a sources we used sinfoni in no - ao mode , with a 250 milli - arcsec pixel scale and a field - of - view ( fov ) of 8@xmath5@xmath178@xmath5 . for group b ,
sources were selected to have a suitable ao reference star , r@xmath3314 mag closer than 30@xmath5 .
hence , observations were taken using the natural guide star ( ngs ) ao mode , using the 100 milli - arcsec pixel scale and a 3@xmath5@xmath173@xmath5 fov . since the main goal of these observations was to map the h@xmath0 emission , we used either the h - band grating , which provides a wavelength coverage of 1.45 - 1.85 @xmath31 m and a spectral resolution of 3000 , or the k - band grating with a spectral resolution of 4000 and wavelength coverage of 1.95 to 2.45 @xmath31 m , depending on the redshift of the source .
the k grating was used for all sources except for j033246 .
the total integration time per source ranged from 2 to 4 hours . since
each source occupied a small fraction of the fov , the surrounding empty regions were used for sky - subtraction , thus achieving a 100% on - source time .
each observation was broken into 1-hour observing blocks ( ob ) , which includes all overheads .
for each ob , a corresponding telluric standard star was observed .
the primary data reduction was performed using the eso sinfoni pipeline version 2.5.2 this pipeline is organized as a set of 6 stand - alone recipes .
the first step evaluates the detector linearity ; this is produced by analyzing flat fields taken with increasing intensity in order to create a map of highly non - linear bad pixels .
then , a master dark frame and a map of hot pixels are created from the median of a series of dark frames .
finally , flat field frames with nearly constant intensity are combined to produce a master flat field and a third bad - pixels map ( bpm ) .
these three bpms are then combined to produce a master bpm .
the pipeline then computes the geometric distortions by placing a fiber at different positions on the detector .
each science frame is therefore corrected for dark current , bad pixels , geometric distortions and then flat fielded and wavelength calibrated .
we then performed sky estimation and subtraction , which is critical when dealing with faint sources , such as those in our sample , at near - ir wavelengths .
since our targets have relatively small spatial extents , the sinfoni fov contains enough source - free area to provide a good sky estimation .
the pipeline provides two methods to estimate and subtract the sky contribution .
the first one , which computes the median of all images on a set as an estimation of the sky , is not very accurate especially when the sky is not stable .
the second method , which we chose for our data reduction , uses the closest frame in time as an estimation of the sky to be subtracted .
all frames are then shifted and co - added .
they contain both spectral and spatial information , and are therefore reconstructed to produce 3d cubes .
the extraction of a 2d image from the data cube for each source was done by creating a pseudo - longslit using the qfitsview software[multiblock footnote omitted ] and its imred analysis option .
the size of the extraction aperture for each source is determined as a compromise between the maximum gathered flux and the minimum sky residual contribution .
then , the apall iraf task was used to extract an integrated 1d spectrum for each source .
finally , additional sky correction , particularly useful to eliminate residuals of sky emission lines , was carried out using the skycorr package @xcite .
flux calibration and telluric corrections were performed using observations of standard stars closer than 2 hours in time and with a @xmath43@[email protected] in airmass .
specifically , we used the fitting utility for sinfoni ( fus ) package developed by dr .
krispian lowe as part of his phd thesis[multiblock footnote omitted ] . in order to estimate the spatial resolution of our observations , for each target we fitted gaussian profiles to the images of the corresponding standard stars . as presented in table 1 , the effective angular resolution for the non - ao targets ranges from 0.64@xmath5 to 0.84@xmath5 , with a mean and median of 0.7 and 0.65 respectively ( corresponding to @xmath20.59 kpc and @xmath20.55 kpc at z@xmath22 ) .
for the galaxies in group b , which correspond to ao - assisted observations , these values are : 0.34@xmath5 for mean and median ( corresponding to @xmath20.28 kpc at z@xmath22 ) .
we note that these estimates come from shorter exposures ( @xmath210s ) of brighter sources than the science exposures and are therefore most likely lower limits to the spatial resolution in our targets .
however , we expect this underestimation to be rather small , @xmath20.1@xmath5 , and thus does not affect any of our conclusions .
with the sinfoni data described above , we were able to study both the galaxy - integrated and the spatially resolved properties of the h@xmath0 emission in our galaxy sample . in this section , we present images of the near - ir continuum and h@xmath0 emission , star formation rates , and test for evidence of agn activity .
we further study the star formation rate surface density , velocity fields and velocity dispersion maps . in table
[ tab_spec_meas ] , we present the spectroscopic properties for the sources in our sample derived from the vlt / sinfoni data , including h@xmath0 fluxes , h@xmath0 luminosities amongst others .
the derived sources properties and associated sfrs , including those derived from ir observations obtained from the literature are listed in table [ tab_spec_der ] .
figure [ source_images ] shows the _ hst _ observed - frame near - ir images for j033246 and j033249 , smmj04431 + 0210 ( n4 ) and smmj2135 - 0102 . in all of them
we also present a zoom - in around the region where the sub - mm emission is detected and a reconstructed 2-d image obtained from the vlt / sinfoni data cube . for rgj0302 and l50879
only the vlt / sinfoni image is presented , as these sources were not observed by _
hst_. in all cases ,
the h@xmath0 emission is overlaid on the vlt / sinfoni image as red contours .
+ + [ [ j033246 ] ] j033246 + + + + + + + j033246 presents an extended morphology with an angular extension of 1@xmath4487 , which corresponds to a physical size of 16 kpc . the h@xmath0 emission is marginally resolved spatially , with an angular diameter of 1@xmath4469@xmath240@xmath441 , which corresponds to a physical size of [email protected] kpc .
interestingly , the region where the h@xmath0 emission originates from appears to be significantly displaced from the rest - frame optical continuum .
this is further discussed in the next section .
[ [ j033249 ] ] j033249 + + + + + + + j033249 also shows an extended morphology , with an angular extension of @xmath21@xmath4447 and a size of 11.3 kpc for region 1 and for @xmath21@xmath4428 or 10.7 kpc the region 2 , as can be seen in fig [ j033249_img ] .
the spatial distribution of h@xmath0 emission is resolved and offset from the galaxy center , with an angular extension of @xmath20@xmath4482@xmath240@xmath441 or [email protected] kpc for region 1 and @xmath20@xmath447@xmath240@xmath441 or [email protected] kpc for region 2 .
the morphology of this system is consistent with that of an early - stage merger with a projected distance between the two components of 11.4 kpc .
[ [ l50879 ] ] l50879 + + + + + + l50879 presents a rather smooth morphology with a spatial extension of @xmath22.32@xmath5 , which corresponds to @xmath210.25 kpc at the redshift of the source .
the spatial distribution of the h@xmath0 line emission shows that most of it is offset from the center of the system as discussed in the following section . for the h@xmath0 emission we measured an angular size of @xmath21@xmath442@xmath240@xmath442 or @[email protected] kpc .
[ [ rgj03020010 ] ] rgj0302 + 0010 + + + + + + + + + + + + rgj0302 + 0010 shows a disturbed morphology , as can be seen in fig [ rgj0302_img ] .
the brighter region has a spatial extension of @xmath21@xmath4415 or 9.8 kpc , while for the fainter one the size is @xmath20@xmath4435 or 3 kpc .
the h@xmath0 distribution is clumpy , presenting a larger central region ( region 1 ) of @xmath20@[email protected] in size , corresponding to @[email protected] kpc .
another region can be identified ( region 2 ) , with an angular size of @xmath20@[email protected] or [email protected] kpc .
[ [ smmj044310210-n4 - 1 ] ] smmj04431 + 0210 ( n4 ) + + + + + + + + + + + + + + + + + + + for smmj04431 + 0210 ( n4 ) , the sinfoni / vlt k band image of this region reconstructed from the data cube shows a rather clumpy morphology , with one strong and prominent source and at least one and possibly two other clumps .
the h@xmath0 emission is strongly concentrated on the southern , brightest , region .
the h@xmath0 emission is spatially resolved , showing a disturbed morphology .
we can identify two marginally - separated components .
the brighter one ( region 1 ) has a angular size of @xmath20@xmath4485 corresponding to 7.0 kpc , while the other one ( region 2 ) has a size of @xmath20@xmath4448 ( 3.9 kpc ) .
the overall h@xmath0 emission extends for @xmath21@xmath444@xmath240@xmath443 , corresponding to [email protected] kpc , uncorrected for lensing amplification . correcting for lensing magnification using a magnification factor of @xmath31=4.4 @xcite , region 1
has a size of @[email protected] kpc and region 2 of @[email protected] kpc . [
[ smmj2135 - 0102 - 1 ] ] smmj2135 - 0102 + + + + + + + + + + + + + figure [ smmj2135_img ] presents the hst / wcf3 f140w image of the region around the source smmj2135 - 0102 ( the `` cosmic eyelash '' ) , located near the massive cluster macs j2135 - 0102 .
the k - band image obtained from the vlt / sinfoni data cubes shows a clumpy morphology , in which at least two bright star - forming regions in the source plane can be clearly seen , corresponding to regions identified as y1 and y2 by @xcite .
these regions are separated by @xmath22.8 kpc and have a spatial extension of @xmath20@xmath445 , corresponding to 2.4 kpc .
we can further distinguish a smaller unresolved star forming region , labeled `` region 3 '' .
for the region y1 we measure an angular size of @xmath20@xmath446@xmath240@xmath441 corresponding to a diameter of @[email protected] kpc , while for the y2 region the size is @xmath20@xmath445@xmath240@xmath441 , corresponding to @[email protected] kpc . for region 3
we measure an angular size of @xmath20@xmath443 or 1.4 kpc , all of them uncorrected for lensing magnification . in order to correct for the distorsions of the size measurements caused by the lensing effects
, we used the hst images .
this is done by comparing the physical sizes of different images of the same source caused by lensing .
from the hst data presented by @xcite we conclude that the spatial magnification corresponds to a factor of @xmath22 . hence , applying this correction factor , we find that the y1 region has an intrinsic size of @xmath21.4 kpc , while for the y2 region the size is @xmath21.2 kpc and 0.7 kpc for region y3 .
sources j033246 , j033249 , l50879 and region 2 of smmj2135 - 0102 show a clear spatial displacement between the h@xmath0 and the continuum emission , as can be seen in figure [ source_images ] .
these spatial offsets range from @xmath20@xmath4413 to 0@xmath4450 , corresponding to @xmath21.2 - 4.2 kpc , relative to the center of the base continuum emission at nearby wavelengths .
these offsets could be explained by off - nuclear star formation , perhaps as a consequence of a recent major merger similar to the one found in ii zw 096 , where 80% of the total infrared luminosity comes from an extremely compact , red , source not associated with either nucleus of the merging galaxies @xcite .
alternately , the offset h@xmath0 distribution could result from a superwind blowing out of the galaxy , similar to the one detected in m82 @xcite . in the latter scenario
, the asymmetry would be due to obscuration of the receding wind by the galaxy , as a consequence of a viewing angle which is not edge - on .
furthermore , the h@xmath0 offset could be explained by the effects of obscuration in the direction of the nucleus .
recently , @xcite also find significant displacements between the positions of the h@xmath45-band continuum and the 870 @xmath31 m emission in smgs .
this suggests that the dusty starburst regions and the less - obscured stellar distributions are not co - located .
lccccc j033129 & 1.482 & @xmath3345 & & & + j033212 & 1.93 & @xmath3339 & & & + j033246 & 1.383 & 27@xmath244 & 45@xmath248 & 106@xmath2438 & [email protected] + j033249 & 2.327 & 42@xmath245 & 56@xmath249 & 300@xmath2430&[email protected] +
region 1 & 2.327 & 17@xmath242 & 20@xmath242 & 211@xmath2439 & + region 2 & 2.327 & 13@xmath244 & 29@xmath243 & 218@xmath2424 & + l50879 & 2.509 & 29@xmath242 & 42@xmath245 & 88@xmath2414 &
+ rgj0302 + 0010 & 2.236 & 54@xmath247 & 41@xmath249 & 173@xmath2459 & [email protected] + region 1 & & 23@xmath248 & 42@xmath2410 & 103@xmath2429 & [email protected] + region 2 & & 25@xmath245 & 47@xmath248 & 144@xmath2415 & [email protected] + smmj04431 + 0210 ( n4 ) & 2.509 & 18@xmath245 & 46@xmath242 & 430@xmath2429 & [email protected] + region 1 & & 7@xmath241 & 112@xmath2415 & 397@xmath2457 & [email protected] + region 2 & & 5@xmath241 & 30@xmath244 & 175@xmath2427 & [email protected] + smmj2135 - 0120 & 2.323 & 24@xmath242 & 62@xmath244 & 127@xmath2416 & [email protected] + region 1 & & 10@xmath242 & 62@xmath244 & 127@xmath2416 & [email protected] + region 2 & & 15@xmath243 & 76@xmath2411 & 85@xmath2426 & [email protected] + region 3 & &
6@xmath241 & 24@xmath242 & [email protected] & [email protected] * * figure [ fig_1dspec ] shows the extracted 1-d spectra for the six sources in our sample detected by our sinfoni observations .
the spectra were obtained using the pseudo - longslits defined in the source images shown in figure [ source_images ] .
the redshift of each source was confirmed by the detection of the h@xmath0 emission line . for the undetected sources , j033219 and j033212 we derived the 3@xmath46 h@xmath0 flux upper limits by measuring the noise properties of the continuum regions surrounding the expected wavelengths of the lines based on the redshift values provided by @xcite . for the detected sources
, we then used these spectra to classify each spatially - resolved region as dominated by either agn or star - formation using the classification scheme based on emission line ratios of @xcite .
ideally , this is done using a combination of line ratios such as [ nii]/h@xmath0 and [ oiii]/h@xmath47 . however , in this case only the h@xmath0 and [ nii ] emission lines are available in our sinfoni data .
hence , we adopt a [ nii]/h@[email protected] flux ratio for sources or regions classified as agn dominated , based on the @xcite classification scheme .
+ + [ [ j033246 - 1 ] ] j033246 + + + + + + + from the sinfoni h - band spectrum of j033246 we derive a redshift of @xmath1=1.383 based on the h@xmath0 and [ nii]@xmath366583 lines , fully consistent with the value reported by @xcite of @xmath1=1.382 from vlt / isaac observations .
the intrinsic h@xmath0 line width is 130@xmath2438 km s@xmath4 , in line with the value previously reported by @xcite of 150@xmath2470 kms@xmath4 .
the rest - frame [ nii]/h@xmath0 line flux ratio is [email protected] , also in accord with the value found by @xcite , which suggests that j033246 is consistent with star - forming activity .
starting from the measured spatially - integrated h@xmath0 line flux , and assuming that it is entirely due to star formation , we can estimate the star formation rate for j033246 using the relation reported by @xcite : @xmath48 = 7.9 \times 10^{-42 } l(h\alpha ) [ erg s^{-1}].\ ] ] we find a sfr of 26@xmath244 m@xmath42yr@xmath4 .
using the measured size and derived sfr , we compute a star formation rate surface of @xmath49=0.5 m@xmath42yr@xmath4kpc@xmath50 .
[ [ j033249 - 1 ] ] j033249 + + + + + + + for j033249 we measure a redshift of @xmath1=2.327 based on the h@xmath0 and [ nii]@xmath366583 lines . the rest frame [ nii]/h@xmath0 line flux ratio is [email protected] , fully consistent with the value reported by @xcite of [email protected] , while for h@xmath0 we find a fwhm of 310@xmath2430 km s@xmath4 , consistent with the value of 360@xmath2450 km s@xmath4 reported by @xcite . from the observed h@xmath0 flux
we derive a sfr of 132@xmath2417 m@xmath42yr@xmath4 . using these estimates
, we compute a star formation rate surface density of @xmath49=7.9 m@xmath42yr@xmath4kpc@xmath50 . [ [ l50879 - 1 ] ] l50879 + + + + + + in the case of l50879 , we measure a redshift of 2.5097 from the h@xmath0 and [ nii]@xmath366583 lines .
the measured rest - frame [ nii]/h@xmath0 line flux ratio is [email protected] , suggesting that l50879 is dominated by star - forming activity .
the intrinsic h@xmath0 line width is 42@xmath245 km s@xmath4 . from the integrated h@xmath0 emission of the galaxy
we derive a sfr of 60@xmath248 m@xmath42yr@xmath4 .
we then compute a star formation rate surface density of @xmath49=2.4 m@xmath42yr@xmath4kpc@xmath50 .
[ [ rgj03020010 - 1 ] ] rgj0302 + 0010 + + + + + + + + + + + + for rgj0302 + 0010 we derive a redshift of @xmath1=2.237 based on the h@xmath0 and nii@xmath366583 lines .
similar results were reported previously using keck / nirspec and vlt / isaac @xcite and nifs at gemini - north @xcite .
the intrinsic fwhm of h@xmath0 is 188@xmath2459 km s@xmath4 .
the rest frame [ nii]/h@xmath0 line flux ratio is measured to be [email protected] , marginally suggesting the presence of agn activity .
this is consistent with the values found by @xcite and @xcite of [email protected] .
similarly , the value of @xmath51(o[iii]/h@xmath47)[email protected] ( o[iii]/h@[email protected] ) previously found by @xcite , further confirms this conclusion regarding the agn activity in this galaxy . from the galaxy - wide integrated h@xmath0 emission we derive an sfr of 136@xmath2420 m@xmath42yr@xmath4 , assuming that the h@xmath0 is only due to star formation , which as we discussed above is unlikely to be the case .
region 1 has [ nii]/h@[email protected] and a line ratio [ o iii]/h@xmath47=9@xmath243 indicating that this this is the region where the agn is located , while region 2 has [ nii]/h@[email protected] , likely suggesting that it is dominated by star - forming activity .
the sfr derived from the h@xmath0 line emission is 75@xmath2420 m@xmath42yr@xmath4 for region 1 , and 105@xmath2414 m@xmath42yr@xmath4 for region 2 . using these estimates of the sizes of h@xmath0 and sfr derived from the h@xmath0 emission , we compute the star formation rate surface densities and find @xmath49=17 m@xmath42yr@xmath4kpc@xmath50 for region 1 and @xmath49=5.7 m@xmath42yr@xmath4kpc@xmath50 for region 2 . [ [ smmj044310210-n4 - 2 ] ] smmj04431 + 0210 ( n4 )
+ + + + + + + + + + + + + + + + + + + for smmj04431 + 0210 ( n4 ) we derive a redshift of @xmath1=2.5095 based on the h@xmath0 and [ nii]@xmath366583 lines , similar to the value found by @xcite of @xmath1=2.5092 from keck / nirspec observations .
the sinfoni k - band spectrum , presented in fig .
[ smmj04431_1d ] , shows a strong h@xmath0 line with a fwhm of 437@xmath2429 km s@xmath4 , somewhat smaller than the value reported by @xcite of 520@xmath2440 km s@xmath4 , and larger than the measurement by @xcite of 350@xmath2460 km s@xmath4 .
the observed [ nii]@xmath366583 line is narrower , having an intrinsic fwhm of 165@xmath2428 km s@xmath4 . the observed [ nii]@xmath366583/h@xmath0 flux ratio is [email protected] .
we find a sfr of 77@xmath2420 m@xmath42yr@xmath4 ( uncorrected for lensing magnification ) , and [email protected] m@xmath42yr@xmath4 corrected for lensing magnification , assuming an amplification factor of @xmath31=4.4 as measured by @xcite .
we find that region 1 is consistent with ionization due to an agn , given the observed flux ratio of [ nii]/h@[email protected] and a sfr of [email protected] m@xmath42yr@xmath4 , while region 2 is probably dominated by star formation processes , as it has a flux ratio of [ nii]/h@[email protected] and sfr of [email protected] m@xmath42yr@xmath4 , all of them uncorrected for lensing magnification .
combining the h@xmath0 sfr with the observed sizes of each region we can compute sfr surface densities of @xmath40@xmath52 of 2.3 m@xmath42yr@xmath4kpc@xmath50 for the entire galaxy , @xmath40@xmath52=2.4 m@xmath42yr@xmath4kpc@xmath50 for region 1 and @xmath40@xmath52=7.9 m@xmath42yr@xmath4kpc@xmath50 for region 2 . [
[ smmj2135 - 0102 - 2 ] ] smmj2135 - 0102 + + + + + + + + + + + + + figure [ smmj2135_1d ] shows the extracted one - dimensional k - band spectrum for each of the three regions in smmj2135 and the galaxy - wide integrated spectrum .
we confirm the redshift of the source at @xmath1=2.323 based on the h@xmath0 and [ nii]@xmath366583 lines , consistent with the value found by @xcite of @xmath1=2.3259 based on gbt / zpectrometer observations of co lines .
the overall [ nii]/h@xmath0 line flux ratio is found to be [email protected] suggesting that the observed emission lines are dominated by star formation activity . from the integrated h@xmath0 emission
we derive a sfr of 82@xmath244 m@xmath42yr@xmath4 , uncorrected for lensing magnification and extinction .
we measure star formation rates of 33@xmath247 m@xmath42yr@xmath4 , 54@xmath248 m@xmath42yr@xmath4 and 20@xmath243 m@xmath42yr@xmath4 for regions y1 , y2 and 3 respectively . using these estimates of the sizes of h@xmath0 present in the previous section and sfr from h@xmath0 emission for these regions
we derive star formation rate surface of @xmath49=16.8 m@xmath42yr@xmath4kpc@xmath50 for region y1 , @xmath49=37.5 m@xmath42yr@xmath4kpc@xmath50 for region y2 , and @xmath49=40.8 m@xmath42yr@xmath4kpc@xmath50 for region 3 .
in addition to the identification and redshift measurements of the galaxy responsible for the sub - mm emission , the vlt / sinfoni data cubes can be also used to study their dynamical properties.velocity fields and velocity dispersion maps are measured from the shift in observed wavelength and the width of the h@xmath0 line across each source .
figure [ fig_velmap ] shows the h@xmath0 velocity maps for four of our targets : j033246 , rgj0302 + 0010 , smmj04431 + 0210 ( n4 ) , and smmj2135 - 0102 .
for the two remaining sources , j033249 and l50879 , the signal to noise ratio of the h@xmath0 line was not high enough to produce a velocity map .
j033246 has the most symmetric velocity field , showing some evidence of rotation , with a gradient of @xmath2250 km / s from one side to the opposite .
smmj04431 + 0210 ( n4 ) is clearly irregular , both in the flux distribution and velocity profile , with no evidence of ordered motions .
finally , both rgj0302 + 0010 and smmj2135 - 0102 present a clumpy structure , with no signs of ordered motions . in about half of the targets in our sample , j033246 ,
rgj0302 + 0010 , smmj04431 + 0210 ( n4 ) and smmj2135 - 0102 , we find velocity offsets of @xmath2 few @xmath17100 km s@xmath4 between distinct galactic - scale regions , with irregular kinematics .
these velocity offsets could be explained by invoking a merger scenario , as presented by e.g. , @xcite .
this is also supported by the clearly disrupted morphologies shown by the deep _ hst _ images available for these sources .
evidence for a connection between high luminosity smgs and major mergers has been previously and extensively reported in the literature ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
we have spatially mapped the h@xmath0 line emission for a sample of six high - redshift luminous smgs . as presented in section 3.1 ,
the h@xmath0 morphologies in our sample tend to be irregular and/or clumpy .
we measure the h@xmath0 emission to be extended on scales @xmath74 kpc , with an average h@xmath0 half - light radius of @xmath33r@xmath53@[email protected] kpc .
these sizes are in good agreement with previous , seeing - limited measurements of the extent of smgs in h@xmath0 ( 416 kpc ; * ? ? ?
* ; * ? ? ?
* ) and nebular lines ( 411 kpc ; * ? ? ?
similarly , recent high - resolution measurements ( 0.3@xmath5 ) using the atacama large millimeter array ( alma ) 870 @xmath31 m of 52 bright smgs in the ultra deep survey ( uds ) field reported by @xcite yielded a median physical half - light diameter of [email protected] kpc . however using hst / wfc3 ,
@xcite reported a median half - light radius of 4.4@xmath54 kpc for a sample of 48 bright smgs in the extended chandra deep field south ( ecdf - s ) .
our h@xmath0 size measurements are also consistent with the expectations based on theoretical simulations of mergers with high gas fractions ( e.g. , @xcite ) . for three galaxies in our sample , we were able to resolve the h@xmath0 emission into two or more galactic - scale extensions / components on @xmath20.5 - 1.5 arcsec ( 3 - 11 kpc ) scales . in two of these three systems
we were able to identify possible signs of agn activity based on the observed [ nii]/h@xmath0 flux ratio .
previous ifu studies of smgs have targeted galaxies showing agn signatures in their h@xmath0 spectra @xcite , while just two systems in our sample display clear h@xmath0 agn signatures . in the remaining sources in our sample we find undisturbed morphologies , with extended spatial extensions ( size @xmath2 8 - 11 kpc ) of narrow - line h@xmath0 emission that contribute a significant fraction ( @xmath240%-90% ) of the galaxy - wide h@xmath0 emission .
furthermore , we also find a lower [ nii]/h@xmath0 flux ratio , thus strongly suggesting that the h@xmath0 component in these systems is associated with star - forming activity .
high - resolution radio continuum observations @xcite found spatial extensions of @xmath28 - 10 kpc in diameter , while a high - resolution far - ir study by @xcite revealed that this emission typically extends out to spatial scales in the range of @xmath25 - 8 kpc . in high - resolution observations of a range of co transitions ( e.g. , * ? ? ? * ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , the size measurements of smgs from co fluxes are in the range of @xmath21 - 16 kpc .
our results indicate that the smg clumps in our sample have high star formation surface densities ( section [ sec:1dspec ] ) , close to those found in local extreme environments , such as in circum - nuclear starbursts and luminous infrared galaxies . in some of our targets
, the h@xmath0 distribution appears to be offset from the observed - frame near - ir continuum emission traced by sinfoni and _ hst _ observations .
considering the much greater spatial extents found for these smgs ( @xmath2 3 - 11 kpc ) in comparison to the @xmath21-kpc sized nuclear starbursts @xcite and the @xmath2100 pc starburst regions observed in local ulirgs ( e.g. , @xcite ) , smgs appear to be undergoing this intense activity over much larger spatial scales .
this is well in line with similar conclusions reported by several studies in the past @xcite .
as described in section [ sec : vmap ] , the smgs studied in this work present a wide range of h@xmath0 kinematical properties . in order to establish if the motions observed in these galaxies are dominated by ordered ( e.g. , rotation ) or random ( as caused by e.g. , a recent major galaxy merger ) dynamics , we employ kinemetry analysis
. this is done by measuring the level of symmetry in the mean of the velocity and the velocity dispersion fields .
the kinemetry measurement procedure described by @xcite uses the method detailed in @xcite to analyze maps of kinematic moments based on the line - of - sight - velocity distribution .
this can be considered as an extension of the surface photometry to the higher - order moments of the velocity distribution .
this technique was developed to study stellar dynamics of local ellipticals @xcite , but has been extended to the dynamics of high-@xmath1 galaxies ( e.g. , * ? ? ?
more recently , @xcite used a technique based on measuring velocity and velocity dispersion asymmetries in order to establish if galaxies are dominated by rotation or random motions .
they applied this technique to smgs similar to the ones in our sample .
asymmetries in moment fields are measured by fitting ellipses , varying the position angle and ellipticity , to the velocity and dispersion fields .
these coefficients , v@xmath55 and @xmath46@xmath55 , are used to establish the level of asymmetry .
the two asymmetry measures are combined as k@xmath55 = @xmath56 . following the definition of @xcite , systems with k@[email protected] are classified as mergers .
any asymmetries in these fields represent deviations from the idealized model ( a rotating thin disk ) and as such the combined asymmetry of the two fields can fully describe how accurately a system is represented by this idealized disk .
this model displays an ordered velocity field , peaking at the semi - major axis of the ellipse and reaching a value of zero at the semi - minor axis , and a centrally peaked velocity dispersion field . in fig .
[ fig_kinemetry ] we compare the levels of asymmetry in the velocity and velocity dispersion for the four smgs in our sample for which velocity maps were obtained . as can be seen on that figure , and perhaps surprisingly , 3/4 of the galaxies in the smg sample ( j033246 , rgj0302 + 0010 and smmj2135 - 0102 ) are found in the region where disk - dominated sources are located .
this is somewhat expected for j033246 , since as discussed above , this source appears to be dominated by rotation .
the other two sources appear to be clumpy in their h@xmath0 morphologies , which might explain why their asymmetry values are lower than expected , as each individual clump is spatially unresolved and thus does not show significant asymmetries .
hence , their classifications based on the derived asymmetry values are probably explained by the limitations of the spatial resolution of our observations .
finally , smmj04431 + 0210 ( n4 ) clearly shows very high asymmetry levels , as can be seen in figure [ fig_velmap ] .
therefore , it is not surprising that this source is also classified as `` merger - dominated '' in this diagram .
hence , the kinemetry analysis also suggests that most of these sources are dominated by random motions .
this is in line with previous observations of similar high - luminosity ir and sub - mm galaxies at @xmath1@xmath22 , such as those reported by @xcite and @xcite among others .
a recent analysis presented by @xcite , which takes a sample of local ir - luminous galaxies and artificially redshifts their h@xmath0 ifu data , shows that post - coalescence mergers may also display kinematics that can be wrongly classified as `` disk like '' , as appears to be the case for some of our sources .
@xmath22 in the sins sample .
the background blue points are template disks and red points are mergers , from the library of @xcite .
while 3 out of the four sources , j033246 , rgj0302 + 0010 and smmj2135 - 0102 are located in the `` disk '' region of the diagram , only in one of them ( j033246 ) did we find evidence for ordered motions . in the other two cases , these are clearly clumpy galaxies in which each individual clump is spatially unresolved .
smmj04431 + 0210 ( n4 ) is clearly in the `` mergers '' part of the diagram .
[ fig_kinemetry ] ] we now compare the star formation rates for the galaxies in our sample , previously derived from far - ir observations , with those obtained from the h@xmath0 observations described here .
the implied star formation rates , including those derived from ir observations obtained from the literature are presented in table [ tab_spec_der ] .
specifically , we use the values reported by @xcite for j033246 and j033249 from spitzer - mips , blast and laboca observations . for rgj0302 + 0010 the values
were derived by @xcite fitting model spectral energy distributions ( seds ) to the observed 850 @xmath31 m and 1.4 ghz fluxes assuming that the local far - ir / radio correlation @xcite holds at higher redshifts . for smmj04331 + 0210
we used the values obtained by @xcite from observed 850@xmath31 m flux densities and assuming a modified grey - body model ( t=40 k ) and a frequency - dependent emissivity .
finally , for smmj2135 - 0120 values were derived by @xcite from herschel , scuba-2 , vla and apex observations .
both sfr measurements for smmj2135 - 0120 were corrected for lensing magnification assuming a factor of @[email protected] @xcite , while a magnification factor of @xmath2=4.4 @xcite was assumed for smmj04431 + 0210 ( n4 ) .
it should be noted that using the same lensing magnification for the h@xmath0 and fir fluxes is not ideal , as these emissions might come from slightly different regions and thus might present small variations in their magnifications . in figure [ fig_starformation
] , we compare the sfr derived from the h@xmath0 luminosity with the fir sfrs for the galaxies in our sample , without correcting the h@xmath0 measurements for extinction . we also include in this comparison a sample of 30 high - redshift far - infrared luminous galaxies from @xcite .
the h@xmath0 determined sfrs are significantly lower than the values determined from the ir luminosity , by factors of @xmath23 - 5 .
this is most likely explained by both the presence of large amounts of dust obscuration in these systems and if there are fully - dust - obscured sf regions in the galaxy that do not contribute much to the balmer emission lines .
unfortunately , the wavelength coverage of our spectroscopic observations does not extend to the @xmath57 emission line , and thus the reddening in these galaxies can not be estimated directly from the balmer decrement , which would test this hypothesis . however
, if we assume an average extinction value of a@[email protected] , as previously used by @xcite and measured by @xcite , we find that the extinction - corrected h@xmath0 sfrs are in good agreement with those derived from the ir .
the scatter seen in figure [ fig_starformation ] could be explained by a combination of the fact that most of the star formation in these galaxies is not occurring in fully - dust - obscured regions , and due to the morphological diversity of sub - mm / ulirgs - selected galaxies .
estimates of the sfr from the h@xmath0 flux in smgs retain the substantial caveat that in the presence of an agn , the blended nuclear emission may result in the broadening and brightening of the h@xmath0 emission , potentially leading to overestimates of the sfr . due to the large gas and dust reservoirs of smgs , which could potentially provide ample fuel to trigger an agn that could contaminate our estimates of sfr and @xmath10sfr . however , we note that only two sources in our sample show some evidence for the presence of an agn , either via elevated [ nii]/h@xmath0 ratios or a broad line component .
previous ifu studies of smgs have mainly targeted galaxies showing strong agn signatures in their h@xmath0 spectra @xcite , whereas the majority of our sample does not display clear h@xmath0 agn signatures . in smgs , agn were found to only contribute at low levels ( @xmath3320% ) to the bolometric output @xcite .
however , even if an agn is not dominating the bolometric output it may still affect the h@xmath0 distribution and dynamics , being energetic enough to drive ionized gas over scales of a few kpc @xcite .
lccccc j033246 & 26@xmath244 & 692@xmath59 & 14.5 & 0.5 & [email protected] + j033249 & 132@xmath2417 & 14084@xmath60 & 8.2 & 7.9 &
+ region 1 & 58@xmath246 & & 6.8 & 5.0 & + region 2 & 46@xmath2414 & & 5.8 & 5.5 & + rgj0302 + 0010 & 171@xmath2422 & 1340@xmath61 & & & [email protected] + region 1 & 73@xmath2425 & & 4.2 & 17.0 &
+ region 2 & 75@xmath2416 & & 3.2 & 5.7 &
+ l50879 & 60@xmath248 & & 10 & 2.4 & + smmj04431 + 0210 ( n4 ) & 77@xmath2420 & 520 & 11.5 & 2.3 & [email protected] + region 1 & 30@xmath244 & & 7.0 & 2.4 & + region 2 & 20@xmath243 & & 3.9 & 7.9 &
+ smmj2135 - 0120 & 82@xmath244 & 398@xmath62 & & & [email protected] + region 1 & 33@xmath247 & &2.8 & 16.8 &
+ region 2 & 54@xmath248 & &2.4 & 37.5 & + region 3 & 20@xmath243 & & 1.4 & 40.8 & ) versus sfr(fir ) .
the _ red points _ are measured values from the present work .
sfr(h@xmath0 ) values were derived from the h@xmath0 luminosities obtained from our sinfoni observations ( not corrected for extinction ) .
the fir sfrs for j033246 , j033249 , rgj0302 + 0010 , smmj04431 + 0210 ( n4 ) and smmj2135 - 0210 were derived from fir luminosities in the literature using the kennicutt relations @xcite . for smmj04431 + 0210 ( n4 )
the sfr values were corrected for lensing magnification assuming a @xmath31=4.4 @xcite .
the sfr values for smmj2135 - 0210 were corrected for lensing magnification using an amplification factor of @[email protected] derived by @xcite .
the _ blue circles _ show measurements of high - redshift far - infrared luminous galaxies from @xcite .
the gray dashed lines correspond to ratios of sfr(h@xmath0 ) to sfr(fir ) of 0.01 , 0.05 , 0.1 , 0.2 , 0.3.[fig_starformation],scaledwidth=50.0% ] we now use the h@xmath0-derived sizes for these smgs to compare the sfr surface densities determined from h@xmath0 and fir luminosities with the molecular gas surface density .
two of our sources have previously been observed in co. due to the uncertainties in the conversion between l@xmath63 and the h@xmath64 mass ( e.g. , * ? ? ? * ) , we consider both @xmath65 values of 4.6 and 0.8 k km s@xmath4 pc@xmath50 @xmath4 , corresponding to spiral galaxies @xcite and to ulirgs & star - forming galaxies at high redshift @xcite , respectively .
for smmj04431 + 0210 ( n4 ) we considered a line luminosity l@xmath63 [email protected] @xmath1710@xmath41 k km s@xmath4pc@xmath50 ( corrected for the lensing magnification ) , and co diameter of 1.8 kpc , as reported by @xcite . with these measurements , we compute gas surface densities for @xmath65=0.8 and @xmath65=4.6 of 3.1@xmath1710@xmath66 m@xmath42pc@xmath50 and 18 @xmath1710@xmath66 m@xmath42pc@xmath50 respectively .
for the galaxy smmj2135 - 0120 we use the line luminosities l@xmath63 reported by @xcite based on evla observations , corrected for lensing magnification .
the reported luminosities are ( 3.9@xmath24 0.4)@xmath1710@xmath67 k km s@xmath4pc@xmath50 for region y1 and ( 7.1@xmath24 0.8)@xmath1710@xmath67k km s@xmath4pc@xmath50 for region y2 .
we use the h@xmath0 size measurements of 1.4 kpc for region y1 and 1.2 kpc for region y2 .
then , we derive gas surface densities , assuming @xmath65=0.8 and @xmath65=4.6 , of 6.3@xmath1710@xmath68 m@xmath42pc@xmath50 and 3.6@xmath1710@xmath66 m@xmath42pc@xmath50 for region y1 and 1.5 @xmath1710@xmath66 m@xmath42pc@xmath50 and 9@xmath1710@xmath66 m@xmath42pc@xmath50 for region y2 , respectively and corrected for lensing magnification . for the other four sources , j033246 , j033249 , rgj0302 + 0010 and l50879 , there are no observations of co emission available so far .
figure [ fig_ks ] shows the gas surface density versus the sfr surface density for the sources in our sample along with local and high - z sources from the literature .
the values from the literature were taken directly from each publication . as can be seen in figure [ fig_ks ] , in all three sources for which we could measure @xmath40gas and @xmath40(sfr ) ,
they appear to be slightly below ( but consistent ) with the relation derived for local and high redshift star forming galaxies ( e.g. , * ? ? ?
* ; * ? ? ?
this is in contrast with previous results reported by @xcite and @xcite , who found a higher relation for both local and high redshift ulirgs and smg .
while here we do not attempt to correct the values previously reported for different assumed @xmath65 values , for our sources we assume a wide range of @xmath65 , from 0.8 to 4.6 , finding that our sources still fall below the ulirg and smg relation and appear to be consistent with that for normal sfgs . while this is certainly interesting , it is hard to reach stronger conclusions given the small sample studied here and the lack of high resolution co data . work to increase the number of sources in our sample and to obtain alma observations for them is currently being carried out .
= 0.8 and @xmath65=4.6 . _ red lines _ show measurements for the smgs in our sample but using sfr between sfr(fir ) and sfr(h@xmath0 ) instead .
the _ blue squares _ present the location of the bzk galaxies in the sample of @xcite , assuming a @xmath0=3.2 value . _
yellow squares _ show the sfrgs and smgs in the @xcite sample , using @xmath65=3.2 for sfgs and 1 for z@xmath71 smgs .
_ green diamonds _ show the smg sample of @xcite , using @xmath65=0.8 . _
black triangles _
show normal spiral galaxies , while the _ black pluses _ present the starburst galaxies from the @xcite sample , which were used to derive the relation shown by the _ dashed black line _ , a power law with a slope of 1.4 .
the _ dotted black line _ indicates the relation found by for @xcite , a power law with a slope of 1.42 and a higher normalization by 0.9 dex.,scaledwidth=50.0% ]
in this paper we used the vlt / sinfoni integral field spectrometer to study , at high resolution , the spatial distribution and kinematics of eight smgs at 1.3@xmath33@[email protected] using the h@xmath0 emission line as a tracer .
these observations allow us to study the gas dynamics and morphologies for six sources in our sample for which the h@xmath0 line was detected at sufficient signal to noise ratio .
overall , we find that these smgs have highly irregular and/or clumpy morphologies .
the h@xmath0 emission has relatively large spatial extent , @xmath69 kpc , with a mean value for our sample of @xmath33r@xmath70@xmath7= 6.4 kpc , in contrast to local ulirgs , which tend to be more compact .
we find that in three cases the h@xmath0 emission is significantly offset from the observed - frame near - ir continuum emission derived from the sinfoni and hst images .
we then analyzed the velocity and velocity dispersion fields in this sample , finding that these smgs in general show large velocity gradients across each system .
the majority of the smgs in our sample are not consistent with being disk - like systems dominated by rotation and instead present irregular and turbulent or clumpy velocity and velocity dispersion fields , which could be explained by a recent major merger .
we find that of the six smgs with resolved spectroscopy , at least three appear to comprised of two or more dynamical sub - components .
the average velocity offsets between these components is @xmath2180 km s@xmath4 across a projected spatial scale of @xmath28 kpc
. the obvious merging / interacting nature of these systems suggests that they are analogous to the typically less luminous and slightly more compact ulirgs in the local universe .
this is confirmed by a kinemetry analysis performed for the sources in our sample , which suggests that most of them do not show strong evidence for the presence of ordered motions .
the smgs in our sample display high sfr surface densities ( @xmath40@xmath52 ) , similar to those found in the most extreme local environments , such as circum - nuclear starbursts and ir - luminous galaxies but over larger spatial extents .
our ifu observations allow us to disentangle agn and starburst - like components ( from [ n ii]/h@xmath0 flux ratios ) .
two of our targets ( rj0302 - 0010 and smmj04431 + 0210 ( n4 ) ) show possible signs of agn activity surrounded by star - forming regions / clumps .
we then further confirm that these extreme smgs at high-@xmath1 appear to follow the same star - formation scaling relations as less luminous `` normal '' star forming galaxies .
smgs in our group b sample are the only ones which present evidence for agn activity .
therefore the data do not support the hypothesis that the agn leads to hotter dust , at least for the sources in our group a , which contain the warm - dust ulirgs . while still very expensive and limited in sample size , rest - frame optical ifu observations of high-@xmath1 extreme star - forming galaxies are starting to reveal the details of the violent star formation episodes in the early universe . in the near future ,
such studies will be possible for much larger samples thanks to e.g. , the kmos multi - object ifu spectrograph at the vlt .
in addition , alma observations will allow studies of the molecular gas contents , fuel for both star formation and for the central supermassive black holes , for these galaxies .
we thank the anonymous referee for very useful comments and suggestions .
support for this work was provided by the center of excellence in astrophysics and associated technologies ( pfb 06 ; et , vo , feb ) , by the fondecyt regular grants 1120061 , 1130528 and 1160999 ( et ) and 1141218 ( feb ) , and by the `` embiggen '' conicyt anillo project act1101 ( et , vo , feb , gcp ) .
we further acknowledge support from the fondecyt postdoctorado grant 3150361 ( gcp ) and the ministry of economy , development , and tourism millennium science initiative through grant ic120009 , awarded to the millennium institute of astrophysics , mas ( feb ) .
irs acknowledges support from stfc ( st / l007sx/1 ) , the erc advanced grant dustygal 3213324 and a royal society wolfson research merit award . , m. s. , chapman , s. c. , tacconi , l. , smail , i. , ivison , r. j. , casey , c. m. , bertoldi , f. , beswick , r. , biggs , a. , blain , a. w. , cox , p. , genzel , r. , greve , t. r. , kennicutt , r. , muxlow , t. , neri , r. , & omont , a. 2010 , , 405 , 219 , m. s. , smail , i. , chapman , s. c. , genzel , r. , ivison , r. j. , tacconi , l. j. , alaghband - zadeh , s. , bertoldi , f. , blain , a. w. , casey , c. m. , cox , p. , greve , t. r. , lutz , d. , neri , r. , omont , a. , & swinbank , a. m. 2013 , , 429 , 3047 , n. , cresci , g. , davies , r. , eisenhauer , f. , frster schreiber , n. m. , genzel , r. , gillessen , s. , lehnert , m. , lutz , d. , nesvadba , n. , shapiro , k. l. , sternberg , a. , tacconi , l. j. , verma , a. , cimatti , a. , daddi , e. , renzini , a. , erb , d. k. , shapley , a. , & steidel , c. c. 2007 , , 671 , 303 , c. m. , scoville , n. z. , sanders , d. b. , lee , n. , cooray , a. , finkelstein , s. l. , capak , p. , conley , a. , de zotti , g. , farrah , d. , fu , h. , le floch , e. , ilbert , o. , ivison , r. j. , & takeuchi , t. t. 2014 , , 796 , 95 , c .- c . ,
smail , i. , swinbank , a. m. , simpson , j. m. , ma , c .- j . , alexander , d. m. , biggs , a. d. , brandt , w. n. , chapman , s. c. , coppin , k. e. k. , danielson , a. l. r. , dannerbauer , h. , edge , a. c. , greve , t. r. , ivison , r. j. , karim , a. , menten , k. m. , schinnerer , e. , walter , f. , wardlow , j. l. , wei , a. , & van der werf , p. p. 2015
, , 799 , 194 , y. , bacon , r. , bureau , m. , davies , r. l. , emsellem , e. , kuntschner , h. , miller , b. , peletier , r. , verolme , e. k. , & de zeeuw , p. t. 2001 , in sf2a-2001 : semaine de lastrophysique francaise , ed . f. combes , d. barret , & f. thvenin , 289 , a. l. r. , swinbank , a. m. , smail , i. , cox , p. , edge , a. c. , weiss , a. , harris , a. i. , baker , a. j. , de breuck , c. , geach , j. e. , ivison , r. j. , krips , m. , lundgren , a. , longmore , s. , neri , r. , & flaquer , b. o. 2011 , , 410 , 1687 , f. et al .
2003 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 4841 , society of photo - optical instrumentation engineers ( spie ) conference series , ed . m. iye & a. f. m. moorwood , 15481561 , n. m. , genzel , r. , bouch , n. , cresci , g. , davies , r. , buschkamp , p. , shapiro , k. , tacconi , l. j. , hicks , e. k. s. , genel , s. , shapley , a. e. , erb , d. k. , steidel , c. c. , lutz , d. , eisenhauer , f. , gillessen , s. , sternberg , a. , renzini , a. , cimatti , a. , daddi , e. , kurk , j. , lilly , s. , kong , x. , lehnert , m. d. , nesvadba , n. , verma , a. , mccracken , h. , arimoto , n. , mignoli , m. , & onodera , m. 2009 , , 706 , 1364 , r. , tacconi , l. j. , gracia - carpio , j. , sternberg , a. , cooper , m. c. , shapiro , k. , bolatto , a. , bouch , n. , bournaud , f. , burkert , a. , combes , f. , comerford , j. , cox , p. , davis , m. , schreiber , n. m. f. , garcia - burillo , s. , lutz , d. , naab , t. , neri , r. , omont , a. , shapley , a. , & weiner , b. 2010 , , 407 , 2091 , r. , tacconi , l. j. , lutz , d. , saintonge , a. , berta , s. , magnelli , b. , combes , f. , garca - burillo , s. , neri , r. , bolatto , a. , contini , t. , lilly , s. , boissier , j. , boone , f. , bouch , n. , bournaud , f. , burkert , a. , carollo , m. , colina , l. , cooper , m. c. , cox , p. , feruglio , c. , frster schreiber , n. m. , freundlich , j. , gracia - carpio , j. , juneau , s. , kovac , k. , lippa , m. , naab , t. , salome , p. , renzini , a. , sternberg , a. , walter , f. , weiner , b. , weiss , a. , & wuyts , s. 2015 , , 800 , 20 , c. , pozzi , f. , rodighiero , g. , delvecchio , i. , berta , s. , pozzetti , l. , zamorani , g. , andreani , p. , cimatti , a. , ilbert , o. , le floch , e. , lutz , d. , magnelli , b. , marchetti , l. , monaco , p. , nordon , r. , oliver , s. , popesso , p. , riguccini , l. , roseboom , i. , rosario , d. j. , sargent , m. , vaccari , m. , altieri , b. , aussel , h. , bongiovanni , a. , cepa , j. , daddi , e. , domnguez - snchez , h. , elbaz , d. , frster schreiber , n. , genzel , r. , iribarrem , a. , magliocchetti , m. , maiolino , r. , poglitsch , a. , prez garca , a. , sanchez - portal , m. , sturm , e. , tacconi , l. , valtchanov , i. , amblard , a. , arumugam , v. , bethermin , m. , bock , j. , boselli , a. , buat , v. , burgarella , d. , castro - rodrguez , n. , cava , a. , chanial , p. , clements , d. l. , conley , a. , cooray , a. , dowell , c. d. , dwek , e. , eales , s. , franceschini , a. , glenn , j. , griffin , m. , hatziminaoglou , e. , ibar , e. , isaak , k. , ivison , r. j. , lagache , g. , levenson , l. , lu , n. , madden , s. , maffei , b. , mainetti , g. , nguyen , h. t. , ohalloran , b. , page , m. j. , panuzzo , p. , papageorgiou , a. , pearson , c. p. , prez - fournon , i. , pohlen , m. , rigopoulou , d. , rowan - robinson , m. , schulz , b. , scott , d. , seymour , n. , shupe , d. l. , smith , a. j. , stevens , j. a. , symeonidis , m. , trichas , m. , tugwell , k. e. , vigroux , l. , wang , l. , wright , g. , xu , c. k. , zemcov , m. , bardelli , s. , carollo , m. , contini , t. , le fvre , o. , lilly , s. , mainieri , v. , renzini , a. , scodeggio , m. , & zucca , e. 2013 , , 432 , 23 , c. m. , alexander , d. m. , swinbank , a. m. , smail , i. , alaghband - zadeh , s. , bauer , f. e. , chapman , s. c. , del moro , a. , hickox , r. c. , ivison , r. j. , menndez - delmestre , k. , mullaney , j. r. , & nesvadba , n. p. h. 2012 , , 426 , 1073 , d. h. , serjeant , s. , dunlop , j. , rowan - robinson , m. , blain , a. , mann , r. g. , ivison , r. , peacock , j. , efstathiou , a. , gear , w. , oliver , s. , lawrence , a. , longair , m. , goldschmidt , p. , & jenness , t. 1998 , , 394 , 241 , h. , armus , l. , surace , j. a. , mazzarella , j. m. , evans , a. s. , sanders , d. b. , howell , j. h. , petric , a. , vavilkin , t. , iwasawa , k. , haan , s. , murphy , e. j. , stierwalt , s. , appleton , p. n. , barnes , j. e. , bothun , g. , bridge , c. r. , chan , b. , charmandaris , v. , frayer , d. t. , kewley , l. j. , kim , d. c. , lord , s. , madore , b. f. , marshall , j. a. , matsuhara , h. , melbourne , j. e. , rich , j. , schulz , b. , spoon , h. w. w. , sturm , e. , u , v. , veilleux , s. , & xu , k. 2010 , , 140 , 63 , r. j. , swinbank , a. m. , swinyard , b. , smail , i. , pearson , c. p. , rigopoulou , d. , polehampton , e. , baluteau , j .-
p . , barlow , m. j. , blain , a. w. , bock , j. , clements , d. l. , coppin , k. , cooray , a. , danielson , a. , dwek , e. , edge , a. c. , franceschini , a. , fulton , t. , glenn , j. , griffin , m. , isaak , k. , leeks , s. , lim , t. , naylor , d. , oliver , s. j. , page , m. j. , prez fournon , i. , rowan - robinson , m. , savini , g. , scott , d. , spencer , l. , valtchanov , i. , vigroux , l. , & wright , g. s. 2010 , , 518 , l35 , s. p. , wilson , g. w. , wang , q. d. , williams , c. c. , scott , k. s. , yun , m. s. , pope , a. , lowenthal , j. , aretxaga , i. , hughes , d. , kim , m. j. , kim , s. , tamura , y. , kohno , k. , ezawa , h. , kawabe , r. , & oshima , t. 2013 , , 431 , 662 , k. l. , genzel , r. , frster schreiber , n. m. , tacconi , l. j. , bouch , n. , cresci , g. , davies , r. , eisenhauer , f. , johansson , p. h. , krajnovi , d. , lutz , d. , naab , t. , arimoto , n. , arribas , s. , cimatti , a. , colina , l. , daddi , e. , daigle , o. , erb , d. , hernandez , o. , kong , x. , mignoli , m. , onodera , m. , renzini , a. , shapley , a. , & steidel , c. 2008 , , 682 , 231 , j. m. , smail , i. , swinbank , a. m. , almaini , o. , blain , a. w. , bremer , m. n. , chapman , s. c. , chen , c .- c . , conselice , c. , coppin , k. e. k. , danielson , a. l. r. , dunlop , j. s. , edge , a. c. , farrah , d. , geach , j. e. , hartley , w. g. , ivison , r. j. , karim , a. , lani , c. , ma , c .- j . , meijerink , r. , michaowski , m. j. , mortlock , a. , scott , d. , simpson , c. j. , spaans , m. , thomson , a. p. , van kampen , e. , & van der werf , p. p. 2015
, , 799 , 81 , j. m. , smail , i. , swinbank , a. m. , chapman , s. c. , geach , j. e. , ivison , r. j. , thomson , a. p. , aretxaga , i. , blain , a. w. , cowley , w. i. , chen , c .- c . , coppin , k. e. k. , dunlop , j. s. , edge , a. c. , farrah , d. , ibar , e. , karim , a. , knudsen , k. k. , meijerink , r. , michaowski , m. j. , scott , d. , spaans , m. , & van der werf , p.
, , 807 , 128 , j. m. , swinbank , a. m. , smail , i. , alexander , d. m. , brandt , w. n. , bertoldi , f. , de breuck , c. , chapman , s. c. , coppin , k. e. k. , da cunha , e. , danielson , a. l. r. , dannerbauer , h. , greve , t. r. , hodge , j. a. , ivison , r. j. , karim , a. , knudsen , k. k. , poggianti , b. m. , schinnerer , e. , thomson , a. p. , walter , f. , wardlow , j. l. , wei , a. , & van der werf , p.
, , 788 , 125 , a. m. , simpson , j. m. , smail , i. , harrison , c. m. , hodge , j. a. , karim , a. , walter , f. , alexander , d. m. , brandt , w. n. , de breuck , c. , da cunha , e. , chapman , s. c. , coppin , k. e. k. , danielson , a. l. r. , dannerbauer , h. , decarli , r. , greve , t. r. , ivison , r. j. , knudsen , k. k. , lagos , c. d. p. , schinnerer , e. , thomson , a. p. , wardlow , j. l. , wei , a. , & van der werf , p. 2014
, , 438 , 1267 , a. m. , smail , i. , longmore , s. , harris ,
a. i. , baker , a. j. , de breuck , c. , richard , j. , edge , a. c. , ivison , r. j. , blundell , r. , coppin , k. e. k. , cox , p. , gurwell , m. , hainline , l. j. , krips , m. , lundgren , a. , neri , r. , siana , b. , siringo , g. , stark , d. p. , wilner , d. , & younger , j. d. 2010 , , 464 , 733 , l. j. , genzel , r. , neri , r. , cox , p. , cooper , m. c. , shapiro , k. , bolatto , a. , bouch , n. , bournaud , f. , burkert , a. , combes , f. , comerford , j. , davis , m. , schreiber , n. m. f. , garcia - burillo , s. , gracia - carpio , j. , lutz , d. , naab , t. , omont , a. , shapley , a. , sternberg , a. , & weiner , b. 2010 , , 463 , 781 , l. j. , genzel , r. , smail , i. , neri , r. , chapman , s. c. , ivison , r. j. , blain , a. , cox , p. , omont , a. , bertoldi , f. , greve , t. , frster schreiber , n. m. , genel , s. , lutz , d. , swinbank , a. m. , shapley , a. e. , erb , d. k. , cimatti , a. , daddi , e. , & baker , a. j. 2008 , , 680 , 246 , l. j. , neri , r. , genzel , r. , combes , f. , bolatto , a. , cooper , m. c. , wuyts , s. , bournaud , f. , burkert , a. , comerford , j. , cox , p. , davis , m. , frster schreiber , n. m. , garca - burillo , s. , gracia - carpio , j. , lutz , d. , naab , t. , newman , s. , omont , a. , saintonge , a. , shapiro griffin , k. , shapley , a. , sternberg , a. , & weiner , b. 2013 , , 768 , 74 , s. x. , brandt , w. n. , luo , b. , smail , i. , alexander , d. m. , danielson , a. l. r. , hodge , j. a. , karim , a. , lehmer , b. d. , simpson , j. m. , swinbank , a. m. , walter , f. , wardlow , j. l. , xue , y. q. , chapman , s. c. , coppin , k. e. k. , dannerbauer , h. , de breuck , c. , menten , k. m. , & van der werf , p. 2013
, , 778 , 179 , a. , kovcs , a. , coppin , k. , greve , t. r. , walter , f. , smail , i. , dunlop , j. s. , knudsen , k. k. , alexander , d. m. , bertoldi , f. , brandt , w. n. , chapman , s. c. , cox , p. , dannerbauer , h. , de breuck , c. , gawiser , e. , ivison , r. j. , lutz , d. , menten , k. m. , koekemoer , a. m. , kreysa , e. , kurczynski , p. , rix , h .- w . , schinnerer , e. , & van der werf , p.
, , 707 , 1201 | we present near - infrared integral - field spectroscopic observations targeting h@xmath0 in eight sub - millimeter galaxies ( smgs ) at @xmath1=1.3 - 2.5 using vlt / sinfoni , obtaining significant detections for six of them .
the star formation rates derived from the h@xmath0 emission are @xmath2100 m@xmath3yr@xmath4 , which account for only @xmath2 20 - 30% of the infrared - derived values , thus suggesting that these systems are very dusty .
two of these systems present [ nii]/h@xmath0 ratios indicative of the presence of an active galactic nucleus ( agn ) .
we mapped the spatial distribution and kinematics of the star forming regions in these galaxies on kpc - scales .
in general , the h@xmath0 morphologies tend to be highly irregular and/or clumpy , showing spatial extents of @xmath23 - 11 kpc .
we find evidence for significant spatial offsets , of @xmath20.1 - 0.4@xmath5 or 1.2 - 3.4 kpc , between the h@xmath0 and the continuum emission in three of the sources . performing a kinemetry analysis
we conclude that the majority of the sample is not consistent with disk - like rotation - dominated kinematics . instead
, they tend to show irregular and/or clumpy and turbulent velocity and velocity dispersion fields .
this can be interpreted as evidence for scenario in which these extreme star formation episodes are triggered by galaxy - galaxy interactions and major mergers .
in contrast to recent results for smgs , these sources appear to follow the same relations between gas and star forming rate densities as less luminous and/or normal star forming galaxies . |
in condensed matter physics , the goal of generalizing su(2 ) lattice fermion or spin models to those with high symmetries of su(@xmath1 ) @xcite or sp(@xmath1 ) @xcite , was originally to employ the systematic @xmath2-expansion to handle strong correlation physics , especially in the cases with doping or frustrations . generally speaking ,
the large symmetries of su(@xmath1 ) and sp(@xmath1 ) enhance quantum spin fluctuations and suppress the antiferromagnetic ( af ) order @xcite .
various exotic quantum paramagnetic phases have been proposed based on the large-@xmath1 method , including various valence bond solid states and quantum spin liquid states @xcite . however , in conventional solid states , the su(@xmath1 ) symmetry is rare and thus the su(@xmath1 ) hubbard or heisenberg models are purely of academic interest . with the rapid development of the ultracold atom experiments , the realization of multi - component fermionic hubbard models with the su(@xmath0 ) or sp(@xmath0 )
symmetry has become a realistic goal ( the number of fermion components due to the hyperfine spin degree of freedom is naturally an even number ) .
it was proposed that the simplest sp(@xmath0 ) and su(@xmath0 ) hubbard models with @xmath3 can be realized in the hyperfine spin-@xmath4 alkali and alkaline - earth atoms @xcite . in these spin-@xmath4 hubbard models , an exact sp(4 )
spin symmetry exists without any fine - tuning of parameters , which is further enlarged to su(4 ) when the interactions do not rely on hyperfine spin .
the alkaline - earth atoms , _
e.g. _ , @xmath5yb and @xmath6sr , have a closed shell of valence electrons and thus their hyperfine spins are simply nuclear spins .
the interactions between the atoms with different hyperfine spins are insensitive to the nuclear spins , leading to the su(@xmath0 ) symmetry with @xmath0 being the number of fermion components @xcite .
recently , significant progress has been made in the experiment of ultracold alkaline - earth fermions with large hyperfine spins .
the @xmath5yb and @xmath6sr atoms have been cooled down to quantum degenerate temperatures @xcite , revealing the su(6 ) and su(10 ) symmetries respectively .
furthermore , an su(6 ) single - band fermionic hubbard model has also been realized with @xmath5yb atoms in a three - dimensional optical lattice@xcite . beyond single - band case ,
the spin - exchange interactions have recently been observed in the two - orbital su(@xmath7 ) and su(@xmath8 ) fermion system respectively@xcite .
also the number of spin components can be tuned experimentally @xcite .
theoretically , the novel symmetries of the multi - component hubbard model can give rise to the novel superfluidity @xcite and exotic quantum magnetism @xcite . in ultracold atom experiments , achieving low enough temperature regime below the spin superexchange scale has been considered a benchmark for simulating strongly correlated quantum systems . despite numerous efforts by experimentalists , achieving this temperature regime is still out of reach and remains one of the most challenging problems in this field .
so far , ultracold fermions in optical lattices have been cooled down to temperature regime below the hopping energy scale , @xmath9 .
one of the promising schemes for further cooling the system into spin superexchange scale @xmath10 is known as interaction - induced adiabatic cooling@xcite , a cooling scheme by adiabatically increasing interactions .
this cooling scheme utilizes the pomeranchuk effect which was originally proposed in @xmath11he systems .
however , for a two - component hubbard model in conventional lattices , the pomeranchuk effect is weak due to the antiferromagnetic correlations in the su(2 ) mott insulator .
it is still controversial whether the system can be cooled down to spin superexchange temperature by pomeranchuk cooling@xcite . as we show below , the multi - component su(@xmath0 ) hubbard model significantly facilitates the pomeranchuk cooling , cooling the system down to the temperature scale of @xmath12 from an initial temperature that is currently accessible in experiments .
this paper extends the previous work reported in ref .
. we have performed detailed determinant quantum monte carlo ( dqmc ) simulations of thermodynamic properties of the half - filled su(@xmath0 ) hubbard model with @xmath3 and 6 in the temperature regime @xmath13 .
we calculated the entropy - temperature relation and isoentropy curves , which show the enhancement of entropy with increasing interaction strength in the intermediate temperature regime , _
i.e. _ , the pomeranchuk effect .
the probability distributions of the onsite occupation number show the enhancement of particle localization as temperature increases in the low and intermediate temperature regimes .
the uniform spin susceptibilities and af structure factors are also calculated .
the rest part of this paper is organized as follows . in section [
sect : model ] , we introduce the definition of the su(@xmath0 ) hubbard model .
a discussion of the charge and spin energy scales of the half - filled su(@xmath0 ) hubbard model is followed in section [ sect : energyscale ] .
the parameters of the dqmc simulations are given in section [ sect : parameters ] . in section [
sect : thermo ] , we present the results of dqmc study on the thermodynamic properties of the half - filled su(4 ) and su(6 ) hubbard models .
in section [ sect : mag ] , the magnetic properties at finite temperatures are investigated .
conclusions are drawn in section [ sect : conclusion ] .
since naturally the number of fermion components due to the hyperfine spin degree of freedom is an even number , we only consider hubbard model with su(@xmath0 ) symmetry . at half - filling
, an su(@xmath0 ) hubbard model is defined by the lattice hamiltonian : @xmath14 where @xmath15 denotes nearest neighbors and the sum runs over sites of a two - dimensional square lattice ; @xmath16 represents spin indices running from 1 to @xmath0 ; @xmath17 is the particle number operator on site @xmath18 defined by @xmath19 ; @xmath20 and @xmath21 are the nearest neighbor hopping integral and the onsite interaction , respectively .
this definition of hubbard hamiltonian eq.([eq : sun ] ) offers several advantages . in the atomic limit ( @xmath22 ) ,
consider a half - filled lattice with @xmath1 particles per site , the energy cost of moving a particle from one site to its neighboring site is @xmath21 , which is independent of @xmath1 . due to
half - filling , the chemical potential @xmath23 vanishes in this grand canonical hamiltonian .
eq.([eq : sun ] ) also has particle - hole symmetry in bipartite lattices , which removes the sign problem in dqmc simulations for an arbitrary value of @xmath0 . in terms of the multiplets of su(@xmath0 ) fermions in the fundamental representation , the generators of the su(@xmath0 ) group can be written as @xmath24 where @xmath16 and @xmath25 run from 1 to @xmath0 .
the generators defined above are not independent of each other , since the diagonal operators satisfy the relation , @xmath26 .
nevertheless , the definition of operators , eq.([eq : su2n_gen ] ) , results in a simple commutation relation @xmath27=\delta_{\beta\gamma } s^{\alpha\delta } -\delta_{\alpha\delta } s^{\gamma\beta}.\ ] ] for convenience , we define the structure factor @xmath28 as s_su(2n)()=_i , j e^i r s_spin(i , j ) , [ eq : struc ] where @xmath29 is the relative vector between sites @xmath18 and @xmath30 . @xmath31 is the su(@xmath0 ) version of equal - time spin - spin correlation functions defined by s_spin(i , j)= _ , s_,i s_,j . [ ss ]
before going to the results of dqmc simulation , let us present a qualitative understanding of the physics of the half - filled hubbard model on a square lattice .
the charge channel and spin channel are characterized by two energy scales @xmath32 and @xmath33 , respectively , which will be discussed in both the weak and strong interaction regimes below .
we first consider the weak interaction limit ( @xmath34 ) . in this regime
, the underlying fermi surface plays an important role , which possesses the diamond shape and thus exhibits perfect nesting .
the key physics then arises from the fermi surface nesting : the spin susceptibility in the non - interacting limit divergences logarithmically , and thus an infinitesimal repulsive interaction generates af long - range - order . in this case , the gapped quasi - particle excitations carry both charge and spin quantum numbers .
the charge and spin energy scales are identical in this regime as @xcite _ s / t=_c / t= 4 ^ 2e^- . certainly , in the weak interaction regime , the system is a weak insulator .
although it is gapped , charge fluctuations can not be neglected . on the contrary , in the atomic limit ( @xmath22 , or , @xmath35 ) ,
the fermi surface completely disappears , and we need to use the local moment picture . at zero temperature ,
charge fluctuations are completely frozen . in the mott - insulating state
, we use the single - particle gap to denote the charge fluctuations , _
i.e. _ , the energy cost by adding or removing a particle from the mott - insulating state .
it is the energy difference between the two energy levels @xmath36 which equals @xmath37 in the atomic limit .
since the hopping process is completely suppressed , the af exchange energy @xmath38 , _
i.e. _ , the spin energy scale @xmath33 is zero in the atomic limit . , ( right ) @xmath39 .
( b ) sketches of a hole hopping in the su(2 ) ( up ) and su(4 ) ( down ) af backgrounds , respectively .
( fig [ fig : hopping ] ( b ) is from ref . [ ] . ) ] let us consider the strong interaction regime in which @xmath21 is large but not strong enough to completely suppress charge fluctuations , _
i.e. _ , @xmath39 .
let us consider the charge energy scale @xmath32 by adding an extra particle ( hole ) onto the background of the mott - insulating state .
as shown in fig .
[ fig : hopping ] ( a ) , the propagation of the particle ( hole ) expands the excitation energy level at @xmath40 in the atomic limit into an energy band .
the charge gap corresponds to the band bottom and thus is lowered to _
c=- , where @xmath41 is the band width and is determined by the hopping process of the extra particle ( hole ) . in fig .
[ fig : hopping ] ( b ) , we compare the hopping process of an extra hole in the half - filled su(2 ) and su(4 ) mott insulators . in the su(4 ) case , there are much more routes for the hole to hop from one site to its neighboring site than it does in the su(2 ) case . typically speaking , the number of hopping process for an extra particle or hole under the half - filled mott - insulating background scales as @xmath1 , thus we estimate @xmath42 .
the mobility of the extra hole is greatly enhanced in the su(@xmath0 ) mott - insulating state .
consequently , the charge energy scale @xmath32 is significantly lowered with increasing @xmath0 .
naively , we could estimate that @xmath32 vanishes at u_c nt .
[ eq : uc ] certainly , due to fermi surface nesting , @xmath32 does not vanish even in the weak interaction regime but becomes exponentially small .
nevertheless , @xmath43 sets up a scale of interaction strength to separate the regimes of the fermi surface nesting and the local moments .
the low energy physics in the strong interaction regime is described by the su(@xmath0 ) generalization of the heisenberg model : @xmath44 } s_{\alpha\beta,\vec{i}}s_{\beta\alpha,\vec{j}},\ ] ] where the su(@xmath0 ) spin operators are defined in eq.([eq : su2n_gen ] ) .
@xmath12 describes the strength of spin superexchange energy scale which can be viewed as the spin energy scale @xmath33 . the second order perturbation theory yields @xmath45 which decreases as @xmath21 increases . noting that @xmath46 in the weak interaction regime , which increases with @xmath21
, there should exist a peak in the intermediate interaction regime .
when the fermion system is deep in the mott - insulting state , in the low temperature regime @xmath47 , charge fluctuations are strongly suppressed , and the physics is dominated by spin superexchange process . therefore quantum spin fluctuations play an important role in determining the magnetic properties of the mott - insulating state at low temperatures . on the contrary , at high temperatures @xmath48 , quantum fluctuations give way to thermal fluctuations , which suppress quantum correlations , and thus interaction effects can be neglected . in the intermediate temperature regime @xmath49 , @xmath50 is high enough to suppress the af correlations , but not sufficient to defreeze charge fluctuations .
both quantum and thermal fluctuations are important in the intermediate temperature regime , and the interplay between them gives rise to interesting phenomena and universal properties @xcite .
the dqmc method is a widely used non - perturbative method for studying strongly correlated fermion systems @xcite .
provided that there is no sign problem , dqmc is known to be a well - controlled and unbiased method , which yields asymptotically exact results .
one of the most remarkable dqmc results is the af long - range order in the ground state of 2d su(2 ) half - filled fermi - hubbard model on a square lattice @xcite . in the subsequent sections
, we will use the dqmc method to simulate thermodynamic properties of the half - filled su(@xmath0 ) hubbard model in different regimes of temperature @xmath50 and interaction @xmath21 .
we will also show how those results are related to the two energy scales @xmath33 and @xmath32 analyzed in previous section . considering the error accumulation from matrix multiplications and simulation time , the lowest temperature in simulations
is set as @xmath51 ( with @xmath52 ) .
the suzuki - trotter decomposition is used in which the error is proportional to the cube of time discretization parameter @xmath53 .
@xmath54 is set from @xmath55 to @xmath56 in the temperature regime @xmath57 , and the convergence with respect to the scalings of @xmath54 has been checked . a hubbard - stratonovich ( hs ) transformation that decomposes the onsite hubbard interaction term in the density channel @xcite preserves the su(@xmath58 ) symmetry of the hubbard model , which can also be generalized to the su(@xmath0 ) hubbard model .
the discrete hs decomposition with an ising field only applies to the spin-@xmath59 case @xcite . for the cases of su(4 ) and su(6 ) , an exact discrete decomposition has been developed in ref .
[ ] , which is explained in appendix and employed in our simulations .
the simulated system is a @xmath60 square lattice with @xmath61 .
we focus on the parameter regimes of @xmath62 and @xmath63 . for a typical data point
, we use 10 qmc bins each of which includes 2000 warm - up steps and 8000 measurement steps .
we collect data once in each time slice . in our simulations , @xmath20 is set to unity and then the hubbard @xmath21 and temperature @xmath50 are given in the unit of @xmath20 .
in this section , we present the simulations of thermodynamic properties of the half - filled su(@xmath0 ) hubbard model with @xmath3 and @xmath7 on a square lattice , including the entropy - temperature relation , and the pomeranchuk effect . for different values of @xmath21 in ( a ) su(4 ) and ( b ) su(6 )
hubbard models .
the system size is @xmath60 with @xmath61.,title="fig : " ] for different values of @xmath21 in ( a ) su(4 ) and ( b ) su(6 ) hubbard models .
the system size is @xmath60 with @xmath61.,title="fig : " ] in cold atom experiments , entropy is actually a more physical quantity than temperature to characterize the system .
below we present the simulated entropy in the half - filled su(@xmath64 ) and su(@xmath7 ) hubbard models .
the parameters of simulation are chosen in the regimes of @xmath65 and @xmath63 , which are of interest in experiment .
the simulated entropy per particle ( not per site ) is defined by @xmath66 , where @xmath67 is the total entropy in the lattice .
the following formula is employed to calculate @xmath68 : = 4+t- ^_t dt^ , where @xmath69 is the entropy at infinite temperature ; @xmath70 denotes the internal energy per particle at temperature @xmath50 . in fig .
[ fig : stu ] , we show the entropy ( per particle ) of su(4 ) and su(6 ) fermions as a function of @xmath50 for various values of @xmath21 . in both cases ,
the curves cross at two _ typical _ temperatures @xmath71 ( low ) and @xmath72 ( high ) which divide the temperature into three different regimes .
let us first look at the su(4 ) case as follows .
_ the low temperature regime @xmath73 _ in this regime , the dependence of the entropy per particle @xmath68 on @xmath21 is non - monotonic , which can be understood by the competition between the spin energy scale @xmath33 and the charge energy scale @xmath32 as explained below . for weak interactions @xmath74 , @xmath75 is insensitive to @xmath21 .
as explained in section [ sect : weakatom ] , the physics in this regime is characterized by the fermi surface nesting @xmath32 and @xmath33 which are equal and are smaller than @xmath76 ( the lowest temperature reached in our simulations ) , and thus interaction effects are unimportant .
@xmath75 is then approximately the same as that in the non - interacting limit .
the non - zero residue entropy is due to the finite size effect which is caused by the degeneracy of single particle states right located on the fermi surface . as @xmath21 increases , @xmath77 increases faster than @xmath33 , while @xmath33 quickly reaches its maximum . in this interaction regime .
the relation @xmath78 holds , and thus increasing @xmath21 freezes charge fluctuations but enhances af correlations . even though there can not be true long - range af ordering at finite temperatures in 2d , the af correlation length scales as @xmath79 .
consequently , @xmath75 drops with increasing @xmath21 and the residue entropy , in principle , approaches zero . if @xmath21 continues to increase and reaches the strong interaction regime , say , @xmath80 , @xmath81 increases while @xmath82 decreases .
thus the relation @xmath83 holds .
@xmath50 is low enough to freeze charge fluctuations but high enough to disorder af correlations .
therefore , increasing @xmath21 ( or equivalently , decreasing @xmath33 ) enhances the entropy at a fixed @xmath50 in the low temperature regime .
_ intermediate temperature regime @xmath84 _ in this temperature regime , @xmath75 monotonically increases with @xmath21 at a fixed @xmath50 . at small @xmath21
the system is in the weak mott - insulating state with small values of @xmath33 and @xmath32 , which is very close to a fermi liquid state .
the system enters the solid - like strong mott state at large @xmath21 , where the physics is mostly local - moment - like .
the increase of entropy with @xmath21 means that the liquid - like state is more ordered than the solid - like state , known as the pomeranchuk effect . in this temperature regime
, @xmath50 is high enough compared to @xmath33 , and thus thermal fluctuations suppress magnetic correlations , while it remains smaller than @xmath32 , such that charge fluctuations are still frozen . in the strong mott - insulating state ,
fermions on each site are nearly independent of each other , and thus the entropy per site is proportional to the logarithmic of the spin degeneracy . in the weak mott - insulating state , we could think there is still a reminiscence of fermi surface , which strongly suppresses the entropy contribution .
therefore , the liquid - like state is more ordered than the solid - like state in the intermediate temperature regime . on the other hand ,
the pomeranchuk effect does not occur in the low temperature regime ( @xmath85 ) , where thermal fluctuations are not strong enough to suppress af correlations . in this case , the af correlations between adjacent sites lift spin degeneracy and lower the entropy in the strong mott - insulating state .
consequently , the pomeranchuk effect is prominent in the intermediate temperature regime .
_ high temperature regime @xmath86 _ in this case , not only spin fluctuations but also charge fluctuations are defrozen by thermal fluctuations @xmath87 .
charge fluctuations can not be completely suppressed by @xmath21 , and contribute most to the entropy .
therefore , increasing @xmath21 lowers the entropy at a fixed @xmath50 in the high temperature regime . as for the su(6 ) case , its behavior of entropy v.s . @xmath21 and @xmath50 is qualitatively identical to the su(4 ) case .
nevertheless , the intermediate temperature regime of the former is broader than the latter , which means that the pomeranchuk effect is more prominent . )
su(4 ) and ( @xmath88 ) su(6 ) hubbard models on a @xmath89 square lattice .
the entropy per particle @xmath68 is indicated on each curve .
( b ) was published in ref .
[ ] , which is reproduced here with a new curve of @xmath90 added .
, title="fig : " ] ) su(4 ) and ( @xmath88 ) su(6 ) hubbard models on a @xmath89 square lattice .
the entropy per particle @xmath68 is indicated on each curve .
( b ) was published in ref .
[ ] , which is reproduced here with a new curve of @xmath90 added . ,
title="fig : " ] as analyzed in section [ sect : entropy ] , in the strong mott - insulating regime the lattice system has larger entropy capacity than in the weak mott - insulating regime , which leads to the pomeranchuk effect .
this effect was first proposed in the @xmath11he system , where , in the low temperature region , increasing pressure adiabatically can further cool the system . in low temperature physics ,
this effect was employed as an effective cooling method named after pomeranchuk .
the similar situation occurs in the hubbard model .
nevertheless , at low temperatures where the af correlations are important , the spin degeneracy is lifted , which reduces the entropy in the mott - insulating state . in this case
, the pomeranchuk effect does not occur .
the dqmc simulations have been performed for the half - filled su(2 ) hubbard model in the literature .
both in 2d and 3d cases , the pomeranchuk effect is not obvious in the range of the entropy per particle @xmath91 between @xmath92 to @xmath93 , even if the interaction achieves @xmath94 @xcite . in the multi - component su(@xmath0 ) hubbard model , the situation is different . due to the increase of the number of fermion components , the pomeranchuk effect
is greatly facilitated @xcite .
the isoentropy curves v.s .
@xmath21 and @xmath50 are plotted in fig . [
fig : pc ] ( a ) and ( b ) for the su(4 ) and su(6 ) hubbard models , respectively . in both cases ,
the pomeranchuk effects are prominent in the intermediate temperature regime with intermediate interaction strengths , which emerge at @xmath95 for the su(4 ) case , and at @xmath96 for the su(6 ) case .
the enhancement of the pomeranchuk effect can be illustrated by comparing the su(2 ) and su(4 ) cases .
when deeply inside the mott - insulating state , in the intermediate temperature regime , the af correlations can be neglected , and the entropy per particle @xmath68 is dominated by the contribution of spin degeneracy . therefore , the entropy capacities can be estimated as @xmath97 , which is smaller than @xmath98 .
on the other hand , in the intermediate interaction regime , the charge gap in the su(4 ) case is significantly smaller than that in the su(2 ) case for the same interaction @xmath21 .
this means that fermions in the su(4 ) hubbard model can be more easily excited to the upper hubbard band than those in the su(2 ) case , which also enhances the entropy capacity of fermions in the su(4 ) case .
for the onsite particle numbers ( a ) @xmath99 , ( b ) @xmath100 , and ( c ) @xmath101 v.s .
@xmath50 and @xmath21 in the half - filled su(4 ) hubbard model . due to the particle - hole symmetry , @xmath102 , and @xmath103 , and @xmath104 .
the lattice size is @xmath60 with @xmath61 .
, title="fig : " ] for the onsite particle numbers ( a ) @xmath99 , ( b ) @xmath100 , and ( c ) @xmath101 v.s . @xmath50 and @xmath21 in the half - filled su(4 ) hubbard model . due to the particle - hole symmetry , @xmath102 , and @xmath103 , and @xmath104 .
the lattice size is @xmath60 with @xmath61 .
, title="fig : " ] for the onsite particle numbers ( a ) @xmath99 , ( b ) @xmath100 , and ( c ) @xmath101 v.s . @xmath50 and @xmath21 in the half - filled su(4 ) hubbard model . due to the particle - hole symmetry , @xmath102 , and @xmath103 , and @xmath104 .
the lattice size is @xmath60 with @xmath61 .
, title="fig : " ] to characterize charge fluctuations , we study the probability distribution of the onsite occupation number . in the su(2 ) case , at half filling the double - occupation number @xmath105 is related to the local moment @xmath106 , which exhibits a slightly non - monotonic behavior as a function of @xmath50 for fixed @xmath21@xcite .
also , in cold atom experiments , this quantity can be measured with _ in situ _ single - site resolution techniques @xcite .
let us consider the su(4 ) case
. at half - filling the most probable configuration of onsite particle number is @xmath107 . at finite @xmath21
, particles are allowed to hop between different sites , leading to charge fluctuations . due to particle - hole symmetry , the probabilities for the occupation numbers @xmath108 and @xmath109 are equal .
thus we only need to calculate @xmath110 with @xmath111 , and @xmath58 , respectively .
they are defined as @xmath112 obviously , they satisfy the relation @xmath113 .
the simulation results of @xmath110 ( @xmath111 and @xmath58 ) as a function of @xmath50 for various @xmath21 are presented in fig .
[ fig : momentum ] . in the weak and intermediate interaction regimes ,
charge fluctuations are significant .
the maximal probability of the exact half - filling @xmath101 only achieves around 70% when @xmath114 .
in contrast , the probabilities of one - particle fluctuation defined by @xmath115 fall in the range between @xmath116 and @xmath117 .
the probabilities of two - particle fluctuation , @xmath118 , are typically as low as a few percent .
each @xmath110 ( @xmath111 and @xmath58 ) exhibits non - monotonic behavior as @xmath50 increases . for example , at low temperatures @xmath119 and @xmath120 fall with the increase of @xmath50 ; then after reaching their minima at the temperature scale around @xmath20 , they grow again .
this indicates that in the temperature regime @xmath121 , on - site charge fluctuations are suppressed with increasing @xmath50 .
this counterintuitive phenomenon reminds us of the pomeranchuk effect , that the system tends to localize fermions in the intermediate temperature regime to maximize the entropy , which mainly comes from the spin degeneracy .
when the temperature further increases , charge fluctuations are activated and grow with @xmath50 .
in this section , we study the magnetic properties of the half - filled su(@xmath0 ) ( @xmath0=4 and 6 ) hubbard model on a square lattice at finite temperatures . including both the uniform spin susceptibilities and the af structure factor . for various @xmath21 in
the half - filled ( a ) su(4 ) and ( b ) su(6 ) hubbard models .
the system size is @xmath60 with @xmath61 .
error bars are smaller than the data points . , title="fig : " ] for various @xmath21 in the half - filled ( a ) su(4 ) and ( b ) su(6 ) hubbard models .
the system size is @xmath60 with @xmath61 .
error bars are smaller than the data points . , title="fig : " ] we now consider the uniform spin susceptibilities @xmath122 .
because the total spin is conserved , @xmath122 can be expressed as the equal - time correlations through the structure factor at @xmath123 : _
su(2n)(t)= s_su(2n)(q=0 ) . where the structure factor @xmath28 is defined by eq.([eq : struc ] ) .
in the 2d half - filled su(2 ) hubbard model @xcite , or , the heisenberg model @xcite , it is known that at high temperatures , @xmath124 behaves as the curie - weiss law which is proportional to @xmath125 . at low temperatures
, @xmath126 is suppressed due to the af correlations , and therefore it exhibits a peak at a low temperature scale @xmath127 .
@xmath127 can be used to roughly characterize the spin energy scale @xmath128 ( in heisenberg model @xmath129 ) .
the simulation results of the uniform spin susceptibilities for the su(4 ) and su(6 ) hubbard models are presented in figs .
[ fig : chi ] ( a ) and ( b ) , respectively . only the weak and intermediate interaction regimes with @xmath63
are considered here . in the su(4 ) case , when @xmath130 , the energy scale of @xmath33 is very small , and thus the peak location is beyond the temperature scope @xmath131 in our simulations .
compared to the su(2 ) case , @xmath33 in the su(4 ) case is significantly weakened .
for example , the peak in the su(2 ) hubbard model is located around @xmath132 when @xmath133 as simulated in ref .
nevertheless , the peak locations in the interaction regime @xmath134 have already become visible at temperatures @xmath135 .
furthermore , the magnitude of @xmath127 and the peak of @xmath136 increases with @xmath21 , which shows the enhancement of af correlations . in the su(6 ) case , the af correlations are further weakened compared to the su(4 ) case : among the curves in fig .
[ fig : chi ] ( b ) , only the one with @xmath137 exhibits a peak visible at temperatures @xmath135 .
this indicates the weakening of the spin energy scale @xmath33 in the intermediate interaction regime with increasing the number of fermion components .
with @xmath138 : ( a ) as a function of @xmath50 for fixed @xmath137 , and ( b ) as a function of @xmath21 for fixed @xmath139 for the half - filled su(4 ) and su(6 ) hubbard models .
the system size is @xmath60 with @xmath61 .
error bars are smaller than the data points.,title="fig : " ] with @xmath138 : ( a ) as a function of @xmath50 for fixed @xmath137 , and ( b ) as a function of @xmath21 for fixed @xmath139 for the half - filled su(4 ) and su(6 ) hubbard models .
the system size is @xmath60 with @xmath61 .
error bars are smaller than the data points.,title="fig : " ] we use the af structure factor @xmath140 defined with staggered wave vector @xmath138 to describe the af correlations . in fig .
[ fig : structure ] ( a ) , the curves of the af structure factor v.s .
@xmath50 are plotted at @xmath137 in both the su(4 ) and su(6 ) cases .
both @xmath141 and @xmath142 increase monotonically with the decrease of @xmath50 , which indicates the development of af correlations . at a fixed @xmath50 ,
@xmath141 is stronger than @xmath142 , which becomes even more prominent at low temperatures .
this is consistent with the picture that increasing the number of fermion components suppresses af correlations . in fig .
[ fig : structure ] ( b ) , the dependence of @xmath140 on @xmath21 at a fixed temperature , @xmath139 , is plotted in both su(4 ) and su(6 ) cases , where the @xmath21-dependencies exhibit a non - monotonic behavior . at small @xmath21 , @xmath141 and @xmath142
increase with @xmath21 . in this weak interacting regime , @xmath143 , and thus increasing @xmath21 enhances the spin energy scale but suppresses charge fluctuations , which facilitates to build up the af correlation .
again , the enhancement of the af correlation is more prominent in the su(4 ) case than in the su(6 ) case .
an interesting feature is that the rates of increase jump at @xmath144 in the su(4 ) case , and at @xmath145 in the su(6 ) case .
this may be due to a rapid increase of the charge energy scale @xmath32 , which indicates the crossover from the weak interaction regime to the intermediate interaction regime .
the af in the weak interaction regime is due to fermi surface nesting , and it evolves to the local moment physics as @xmath21 enters the intermediate interaction regime . at large @xmath21 , @xmath146 decreases with the increase of @xmath21 .
thermal fluctuations are described by the parameter @xmath147 , and thus increasing @xmath21 effectively enhances thermal fluctuations .
on the other hand , the zero temperature projector qmc results @xcite show that the af long - range orderings reach the maxima around @xmath148 and @xmath8 in the su(4 ) and su(6 ) cases , respectively .
the af ordering is then suppressed by further increasing @xmath21 , which is an effect of quantum fluctuations . in the su(6 ) case , the af long - range order is even completely suppressed around @xmath149 at zero temperature . this is because when deeply inside the mott - insulating regime , the number of superexchange processes between two adjacent sites increases rapidly with the number of fermion components , which indicates the enhancement of quantum fluctuations . combining both effects of quantum and thermal fluctuations , the af correlations are weakened with increasing @xmath21 in the strong interaction regime .
in summary , we have performed a systematic dqmc simulation study of thermodynamic properties of the half - filled su(@xmath0 ) hubbard model on a square lattice .
various thermodynamic behaviors including the entropy - temperature relation , the isoentropy curve and the probability distribution of the onsite occupation number have been simulated , which demonstrate the pomeranchuk effect is facilitated with increasing fermion components . based on the charge and spin energy scales ,
we have analyzed the thermodynamic properties in weak and strong interaction regimes . in the weak interaction regime
, the physics is characterized by the fermi surface nesting , while in the strong interaction regime the physics is mostly in the local moment picture . additionally , in our simulations the uniform spin susceptibilities and the af structure factors both exhibit qualitatively different behaviors in weak and strong interaction regimes .
theoretical analysis as well as dqmc simulations show that the interaction strength separating weak and strong interaction regimes increases with the number of fermion components .
z.z . , c.w . , and y.w .
acknowledges financial support from the national natural science foundation of china under grant no .
11328403 , and the fundamental research funds for the central universities .
is supported by the nsf dmr-1410375 and afosr fa9550 - 14 - 1 - 0168 .
z.c . acknowledges the support from sfb foqus ( fwf project no.f4006-n16 ) and the erc synergy grant uquam .
for the su(2 ) case , the hs transformation is usually performed by using the discrete ising fields @xcite
. however , the decomposition in spin channel can not be easily generalized to the su(@xmath0 ) case due to the increase of spin components .
instead , we choose an discrete hs decomposition in the density channel at the price of involving complex numbers as used in ref . [ ] .
the hs transformation for a half - filled su(@xmath0 ) hubbard model reads : @xmath150 where two sets of discrete hs fields @xmath151 and @xmath152 are employed . however , the hs decomposition with an error of order @xmath153 in ref .
[ ] is not exact . in ref .
[ ] , a new hs decomposition was proposed with a new set of parameters , which is exact for the su(@xmath64 ) and su(@xmath7 ) hubbard interactions .
the ising fields are defined as follows @xmath154 where @xmath155 , and @xmath156 . | we have systematically studied the thermodynamic properties of a two - dimensional half - filled su(@xmath0 ) hubbard model on a square lattice by using the determinant quantum monte carlo method .
the entropy - temperature relation , the isoentropy curve , and the probability distribution of the onsite occupation number are calculated in both su(4 ) and su(6 ) cases , which exhibit prominent features of the pomeranchuk effect .
we analyze these thermodynamic behaviors based on charge and spin energy scales . in the charge channel , the interaction strength that marks
the crossover from the weak to strong interaction regimes increases with the number of fermion components . in the spin channel ,
increasing the number of fermion components enhances quantum spin fluctuations , which is shown in the simulations of uniform spin susceptibilities and antiferromagnetic structure factors . |
for electrons emitted from a surface into a free space , the amount of current emitted at low - current regime is dominated by the emission mechanism , which is known as source - limited emission .
the mechanism can be divided into three types , namely thermionic emission , field emission , and photoemission , which is respectively , described by the richardson - laue - dushman ( rld ) law @xcite , the fowler - nordheim ( fn ) law @xcite , and the fowler - dubridge ( fd ) law @xcite .
a good overview can be found in a recent paper by jensen @xcite . in particular , all three emission mechanisms can be combined in a generalized model @xcite . at high - current regime , the amount of the emitted current will be influenced by the space charge effect , and it is known as the space - charge - limited ( scl ) current , which describes the maximum current density allowed for steady - state electrons emitted from the cathode and transported across the gap . for a one - dimensional ( 1d ) gap of spacing @xmath1 and a dc voltage of @xmath2 ,
the scl current is governed by the1d classical child - langmuir ( cl ) law @xcite , given by @xmath3 where @xmath4 , @xmath5 and @xmath6 is , respectively , vacuum permittivity , electron charge and electron mass .
scl electron flow occurs when the charge of the emitted electrons is sufficient to suppress the electric field at the cathode to zero , and the electrostatic potential distribution function @xmath7 is @xmath8 the transition from source - limited emission to scl current is important in the development of cathodes such as field emitters and photocathode , which is , respectively , based on field emission and photoemission .
the transition from the field emission to scl current has been developed for large @xcite and small @xcite gaps . for the transition from photoemission to scl current
, the cl law can not be used directly , because the pulse length of the electron photoemission is normally much smaller than the electron transit time across the gap . this short - pulse effect on the cl law
had been developed in a 1d classical model @xcite given by @xmath9 where @xmath10 is the normalized pulse length and @xmath11 is the transit time at the scl condition .
a more recent model has also been developed to include the quantum effects when the electron de broglie wavelength is comparable or smaller than the electron pulse width @xcite .
recently , significant efforts have been made in using ultrafast laser to induce electron photo - field emission from metallic tips @xcite . in most studies ,
space charge effect has been ignored , which may be important at the high - current regime operating at the high laser fields @xcite .
depending on the operating conditions , the emission process due to ultrafast laser excitation can be complicated , which leads to multiphoton emission @xcite , optical tunneling @xcite , strong field photoemission @xcite , and above threshold photoemission @xcite .
the onset of the optical tunneling from the multiphoton emission can be determined by a formula @xcite , which is at about 9.81 @xmath12 for an 8 femtosecond laser pulse . in this paper , we only focus on the photoemission process , and are interested to develop a simple 1d model to show the transition from the ultrafast - laser - induced multi - photon emission to the space charge limit at the ultrafast time scale . due to the long relaxation time scale ( ps ) ,
the multiphoton emission process is governed by the non - equilibirum heating model @xcite , which had shown that the emission is time - dependent at a time scale less than 1 ps for metals .
note this phenomenon was verified by an experiment in 2011 @xcite .
while is able to account for the effect of short pulse , yet it has assumed that the current injected into the gap is time - independent . here , from @xmath13 ( beginning of the laser pulse ) to @xmath14 ( end of the pulse ) , the injected current density from the cathode at @xmath15 is emitted as a function of time , given by @xmath16 . to solve this problem , we need to find the spatial variation of the injected electron density into the gap , @xmath17 for @xmath18 , where @xmath19 ( @xmath20 ) is the position of the beam front at at @xmath14 .
once this @xmath17 is obtained , we may use the similar approach in the short pulse model @xcite to calculate the space charge limited electron flow for the ultrafast - laser - induced photo - field emission .
it is important to note that the model presented here has ignored the sharpness of the tip .
thus the prediction can be considered only as a zero - order estimation .
while the 2d cl law has been developed @xcite , it can not be applied directly as the emitting area ( in a sharp tip ) is much smaller than the gap spacing .
the model is a quantitative one that is able to compare experiments in using ultrafast laser to excite multi - photon electron emission from a flat surface . in future
, it can be extended to a non - uniform and at least a 2d model to account for the sharpness of the tip in order to compare quantitatively with the experiments using sharp tips .
note the latter is not a trivial task , as there is no protrusive cl law which is valid for even steady - state electron emission from a sharp tip .
consider a gap of spacing @xmath1 with an external fixed electric field @xmath21 .
the cathode ( at @xmath22 = 0 ) is excited by an ultrafast laser to induce electron emission with a time - varying current density @xmath16 .
the emitted electrons will travel under the influence of the electric field ( including space charge field ) to reach a distance of @xmath19 ( @xmath23 ) within the laser pulse - length @xmath24 , which is smaller than the transit time @xmath25 .
note the laser profile is a step - like function with a temporal duration of @xmath26 . in the region of @xmath18 ,
the electrostatic potential is @xmath27 where @xmath28 is the space - charge electrostatic potential .
the value of @xmath19 can be calculated by solving the equation of motion for @xmath29 given by @xmath30\nonumber\\ & = & \frac em[f_{\rm dc}+\frac{\rm d}{{\rm d}x}\delta\phi(x)],\end{aligned}\ ] ] with initial conditions : @xmath31 and @xmath32 , where the primed symbol indicates the first derivative . here
, the initial velocity @xmath33 is kept to be reasonably small ( but not equal to zero ) to avoid the difficulty in the numerical integration of poisson equation ( see below ) at @xmath22 = 0 . in general
, we have @xmath34 in our calculation . near the cathode surface
, we consider a surface potential barrier , @xmath35 where @xmath36 is the fermi energy , @xmath37 its work function and the third term the classical image charge potential .
using , we obtain that the time - dependent tunneling current density @xmath16 , which is calculated from using formula @xmath38 here , @xmath39 is the electron tunneling probability at energy level @xmath40 through the surface potential barrier ( cf . ) , which is based on the modified wentzel - kramers - brillouin ( wkb ) method , @xmath41 is the time - varying non - equilibrium electron distribution function based on our previous model @xcite , @xmath40 is a dummy variable in the integration , and @xmath42 is the reduced plank constant .
the details on how to obtain @xmath39 and @xmath41 can be found in ref . [ 18 ] and [ 22 ] .
using , the emitted charge density ( per unit area ) within the laser pulse - width is calculated by @xmath43 to determine @xmath28 in the region @xmath18 , we solve the poisson equation , @xmath44 where @xmath45 is the velocity profile of the electron flow that can be obtained using the energy conservation , @xmath46.\ ] ] here , @xmath17 is the current density profile obtained at the end of the pulse at @xmath47 , which is related to , and it can be expressed as @xmath48,\ ] ] where @xmath49 is the inverse function of @xmath29 solved in . combining to
, we have @xmath50}{\sqrt{\frac{2e}{m}(f_{\rm dc}x+\delta\phi(x))}}.\end{aligned}\ ] ] the boundary conditions for solving are the zero space potential at the cathode ( @xmath22 = 0 ) and the continuous electric field at the beam front ( @xmath51 ) , which are respectively , @xmath52 and @xmath53 note is reduced from @xmath54 .
finally , the electrostatic potential in the vacuum region of @xmath55 ( in front of the electron beam ) is @xmath56 from the equations above , it is required to solve @xmath28 , @xmath19 and @xmath29 numerically to determine the scl current density @xmath16 or the scl charge density @xmath57 consistently . here
we construct a _ numerical _ algorithm to perform the calculation iteratively .
on the first step , we assume @xmath58 in to obtain the electron current density ( without space charge effects ) @xmath59 from , which gives the first estimated value for @xmath16 . by substituting @xmath60 into , we also obtain the first estimated time - profile @xmath29 and hence its inverse function @xmath49 . here the value of @xmath19 is determined by @xmath61 ( @xmath62 ) . with this @xmath59
, we can solve the poisson equation from for the new @xmath63 , which can be used to estimate the new current density @xmath64 .
here , @xmath64 is different from @xmath59 due to the finite value of @xmath63 as the space charge effect has been included .
the iterative process will continue until a convergence is reached which is determined by @xmath65 , where @xmath66 is the iterative step number . once the convergence is reached , we can determine the space charge potential @xmath28 , and compute the scl charge density @xmath57 according to , which can be compared to our previous work that excluded the space charge effects completely @xcite .
it is important to note that this approach has ignored completely the space charge field within the short time scale less than @xmath26 .
the approach is similar to the short pulse model with constant current density @xcite , which had been confirmed with particle - in - cell ( pic ) code .
our results will also be compared with pic simulation with a time - dependent emission current , which shows rather good agreement . before presenting the results
, we are interested to calculate the saturation of the scl current density at high fields due to the suppression of the @xmath67 electric field towards zero at the cathode .
thus the space charge potential @xmath28 must be large enough to suppress the applied dc field @xmath68 , given by @xmath69 when the condition above is fulfilled , we define a critical ( scl ) current density , @xmath70 where @xmath71 is an enhancement factor over the current density @xmath72 obtained from the time - dependent ultrafast laser emission model ( without space charge effects ) @xcite . here , @xmath71 can be expressed by @xmath73 where the @xmath74 is the space charge potential obtained by solving the poisson equation using @xmath75 .
here , we will not present the calculated @xmath41 in this paper , which is due to a step - function like ultrafast profile , and it has been shown elsewhere ( see figure 1 in ref . [ 18 ] ) .
the property of the @xmath41 for a gaussian - like ultrafast laser profile can also be found in a recent paper ( see figure 1 in ref . [ 22 ] ) . from these calculated @xmath41 ,
we solve ( 6 ) , ( 7 ) and ( 12 ) to obtain a self - consistent emission density including the space charge effect , which is linked to the emission process and also the non - equilibrium laser - metal interaction at the ultrafast time scale . a more recent model on the non - equilibrium laser - metal interaction
can also be found in ref .
[ 28 ] . in figure 1 , we present the emitting charge density @xmath57 as a function of the laser field @xmath76 with the following parameters : @xmath24 = 50 fs , @xmath77 , @xmath78 and @xmath79 . in this figure , we see that at the low laser field @xmath80 , both @xmath81 values calculated by the models with ( red squares ) and without space - charge effect ( black circles ) are nearly identical , which indicates space - charge effects are not important in this range of the low laser field @xmath76 . around @xmath82 ,
we observe a smooth transition into the scl regime ( space - charge limits indicated in open square ) ( also cf .
the inset of figure 1 which zooms in the shaded oven zone ) . in comparison with the time - independent model ( dashed purple line ) by valfells @xcite , our results ( time - dependent ) of sc limit are slightly higher . above the threshold laser field ,
increasing laser power will not further increase the emitted charge , and the limit is determined by the scl current . at this regime , the amount of charge density remains as a constant , which is simply controlled by @xmath83 , where @xmath84 is is the capacitance per area of the gap . to compare our results with pic simulation , we use a 2d pic code called vorpal @xcite . in the simulation , we inject a time - dependent electron current density based on our model @xcite , and we determine the scl current when reflection of electrons is detected @xcite .
we have a large emitting area so that the electron flow has a uniform space distribution in the transverse directions in order to be comparable with our 1d model .
the comparison shows that our model ( open square ) is slightly lower than the pic simulation results ( triangle ) as shown in the inset of figure 1 . because both our method and pic simulation give higher emission densities than valfells formula does , this demonstrates or at least inspires one to ponder whether it is possible to achieve a higher upper limit of space - charge limited emission density in time - dependent injection current case . in figure 2 ,
we plot the space charge potential @xmath28 in the region where electrons distribute @xmath85 for @xmath86 = 2 gv / m [ figure 2(a ) ] and 5 gv / m [ figure 2(b ) ] based on the same parameters in figure 1 .
it is clear that the space charge effect can be ignored for @xmath87 , where the magnitude of @xmath28 is much less than than the one based on @xmath86 = 5 gv / m by 2 orders of magnitude .
the corresponding emitting current density is plotted in panels 2(c ) and 2(d ) , respectively .
this figure confirms our observation above that the critical point for the space - charge effect to dominate is around @xmath88 5 @xmath89 under the physical situation investigated in figure 1 . in figure 3 , we study the dependence of our results by varying the work function @xmath37 , dc field @xmath68 and laser pulse length @xmath24 . here , the gap spacing is fixed at @xmath1 = 1 cm . in figure 3(a - c ) , we present the cases at @xmath24 = 10 fs for @xmath68 = 1 mv / m , 1 gv / m and 3 gv / m ( left to right ) . on each panel of figure 3 , we show three calculations at different work functions @xmath37 = 2.2 , 4.4 and 5.4 @xmath90 .
figure 3(d - f ) are presented similarly to figure 3(a - c ) , except at a longer laser pulse @xmath24 = 100 @xmath91 . by comparing the six panels of figure 3
, we make the following observations and discussions . from the results of figure 3
, we can see that higher work function ( like 5.4 ev ) will enter space charge limited ( scl ) regime at a higher laser field . however , the difference between 5.4 ev and 4.4 ev is not significant , both cases of which will be in the scl regime at @xmath92 5 gv / nm .
thus , we expect that experiments using gold with a higher work function ( 5.4 ev ) than tungsten ( 4.4 ev ) will have a less space charge effect at a fixed laser field . as the laser field increases from @xmath86 = 0.1 to 100 @xmath93 , the emission charge density @xmath57 increases within the small @xmath86 range and it gradually saturates to the scl regime at the high @xmath86 range as expected .
the critical value of @xmath86 that space charge effect becomes important ( thus also the threshold to reach space - charge limit ) is dependent on all three parameters @xmath94 , @xmath37 and @xmath26 .
in general , it is easy to reach the scl regime ( critical value of @xmath86 is small ) for small @xmath37 , and low @xmath94 . at small @xmath37 ,
it is easy to have a large emission current at a fixed field , so space charge limit may also be reached at low @xmath86 values , naturally consistent with our expectation because lower work function makes easier the tunnelling of electrons .
for example , the critical value to reach the the scl regime ( show in figure 3a ) is about @xmath86 1 to 2 gv / m for @xmath37 = 2.2 ev as compared to @xmath86 = 5 to 10 gv / m for @xmath37 = 4.4 to 5.4 ev . at small @xmath94 and fixed @xmath1 , scl current is small , so it is easy to reach the scl regime for a given emission current . as @xmath94 increases
, emission is greatly promoted because tunnelled probability increases shown in .
however , the scl current density also increases with @xmath68 .
for example , the critical value to reach the scl regime ( @xmath37 = 2.2 ev case ) is about @xmath76 = 1 gv / m ( at @xmath94 = 1 mv / m in figure 3(a ) ) as compared with @xmath76 = 2 gv / m ( at @xmath94 = 3
gv / m in figure 3(c ) ) .
thus the dependence on the dc field is not as sensitive as the work function .
the total amount of the scl charge density @xmath57 ( within the laser pulse duration ) is found to be nearly identical for various @xmath26 = 10 to 100 fs ( see the blue dashed lines labelled sc limit ) . while the scl charge current density
is expected to increase with small @xmath26 according to both short pulse @xcite and our model here , but the total amount of scl emitted charge integrated over the small laser pulse length will cancel the short pulse enhancement on current density .
this is understandable as the total amount of charge emitted under scl will depend on the surface electric field on the cathode @xcite . the critical value for the laser field to reach the space - charge limit will be lower at longer pulse duration of the electron flow . finally , it is important to note that the scl current density calculated here is based on a 1d model that does not account for the sharpness of the tip as mentioned in the introduction .
it is expected that the value to be enhanced by a factor of 20 to 100 for a very sharp tip ( see figure 8 in @xcite ) .
base on this , if we assume the scl current density is enhanced by a factor of 50 , then the threshold of the laser field to reach the scl current may be increased from 5 gv / m ( for a flat surface in figure 1 ) to about 10 gv / m ( for a sharp tip ) in order to reach the higher scl current density .
note this estimation is only a zero - order approximation , and accurate results will require a protrusive scl current model beyond the scope of this paper .
in summary , we have developed a 1d model to study the space - charge - limited ( scl ) emission under the ultrafast laser induced electron emission due to multiphoton absorption and non - equilibrium laser heating on a flat metal surface .
our model indicates that the space - charge effect may _ not _ be negligible for the high laser field and the threshold to reach the scl regime is determined for various parameters such as work function , dc applied field and laser pulse length .
smooth transition from the source - limited emission to scl emission is obtained .
the calculated results are compared with pic simulation and a short pulse scl current model @xcite .
it is found that the space charge effect is more dominant for materials with a lower work function like tungsten ( 4.4 ev ) as compared with gold ( 5.4 ev ) .
however for a flat surface , both materials will reach the space charge limited regime at sufficiently high laser field such as @xmath95 5 gv / m with a laser pulse length of tens to a hundred femtoseconds .
the ` vorpal ` simulation work herein was performed in axle and fuji servers of a - star computational resource centre(a*crc ) , singapore , supported by institute of high performance computing , agency for science , technology and research(ihpc , a*star ) , singapore .
some ` c++ ` code was run under high performance computing centre , nanyang technological university(hpc , ntu ) .
part of this project was benefitted from l y s 2011 visit to hong kong university of science and technology .
this project was funded by the singapore moe tier 2 grant ( moe2008-t2 - 01 - 033 ) and usa afoar aoard grant ( 11 - 4069 ) .
l y thanks wu lin , christine roark , koh wee shing , chen shih - hung and liu yao - li for enlightening guidance for this work , liang shijun for his assistance to run the relevant code , and also the considerate editorial work of editor m johnston .
l k a proposed this starting idea including the iterative numerical algorithm and supervised the whole work . l y developed the numerical algorithm , performed the calculation and ` vorpal ` simulation , plotted all the graphs from ` origin ` , and wrote the manuscript draft .
l k a revised the manuscript .
10 yanagisawa h , hengsberger m , leuenberger d , klckner m , hafner c , greber t and osterwalder j 2011 energy distribution curves of ultrafast laser - induced field emission and their implications for electron dynamics ( 8) 087601 liu y , wu l and ang l k june-17 2011 saturation of ultrafast laser excited electron emission from metallic nanotip at high current regime presented at _
east asian postgraduate workshop on nanoscience and technology _ , the hong kong university of science and technology ( hkust ) , * 109 - 110 * ; liu y and ang l k may-22 2013 space charge effect of time - dependent emission current excited from ultrafast laser presented at _ 14th ieee international vacuum electronics conference ( ivec ) _ , paris , poster session * i * | a one - dimensional ( 1d ) model has been constructed to study the transition of the time - dependent ultrafast laser photo - electron emission from a flat metallic surface to the space - charge - limited ( scl ) current , including the effect of non - equilibrium laser heating on metals at the ultrafast time scale . at a high laser field , it is found that the space charge effect can not be ignored and the scl current emission is reached at a lower value predicted by a short - pulse scl current model that assumed a time - independent emission process .
the threshold of the laser field to reach the scl regime is determined over a wide range of operating parameters .
the calculated results agree well with particle - in - cell ( pic ) simulation .
it is found that the space charge effect is more important for materials with lower work function like tungsten ( 4.4 ev ) as compared to gold ( 5.4 ev ) .
however for a flat surface , both materials will reach the scl regime at the sufficiently high laser field such as @xmath0 with a laser pulse length of tens to a hundred femtoseconds . |
the presence of magnetic fields on galactic and extragalactic scales is a major unsolved puzzle in modern astrophysics .
although the observational evidence supporting the existence of magnetic fields in large - scale structure is overwhelming , there is no consensus as to their origins .
the standard paradigm for the creation of these fields is the dynamo mechanism , in which a small initial seed field is amplified by turbulence ( the @xmath1 effect ) and/or by differential rotation ( the @xmath2 effect ) . in principle , once a seed field is in place , it should be possible to follow its evolution and amplification from the collapse of structure and the operation of any relevant dynamos . in recent papers , it has been argued that magnetic fields are generated in the early universe due to induced vorticity @xcite . in this paper , we put forward a new mechanism for the generation of seed fields from cosmological perturbations .
we show that the evolution of cosmological perturbations in the pre - recombination era produces charge separations and currents on all scales , both of which contribute to magnetic fields , in addition to and independent of any fields generated via vorticity .
these seed fields persist until the onset of significant gravitational collapse , at which point field amplification and dynamo processes may magnify such seeds to the @xmath3 fields observed today .
the generation of magnetic fields from charge separations and currents in the early universe is a necessary consequence of structure formation .
this paper calculates the magnitude of these seed fields , and it is shown that these seed fields may be sufficiently strong to account for all of the observed magnetic fields in large - scale structures .
the layout of the paper is as follows : section 2 gives an overview of the observational evidence for magnetic fields along with a brief theoretical picture of their generation .
section 2 also contains an explanation of the novel idea that the early stages of structure formation in a perturbed universe generate magnetic fields .
section 3 presents a detailed treatment of cosmological perturbations , with a specific view towards the creation and evolution of local charge separations and currents .
section 4 explicitly calculates the seed magnetic fields that arise via this mechanism as a function of scale and epoch .
finally , section 5 compares the results of this mechanism with competing theories .
it also discusses avenues for future investigation of this topic , including possible observational signatures of the fields that would arise from this mechanism .
magnetic fields with strength @xmath4 are seen in all gravitationally bound or collapsing structures in which the appropriate observations have been made @xcite .
the four major methods used to study astrophysical magnetic fields are synchrotron radiation , faraday rotation , zeeman splitting , and polarization of starlight .
these observational techniques are detailed in depth in @xcite , with faraday rotation often proving the most fruitful of the above methods .
magnetic fields have been found in many different types of galaxies , in rich clusters , and in galaxies at high redshifts .
spiral galaxies , including our own , appear to have relatively large magnetic fields of @xmath5 on the scale of the galaxy @xcite , with some ( such as m82 ) containing fields up to @xmath6 @xcite .
elliptical and irregular galaxies possess strong evidence for magnetic fields ( of order @xmath7 ) as well @xcite , although they are much more difficult to observe due to the paucity of free electrons in these classes of galaxies .
coherence scales for magnetic fields in these galaxies , as opposed to spirals , are much smaller than the scale of the galaxy , of order @xmath8 to @xmath9 pc .
furthermore , galaxies at moderate @xmath10 and high redshifts @xmath11 have been observed to require significant @xmath12 magnetic fields to explain their observed faraday rotations @xcite .
magnetic fields are also observed in structures larger than individual galaxies .
the three main types of galaxy clusters are those with cooling flows , those with radio - halos , and those devoid of both .
galaxy clusters with cooling flows are observed to have fields of @xmath13 to @xmath14 @xcite , the coma cluster ( a prime example of a radio - halo cluster ) is observed to have a field strength @xmath15 @xcite , while clusters selected to have neither cooling flows nor radio halos still exhibit indications of strong @xmath16@xmath17 fields @xcite . there
even exists evidence for magnetic fields on extracluster scales .
an excess of faraday rotation is observed for galaxies lying along the filament between the coma cluster and the cluster abell 1367 , consistent with an intercluster magnetic field of @xmath18@xmath19 @xcite . on the largest cosmological scales ,
there exist only upper limits on magnetic fields , arising from observations of the cosmic microwave background @xcite and from nucleosynthesis @xcite , setting limits that on scales @xmath20 , field strengths are @xmath21 .
observational evidence for magnetic fields is found in galaxies of all types and in galaxy clusters , both locally and at high redshifts , wherever the appropriate observations can be made .
a review of observational results can be found in @xcite . the theoretical picture of the creation of these fields ,
however , is incomplete .
fields of strength @xmath7 can be explained by the magnification of an initial , small seed field on galactic ( or larger ) scales by the dynamo mechanism @xcite .
a protogalaxy ( or protocluster ) containing a magnetic field can have its field strength increased by many orders of magnitude through gravitational collapse @xcite , and can also be further amplified via various dynamos .
dynamos which can amplify a small seed field into the large fields observed today involve helical turbulence @xmath22 and/or differential rotation @xmath23 .
various types of these dynamos include the mean - field dynamo @xcite , the fluctuation dynamo @xcite , and merger - driven dynamos @xcite , among others .
however , the dynamo mechanism does not explain the origin of such seed fields .
while the initial seeds that grow into magnetic fields are anticipated to be small , they must still come from somewhere @xcite , as their existence is not explained by the dynamo mechanism alone .
there are many mechanisms that can produce small - strength magnetic fields on astrophysically interesting scales , either through astrophysical or exotic processes ( see * ? ? ?
* for a detailed review ) .
exotic processes generally rely on new physics in the early universe , such as a first - order qcd phase transition @xcite , a first order electroweak phase transition @xcite , broken conformal invariance during inflation @xcite , specific inflaton potentials @xcite , or the presence of charged scalars during inflation @xcite .
astrophysical mechanisms , in contrast , are generally better grounded in known physics , although they have difficulty generating sufficiently strong fields on sufficiently large scales .
the difference in mobility between electrons and ions admits , under appropriate circumstances , the creation of large - scale currents and magnetic fields .
seed magnetic fields arising from this difference in mobility between electrons and ions may be astrophysically generated via a variety of phenomena , including radiation - era vorticity @xcite , vorticity due to gas - dynamics in ionized plasma @xcite , stars @xcite , or from active galactic nuclei @xcite .
although there are many candidates for producing the seed magnetic fields required by the dynamo mechanism , none has emerged as a definitive solution to the puzzle of explaining their origins .
the mechanism proposed in this paper is that seed magnetic fields are generated by the scattering of photons with charged particles during the radiation era . unlike the mechanism of @xcite , which is disfavored @xcite due to its requirement of substantial primordial vorticity (
although see * ? ? ?
* ; * ? ? ?
* for an argument that some vorticity is necessary ) , the fields of interest here are generated by the earliest stages of structure formation , _ requiring no new physics_. ions ( taken to be protons , for simplicity ) and electrons are treated as separate fluids , with opposite charges but significantly different masses .
the mass - weighted sums of their density and velocity fields will determine the evolution of the net baryonic component of the universe , and should agree with previous treatments ( e.g. * ? ? ? * ) .
the differences of their density and velocity fields , however , provides a description of a local charge separation and a local current density , both of which contribute to magnetic fields . since cosmological perturbations , which serve as seeds for structure formation , exist on all scales
, it is expected that seed magnetic fields will be generated on all scales by this mechanism . as the magnetic fields generated via this mechanism
are in place at very early times , they possess the advantage over other astrophysical mechanisms that these fields will have optimally large amounts of time to be affected by amplification effects and various dynamos .
the remainder of this paper focuses on calculating the magnitude of the magnetic fields generated by this process and discussing their cosmological ramifications .
although the early universe is isotropic and homogeneous to a few parts in @xmath24 @xcite , it is these small density inhomogeneities , predicted by inflation to occur on all scales @xcite , that lead to all of the structure observed in the universe today . as it is the early epoch of structure formation that is of interest for the creation of magnetic fields ,
we calculate the evolution of inhomogeneities in the linear regime of structure formation .
the most sophisticated treatment of cosmological perturbations in the linear regime to date is that of @xcite , which provides evolution equations for an inhomogeneous universe containing a cosmological constant , dark matter , baryons , photons , and neutrinos .
this section extends their treatment to encompass separate proton and electron components .
the mass - weighted sum of protons and electrons will recover the usual baryon component , whereas the difference of the density fields represents a charge separation , and the difference of the velocities is a net current .
the dynamics of any cosmological fluid can be obtained , in general , from the linear einstein equations ( see * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* for earlier treatments ) .
although the choice of gauge does not impact the results , the conformal newtonian gauge leads to the most straightforward calculations .
when tensor modes are unimportant , the metric is given by @xmath25\mathrm{,}\ ] ] where @xmath26 when gravitational fields are weak .
the linearized einstein equations are then @xmath27 where @xmath28 is the shear stress , which is negligible for non - relativistic matter but important for photons and neutrinos .
a cosmological fluid that is either uncoupled to the other fluids or mass - averaged among uncoupled fluids in the early universe obeys @xmath29 where @xmath30 is defined as the local density relative to its spatial average @xmath31 ; @xmath32 , where @xmath33 is the local peculiar velocity ; and @xmath34 is the sound speed of the fluid . for individual components with additional inter - component interactions , equation ( [ evolution ] )
must be augmented to include these interactions .
examples of such interactions include momentum transfer between photons and charged particles and coulomb interactions between protons and electrons . for protons , electrons , and cold dark matter ( cdm ) ,
an equation of state @xmath35 is assumed , and for radiation and neutrinos , @xmath36 .
the master equations for each component of interest are computed explicitly in subsections 3.13.5 .
the evolution equations which appear in this section for baryons , photons , light neutrinos , and cold dark matter are derived in much greater detail , including photon polarization , hydrogen and helium recombination , in both synchronous and conformal newtonian gauges , in @xcite .
the cold dark matter component ( denoted by the subscript @xmath37 ) is collisionless and pressureless ; its evolution can be simply read off from equation ( [ evolution ] ) with @xmath38 to be @xmath39 any cold ( i.e. nonrelativistic ) , collisionless component will behave according to the dynamics given by equation ( [ cdm ] ) . for massless ( or nearly massless ) particles ,
pressure is non - negligible , and the appropriate equation of state has @xmath40 .
additionally , the shear term ( @xmath28 ) may be important as well .
the only accurate way to compute the evolution of such a component of the universe is by integration of the boltzmann equation , which is given for light neutrinos ( denoted by subscript @xmath41 ) by @xmath42 \mathrm{,}\ ] ] in fourier space .
the approximation that neutrinos are massless and uncoupled is very good from an age of the universe of approximately @xmath43 s until the epoch of recombination .
the evolution equations for light neutrinos are then @xmath44 \mathrm{,}\end{aligned}\ ] ] where @xmath45 is related to @xmath46 by @xmath47 , and the index @xmath48 governs the final equation for @xmath49 .
@xmath50 is defined by the expansion of the perturbations in the distribution function , @xmath46 , @xmath51 where @xmath52 are the legendre polynomials .
equations ( [ boltz][fnu ] ) are valid for any non - collisional species behaving as radiation .
photons ( denoted by subscript @xmath53 ) , although similar to light neutrinos , evolve differently due to their large coupling to charged particles .
the differential scattering cross - section of photons with electrons , is given by @xmath54 where @xmath55 is the thomson cross - section @xcite .
photons also scatter with protons , but with a cross - section suppressed by a factor of the squared ratio of electron to proton mass , @xmath56 .
the polarization - summed phase - space distribution for photons , @xmath57 , is the same as the distribution function for neutrinos ( see equation [ fnu ] ) , with a non - zero difference between the two linear polarization components denoted by @xmath58 .
the linearized collision operators for thomson scattering @xcite yield the set of master equations for photons , @xmath59 - a n_e \sigma_t
\mathcal{f}_{\gamma l } \mathrm { , } \nonumber\\ \dot \mathcal{g}_{\gamma m } & = & \frac{k}{2m+1 } [ m \mathcal{g}_{\gamma ( m-1 ) } - ( m+1 ) \mathcal{g}_{\gamma ( m+1 ) } ] \nonumber\\ & & + a n_e \sigma_t \left [ \frac{1}{10 } \mathcal{f}_{\gamma m } - \frac25 \mathcal{g}_{\gamma m } \right ] \mathrm{,}\end{aligned}\ ] ] where @xmath60 , @xmath61 , @xmath62 , the indices @xmath48 and @xmath63 are valid for @xmath64 and @xmath65 , and the subscript @xmath66 denotes the baryonic component , which is the mass - weighted sum of the electrons and protons .
electron - photon scattering is so dominant over proton - photon scattering as to render the latter negligible , but the electron - proton coupling ( via electromagnetism ) is sufficiently strong that , to leading order , those two fluids move in kinetic equilibrium .
the net behavior of the baryonic component can be derived from combining the mass - weighted contributions of the proton and electron fluids .
both protons and electrons contain all of the terms present in the cdm equations ( cf .
section 3.1 ) , but additionally contain important terms arising from the coulomb interaction and from thomson scattering .
the coupling of the coulomb interaction to density inhomogeneities can be calculated through a combination of the electromagnetic poisson equation , @xmath67 where @xmath68 is the electric charge density , and the euler equation , @xmath69 with @xmath70 as the charge - to - mass ratio of the particle in question and @xmath71 the collision operator .
the coulomb contribution appears as @xmath72 in the evolution equation for @xmath73 , where @xmath74 denotes a species of particle with a mass @xmath75 and charge @xmath76 .
the evolution equations are therefore @xmath77 for electrons ( denoted by subscript @xmath78 ) , and @xmath79 for protons , denoted by subscript @xmath80 , where @xmath81 characterizes the rate of momentum transfer due to photon scattering with charged particles .
the damping coefficient for electrons @xmath82 is given by @xmath83 and the analogous quantity for protons is smaller by a factor @xmath84 .
note the difference in the sign of the final terms in the equations for @xmath85 and @xmath86 , which will prove important below . from equations ( [ electrons ] ) and
( [ protons ] ) for electrons and protons the dominant gravitational and electromagnetic combinations can be constructed separately .
the remainder of this subsection details the evolution of baryons in the linear regime of a perturbed universe .
baryonic matter can be treated as the combination of electrons and protons , thus the mass weighted sum of proton and electron overdensities gives rise to the baryonic perturbations , @xmath87 where @xmath88 . the evolution of baryonic matter follows from a mass - weighted combination of the equations for electrons ( [ electrons ] ) and protons ( [ protons ] ) .
to the extent that electrons and protons move together ( the tight - coupling approximation ) and the difference @xmath89 is negligibly small ( for baryonic evolution this is an excellent approximation ; see section 3.5 ) , the evolution equations for baryonic matter are @xmath90 where @xmath91 . in the tight coupling approximation , the baryon - photon coupling term in equation ( [ baryons ] )
is driven by the electron - photon interaction , as has been shown by @xcite and subsequent authors .
the equations in ( [ baryons ] ) are identical to the equations for the evolution of baryonic inhomogeneities derived in @xcite . from equations ( [ electrons ] ) and
( [ protons ] ) , a charge _ difference _ component as well as a sum component can be obtained .
as the limits on a net electric charge asymmetry in the universe are very strict @xcite , we expect that any differences in densities and/or velocities of protons and electrons will not be strong enough to significantly alter the evolution of the other components of the universe .
it is therefore expected that the analysis of baryonic matter , detailed in section 3.4 , will be unaffected by differences in proton and electron densities and velocities .
however , even a small separation may still have important cosmological implications .
the charge difference component ( denoted by subscript @xmath92 ) is the difference between the proton and electron components , such that @xmath93 and @xmath94 .
the gravitational potential does not affect the evolution of these quantities ; gravity acts equivalently on electrons and protons .
however , if @xmath95 and @xmath96 are nonzero , resulting electromagnetic fields will act differently on oppositely charged particles .
the master equations for the charge - asymmetric component , obtained from equations ( [ electrons ] ) and ( [ protons ] ) are @xmath97 where the approximations @xmath98 and @xmath99 have been utilized where applicable .
the term @xmath100 in equation ( [ a ] ) arises from the coulomb force acting on charged particles , while the final term , @xmath101 , arises from the difference in thomson scattering between protons and electrons .
this final term provides a source for the generation of a charge separation independent of and in addition to any initial charge asymmetry , and will create a local charge asymmetry _ even when there is none initially_. in the evaluation of equation ( [ a ] ) , the electromagnetic terms dominate the cosmological terms , such that an excellent approximation in the pre - recombination universe is @xmath102 for some purposes , it is useful to express the set of equations found in equation ( [ a ] ) as a single ordinary differential equation .
this can be accomplished by setting @xmath103 , and again by neglecting the unimportant cosmological terms @xmath104 and @xmath105 .
many of the coefficients in equation ( [ a ] ) are functions of @xmath106 , but the derivatives in equation ( [ a ] ) are with respect to conformal time , @xmath107 .
a change of variables can be performed , using the relation that @xmath108^{1/2 } % \mathrm{seconds } \nonumber\\ & = & n a^2 t_0 \mathrm { , } \qquad n \simeq 72.2 \mathrm { , } \label{ttoa}\end{aligned}\ ] ] in the radiation era , where @xmath109 is the age of the universe today and @xmath110 , to express all derivatives as derivatives with respect to @xmath106 , denoted by primes ( instead of dots )
. the evolution of @xmath111 can be tracked by evolving equation ( [ solve ] ) below , @xmath112 where the subscript @xmath113 denotes the present value of a quantity .
this is simply the equation of a damped harmonic oscillator , with coefficients that change slowly with time compared to damping or oscillation times .
the behavior can be characterized as overdamped at the earliest times , critically damped when @xmath114 , and underdamped at late times .
of all the terms in equation ( [ solve ] ) , only @xmath115 , @xmath116 , and @xmath117 ( and derivatives ) are functions of @xmath106 ; all other quantities are constant coefficients .
although there does not exist a simple analytic form for @xmath118 in general , at sufficiently early times there exists the simple approximation @xmath119 valid when the following condition is met : @xmath120 equation ( [ thetadiff ] ) is an approximation for a flat @xmath121cdm cosmology with cosmological parameters @xmath122 , @xmath123 , @xmath124 , and a helium-4 mass fraction of @xmath125 .
these parameters are used in all subsequent analyses for the calculation of cosmological quantities .
the approximation in equation ( [ thetadiff ] ) breaks down at later times .
when this occurs , numerical methods must be used to obtain the quantity @xmath118 .
the software package @xcite is ideal for performing this computation , as it performs the numerical evolution of equations ( [ cdm],[nus],[photons ] , and [ baryons ] ) concurrently .
computational results for the quantities @xmath126 and @xmath116 are given by , which are valid at all times in the linear regime of structure formation .
it is found that when the approximation in equation ( [ thetadiff ] ) breaks down , the quantity @xmath118 grows more slowly initially , and proceeds to oscillate at a roughly constant amplitude at later times .
these oscillations in the quantity @xmath118 are closely related to the acoustic oscillations between baryons and photons observed in the cosmic microwave background @xcite . numerical integration of equation ( [ solve ] ) can be accomplished in various ways ( see * ? ? ? * for examples ) . at sufficiently late times ( when @xmath127 ) , numerical results indicate that the quasi - equilibrium solution @xmath128 obtained by neglecting the first two terms in equation ( [ solve ] ) , is an excellent approximation . with @xmath129 and @xmath130 given by in units of @xmath131 , the prefactor in equation ( [ dsoln ] ) can be written as @xmath132 where @xmath133 is the classical electron radius .
the quantity @xmath134 then follows directly from equation ( [ a ] ) to be @xmath135 the solutions in equations ( [ dsoln ] ) and ( [ tsoln ] ) remain valid up to the epoch of recombination @xmath136 . the results of numerically integrating the equations for @xmath117 and @xmath134 on various length scales through recombination using the code @xcite are presented as a function of redshift @xmath137 in figure [ dtfig ] for @xmath138 , 1 , 0.1 , and @xmath139 .
the amplitude @xmath140 rises as @xmath141 initially , then ceases to grow when the scale of interest enters the horizon , and oscillates at an amplitude which first continues to rise slowly , then falls , eventually matching on to the equilibrium solution @xmath142 .
the magnitude of the charge separation is small ( of order @xmath143 in figure [ dtfig ] ) because the constant in equation ( [ constant ] ) is small compared to the hubble length at decoupling . in the conformal newtonian gauge
the mode amplitudes are defined to be of order 1 outside the horizon , and need to be multiplied by the cobe normalization @xmath144 of @xcite to obtain the physical quantities of magnetic field strength and magnetic spectral density , as in section 4 .
figure [ figp_q ] shows the spectral density of the charge asymmetry , @xmath145 , plotted as a function of @xmath146 . for comparison
, the spectral density of baryonic matter , @xmath147 , is also shown .
note the identical scales , @xmath146 , at which @xmath148 and @xmath149 change signs .
the results thus far are accurate up through the epoch of recombination . at this point , however , the universe transitions from a state with ionization fraction @xmath150 to a state where the ionization fraction is small , @xmath151 @xcite , and the universe is transparent to radiation . the calculations of @xcite suggest that silk damping affects only scales @xmath152 . in the absence of interactions with photons
, a charge separation would continue to evolve as @xmath153 where @xmath154 is the ratio of the electric to gravitational forces , @xmath155 a residual charge separation without additional interactions would oscillate ( plasma oscillations ) with proper frequency @xmath156 , with a further slow adiabatic decay of the oscillation amplitude
. however , there are many other effects that begin to become important after recombination , including gravitational collapse , dynamo effects , and continued electron - photon scattering , and we do not attempt to compute the subsequent evolution in detail in this paper .
with the results derived in section 3.5 for @xmath157 and @xmath140 , values for the local current densities and local charge separations can be obtained at any time in the pre - recombination universe on all scales .
both @xmath140 and @xmath157 will contribute to magnetic fields , as currents create magnetic fields directly , and the bulk motion of a region of net charge will also produce a magnetic field . for each comoving distance scale
( given by the value of @xmath146 ) and each epoch ( determined by the scale factor @xmath106 ) of the universe , there will be a magnetic field amplitude associated with that scale .
this field may serve as the seed for the large - scale magnetic fields observed today .
an expression for magnetic fields can be derived from the currents arising from the relative motion of the protons and electrons in the universe .
magnetic fields can be derived from maxwell s equations @xmath158 with the current density @xmath159 given by @xmath160 \mathrm{,}\ ] ] where @xmath161 , and the displacement current is neglected . in the fourier domain , the curl of maxwell s equation for @xmath162 gives a direct expression for the magnetic field as a convolution @xmath163 \mathrm { , } \end{aligned}\ ] ] from which the second moment of the magnetic field @xmath164 is @xmath165 where @xmath166 is the dirac delta function and @xmath167 is the magnetic field power spectrum .
note that the direction parallel to @xmath168 does not contribute to magnetic fields .
therefore , the direction perpendicular to @xmath168 is projected out in equation ( [ bmoment ] ) . the power spectrum , @xmath167 ,
is then given by @xmath169 \mathrm { , } \end{aligned}\ ] ] where @xmath170 is the angle between the vectors @xmath168 and @xmath171 .
the second moment of two general quantities , @xmath172 and @xmath173 , is expressed as the power @xmath174 , defined by @xmath175 where @xmath176 .
all terms in equation ( [ p_b ] ) make comparable contributions to the total .
note that although the velocities @xmath177 and @xmath178 ( or @xmath179 and @xmath180 ) are derived from potentials , they appear in convolution with density inhomogeneities .
for this reason , there is no need for explicitly second - order quantities ( such as vorticity or anisotropic stress ) to obtain a nonzero magnetic field , an effect similar to the calculation of the kinetic sunyaev - zeldovich effect by @xcite .
the spectral density , @xmath181 , obtained by numerically integrating @xmath134 and @xmath117 in equations ( [ dsoln ] ) and ( [ tsoln ] ) , provides both a measure of the magnetic field strength on a given scale @xmath146 and a measure of the energy stored in magnetic fields , @xmath182 .
the results for the spectral density of the magnetic field on comoving scales from @xmath183 to @xmath184 at the epoch of recombination are shown in figure [ bgraph ] .
the peak of the density corresponds to a typical magnetic field strength of @xmath185@xmath186 on a comoving scale of @xmath187 .
the major result of this paper has been to demonstrate that seed magnetic fields of cosmologically interesting strengths and scales are necessarily generated by the same processes that cause structure formation .
as overdensities in the early universe slowly grow during the radiation era , the differing photon interactions with protons and electrons create charge separations and currents of small magnitudes on all scales .
these charge separations and currents grow in magnitude as the universe evolves , causing magnetic fields to grow as well . the magnetic power in a given mode peaks at approximately the time of horizon crossing , oscillating and falling slowly in amplitude after that .
the net result is that , at the epoch of recombination ( and hence prior to any field amplification due to gravitational collapse or dynamo effects ) , seed magnetic fields of magnitude @xmath0 are created by the simple dynamics of charged particles .
other recent discussions of the origins of large - scale magnetic fields have mentioned the possibility that the evolution of cosmological perturbations could be responsible for their generation @xcite .
these groups have focused on the effects of vorticity , which arises at second order in cosmological perturbation theory .
in addition to the charge separations and currents induced by photon scattering and the coulomb interaction , as discussed in this work , @xcite also considered the contribution of photon anisotropic stress to magnetic fields , but found that it is uninteresting on large scales .
the overall strength and scale of the magnetic fields reported by these authors seem to differ by large factors from one another .
the fields derived in @xcite and @xcite appear to be unreasonably large , and peak at scales where silk damping ought to be a dominant effect @xcite .
as shown in the subsequent paragraph , their value for a magnetic field strength of @xmath188 on @xmath189 scales is difficult to reconcile with other estimates .
it also appears that the fields derived in @xcite are insignificantly small , as their analysis of vorticity may not include the dominant terms in the creation of magnetic fields . for reasons discussed in the next paragraph
, it appears that the estimates for magnetic fields generated by vorticity are most accurately given by @xcite .
when their results are considered on @xmath190 scales , the fields generated are found to be of physically reasonable strengths similar to our own , of order @xmath191 .
their spectrum is similar to the spectrum of figure [ bgraph ] as well , although their spectrum peaks at a slightly higher amplitude , is narrower in shape , and reaches its maximum at a smaller scale .
the approximate magnitude of the field strengths arising from the full calculations of section 4 can be understood intuitively . from the quasi - equilibrium solution in equation ( [ dsoln ] ) ,
the fractional charge asymmetry @xmath192 is of order @xmath143 at the scale of the peak of the power spectrum .
when this information is combined with the fact that the streaming velocity at decoupling is @xmath193 , a current that yields magnetic field strengths of @xmath194 is produced .
the field strengths expected from vorticity are somewhat less intuitive .
although for an ideal fluid a rotational velocity is not generated from an initially longitudinal velocity if there is no vorticity at linear order @xcite , this does not remain the case if the imperfect nature of the fluid is properly taken into account @xcite .
a reasonable `` upper - bound '' estimate of the magnetic field that arises from vorticity can be calculated from the angular momentum of a protogalaxy , yielding an angular velocity @xmath23 and therefore a magnetic field .
@xcite showed that magnetic fields of magnitude @xmath195 would arise in the radiation era due to vorticity . from @xcite , an estimate of the angular momentum in the baryonic component of galaxies
is @xmath196 , which yields a magnetic field of @xmath197 for a protogalaxy at recombination .
this estimate of the field strength is reasonable as an upper bound , as the angular momentum from @xcite is for the entire protogalactic system , but the associated magnetic field energy may appear distributed across many smaller - scale modes .
furthermore , this estimate is comparable to and consistent with the field strengths found in @xcite .
it is therefore our assessment that their paper is the most correct of the works detailing magnetic fields arising from the vorticity generated by cosmological perturbations .
it should not be surprising that the calculational frameworks of this paper and those involving magnetic fields arising from vorticity can lead to comparable results , as both calculations rely on second order quantities .
the calculation of a vorticity in the context of cosmological perturbation theory requires a velocity term that is explicitly second order , whereas section 4 of this paper requires the convolution of two first order quantities . on dimensional grounds ,
the two methods should yield comparable results .
the results in this paper rely only on fields known to exist , and whether the contribution from charge separations and currents or from vorticity dominates can vary both with scale and from system to system .
the usual evolution of cosmological perturbations leads in a straightforward manner to charge separations and currents , which necessarily arise from the differing interactions between protons and electrons with photons .
as has been shown in section 4 , magnetic fields follow directly from these sources . a seed field of @xmath198 , as generated via the charge separations and currents induced by structure formation , would provide an excellent candidate for the origins of cosmological magnetic fields . while field amplification due to gravitational collapse is negligible at the epoch of recombination , this will not be the case at all times . at recombination
, the universe has only been matter - dominated for a brief time , and thus density perturbations have only grown by a small amount in that time , leading to an insignificant amplification of the field strength . as magnetic flux is frozen in , nonlinear collapse causes @xmath199 to increase by many orders of magnitude @xcite .
however , the major source of amplification of an initial seed field comes from dynamo effects , as discussed in section 2 .
the key to solving the puzzle of the origin of cosmic magnetic fields lies in determining whether the seed fields produced by a given mechanism can be successfully amplified into the @xmath200 fields observed today .
a major problem with many of the mechanisms that produce seed fields is that they produce weak fields at times insufficiently early for dynamo amplification to produce fields as large as @xmath201 g .
the biermann mechanism , for instance , can produce seed fields of order @xmath202 g , but only at a redshift of @xmath203 .
although those initial fields are larger than the @xmath204 g fields produced by the growth of cosmic structure , the fact that magnetic fields from structure formation are in place at @xmath205 makes them an extremely attractive candidate for the seeds of cosmic magnetic fields . as argued by @xcite , a seed field as small as @xmath206 g at recombination could possibly be amplified into a @xmath207 g field today . clearly , more work on understanding dynamo amplification is necessary before a definitive solution to the puzzle of cosmic magnetic fields can emerge .
one interesting mechanism worth investigating further is for the cosmic seed fields generated by density perturbations to seed supermassive black holes .
it is known that the magnetic field energy in active galactic nuclei and quasars is comparable to the magnetic field energy in an entire galaxy .
however , these structures can not generate their own magnetic fields from nothing ; they require a pre - existing seed field .
it therefore appears to be a reasonable possibility that the seed fields generated by cosmic structure formation could provide the necessary fields to seed supermassive black holes .
the resultant amplification via collapse and dynamo effects could explain the origin of large - scale magnetic structures in the universe .
if large - scale magnetic fields exist at the epoch of recombination , they may be detectable by upcoming experiments .
the results shown in figure [ bgraph ] provide a prediction of large - scale magnetic fields at the epoch of the cosmic microwave background .
sufficiently large magnetic fields on large scales at recombination may be detectable by planck @xcite , although current estimates of their sensitivity indicate that the field strengths predicted in this paper @xmath208 would be significantly out of range of planck s capabilities @xmath209 .
nonetheless , a knowledge of the field strengths at recombination allow for predictions of cmb photon polarizations and faraday rotation , both of which may be , at least in principle , observable .
it is also of interest to note that any primordial charge asymmetry or large - scale currents ( and therefore magnetic fields ) created in the very early universe will be driven away by these dynamics .
equation ( [ solve ] ) has an approximate solution for @xmath140 which is critically ( exponentially ) damped at redshift @xmath210 , capable of reducing any pre - existing charge or current by as much as a factor of @xmath211 .
as this factor is extraordinarily large , it is easy to conclude that any initial @xmath117 or @xmath134 will be driven quickly to the value given by equations ( [ dsoln ] ) and ( [ tsoln ] ) at the epoch of critical damping .
this result is independent of @xmath146 and ought to be applicable even to a charge asymmetry on large scales , perhaps approaching that of the present horizon . as a result of this reduction of pre - existing charges or currents ,
the late - time @xmath212 solutions obtained in this paper for charge separations , currents , and magnetic fields ought to be independent of the initial conditions for @xmath117 , @xmath134 , and @xmath199 in the early universe .
overall , the dynamics of ions , electrons , and photons during the radiation era necessarily leads to charge separations and currents on all scales , which in turn generate magnetic fields .
these fields supersede any pre - existing fields and are in place prior to substantial gravitational collapse .
thus , the creation of charge separations and currents from the evolution of cosmological perturbations emerges as a promising and well - motivated new candidate to explain the origins of cosmic magnetic fields .
we thank ed bertschinger for providing the source code to the program , which proved invaluable in our analysis , and steve detweiler for his input towards solving the differential equations in this paper .
we also thank ethan vishniac for his extremely helpful comments on the first draft of this paper . e.r.s .
acknowledges the university of florida s alumni fellowship program for partial support .
this research has made use of nasa s astrophysics data system . | this paper examines the generation of seed magnetic fields due to the growth of cosmological perturbations . in the radiation era , different rates of scattering from photons induce local differences in the ion and electron density and velocity fields .
the currents due to the relative motion of these fluids generate magnetic fields on all cosmological scales , peaking at a magnitude of @xmath0 at the epoch of recombination .
magnetic fields generated in this manner provide a promising candidate for the seeds of magnetic fields presently observed on galactic and extra - galactic scales . |
cell organelles have a variety of morphologies . in some organelles , lipidic necks or pores
connect biomembranes such that they hold a nonzero genus .
for example , a nuclear envelope consists of two bilayer membranes connected by many lipidic pores , which is supported by protein complexes .
the outer nuclear membrane is also connected by the endoplasmic reticulum , which consists of tubular networks and flat membranes .
the fusion and fission of the membrane tubes change their genus .
it is important to understand the topology dependence of their morphologies . since lipid bilayer membranes are basic components of biomembranes , lipid vesicles in a fluid phase are considered a simple model of cells and organelles .
the morphologies of genus-0 vesicles have been intensively studied experimentally and theoretically @xcite . a red - blood - cell shape ( discocyte ) as well as
prolate and stomatocyte , can be reproduced by minimization of the bending energy with area and volume constraints .
other shapes such as a pear and branched tubes are obtained by the addition of spontaneous curvature or area - difference elasticity ( ade ) @xcite .
in particular , the ade model can reproduce experimentally observed liposome shapes very well @xcite .
in contrast to the genus-0 vesicles , vesicles with a nonzero genus have been much less explored .
vesicle shapes with the genuses @xmath3 and @xmath4 were studied in the 1990s @xcite . for @xmath3 , a conformational degeneracy was found in the ground state of the bending energy , where the vesicles can transform their shapes if the vesicle volume is allowed to freely vary @xcite . for @xmath5 , a conformational degeneracy is obtained even with a fixed volume @xcite .
phase diagrams of genus-1 and genus-2 vesicles were constructed by jlicher et al .
@xcite for symmetric shapes .
recently , we found that nonaxisymmetric shapes such as elliptic and handled discocytes also exist in equilibrium for genus-1 vesicles @xcite .
vesicles with @xmath6 were observed for polymersomes , and the budding of hexagonally arranged pores was investigated in ref .
@xcite . to our knowledge , this paper is the only previous study of the shape transition of vesicles with @xmath7 .
thus , vesicle shapes for @xmath7 have been little explored . in this study
, we investigate the vesicle shapes for @xmath8 by using dynamically triangulated membrane simulation with the ade model .
in particular , we focus on the genus-5 vesicles , where various shapes including cubic and five - fold symmetrical shapes are formed
. we will show that the arrangements of the pores remarkably change the shape transitions .
the morphologies of fluid vesicles are simulated by a dynamically triangulated surface method @xcite . since the details of the potentials are described in ref .
@xcite and the general features of the triangulated membrane can be found in ref .
@xcite , the membrane model is briefly described here .
a vesicle consists of @xmath9 vertices with a hard - core excluded volume of diameter @xmath10 .
the maximum bond length is @xmath11 . the volume @xmath12 and surface area @xmath13 are maintained by harmonic potentials @xmath14 and @xmath15 with @xmath16 and @xmath17 , where @xmath18 is the thermal energy .
the deviations in the reduced volume @xmath19 from the target values are less than @xmath20% , where @xmath21 .
a metropolis monte carlo ( mc ) method is used for vertex motion and reconnection of the bonds ( bond flip ) .
the bending energy of a single - component fluid vesicle is given by @xcite @xmath22 where @xmath23 and @xmath24 are the principal curvatures at each point in the membrane .
the coefficient @xmath25 is the bending rigidity .
the spontaneous curvature and gaussian bending energy are not taken into account since the spontaneous curvature vanishes for a homogeneous bilayer membrane and the integral over the gaussian curvature @xmath26 is invariant for a fixed topology . in the ade model
, the ade energy @xmath27 is added as follows @xcite : @xmath28 the areas of the outer and inner monolayers of a bilayer vesicle differ by @xmath29 , where @xmath30 is the distance between the two monolayers .
the area differences are normalized by a spherical vesicle as @xmath31 and @xmath32 to display our results .
the spherical vesicle with @xmath33 has @xmath34 and @xmath35 .
the mean curvature at each vertex is discretized using dual lattices @xcite . in the present simulations , we use @xmath36 and @xmath37 .
these are typical values for phospholipids @xcite .
most of the simulations are performed with the bending and ade potentials under the volume and area constraints . in some of the simulations , @xmath38 and @xmath39
are employed in order to simulate the vesicles without the ade energy and volume constraints , respectively . in the long time
limit , the area difference is relaxed to @xmath40 , although it does not occur on a typical experimental time scale .
the canonical mc simulations of the ade model are performed from different initial conformations for various values of @xmath41 and @xmath42 . to obtain the thermal equilibrium states , a replica exchange mc ( remc ) method @xcite with @xmath43 to @xmath44 replicas
is employed for the genus-5 vesicles .
different replicas have different values of @xmath42 or @xmath45 and neighboring replicas exchange them by the metropolis method . ,
( b),(c ) with the volume constraint at @xmath38 , and ( d ) , ( e ) with the volume constraint at @xmath46 and @xmath47 . ( a ) circular - cage stomatocyte at @xmath48 and @xmath49 .
( b ) stomatocyte at @xmath50 and @xmath51 .
( c ) spherical stomatocyte at @xmath52 and @xmath53 .
( d ) cube at @xmath54 and @xmath55 .
( e ) discocyte at @xmath54 and @xmath56 .
the top and front views are shown . in the front view
, the vesicles are cut along a plane shown as a ( red ) straight line on the left side in the top view .
their front halves are removed and the cross sections are indicated by the thick ( red ) lines . ]
obtained from the simulation with the volume constraint at @xmath38 .
the contour lines of the probability distribution @xmath57 are obtained by the simulation without the volume constraint at @xmath38 .
the circles and squares with the dashed lines represent the phase boundaries of @xmath42 for stable and metastable discocytes , respectively .
the error bars are shown at several data points for @xmath58 and at all data points for the phase boundaries . ]
we intensively investigate vesicles with the genus @xmath59 ( figs . [
fig : snap_stov4][fig : v5 ] ) .
we categorize the vesicle shapes as stomatocyte and discocyte .
a spherical invagination with a narrow neck into the inside of a spherical vesicle is a typical stomatocyte . here
, we also include a nonspherical invagination [ see fig .
[ fig : snap_stov4](a ) ] and an invagination with wide necks [ see fig . [
fig : snap_stov4](d ) ] into the stomatocyte . on the discocyte
, five pores can be aligned in circular , straight , or branched lines [ see figs . [
fig : snap_stov4](e ) , [ fig : v6](b ) , and [ fig : v5](c ) ] .
when the volume constraint and the ade potential are removed ( @xmath39 and @xmath38 ) , the vesicle shape is determined by the bending energy with the topological constraint .
as mentioned in the introduction , the lowest bending - energy states of the vesicles with @xmath60 are not a single conformation but the vesicles transform shapes without changing the bending energy @xcite . at @xmath59 ,
the lowest energy states are a circular - cage stomatocyte , where @xmath1 pores are aligned in a circular line and the pore size largely fluctuates [ see fig .
[ fig : snap_stov4](a ) ] .
the vesicle shapes are distributed around @xmath61 and @xmath62 as shown in fig .
[ fig : phase ] . note that the simulations are performed at a finite temperature ( @xmath36 ) so that the obtained shapes contain thermally excited states around the lowest energy states .
stomatocytes are stable shapes with the volume constraint for @xmath38 . a valley in the free - energy landscape in the @xmath41@xmath63 space is extended from the lowest energy state to low reduced volumes along the solid ( red ) line in fig .
[ fig : phase ] .
this line is calculated using the remc method for @xmath41 with @xmath64 without the ade potential . with decreasing @xmath41
, it is found that the vesicle transforms from the circular - cage shape to a spherical stomatocyte [ see fig .
[ fig : snap_stov4](c ) ] , where @xmath1 pores are distributed on the vesicle surface .
this transformation occurs as a continuous change . in the transient region at @xmath65 , the vesicle has intermediate shapes , in which the positions of one or two pores often deviate from a plane [ see fig . [
fig : snap_stov4](b ) ] .
this shape transformation in the stomatocytes is characteristic for vesicles with @xmath66 but is not obtained for @xmath67 .
thus , the arrangement of the pores appears as a new factor for the high - genus vesicles .
next , we describe the stomatocyte
discocyte transition as a function of @xmath42 using the ade model .
it drastically changes above or below the critical reduced volume @xmath68 of the change in the pore arrangement on the stomatocyte .
figure [ fig : v6 ] shows the free energy and shapes of the vesicles at @xmath69 calculated by the remc simulation .
the lowest free - energy state is the circular - cage stomatocyte at @xmath70 [ see fig . [
fig : v6](c ) ] .
we takes this to be the origin of the free energy , @xmath71 , in this study . with increasing @xmath42 , one of the pores in
the cage stomatocyte gradually opens [ see fig . [
fig : v6](a ) ] , and subsequently , a discocyte with a straight line of @xmath2 pores is formed [ see fig . [
fig : v6](b ) ] . for genus-5 vesicles at @xmath69 and @xmath46 .
( a),(b ) snapshots at @xmath72 ( a ) @xmath73 and ( b ) @xmath74 .
( c ) free - energy profile @xmath75 .
( d ) mean asphericity @xmath76 .
( e ) mean area difference @xmath58 compared to the intrinsic area difference @xmath42 .
the error bars are shown at several data points .
, title="fig : " ] for genus-5 vesicles at @xmath69 and @xmath46 .
( a),(b ) snapshots at @xmath72 ( a ) @xmath73 and ( b ) @xmath74 .
( c ) free - energy profile @xmath75 .
( d ) mean asphericity @xmath76 .
( e ) mean area difference @xmath58 compared to the intrinsic area difference @xmath42 .
the error bars are shown at several data points .
, title="fig : " ] for genus-5 vesicles at @xmath52 and @xmath46 .
( a ) snapshot of a pentagonal pyramid at @xmath77 .
( b ) snapshot of a concave pentagonal pyramid at @xmath78 .
( c ) snapshot of a discocyte at @xmath78 .
( d ) free - energy profile @xmath75 .
the ( blue ) solid line represents @xmath75 of the spherical and pyramidal stomatocytes ; the ( black ) dashed line represents @xmath75 of the discocyte .
( e ) mean asphericity @xmath76 .
( f ) mean area difference @xmath58 compared to the intrinsic area difference @xmath42 .
the solid lines in ( e ) and ( f ) represent the data in equilibrium and the dashed lines represent the data averaged for either pyramid or discocyte .
the error bars are shown at several data points .
, title="fig : " ] for genus-5 vesicles at @xmath52 and @xmath46 .
( a ) snapshot of a pentagonal pyramid at @xmath77 .
( b ) snapshot of a concave pentagonal pyramid at @xmath78 .
( c ) snapshot of a discocyte at @xmath78 .
( d ) free - energy profile @xmath75 .
the ( blue ) solid line represents @xmath75 of the spherical and pyramidal stomatocytes ; the ( black ) dashed line represents @xmath75 of the discocyte .
( e ) mean asphericity @xmath76 .
( f ) mean area difference @xmath58 compared to the intrinsic area difference @xmath42 . the solid lines in ( e ) and ( f ) represent the data in equilibrium and the dashed lines represent the data averaged for either pyramid or discocyte .
the error bars are shown at several data points .
, title="fig:",width=283 ] the vesicle shapes are quantified by a shape parameter , asphericity @xmath79 , and the area difference @xmath63 .
the asphericity @xmath79 is defined as @xcite @xmath80 where @xmath81 are the eigenvalues of the gyration tensor of the vesicle .
the asphericity is the degree of deviation from a spherical shape : @xmath82 for spheres , @xmath83 for thin rods , and @xmath84 for thin disks @xcite .
the spherical stomatocyte and discocyte have @xmath85 and @xmath86 , respectively .
the circular - cage stomatocyte has @xmath87 . as the stomatocyte transforms into the discocyte , @xmath79 and @xmath63 changes abruptly . the transition point ( @xmath88 )
is estimated from the maximum slope of the @xmath89 curves in fig .
[ fig : v6](d ) .
when the discocyte vesicle transforms into the open stomatocyte , the mirror symmetry breaks , and the slope of @xmath90 changes .
this is a second - order type of transition , but it is rounded by the thermal fluctuations .
this characteristic of the transition is the same as the stomatocyte
discocyte transition at @xmath91 and @xmath92 .
the stomatocyte discocyte transformation becomes a discrete transition below the critical reduced volume @xmath68 .
we obtained the coexistence of the stomatocyte and discocyte using canonical mc simulations in the right region of the dashed ( gray ) line with squares in fig .
[ fig : phase ] .
the final shapes are determined by hysteresis from the initial conformations .
for example , cubic and discocyte vesicles are obtained at @xmath54 and @xmath47 [ see figs . [
fig : snap_stov4](d ) and ( e ) ] .
however , the canonical simulation can not determine which shape is more stable .
both shapes can be maintained without transformation into the other shape in this region .
when we use the remc simulation for @xmath42 , the exchange between the stomatocyte and the discocyte does not occur in equilibrium owing to a large free - energy barrier . to overcome such an energy barrier ,
one more order parameter is often added and the remc or alternative generalized ensemble method is performed in a partially or fully two - dimensional parameter space .
previously , we employed the asphericity as an additional order parameter for a constant radius of gyration in order to overcome the energy barrier between the two free - energy valleys of the discocyte and prolate of a genus-0 vesicle @xcite . for the current transition , however , we did not succeed with similar strategies
. therefore , we take a detour via the stomatocyte discocyte continuous transition at @xmath69 .
this type of detour has been used in free - energy calculations @xcite . for liquid
gas phase transitions the barrier can be avoided via supercritical fluids . for membranes ,
an external order field was used to investigate the formation energy of a fusion intermediate @xcite . here
, we use three remc simulations to estimate the free - energy difference @xmath93 between the stomatocyte at @xmath94 and the discocyte at @xmath95 for @xmath52 .
first , the free - energy difference between the stomatocytes at @xmath52 and @xmath96 is calculated as @xmath97 from the remc simulation used for calculating the solid line in fig .
[ fig : phase ] .
second , the difference @xmath98 between the stomatocyte at @xmath70 ( the free - energy minimum for @xmath99 ) and the discocyte at @xmath95 for @xmath69 is calculated as shown in fig .
[ fig : v6](c ) .
third , the difference @xmath100 between the discocytes at @xmath52 and @xmath96 for @xmath95 is calculated from the remc simulation for @xmath41 with @xmath101 and @xmath95 .
hence , the free - energy difference at @xmath52 is calculated as @xmath102 .
we simulate the stomatocyte and discocyte at @xmath52 separately using the remc simulations [ see the lower ( blue ) and upper ( black ) dashed lines in figs . [
fig : v5](e ) and ( f ) ] and obtain the equilibrium states by averaging them with the weight @xmath103 as shown in fig . [ fig : v5](d ) .
this calculation clarifies that the stomatocyte discocyte transformation is a discrete transition at @xmath104 for @xmath52 .
[ see the solid ( red ) lines in figs .
[ fig : v5](e ) and ( f ) ] . next , we investigate the vesicle shapes in detail for @xmath52 . with an increase in @xmath42 ,
the spherical stomatocyte transforms into a pentagonal pyramid [ see fig .
[ fig : v5](a ) ] . with a further increase , the side faces of the pyramid become concave , and the vesicle shape
is represented by a circular toroid connected to a sphere via five narrow necks [ see fig . [
fig : v5](b ) ] .
this concave necked shape prevents the opening of the bottom pore into a discocyte . on the discocytes ,
pores are aligned in circular , straight , or branched lines [ see figs . [
fig : snap_stov4](e ) , [ fig : v6](b ) , and [ fig : v5](c ) ] . at @xmath52 ,
these three alignments coexist . as @xmath41 decreases and increases ,
the circular and straight pore alignments appear more frequently , and the circular and straight alignments only exist for @xmath54 and @xmath96 , respectively . and
( a),(b ) budded stomatocytes at @xmath72 ( a ) @xmath105 and ( b ) @xmath106 .
the bird s eye and front views are shown .
( c ) budded discocyte at @xmath107 . ]
dependence of the vesicle shapes at @xmath52 and @xmath46 . ( a ) snapshots of genus-3 vesicles at @xmath108 .
( b ) snapshots of genus-8 vesicles at @xmath108 .
( c ) upper ( @xmath109 ) and lower ( @xmath110 ) boundaries of the coexistence regions as a function of the genus @xmath2 .
the error bars are shown at all data points .
, title="fig : " ] dependence of the vesicle shapes at @xmath52 and @xmath46 .
( a ) snapshots of genus-3 vesicles at @xmath108 .
( b ) snapshots of genus-8 vesicles at @xmath108 .
( c ) upper ( @xmath109 ) and lower ( @xmath110 ) boundaries of the coexistence regions as a function of the genus @xmath2 .
the error bars are shown at all data points .
, title="fig : " ] our simulations have revealed that the stomatocyte discocyte transformation changes from a continuous to a discrete transition as the stomatocyte changes from the circular - cage to the spherical shape . the crucial difference between these two transitions is the change in the pore alignments . during the circular - cage to discocyte transformation ,
the pore alignment does not change [ see figs . [
fig : snap_stov4](a ) and [ fig : v6](a),(b ) ] .
in contrast , the transformation from the cubic to discoidal shapes at @xmath54 is accompanied by the realignment of the pores from cubic symmetry to five - fold symmetry [ see figs . [ fig : snap_stov4](d ) and ( e ) ] . the transformation from the pyramidal to the discoidal shape at @xmath52 is not accompanied by pore realignment , but the circular alignment of five pores stabilizes the half - open stomatocyte [ see fig .
[ fig : v5](b ) ] .
thus , the pore alignments are the origin of the discrete transitions . at @xmath111
, outward buddings occur for both stomatocytes and discocytes .
since the buds are connected by multiple membrane necks , the bud shapes are not always spherical .
for the stomatocytes , the circular - toroidal and spherical compartments are divided and the toroidal compartment elongates at @xmath112 [ see fig . [
fig : bud](a ) ] . with increasing @xmath42 ,
budding occurs in narrow regions of the toroid [ see fig . [
fig : bud](b ) ] .
discocytes become divided along the pore alignments [ see fig . [
fig : bud](c ) ] . with further increasing @xmath42 ,
more buds are formed .
these budded shapes coexist in a wide range of @xmath42 because of the free - energy barrier for the formation and removal of the narrow necks .
the discrete transitions from stomatocytes to discocytes are obtained for @xmath66 ( see fig . [
fig : disn ] ) .
the features described in the previous section for @xmath59 are general for @xmath66 .
a @xmath2-fold pyramid - shaped vesicle coexists with a discocyte [ see the triangular and concave octagonal pyramids in figs . [
fig : disn](a ) and ( b ) ] . for @xmath113 , stomatocytes transform into discocytes at @xmath114 for @xmath52 . for @xmath115 ,
the coexistence region has no upper @xmath42 limit . at @xmath111 , the vesicles exhibit budding , as for @xmath59 .
we revealed that the stomatocyte discocyte transformation is a discrete shape transition at low reduced volumes @xmath41 for @xmath116 , while it is a continuous transformation at high @xmath41 .
this discreteness is caused by the alignment of @xmath1 pores in the stomatocytes . at high @xmath41 ,
the pores are aligned in a circular line , while the pores are distributed on the entire surface at low @xmath41 .
as the intrinsic area difference @xmath42 increases , this spherical stomatocyte transforms into polyhedral shapes .
the transformations of these vesicles from polyhedral to discoidal shapes are the first - order transition .
we found a continuous transformation from circular - cage to spherical stomatocytes .
however , for vesicles with @xmath6 , this transformation may be a discrete transition since the symmetry of the pores is changed .
the phase behavior at @xmath6 is an interesting problem for further studies .
polyhedral vesicles are formed under various conditions : gel - phase membranes @xcite , phase - separated membranes @xcite , the assembly of protein rods @xcite , and fluid membranes with the accumulation of specific lipids or defects on the polyhedral edges or vertices @xcite .
however , such vesicles typically have multiple ( meta-)stable shapes and it is difficult to produce only a specific type of polyhedral shape . for high - genus vesicles , a polyhedral vesicle has @xmath1 faces .
the vesicles can only form a triangular pyramid for @xmath113 and a cube and pentagonal pyramid for @xmath59 .
thus , the possible polyhedra are limited and the shape can be controlled by @xmath41 and @xmath42 . recently
, the assembly and packing of polyhedral objects have received growing attentions @xcite .
cubic and other polyhedral vesicles can be interesting building blocks , since their shapes are deformable and can be controlled by the osmotic pressure .
a mitochondrion consists of two bilayer membranes .
the inner membrane has a much larger surface area than the outer one and forms many invaginations called cristae . as a model of such a confinement of an outer vesicle , the morphology of a genus-0 vesicle under spherical confinement has been recently studied @xcite .
the confinement induces various shapes such as double and quadruple stomatocytes , a slit vesicle , and vesicles of two or three compartments .
very recently , bouzar et al . reported that the confinement transforms axisymmetric toroids into asymmetric shapes for genus-1 vesicles @xcite .
for higher - genus vesicles , the confinement should similarly stabilize the stomatocytes with respect to the discocytes since the stomatocytes are more compact . | the morphologies of genus-2 to -8 fluid vesicles are studied by using dynamically triangulated membrane simulations with area - difference elasticity .
it is revealed that the alignments of the membrane pores alter the vesicle shapes and the types of shape transitions for the genus @xmath0 . at a high reduced volume , a stomatocyte with a circular alignment of @xmath1 pores continuously transforms into a discocyte with a line of @xmath2 pores with increasing intrinsic area difference .
in contrast , at a low volume , a stomatocyte transforms into a ( @xmath1)-hedral shape and subsequently exhibits a discrete phase transition to a discocyte . |
transliterations play an important role in multilingual entity reference resolution , because proper names increasingly travel between languages .
this process tends to create a substantial number of out of vocabulary ( oov ) words in the multilingual analysis of news and social media .
when gangnam style " topped the music charts of more than 30 countries , a word imported from korean suddenly became part of the language spoken by millions of people around the world .
news events like the catastrophic failure at nuclear power plant bring words associated with new people and places ( fukushima `` ) across languages into common use .
these words do not reside in standard vocabulary lexicons , but are generated as needed via a process of transliteration . detecting transliterated word pairs contributes to many language processing tasks , including entity resolution , translation , topic classification and sentiment analysis , as well as facilitates studying linguistic phenomenon like cross - language morphologic evolution .
however , previous transliteration systems generally focus on a small number of language pairs .
further , they only consider morphological similarity even in translation systems , creating a problem of ' ' false friends " of word pairs which look / sound alike but mean different things . in this paper , we target the problem of generating transliterations between arbitrary pairs of 69 languages , and detecting borrowed words and entities across these languages .
we train both transliteration models and semantic word embeddings in an unsupervised manner using large - scale corpus from wikipedia . as an example , figure [ obama ] shows our transliterations of the name _ obama _ into 25 non - latin scripts .
we provide both our transliteration ( constructed from scratch ) and lowest edit - distance match appearing in 100,000 most frequent wikipedia lexicons for these languages .
our closest match proves to equal the gold standard for 20 out of 23 languages where it appears in the lexicon .
further , our constructed best differs from the gold standard name within at most one character substitution for 22 out of 25 languages .
our major contributions are : * _ training methods for transliteration _ we used wikipedia to build a training set for transliteration , starting from the cross - language links between personal and place names in wikipedia .
we collect a dataset with 576,403 items contributing one or more transliterations from english to other languages yielding reasonable training and testing sets to learn transliterations .
but this is a very dirty training set , because many such translation pairs are not transliterations ( e.g. _ estados unidos _ for _ united states _ ) .
we develop unsupervised methods to distinguish true transliterations from false , and thus clean the training set .
* _ accurate transliteration via substring matching _
we use an expectation maximization approach to use statistics of string alignments to train improved cost matrices via a bayesian probability model .
our methods employ substring matching instead of single - character transition matrices , enabling the recognition of phonemes , character bigrams , and beyond .
+ we have trained models that permit us to construct transliterations for any string between all pairs of 69 languages .
we evaluate our work against a recently - published transliteration system @xcite which has been integrated into the moses statistical machine translation system .
we compare our transliteration to moses on the four languages it supports ( arabic , chinese , hindi , and russian ) , outperforming it in 61 of 64 standards over the set of languages . *
_ distinguishing translations from false friends _ similarly spelled or sounding words can have substantially different meanings .
such pairs that span language boundaries are called false friends , e.g. _ ropa _ in spanish means _ clothes _ , not _ rope)_. by coupling transliteration pair analysis with semantic tests using distributed word embeddings , we can generate comprehensive lexicons of true and false friends .
our methods get very good results in tests against human annotated standards for french ( f1=0.890 ) and spanish ( f1=0.825 ) . +
we use our approach to generate lexicons of true and false friends between english and 69 languages .
we show that the lexically - closest cohort of word pairs has a higher probability of being true friends than words that are more lexically distant in 68 out of 69 languages , indicating our methods provide a good signal to identify borrowed words .
we provide a demo that can transliterate any english string to non - latin languages .
our transliteration code , corpora , weight matrices , and false - friend lexicons for all 69 languages will be made publicly available upon acceptance of this paper .
the rest of this paper is organized as follows .
we review related work in section [ related - work ] . in section [ data - collection ] , we describe the procedure of collecting data . section [ training - model ] talks about our model of learning character - based transliteration cost matrices .
we analyze the performance of our model in section [ results - section ] .
section [ friend - detection ] discusses the application of detecting true and false friends .
finally , we conclude with discussion and ideas for future work .
transliteration research first associates with the field of orthographic similarity detections since sound similarities co - exist with orthographic similarities @xcite .
this work shows reasonableness of character - based transliteration between close languages ( i.e. languages sharing characters ) but does not discuss on distant language pairs . similarly , work on cognate identification
also focus on close language pairs @xcite .
however , we believe multilingual transliterations contribute to even distant languages ( e.g. english and japanese ) when handling oov words and resolving ambiguities .
further transliteration researches divide into two branches .
one tries to study delicate sound changing rules of specific languages @xcite .
especially , an excellent ideas of using wikipedia external links is proposed in @xcite and achieve promising results in english - hebrew transliteration using moses @xcite .
however , all these systems are supervised and require extra linguistic background knowledge during processing .
plus , only one among this work evaluates transliteration on up to 4 languages and it is hard to generalize for multiple languages .
the other branch learns from only sequence of characters .
one of the great advantages against sound based transliteration is that multilingual texts are much easier to obtain .
compares phonetic based systems with spelling based systems on transliterations between english and arabic .
makes transliteration model based on similar spelling rules in close languages .
recent work of is integrated in moses as a module , providing an unsupervised character - based transliteration training model .
proposed a proper name transliteration model on several language pairs .
however , we believe utilizing character - based transliteration model can provide us with even more valuable information in natural language processing tasks .
inspire us to use wikipedia external links to build an aligned multilingual corpus .
however , their work requires language - specific knowledge , for instance , discarding vowels and filtering out junk data using pre - defined consonant matching . in our task
, we use the names of entities ( people and places ) to create a training set for transliteration .
@xcite state that over 40% of the brands choose to create corresponding foreign names via transliterations , by querying freebase in categories containing 3,388,225 entries , we create a precise multilingual transliteration dictionary through wikipedia page titles .
we then perform a rough clean up procedure to ( 1 ) unify punctuation by converting hyphens , dots , comma to underscores and , ( 2 ) remove entries which do not adhere to the ( first name , last name ) or whole name format our final collection contains 576,403 english entries with multilingual mapping . as an additional resource
, we query google translation api to get formal translations of certain english proper nouns to all 69 languages . to reduce machine translation error , we manually pick 1,373 entities without no multi - sense ambiguities from the names of people @xcite , countries and capital cities @xcite , resulting in more than 70,000 pairs of proper name transliteration from english .
table [ cs ] shows statistics of final data size in each language .
80% of the final data will be used for training , 10% is for tuning and the remaining 10% is for testing .
.languages with the largest and smallest set of translated entities , i.e. reflecting the availability of training data . [ cols= " > ,
> , > , > " , ] our methods yield substantial agreement with these published standards , demonstrating the general soundness of our approach . emboldened by these results
, we performed a search for lexically / semantically similar words between english and all 69 of our transliterated languages .
the results appear in table [ truefriends ] . for each language , we report the number of false friends we identify ( column n ) .
the other three columns reflect different notions of true friends : single - word translations according to google ( tp ) , near neighbors in embedding test ( ep ) , and those which survive both of these semantic tests ( b ) . without a language - specific analysis of each of the classes , it is difficult to determine which of these reflect language borrowings most accurately .
the quality of the word embeddings vary substantially by language , as does the quality of google s translation support .
our preferred measure of quality is the ratio of word pairs which survive both tests
( b ) over all that having google translations ( b+tp ) .
the 33 languages colored red and green all have a ratio of @xmath0 , indicating the highest quality embeddings .
the red languages denote the five with the best embeddings , with the poorest five ( in yellow ) reflect languages with excessively small training data ( amharic and khmer ) .
our methods have a particularly difficult time with vietnamese , which bases a misleading similarity to latin languages at the character level .
figure [ plot ] provides a deeper assessment of our cross - language scan . for each language
, we identified which words in its 100,000 word vocabulary were lexically very similar ( edit distance @xmath1 , which is decided by initial value of substitution ) to a word in the english vocabulary .
we then considered the next closest 10,000 word pairs , which should also be enriched in real transliterations ( by contrast , only 0.01% random word pairs have a translation link ) but less enriched than the initial cohort .
indeed , figure [ plot ] ( left ) shows this to be true for 68 of 69 languages , denoted by points in the upper left triangle . to establish that our embedding test accurately eliminates false friends
, we pruned the lower half of each cohort according to the embedding test , i.e. retained only those words whose distance in embedding space was below the median value .
figure [ plot ] ( right ) shows that this action dramatically shifts each language up and to the right . with the exception of three outlier languages ( vietnamese , latin , and maltese ) , well over 50% of our closest cohort are now true friends ( translations ) . for somewhat more than half of the languages ,
the lexicographically second cohort is now rich in true friends to the 50% level .
in this paper , we have developed transliteration models that accurately identify borrowed out of vocabulary ( oov ) words , for 69 different languages .
we have evaluated our transliterations against published gold standards when available and against intrinsic measures when such standards are not available .
further , we demonstrated that adding word embeddings to provide a semantic test enables us to distinguish true borrowings from false friends .
there are several directions to improve the future quality of our transliteration model : * _ phonetic information _
our models improve with additional training data , particularly for resource - poor languages . an exciting way to increase this volume would be aligning speech translations as represented in a phonetic dictionary or sound system ( e.g. ipa ) as suggested in @xcite . * _ multiple transliteration _
though english wikipedia has the richest resources in the world , it is not guaranteed that english is the source language of borrowed names .
currently we employ a star network of transliteration pairs centered through english . a richer graph with other important languages (
e.g. russian and chinese ) would improve performance .
* _ longer - range dependencies _ as we target transliteration , our model should utilize longer range dependencies .
observe that a silent e at the end of english words changes the pronunciation of vowels earlier in the word , so the li " is different in lit " and like " . under context
the moses system with optimization exploits such phenomena , but we believe with we can learn such pronunciation features from the text itself .
this research was partially supported by nsf grants dbi-1060572 and iis-1017181 , and a google faculty research award .
nasreen abduljaleel and leah s larkey . 2003 .
statistical transliteration for english - arabic cross language information retrieval . in _ proceedings of the twelfth international conference on information and knowledge management _ , pages 139146 .
yaser al - onaizan and kevin knight .
machine transliteration of names in arabic text . in _ proceedings of the acl-02 workshop on computational approaches to semitic languages _ , pages 113 .
association for computational linguistics .
roger boada , rosa sanchez - casas , jose m gavilan , jose e garcia - albea , and natasha tokowicz .
effect of multiple translations and cognate status on translation recognition performance of balanced bilinguals .
, 16(01):183197 .
ondrej bojar , christian buck , christian federmann , barry haddow , philipp koehn , johannes leveling , christof monz , pavel pecina , matt post , herve saint - amand , et al .
findings of the 2014 workshop on statistical machine translation . in _ proceedings of the ninth workshop on statistical machine translation _ , pages 1258 .
association for computational linguistics baltimore , md , usa . mauro cettolo , christian girardi , and marcello federico . 2012 .
wit3 : web inventory of transcribed and translated talks . in _ proceedings of the 16th conference of the european association for machine translation ( eamt ) _ , pages 261268 .
june np francis , janet py lam , and jan walls . 2002 .
the impact of linguistic differences on international brand name standardization : a comparison of english and chinese brand names of fortune-500 companies . , 10(1):98116 .
gumwon hong , min - jeong kim , do - gil lee , and hae - chang rim . 2009 .
a hybrid approach to english - korean name transliteration . in _ proceedings of the 2009 named entities workshop : shared task on transliteration _ , pages 108111 .
association for computational linguistics .
diana inkpen , oana frunza , and grzegorz kondrak .
2005 . automatic identification of cognates and false friends in french and english . in _ proceedings of the international conference recent advances in natural language processing _ , pages 251257 .
jagadeesh jagarlamudi and hal daum iii .
2012 . regularized interlingual projections : evaluation on multilingual transliteration . in
_ proceedings of the 2012 joint conference on empirical methods in natural language processing and computational natural language learning _ , pages 1223 .
association for computational linguistics .
amit kirschenbaum and shuly wintner .
lightly supervised transliteration for machine translation . in _ proceedings of the 12th conference of the european chapter of the acl ( eacl 2009 ) _ , pages 433441 .
association for computational linguistics , march .
amit kirschenbaum and shuly wintner .
a general method for creating a bilingual transliteration dictionary . in _ proceedings of the seventh conference on international language resources and evaluation ( lrec10 )
_ , pages 273276 .
european language resources association ( elra ) , may .
philipp koehn , hieu hoang , alexandra birch , chris callison - burch , marcello federico , nicola bertoldi , brooke cowan , wade shen , christine moran , richard zens , et al . 2007 .
moses : open source toolkit for statistical machine translation . in _ proceedings of the 45th annual meeting of the acl on interactive poster and demonstration sessions _ , pages 177180 .
association for computational linguistics .
grzegorz kondrak and bonnie dorr .
identification of confusable drug names : a new approach and evaluation methodology . in
_ proceedings of the 20th international conference on computational linguistics _
, page 952 .
association for computational linguistics .
gideon s mann and david yarowsky . 2001 .
multipath translation lexicon induction via bridge languages . in _ proceedings of the second meeting of the north american chapter of the association for computational linguistics on language technologies _ , pages 18 .
association for computational linguistics .
matt post , chris callison - burch , and miles osborne .
2012 . constructing parallel corpora for six indian languages via crowdsourcing . in _ proceedings of the seventh workshop on statistical machine translation _
, pages 401409 .
association for computational linguistics .
michel simard , george f foster , and pierre isabelle .
1993 . using cognates to align sentences in bilingual corpora . in _ proceedings of the 1993 conference of the centre for advanced studies on collaborative research : distributed computing - volume 2 _ , pages 10711082 .
ibm press .
prayut suwanvisat and somboon prasitjutrakul .
thai - english cross - language transliterated word retrieval using soundex technique . in _ proceesings of the national computer science and engineering conference , bangkok , thailand_. paola virga and sanjeev khudanpur . 2003 .
transliteration of proper names in cross - lingual information retrieval . in _ proceedings of the acl 2003 workshop on multilingual and mixed - language named entity recognition - volume 15 _ , pages 5764 .
association for computational linguistics . | transliterations play an important role in multilingual entity reference resolution , because proper names increasingly travel between languages in news and social media .
previous work associated with machine translation targets transliteration only single between language pairs , focuses on specific classes of entities ( such as cities and celebrities ) and relies on manual curation , which limits the expression power of transliteration in multilingual environment .
by contrast , we present an unsupervised transliteration model covering 69 major languages that can generate good transliterations for arbitrary strings between any language pair .
our model yields top-(1 , 20 , 100 ) averages of ( 32.85% , 60.44% , 83.20% ) in matching gold standard transliteration compared to results from a recently - published system of ( 26.71% , 50.27% , 72.79% ) .
we also show the quality of our model in detecting true and false friends from wikipedia high frequency lexicons .
our method indicates a strong signal of pronunciation similarity and boosts the probability of finding true friends in 68 out of 69 languages . |
many real - life systems can be modeled using markov chains .
fields of application are queueing theory , telecommunications , option pricing , etc . in most interesting situations ,
analytic formulas are not available and the state space of the chain is so large that classical numerical methods would require a considerable computational time and huge memory capacity .
so monte carlo ( mc ) simulation becomes the standard way of estimating performance measures for these systems .
a drawback of mc methods is their slow convergence , with respect to the number of random points used .
various techniques have been developed , in order to reduce the variance of the approximation , including stratified sampling and latin hypercube sampling @xcite .
it is shown in a series of papers @xcite that each step of a mc simulation of a markov chain amounts to approximating the measure of a subdomain of the @xmath1-dimensional unit hypercube @xmath2 .
the techniques presented here use stratified samples for calculating this approximation . among stratification strategies ,
we first consider the simple approach ( smc ) : the unit hypercube is divided into @xmath0 subcubes having the same measure , and one random point is chosen in each subcube . for latin hypercube sampling ( lhs ) , the projections of the points on each coordinate axis are evenly distributed : one projection in each of the @xmath0 subintervals that uniformly divide the unit interval @xmath3 .
then we propose an hybrid method between smc and lhs , that has properties of both approaches , with one random point in each subcube and one projection in each subinterval ; we call this technique _ sudoku _ sampling ( ss ) due to the properties of the points recalling a sudoku grid
. the improved accuracy of stratified methods may be lost for problems in which we have to approximate the measure of subdomains with irregular boundaries .
it is necessary to take special measures to make optimal use of the greater uniformity associated with stratified samples .
this is achieved in @xcite through the additional effort of reordering the copies of the chain at each time step .
this type of sorting was initiated by @xcite in the context of quasi - monte carlo ( qmc ) methods .
this paper is organized as follows . in section [ sec : integration ] we present smc , lhs and ss methods for numerical integration .
we recall variance bounds for smc and ss and we establish a new bound for the variance of lhs approach in the restrictive case of the approximate calculation of the measure of an interval in dimension @xmath1 . in section [ sec : simulation ] , we propose a mc simulation of markov chains using stratified samples in the context of discrete markov chains with continuous multi - dimensional state space . the results of numerical experiments are presented in section [ sec : illustrations ] .
we compute the values of european and asian options and we compare the variance of the results and the efficiency of the approaches .
it is shown that both smc and ss strategies outperform mc or lhs approaches .
finally , we give some perspectives for future work .
let @xmath4 be a given dimension ; then @xmath5 is the @xmath1-dimensional half - open unit hypercube and @xmath6 denotes the @xmath1-dimensional lebesgue measure .
if @xmath7 is a square - integrable function defined on @xmath5 , we want to approximate @xmath8 for the usual mc approximation , @xmath9 are independent random variables uniformly distributed over @xmath5
. then @xmath10 is an unbiased estimator of @xmath11 .
when @xmath12 , for some measurable @xmath13 , one has @xmath14 a simple stratified sampling ( smc ) method was proposed in @xcite and further analyzed in @xcite . for @xmath15 , put @xmath16 let @xmath17 be independent random variables , with @xmath18 uniformly distributed over @xmath19 .
then @xmath20 is another unbiased estimator of @xmath11 . in @xcite , we have analysed the following case : we consider a function @xmath21 and we define @xmath22 then for @xmath23 we obtain @xmath24 if @xmath25 is of bounded variation @xmath26 in the sense of hardy and krause ( we refer to @xcite for this concept ) .
latin hypercube sampling ( lhs ) was introduced in @xcite and examined studiously in @xcite .
let @xmath27 and @xmath28 be independent random variables , where @xmath29 is uniformly distributed over @xmath30 . if @xmath31 are independent random permutations of @xmath32 and @xmath33 , then each @xmath34 is uniformly distributed over @xmath5 .
consequently , @xmath35 is another unbiased estimator of @xmath11 .
we have proposed in @xcite a combination of smc and lhs : we construct @xmath15 random points in @xmath5 such that in every interval @xmath36 ( for @xmath37 and @xmath38 ) or @xmath19 ( for @xmath39 ) lies only one point of the set ( property @xmath40 ) .
this is achieved as follows .
if @xmath41 , we put @xmath42 .
let @xmath43 be random bijections @xmath44 and @xmath45 be random variables uniformly distributed on @xmath46 ; all these variables are assumed to be mutually independent .
we put @xmath47 the point set @xmath48 has property @xmath40 .
if @xmath49 is defined by @xmath50 it is an unbiased estimator of @xmath11 .
the following variance bound is established in @xcite .
let @xmath13 be such that , for all @xmath51 , with @xmath37 , @xmath52 where @xmath53 are lipschitz continuous functions @xmath54 .
then , for @xmath55 , we have @xmath56 where @xmath57 is a lipschitz constant ( for the maximum norm ) for all the @xmath53 .
we prove a similar result for lhs , in a very restrictive case .
if @xmath58 is a subinterval of @xmath5 , then the variance of the lhs approximation @xmath59 satisfies ( for @xmath60 ) @xmath61 we have @xmath62 for @xmath63 with @xmath64 , we set @xmath65 .
if @xmath66 , then @xmath67 we may assume that @xmath58 is a closed interval : @xmath68,\ ] ] with @xmath69 and @xmath70 .
let us note @xmath71 : = \{1,2 , \ldots , s \}$ ] and , for @xmath72 $ ] , denote @xmath73 \setminus h$ ] .
then @xmath74 } \sum_{\epsilon_i = \pm 1 , i \in h^c } \prod_{h \in h } n_h ( n_h+x_h^{-1 } + x_h^{+1 } -1 ) \prod_{i \in h^c } x_i^{\epsilon_i } ( n_i+ x_i^{-\epsilon_i } ) \\ & = & \frac{1}{n^{2s } } \prod_{i=1}^s ( n_i ( n_i + x_i^{-1 } + x_i^{+1}-1 ) + ( n_i + x_i^{-1 } ) x_i^{+1 } + ( n_i + x_i^{+1 } ) x_i^{-1 } ) .\end{aligned}\ ] ] hence @xmath75 since @xmath76 we obtain @xmath77 and the result follows .
in this section , we use the previous stratification techniques for markov chains simulation . let @xmath78 ; we consider an homogeneous markov chain @xmath79 with state space @xmath80 , evolving according to the stochastic recurrence : for @xmath81 @xmath82 here @xmath83 is a sequence of i.i.d .
uniform random variables over @xmath84 ( for @xmath85 ) and each @xmath86 is a measurable map
. the distribution @xmath87 of @xmath88 is known , and our aim is to approximate the distribution @xmath89 of @xmath90 .
the standard iterative monte carlo scheme proceeds as follows .
a large number @xmath0 of samples @xmath91 are drawn from the initial distribution @xmath87 ; then we generate @xmath0 sample paths of the chain as follows . for @xmath81 and for each @xmath92 @xmath93 where @xmath94 are pseudo - random numbers simulating i.i.d .
uniform random variables over @xmath84 , independent from all variables introduced previously .
qmc variants have been proposed to improve the accuracy of the method @xcite .
the pseudo - random numbers @xmath95 are replaced with quasi - random numbers ; in order to benefit from the great uniformity of quasi - random points , one possibility is to sort the states @xmath96 by position in every step . since qmc methods do not give confidence intervals , randomized qmc algorithms have also been introduced in @xcite , with randomized quasi - random points . in the present paper , we propose a scheme using the sampling strategies presented in section [ sec : integration ] .
let @xmath97 denote the set of all nonnegative measurable functions on @xmath98 . from ( [ eq : chaintransition ] )
, we obtain @xmath99 let @xmath100 be an integer and put @xmath101 . for each @xmath81 , we are looking for an approximation of @xmath89 of the form @xmath102 where @xmath103 is a subset of @xmath98 to be determined .
we first sample a point set @xmath104 of @xmath0 states from the initial probability distribution @xmath87 .
once we have calculated a point set @xmath105 such that @xmath106 _ approximates _
@xmath89 , we compute @xmath107 in two steps : we first sort the states of @xmath105 according to their successive coordinates , then we perform a numerical integration using a stratified sample . [
[ step-1-relabeling - the - states ] ] step 1 : relabeling the states + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we label the states @xmath108 using a multi - dimensional index @xmath109 with @xmath110 , such that : : : if @xmath111 then @xmath112 , : : if @xmath113 then @xmath114 , : : @xmath115 : : if @xmath116 then @xmath117 . in the case
@xmath118 , this reduces to simply sort the states by increasing order .
if @xmath119 , the @xmath0 states are first sorted in @xmath120 batches of size @xmath121 according to their first coordinates ; then each batch is sorted in subgroups of @xmath120 batches of size @xmath122 by order of the second coordinates , and so on . at the last step of the sorting , subgroups of size @xmath123
are ordered according to the last coordinate of the state .
this type of nested sorting was introduced in @xcite for the qmc simulation of the boltzmann equation : since the algorithm is described by a series of numerical integrations , the sorting tends to reduce the number of the jumps of the integrand .
[ [ step-2-using - stratified - samples - for - transition ] ] step 2 : using stratified samples for transition + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we define a probability measure @xmath124 on @xmath98 by replacing @xmath89 with @xmath125 in eq .
[ eq : caractransition ] : @xmath126 to obtain a uniform approximation of @xmath127 , similar to ( [ eq : defapprox ] ) , we use a quadrature with stratified samples : let @xmath128 be pseudo - random numbers simulating stratified variables on @xmath129 as described in section [ sec : integration ] , independent from all variables introduced previously . for @xmath109 with @xmath110 , let @xmath130 be the indicator function of the interval @xmath131
. for @xmath132 , denote @xmath133 then we have : @xmath134 we obtain @xmath135 by @xmath136 the second step of the algorithm may be written as follows . for @xmath137 let @xmath138 and @xmath139 ; for @xmath140 , let @xmath141 . then @xmath142 ( compare with eq .
[ eq : mctransition ] ) . here
the states are labeled using a multi - dimensional index @xmath143 with @xmath144 .
the first @xmath1 components of @xmath145 are used to select the state of the chain that perform a transition , while the remaining @xmath146 components are used to determine the new state .
in this section , we compare the stratified strategies with the standard mc scheme in numerical experiments in the black - scholes model and under the risk - neutral measure , the asset price @xmath147 at time @xmath148 obeys the stochastic differential equation : @xmath149 , where @xmath150 is the risk - free interest rate , @xmath151 the volatility parameter and @xmath152 is a standard brownian motion .
the solution of this equation is given by @xmath153 let @xmath154 be the maturity date and @xmath155 the strike price .
we want to estimate the value of the call option : @xmath156 $ ] . to formulate the problem as a markov chain , we discretize the interval @xmath157 $ ] using observation times @xmath158 .
the discrete version of ( [ eq : st ] ) can be written as : for @xmath81 @xmath159 where @xmath160 . in this example @xmath161 .
we choose the following parameters : @xmath162 , @xmath163 , @xmath164 , @xmath165 , @xmath166 , @xmath167 and @xmath168 , for @xmath169 .
we want to compare the variances of the mc , lhs , smc and ss estimators of @xmath170 .
we replicate the calculation independently @xmath171 times and we compute the sample variance . figure [ fig : europ_var ] shows the results as functions of @xmath0 , for @xmath172 , in log - log scale ( base @xmath173 ) .
copies of the calculation of @xmath170 as a function of @xmath0 .
mc ( @xmath174 ) , lhs ( @xmath175 ) , smc ( @xmath176 ) and ss ( @xmath177 ) outputs , in log - log scale ( base @xmath173).,scaledwidth=90.0% ] it is clear that smc and ss produce smaller variances than mc and lhs ( for the same @xmath0 ) .
when comparing the results of smc and ss , we can see that the later approach outperforms the former . at each step of the ss algorithm ,
the mapping @xmath178 is one - to - one , so that each state is considered exactly once for a transition . assuming that @xmath179 , linear regression is used to evaluate @xmath180 .
the outputs are listed in table [ tab : europ ] .
the convergence rates are close to those established for numerical integration in dimension @xmath173 .
.european option : order @xmath180 of the variance of the calculation of @xmath170 .
[ cols="^,^,^,^",options="header " , ] as before , smc and ss stratification techniques give smaller variances and better convergence rates .
but the advantage of the ss algorithm compared to the smc is lost . at each step of the ss algorithm , the mapping @xmath181 is not necessarily one - to - one .
the efficiencies of the four methods are reported in figure [ fig : asian_eff ] .
smc and ss calculations give similar results and outperform mc and lhs outputs .
copies of the calculation of @xmath182 as a function of @xmath0 .
comparison of mc ( @xmath174 ) , lhs ( @xmath175 ) , smc ( @xmath176 ) and ss ( @xmath177 ) outputs , in log - log scale ( base @xmath173 ) . ]
we have proposed upper bounds for the variance , when we approximate the integral of an indicator function of a subdomain of @xmath5 with stratified monte carlo techniques .
we have proposed strategies for simulating markov chains using stratified samples and we have shown on examples that this approach could lead to better efficiency than naive monte carlo simulation . the variance bound of the lhs approximation is obtained in a very restrictive case and should be extended to less specific subdomains of @xmath5 . the analysis of stratified simulation of markov chains remains undone and will be the subject of future work .
[ 73 ] r. c. h. cheng , t. davenport , the problem of dimensionality in stratified sampling , management science 35 ( 1989 ) 12781296 .
r. el - haddad , r. fakhereddine , c. lcot , stratified monte carlo integration , in : k. k. sabelfeld , i. dimov ( eds . ) , monte carlo methods and applications , de gruyter , berlin , 2013 , pp .
r. el - haddad , r. fakhereddine , c. lcot , g. venkiteswaran , extended latin hypercube sampling for integration and simulation , in : j. dick , f.y .
kuo , g.w .
peters , i.h .
sloan ( eds . ) , monte carlo and quasi - monte carlo methods 2012 , springer , berlin , 2014 , pp . 317330 .
r. el - haddad , c. lcot , p. lecuyer , quasi - monte carlo simulation of discrete - time markov chains on multidimensional state spaces , in : a. keller , s. heinrich , h. niederreiter ( eds . ) , monte carlo and quasi - monte carlo methods 2006 , springer , berlin , 2008 , pp . 413429 .
r. el - haddad , c. lcot , p. lecuyer , n. nassif , quasi - monte carlo methods for markov chains with continuous multi - dimensional state space , mathematics and computers in simulation 81 , ( 2010 ) , 560567 .
m. evans , t. swartz , approximating integrals via monte carlo and deterministic methods , oxford university press , oxford , 2000 .
g. s. fishman , monte carlo , springer , new york , 1996 .
p. glasserman , monte carlo methods in financial engineering , springer , new york , 2004 .
s. haber , a modified monte - carlo quadrature , mathematics of computation 20 ( 1966 ) 361368 . c. lcot , a direct simulation monte carlo scheme and uniformly distributed sequences for solving the boltzmann equation , computing 41 ( 1989 ) 4157 . , a quasi - monte carlo scheme using nets for a linear boltzmann equation , siam journal on numerical analysis 35 ( 1998 ) 5170 .
c. lcot , b. tuffin , comparison of quasi - monte carlo - based methods for the simulation of markov chains , monte carlo methods and applications 10 ( 2004 ) 377384 . , quasi - monte carlo methods for estimating transient measures of discrete time markov chains , in : h. niederreiter ( ed . ) , monte carlo and quasi - monte carlo methods 2002 , springer , berlin , 2004 , pp . 329343 .
p. lecuyer , efficiency improvement and variance reduction , in : j. d. tew , s. manivannan , d. a. sadowski , a. f. seila ( eds . ) , proceedings of the 1994 winter simulation conference , ieee press , 1994 , pp .
p. lecuyer , c. lcot , a. larchevque - gaudet , on array - rqmc for markov chains : mapping alternatives and convergence rates , in : p. lecuyer , a. b. owen ( eds . ) , monte carlo and quasi - monte carlo methods 2008 , springer , berlin , 2009 , pp .
p. lecuyer , c. lcot , b. tuffin , randomized quasi - monte carlo simulation of markov chains with an ordered state space , in : h. niederreiter , d. talay ( eds . ) monte carlo and quasi - monte carlo methods 2004 , springer , berlin , 2006 , pp .
p. lecuyer , c. lcot , b. tuffin , a randomized quasi - monte carlo simulation method for markov chains , operations research 56 ( 2008 ) 958975 . m. d. mckay , r. j. beckman , w. j. conover , a comparison of three methods for selecting values of input variables in the analysis of output from a computer code , technometrics 21 ( 1979 ) 239245 .
h. niederreiter , random number generation and quasi - monte carlo methods , siam , philadelphia , pennsylvania , 1992 .
a. b. owen , monte carlo variance of scrambled net quadrature , siam journal on numerical analysis 34 ( 1997 ) 18841910 .
m. stein , large sample properties of simulations using latin hypercube sampling , technometrics 29 ( 1987 ) 143151 . | we present several monte carlo strategies for simulating discrete - time markov chains with continuous multi - dimensional state space ; we focus on stratified techniques .
we first analyze the variance of the calculation of the measure of a domain included in the unit hypercube , when stratified samples are used .
we then show that each step of the simulation of a markov chain can be reduced to the numerical integration of the indicator function of a subdomain of the unit hypercube .
our approach for markov chains simulates @xmath0 copies of the chain in parallel using stratified sampling and the copies are sorted after each step , according to their successive coordinates .
we analyze variance reduction on examples of pricing of european and asian options : enhanced efficiency of stratified strategies is shown .
stratified sampling , monte carlo simulation , markov chains |
due to large computational costs , @xmath1 , @xmath2 quark mass in current lattice qcd simulations is heavier than the physical value .
one then has to extrapolate results obtained at heavier quark mass to those at the physical point , using some formula such as chiral perturbation theory(chpt ) .
it is not clear , however , whether chpt can be applied at rather heavy quark masses employed so far in lattice qcd simulations .
this situation is gradually changing , thanks to improvements for both computers and algorithms . indeed
the first important message from the annual lattice conference of this year , lattice 2008 , is that full qcd simulations near the physical light quark mass becomes possible . in table[tab :
full_qcd ] , parameters such as number of flavors , the lattice spacing @xmath3 , the spatial extension of the lattice @xmath4 , the minimum pion mass @xmath5 and @xmath6 , are listed for recent large scale full qcd simulations . in the left half of the list , the conventional quark action such as the wilson quark action or the staggered quark action is employed for simulations , while the chirally symmetric quark action such as the overlap quark action or the domain - wall quark action is used in the right half .
let me point out here that the minimum pion mass in the full qcd simulations now becomes below 200 mev for conventional actions and close to 300 mev for chirally symmetric actions , though finite size effects could be sizable in some cases where @xmath7 .
[ tab : full_qcd ] .(incomplete ) list of recent full qcd simulations .
[ cols="^,^,^,^,^,^,^,^,^,^ " , ] results for hadron spectrum from one of the most extensive calculations have been reported in ref.@xcite , where both chiral and continuum extrapolations have been performed . in this calculation , the minimum pion mass @xmath8 mev is light enough , and @xmath9 is always satisfied . in fig.[fig : spectrum ] , the hadron spectrum obtained in the continuum limit is compared with experiment . here
the overall scale , the light quark mass and the strange quark mass are fixed by @xmath10 , @xmath11 and @xmath12 . as can be seen from the figure , the agreement between lattice qcd and experiment is excellent .
, @xmath12 and @xmath10 are used as inputs to fix free parameters of qcd.,width=207 ] it is also reported recently that the pion mass can be smaller than 160 mev in the 2 + 1 flavor full qcd simulation at @xmath13 fm@xcite . in fig .
[ fig : pacs - cs ] , the pion mass squired divided by the light quark mass , @xmath14 , is plotted as a function of the light quark mass , where black solid circles are previous results from cp - pacs / jlqcd collaborations@xcite obtained at heavier quark masses while red solid circles are newt ones from pacs - cs collaboration@xcite . as quark mass decreases , this ratio increasingly deviates from the straight line obtained from the fit with black circles , suggesting an existence of the chiral logarithm in the ratio .
note however that @xmath15 in this calculation , so that the finite size effect might be sizable at the lightest quark mass . to reduce the magnitude of the finite size effect less than a % level , gauge configurations at @xmath16 fm with @xmath17 mev corresponding @xmath18 are currently accumulated by this group .
this simulation indeed will become the real qcd calculation . as a function of @xmath19 at @xmath13 fm and @xmath20 fm .
red circles correspond to @xmath21 , 570 , 412 , 296 , 156 mev from right to left .
the vertical dashed line is the physical point .
, width=207 ] the second message of this year from lattice qcd is that nlo su(2 ) chiral perturbation theory(chpt ) seems to work at @xmath22 mev while nlo su(3 ) chpt seems to fail at the physical strange quark mass .
the first positive statement can be established by both 2-flavor full qcd simulations with the small enough quark mass and use of chirally symmetric quark such as overlap or domain - wall quark .
the second negative statement has been concluded from analyses in 2 + 1 flavor full qcd simulations . as a consequence su(2 ) chpt analysis is applied even to data in 2 + 1 flavor qcd . in fig.[fig : chpt ] @xmath23(top - left ) and @xmath24(bottom - left ) are plotted as function of the quark mass @xmath25 with the 2-flavor overlap fermion at @xmath26 fm@xcite .
the lightest point corresponds to @xmath27 mev while the heaviest to 720 mev .
the lightest 3 points at @xmath28 mev are fitted by nlo su(2 ) chpt formula given by @xmath29+c_3 m , \qquad f_\pi = f\left[1-m \ln m\right ] + c_4 m , \label{eq : chpt}\end{aligned}\ ] ] where @xmath30 ( green dash - dotted line ) , @xmath31 ( blue dashed line ) or @xmath32 ( red solid line ) .
differences among these three choices are higher order of chpt in eq.([eq : chpt ] ) . the fact that all three fits work reasonably well at @xmath28 mev establishes the validity of the nlo su(2 )
chpt at @xmath33 mev .
it is also shown in ref.@xcite that all data points up to the one at @xmath34 mev can be fitted by the nnlo formula of the chpt with @xmath35 ,
whose nlo fit well describes data even beyond the fitted range .
the convergence behaviour of the chiral expansion is also plotted for @xmath23(top - right ) and @xmath24(bottom - right ) in fig.[fig : chpt ] , which shows , for example , that the nlo contribution is a -10% ( + 28% ) level for @xmath23 ( @xmath24 ) at @xmath36 mev , while the nnlo correction gives about + 3%(18% ) .
although the chiral expansion is at least convergent for both quantities , the nnlo contribution for @xmath24 becomes significant already at @xmath36 mev , corresponding to the @xmath37 meson mass region for the 3 flavor case . in the 2 + 1 flavor full qcd simulations with the domain - wall quark , it is reported@xcite that the nlo su(3 ) ( partially quenched ) chpt can not explain the quark mass dependence of mass and decay constant in the pseudo - scalar meson sector around the physical strange quark mass region .
instead the formula in the nlo su(2 ) chpt plus the heavy strange quark theory works better even for these data in the 2 + 1 flavor full qcd .
this fact is consistent with the conclusion in the previous paragraph that the nnlo correction in the su(2 ) chpt can not be neglected at @xmath36 mev for these quantities .
( top ) and @xmath24 ( bottom ) as a function of @xmath25 at @xmath26 fm .
right : nnlo chiral fit with @xmath38 using all data points for @xmath39(top ) and @xmath40(bottom ) .
solid curves are the full nnlo fit , while dashed ones represent the nlo contributions . ,
title="fig:",width=226 ] ( top ) and @xmath24 ( bottom ) as a function of @xmath25 at @xmath26 fm .
right : nnlo chiral fit with @xmath38 using all data points for @xmath39(top ) and @xmath40(bottom ) .
solid curves are the full nnlo fit , while dashed ones represent the nlo contributions . ,
title="fig:",width=257 ] the 3rd message of this year from lattice qcd is that it becomes difficult to change topological charge in full qcd simulations at the lighter quark mass and/or near the continuum limit .
one of possible solutions to this problem recently investigated is qcd with fixed topological charge .
for example , the two point function between flavor singlet pseudo - scalar densities @xmath41 at fixed topological charge @xmath42 behaves at large separation as @xmath43 where @xmath44 is the space - time volume , @xmath45 is the quark mass , @xmath46 is the expectation values at fixed @xmath42 , @xmath47 and @xmath48 are the topological susceptibility and the 4-th cumulant defined in the ordinary qcd at @xmath49 .
this formula tells us that one can extract informations about qcd at @xmath49 such as @xmath50 or @xmath51 from behaviours of correlation functions calculated at fixed @xmath42 .
this strategy is suitable for lattice qcd since it is easier to fix the topological charge @xmath42 than to change it correctly during simulations , in particular , at small quark mass and/or near the continuum limit .
the jlqcd has taken this strategy for the dynamical overlap project , where the fixed topology not only guarantees the locality of the overlap dirac operator but also makes full qcd simulations with overlap quarks much faster . in fig.[fig : chi_t ] , the topological susceptibility @xmath50 extracted from eq.([eq : chi_t ] ) at @xmath52 is plotted as a function of the light quark mass @xmath53 for 2 flavor@xcite and 2 + 1 flavor@xcite qcd .
the results show a clear linear dependence on the sea quark mass expected from the chpt as @xmath54 for 2 flavor and @xmath55 for 2 + 1 flavor , where @xmath56 is the chiral condensate at massless limit and @xmath57 is the ( fixed ) strange quark mass .
the fits to data using this formula , denoted solid lines in the figure , yield , after a renormalization , @xmath58 ^ 3 $ ] @xmath59 and @xmath60 ^ 3 $ ] @xmath61 . as a function of sea quark mass @xmath53 for @xmath62 ( circles ) and @xmath63 ( triangles ) at @xmath52 .
the solid lines are fits with the form of chpt.,width=219 ]
it has been shown that the zero - viscosity ( perfect ) hydro calculations explain the rhic experimental results well .
this indicates that in heavy - ion collision , after thermalization , the system expands and cools with constant entropy . in this case ,
the equation of state ( eos ) along the line of the constant entropy per baryon number , called the isentropic eos , becomes more relevant than the ordinary eos as a function of temperature at the fixed chemical potential @xmath64 .
the entropy per baryon number , @xmath65 is roughly equal to 300 at rhic , 45 at sps and 30 at ags .
the lattice qcd at non - zero chemical potential , however , is notoriously difficult to be simulated .
so far there is no established method to simulate the full qcd at arbitrary value of the chemical potential @xmath64 .
if the chemical potential is small , the taylor expansion in terms of @xmath64 works well@xcite .
the milc collaboration has calculated the isentropic eos using the taylor expansion method in 2 + 1 flavor qcd with the staggered quarks at @xmath66 220 mev@xcite . in fig.[fig : isentr](left ) their result for @xmath67 where @xmath68 denotes the pressure is plotted as a function of @xmath69 along lines of the constant @xmath70 30 , 45 , 300 and @xmath71 , at @xmath72 ( open symbols ) and @xmath73 ( solid symbols ) where @xmath74 is the number of the lattice sites in the temporal direction which controls the lattice spacing as @xmath75 .
the result shows that the lattice discretization errors are small for this quantity and @xmath67 becomes larger at fixed @xmath69 as @xmath65 decreases . in fig.[fig :
isentr](right ) the result for @xmath76 , obtained by rbc - bielefeld collaborations in 2 + 1 flavor qcd with the staggered quarks at @xmath66 220 mev using the taylor expansion method@xcite , is plotted as a function of @xmath77 on lines of the constant @xmath78 at @xmath72 ( filled ) and @xmath73 ( open ) , where @xmath77 is the energy density .
form this quantities the velocity of sound @xmath79 , which is important for the hydrodynamics calculation in heavy ion collisions , can be extracted as @xmath80 .
although discretization errors are observed , the dependence on @xmath65 seems small in this ratio . as a function of temperature @xmath69 along the lines of constant entropy per baryon number in 2 + 1
flavor qcd@xcite .
right : the ratio of pressure and energy density @xmath76 as a function of @xmath77 along the lines of constant entropy per baryon number 2 + 1 flavor qcd@xcite.,title="fig:",width=219 ] as a function of temperature @xmath69 along the lines of constant entropy per baryon number in 2 + 1 flavor qcd@xcite .
right : the ratio of pressure and energy density @xmath76 as a function of @xmath77 along the lines of constant entropy per baryon number 2 + 1 flavor qcd@xcite.,title="fig:",width=219 ] the finite temperature phase `` transition '' in 2 + 1 flavor qcd at the physical light ( u , d ) quark mass and the physical strange quark mass seems a cross - over at zero or small chemical potential , while the phase transition is expected to become of first order at large chemical potential . between the cross - over region and the first order region
, there must exist an unique second order phase transition point , which is called the critical end point of qcd .
important questions to be answered are as follows .
does the critical end point indeed exist ?
if so at which temperature and density does it occur ? to answer these questions , lattice qcd simulations at large @xmath64 are required .
unfortunately the taylor expansion method mentioned before does not work at such large @xmath64 .
the reweighting method , which looked promising on small volumes , may not be reliable to answer these questions@xcite . a new method appears@xcite , but we have to wait for a while , in order to see whether the method indeed works for this problem .
one established method to investigate the existence of the critical end point is the imaginary chemical potential method , which i will discuss in this subsection . at a given chemical potential @xmath64 in 2 + 1 flavor qcd with the fixed strange quark mass @xmath57
, the second order phase transition appears at @xmath81 where @xmath53 is the light ( u , d ) quark mass .
furthermore the phase transition is of first order at @xmath82 while it becomes a cross - over at @xmath83 .
the statement that the phase transition in 2 + 1 flavor qcd at @xmath84 is a cross - over at physical @xmath53 and @xmath57 correspond to the fact that @xmath85 at @xmath86 where @xmath87 and @xmath88 are physical light and strange quark masses , respectively .
if @xmath89 becomes larger than @xmath90 at larger @xmath64 , it is expected that @xmath89 at @xmath88 becomes equal to @xmath91 at some @xmath92 ; @xmath93 .
this @xmath94 corresponds to the value of the chemical potential at the critical end point of qcd .
if this scenario is correct , it is expected that @xmath89 increases from @xmath90 , so that @xmath95 at @xmath84 .
( note that @xmath96 at @xmath84 . )
the imaginary chemical potential method avoids the complex phase problem of lattice qcd with non - zero @xmath64 by taking @xmath97 , and lattice qcd with real @xmath98 can be easily simulated .
if @xmath89 can be analytically continued to @xmath99 , we have @xmath100 therefore @xmath101 can be extracted from @xmath102 .
the latest result@xcite in 3 flavor qcd with @xmath103 at @xmath72 yields @xmath104 which strongly indicates that the expected critical endpoint is absent at small @xmath105 .
of course it is necessary to check this unexpected behaviour by other methods for the definite conclusion .
2n state with anti - periodic boundary condition at @xmath106 mev(red bars ) and periodic boundary condition ( blue crosses ) at @xmath107 in quenched qcd .
right : the central potentials for @xmath108 ( red ) and @xmath109 ( blue ) @xmath110 @xmath111 states in quenched qcd.,title="fig:",width=226 ] 2n state with anti - periodic boundary condition at @xmath106 mev(red bars ) and periodic boundary condition ( blue crosses ) at @xmath107 in quenched qcd .
right : the central potentials for @xmath108 ( red ) and @xmath109 ( blue ) @xmath110 @xmath111 states in quenched qcd.,title="fig:",width=226 ] the force between nucleons ( nuclear force ) play a central role in nuclear physics . in a recent paper@xcite , which has received general recoginition@xcite , the potential between nucleons has been calculated from lattice qcd in the quenched approximation .
the result qualitatively reproduces all the features of phenomenological potentials that the force at medium to long range is attractive while it becomes repulsive at short distance forming a characteristic repulsive core .
the potential between nucleons , @xmath112 , is extracted from the wave function @xmath113 through the schrdinger equation as @xmath114\varphi_e ( { \bf x})}{\varphi_e({\bf x } ) } , \quad h_0 = \frac{-{\bf \nabla}^2}{2\mu } , \quad \varphi_e({\bf x } ) = \langle 0 \vert n({\bf x } , 0 ) n({\bf 0 } , 0 ) \vert 2n , e\rangle,\end{aligned}\ ] ] where @xmath115 is a nucleon interpolating field , @xmath116 is a 2 nucleon ( 2n ) eigenstate in qcd with energy @xmath117 and @xmath64 is a reduced mass of 2n system ( @xmath118 ) . in ref.@xcite
, @xmath107 has been used to extract the potential , but the potential from this definition may depend on the energy @xmath117 of the 2n state . in fig.[fig :
pot ] ( left ) , the potentials obtained at @xmath107(blue ) and at @xmath106 mev(red ) are compared in quenched qcd@xcite .
the potentials are almost identical between two energies , tough statistical fluctuations becomes sizable at large @xmath119 in the case of @xmath106 mev , partly due to the contamination from higher energy excited states .
it seems that the energy dependence of the potential is weak at this energy range .
an important applications of this method is to calculate the potential between nucleon and hyperon or between two hyperons , which can not be extracted directly from scattering experiments . in fig.[fig : pot ] ( right ) .
the potentials between @xmath120 and @xmath121 in the @xmath110 channel for @xmath108 ( red ) and @xmath109(blue ) states in quenched qcd are presented@xcite .
while overall features are similar to those of nuclear potentials , stronger spin dependence of the potentials is observed in this case . 00
pacs - cs collaboration : s. aoki , _ et al .
_ , arxiv:0807.1661[hep - lat ] .
milc collaboration : c. bernard , _ et al . _ , pos(lattice 2007 ) , 090 ( 2007 ) .
_ et al . _ , science 322 , 1224 ( 2008 ) .
l. del debbio , l. guisti , m. lscher , r. petoronzio , n. tantalo , jhep0702 , 056 ( 2007 ) .
etm collaboration : b. blossier , _ et al .
_ , jhep 0804 , 020 ( 2008 ) .
g. von hippel , r. sommer , j. heitger , s. schaefer , n. tantalo , pos(lattice 2008 ) , 227 ( 2008 ) .
m. gckeler , _ et al .
_ , arxiv:0810.5337[hep - lat ] .
rbc - ukqcd collaborations : c. allton , _ et al . _ , arxiv:0804.0473[hep - lat ] .
jlqcd collaboration : s. hashimoto , _ et al .
_ , pos(lattice 2007 ) , 101 ( 2007 ) . y. aoki , _ et al .
rev d72 , 114505 ( 2005 ) .
jlqcd collaboration : s. aoki , _ et al .
d78 , 014508 ( 2008 ) .
cp - pacs / jlqcd collaborations : t. ishikawa , _ et al . _ ,
d78 , 0011502 ( 2008 ) .
jlqcd collaboration : j. noaki , _
et al/ _ , phys .
lett . 101 , 202004 ( 2008 ) .
r. brower , s. chandrasekharan , j. negele , u .- j .
wiese , phys .
b560 , 64 ( 2003 ) .
s. aoki , h. fukaya , s. hashimoto , t. onogi , phys .
d76 , 065608 ( 2007 ) .
jlqcd and twqcd collaborations : s. aoki , _ et al .
b665 , 294 ( 2008 ) .
jlqcd and twqcd collaborations : t. w. chiu , _ et al .
_ , pos(lattice 2008 ) 072 .
c. r. alton , _ et al .
d66 , 074507 ( 2002 ) . c. bernald , _ et al .
d77 , 013403 ( 2008 ) .
milc collaboration : s. basak , _ et al .
_ , pos(lattice 2008 ) , 171 ( 2008 ) .
rbc - bielefeld collaborations : c. miao , c. schmidt , arxiv:0810.0375[hep - lat ] . s. ejiri , phys .
d73 , 054502 ( 2006 ) .
s. ejiri , phys .
d77 , 014508 ( 2008 ) ; arxive:0804.3227[hep - lat ] . p. de forcrand , o. philipsen , pos(lattice 2008 ) , 208 ( 2008 ) .
n. ishii , s. aoki , t. hatsuda , phys .
, 0022001 ( 2007 ) . in _
research highlights 2007 _ , nature 450 , 1130 ( 2007 ) .
h. nemura , n. ishii , s. aoki , t. hatsuda , arxiv:0806.1094[nucl - th ] .
s. aoki , _ et al .
_ , arxiv:0812.0673[hep - lat ] . | some topics from recent progresses in lattice qcd are reviewed .
full qcd simulations , hadron spectrum , chiral perturbation theory , topological charge , qcd at finite density , isentropic equation of state , qcd critical end point , nuclear potential 11.15.ha , 12.38.gc , 11.30.rd , 12.38.qk , 12.39.fe , 13.75.ev |
the parity - violating ( pv ) nucleon - nucleon interaction has been a subject of interest in nuclear and particle physics for some time . to date
, pv observables generated by this interaction remain the only experimental windows on the @xmath11 , nonleptonic weak interaction . since the 1970 s
, the pv nn interaction has been studied in a variety of processes , including @xmath12 and @xmath13nucleus scattering , @xmath14-decays of light nuclei , the scattering of epithermal neutrons from heavy nuclei , and atomic pv ( for a review , see refs .
the on - going interest in the subject has spawned new pv experiments in few - body systems , including high - energy @xmath12 scattering at cosy , @xmath15 at lansce@xcite , @xmath16 at jlab@xcite , and the rotation of polarized neutrons in helium at nist .
the theoretical analysis of these pv observables is complicated by the short range of the low - energy weak interaction .
the compton wavelength of the weak gauge bosons ( @xmath17 fm ) implies that direct @xmath18 and @xmath19 exchange between nucleons is highly suppressed by the short - range repulsive core of the strong nn interaction . in the conventional framework ,
longer range pv effects arise from the exchange of light mesons between nucleons .
one requires the exchange of the @xmath20 , @xmath21 and @xmath22 in order to saturate the seven spin - isospin channels associated with the quantum numbers of the underlying four - quark strangeness - conserving pv interaction , @xmath23 ( henceforth , the @xmath11 will be understood ) .
these exchanges are parameterized by pv meson - nucleon couplings , @xmath24 , whose values may be extracted from experiment . at present , there appear to be discrepancies between the values extracted from different experiments .
in particular , the values of the isovector @xmath25 coupling , @xmath5 , and the isoscalar @xmath26 coupling , @xmath27 as extracted from @xmath28 scattering and the @xmath14-decay of @xmath29f , do not appear to agree with the corresponding values implied by the anapole moment of @xmath30cs as measured in atomic pv @xcite .
the origin of this discrepancy is not understood .
one possibility is that the use of @xmath21 and @xmath22 exchange to describe the short - range part of the pv nn interaction is inadequate .
an alternate approach , using effective field theory ( eft ) , involves an expansion of the short - range pv nn interaction in a series of four - nucleon contact interactions whose coefficients are _ a priori _ unknown but in principle could be determined from experiment .
the use of @xmath21 and @xmath22 exchange amounts to adoption of a model rather than the use of experiment to determine the coefficients of the higher - derivative operators in this expansion .
whether or not the application of eft to nuclear pv can yield a more self - consistent set of pv low - energy constants than the meson - exchange approach remains to be seen .
a comprehensive analysis of nuclear pv observables using eft has yet to be performed .
the least ambiguous element shared by both approaches involves the long - range @xmath20-exchange interaction . at leading order in the derivative expansion , the pv @xmath25 interaction is a purely isovector , yukawa interaction .
the strength of this interaction is characterized by the same constant
@xmath5 in both the eft and meson - exchange approaches . at the level of the standard model ( sm )
, @xmath5 is particularly sensitive to the neutral current component of @xmath31 . in this respect ,
the result of @xmath29f pv @xmath14-decay measurement is puzzling : @xmath32 where @xmath33 gives the scale of the @xmath24 in the absence of neutral currents@xcite .
this result is especially significant , since the relevant two - body nuclear parity - mixing matrix element can be obtained by isospin symmetry from the @xmath34-decay of @xmath29ne @xcite .
the result in eq .
( [ eq : flourine ] ) is , thus , relatively insensitive to the nuclear model .
theoretical calculations of @xmath5 starting from @xmath31 have been performed using su(6)@xmath35 symmetry and the quark model @xcite , the skyrme model @xcite , and qcd sum rules @xcite . as a benchmark for comparison with experiment
, we refer the su(6)@xmath35/quark model analysis of refs .
hereafter referred to as ddh , fcdh . in order to connect weak vector meson and pion couplings the predictions relating pion couplings alone to hyperon decay data rely only su(3 ) . ]
these authors quote a best value `` and reasonable range '' for the @xmath24 : @xmath36 where here the best value `` is more aptly described as an educated guess , while the reasonable range '' indicates a set of numbers such that theory would be very hard - pressed to explain were the experimental value not found to be within this band .
neverthess , the difference between the best value " of eq .
( [ eq : fcdhbest ] ) and the @xmath29f result would appear to call for an explanation , and in the following note we comment on a possible source of the discrepancy . in general ,
the problem of relating the fundamental weak quark - quark interaction to the low - energy constants which parameterize hadronic matrix elements of that interaction is non - trivial . in the framework of eft
, one may define these constants at tree - level in the hadronic effective theory .
the quantities extracted from experiment in the conventional analysis , however , are not the tree - level parameters , but rather renormalized couplings . denoting the latter as @xmath37 , one has @xmath38 where @xmath39 is the coefficient of the leading - order , pv yukawa interaction in the effective theory , @xmath40 and @xmath41 denotes chiral loop renormalizations of the nucleon and pion wavefunctions , respectively , and @xmath42 denotes contributions from chiral loops and higher - dimension operators to the yukawa interactions ( only the finite parts of these couplings are implied ; loop divergences are cancelled by the corresponding pole terms in @xmath39 and the @xmath43 ) . at leading order in @xmath44 ,
one has @xmath45 , @xmath46 , and @xmath47 .
the renormalized coupling appears as the coefficient in the one - pion - exchange ( ope ) pv nn potential @xmath48 \ \ \ , \ ] ] where @xmath49 is the strong @xmath25 coupling and @xmath50 . neglecting the effects of three - body pv forces and @xmath51-exchange interactions , it is @xmath37 to which the result in eq .
( [ eq : flourine ] ) corresponds .
the relationship between @xmath37 and the coupling obtained by computing @xmath52 in a microscopic model is not immediately transparent . in what follows ,
we make several observations about this relationship .
we first show that @xmath53 and @xmath42 are substantial , so that @xmath37 differs significantly from @xmath39 . to that end
, we compute all of the chiral corrections to the pv yukawa interaction through @xmath54 , where @xmath55 .
we work to leading order in @xmath56 in heavy baryon chiral perturbation theory ( hbchpt ) . of particular significance
is the dependence of @xmath42 on other low - energy constants parameterizing pv @xmath51 production and the pv @xmath57 transition .
we subsequently reexamine the su(6)@xmath35/quark model calculation of refs .
@xcite and argue that most if not all of the chiral loop effects which renormalize @xmath5 are not included in the microscopic calculation .
thus , the relationship between @xmath37 and microscopic calculations remains ambiguous at best .
this ambiguity is unlikely to be resolved until an unquenched lattice qcd calculation of @xmath5 using light quarks becomes tenable . in the meantime
, one should not necessarily view a discrepancy between the experimental value of @xmath37 and microscopic model calculations as disturbing .
our discussion of these observations is organized as follows . in section 2
we summarize our conventions and notation , including the pv chiral lagrangians relevant to @xmath5 renormalization .
section 3 gives a discussion of the loop calculations . in section 4
we comment on the scale of the loop corrections and provide simple estimates of some of the new pv low - energy constants appearing in the analysis .
section 5 gives our discussion of the relationship between @xmath37 and the calculation of refs .
section 6 summarizes our conclusions .
some technical details are relegated to the appendices .
we follow standard hbchpt conventions @xcite and introduce @xmath58 with @xmath59 mev being the pion decay constant .
the chiral vector and axial vector currents are given by @xmath60 for the @xmath61 , we use the isospurion formalism@xcite , treating the @xmath61 field @xmath62 as a vector spinor in both spin and isospin space with the constraint @xmath63 .
the components of this field are @xmath64 the field @xmath65 also satisfies the constraints for the ordinary schwinger - rarita spin-@xmath66 field , @xmath67 we eventually convert to the heavy baryon expansion , in which case the latter constraint becomes @xmath68 with @xmath69 the heavy baryon velocity .
the relativistic parity - conserving ( pc ) lagrangian for @xmath20 , @xmath70 , @xmath61 interactions needed here is @xmath71 t_j^\nu \\ & & + g_{\pi n\delta } [ \bar t^\mu_i ( g_{\mu\nu } + z_0 \gamma_\mu\gamma_\nu ) \omega_i^\nu n + h.c.]\ \ \ , \end{aligned}\ ] ] where @xmath72/2 $ ] while @xmath73 and @xmath74 are the gauge and chiral covariant derivatives , respectively .
explicit expressions for the fields and the transformation properties can be found in @xcite . here
, @xmath75 is an off - shell parameter , which is not relevant in the present work @xcite . in order to obtain proper chiral counting for the nucleon
, we employ the conventional heavy baryon expansion of @xmath76 , and in order to cosistently include the @xmath61 we follow the small scale expansion proposed in @xcite . in this approach energy - momenta and the delta and nucleon mass difference @xmath77 are both treated as small expansion parameters in chiral power counting .
the leading order vertices in this framework can be obtained via @xmath78 where @xmath79 is the original vertex in the relativistic lagrangian and @xmath80 are projection operators for the large , small components of the dirac wavefunction respectively .
we collect some of the relevant terms below : @xmath81n -i { \bar t}^\mu_i [ iv\cdot d^{ij } -\delta^{ij }
\delta + g_1 s\cdot a^{ij } ] t_\mu^j \\ & & + g_{\pi n\delta}[{\bar t}^\mu_i \omega_\mu^i n + \bar n \omega_\mu^{i\dag } t_i^\mu]\end{aligned}\ ] ] where @xmath82 is the pauli - lubanski spin operator and @xmath83 .
the pv analog of eq . ( [ eq : lpc ] ) can be constructed using the chiral fields @xmath84 defined as @xcite : @xmath85 we find it convenient to follow the convention in ref . @xcite and separate the pv lagrangian into its various isospin components .
the hadronic weak interaction has the form @xmath86 where @xmath87 denotes either a charged or neutral weak current built out of quarks . in the standard model , the strangeness conserving charged currents are pure isovector , whereas the neutral currents contain both isovector and isoscalar components .
consequently , @xmath88 contains @xmath89 pieces and these channels must all be accounted for in any realistic hadronic effective theory .
we quote here the relativistic lagrangians , but employ the heavy baryon projections , as described above , in computing loops .
it is straightforward to obtain the corresponding heavy baryon lagrangians from those listed below , so we do not list the pv heavy baryon terms below . for the @xmath90 sector
we have @xmath91\gamma^\mu n \\ \nonumber & & \ \ \
-{h^2_a\over 2 } { \cal i}^{ab } \bar n [ x_r^a a_\mu x_r^b -x_l^a
a_\mu x_l^b]\gamma^\mu\gamma_5 n \ ; , \end{aligned}\ ] ] where @xmath92 is a matrix coupling the @xmath93 to @xmath94 . the above lagrangian
was first given by kaplan and savage ( ks)@xcite .
however , the coefficients used in our work are slightly different from those of ref .
@xcite since our definition of @xmath95 differs by an overall phase .
the term proportional to @xmath3 contains no derivatives . at leading - order in @xmath96
, it yields the pv @xmath2 yukawa coupling traditionally used in meson - exchange models for the pv nn interaction @xcite . unlike the pv yukawa interaction , the vector and axial vector terms in eqs .
( [ n1]-[n3 ] ) contain derivative interactions .
the terms containing @xmath97 and @xmath98 start off with @xmath99 interactions , while all the other terms start off as @xmath2 .
such derivative interactions have not been included in conventional analyses of nuclear and hadronic pv experiments .
consequently , the experimental constraints on the low - energy constants @xmath100 , @xmath101 are unknown .
it is useful to list the first few terms obtained by expanding the lagrangians in eqs .
( [ n1]-[n3 ] ) in @xmath96 . for the present purposes ,
the following terms are needed : @xmath102&\\ { \cal l}^{\pi nn}_{v } = -{h_v^0 + 4/3 h_v^2\over \sqrt{2}f_\pi } [ \bar p\gamma^\mu n d_\mu\pi^+ + \bar n \gamma^\mu p d_\mu \pi^- ] & \\ \nonumber { \cal l}^{\pi nn}_{a } = i{h_a^1+h_a^2\over f^2_\pi}\bar p\gamma^\mu \gamma_5 p ( \pi^+ d_\mu \pi^--\pi^- d_\mu \pi^+ ) & \\ \nonumber + i{h_a^1-h_a^2\over f^2_\pi}\bar n\gamma^\mu \gamma_5 n ( \pi^+ d_\mu \pi^--\pi^- d_\mu \pi^+ ) & \\
+ i{\sqrt{2}h_a^2\over f^2_\pi}\bar p\gamma^\mu \gamma_5 n \pi^+ d_\mu \pi^0 -i{\sqrt{2}h_a^2\over f^2_\pi}\bar p\gamma^\mu \gamma_5 n \pi^+ d_\mu \pi^0 & \ ; .\end{aligned}\ ] ] for the pv @xmath25 yukawa coupling we have also kept terms with three pions . the corresponding pv lagrangians involving a @xmath103 transition are somewhat more complicated .
we relegate the complete expressions to appendix a , and give here only the leading terms required for our calculation . as noted in ref .
@xcite , the one - pion @xmath104 pv lagrangian vanishes at leading order in the heavy baryon expansion .
the two - pion terms are @xmath105 where the couplings @xmath106 are defined in terms of the various su(2 ) pv low - energy constants in appendix a. the pv @xmath107 lagrangians , also listed in appendix a , contain terms analogous to the yukawa , @xmath108 , and @xmath109 terms in eqs .
( [ n1]-[n3 ] ) . since we compute corrections up to one - loop order only , and since the initial and final states are nucleons , the pv @xmath110 terms ( @xmath109-type ) are not relevant here .
the leading , single-@xmath20 yukawa and @xmath108-type interactions are @xmath111 where the coefficients are given in appendix a. one may ask whether there exist additional pv effective interactions that could contribute at the order to which we work . in the pionic sector
there exists one cp - conserving , pv lagrangian : @xmath112 at leading order in @xmath96 , @xmath113 contains five pions .
its lowest order contribution appears at two - loop order at best , so we do not consider it here .
similarly , one may consider possible contributions from two - derivative operators .
there exists one cp - conserving , pv operator : @xmath114 n\ \ \ .\ ] ] there exist three independent pc , two - derivative operators @xcite .
for example , one may choose the following three : @xmath115 @xmath116 @xmath117 n\ ; .\ ] ] as we discuss in appendix b , none of the two - derivative operators in eqs .
( [ eq : twoderivpv ] ) - ( [ eq : twoderivpc ] ) contribute to the renormalization of @xmath3 at the order to which we work in the present analysis .
the leading order loop corrections to the yukawa interaction of eq .
( [ pi - n - n ] ) are generated by the diagrams of figs . 1 - 2 .
as we discuss in appendix b , the contributions from many of the diagrams which nominally renormalize @xmath5 vanish at the order at which we truncate . in particular , none of the vector ( @xmath108-type ) @xmath25 and @xmath107 terms contribute to this order . in what follows , we discuss only the non - vanishing yukawa and @xmath109-type contributions .
details regarding the vanishing of the other contributions appear in appendix b. following the conventional practice , we regulate the loop integrals using dimensional regularization .
the pole terms proportional to @xmath118 are cancelled by appropriate counterterms .
we identify only the terms nonanalytic in quark masses with the loops .
all other analytic terms are indistinguishable from finite parts of the corresponding counterterms .
the nonvanishing contribution from fig .
1(a ) arises from the insertion of the @xmath119 part of the yukawa interaction of eq .
( [ n2 ] ) .
the nonanalytic term is @xmath120 where @xmath121 and @xmath122 is the subtraction scale introduced in dimensional regularization . for simplicity
, we show here only the contributions for @xmath123 .
the terms for @xmath124 are equal in magnitude and opposite in sign since it is the hermitian conjugate of the @xmath123 piece .
this property holds to all orders of chiral expansion . the nonvanishing contribution from fig .
1 ( b ) arises from strong vertex correction to the leading order @xmath25 yukawa interaction : @xmath125 the terms in figs .
1(c1)-(c2 ) are generated by the pv axial @xmath126 couplings proportional to the @xmath101 .
we have @xmath127 the contribution from @xmath98 to these two diagrams cancels out , leaving only the dependence on @xmath97 .
we note that although this term is propotional to @xmath128 and , thus , nominally suppressed , the coefficient of @xmath97 is fortuitously large ( @xmath129 ) .
the two pion vertex in fig .
1 ( d1)-(d2 ) comes from the chiral connection @xmath130 : @xmath131 the leading contribution involving @xmath61 intermediate states arises from fig .
the corresponding amplitude receives contributions from three different isospin combinations for the @xmath61 intermediate states .
their sum reads @xmath132 \tau^+ \ ; .\ ] ] the corrections generated by the pv @xmath133 vertices are @xmath134 h_a^\delta \tau^+ \ ; , \ ] ] where @xmath135 is defined as @xmath136 summing all the non - vanishing loop contributions yields the following expression for @xmath42 : @xmath137\\ & & + { 2\over 3}{g_{\pi n\delta } \over f_\pi \lambda_\chi^2 } [ ( \delta^2-{3\over 2}m_\pi^2)\delta \ln ( { \mu\over m_\pi})^2 -2 ( \delta^2-m_\pi^2)^{3/ 2 } \ln { \delta + \sqrt{\delta^2-m_\pi^2}\over m_\pi } ] h_a^\delta\end{aligned}\ ] ] the final nonvanishing corrections arise from @xmath70 and @xmath20 wavefunction renormalization .
these corrections , which have been computed previously @xcite , generate deviations from unity of @xmath138 and @xmath41 appearing in the expression for @xmath37 in eq .
( [ eq : hpieff ] ) . in the case of @xmath138 ,
the nonvanishing contributions arise from figs .
1(e1)-(e2 ) and 2(c1)-(c2 ) : @xmath139 % + { 16 } g_{\pi n\delta}^2 { \delta^2-m_\pi^2\over \lambda_\chi^2 } \ \ \ .\end{aligned}\ ] ] the pion s wavefunction renormalization arises from fig .
2 ( k ) @xcite : @xmath140 numercially , the loop contributions to @xmath41 are small compared to those entering @xmath138 .
note the one loop renormalization of @xmath3 from the pv yukawa @xmath25 and @xmath107 vertices is already at the order @xmath141 . an additional loop will introduce a factor of @xmath141 .
loops containing the axial vector @xmath99 and @xmath142 vertices and one strong @xmath2 or @xmath143 vertex are of @xmath144 . to obtain contributions of @xmath145
, one would require the insertion of operators carrying explicit factors of @xmath44 into one loop graphs .
we find no such contributions .
we may estimate the numerical importance of the loop corrections to @xmath39 by taking @xmath146 gev , @xmath147 @xcite and @xmath148 @xcite and by choosing @xmath149 gev is logarithmic , one may choose other values , such as @xmath150 , without affecting the numerical results significantly ] . with these inputs ,
the value of @xmath53 is completely determined .
the vertex corrections , which appear as @xmath42 in eq .
( [ eq : hpieff ] ) , depend on the pv couplings @xmath5 , @xmath97 , @xmath151 , and @xmath135
. we obtain @xmath152 note that the effect of the wavefunction renormalization corrections is to reduce the dependence on @xmath39 by roughly 50% .
in addition , the dependence of @xmath37 on @xmath97 and @xmath151 is non - negligible .
their coefficients are only a factor of two smaller than that of @xmath39 .
although these contributions arise at @xmath153 , they are fortuitously enhanced numerically .
thus , in a complete anaysis of the ope pv interaction one should not ignore these constants . at present
, one has no direct experimental constraints on the parameters @xmath97 , @xmath151 , and @xmath135 , as a comprehensive analysis of hadronic pv data including the full chiral structure of the pv hadronic interaction has yet to be performed .
consequently , one must rely on theoretical input for guidance regarding the scale of the unknown constants .
estimates of @xmath97 are given by the authors of ref .
these authors observe that the usual pole dominance approximation for p - wave non - leptonic hyperon decays typically underpredicts the experimental amplitudes by a factor of two .
the difference may be resolved by the inclusion of local , parity - conserving operators having structures analogous to the @xmath109-type terms in eq .
( [ n2 ] ) .
the requisite size of the @xmath154 contact terms may imply a scale for the analogous @xmath155 pv terms .
if so , one might conclude that @xmath97 should be on the order of @xmath156 .
on the other hand , a simple factorization estimate leads to @xmath157 .
while the sign of @xmath97 is fixed in the factorization approximation , the sign of the larger value is undetermined .
thus , it is reasonable to conclude that @xmath97 may be large enough to significantly impact @xmath37 , though considerably more analysis is needed to yield a firm conclusion .
the @xmath107 yukawa coupling @xmath151 has been estimated in ref .
@xcite using methods similar to those of ref .
the authors quote a best value `` of @xmath158 , with a reasonable range '' of @xmath159 . in ref .
] navely , subsitution of the best value into eq .
( [ eff ] ) would increase the value of @xmath37 , whereas the @xmath29f result would seem to require a reduction .
as we argue below , however , the relationship between the couplings computed in refs . @xcite and the parameters appearing in eq .
( [ eff ] ) is somewhat ambiguous .
direct substitution of the theoretical value into @xmath37 may not be entirely appropriate . to date , no theoretical estimate of the @xmath109-type @xmath133 coupling has been performed . a simple estimate of the scale is readily obtained using the factorization approximation . to that end
, we work with tree - level form of @xmath31 . neglecting short - distance qcd corrections and terms containing strange quarks ,
one has @xmath160 where @xmath161 and @xmath162 denote the third components of the octet of vector and axial vector currents , respectively , and @xmath163 consider now the first term in the expression for @xmath135 given in eq .
( [ eq : hadelta ] ) . in the factorization approximation
, @xmath31 contributes only to the antisymmetric combination @xmath164 the neutral current contribution to this combination , which arises only from the term containing @xmath161 , is @xmath165 where @xmath166 is the axial vector @xmath167 form factor at the photon point .
after fierz re - ordering , the charged current component of @xmath31 contributes roughly @xmath168 yielding a total factorzation contribution of about @xmath169 .
thus , one would expect the scale of the axial vector @xmath133 couplings to be on the order of a few @xmath170 .
in the particular case of the combination appearing in @xmath135 , however , the sum of factorization contributions cancels identically .
as one sees from the expressions for the @xmath171 given in appendix a , isospin requires @xmath172 the factorization contributions independently satisfy this sum rule . the second combination of constants appearing in eq .
( [ eq : hadelta ] ) , @xmath173 also vanishes in the factorization approximation , even though the individual couplings do not .
the third pair of couplings received no factorization contributions .
thus , one has @xmath174 in this approximation . in principle
, non - factorization contributions yield a non - zero value for @xmath135 .
although we have not evaluated these contributions , we do not expect the scale to be significantly larger than the factorization value for the individual @xmath171 couplings .
consequently , we estimate a reasonable range for @xmath135 of @xmath175 . these theoretical estimates suggest considerable ambiguity in the prediction for @xmath37 . in principle , some of this ambiguity might be removed by performing the comprehensive analysis of hadronic pv suggested above , in which the various constants would be determined entirely by experiment .
the viability of such a program remains to be seen .
the results in eqs . ( [ eq : deltahpi]-[eq : piwaveren ] ) embody the full su(2 ) chiral structure at @xmath176 of @xmath52 at leading order in the pion momentum .
any microscopic calculation of this matrix element which respects the symmetries of qcd should display the dependence on light quark masses appearing in @xmath37 . in principle , an unquenched lattice qcd calculation with light quarks would manifest this chiral structure . in practice , however , unquenched calculations remain difficult , and even quenched calculations require the use of heavy quarks . for a lattice determination of @xmath52 , the expressions in eqs .
( [ eq : deltahpi]-[eq : piwaveren ] ) could be used to extrapolate to the light quark limit , much as the chiral structure of baryon mass and magnetic moment can be used for similar extrapolations @xcite . in the absence of a first principles
qcd calculation , one must rely on symmetries and/or models to obtain the pv @xmath2 coupling .
a variety of such approaches have been undertaken , including the su(6)@xmath35/quark model calculation of refs .
@xcite , the skyrme model @xcite , and qcd sum rules @xcite . to date
, the ddh / fcdh analysis remains the most comprehensive and has become the benchmark for comparison between experiment and theory .
consequently , we focus on this work as a case study " in the problem of matching microscopic calculations onto hadronic effective theory .
the ddh / fcdh approach relies heavily on symmetry methods in order to relate the pv @xmath11 matrix elements to experimental @xmath154 nonleptonic hyperon decay amplitudes .
all the charged current ( cc ) contributions to the @xmath177 @xmath178 amplitudes , where @xmath179 is a pseudoscalar meson , can be related using su(3 ) arguments .
likewise , the neutral current ( nc ) component of the effective weak hamiltonian belonging to the same multiplets as the cc components ( _ i.e. _ those arising from a product of purely left - handed currents ) can also be related via su(3 ) .
the remaining nc contributions to the @xmath11 pv amplitudes are computed using factorization and the mit bag model .
the ddh approach also employs su(6)@xmath35 symmetry arguments in order to calculate parity - violating vector meson couplings .
although one requires only su(3 ) to determine the pseudoscalar couplings , we refer below to the general su(6)@xmath35 formalism used in refs .
@xcite .
the general su(6)@xmath35 analysis employed by ddh / fcdh introduces five reduced matrix elements : @xmath180 , @xmath181 , and @xmath182 .
these constants correspond to su(6)@xmath35 components of the weak hamiltonian : @xmath183_{35 } \sim c_v\ ] ] @xmath184_{280,\overline{280 } } \sim b_t , b_v\ ] ] @xmath185_{280,\overline{280 } } \sim a_t , a_v\ ] ] one may represent these different components of @xmath31 diagramatically as in fig .
3 . the components shown in fig . 3a , b correspond to @xmath181 and @xmath182 , respectively . in practice
, these contributions are determined entirely from empirical hyperon decay data .
the term in fig .
3a corresponds to @xmath180 and is computed in refs .
@xcite using factorization .
the pv @xmath2 yukawa coupling can be expressed in terms of these su(6)@xmath35 reduced matrix elements plus an additional factorization / quark model term . temporarily neglecting short - distance qcd corrections to @xmath31
, one has @xmath186 where @xmath187 and @xmath188 are the cabibbo and weinberg angles , respectively , and @xmath189 denotes a fierz / factorization contribution . the first term on the rhs of eq .
( [ eq : su6bare ] ) gives the cc contribution , while the remaining terms arise from weak nc s .
including short - distance qcd renormalization of @xmath31 leads to a modification of eq .
( [ eq : su6bare ] ) : @xmath190\gamma(k ) + \sin^2\theta_c\right\}{\rho\over \sin^2\theta_c } g_\pi \nonumber \\ \label{eq : su6qcd } & & + \sin^2\theta_c ( b_1+b_2)\ \ \ , \end{aligned}\ ] ] where @xmath191 and @xmath192 , @xmath193 and @xmath194 are summed leading log ( renormalization group ) factors dependent on @xmath195 \ln{\mws\over\mu^2 } \ \ \ .\ ] ] the overall scale factor @xmath21 appearing in eq .
( [ eq : su6qcd ] ) was introduced in ref.@xcite in order to account for various theoretical uncertainties entering the analysis .
the appearance of @xmath182 , @xmath196 , and @xmath197 in @xmath198 and @xmath199 relies on _ tree - level _ su(6)@xmath35 symmetry long - distance chiral corrections of the types shown in fig .
4 have not been explicitly included .
inclusion of such corrections would necessitate a reanalysis of the @xmath154 amplitudes in much the same way that one treats the octet of baryon axial vector currents @xcite or magnetic moments @xcite .
for example , letting @xmath200 denote the amplitude for @xmath201 one has at tree - level @xmath202 including the leading chiral corrections would yield the modification @xmath203 where @xmath204 denotes vertex corrections and possible contributions from higher - dimension operators .
similar corrections would appear in the su(6)@xmath35 symmetry terms in eqs .
( [ eq : su6bare ] , [ eq : su6qcd ] ) . given the absence of these corrections from the ddh / fcdh analysis , the symmetry components @xmath205 do not formally embody the subleading chiral structure of @xmath37 .
the _ numerical _ impact of applying chiral corrections to the ddh / fcdh tree - level su(6)@xmath35 analysis is much less clear , since some of the chiral modifications can be absorbed into renormalized values of the chiral couplings , which are determined empirically .
nevertheless , the potentially sizeable effect of the su(2 ) chiral corrections on @xmath37 should give one pause .
a related issue is the degree to which ambiguities introduced by kaon and @xmath206 loops in su(3 ) hbchpt could plague an analysis of the @xmath154 amplitudes
. here recent work by donoghue and holstein argues that finite nucleon size call for long - distance regularization of such heavy meson loops , which substantially reduces their effects@xcite .
results are then similar to what arises from use of a cloudy bag approach to such matrix elements@xcite . a comprehensive study of such issues and their impact on the ddh / fcdh calculation of @xmath5 goes beyond the scope of the present work .
nevertheless , the sizeable impact of the chiral corrections in @xmath37 and the use of tree - level symmetry arguments in refs .
@xcite points to a possibly significant mismatch between @xmath37 and @xmath207 .
the remaining terms in the ddh / fcdh analysis involving the parameters @xmath206 and @xmath189 are determined by explicit mit bag model calculations .
one may ask whether the latter effectively includes any part of the subleading chiral structure of @xmath37 . in order to address this question ,
we make three observations : * 1 .
sea quarks and gluons generate @xmath182*. the parameter @xmath182 vanishes identically in any quark model in which baryons consist solely of three constituent quarks .
the @xmath154 hyperon decay data , however , clearly implies that @xmath208 . in order to obtain a nonzero value in a quark model ,
one requires the presence of sea quarks and gluons .
it is shown in @xcite for example , that @xmath208 when gluons are added to the mit bag model .
similarly , one would expect contributions from the @xmath209 pairs in the sea . since relativistic quark models already contain @xmath209 pairs in the form of z - graphs `` @xcite ,
it is likely that disconnected @xmath209 insertions ( see fig .
5b ) give the dominant sea quark contribution to @xmath182 . in a chirally corrected analysis of nonleptonic decays ,
the long - distance part of the disconnected @xmath209 insertions appear explicitly in the guise of pseudoscalar loops , while the short - distance contributions are subsumed into the value of @xmath182 and possible higher dimension operators .
quenched '' quark models without explicit pionic degress of freedom generally do not include the long - distance physics of disconnected insertions .
the @xmath210-dependence is different*. in conventional hbchpt analyses of hadronic observables , one only retains the loop contributions non - analytic in the light quark mass .
the constituent quark model ( without explicit pions ) has a difficult time producing these non - analytic contributions .
the simplest , illustrative example is the nucleon isovector charge radius , @xmath211 , which is singular in the chiral limit @xcite .
this chiral singularity , of the form @xmath212 , is produced by @xmath20 loops .
relativistic quark models , such as the mit bag model , yield a finite value for @xmath211 as @xmath213 .
one can not produce the chiral singularity in a quark model without including disconnected @xmath209 insertions dressed as mesons .
the corresponding argument in the case of @xmath37 is less direct , but still straightforward . in the limit of a degenerate @xmath70 and @xmath61 , the non - analytic terms in @xmath37 have quark mass - dependences of the form @xmath214 or @xmath128 .
as we show in appendix c , bag model matrix elements of the four quark operators appearing in @xmath31 have a taylor series expansion about @xmath215 .
thus , the parameters @xmath206 and @xmath189 can not contain the non - analytic structures generated by the diagrams in figs . 1 - 2 .
graphs are missing*. this observation is simply a diagrammatic summary of the previous two observations . for simplicity , consider a subset of the quark - level diagrams associated with the appearance of @xmath101 in @xmath37 .
typical contributions to the axial @xmath99 pv vertex are shown in fig .
5a . the corresponding loop contributions to @xmath37 appear in fig .
those in fig .
5b involve disconnected @xmath209 insertions , which do not occur in the constituent quark model . the contribution of fig .
5c involves z - graphs , which are produced in a relativistic quark model terms of fig .
3b ] . in principle
, the @xmath216 intermediate state could contain an @xmath217 pair .
as argued previously , however , the z - graphs implicit in the mit bag model calculation of @xmath5 do not produce the nonanalytic structure of the corresponding @xmath20 loop .
apparently , only an unquenched quark model , which generates the disconnected insertions of fig .
5b , could produce the requisite nonanalytic terms . from this case study `` of the ddh / fcdh calculation of @xmath5 , we conclude that the su(6)@xmath35/quark model approach used in refs .
@xcite does not incorporate the chiral structure of @xmath37 . were the numerical impact of the chiral corrections negligible
, this observation would not be bothersome .
the actual impact of the chiral corrections , however , is significant .
thus , it is perhaps not surprising that the @xmath29f result and the ddh / fcdh best value '' do not agree .
with the confirmation of the electroweak sector of the standard model at the 1% level or better in a variety of leptonic and semi - leptonic processes , one has little reason to doubt its validity in the purely hadronic domain .
similarly , the predictions of qcd in the perturbative regime have been confirmed with a high degree of confidence .
thus , one may justifiably consider @xmath31 , the effective hamiltonian including its perturbative strong interaction correction , to be well understood .
moreover , the precision available with present and future hadronic pv experiments is unlikely to match the levels achieved in leptonic and semileptonic processes .
consequently , one has little hope of detecting small deviations in @xmath31 from its sm structure due to new physics " . on the other hand ,
much about qcd in the non - perturbative regime remains mysterious : the mechanism of confinement , the dynamics of chiral symmetry breaking , the role or sea quarks in the low - energy structure of the nucleon , and so forth .
each of these issues bears on one s understanding of _ matrix elements _ of @xmath31 . in this sense
, the low - energy , pv hadronic weak interaction constitutes a probe of the dynamics of low - energy qcd , in a manner analogous to the probe provided by the electromagnetic interaction . from a phenomenological standpoint , the matrix element
one may hope to extract from hadronic pv observables with the least ambiguity is @xmath52 . in this study , we have argued that any theoretical interpretation of this matrix element must take into account the consequences of chiral symmetry .
indeed the chiral corrections to the tree - level , pv @xmath25 yukawa coupling are not small . at @xmath54 , the effective coupling measured in experiments , @xmath37 ,
depends not only on the bare coupling , @xmath39 , but also on new ( and experimentally undetermined ) pv low - energy constants , @xmath97 , @xmath135 , and @xmath151 , as well .
furthermore , the coefficients of @xmath39 , @xmath97 , and @xmath151 are comparable in magnitude . at present
, one has only simple theoretical estimates of the magnitudes of the @xmath97 and @xmath135 in addition to the fcdh calculation of @xmath151 .
these estimates suggest that the new pv couplings appearing in @xmath37 could be as large as @xmath39 .
since no experimental constraints have been obtained for the new couplings , there exists considerable latitude in the theoretical expectation for @xmath37 . for two decades
now , the benchmark theoretical calculation of @xmath52 has been the su(6)@xmath35/quark model approach of ref .
@xcite , updated in ref .
we have argued , however , that the ddh / fcdh calculation does not manifest the general strictures of chiral invariance obtained in the present analysis . at the quark level , this chiral structure reflects the importance of the disconnected `` @xmath209 components of the sea .
while relativistic quark models contain @xmath209 sea quark effects in the guise of z - graphs or lower - component wavefunctions , the most common quenched '' versions do not include explicit disconnected pairs pairs may , however , hide in the effective parameters of the quark model , such as the string tension @xcite . ] .
given the size of the chiral corrections associated in part with the disconnected insertions , it may then be not so surprising to find a possible discrepancy between the experimental value for @xmath37 and the ddh / fcdh best value " .
applying chiral corrections to the su(3 ) analysis of @xmath154 hyperon decays may help to close the gap between @xmath37 and @xmath207 .
presumably , similar corrections should be applied in other approaches not containing explicit pionic degress of freedom . in the longer run
, one may be able to use the chiral structure of @xmath37 to extrapolate an unquenched lattice calculation with heavy quarks into the physical regime .
it is a pleasure to thank j.l .
goity and n. isgur for useful discussions .
this work was supported in part under u.s .
department of energy contract # de - ac05 - 84er40150 , the national science foundation , and a national science foundation young investigator award
. 99 w. haeberli and b.r .
holstein in _ symmetries and fundamental interactions in nuclei _ ,
haxton and e.m .
henley , eds .
, world scientific , singapore , 1995 , p. 17 .
e. g. adelberger and w. c. haxton , ann .
nucl . part .
35 , 501 ( 1985 ) .
snow et al . , nucl .
a440 * , 729 ( 2000 ) .
jlab loi 00 - 002 , w. van oers and b. wojtsekhowski , spokesmen . c. s. wood et al . ,
science 275 , 1759 ( 1997 ) .
d. tadic , phys . rev . * 174 * , 1694 ( 1968 ) ; b.h.j .
mckellar , phys .
b26 * , 107 ( 1967 ) ; e. fischbach , phys . rev . *
170 * , 1398 ( 1970 ) ; w. kummer and m. schweda , acta phys
* 28 * , 303 ( 1968 ) .
b. desplanques , j. f. donoghue and b. r. holstein , ann .
124 , 449 ( 1980 ) .
g. b. feldman et al .
c 43 , 863 ( 1991 ) .
n. kaiser and ulf .- g .
meissner , nucl .
phys . a 499 , 699 ( 1989 ) ; _ ibid . _ a 510 , 759 ( 1990 ) ; ulf .- g .
meissner and h. weigel , phys .
b 447 , 1 ( 1999 ) .
e. m. henley , w .- y .
p. hwang and l. s. kisslinger , phys .
b 367 , 21 ( 1996 ) ; _ erratum - ibid . _
b 440 , 449 ( 1998 ) .
e. jenkins and a. v. manohar , phys .
b 255 , 558 ( 1991 ) ; b 259 , 353 ( 1991 ) .
meissner , int .
e 1 , 561 ( 1992 ) .
t. r. hemmert , b. r. holstein and j. kambor , j. phys .
g 24 , 1831 ( 1998 ) .
shi - lin zhu , s. puglia , m. j. ramsey - musolf , and b. holstein , [ hep - ph/0002252 ] to appear in phys .
d. d. b. kaplan and m. j. savage , nucl .
a 556 , 653 ( 1993 ) .
n. fettes , u .- g .
meissner , s. steininger , nucl .
phys . a 640 , 199 ( 1998 ) . v. bernard et al .
phys . a 635 , 121 ( 1998 ) .
j. f. donoghue , e. golowich and b. r. holstein , _ dynamics of standard model _ , cambridge university press , 1992 , new york .
e. caso et al . , particle data group , euro .
j. c 3 , 1 ( 1998 ) .
a. w. thomas et al . ,
a 663 , 973 ( 2000 ) ; e. j. hackett - jones , d. b. leinweber and a. w. thomas , hep - lat/0004006 .
e. jenkins et al . , phys . lett .
b 302 , 482 ( 1993 ) ; b 388 , 86 ( 1996 ) [ e ] ; ulf - g . meissner and s. steininger , nucl . phys .
b 499 , 349 ( 1997 ) ; l. durand and p. ha , phys . rev .
d 58 , 013010 ( 1998 ) ; s. puglia and m. j. ramsey - musolf , hep - ph/9911542 , phys .
d ( in press ) . j.f .
donoghue and b.r .
holstein , phys .
b436 * , 331 ( 1998 ) ; j.f .
donoghue , b.r .
holstein , and b. borasoy , phys .
rev . p**d59 * * , 036002 ( 1999 ) .
t. yamaguchi et al .
a500 * , 429 ( 1989 ) ; k. kubodera et al .
* a439 * , 695 ( 1985 ) ; s. theberge et al .
, phys . rev .
* d22 * , 2838 ( 1980 ) ; a.w .
thomas , j. phys .
* g7 * , l283 ( 1981 ) .
j. f. donoghue and e. golowich , phys .
b 69 , 437 ( 1977 ) . p. geiger and n. isgur , phys .
d 55 , 299 ( 1997 ) .
m. a. b. beg and a. zapeda , phys .
d 6 , 2912 ( 1972 ) ; j. gasser , m. e. sainio and a. svarc , nucl .
b 307 , 779 ( 1988 ) . p. geiger and n. isgur , phys . rev .
d41 , 1595 ( 1990 ) .
donoghue , e. golowich , and b. holstein , phys .
131 , 319 ( 1986 ) .
here we present the full expressions for some of the pv lagrangians not included in the main body of the paper .
the analogues of eqs .
( [ n1]-[n3 ] ) are @xmath218 t_c^\mu \\ & & + g_1 \bar n [ a_\mu , x_-^a]_+ t^\mu_a+g_2 \bar n [ a_\mu , x_-^a]_- t^\mu_a + { \hbox{h.c.}}\\ \nonumber & & \\
\label{d2 } \nonumber { \cal l}^{\pi\delta n}_{\delta i=1 } & = & f_2 \epsilon^{ab3 } \bar n
i\gamma_5 [ a_\mu , x_+^a]_+ t^\mu_b + f_3 \epsilon^{ab3}\bar n i\gamma_5[a_\mu , x_+^a]_- t^\mu_b \\ \nonumber & & + { g_3\over 2}\bar n [ ( x_l^a a_\mu x_l^3-x_l^3 a_\mu x_l^a)- ( x_r^a a_\mu x_r^3-x_r^3 a_\mu x_r^a ) ] t^\mu_a\\ \nonumber & & + { g_4\over 2 } \{\bar n [ 3x_l^3 a^\mu ( x_l^1 t^1_\mu + x_l^2 t^2_\mu ) + 3(x_l^1 a^\mu x_l^3 t^1_\mu + x^2_l a^\mu x^3_l t^2_\mu ) \\ & & -2 ( x_l^1 a^\mu x_l^1 + x_l^2 a^\mu x_l^2 - 2x_l^3 a^\mu x_l^3)t^3_\mu ]
-(l\leftrightarrow r ) \ } + { \hbox{h.c . } } \\ \nonumber & & \\ \label{d3 } \nonumber { \cal l}^{\pi\delta n}_{\delta i=2 } & = & f_4 \epsilon^{abd } { \cal i}^{cd}\bar n i\gamma_5 [ x_l^a a_\mu x_l^b + x_r^a a_\mu x_r^b]t^\mu_c \\ \nonumber & & + f_5 \epsilon^{ab3 } \bar n i\gamma_5 [ x_l^a a_\mu x_l^3+x_l^3 a_\mu x_l^a + ( l\leftrightarrow r)]t^\mu_b \\ & & + g_5 { \cal i}^{ab}\bar n [ a_\mu , x_-^a]_+ t^\mu_b + g_6 { \cal i}^{ab}\bar n [ a_\mu , x_-^a]_- t^\mu_b + { \hbox{h.c.}}\ \ \ , \end{aligned}\ ] ] where the terms containing @xmath219 and @xmath220 start off with one- and two - pion vertices , respectively . in the heavy baryon expansion , the terms containing the @xmath219 start to contribute at @xmath221 .
the leading order term vanishes since @xmath222 .
since we work only to lowest order in the @xmath56 expansion , we obtain no contribution from the terms containing the @xmath219 . for the pv @xmath223
effective lagrangians we have @xmath224 @xmath225 -2(\bar t^1 \gamma^\mu t^1 + \bar t^2 \gamma^\mu t^2 -2\bar t^3 \gamma^\mu t^3 ) tr ( a_\mu x_+^3 ) \ } & \\ \nonumber + k_2 \ { 3[(\bar t^3 \gamma^\mu\gamma_5 t^1 + \bar t^1 \gamma^\mu\gamma_5 t^3 ) tr ( a_\mu x_-^1 ) + ( \bar t^3 \gamma^\mu\gamma_5 t^2 + \bar t^2 \gamma^\mu\gamma_5 t^3 ) tr ( a_\mu x_-^2 ) ] & \\ \nonumber -2(\bar t^1 \gamma^\mu\gamma_5 t^1 + \bar t^2 \gamma^\mu\gamma_5 t^2 -2\bar t^3 \gamma^\mu\gamma_5 t^3 ) tr ( a_\mu x_-^3 ) \ } & \\ \nonumber + j_3 \ { \bar t^a \gamma^\mu [ a_\mu , x_+^a]_+ t^3 + \bar t^3 \gamma^\mu [ a_\mu , x_+^a]_+ t^a \ } & \\ \nonumber + j_4 \ { \bar t^a \gamma^\mu [ a_\mu , x_+^a]_- t^3 - \bar t^3 \gamma^\mu [ a_\mu , x_+^a]_- t^a \ } & \\ \nonumber + k_3 \ { \bar t^a \gamma^\mu\gamma_5 [ a_\mu , x_-^a]_+ t^3 + \bar t^3 \gamma^\mu\gamma_5 [ a_\mu , x_+^a]_+ t^a \ } &
\\ + k_4 \ { \bar t^a \gamma^\mu\gamma_5 [ a_\mu , x_-^a]_- t^3 - \bar t^3 \gamma^\mu\gamma_5 [ a_\mu , x_+^a]_- t^a \}&\ ; , \end{aligned}\ ] ] @xmath226\gamma^\mu t_i & \\
\nonumber + k_5 { \cal i}^{ab } \bar t^i [ x_r^a a_\mu x_r^b -x_l^a a_\mu x_l^b]\gamma^\mu\gamma_5 t_i & \\ + k_6 \epsilon^{ab3 } [ \bar t^3 i\gamma_5 x_+^b t^a + \bar t^a i\gamma_5 x_+^b t^3 ] & \ ; , \end{aligned}\ ] ] where we have suppressed the lorentz indices of the @xmath61 field , i.e. , @xmath227 .
the vertices with @xmath228 contain two pions .
all other vertices contain one pion when expanded to the leading order . at first sight the leading order term with @xmath229 in ( [ ddd3 ] )
has no pions .
however such a term cancels its hermitian conjugate exactly . the constants @xmath230 are the pv @xmath231 yukawa coupling constants . in section 2 ,
the leading terms in the above lagrangians were expressed in terms of effective @xmath232 and @xmath107 coupling constants . these constants may be expressed in terms of the @xmath219 , @xmath220 , @xmath233 , @xmath234 and @xmath235 as follows : @xmath236 @xmath237 it is interesting to note there is only one independent pv yukawa coupling constant @xmath151 for @xmath223 interactions .
as noted in section 3 , a large number of graphs which nominally contribute to @xmath37 actually vanish up to @xmath54 . here , we summarize the the reasons why .
consider first the corrections due to the pv vector @xmath25 vertices . for fig .
1 ( b ) we have @xmath238 where we have used @xmath239 .
since we are working to leading order in the @xmath56 expansion , this amplitude does not contribute .
the pv vector interactions also appear in figs .
1(j1,j2 ) .
the corresponding amplitude is @xmath240_+\over v\cdot k v\cdot ( k+q ) ( k^2-m_\pi^2 ) } = 0 \;\ ] ] this integral vanishes because it is proportional to @xmath241_+$ ] , which vanishes because @xmath242 .
all other possible insertions of pv vector @xmath25 vertices vanish for similar reasons as either ( [ m_b1 ] ) or ( [ m_j ] ) . in what follows
, we refer only to insertions involving the pv @xmath25 yukawa and @xmath126 axial couplings .
the propagator corrections in fig . 1 ( g1)-(h2 ) vanish after integration since their amplitude of ( g1)-g2 ) goes as @xmath243 while the amplitude of ( h1)-(h2 ) goes as @xmath244 the amplitude of fig .
1 ( i1)-(i4 ) contains a vanishing integral @xmath245 figs . 1 ( j1)-(j2 )
do not contribute for the pv yukawa coupling @xmath3 due to charge conservation .
the remaining non - zero diagrams are figs . 1 ( a)-(f2 ) where the insertions in loops are of the yukawa or axial interations
( f1 ) , ( f2 ) arises from the insertion of the counter terms of mass and wave function renormalization .
fig . 1 ( e1)-(e2 ) and fig .
2 ( c1)-(c2 ) contribute to the wave function renormalization in eq .
( [ z - n ] ) . due to
the heavy baryon projection @xmath246 the one pion pv @xmath247 vertex does not contribute in the leading order of heavy baryon expansion .
hence , the chiral loop corrections from fig . 2 ( d1)-(g4 ) are of higher order .
2 ( h1)-(j2 ) vanishes after integration for reasons similar to ( [ m_j ] ) .
the remaining , non - vanishing diagrams are discussed explicitly in section 3 .
as pointed out in section 2 , both pc and pv two - derivative operators which conserve cp do not contribute to @xmath3 renormalization .
for example , there exists one cp - conserving , pv such operator : @xmath248 n\ \ \ .\ ] ] after expansion , the leading term starts with three pions .
it contributes via figure 1 ( a ) , at the order of @xmath249 .
moreover the loop integration yields a factor @xmath250 and leads to zero after contraction with @xmath251 .
another possibility comes from insertions of pc two - derivative nucleon pion operators .
there are three pc operators which conserve cp : @xmath115 @xmath116 @xmath252 n\ ; .\ ] ] note the first two operators are symmetric in the lorentz indices .
only the last one arises from the antisymmetric operators listed in eq .
( [ eq : twoderivpc ] ) .
the first one starts off with one pion .
the relevant feynman diagrams are figure 1 ( c1)-(c2 ) , where the pv vertex is associated with @xmath101 .
note these diagrams do not contribute at leading order of hbchpt due to the presence of the @xmath253 .
the remaining two operators start off with two pions .
the relevant diagrams are figure 1 ( d1)-(d2 ) .
after integration the contribution of the third operator reads @xmath254 so its contribution is zero .
in contrast the second operator yields @xmath255 note @xmath256 .
so its contribution is of order @xmath257 . in short ,
none of the two - derivative operators contribute to the renormalization of @xmath3 at the order to which we work .
here , we show that the four - quark bag model integrals relevant to the calculation of the ddh / fcdh parameters @xmath206 and @xmath189 have a taylor expansion in light quark mass around @xmath215 . we write a bag model quark wavefunction as @xcite @xmath258 where @xmath259 denotes a two - component pauli spinor and where wave function normalization yields @xmath260 where the the radial integration runs from 0 to the bag radius , @xmath261 .
the four quark matrix elements of interest here can depend three different integrals : @xmath262 the quark radial wave functions are @xmath263 where @xmath264 @xmath265 @xmath266 @xmath267 b is the bag constant and @xmath268 is a phenomenological parameter involved with the center of mass motion of the bag .
it is straightforward to show that the bag model integrals in eq .
( [ eq : bagint ] ) have a taylor expansion about @xmath215 .
the argument proceeds by noting that the quantities @xmath70 , @xmath261 , @xmath270 , @xmath271 and the argument of the spherical bessel functions all have taylor series in @xmath210 about @xmath215 .
the existence of this expansion can be seen be an explicit , iterative construction .
first , expand @xmath271 and @xmath261 : @xmath272 now let @xmath215 in eqs .
( [ eq : tanp],[eq : pndef ] ) . doing so eliminates all dependence on @xmath261 and determines @xmath273 .
next , set @xmath215 in eqs .
( [ eq : n ] , [ eq : r ] ) with @xmath274 .
doing so determines @xmath275 .
now expand eqs .
( [ eq : tanp],[eq : pndef ] ) to first order in @xmath210 .
this step yields @xmath276 in terms of @xmath273 and @xmath275 .
expanding eqs .
( [ eq : n ] , [ eq : r ] ) to first order in @xmath210 then determines @xmath277 in terms of @xmath273 , @xmath276 , and @xmath275 and so forth . note that at any step of the recursion , the argument of any transcendental function is @xmath273 .
hence , at any order , a solution for the @xmath278 and @xmath279 exists .
the expansion of the bag model integrals continues by computing their derivatives with respect to @xmath210 and using the expansions of @xmath70 , @xmath261 , etc . in terms of @xmath210 as constructed above .
taking @xmath280 derivatives of one of the integrals in eq .
( [ eq : bagint ] ) yields new intregrals involving powers of @xmath281 times products of the bessel functions and their derivatives . using the standard bessel function recursion relations ,
the derivatives of the @xmath282 can always be expressed in terms of other spherical bessel functions . since the @xmath282 and their derivatives are finite at the origin , and since the radial bag integration is bounded above by @xmath261 , the @xmath280th derivative of any of the integrals in eq .
( [ eq : bagint ] ) is finite .
thus , each of the integrals in eq .
( [ eq : bagint ] ) can be expanded in a taylor series about @xmath215 .
meson - nucleon intermediate state contributions to the pv @xmath25 vertex @xmath3 .
the shaded circle denotes the pv vertex .
the solid and dashed lines correspond to the nucleon and pion respectively .
figure 3.diagrammatic representation of the su(6)@xmath35 components of @xmath283 . figs .
3a - c correspond , respectively , to @xmath181 , @xmath182 , and @xmath180 .
the wavy line denotes the action of @xmath31 .
quark line diagrams for the renormalization of @xmath3 due to the axial pv @xmath126 interaction . as in fig .
3 , the wavey line denotes the action of @xmath31 . fig .
5a shows a typical contribution to @xmath101 .
figs 5b , c denote the corresponding loop corrections to @xmath3 .
5(b ) contains the disconnected @xmath209 insertions , while fig .
5(c ) gives a z - graph contribution . | we construct the complete @xmath0 parity - violating ( pv ) @xmath1 interaction lagrangian with one derivative , and calculate the chiral corrections to the pv yukawa @xmath2 coupling constant @xmath3 through @xmath4 in the leading order of heavy baryon expansion .
we discuss the relationship between the renormalized @xmath5 , the measured value of @xmath5 , and the corresponding quantity calculated microscopically from the standard model four - quark pv interaction . 0.5 true cm pacs indices : 21.30.+y , 13.40.ks , 13.88.+e , 11.30.er = -1 cm = -0.7 cm = -0.7 cm # 1#1 # 1#1 # 1#2 # 1@xmath6 # 1 # 1#1 # 1^#1 # 1#1 # 1#1 # 1#1 # 1_phys .
rev . _ * # 1 * # 1_phys .
rev . _
* c#1 * # 1_phys .
rev .
_ * d#1 * # 1_phys .
rev .
lett . _ * # 1 * # 1_nucl .
phys . _ * a#1 * # 1_nucl .
phys . _ * b#1 * # 1_ann . of phys . _ * # 1 * # 1_phys
. reports _ * # 1 * # 1_phys
. lett . _ * b#1 * # 1_z .
fr phys .
_ * a#1 * # 1_z .
fr phys .
_ * c#1 * 0q_^0 shi - lin zhu@xmath7 , s.j .
puglia@xmath7 , b.r .
holstein@xmath8 , m. j. ramsey - musolf@xmath9 @xmath7 department of physics , university of connecticut , storrs , ct 06269 usa + @xmath10 theory group , thomas jefferson national accelerator facility , newport news , va 23606 usa + @xmath8 department of physics and astronomy , university of massachusetts , amherst , ma 01003 usa |
it has often been considered that the narrow - line regions ( nlrs ) in active galactic nuclei ( agns ) are photoionized by the radiation from central engines . however , this photoionization scenario has sometimes been conflicted with several serious problems .
one of such problems is that any single - zone photoionization models underpredict the [ o iii]@xmath04363/[o iii]@xmath05007 flux ratio , @xmath1 ( e.g. , simpson et al .
1996 ) . to solve this problem ,
some models have been proposed ; e.g. , two component models with high - density gas clouds ( e.g. , filippenko 1985 ) or those with shock - heated regions ( e.g. , dopita & sutherland 1995 ) .
in addition , some previous studies reported that type 1 and type 1.5 seyfert nuclei ( s1s and s1.5s , respectively ) tend to have higher @xmath1 values than type 2 seyfert nuclei ( s2s ) ( e.g. , heckman & balick 1979 ) . in order to know where and how the [ o iii]@xmath04363 emission is radiated , and in order to solve the underprediction problem of @xmath1 , it seems critically important to investigate why s1s show higher @xmath1 than s2s
therefore , we examine how the observed value of @xmath1 depends on the agn type based on a large sample of agns compiled from the literature .
then , we compare @xmath1 with various parameters to know the nature of the [ o iii]@xmath04363 emitting regions in agns ( see nagao , murayama , & taniguchi 2001 for more details ) .
in order to investigate statistical properties of @xmath1 , we compiled @xmath1 of 82 s1s , 54 s1.5s , and 78 s2s , from the literature .
the histograms of @xmath1 for each seyfert type are shown in figure 1 .
it is clearly shown that the s2s exhibit lower @xmath1 than the s1s and the s1.5s . according to the kolmogorov - smirnov statistical test
, the probability which the distribution of @xmath1 of s2s and that of s1s or s1.5s come from the same underlying population is less than 10@xmath2 .
to explore the origin of this agn - type dependence of @xmath1 and the nature of the [ o iii]@xmath04363 emitting regions , we compare @xmath1 with various parameters .
the results are as follows : * _ the higher-@xmath1 objects show the hotter mid - infrared ( mir ) colors . _
the hotter mir colors are thought to be attributed to the hotter dusty grains located at the inner surface of the dusty tori , which can be seen if we see the torus from a favored viewing angle ( i.e. , a nearly face - on view ) .
therefore , this means that the higher-@xmath3 objects are seen from a more face - on view toward dusty tori than the lower-@xmath1 objects . * _ the higher-@xmath1 objects show the stronger [ fe vii]@xmath06087 and [ fe x]@xmath06374 emission .
_ since a large fraction of high - ionization emission lines is thought to arise from dense gas clouds obscured by the dusty tori ( e.g. , murayama & taniguchi 1998a ) , the higher-@xmath1 can be attributed to the significant flux contribution from such a high - density gas cloud . * _ the higher-@xmath1 objects tend to show the larger fwhm([o iii]@xmath04363)/ fwhm([o iii]@xmath05007 ) ratios .
_ this also suggests that the [ o iii]@xmath04363 emitting regions are located at inner regions compared to the [ o iii]@xmath05007 emitting regions . * _ the s1s have wider fwhm([o iii]@xmath04363 ) and larger ratio of fwhm([o iii]@xmath04363)/fwhm([o iii]@xmath05007 ) than the s2s .
_ this suggests that the [ o iii]@xmath04363 emitting regions are located at inner regions compared to the [ o iii]@xmath05007 emitting regions and have an anisotropic property .
all of these facts can be consistently understood when we introduce the high - density gas clouds located close to the nucleus , which emit a large fraction of the [ o iii]@xmath04363 emission . since these clouds may suffer significantly from the obscuration by dusty tori in s2s
, we conclude that the agn - type dependence of @xmath1 is attributed to the viewing - angle dependence of the visibility of the high - density gas clouds .
to examine whether or not the viewing - angle dependence of the visibility of the high - density gas can account for the observed difference in @xmath1 between s1s and s2s in the framework of the photoionization scheme , we carry out dual - component photoionization model calculations following the manner of murayama & taniguchi ( 1998b ) .
this method takes account of such high - density gas clouds as a strong [ o iii]@xmath04363 emitter , in addition to the typical nlr component .
we perform model calculations using the spectral synthesis code @xmath4 version 90.04 ( ferland 1996 ) .
we assume uniform density gas clouds with a plane - parallel geometry .
the dense component ( dc ) is assumed to be truncated clouds to avoid unusually strong [ o i ] emission .
here we assume @xmath5 @xmath6 .
we perform several model runs covering @xmath7 @xmath8 @xmath6 .
the ionization parameter of the nlr component is assumed as @xmath9 = 10@xmath10 .
the ionization parameter and the hydrogen column density of dc are determined by the following two conditions : ( [ fe x]/[fe vii])@xmath11 = 0.8 and ( [ fe vii]/[o iii]@xmath05007)@xmath11 = 1.0 .
the former ratio is the typical value of seyfert galaxies ( nagao , taniguchi , & murayama 2000 ) , and the latter condition is introduced by murayama & taniguchi ( 1998b ) . as a result , @xmath12 and
@xmath13 cm@xmath10 are adopted .
the calculations are stopped when the gas temperature falls to 4000k for the nlr component .
we set the gas - phase element abundances to be solar ones .
we adopt the power - law continuum as @xmath15 = 1.5 between 10 @xmath16 m and 50 kev in the form @xmath17 .
the spectral index is set to @xmath15 = 2.5 at lower energy ( i.e. , @xmath18 m ) and to @xmath15 = 2 at higher energy ( i.e. , @xmath19 50 kev ) .
the fraction of dc to the nlr component is treated as a free parameter in our calculations .
we present our model calculations and compare them with the observations in a diagram of @xmath1 versus [ fe vii]/[o iii]@xmath05007 ( figure 2 ) .
we find that the model grids are roughly consistent with the observations if we take the effects of the correction for the extinction into account .
though the dispersion of observation appears to be larger than the model grids , this may be attributed to the fact that the parameters , e.g. , @xmath9 and @xmath20 , are different from object to object .
it is shown that the @xmath1 of the s1s can be explained by introducing a 5% 20% contribution from dc while the @xmath1 of the s2s can be explained by introducing a 0% 2% contribution from dc . through the dual - component photoionization model calculations ,
we conclude that ( 1 ) the origin of the agn - type dependence of @xmath1 is the viewing - angle dependence ( i.e. , the agn - type dependence ) of the visibility of the dense gas clouds which can be hidden by dusty tori , and ( 2 ) the underprediction problem of @xmath1 is solved taking such high - density gas clouds into account .
dopita , m. a. , & sutherland , r. s. 1995 , , 455 , 468 ferland , g. j. 1996 , hazy : a brief introduction to cloudy ( lexington : univ .
kentucky dept .
phys . astron . )
filippenko , a. v. 1985 , , 289 , 475 heckman , t. m. , & balick , b. 1979 , , 79 , 350 murayama , t. , & taniguchi , y. 1998a , , 497 , l9 murayama , t. , & taniguchi , y. 1998b , , 501 , l115 nagao , t. , murayama , t. , & taniguchi , y. 2001 , , 549 , 155 nagao , t. , taniguchi , y. , & murayama , t. 2000 , , 119 , 2605 simpson , c. , ward , m. , clements , d. l. , & rawlings , s. 1996 , , 281 , 509 | the emission - line flux ratio of [ o iii]@xmath04363/[o iii]@xmath05007 is a useful diagnostic for the ionization mechanism and physical properties of narrow - line regions in active galactic nuclei .
however , it is known that simple photoionization models underpredict this ratio . in this contribution
, we report on some pieces of evidence that a large fraction of the [ o iii]@xmath04363 emission arises from the dense gas , which can be obscured by dusty tori . taking this dense component into account
, we show that the flux ratio of [ o iii]@xmath04363/[o iii]@xmath05007 can be explained by two - component photoionization models .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in |
since the discovery of high - tc superconductivity , @xcite there has been a growing interest in the properties of correlated two - dimensional ( 2d ) electronic systems .
@xcite within this context , the hubbard model @xcite has received a lot of attention since it is considered one of the simplest models still containing the relevant physics .
@xcite renewed interest in the hubbard hamiltonian also comes from recent experiments @xcite with cold fermionic atoms in optical lattices which open the possibility for direct simulations of the model with lattice emulators .
@xcite hubbard - like models are also relevant to describe electronic properties within the active research field of graphene .
@xcite the repulsive hubbard hamiltonian is a very interesting model in theoretical physics . on the one hand , neither its hopping ( one - body ) nor its on - site interaction ( two - body ) terms favor any interesting magnetic ordering . on the other hand ,
when both of them combine into the full hamiltonian a rich variety of interesting phenomena is displayed , for example , correlation - driven metal - insulator transitions , @xcite ferromagnetism , @xcite deviations from the standard fermi - liquid results , @xcite long - wavelength collective modes @xcite and spatially inhomogeneous phases .
@xcite the dimensionality of the model also challenges the theoretical tools at our disposal .
exact analytical solutions exist in the one - dimensional ( 1d ) case @xcite whereas the present knowledge of the basic quantum mechanical properties of the 2d hubbard hamiltonian relies , to a large extent , on numerical techniques applied to the hamiltonian itself or to its strong coupling approximations , i.e. , the t - j , t - j@xmath4 and heisenberg models .
@xcite in particular , for the case of the full 2d hubbard hamiltonian , a very efficient lanczos algorithm , @xcite based on the classification of all the irreducible representations of the space group , has allowed systematic studies in the @xmath1 lattice .
going beyond the present limits of exact diagonalization ( ed ) techniques requires a truncation strategy .
a key issue is then how to truncate the model space while still being able to retain the most important degrees of freedom relevant for the description of a particular ground and/or excited state .
nowadays there are several methods at our disposal , some of them already heavily used to study 1d and 2d hubbard models with variable degree of success .
one that has been used with great success is the quantum monte carlo @xcite ( qmc ) approach .
another is the density matrix renormalization group @xcite ( dmrg ) scheme that represents a very powerful and general decimation prescription . currently , the dmrg algorithm is understood as an energy minimization within a class of low entanglement wavefunctions known as matrix product states @xcite ( mps ) establishing an exciting link with quantum information perpectives . @xcite a very flexible entanglement encoding is also provided by the rapidly expanding research area of tensor network states @xcite ( tns ) .
variational principles also offer very powerful methods to study hubbard - like models .
for example , the dynamical variational principle , @xcite expressed in the language of green s functions and self - energies , @xcite provides us with the variational cluster approximation @xcite ( vca ) , the dynamical impurity approximation @xcite ( dia ) and the dynamical mean field theory @xcite ( dmft ) . within this context
, the self - energy - functional theory @xcite ( sft ) has emerged as a conceptual framework in which the vca , dia and dmft , as well as several extensions of them , can be specified by the choice of a reference system . in particular , the cluster extensions to dmft have provided important insights into the physics of the 2d hubbard model in aspects such as the mott - hubbard transition , the pseudogap in doped systems , and the phase diagram itself.@xcite dmft and its cluster extensions are particularly valuable as they have been shown to be complementary to finite size simulations , @xcite including ours . here , we also refer the interested reader to recent work @xcite where a hierarchy of truncated configuration interaction ( ci ) expansions has been considered as a solver for quantum impurity models and dmft . in the present work ,
we explore an alternative avenue not only to describe ground state properties of the 2d hubbard model but also to access excitation spectra which represent a basic fingerprint of quantum mechanical correlations in the considered lattices . a first step in this direction , based on symmetry - projected configuration mixing ideas originally employed in microscopic nuclear structure theory ,
@xcite was undertaken for the 1d hubbard model @xcite and is extended in the present work to the 2d hubbard hamiltonian with periodic boundary conditions ( pbc ) . for a given single - electron space
, we construct the most general unitary hartree - fock ( hf ) transformation .
@xcite since this hf - transformation mixes all the spin and linear momentum quantum numbers of the single - electron basis states , the corresponding slater determinant deliberately breaks the original spin and translational symmetries of the 2d hubbard hamiltonian .
therefore , as such , our symmetry - broken slater determinant can be considered as a convenient mean - field starting point enlarging the space of trial wave functions .
@xcite we restore the broken translational and spin symmetries with the help of linear and angular momentum projection operators .
this symmetry restoration recovers the multi - determinantal character in our trial state keeping good spin and linear momentum quantum numbers .
the ritz variational principle @xcite is then applied to the projected energy , i.e. , ours is a variation - after - projection ( vap ) scheme .
this procedure provides us with the optimal ( variational ) representation of a ground state , with well defined spin and linear momentum quantum numbers , via a single symmetry - projected configuration .
our vap scheme is also very close in spirit to projected quasiparticle theory @xcite ( pqt ) and is related to other variational approaches.@xcite in order to describe excited states with well defined quantum numbers , we construct a truncated basis consisting of a few ( orthonormalized ) symmetry - projected states throughout a chain of vap calculations .
this can be easily done , still with low computational cost , due to the simple structure of our projected wave functions .
finally , a further diagonalization of the 2d hubbard hamiltonian is performed within such a basis . with this configuration mixing procedure
we may account , in a similar fashion , for additional correlations in both ground and excited states .
in addition , our theoretical framework can be used to study important dynamical properties of the 2d hubbard hamiltonian like spectral functions .
@xcite in this paper we have three main goals .
first , we present the methodology of a vap configuration mixing scheme , originally devised for the nuclear many - body problem , but not yet explored for the 2d hubbard model . therefore , in sec . [ theory ] we introduce our theoretical formalism .
symmetry restoration is described in sec .
[ formalism - h2d ] while our configuration mixing scheme is outlined in sec .
[ formalism - h2d - excited ] . for the reader
s convenience , the key ingredients of our approximations are stressed in these two sections while , to make our presentation self - contained , more technical details can be found in appendices [ app-1 ] and [ app-2 ] , respectively .
our second goal is to show how our theoretical framework can be used to access the spectral weight of states with different linear momentum quantum numbers . to this end
, the computation of hole and particle spectral functions is briefly described in sec . [ spectral - h2d ] and more details
are given in appendix [ app-3 ] .
our third goal is to test the performance of our approximation for a selected set of illustrative examples .
the results of our calculations for the half - filled @xmath0 , @xmath1 and @xmath2 lattices are discussed in sec .
[ results ] .
there , we pay attention to the properties of ground and excited states but also discuss hole and particle spectral functions as well as the corresponding density of states ( dos ) .
in addition , in the case of the @xmath1 lattice , we consider doped systems with 14 and 15 electrons . finally , sec .
[ conclusions - work ] is devoted to the concluding remarks and work perspectives .
in what follows , we describe the theoretical framework used in the present study .
first , symmetry restoration and configuration mixing are presented in secs .
[ formalism - h2d ] and [ formalism - h2d - excited ] .
the computation of spectral functions is briefly described in sec .
[ spectral - h2d ] .
we consider the following one - band version of the 2d hubbard hamiltonian @xcite @xmath5 where the first term represents the nearest - neighbor hopping ( t @xmath6 0 ) , with unit hopping vectors @xmath7 and @xmath8 , and the second is the repulsive on - site interaction ( u @xmath6 0 ) .
the operators @xmath9 and @xmath10 create and destroy a particle with spin - projection @xmath11 ( also denoted as @xmath12 ) along an arbitrary chosen quantization axis on a lattice site @xmath13=(@xmath14 ) .
they satisfy the usual anticommutation relations for fermion operators.@xcite here , and in what follows , the lattice indices run as @xmath15 and @xmath16 with @xmath17 and @xmath18 being the number of sites along the x and y directions , respectively .
the total number of sites is given by @xmath19 .
we assume pbc , i.e. , the sites @xmath20 and 1 , with i = x , y , are identical .
furthermore , we assume a lattice spacing @xmath21=1 .
next , we apply the 2d fourier transform @xmath22 to obtain operators with momentum @xmath23 .
the hamiltonian ( [ ham - hub-2d ] ) can be easily written in terms of these new operators .
the quantum numbers @xmath24 , with i = x , y , take the allowed values @xmath25 inside the brillouin zone ( bz ) .
@xcite equivalently , they can take all integer values between 0 and @xmath26 . in the hf - approximation ,
the ground state of an @xmath27-electron system is represented by a slater determinant @xmath28 in which the energetically lowest @xmath27 single - electron states ( hole states @xmath29 , @xmath30 , ) are occupied while the remaining @xmath31 states ( particle states @xmath32 , @xmath33 , ) are empty .
the hf - quasiparticle operators are given by @xmath34 where @xmath35 is a general @xmath36 unitary transformation .
@xcite in eq .
( [ hf - transformation ] ) @xmath37 is a shorthand notation for the set @xmath38 .
the transformation ( [ hf - transformation ] ) mixes all the linear momentum quantum numbers as well as the spin projection of the states ( [ fourier-2d ] ) . as a consequence ,
@xmath39 deliberately breaks rotational ( in spin space ) and translational invariances . to restore the spin quantum numbers we explicitly use the projection operator @xmath40 where @xmath41 is the rotation operator in spin space , @xmath42 stands for the set of euler angles and @xmath43 are wigner functions .
@xcite the form ( [ proj - s ] ) has been frequently used for total angular momentum projection in nuclear physics .
@xcite this form has also been adopted in the study of the 1d hubbard model @xcite and more recently within pqt in quantum chemistry .
@xcite the linear momenta , @xmath44 and @xmath45 , are restored with the projector @xmath46 where @xmath47 is the generator of the considered lattice translations .
note that this operator neither has vector character nor corresponds to the true linear momentum operator .
it is associated with the quasi - momentum resulting from translational invariance of the lattice .
we will refer to it , however , as linear momentum for simplicity .
the projector ( [ projector - lm ] ) represents the 2d limit of the general operator restoring galilei invariance .
@xcite note that , at variance with atomic nuclei , lattice systems can have solutions with linear momenta different from zero . in what follows ,
we introduce the shorthand notation @xmath48 for the set of ( symmetry ) quantum numbers @xmath49 , i.e. , @xmath50 .
we then use the following symmetry - projected wave function @xmath51 are variational parameters .
note that , through the action of the projection operator @xmath52 , the multi - determinantal character of the state characterized by the quantum numbers @xmath53 and @xmath54 is recovered and written in terms of the quantum numbers @xmath53 and @xmath55 .
@xcite in practice , the integration over the set of euler angles in eq.([proj - s ] ) is discretized . for the integrals in @xmath56 and @xmath57
we have used 8 grid points whereas for the @xmath58-integration we have used 16 points .
therefore , the total number of grid points to be used for the projection operator @xmath52 is @xmath59 . for a given symmetry @xmath53 , the energy ( independent of @xmath55 ) associated with the state ( [ sp - wf - vampir ] ) @xmath60 is given in terms of the @xmath61 hamiltonian @xmath62 and norm @xmath63 matrices
( see appendix [ app-1 ] ) .
it has to be minimized with respect to the coefficients @xmath64 and the hf - transformation @xmath35 .
the variation with respect to the mixing coefficients yields the following generalized eigenvalue equation @xmath65 with the constraint @xmath66 ensuring the orthogonality of the solutions .
the unrestricted minimization of the energy ( [ energy - vampir ] ) with respect to the underlying hf - transformation @xmath35 can be carried out via the thouless theorem .
@xcite the corresponding variational equations assume the form @xmath67 with @xmath68_{ph}\end{aligned}\ ] ] here , the @xmath69 and @xmath70 matrices @xmath71 and @xmath72 are obtained via the cholesky decompositions .
@xcite the particle - hole kernels @xmath73 and latexmath:[${\cal{r}}_{\sigma { \sigma}^{'}}^{\theta;ph } = \langle { \cal{d } } | \hat{p}_{\sigma { \sigma}^{'}}^{\theta } { \hat{b}}^{\dagger}_{p } { \hat{b}}_{h } stressed that , for a given symmetry @xmath53 , we only retain the energetically lowest solution of eqs.([hw-1 ] ) and ( [ gradient ] ) . both the hf - transformation @xmath35 and the mixing coefficients @xmath75
are essentially complex , therefore one needs to minimize @xmath76 real variables .
we use a quasi - newton method for such a minimization .
@xcite the variational procedure already described is known in nuclear structure physics as the vampir ( i.e. , variation after mean field projection in realistic model spaces ) .
@xcite note , that particle number projection @xcite is not carried out in the present study since the considered slater determinants conserve the number of electrons .
an accurate description of excited states in a many - fermion system is much more difficult even when one is usually interested in just a small fraction of the low - lying spectrum . here
, the main difficulty in the optimization of excited states is ensuring orthogonality among them and with respect to the ground state .
for this , we simply use a gram - schmidt orthogonalization .
our goal in this section is to construct , throughout a chain of vap calculations , a basis of a few ( orthonormalized ) states with well defined quantum numbers @xmath53 .
suppose we have generated a ground state solution @xmath77 out of eqs .
( [ hw-1 ] ) and ( [ gradient ] ) in sec .
[ formalism - h2d ]
. then we write the first excited state wave function as @xmath78 has a form similar to eq.([sp - wf - vampir ] ) but with different coefficients @xmath79 and underlying hf - transformation @xmath80 .
the label 2 distinguishes them from the ones ( i.e. , @xmath81 and @xmath82 ) corresponding to the reference ground state we already have .
both @xmath83 and @xmath84 can be obtained by requiring that @xmath85 and @xmath86 .
they are given in terms of the projector ( i.e. , @xmath87 ) @xmath88 as follows @xmath89 the first excited state is obtained varying the energy functional for ( [ example - excvampir ] ) with respect to @xmath79 and @xmath80 . for the second excited state , we introduce a new state @xmath90 , again with the same form as in eq.([sp - wf - vampir ] ) , and write @xmath91 with coefficients @xmath92 , @xmath93 and @xmath94 such that @xmath95 is orthogonal to the previous solutions @xmath96 [ eq.([sp - wf - vampir ] ) ] and @xmath97 [ eq.([example - excvampir ] ) ] as well as @xmath98 . the second excited state is obtained varying the energy functional for ( [ example - excvampir-2 ] ) with respect to @xmath99 and @xmath100 .
let us have a more general situation in which , by successive variation , @xmath101 orthonormalized solutions ( for example , @xmath102 and @xmath97 ) @xmath103 in ( [ cuco-1 ] ) has the same form as ( [ sp - wf - vampir ] ) .
one then writes the ansatz for the @xmath104th state wave function ( for example , @xmath95 ) as @xmath105 having again the form ( [ sp - wf - vampir ] ) . requiring orthonormalization with respect to all the previous @xmath106 solutions ( [ cuco-1 ] ) the coefficients @xmath107 and @xmath108 in eq.([cuco-2 ] ) read @xmath109 in terms of the projector
( i.e. , @xmath110 ) @xmath111 the energy for the state ( [ cuco-2 ] ) takes the form @xmath112 with kernels @xmath113 and @xmath114 accounting for the fact that @xmath104 - 1 linearly independent solutions have been removed from the variational space .
their expressions are slightly more involved @xcite than the ones required in eq.([energy - vampir ] ) but still straightforward .
they require the knowledge of the symmetry - projected matrix elements between two different slater determinants @xmath115 and @xmath116 ( see appendix [ app-1 ] ) .
the variation of the energy ( [ ojo - ojo ] ) with respect to @xmath117 yields an equation similar to ( [ hw-1 ] ) with the constraint @xmath118 .
the unrestricted minimization of the energy ( [ ojo - ojo ] ) with respect to @xmath119 , via the thouless theorem , leads to variational equations similar to ( [ gradient ] ) but with kernels @xmath120 and @xmath121 that require symmetry - projected particle - hole matrix elements between two different slater determinants @xmath115 and @xmath116 ( see appendix [ app-2 ] ) .
the procedure outlined in this section is known in nuclear structure physics as excited vampir . @xcite
it provides a ( truncated ) basis of m ( orthonormalized ) states @xmath122 , with a well defined symmetry @xmath53 , still keeping low computational cost .
this is doable due to the simple structure of the projected states defining such a basis in combination with a fast minimization scheme .
@xcite our method can also be extended to use general hartree - fock - bogoliubov ( hfb ) transformations .
@xcite however , this requires an additional projection of the particle number , which increases the numerical effort by about one order of magnitude and has hence not been used in the present paper .
it should be noticed that the ground state @xmath102 [ eq.([sp - wf - vampir ] ) ] is written as a projection operator acting on a single determinant , the first excited state @xmath97 [ eq.([example - excvampir ] ) ] as a projection operator acting on two determinants , and so on . because this allows excited state wave functions to be described at a higher level of quality than is the ground state wave function ,
our final step is to diagonalize the 2d hubbard hamiltonian in the basis of the states @xmath122 .
@xmath123 c_{j \alpha}^{\theta } = 0\end{aligned}\ ] ] for ground and excited states , the resulting wave functions @xmath124 . in the present work , as a first step ,
we have restricted ourselves to test the performance of our approximation with @xmath104=5 ( orthonormalized ) states .
as we will see , this turns out to be a reasonable starting point for , at least , a qualitative description of the considered lattices .
an interesting issue is the evolution of the energy of each state with the number @xmath104 of transformations included in the prescription described in this section .
we observe that , for the lattices considered in the present study , the energy of the ground and the first couple of excited states remains unchanged when @xmath104 goes from 1 to 5 ( the changes in the energy per site are of the order @xmath125 ) .
this is partly because the main correlations have already been accounted for with a single symmetry - projected determinant .
therefore , the excited configurations obtained constitute reasonably good approximations to the true excited states of the considered system .
we produce @xmath104=5 symmetry - projected determinants in order to obtain the low - lying spectrum .
for the systems considered in this work , these states turn out to be weakly coupled through the hamiltonian . however , this can not be anticipated _ a priori _ and the diagonalization eq.([diagonalizationfinal ] ) should always be carried out .
preliminary results for larger square lattices ( i.e. , @xmath126 and @xmath127 ) as well as for other hamiltonians ( i.e. , the @xmath128 and @xmath129 hubbard models ) indicate that there are cases in which the diagonalization eq.([diagonalizationfinal ] ) brings a sizeable amount of additional correlations .
let us assume that for an even number @xmath27 of electrons we already have the ground state wave function @xmath130 , out of the calculations described in sec .
[ formalism - h2d ] .
since for all the considered lattices with an even number of electrons the ground state has spin s=0 , but not neccessarily linear momenta zero , we write its quantum numbers as @xmath131 .
usually , spectral functions are computed within a green s function perspective .
@xcite the key point is then to approximate the ground states of the @xmath3-electron systems by a suitable ansatz . in the present study ,
we approximate @xcite the ground state of the ( @xmath27 - 1)-electron system , with the symmetry @xmath132 , by @xmath133 where the index i runs as @xmath134 , the hole index h as @xmath135 and @xmath136 . in eq.([hole - wf - spectral ] ) , we write @xmath137 to explicitly indicate that holes are made on @xmath138 different slater determinants .
the determinants @xmath139 and @xmath140 correspond to the ground and lowest energy ( @xmath141 ) states obtained for the @xmath27-electron system out of the calculations described in sec .
[ formalism - h2d ] . in the present study , we have restricted ourselves to a maximum of @xmath142 hf - transformations . the coefficients @xmath143 in eq.([hole - wf - spectral ] ) are obtained by solving the equation @xmath144 that yields @xmath145 hole solutions @xmath146 with energies @xmath147 . with all the previous ingredients
, one can compute @xcite the hole spectral function as @xmath148 in terms of the reduced matrix element @xmath149^{-1 } n^{i 1}(\omega,{\boldsymbol{j } } ) \end{aligned}\ ] ] where @xmath150 .
the indices i , @xmath151 and @xmath152 run as in ( [ hole - wf - spectral ] ) , @xmath153 and @xmath154 run as in eq.([values - momenta ] ) and @xmath155 .
details for the computation of the kernels @xmath156 and @xmath157 in eq.([hw - holes - h ] ) as well as @xmath158^{-1}$ ] and @xmath159 in eq.([hole - spectra function ] ) can be found in appendices [ app-3 ] and [ app-1 ] , respectively .
the occupation number @xmath160 of a basis state ( [ fourier-2d ] ) in the @xmath27-electron ground state can be computed as @xmath161 the ( @xmath27 + 1)-electron system , with the symmetry @xmath162 , is approximated by @xcite @xmath163 where the index i runs again as in ( [ hole - wf - spectral ] ) .
the particle index p takes the values @xmath164 and @xmath136 . in this case , the coefficients @xmath165 are obtained by solving the equation @xmath166 that yields @xmath167 particle solutions @xmath168 with energies @xmath169 . the particle spectral function
is then written as @xmath170 in terms of the reduced matrix element @xmath171 where , in this case , @xmath172 .
the indices i , @xmath173 and @xmath174 , @xmath175 run as in ( [ particle - wf - spectral ] ) , @xmath176 and @xmath177 run as in eq.([values - momenta ] ) and @xmath178 .
details for the computation of the kernels @xmath179 and @xmath180 as well as @xmath181 in eq.([particle - spectra function ] ) can be found in appendix [ app-3 ] .
finally , the dos can be computed as @xmath182\end{aligned}\ ] ] where the indices @xmath146 and @xmath168 are absorbed into the continuous variable @xmath183 . due to the finite size of the system the spectral functions consist of a finite number of @xmath184 functions with different weights .
therefore , we introduce an artificial width @xmath185 for each state using a lorentzian . in all cases
our dos is normalized to 2 @xmath186 @xmath187 .
in this section , we discuss the results of our study .
we have considered the @xmath188 half - filled lattice as a prototypical system where one can obtain the full spectrum by means of ed .
this allows us to callibrate our approximation not only for ground state properties but also for excited states .
next , we have considered the well - studied half - filled @xmath1 lattice , which constitutes the largest square lattice for which exact ground state energies are available in the literature .
other approximation schemes have also been tested for this lattice in previous works .
results have already been published for doped systems with 14 and 15 electrons in this lattice , which motivated us to also perform calculations for them in the present study .
last , we consider the half - filled @xmath2 lattice as a prototype of a system where ed is no longer feasible .
many of the results to be discussed in what follows correspond to u=4 t taken as a representative on - site repulsion for which studies are available .
nevertheless , let us stress that our formalism can be used for any 2d hubbard hamiltonian of the form ( [ ham - hub-2d ] ) with arbitrary u and/or t values .
let us start by considering the rectangular @xmath0 lattice .
the first five solutions obtained at half - filling via eq.([diagonalizationfinal ] ) , for each of the linear momentum quantum numbers ( 0,0 ) , ( 0,1 ) , ( 0,2 ) , ( 0,3 ) , ( 1,0 ) , ( 1,1 ) , ( 1,2 ) and ( 1,3 ) and the spins s=0,1 , and 2 , are plotted in panel a ) of fig.[2by4_lattice_spectrum_ne8 ] for u=4 t .
the first excited state corresponds to a @xmath189 configuration [ with linear momenta @xmath190 .
the energies @xmath191 of the 120 solutions shown in the figure , have been compared to the ones obtained using an ed .
@xcite the comparison reveals that both spectra follow the same qualitative trend and can hardly be distinguished . therefore in panel b ) of the same figure , we have plotted the absolute errors @xmath192 for each of the predicted 120 states .
our approximation fairly reproduces the exact ground state energy -10.2529
t for this system .
for all the 40 s=0 and s=1 solutions considered the absolute errors remain very small , the largest deviation being 0.047 t for the second state with the symmetry @xmath193 .
the previous results are encouraging if one takes into account that , even for this relatively small lattice , the number of variational parameters in our approximation @xmath194(@xmath195 and @xmath194(@xmath196 is about half of the dimensions @xmath197(@xmath198 and @xmath197(@xmath199 of the restricted hilbert spaces . on the other hand , @xmath194(@xmath200 is larger than @xmath197(@xmath201 and therefore our solutions reproduce the ed ones for s=2 states . in panel
a ) of fig . [ densitystates_halfilling_2by4_u4 ] , we have plotted the dos @xmath202 [ eq.([densityofstates ] ) ] for the half - filled @xmath0 lattice at u=4 t .
the calculations have been carried out by approximating the ( @xmath27 @xmath203 1)-electron systems [ see eqs .
( [ hole - wf - spectral ] ) and ( [ particle - wf - spectral ] ) ] with @xmath138=1 ( red curve ) and @xmath138=5 ( blue curve ) hf - transformations along the lines described in sec .
[ spectral - h2d ] .
we have introduced a shift equal to the chemical potential at half - filling ( @xmath204 ) so that the dos in fig.[densitystates_halfilling_2by4_u4 ] appears to be symmetric around @xmath183-u/2=0 .
this convention , i.e. , to plot dos and spectral functions vs. @xmath183-u/2 will be adopted in the rest of the paper .
the dos shows the hubbard gap , @xmath205 , characteristic of finite size lattices .
we note , however , that previous studies within the framework of the dynamical cluster approximation ( dca ) have shown that the gap is preserved at sufficiently low temperatures even in the thermodynamic limit ( tdl ) .
@xcite on the other hand , the nonperturbative study of ref .
59 has concluded that for the half - filled hubbard model the gap persists for any finite value of the on - site repulsion u , the only singular point being u=0 t . from panel a ) of fig.[densitystates_halfilling_2by4_u4 ]
one realizes that , even for this small lattice , the fine details of the energy distribution of @xmath202 can only be obtained using a larger number @xmath138=5 of hf - transformations to describe the ( @xmath27 @xmath203 1)-electron systems . using @xmath138=5 transformations , eqs.([hw - holes - h ] ) and ( [ hw - particles - h ] )
provide us with 80 hole and particle solutions while only 16 solutions are obtained with @xmath138=1 .
therefore , contributions to @xmath202 with a more collective nature can be better accounted for in the former case ( i.e. , @xmath138=5 ) .
this is further corroborated by comparing our dos , computed with @xmath138=5 transformations , with the one obtained using an ed , performed with an in - house code , shown in panel b ) of the figure .
note that we have intentionally used a small broadening @xmath185=0.05 t to retain as much structure as possible in our dos as well as to emphasize the differences with the ed one . as can be observed
there is excellent agreement in the position and relative heights of all the prominent peaks . the hole ( blue ) and particle ( black ) spectral functions , computed with @xmath138=5 hf - transformations , are displayed in panel c ) of the same figure .
we have not included the ones provided by the ed since they are quite similar to ours .
their structure is dominated by a main peak but less prominent ones are also visible in the figure .
the momenta @xmath206 and @xmath207 , at the noninteracting fermi surface @xmath208 [ see , eq.([eq - seenergies ] ) of appendix [ app-1 ] ] , have the largest spectral weight near @xmath183-u/2=0 .
on the other hand , the momenta @xmath209 and @xmath210 [ @xmath211 and @xmath190 inside ( outside ) the noninteracting fermi surface contribute mostly to hole ( particle ) states . in fig .
[ occupations_halfilling_2by4_u4 ] , we display the occupation numbers of the basis states [ see eq.([fourier-2d ] ) ] in the @xmath212 ground state of the half - filled @xmath0 lattice .
results are shown for the on - site repulsions u=4 t , 20 t , 40 t , 64 t , 80 t and 120 t .
the calculations were performed using 80 hole solutions @xmath213 ( i.e. , @xmath138=5 hf - transformations ) in eq.([occupation - numbers ] ) .
the evolution of the occupations clearly depict the transition to the strong coupling regime where the hubbard hamiltonian @xcite can be mapped into the af heisenberg model .
@xcite in fact , for u @xmath214 64 t the results look very similar to the uniform distribution , with occupations @xmath215 , expected in the limit u @xmath216 . in panel
a ) of fig.[4by4_lattice_spectrum_ne16 ] , we show the energies @xmath191 obtained , via eq.([diagonalizationfinal ] ) , for the half - filled @xmath1 lattice at u=4 t .
results are only shown for the six essentially different pairs of linear momentum quantum numbers ( 0,0 ) , ( 1,0 ) , ( 1,1 ) , ( 2,0 ) , ( 2,1 ) and ( 2,2 ) .
for each of them we have plotted the energies of the first five solutions with spins s=0,1 , and 2 . in this case , the number of variational parameters in our approximation is @xmath194(@xmath217)=512 , @xmath194(@xmath218)=516 and @xmath194(@xmath219)=520 while the dimensions of the restricted hilbert spaces are @xmath220 , @xmath221 and @xmath222 , respectively .
the energy -13.5898 t of our @xmath212 ground state accounts for 99.76 @xmath223 of the exact one , @xcite -13.6219 t . in order to put our result in perspective
, the relative error 0.24 @xmath223 in our ground state energy per site @xmath224 at u=4 t should be compared , for example , with the value 0.70 @xmath223 recently reported @xcite within the framework of the variational mc ( vmc ) approximation using an ansatz , consisting of the product of a correlator product state tensor network and a pfaffian wave function , with 524,784 variational parameters . note ,
that the dmrg formalism in momentum space @xcite ( kdmrg ) predicts a relative error of 0.37@xmath223 .
we have also studied our ground state energy per site @xmath224 as a function of the interaction strength u=2 t , 6 t , 8 t , 10 t , 12 t and 16 t and found relative errors always smaller than 0.4 @xmath223 .
coming back to the spectrum shown in panel a ) of fig.[4by4_lattice_spectrum_ne16 ] , we note that the first excited state corresponds to a @xmath225 configuration [ with linear momenta @xmath190 .
in fact , similar to the half - filled @xmath0 lattice , the first excited state for each combination @xmath226 has spin s=1 , exception made of @xmath209 .
the @xmath0 lattice displays a low - lying s=0 singlet ( see fig.[2by4_lattice_spectrum_ne8 ] ) while an s=2 quintet appears in the @xmath1 lattice .
the excitation energies , referred to the @xmath212 configuration , of these low - lying s=1 and s=2 states are shown in panel b ) of fig.[4by4_lattice_spectrum_ne16 ] as functions of the linear momentum quantum numbers .
the shape of the curve does not fully agree with the one obtained with the spin - density wave ( sdw ) approximation @xcite mainly due to the absence of degeneracy between the @xmath185=(0,0 ) and @xmath227=(2,2 ) as well as the two peaks for the @xmath228=(1,0 ) and @xmath229=(2,1 ) points .
much of this discrepancy could , however , be due to finite size effects .
@xcite note that the two peaks at @xmath228 and @xmath229 , resulting from a kinetic - energy gap of 2 t , are already visible for the fermi gas ( u=0 t ) . the dos @xmath202 [ eq.([densityofstates ] ) ] for the half - filled @xmath1 lattice at u=4 t is shown in panel a ) of fig.[densitystates_halfilling_4by4_u4 ] .
the calculations have been performed using @xmath138=5 hf - transformations along the lines described in sec .
[ spectral - h2d ] . in this case , eqs.([hw - holes - h ] ) and ( [ hw - particles - h ] ) provide us with 160 hole and particle solutions .
a lorentzian folding of width @xmath185=0.2 t has been used .
similar to the case of the half - filled @xmath0 lattice , the hubbard gap , @xmath230 , remains present in this larger system .
the hole ( blue ) and particle ( black ) spectral functions shown in panel b ) of the figure show that the momenta @xmath231 , @xmath206 and @xmath207 at the noninteracting fermi surface have the largest spectral weight near @xmath183-u/2=0 .
moreover , the spectral weight due to these momenta at the fermi surface is particle - hole symmetric . in panel
a ) of fig.[densitystates_halfilling_4by4_u8 ] , we have plotted the dos @xmath202 [ eq.([densityofstates ] ) ] for the half - filled @xmath1 lattice at u=8 t ( i.e. , an on - site repulsion equal to the noninteracting bandwidth w=8 t ) computed with @xmath138=5 hf - transformations .
a lorentzian folding of width @xmath185=0.2 t has been used . the corresponding hole ( blue ) and particle ( black ) spectral functions are also displayed in panel b ) of the figure .
our dos and spectral functions [ in particular , the ones corresponding to the linear momenta @xmath207 , @xmath232 and @xmath209 ] can be compared with the ones , obtained using the lanczos method , shown in fig.2 of ref . 57 . as can be observed ,
the main qualitative features of the particle - hole symmetric dos are well reproduced , namely the two prominent peaks at @xmath183-u/2 @xmath234 2 t and 3 t ( -2 t and -3 t ) a lump peaked around @xmath183-u/2 @xmath234 5 t ( -5 t ) and a smaller satellite peak in the neighborhood of @xmath183-u/2 @xmath234 8 t ( -8 t ) .
in agreement with the results of ref .
57 , the upper and lower bands as well as the hubbard gap are also clearly visible in fig.[densitystates_halfilling_4by4_u8 ] .
we have also studied the evolution of the dos for the half - filled @xmath1 lattice as a function of u. to this end , in addition to the cases u=4 t and 8 t shown in figs .
[ densitystates_halfilling_4by4_u4 ] and [ densitystates_halfilling_4by4_u8 ] , calculations have also been performed for u=2 t , 12 t and 20 t . in good agreement with previous studies ,
@xcite we observe that the hubbard gap persists for increasing values of u. in our calculations , a pronounced suppression in the dos around @xmath183-u/2=0 is observed with the dos fully vanishing around u=8 t ( i.e. , around the noninteracting bandwidth w=8 t ) which is precisely the region where a sizeable hubbard gap is developed .
@xcite let us now consider two examples of a doped @xmath1 lattice . in fig.[4by4_lattice_spectrum_ne15
] , we show the spectrum in the case of 15 electrons at u=4 t . for each of the linear momentum quantum numbers ( 0,0 ) , ( 1,0 ) , ( 1,1 ) , ( 2,0 ) , ( 2,1 ) and ( 2,2 ) , we plot the energies of the first five solutions of eq.([diagonalizationfinal ] ) for the spins s=1/2 and 3/2 .
the number of variational parameters in our approximation @xmath194(@xmath235=512 and @xmath194(@xmath236=516 should be compared with the dimensions @xmath197(@xmath237 2 @xmath238 and @xmath197(@xmath239 2 @xmath238 of the restricted hilbert spaces .
the first noticeable feature in fig.[4by4_lattice_spectrum_ne15 ] is that the four - fold degenerate @xmath240 ground state has non - zero linear momenta @xmath241 .
a finite linear momentum for the one - hole ground state has also been predicted in previous studies @xcite using a variety of approximations for lattices of different sizes .
our numerical calculations also predict a two - fold degenerate ( 1/2,2,0 ) configuration whose energy is almost the same as the ground state one . for the noninteracting system ( u=0 t ) , the lowest - lying s=1/2 and s=3/2 states with
linear momentum quantum numbers @xmath242 , @xmath243 , @xmath244 , and @xmath245 are degenerate and the same is also true for the configurations @xmath246 and @xmath247 .
therefore , the huge degeneracy observed in the noninteracting case is already partially lifted at u=4 t . in fig.[oneholeprogress ]
, we compare the ground state energy of the @xmath1 lattice with 15 electrons with the exact one @xcite for u=4 t .
the energy -14.5469 t predicted within our symmetry - projected configuration mixing approach , via eq.([diagonalizationfinal ] ) , accounts for 99.19 @xmath223 of the exact result .
it is interesting to note that linear momentum plus @xmath248 projection already accounts for 98.41 @xmath223 of the exact solution .
nevertheless , full spin projection , while also recovering the total spin quantum number , still brings a sizeable amount of correlations .
the ( shifted ) differences @xmath249-u/2 , where @xmath250 is the ground state energy of the half - filled lattice ( fig.[4by4_lattice_spectrum_ne16 ] ) and @xmath251 represents the energy of each of the lowest - lying s=1/2 states in fig.[4by4_lattice_spectrum_ne15 ] , compare very well with the position of the first prominent peak in the hole spectral functions shown in panel b ) of fig.[densitystates_halfilling_4by4_u4 ] .
for example , the variational approach predicts @xmath252-u/2=-1.044 t while the corresponding peak in the hole spectral function is predicted to be at @xmath183-u/2= -1.010 t .
the same is also true for the configuration with linear momenta ( @xmath253,0 ) for which the variational approach predicts @xmath254-u/2=-1.052 t whereas the position of the corresponding peak in the hole spectral function is predicted to be at @xmath183-u/2= -1.012 t .
this leads to the conclusion that the lowest - lying s=1/2 states in the spectrum of fig.[4by4_lattice_spectrum_ne15 ] are reasonably well described by a wave function of the form ( [ hole - wf - spectral ] ) .
this is remarkable , since no orbital relaxation is accounted for in this wave function , i.e. , the determinants @xmath255 in eq.([hole - wf - spectral ] ) correspond to the ones obtained at half - filling .
the spectrum obtained , via eq.([diagonalizationfinal ] ) , for 14 electrons at u=4 t is displayed in fig.[4by4_lattice_spectrum_ne14 ] .
the number of variational parameters in our approximation @xmath194(@xmath256=504 , @xmath194(@xmath257=508 and @xmath194(@xmath258=512 should be compared with the dimensions @xmath197(@xmath259 , @xmath197(@xmath260 and @xmath197(@xmath261 of the restricted hilbert spaces .
the ground state corresponds to the @xmath262=(0,2,2 ) configuration [ linear momenta ( @xmath263 with energy -15.5872 t , while the exact one is -15.7446 t .
@xcite on the other hand , the vmc approximation @xcite predicts a ground state energy of -15.5936 t .
thus , both methods , ours and vmc , yield essentially the same relative error of around 1 @xmath223 in the ground state energy per site . on the other hand , a relative error in the ground state energy per site of 0.45 @xmath223 is obtained within the kdmrg approximation .
@xcite our calculations also predict two other close - lying ( 0,2,2 ) solutions ( with energies -15.5747 t and -15.5777 t ) which can not be distinguished in fig.[4by4_lattice_spectrum_ne14 ] and therefore appear , together with the ground state , as a single thick black line .
the energy -15.5743 t of the @xmath264=(0,0,0 ) configuration is also close to the actual ground state . as a result ,
the symmetry of the @xmath185 and @xmath227 points is almost recovered in panel b ) of fig.[4by4_lattice_spectrum_ne14 ] where the energies of the lowest - lying s=0,1,2 states for each linear momentum combination are shown , referred to the @xmath265 ground state for this system .
note that the configurations ( 0,1,1 ) and ( 0,2,0 ) ( i.e. , the points @xmath266 and @xmath267 in panel b ) of fig.[4by4_lattice_spectrum_ne14 ] ) have very small excitation energies , 0.0552 t and 0.1242 t , respectively .
the two peaks at the points @xmath228 and @xmath229 are also present in the system with @xmath268 electrons at u=0 t .
the spectrum in panel a ) of fig.[4by4_lattice_spectrum_ne14 ] exhibits an increase in the density of energy levels , compared to the one at half - filling , pointing to its very correlated nature . the dos @xmath202 [ eq.([densityofstates ] ) ] for 14 electrons at u=4 t is shown in panel a ) of fig.[densitystates_4by4_ne14_u4 ] .
the calculations have been carried out by approximating the ( @xmath27 @xmath203 1)-electron systems with @xmath138=5 hf - transformations along the lines described in sec .
[ spectral - h2d ] .
a lorentzian folding of width @xmath185=0.2 t has been used . the hole ( blue ) and particle ( black ) spectral functions are displayed in panel b ) of the figure . in this case , eqs.([hw - holes - h ] ) and ( [ hw - particles - h ] ) provide us with 140 hole and 180 particle solutions .
the chemical potential is now located around @xmath183-u/2=-1.2 t . the comparison with panel b ) of fig.[densitystates_halfilling_4by4_u4 ] reveals that the structure of the hole states for @xmath183-u/2 @xmath269 -2 t [ i.e. , those with linear momenta @xmath270 , @xmath210 and @xmath209 ] remains to a large extent intact . on the other hand ,
a large fraction of the particle spectral weight observed at half - filling for 1 t @xmath269 @xmath183-u/2 @xmath269 3 t is removed .
this depletion occurs in favor of new states near @xmath183-u/2=0 .
the spectral decomposition of the dos clearly shows that it is states around the fermi surface that suffer the most pronounced changes with respect to half - filling . as a result
, the particle - hole symmetry in the dos , observed in fig.[densitystates_halfilling_4by4_u4 ] , is suppressed for this doped lattice and the original gap dissapears . in panel
a ) of fig.[densitystates_4by4_ne14_u8 ] , we have plotted the dos @xmath202 [ eq.([densityofstates ] ) ] for the @xmath1 lattice with 14 electrons at u=8 t .
the calculations were performed with @xmath138=5 hf - transformations and a folding @xmath185=0.2 t has been used .
the hole ( blue ) and particle ( black ) spectral functions are displayed in panel b ) of the figure .
our dos and spectral functions can be compared with the ones , obtained using the lanczos method , shown in figs.3 and 4 of ref.57 .
the main qualitative features of the dos are well reproduced , namely the prominent peaks around @xmath183-u/2=-4 t , -3 t , -2 t and 4 t . as can be noted from panel b ) of fig.[densitystates_4by4_ne14_u8 ] , the chemical potential is now located around @xmath183-u/2=-2.4 t .
one of the main features of the dos is that it displays a pronounced pseudogap which , as can be seen from figs .
[ densitystates_halfilling_4by4_u8 ] and [ densitystates_4by4_ne14_u8 ] , results mainly from pulling particle strenghts into the ( half - filling ) gap combined with sizeable contributions of particle states around @xmath183-u/2=4 t .
we note , that the pseudogap problem in the doped 2d hubbard model has also received attention within the framework of quantum cluster approaches .
@xcite we have also performed calculations for the @xmath1 lattice with 14 electrons at u=2 t , 12 t and 20 t . from these calculations , and the results already discussed above for the cases u=4 t and 8 t , we conclude that upon dopping with two holes the original gap observed at half - filling dissapears for u=2 t and 4 t , a pseudogap is developed around the noninteracting bandwidth w=8 t while for the larger interaction strengths u=12 t and u=20 t the gap is not filled .
similar conclusions have been obtained within the lanczos framework .
@xcite finally , let us turn our attention to the half - filled @xmath2 lattice at u=4 t .
the dimensions @xmath197(@xmath271 , @xmath197(@xmath272 and @xmath197(@xmath273 of the corresponding restricted hilbert spaces are far too large for a brute force diagonalization to be feasible .
other approximate methods are then called for , not only to describe ground state properties but also to access the excitation spectrum in this relatively large lattice for which information is rather scarce . in this case , the number of variational parameters in our approximation is @xmath194(@xmath256=2592 , @xmath194(@xmath257=2596 and @xmath194(@xmath258=2600 , respectively .
therefore , the half - filled @xmath2 lattice represents a very challenging testing ground for our symmetry - projected configuration mixing approximation . in fig.[6by6_lattice_spectrum_ne36 ] , we show the energies @xmath191 obtained , via eq.([diagonalizationfinal ] ) , for the ten essentially different pairs of linear momentum quantum numbers ( 0,0 ) , ( 0,1 ) , ( 0,2 ) , ( 0,3 ) , ( 1,1 ) , ( 1,2 ) , ( 1,3 ) , ( 2,2 ) , ( 2,3 ) and ( 3,3 ) . for each of them
we have plotted the first five solutions with spins s=0,1 , and 2 .
the ground state corresponds to the @xmath262=(0,0,0 ) configuration with energy @xmath250=-30.5766 t .
this can be compared with the energy obtained using state - of - the - art auxiliary - field mc , -30.89(1)t .
@xcite from fig.[6by6_lattice_spectrum_ne36 ] , we realize that the first excited state corresponds to a @xmath274 configuration [ with linear momenta @xmath190 . the energy difference 0.1331
t between this @xmath275-configuration and the ground state is smaller than the corresponding value 0.1651 t for the half - filled @xmath1 system . on the other hand ,
similar to the half - filled @xmath0 and @xmath1 lattices , most of the first excited states for each combination @xmath226 have spin s=1 , exception made of @xmath209 and @xmath276 for which an s=2 quintet appears .
note that our calculations predict the lowest - lying ( 1,0,1 ) , ( 1,1,1 ) and ( 1,2,3 ) solutions to be quite close in energy .
as with the other half - filled lattices studied , we find only a handful of excited states within an energy window of t from the ground state . the dos @xmath202 [ eq.([densityofstates ] ) ] for the half - filled @xmath2 lattice at u=4 t is shown in panel a ) of fig.[densitystates_halfilling_6by6_u4 ] .
the calculations have been carried out with @xmath138=5 hf - transformations along the lines described in sec .
[ spectral - h2d ] . in this case , eqs.([hw - holes - h ] ) and ( [ hw - particles - h ] ) provide us with 360 hole and particle solutions . a lorentzian folding of width @xmath185=0.2 t has been used .
the dos for this large lattice shows a clear hubbard gap @xmath205 . the fact that the gap remains intact in going , at half - filling , from the @xmath1 to the @xmath2 lattice is consistent with previous studies within the dca which show that it is preserved even in the tdl .
@xcite we note , that the dmft ( i.e. , @xmath277 ) does not predict a gap for u=4 t . it is only for a larger number @xmath278 of clusters that the gap starts to develop in these studies at the tdl .
@xcite back to our dos , we see again from the spectral decomposition , shown in panel b ) of fig.[densitystates_halfilling_6by6_u4 ] , that it is states at [ i.e. , @xmath279 , @xmath280 , @xmath207 and @xmath206 ] or close to the noninteracting fermi surface that contribute the largest spectral weight to the prominent hole and particle peaks around @xmath183-u/2=-t and @xmath183-u/2=t . in general , the dos for this finite size lattice is still highly peaked , though features should smooth out as one approaches the tdl . systematic studies for the @xmath126 and @xmath127 lattices are in progress and will be presented elsewhere .
how to accurately describe many - fermion systems with approximate methods , which truncate the complete expansion of the wave functions to a numerically feasible number of configurations , is a central question in nuclear structure theory , quantum chemistry , and condensed matter physics . to this end , in the present study we have explored an alternative avenue for the 2d hubbard model .
the main accomplishments of the present study are : * we have presented a powerful methodology of a vap configuration mixing scheme , originally devised for the nuclear many - body problem , but not yet used to study ground and excited states , with well defined quantum numbers , of the 2d hubbard model with nearest - neighbor hopping and pbc .
our scheme relies on the ritz variational principle to construct , throughout a chain of vap calculations , a truncated basis consisting of a few ( orthonormalized ) symmetry - projected hf states .
the simple structure of the projected wave functions employed , combined with a fast minimization algorithm , allows to keep low computational cost in building our basis .
a further diagonalization of the hamiltonian within such a basis allows to account , in a similar fashion , for residual correlations in the ground and excited states . *
due to the simple structure of the wave functions in our approximation , we can construct an ansatz [ eqs .
( [ hole - wf - spectral ] ) and ( [ particle - wf - spectral ] ) ] , whose flexibility is well - controlled by the number of hf - transformations included , to approximate the ground state of the @xmath3-electron system .
this allows us to determine one - electron affinities and ionization potentials as well as to access the spectral weight of states with different linear momentum quantum numbers in the calculation of spectral functions and the corresponding density of states .
* we have shown that our approximation gives accurate results , as compared with exact energies , for the @xmath0 and @xmath1 lattices .
we have also provided the low - lying spectrum of the @xmath2 lattice which , to the best of our knowledge , has not been reported in the literature .
our ground state energy for this lattice compares well with results from state - of - the - art auxiliary - field monte carlo calculations .
regarding the physics of the 2d hubbard model , we have discussed the trends , in going from the @xmath188 to the @xmath1 and @xmath2 half - filled lattices , of both the low - lying spectra and the spectral functions as well as the corresponding density of states .
we have found that the ground states correspond to configurations with spin zero and linear momenta ( 0,0 ) .
we have also found that most of the lowest - lying excited states display spin s=1 .
the doped systems with 14 and 15 electrons in the @xmath1 lattice have also been considered .
the ground states of such systems correspond to configurations with linear momenta different from zero .
special attention has been paid to the spectral weight of states with different linear momentum quantum numbers .
we have compared the dos predicted within our approximation with the one obtained using an exact diagonalization for the half - filled @xmath0 lattice and found an excellent agreement between the two .
our results for the half - filled @xmath1 lattice , at different on - site repulsions , agree qualitatively well with the ones obtained using the lanczos method .
@xcite for all the considered half - filled lattices , a hubbard gap is predicted within our approximation .
in particular , the fact that this gap persists in going from the @xmath1 to the @xmath2 system is consistent with previous studies within cluster extensions to dynamical mean field theory which show that it is preserved even in the thermodynamic limit .
as opposed to the half - filled case , the particle - hole symmetry in the dos is removed when doping is present in the system . from the calculations for 14 electrons in the @xmath1 lattice we conclude that for on - site repulsions smaller than the noninteracting bandwidth the ( half - filling ) gap dissapears , a pseudogap develops around u=8 t while the gap is not filled for larger u values .
these results agree well with similar conclusions extracted from lanczos calculations .
@xcite we have also found the remarkable result that all the lowest - lying s=1/2 states in the spectrum of the @xmath1 lattice with 15 electrons can be reasonably well described by a wave function of the form ( [ hole - wf - spectral ] ) in which no orbital relaxation is accounted for .
one important feature of the scheme presented in this study is that it leaves ample space for further improvements and research .
first , the number of symmetry - projected configurations in our basis set can be easily increased .
second , we could still incorporate particle number symmetry breaking and restoration in our configuration mixing scheme to access even more correlations .
our methods could be useful even for more complicated lattices like the honeycomb one .
an extension of the considered 2d t - u hubbard hamiltonian to the @xmath282 case is also straightforward , allowing the study of several interesting issues like indications of spin - charge separation in 2d systems ( see , for example , ref .
65 and references therein ) .
last but not least , not only the configuration mixing scheme applied in the present work but also the full hierarchy of approximations discussed in ref .
43 can be implemented for the molecular hamiltonian in the realm of quantum chemistry , within the already successful pqt .
@xcite work along these avenues is in progress .
we would like to stress that the cost of the symmetry - projected calculations described in this work has the same scaling as mean - field methods .
@xcite this statement is true as long as the number of grid points required in the symmetry restoration remains relatively constant ( this is usually the case for spin and number projection ) .
note , however , that the restoration of translational symmetry in hubbard lattices with pbc requires a number of grid points equal to @xmath187 .
this makes the cost of our calculations @xmath283 more expensive than the hartree - fock method , which is still a very reasonable scaling .
the computational effort is mainly concentrated in looping over grid points for the evaluation of matrix elements .
this task is trivially parallelizable and one can thus easily reach clusters larger than the ones considered in this study .
preliminary calculations for the half - filled and doped @xmath126 and @xmath127 lattices are in progress .
a discussion of the limitations of our method is also in order here .
evidently , our method relies on the hamiltonian having good symmetries .
the lower the number of symmetries of a given hamiltonian , the lower the correlations that can be accounted for by means of symmetry restoration . on the other hand ,
the most interesting quantum behavior is found in systems where symmetries are present .
second , it is known that the correlation energy per particle obtained with approaches based on a single symmetry - projected determinant @xcite decays as the lattice size increases .
we observe that the error in the energy per site is larger in the case of the half - filled @xmath2 lattice than in the @xmath1 one , although in both cases we have restricted the present study to @xmath104=5 transformations .
one can , however , increase the number of transformations to maintain the quality of our wave functions . in principle , if the number of transformations is equal to the size of the restricted hilbert subspace the method becomes exact . in practice
, one can only hope that the number of transformations needed to access the relevant physics of the considered lattices is relatively low .
finally , we believe that the finite size calculations discussed in the present work are complementary to other approaches where impurity solvers play an important role . however , the symmetries to be broken and restored in the impurity - bath and bath hamiltonians will depend on the details of the case considered .
@xcite this work is supported by the national science foundation under grants che-0807194 and che-1110884 , and the welch foundation ( c-0036 ) .
the authors would like to thank both prof .
shiwei zhang and dr .
eric neuscamman for providing us with unpublished results .
we also thank dr .
donghyung lee for providing us with exact diagonalization energy results for the @xmath0 lattice .
in this appendix , we present the expressions for the matrix elements @xmath284 and @xmath285 required to compute the kernels @xmath113 and @xmath114 in eq.([ojo - ojo ] ) of sec . [ formalism - h2d - excited ] .
note that the matrix elements required in eq.([energy - vampir ] ) of sec .
[ formalism - h2d ] are just a particular case where both slater determinants are the same . here , and in what follows , we keep our notation as close as possible to the one already used for the 1d hubbard model .
@xcite both @xmath286 and @xmath287 read @xmath288 where , @xmath289 and @xmath290 , respectively . for the gauge - rotated norm @xmath291 the determinant has to be taken over the @xmath69 dimensional occupied part of the matrix @xmath292 with @xmath293 the gauge - rotated hamiltonian takes the form @xmath294 with @xmath295^{-1 } { \cal{d}}_{{\boldsymbol{\alpha } } \sigma , h^{'}}^{i } \nonumber\\ { \gamma}_{{\boldsymbol{\alpha } } \sigma , { \boldsymbol{\gamma } } { \sigma}^{'}}^{i k } ( \omega,{\bf{j } } ) & = { \delta}_{\sigma { \sigma}^ { ' } } { \delta}_{{\alpha}_{x } { \gamma}_{x } } { \delta}_{{\alpha}_{y } { \gamma}_{y } } \epsilon({\bf{k}}_{\alpha } ) + \frac{u}{n_{sites } } \sum_{{\boldsymbol{\beta } } { \boldsymbol{\delta } } } { \delta}_{{\alpha}_{x}+ { \beta}_{x } - { \gamma}_{x } - { \delta}_{x } } ^{0 , \pm n_{x } } { \delta}_{{\alpha}_{y}+ { \beta}_{y } - { \gamma}_{y } - { \delta}_{y } } ^{0 , \pm n_{y } } \times \nonumber\\ & \times \big [ { \delta}_{\sigma { \sigma}^ { ' } } { \rho}_{{\boldsymbol{\delta } } -\sigma,{\boldsymbol{\beta } } -\sigma } ^{k i } ( \omega,{\bf{j } } ) - \left ( 1- { \delta}_{\sigma { \sigma}^ { ' } } \right ) { \rho}_{{\boldsymbol{\delta } } \sigma,{\boldsymbol{\beta } } -\sigma } ^{k i } ( \omega,{\bf{j } } ) \big ] \end{aligned}\ ] ] where the product of generalized kronecker deltas in @xmath296 results from the transformation of the on - site interaction term in eq.([ham - hub-2d ] ) to the momentum representation . as a consequence of the pbc
, @xmath297 is one if @xmath298 is either 0 or @xmath299 and zero else .
in this appendix , we present the expressions for the matrix elements @xmath300 and @xmath301 required to compute the kernels @xmath120 and @xmath121 defining the variational equations discussed in sec . [ formalism - h2d - excited ] .
note that the matrix elements required in eq.([gradient ] ) of sec .
[ formalism - h2d ] are just a particular case where both slater determinants are the same .
we obtain @xmath302 where , as in appendix [ app-1 ] , @xmath289 and @xmath290 , respectively . on the other hand , @xmath303^{-1 } { \cal{x } } _ { h^{'}p}^{ik}(\omega,{\bf{j } } ) \end{aligned}\ ] ] with the indices @xmath29 ( @xmath32 ) running over all the occupied ( unoccupied ) states in @xmath304 .
the inverse @xmath305^{-1}$ ] is taken over the occupied part of the matrix ( [ keepx ] ) .
finally , @xmath306_{hp}\end{aligned}\ ] ] with the functions @xmath307 and @xmath308 ) defined , for all the occupied @xmath29 and unoccupied @xmath32 states in @xmath304 , as @xmath309^{-1 } { \cal{d}}_{{\boldsymbol{\alpha } } \sigma , h^{'}}^{i } \nonumber\\ { \overline{\cal{w}}}_{{\boldsymbol{\gamma } } { \sigma}^{'},p}^{ki}(\omega,{\bf{j } } ) & = & \sum_{{\boldsymbol{\delta } } { \sigma}^ { " } { \sigma}^ { ' '' } } \big [ 1- { \rho}^{k i } ( \omega,{\bf{j } } ) \big]_{{\boldsymbol{\gamma } } { \sigma}^{'},{\boldsymbol{\delta } } { \sigma}^ { " } } { \cal{s}}_{{\boldsymbol{\delta } } { \sigma}^{"}{\sigma}^{'''}}(\omega,{\bf{j } } ) { \cal{d}}_{{\boldsymbol{\delta } } { \sigma}^{'''},p}^{k * } \end{aligned}\ ] ]
in this appendix , we present the computation of the kernels @xmath156 and @xmath157 required in eq.([hw - holes - h ] ) as well as of the kernels @xmath179 and @xmath180 in eq.([hw - particles - h ] ) .
both @xmath156 and @xmath157 read | we present a symmetry - projected configuration mixing scheme to describe ground and excited states , with well defined quantum numbers , of the two - dimensional hubbard model with nearest - neighbor hopping and periodic boundary conditions .
results for the half - filled @xmath0 , @xmath1 , and @xmath2 lattices , as well as doped @xmath1 systems , compare well with available results , both exact and from other state - of - the - art approximations .
we report spectral functions and density of states obtained from a well - controlled ansatz for the @xmath3-electron system .
symmetry projected methods have been widely used for the many - body nuclear physics problem but have received little attention in the solid state community . given their relatively low ( mean - field ) computational cost and the high quality of results here reported , we believe that they deserve further scrutiny . |
over the last decade , extragalactic infrared ( ir ) surveys have become a key component in our quest to understand galaxy formation .
the reprocessing of stellar and nuclear radiations by dust and the resulting emission of this energy at thermal ir wavelengths ( 8@xmath1m@xmath7@xmath12@xmath71000@xmath1 m ) plays a major role in shaping the panchromatic appearance of star - forming galaxies and active galactic nuclei ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the direct characterization of the properties of these sources at ir wavelengths is therefore crucial for reliably estimating the fraction of their bolometric energy output absorbed by dust and converted into lower energy photons .
it also provides unique constraints on the nature of the different physical processes driving their on - going activity ( e.g , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the discovery of a large number of distant sources radiating copious amounts of energy in the ir ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) revealed that critical phases of the cosmic star formation and nuclear accretion history happened within dust - embedded environments hidden behind large amounts of absorbing material .
a significant fraction of the growth of structures over cosmic ages thus occurred in galaxies affected by substantial reddening and which characterization at shorter wavelengths ( e.g. , uv , optical ) can be severely biased if their extinction is not properly estimated . at large cosmological distances ,
highly obscured sources can even be barely detectable below @xmath01@xmath1 m and the observations in the ir or the millimeter are sometimes the only way to unveil the presence of on - going star formation and/or nuclear activity in these galaxies ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the infrared spectral energy distributions ( sed ) of star - forming galaxies and active galactic nuclei ( agn ) usually peak at @xmath080200@xmath1 m .
the most reliable estimate of the energy absorbed by dust in these sources would thus require observations in the far - ir or the submillimeter .
unfortunately though , data taken at such wavelengths have been so far hampered by source confusion as well as the very modest sensitivity of current detectors and the difficulty to access this frequency range from the ground .
for instance , observations of the distant universe performed at 70@xmath1 m and 160@xmath1 m with the _ spitzer space telescope _
@xcite or in the submillimeter with ground - based cameras of bolometers have been so far limited to the identification of galaxies at the very bright end of the luminosity function ( i.e. , l@xmath13@xmath310@xmath14l@xmath15 ) , with biases due to dust temperature selection effects ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
detecting the more typical high redshift sources at the peak of their ir sed has not been achieved yet , and our knowledge of the far - ir luminosity function of galaxies in the distant universe is still affected by substantial uncertainties . this strong limitation can be partially remedied with observations performed at mid - ir wavelengths ( i.e. , 5@xmath1m@xmath7@xmath12@xmath760@xmath1 m ) .
although cameras operating in the mid - ir are not as efficient as optical detectors they have been far more sensitive than in the far - ir , allowing the probe of the ir luminosity function down to much fainter levels .
the observed correlations between the mid - ir and the far - ir luminosities of star - forming galaxies allow reasonable estimates of their total ir luminosity ( l@xmath16=l@xmath17 ) with extrapolations from fluxes measured in the mid - ir @xcite . at such wavelengths , starbursts and agns exhibit also quite disctinct spectral signatures , providing invaluable clues to quantify the relative contribution of star - forming and nuclear activity occurring within individual sources .
observations of the deep universe with the 24@xmath1 m bandpass of the _ multi - band imaging and photometer for spitzer _ ( mips , * ? ? ?
* ) and with the _ infrared spectrograph _
( irs , * ? ? ?
* ) have led to tremendous progress in our understanding of the bolometric properties of distant sources ( e.g. , * ? ? ?
* ; * ? ? ?
? * ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the _ spitzer _
data provide tight constraints on the different mechanisms ( merging , feedback , environmental effects ) that regulate the activity of star formation and nuclear accretion in galaxies ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
they also enable the identification of a large number of obscured agns that must have played a key role in driving the coeval growth of bulges and super - massive black holes but whose contribution is usually underestimated by surveys performed at other wavelengths ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . in spite of this wealth of results recently achieved with _
at mid - ir wavelengths , a coherent and comprehensive picture of the deep ir universe is still missing though . given the variety of starburst and agn seds in the mid - ir ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , observations at these wavelengths are affected by quite complex @xmath8-correction effects and their sensitivity decreases rapidly beyond @xmath5@xmath01 .
hence , wide and shallow surveys are only sensitive to the very bright end of the mid - ir luminosity function , while the deeper observations usually performed at the expense of the covered area suffer from limited statistics .
also , the identification of mid - ir selected sources at high redshift can be extremely difficult given their faintness at optical wavelengths . as a result
the exact role that dusty galaxies played with respect to the less luminous but much more numerous sources at the faint end of the luminosity function is still under debate ( but see * ? ? ?
* ) . in the attempt to unify the different pictures of galaxy evolution recently achieved by optical and ir surveys we undertook deep 24@xmath1 m imaging over the whole area of the cosmos field using the mips instrument on - board _
spitzer_. the _ cosmic evolution survey _ ( cosmos ) is designed to probe the evolution of galaxies and agns up to @xmath5@xmath06 over a sky region large enough to address the role of environment and large scale structures @xcite .
it is based on deep multi - wavelength observations performed over @xmath02deg@xmath18 from the x - ray to radio wavelengths @xcite , with additional support from high spatial resolution imaging obtained with the _ hubble space telescope _
intensive spectroscopic follow - up was also carried out with the _ very large telescope _
@xcite while photometric redshifts with unprecedented accuracy were determined for more than @xmath0900000 @xmath6band selected galaxies in the field @xcite .
the quality of the data sets available in cosmos ( and in particular the depth with respect to the covered area ) is therefore very well suited for probing the multi - wavelength properties of galaxies with minimized cosmic variance and over a wide range of redshifts and luminosities . in this paper
we present our very first analysis of the mid - ir selected galaxy population detected down to @xmath2@xmath080@xmath1jy in cosmos .
data are described in sect.2 and our 24@xmath1 m source number counts are shown in sect.3 .
using the photometric redshifts of @xcite and @xcite we then determine the 24@xmath1 m redshift distributions in sect.4 and we analyze the number counts as a function of cosmic time in sect.5 to constrain the history and the build - up of the cosmic infrared background ( cib ) . in sect.6
we present comparisons between the observed 24@xmath1 m source redshift distributions and the predictions from both phenomenological and semi - analytical scenarios of galaxy formation .
the results are discussed in sect.7 , which allows us to quantify how well the contribution of dusty luminous sources as a general population is currently understood in models of galaxy evolution .
future papers will be dedicated to the analysis of the multi - wavelength properties of the cosmos mid - ir selected galaxies , their evolution , their clustering and their connection with sources selected with other criteria . throughout this paper
we assume a @xmath19cdm cosmology with h@xmath20=70 km s@xmath21mpc@xmath21 , @xmath22=0.3 and @xmath23 @xcite .
magnitudes are quoted in ablog@xmath24 [ f@xmath25(mjy)]+16.4 . ] unless the use of the vega system is explicitly stated ( e.g. , sect.[sec : compare_lacey ] ) .
the cosmos field was observed at 24@xmath1 m with the mips instrument on - board _
spitzer _ as part of two general observer programs ( pi d.sanders ) .
the first observations ( go2 , pid 20070 ) were carried out in january 2006 with the mips medium scan mode and two scan passes per sky pixel . as described in more detail by @xcite this shallow imaging covered a total area of @xmath04deg@xmath18 centered on cosmos , with a median integration time of 80s per pixel .
this first program was also used to test in a small region of the field the feasibility to improve the 24@xmath1 m sensitivity with longer integrations despite the cosmos mid - ir background being @xmath02@xmath26 higher than in other typical `` cosmological '' fields ( e.g. , _ hubble / chandra deep fields _ , lockman hole ) . after validating this additional step ,
deeper imaging was performed in 2007 over the whole cosmos field as part of a second program using the mips slow scan mode ( go3 , pid 30143 ) .
these data yielded a second and independent 24@xmath1 m coverage over a total area of @xmath33deg@xmath18 . across the nominal 2deg@xmath18 of cosmos the combination of the go2 and go3 observations results in a median integration time of @xmath03360s per sky pixel .
note that the mips scan mode provides simultaneous imaging at 24 , 70 and 160@xmath1 m ; the observations of the cosmos field at 70/160@xmath1 m are presented by frayer et al .
( submitted to aj ) .
all the data set was reduced following the procedure outlined in @xcite . for each observing campaign a background pattern derived for every scan mirror position of the instrument was scaled and subtracted from all individual frames .
these background - subtracted images were then interpolated to a common grid of pixels using the mopex package @xcite .
they were finally coadded with our own idl routines using the median of all the pixels associated with each sky position .
we stress that the use of a median combination was necessary in order to reject moving sources present in the field at the time of the observations . a forthcoming paper ( aussel et al . in prep .
) will describe in more detail the data reduction .
it will also provide a list of the 24@xmath1m detected asteroids in cosmos .
the 24@xmath1 m sources in our mosaic were detected using the automatic procedure of the sextractor software @xcite and we measured their flux densities with the point spread function ( psf ) fitting technique of the daophot package @xcite .
given the full - width at half maximum ( fwhm ) of the mips-24@xmath1 m psf ( @xmath06 ) , most of the sources are not resolved in our data . in deep and crowded images such as the ones typically obtained in the mips deep surveys , the psf fitting thus provides more reliable flux measurements than aperture photometry .
furthermore , daophot performs simultaneous fits to multiple objects , which ensures a proper separation of blended sources especially when the fluxes are close to the confusion limit .
no prior information from any other wavelength was used for the source detection itself .
however , several psf fitting iterations were necessary to achieve a clean extraction of the blended sources that were not properly identified by sextractor in the input 24@xmath1 m image . in brief ,
these blended cases were located ( with sextractor again ) as positive detections in the residual mosaic obtained after the first psf fitting .
their positions were added to the initial list of 24@xmath1 m sources determined with sextractor and another extraction of the initial cosmos 24@xmath1 m image was then performed .
three iterations of this process were carried out , until the residual mosaic was cleaned from any positive signal above the detection threshold chosen for sextractor .
the result of this psf fitting is illustrated in figure [ fig : daophot ] .
it shows a sub - region of our cosmos 24@xmath1 m mosaic along with the corresponding residual image obtained after subtracting all sources extracted with daophot . to run this psf fitting we built an empirical psf from the brightest point sources identified in the mosaic and we subsequently fit this psf to every single object detected by sextractor .
daophot performs this fit by scaling the empirical psf within a fixed circular aperture centered on the source and it provides the flux of the fitted psf _ enclosed within this aperture_. therefore a correction must be applied to account for the spatially - extended wings of the psf lying beyond the radius of the chosen aperture .
this correction was determined using the mips-24@xmath1 m stiny tim psf models provided by the _
spitzer science center_. these models allow to account for the fraction of energy lying at large radii where the signal to noise in our data is not large enough to constrain the exact profile of the point spread function .
relying on such corrections we believe that our flux absolute calibration should be accurate within a few percent , which is consistent with the results obtained by @xcite and @xcite as part of the calibration of the mips 24@xmath1 m detector . following the convention adopted by the
_ spitzer science center _ we assumed a stellar 10000k black - body spectrum as the reference underlying sed for the 24@xmath1 m flux density measurements .
we tested this calibration by cross - correlating our source list with the stars of the _ two micron all sky survey _ ( 2mass ) catalog @xcite .
our comparison did not reveal any systematic deviation from the relation @xmath27 = 0 $ ] ( vega , with [ 24 ] denoting the equivalent magnitude measured in the 24@xmath1 m mips bandpass ) , which is expected for most of the stellar spectral types when no dust emission excess is detected above the photospheric stellar emission ( e.g. , * ? ? ?
this cross - correlation also allowed us to register the absolute astrometric calibration of our 24@xmath1 m mosaic to the 2mass astrometric system .
this was performed by applying a systematic offset of @[email protected] to the astrometry derived from the _ spitzer _ data products .
the noise in the final 24@xmath1 m mosaic was characterized using two different methods .
first , we computed the dispersion of a set of random flux measurements performed over blank field regions of the cosmos 24@xmath1 m image within the same aperture as the one used for the psf fitting .
similar to the photometry measurements described above a flux correction was applied to account for the extent of the psf lying outside of the aperture , which led to an equivalent sensitivity limit of 1@xmath30=18@xmath1jy .
second , we also performed a number of simulations adding artificial point - like sources in blank field regions of the mosaic .
we measured the fluxes of these sources with the psf fitting of daophot and the distribution of the differences between their measured and input fluxes was fitted with a gaussian function .
our best fit was obtained for a 1@xmath30 standard deviation varying between 14@xmath1jy and 16@xmath1jy depending on the range of input fluxes . using these simulations we finally examined the robustness of the error measurements provided by our psf fitting by comparing the daophot uncertainties obtained for the simulated point sources with the errors directly estimated from the differences between their measured and input fluxes .
we applied a systematic scaling of 3.8 to all the daophot errors so as to get 68% of the distribution of these differences within the daophot 1@xmath30 uncertainties .
the completeness of our 24@xmath1 m source extraction was determined using the same point source simulations as the ones described in the previous paragraph .
we inserted within the mosaic a set of empirical psfs that were scaled to provide a flux distribution representative of the one characterizing the 24@xmath1 m extragalactic source population .
these psfs were subsequently extracted following the same method as the one employed to detect and to measure the flux of the sources already present in the 24@xmath1 m image .
we found that our source extraction is more than 90% complete above a 24@xmath1 m flux @xmath2@xmath080@xmath1jy , which should thus be considered as a safe and conservative flux limit for any analysis requiring an unbiased 24@xmath1 m selection of sources in the cosmos field . according to the simulations though ,
our psf fitting is still reliable down to fluxes as faint as @xmath060@xmath1jy and in spite of a lower completeness ( @xmath075% ) our source extraction should suffer from a relatively small contamination by fake objects at these flux levels .
based on this characterization we defined our final cosmos mips-24@xmath1 m catalog as the population of 24@xmath1 m sources brighter than 60@xmath1jy lying within the area covered by the cosmos photometric redshift catalog of ( * ? ? ?
* , see section [ sec : photoz ] ) but outside of the masked regions defined from the subaru optical observations of the cosmos field .
these masked areas cover a surface as large as 0.26deg@xmath18 but the optical photometry in these regions is affected by larger uncertainties due to the presence of very bright or even saturated objects .
in particular , the photometric redshifts in these areas are less accurate .
keeping these regions would thus compromise our analysis of the mips - selected galaxies given the large fraction of 24@xmath1 m sources associated with faint optical counterparts .
our 24@xmath1 m catalog represents a total of 39413 sources covering an effective area of 1.68deg@xmath18 . in the remaining of this section 2
we will describe the identification of the 24@xmath1 m source counterparts down to @xmath2@xmath3160@xmath1jy .
this characterization will supplement other studies carried out by our team using selections performed at other wavelengths .
however , the analysis reported in some other sections of this current paper assume an unbiased 24@xmath1 m selection and only sources above 80@xmath1jy will be considered in this case . to ensure the most reliable and most complete identification of the 24@xmath1 m source counterparts , we first correlated the 24@xmath1 m data with the @xmath32band cosmos catalog of mccracken et al .
( submitted to apj ) .
as we will see later a non - negligible fraction of the 24@xmath1 m sources at @xmath5@xmath31 are associated with extremely faint optical counterparts .
a direct cross - correlation between the mips-24@xmath1 m and the optical catalogs of cosmos would thus result in a lower rate of identifications .
furthermore , the larger uncertainties affecting the determination of source centroids at faint 24@xmath1 m fluxes , the width of the 24@xmath1 m psf and the very high density of optical sources in our data ( e.g. , @xmath045 galaxies arcmin@xmath33 for @xmath34@xmath3525mag in the subaru cosmos observations , @xcite ) could lead to lots of wrong associations if these identifications had to rely on a direct match between the 24@xmath1 m and optical source catalogs . on the contrary , the cosmos @xmath32band catalog is deep enough ( @[email protected] at 5@xmath30 , mccracken et al . ) to detect counterparts for the majority of the 24@xmath1 m selected sources down to our mips sensitivity limit .
the density of @xmath32band sources is also substantially smaller ( @xmath07 galaxies arcmin@xmath33 for @xmath36@xmath3522mag ) than in the optical bands and the near - ir wavelengths are closer " to the 24@xmath1 m bandpass , which minimizes the risk of wrong associations between mid - ir and @xmath32-selected sources . finally we note that these @xmath32band observations of cosmos were executed under excellent seeing conditions ( @xmath00.7 at 2.2@xmath1 m ) .
compared to the other cosmos near - ir data taken with e.g. , the irac instrument ( [email protected] at 3.6@xmath1 m , @xcite ) , the psf in the @xmath32band image is much narrower and allows more robust identifications in the case of sources blended at 24@xmath1 m .
we correlated the 24@xmath1 m and @xmath32band catalogs with a matching radius of 2 .
we found that this distance was large enough to enable the identification of most of the 24@xmath1 m sources , taking into account the width of the 24@xmath1 m psf ( fwhm@xmath06 ) and the uncertainty on the astrometric position of each single 24@xmath1 m source .
this radius was nonetheless sufficiently small to minimize the fraction of multiple matches as well as the association between objects randomly - aligned on the sky . among the 39413 mips24@xmath1 m sources of the catalog
, we identified a single @xmath32band counterpart for 33146 objects ( 84% of the sample ) while two possible matches were obtained for 4470 other 24@xmath1 m detections ( @xmath011.5% ) . for theses cases we kept
the closest possible counterpart when its centroid was at least twice closer to the 24@xmath1 m source than the second possible match ( 2320 objects ) while additional priors from the cosmos irac-3.6@xmath1 m observations were considered to assign the most likely counterpart to the other 24@xmath1 m detections .
no association was found for the remaining 1797 sources ( 4.5% of the catalog ) .
a cross - correlation between these unidentified sources and our 3.6@xmath1 m catalog using the same matching radius revealed irac counterparts for 437 objects , a rare population that must be characterized by extremely steep seds between 2@xmath1 m and 24@xmath1 m .
for the other sources a visual inspection of our data showed that some of them may have counterparts either fainter than the detection thresholds of our @xmath32/irac cosmos catalogs or with a centroid located slightly beyond our 2 matching radius .
many of the unidentified objects have also a signal / noise below 5@xmath30 at 24@xmath1 m . at the faintest 24@xmath1 m flux
leves some could originate from a small contamination of our catalog by fake sources .
it should however not significantly bias our subsequent results given the small fraction that they represent among the whole sample of 24@xmath1 m sources . in summary
we identified secure @xmath32band counterparts for @xmath090% of the cosmos 24@xmath1 m catalog down to 60@xmath1jy , while less robust identifications were still found for another 5.5% of the mips-24@xmath1 m source population .
finally we note that this cross - identification with near - ir data also revealed 53 pairs of 24@xmath1 m sources associated with the same @xmath32band counterpart .
a careful inspection of the 24@xmath1 m mosaic revealed that they correspond to single objects incorrectly deblended by the psf fitting performed in the mips data .
these `` double '' components were merged together to form a unique 24@xmath1 m source , reducing the mips - selected catalog to 39360 objects .
the distance of the 24@xmath1 m sources was characterized using the photometric redshifts derived by @xcite and @xcite for the optically and x - ray selected sources of the cosmos field .
these photometric redshifts were determined with deep photometry ( e.g. , capak et al .
2009 , in preparation ) performed in a total of 30 large , medium and narrow band filters over a wavelength range covering the far - uv at 1550 up to the mid - ir at 8.0@xmath1 m .
this large amount of data and the high quality of the observations used in their work resulted in a substantial improvement of photometric redshift accuracy with respect to the determination of redshifts performed in other fields .
the catalog of ( * ? ? ? * , hereafter i09 ) provides redshifts for 933789 sources selected at @[email protected] from the subaru / suprime cam observations of cosmos , with a dispersion as small as @xmath37 for @xmath34@xmath3524mag and @[email protected] . this unprecedented accuracy in photometric redshift determinations
was mostly achieved thanks to the high sampling of the observed spectral energy distribution of sources around the position of their balmer break as well as a new method accounting for the contribution of emission lines to their measured flux densities . as shown by @xcite the uncertainties affecting their redshift estimates depend primarily on the redshift and the apparent @xmath6band magnitude of sources ( see their figure 9 ) , with errors naturally increasing for fainter and more distant galaxies .
however a comparison with faint spectroscopic samples obtained in the cosmos field ( @xcite ) revealed a dispersion @xmath38 as low as 0.06 for sources with 23mag@xmath35@xmath34@xmath3525mag at 1.5@xmath7@xmath5@xmath73 .
these small uncertainties will be critical for our analysis of the mips - selected galaxy population given the large fraction of 24@xmath1 m sources associated with faint ( i.e. , @xmath34@xmath324mag ) optical counterparts at @xmath5@xmath31 .
the redshifts provided by i09 were determined using a library of star - forming and passive galaxy templates and they are not adapted to galaxies dominated in their optical / near - ir continuum by the contribution of an active galactic nucleus ( such as the type 1 agns and optically - luminous quasars ) .
this issue must be treated with care given the non - negligible contribution of agns to ir - selected galaxy samples ( e.g. , * ? ? ?
* ; * ? ? ?
the redshifts of the 24@xmath1 m sources also identified with an x - ray counterpart were therefore supplemented using the catalog of @xcite , which provides photometric redshifts for 1542 x - ray selected sources detected in the @xmath39-newton observations of cosmos @xcite .
@xcite used an optimized set of templates with hybrid combinations of agn and non - active galaxy seds , which allowed them to derive photometric redshifts with an accuracy comparable to that achieved by @xcite .
they also made an extensive use of multi - epoch observations of cosmos to correct for the time - variability of type 1 agns .
the photometric redshifts of the full @xmath39-selected sample mostly lie at 0.5@xmath7@[email protected] but extend up to @xmath5@xmath03 , with an accuracy as good as @xmath40 for faint sources ( @[email protected] ) at @xmath5@xmath01 @xcite .
we stress that the redshifts of i09 were also used in the case of the 24@xmath1 m sources associated with obscured nuclei ( such as sources characterized by a rising power - law continuum in the irac bands ) but with no x - ray detection .
their emission at optical wavelengths is usually dominated by the contribution of their host galaxy and the photometric redshifts of i09 should not be significantly less accurate than those determined for star - forming sources .
this argument is actually reinforced by the rather good agreement that we observed betweeen the redshifts of @xcite and those of @xcite for the type 2 agns common to both samples . to derive the photometric redshifts of the 24@xmath1 m sources we first correlated the list of their @xmath32band counterparts with the catalog of @xcite using a matching radius of 1 .
we found a single optical counterpart at @[email protected] for 35347 @xmath32band sources ( 90% of the full 24@xmath1 m source catalog ) .
double matches were obtained for 361 other @xmath32band objects . in those cases , we noticed that the difference between the matching distances of the first and the second closest matches was large enough to consider the first one as the most likely optical counterpart .
the catalog of i09 is limited to @[email protected] and no optical match could be found for the remaining 1855 @xmath32band sources ( i.e. , @[email protected] ) . from this match between the @xmath32 and the optical bands we
then identified 1129 optical counterparts also associated with a source detected by @xmath39 . among those ,
1111 are included in the analysis presented by @xcite and we thus assigned to these objects the photometric redshifts provided in their catalog of x - ray sources .
the identification of the optical counterparts associated with the remaining 18 @xmath39 objects is considered less secure @xcite . for these objects we kept the original photometric redshift derived by i09 .
finally we considered the list of the 1797 mips-24@xmath1 m sources for which we did not find any @xmath32band counterpart ( see sect.[sec : match ] ) .
we correlated this list with the catalog of @xcite using a matching radius of 2 , assuming that some of these sources could have counterparts detected in the very deep subaru @xmath6band observations of cosmos @xcite but not reported in our @xmath32band catalog . among these 24@xmath1 m sources ,
667 objects were identified with one single counterpart in the optical catalog .
we note that some of them seem qto be trully detected in the @xmath32band image of cosmos ( mccracken et al . ) , although their associated signal to noise is very low .
as previously reported some of them also have counterparts at 3.6@xmath1 m in the irac data .
these by - eye " detections gave us further confidence in the reliability of their identification in the @xmath6band catalog of i09 .
to summarize , we derived photometric redshifts for @xmath092% objects of the initial 24@xmath1 m catalog selected above 60@xmath1jy .
sources with counterparts in the @xmath32band but no identification at @[email protected] account for a fraction of 5% , while 3% of the catalog correspond to 24@xmath1 m sources with no identification in the optical nor in the near - ir .
if we rather consider a selection above 80@xmath1jy , we find a total of 29410 sources in the initial mips-24@xmath1 m catalog and 94% of these objects are identified with a redshift .
independently of their possible association with a @xmath32band or 3.6@xmath1 m counterpart , the 24@xmath1 m sources with no redshift are very faint at optical wavelengths .
previous studies have shown that they mostly correspond to highly - obscured luminous galaxies at 1@xmath7@xmath5@xmath73 rather than low redshift sources populating the faint end of the luminosity function ( e.g. , * ? ? ?
* ; * ? ? ?
* ) . by construction
an infrared - selected sample of sources is biased toward dusty galaxies .
the determination of photometric redshifts for these objects is subject to more uncertainties if the effect of dust extinction is not properly taken into account .
furthermore , a large fraction of mid - ir sources in the distant universe are associated with very faint optical counterparts also characterized by less accurate redshifts .
the uncertainties affecting the photometric redshifts of the cosmos mips-24@xmath1 m sources should thus be carefully assessed before analyzing our sample further .
@xcite presented a detailed comparison between their photometric redshifts and the spectroscopic redshifts of mips-24@xmath1 m sources determined by @xcite and kartaltepe et al .
( in prep . ) with spectroscopic follow - ups of the cosmos field .
up to @[email protected] i09 found median dispersions @xmath38 as low as 0.01 and 0.05 for 24@xmath1 m sources with 22.5mag@xmath35@xmath34@xmath3524mag and 24mag@xmath35@xmath34@xmath3525mag respectively . for mid - ir galaxies at higher redshifts and/or with fainter optical counterparts
there is however no spectroscopic sample currently available in cosmos to allow such comparison . to characterize the uncertainties of their photometric redshifts we relied on their associated probability distribution functions
pdf(z ) as provided by i09 . for
their optically - selected sample @xcite showed that the 1@xmath30 uncertainties and negative 1@xmath41 photometric redshift uncertainties using the relation @xmath42 , with @xmath43 the most likely photometric redshift estimated from the minimum of the merit function @xmath44 .
for each object the final 1@xmath30 uncertainty is taken as the maximum between 1@xmath45 and 1@xmath41 . ]
derived from these pdf(z ) provide a reliable estimate of the dispersion @xmath38 obtained from the comparison with spectroscopic redshifts . in figure [ fig : z_uncert ] we show these 1@xmath30 uncertainties restricted to the mips-24@xmath1 m sources as a function of their photometric redshifts , along with their median value calculated in three different bins of @xmath6band magnitude . up to the highest redshift bins characterizing our sample ( i.e. , @xmath5@xmath73 ) we see that the photometric redshifts are still quite robust up to @xmath34@xmath3525mag ( 1@[email protected] ) . at fainter magnitudes ,
the uncertainties are somehow larger but the dispersions are still as low as 1@[email protected] up to @xmath5@xmath02 where most of the faint sources in our sample are located ( see sect.3 ) .
the larger uncertainties that may affect the photometric redshifts of faint optical sources can be produced by the presence of multiple peaks in their pdf(z ) . to quantify the contribution of these potential catastrophic failures we considered the fraction of mips-24@xmath1 m sources with a pdf(@xmath5 ) showing a secondary local maximum stronger than 10% of the value of its first peak .
these fractions are illustrated as a function of redshift and for different bins of optical magnitudes in figure [ fig : double_z ] .
they show that up to @[email protected] the number of possible outliers is negligible ( @xmath75% ) for most of our sample .
fractions reaching @xmath025% may be observed at @xmath34@xmath924mag and 0.8@xmath7@[email protected] , yet these estimates only affect a very small number of sources ( see bottom panel of figure [ fig : double_z ] ) .
the differential 24@xmath1 m source number counts normalized to the euclidean slope and derived from our cosmos data are shown in figure [ fig : counts ] .
they are compared with the counts obtained by @xcite based on the mips _ guaranteed time observer _ programs and by @xcite using deep observations in the elais - n1 field . within the uncertainties
our results show a quite good agreement with these two other surveys , and the variations between the three determinations may result from the effect of cosmic variance between the different fields observed with _
spitzer_. at fluxes brighter than @xmath03mjy though , the catalog that we derived outside of the optically - masked regions of cosmos underestimates the density of sources compared to the counts derived over the whole area of the field .
this is not surprising given that many of the 24@xmath1 m bright objects are associated with optically - bright counterparts in these masked regions .
such objects are either stars or very low redshift galaxies , which should not affect too much our analysis of the general extragalactic 24@xmath1 m source population .
similar to what was observed in other _
24@xmath1 m deep surveys @xcite , the counts measured from our cosmos data show a prominent break at flux densities @[email protected] .
our source extraction is highly reliable at these flux levels and this feature can not be artificially produced by an incompleteness of our 24@xmath1 m catalog . such a rapid turn - over " in the euclidean - normalized source number counts implies their rapid convergence at fainter fluxes ( @[email protected] ) .
hence , a large fraction of the 24@xmath1 m background must be produced by sources readily detected in our cosmos mips data .
for instance @xcite assumed a straight extrapolation of the faint - end slope of the counts measured down to @xmath2@xmath060@xmath1jy and they argued that mips-24@xmath1 m sources detected above 60@xmath1jy resolve as much as @xmath070% of the 24@xmath1 m background intensity . in sect.[sec : background ] we will use our photometric redshift identifications in cosmos to explore the build - up of this mid - ir background light _ as a function of cosmic time_. finally , we note that the persistent shoulder " appearing at @xmath2@xmath03mjy in the normalized counts that were derived from large area surveys such as the _ spitzer wide - area infrared extragalactic survey _ ( swire , * ? ? ?
* ) is also visible in cosmos ( although our data alone can not formally exclude an effect of cosmic variance ) . to our knowledge this characteristics of the 24@xmath1 m counts has not been explained so far .
we now explore the redshift distribution of the cosmos mips-24@xmath1 m sources based on the identifications of their photometric redshifts described in sect.[sec : photoz ] .
we define @xmath46 the redshift distribution of the 24@xmath1m selected sources with @xmath2@xmath9@xmath47 and associated with optical counterparts at @xmath6@xmath35@xmath48 , and we note @xmath49 the redshift distribution of all the cosmos optically selected sources with @xmath6@xmath35@xmath48 . given the relatively small number of agn - dominated sources in the general population of optically - selected galaxies@xmath3524mag are detected in the @xmath39 observations of cosmos down to f@xmath50 ergcm@xmath33s@xmath21 .
] , @xmath49 was computed directly using the photometric redshift catalog of @xcite . in figure
[ fig : histo_z ] we show the redshift distributions @xmath46 calculated down to flux limits of 80@xmath1jy and 300@xmath1jy at 24@xmath1 m and for @xmath6band magnitudes limits of 24 , 25 and 26.5mag .
sources brighter than @xmath34@xmath3120mag were excluded from all these estimates to minimize the effect of large scale structures at low redshift . our redshift distribution at @[email protected] should actually be treated with caution given the limited comoving volume sampled by our survey at these small distances .
we see that to the depth of our mips observations ( @xmath2@xmath980@xmath1jy ) almost all 24@xmath1 m sources at @xmath5@xmath71 are associated with optical counterparts brighter than @xmath34@xmath024mag .
our photometric redshifts are very accurate at these optical magnitudes . given the statistics of our sample beyond @[email protected] our redshift distributions at 0.5@xmath7@xmath5@xmath71 should thus be highly robust independently of the 24@xmath1 m flux limit considered above 80@xmath1jy . at @xmath5@xmath31 ,
the contribution of fainter sources becomes more significant .
for instance , 45% and 33% of the mips-24@xmath1 m sources with respectively @xmath2@xmath980@xmath1jy and @xmath2@xmath9300@xmath1jy at @xmath5@xmath91 are associated with counterparts fainter than @xmath34@xmath3124mag . as we showed in sect.[sec : photoz ] their photometric redshifts are still quite reliable though ( see figures [ fig : z_uncert ] & [ fig : double_z ] ) .
our redshift distribution estimates are mainly affected by the errors on the cosmos photometric redshifts , the cosmic variance and the errors on the 24@xmath1 m flux densities . to quantify the impact of all these uncertainties we first divided the cosmos field into 4 different regions of similar areas ( @xmath00.4deg@xmath18 ) and we derived the possible range of redshift distributions in each region with a set of 5000 monte carlo simulations .
for each of these simulations we considered our _ full _
24@xmath1 m catalog down to @xmath2@xmath3160@xmath1jy and we modified the 24@xmath1 m flux and redshift of each single source according to their 24@xmath1 m flux uncertainty and their 1@xmath30 photometric redshift dispersion characterized in sect.2 . assuming a given 24@xmath1 m flux limit , an average and independent redshift distribution was estimated for the 4 different regions and the dispersion between these 4 estimates was taken as the final uncertainty affecting the global 24@xmath1 m source redshift distribution .
our different estimates are reported in table1 .
they are also shown in the bottom panels of figure [ fig : histo_z ] for 24@xmath1 m flux limits of 80@xmath1jy and 300@xmath1jy .
they suggest that our global redshift distributions inferred from the 1.68deg@xmath52 area observed with mips should be robust within @xmath020% up to @xmath5@xmath03 .
it is important to note that the redshift distribution characterizing the galaxy population selected in the mid - ir differs quite significantly from the one observed at optical wavelengths .
figure [ fig : i_dist ] illustrates how the optically and mid - ir selected sources differ in their occupation of the redshift/@xmath6band magnitude parameter space .
for instance we find that only @xmath027% of the optically - selected galaxies with @xmath34@xmath3524mag are located at @xmath5@xmath31 @xcite while this fraction rises to more than 50% for the 24@xmath1 m sources with @xmath2@xmath980@xmath1jy .
this large proportion of high - redshift mid - ir galaxies is mostly located at 1@xmath7@[email protected] and it has already been seen in other mips high redshift surveys ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
? * ; * ? ? ?
it results primarily from the strong evolution that infrared luminous galaxies have undergone with time ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
given the broad - band aromatic features commonly observed at 6@xmath1m@xmath7@xmath12@xmath720@xmath1 m in the spectra of star - forming galaxies it may also be due to galaxy @xmath8-corrections boosting the mid - ir detectability of sources at 1.5@xmath7@[email protected] when these features are redshifted into the mips-24@xmath1 m bandpass ( e.g. , * ? ? ?
lccc 0.0@xmath35@[email protected] & 679@xmath5391 & 310@xmath5363 & 110@xmath5320 + 0.2@xmath35@[email protected] & 4722@xmath53203 & 2356@xmath5364 & 697@xmath5362 + 0.4@xmath35@[email protected] & 6317@xmath53275 & 3296@xmath5371 & 997@xmath5344 + 0.6@xmath35@[email protected] & 9991@xmath531328 & 4964@xmath53796 & 1604@xmath53353 + 0.8@xmath35@[email protected] & 14055@xmath532001 & 7144@xmath531193 & 2148@xmath53377 + 1.0@xmath35@[email protected] & 8981@xmath53904 & 3743@xmath53471 & 865@xmath53117 + 1.2@xmath35@[email protected] & 5158@xmath53465 & 1704@xmath53167 & 367@xmath5354 + 1.4@xmath35@[email protected] & 6360@xmath53144 & 2356@xmath53155 & 531@xmath5378 + 1.6@xmath35@[email protected] & 4672@xmath53288 & 1993@xmath53117 & 452@xmath5367 + 1.8@xmath35@[email protected] & 4735@xmath53264 & 2167@xmath53203 & 497@xmath5330 + 2.0@xmath35@[email protected] & 3753@xmath53203 & 1683@xmath53150 & 411@xmath5362 + 2.2@xmath35@[email protected] & 1933@xmath5365 & 815@xmath5352 & 205@xmath5313 + 2.4@xmath35@[email protected] & 1423@xmath5393 & 598@xmath5351 & 167@xmath5327 + 2.6@xmath35@[email protected] & 978@xmath53110 & 406@xmath5354 & 126@xmath5331 + 2.8@xmath35@[email protected] & 535@xmath5389 & 222@xmath5332 & 84@xmath5321 + 3.0@xmath35@[email protected] & 335@xmath5334 & 125@xmath538 & 49@xmath536 + 3.2@xmath35@[email protected] & 203@xmath5330 & 103@xmath5317 & 43@xmath535 + 3.4@xmath35@[email protected] & 138@xmath537 & 67@xmath5314 & 36@xmath5318 + 3.6@xmath35@[email protected] & 73@xmath537 & 42@xmath538 & 18@xmath537 + 3.8@xmath35@[email protected] & 53@xmath538 & 31@xmath539 & 16@xmath535 + 4.0@xmath35@[email protected] & 46@xmath5310 & 34@xmath537 & 16@xmath533 + to better characterize this large contribution of mid - ir selected sources at high redshift , we considered the relative distribution @xmath54
. first , this quantity allows us to check if any feature of the mips-24@xmath1 m galaxy redshift distribution does not artificially result from a bias already present in the underlying optically selected catalog of photometric redshifts . besides
, it can be seen as the fraction of optical sources with @xmath6@xmath35@xmath48 and with a mid - ir detection above a 24@xmath1 m flux limit @xmath47 .
our different estimates of @xmath55 are illustrated in the bottom panels of figure [ fig : i_dist ] for the same 24@xmath1 m flux and @xmath6band magnitude limits as the ones considered previously .
we see that independently of these limits the fraction of optical sources detected in the mid - ir progressively rises with redshift , which must be related to the well - known increase of star formation density with lookback time .
the fraction peaks at @xmath5@xmath02 before gradually declining at higher redshifts , probably because of the decreasing number of 24@xmath1 m detections at @xmath5@xmath32 .
more interestingly though , we observe a substantial drop of @xmath55 between @xmath5@xmath01 and @[email protected] followed by a steep increase from @[email protected] and @xmath5@xmath02 .
this effect is clearly apparent even with a cut at @xmath6@xmath3524mag .
therefore it can not originate from systematics affecting our photometric redshifts .
it is rather due to the break characterizing the 24@xmath1 m source redshift distributions around @[email protected] and that does not appear in the smooth distributions observed at optical wavelengths ( e.g. , i09 ) .
note that this break also seems to be more prominent as the 24@xmath1 m flux limit increases .
the redshift distribution derived at @xmath2@xmath9300@xmath1jy looks even like a two - component function with a main and narrow peak of sources at @xmath5@xmath01 followed by a secondary but broader peak at 1.5@xmath7@[email protected] ( e.g. , figure [ fig : histo_z ] , right panel ) .
this characteristic of the 24@xmath1 m redshift distribution could be artificially produced by a relative increase of mid - ir detections at @xmath5@xmath02 due to the rest - frame 7.7@xmath1 m aromatic feature enhancing the apparent 24@xmath1 m flux of star - forming galaxies when it is redshifted into the mips-24@xmath1 m bandpass .
on the other hand it may also reflect a smaller rate of detections at @[email protected] given the possible effect of the 9.7@xmath1 m silicate absorption .
if this break is trully produced by @xmath8-correction effects , the stronger effect observed at brighter 24@xmath1 m flux might also suggest more prominent features in the more luminous mid - ir selected galaxies .
the cosmic infrared background ( cib ) refers to the extragalactic background light ( ebl ) detected between @xmath08@xmath1 m and @xmath01000@xmath1 m .
it can be seen as a fossil record of all the radiations emitted at ir wavelengths along the process of galaxy formation ( e.g. , * ? ? ?
its integrated energy power is comparable to that of the background produced by the stellar component of galaxies in the optical / near - ir wavelength range @xcite .
understanding how the different populations of distant sources contributed to the build - up of the cib as a function of redshift can thus provide significant constraints to our general understanding of the cosmic growth of structures . using stacking analysis at 70@xmath1 m and
160@xmath1 m @xcite showed that 24@xmath1m selected sources with @xmath2@xmath960@xmath1jy contribute more than 70% of the cib .
the large statistics available in the cosmos field and the high rate ( @xmath095% ) of redshift identifications in our 24@xmath1 m catalog give us therefore a unique opportunity to resolve the different contributions of galaxies to the build - up of a large fraction of the ebl . in this current paper
we only explore the history of the monochromatic background light at the wavelength of our mips observations , using the differential 24@xmath1 m source number counts as a function of lookback time . resolving the history of mid - ir source number counts
already provides tight constraints on the different models of galaxy formation , especially when they correctly reproduce integrated measures like the total counts or the diffuse background emission but differ in their predictions of the flux / redshift distribution of galaxies responsible for these quantities .
this issue will actually be discussed in sect.[sec : compare_model ] where we will compare our data with the predictions of different models to quantify their reliability in the description of ir galaxy evolution .
we defer to a subsequent publication the analysis of the total background intensity produced by 24@xmath1m selected galaxies over the entire spectral range of the cib . to characterize this decomposition of the 24@xmath1 m counts as a function of redshift we first defined @xmath56 and @xmath57 the differential counts of the 24@xmath1 m sources identified up to a given redshift @xmath58 or up to a given @xmath6band magnitude @xmath48 , respectively . in figure
[ fig : counts_imag ] we illustrate these quantities computed for @xmath58=0.5 , 1 , 1.5 , 2 and 2.5 ( left panel ) and for @xmath48=21 , 22 , 23 , 24 , 25 and 26.5mag ( right panel ) . for comparison
we also show the total number counts derived from the cosmos field and discussed in sect.[sec : counts ] ( see also figure [ fig : counts ] ) .
we see that the brightest 24@xmath1 m sources are mostly associated with low redshift galaxies characterized by bright optical counterparts . however
a non - negligible fraction of the mips detections at @xmath2@xmath31mjy are also located at much larger distances .
albeit rare in space density , these sources compose the bright end of the ir luminosity function at high redshift and an accurate estimate of their contribution can only be determined with large area surveys . at fainter fluxes
, we also note that a substantial fraction of the 24@xmath1 m sources are associated with faint and distant optical counterparts at @xmath34@xmath924mag .
this conclusion reinforces that discussed above for the redshift distributions .
it emphasizes once again the difficulties to achieve a complete characterization of the 24@xmath1m - selected galaxy population with current optical / near - ir follow - ups , and in particular to obtain a full census of the sources at @[email protected] making the bulk of the mid - ir background light . at such faint optical magnitudes
, spectroscopy has been very challenging so far while photometric redshifts also need to be carefully treated given the rapid increase of their uncertainties beyond @xmath59@xmath025mag ( e.g. , i09 ) . to also account for the contribution of galaxies fainter than our 24@xmath1 m detection limit we extrapolated the 24@xmath1 m number counts at faint fluxes following the same approach as considered by @xcite .
we used power - law functions @xmath60@xmath61@xmath62@xmath63 for the full 24@xmath1 m catalog and for the different sub - samples of 24@xmath1 m sources selected per bin of redshift or @xmath6band magnitude as described above . for each of these extrapolations the power - law index @xmath64 was only constrained with the counts observed above @xmath2=80@xmath1jy to minimize the impact of the incompleteness of our source extraction below this limit . for the total counts we derived @[email protected] .
this is similar to the results obtained by @xcite , @xcite and @xcite who measured indexes between 1.5 and 1.6 using sources detected down to @xmath030@xmath1jy in the elais - n1 field .
this good agreement suggests that the extrapolations obtained for the other sub - samples should also provide a fair representation of the flux and redshift distributions characterizing the mips sources below our detection limit at 24@xmath1 m . finally , we integrated the total extrapolated number counts to estimate the total 24@xmath1 m background light ( referred as @xmath65 hereafter ) .
we also integrated the extrapolated functions @xmath66 for each sub - sample considered above to define @xmath67 and @xmath68 as the background light produced above a 24@xmath1 m flux @xmath69 and up to the redshift @xmath58 ( eq.1 ) or by sources brighter than the @xmath6band magnitude @xmath48 ( eq.2 ) whatever the choice of the redshift or the @xmath6band magnitude limit . ] .
these two quantities are respectively illustrated in figures [ fig : resolv_imag ] & [ fig : resolv_z ] for 24@xmath1 m fluxes of 0.06 , 0.08 , 0.15 , 0.3 and 1mjy as well as for @xmath69=0 .
they were normalized to the total 24@xmath1 m background intensity @xmath65 determined from the extrapolation of the total source number counts .
they can thus be read as the fraction of the 24@xmath1 m background resolved by mips - selected sources as a function of their flux @xmath69 , their redshift and their optical magnitude .
the bottom panel of figure [ fig : resolv_z ] also illustrates the _ differential _
24@xmath1 m background intensity produced as a function of redshift and computed for the same 24@xmath1 m fluxes as in the upper panel .
note that because of the magnitude limit imposed by our optical counterpart identifications @xmath70 was only computed up to @xmath6=26.5mag .
furthermore our determination of @xmath71 is necessarily underestimated at high redshift given the incompleteness of our identifications at faint fluxes .
however we believe that these two effects should affect our results by at most 10% given our estimate of @xmath72@xmath0@[email protected] for @xmath74 , @xmath75 and @xmath76 .
while mid - ir shallow surveys ( e.g. , @[email protected] ) only resolve up to @xmath030% of @xmath65 the analysis of figures [ fig : resolv_imag ] & [ fig : resolv_z ] reveals that the mips-24@xmath1 m sources brighter than @xmath69@xmath060@xmath1jy resolve up to @xmath070% of the total 24@xmath1 m background . this estimate is consistent with that of @xcite but it is slightly higher than the results obtained by @xcite in the elais - n1 field . since a non - negligible fraction of the background is produced by bright sources ( e.g. , @xmath310% at @xmath69@xmath31mjy ) this disagreement
could be simply explained by the small area of the sky covered in their analysis ( @xmath0185arcmin@xmath18 ) and the small number of bright 24@xmath1 m detections in their data .
furthermore up to @xmath010% of the 24@xmath1 m background originates from optical sources fainter than the limit of our photometric redshift catalog ( @xmath6=26.5mag ) , but more than 50% of @xmath65 is due to sources detected above 80@xmath1jy at @[email protected] and associated with optical counterparts brighter than @xmath6=24mag .
although their contribution to the integrated cib would require a careful analysis of their sed in the far - ir , our picture is therefore broadly consistent with previous claims supporting a large contribution of infrared - luminous galaxies at 0.5@xmath7@[email protected] to the ebl ( e.g. , * ? ?
* ; * ? ? ?
conversely , we also find that up to @xmath030% of the total 24@xmath1 m background must be produced by optically - faint ( @xmath6@xmath924mag ) galaxies at @xmath5@xmath31 .
the exact redshift distribution of these sources can not be constrained with our data given our 24@xmath1 m flux limit of 80@xmath1jy as well as the incompleteness of our redshift determinations at faint 24@xmath1 m fluxes .
however , the flattening of @xmath71 observed in figure [ fig : resolv_z ] suggests that they should be mostly located at 1@xmath7@xmath5@xmath73 .
finally , the bottom panel of figure [ fig : resolv_z ] reveals that the relative contribution of galaxies to the 24@xmath1 m background increases up to @xmath5@xmath01 and rapidly declines at higher redshift .
although this trend clearly depends on our extrapolations of the counts below 80@xmath1jy it is again consistent with the picture suggested by phenomenological models of ir galaxy evolution ( e.g. * ? ? ?
* ; * ? ? ?
* ) and by the analysis of the history of the cib recently performed with other methods ( e.g. , * ? ? ?
nonetheless , the contribution of galaxies with @xmath2@xmath980@xmath1jy seems to be roughly constant up to @xmath5@xmath01 .
hence , the peak " observed at @xmath5@xmath01 must be produced by a strong evolving ( albeit not dominant ) contribution of galaxies fainter than 80@xmath1jy and with rather low infrared luminosities .
constraining the physics of galaxy formation with semi - analytical models has become over the last decade one of the major challenges of modern cosmology . based on the now - popular cold dark matter ( cdm ) paradigm of structure formation @xcite these models compute the evolution of the stellar content of galaxies from the hierarchical assembly of dark matter haloes , using detailed simulations to account for a variety of physical processes such as the radiative cooling and the shock - heating of gas , the energy feedback from supernovae and active nuclei , the chemical enrichment of the interstellar medium and galaxy mergers ( e.g. , * ? ? ? * ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
although many quantitative aspects of galaxy evolution are still not correctly predicted with these scenarios they have recently been successful in reproducing several key properties of present - day galaxies such as their large - scale clustering or the bimodality of their optical color distribution @xcite .
because of the additional difficulty to properly account for dust extinction , the determination of accurate predictions of the deep ir universe with these models has been however barely explored so far .
simulating the evolution of galaxy ir properties requires not only a correct estimate of the history of the physical processes powering their energy output over relatively short time scales ( e.g. , star - forming activity , nuclear accretion ) , but an accurate treatment of their reprocessing of uv / optical radiation at longer wavelengths is also critical . since the launch of _ spitzer _ @xcite published the only single study ( to our knowledge ) describing some of the expected properties of 24@xmath1m - selected galaxies with the use of a semi - analytical model based on the hierarchical growth of dark matter haloes .
their simulations were performed with the galform galaxy formation model @xcite .
they used the grasil spectro - photometric code @xcite to calculate seds and to account for the reprocessing of radiation by dust , which allows for predictions of the multi - wavelength properties of all simulated galaxies as a function of cosmic time .
figure [ fig : compare_lacey ] shows their model predictions for the evolution of the galaxy median redshift as a function of 24@xmath1 m flux , along with their estimates of the 10@xmath77 and 90@xmath77 percentiles characterizing the associated redshift distributions .
they are compared with the same quantities estimated from our cosmos data , where the reported error bars account for the combination of flux and redshift uncertainties , the cosmic variance and the poisson noise as described in sect.[sec : z_distrib ] .
interestingly enough , we note that the model of @xcite shows a statistically - significant overestimate of the median redshift of galaxies at all fluxes probed by our survey .
this overestimate is particularly noticeable at fluxes below @xmath00.5mjy . given that their predictions agree reasonably well with the total source number counts observed at such flux levels ( see their figure 2 ) , it implies an excess of faint sources predicted above @xmath5@xmath01 as well as a lack of sources predicted at lower redshifts .
this disagreement was already noticed by @xcite from the comparison between their model and luminosity functions previously derived from _
spitzer _ data .
it is also clearly apparent in figure [ fig : compare_z_dist_lacey ] , where we compare the cosmos redshift distribution measured above @xmath69=83@xmath1jy with the result of the simulations that they obtained for the same 24@xmath1 m flux limit .
on the other hand we also note that the 90@xmath77 percentile of the distribution observed in cosmos at bright fluxes seems to extend up to higher redshifts compared to the model predictions
. this could be interpreted as an underestimate of the bright end of the simulated ir luminosity function at 1.5@xmath7@[email protected] .
it could also be due to an excess of sources at much lower redshifts since the model of @xcite overestimates the differential 24@xmath1 m source number counts above @xmath01mjy .
we will investigate these different possible interpretations in future papers where we will present the ir luminosity functions derived from our mips imaging of the cosmos field . in figures [ fig : compare_z_vs_k_lacey ] and
[ fig : compare_z_vs_r_lacey ] we now confront the observed and the predicted distributions of 24@xmath1 m sources as a function of their near - ir and optical photometry .
following the same selection criteria as those considered by @xcite we restrict our sample to sources brighter than @xmath2@xmath3183@xmath1jy .
we split these sources in different bins of @xmath32band and @xmath78band magnitude and we compare the redshift distributions observed in these bins with the corresponding predictions derived from the simulations=@xmath79 and @xmath80=@xmath81 .
the transformation between our set of filters ( @xmath78 & @xmath32 ) and that used by lacey et al .
( @xmath82 & @xmath11 ) was however neglected .
the range of distributions shown for the cosmos data ( grey shaded regions ) accounts for our best estimate of cosmic variance , flux and redshift uncertainties ( see sect.[sec : z_distrib ] ) .
the photometric errors affecting the @xmath78band and @xmath32band data were not taken into account but their effect should be limited given the adopted bins of magnitudes and the depth of the cosmos observations ( @[email protected] ab and @[email protected] ab at 5@xmath30 : @xcite , mccracken et al . submitted ) . as already noted by @xcite ,
our figures [ fig : compare_z_vs_k_lacey ] & [ fig : compare_z_vs_r_lacey ] show that the excess of simulated sources appearing in their model occurs mostly around @xmath5@xmath02 .
it originates from galaxies fainter than @[email protected] ( ab ) while most of the 24@xmath1 m sources in our selection ( @xmath2@xmath983@xmath1jy ) are typically brighter than this magnitude limit ( see also , e.g. , * ? ? ?
on the other hand our comparison in the @xmath78band reveals that these simulated sources are not particularly fainter than the typical 24@xmath1 m high redshift sources at optical wavelengths .
in fact the redshift distribution of the 24@xmath1 m sources with faint optical fluxes seems to be reasonably well predicted ( although the uncertainties due to the incompleteness of our redshift identification beyond @xmath83=26.5mag prevents reaching definite conclusions ) .
the excess of galaxies in the simulations originates from sources lying among the brightest optical counterparts associated with the 24@xmath1 m galaxy population at @xmath5@xmath02 ( i.e. , @xmath78@xmath725mag vega ) and it may thus be related to a too large number of blue galaxies predicted by the models . similarly we find that the lack of simulated galaxies noted by @xcite at @xmath5@xmath01 is related to an underestimate of sources with relatively bright near - ir counterparts , while the predicted number of galaxies with fainter @xmath32band luminosities ( @[email protected] vega ) is clearly too large at this redshift . given the slightly better agreement found at optical wavelengths , these results reveal again some discrepancies not only in the absolute optical / near - ir magnitudes of the predicted galaxies but also in their colors . we emphasize that this strong lack of 24@xmath1 m sources with bright near - ir luminosities ( i.e. , @[email protected] vega ) is found at all redshifts from @[email protected] to @[email protected] .
since the @xmath32band emission of these distant sources should probe the rest - frame properties of their evolved stellar populations this may reflect a systematic underestimate of the density of massive galaxies with on - going star formation in the distant universe .
unfortunately though , we do not have access to the mock catalogs where the simulated 24@xmath1 m sources were extracted from . more detailed comparisons between the physical properties of the observed and the modeled galaxies will have to be performed to better understand the origin of the discrepancies discussed above . as shown above the understanding of the deep ir universe with semi - analytical simulations still remains an important challenge for extragalactic astronomy .
in the past few years , simpler descriptions accounting for the strong evolution of galaxy ir properties and based on phenomenological " models have been proposed in the literature . assuming a backward evolution of the local luminosity function and a library of galaxy spectral energy distributions , these models are usually constrained to reproduce a number of quantities like the cosmic infrared background and/or galaxy ir properties such as the source number counts and redshift distributions .
although they can not be used to follow the history of the physical ingredients driving the evolution of galaxies and the change of their properties with time ( e.g. , gas accretion , stellar mass build - up , interactions , feedback , ... ) , they do not require a formalism as complex as the one needed in semi - analytical models .
they can also be useful in planing the overall characteristics ( e.g. , source counts , redshift distribution , ... ) of new ir or submillimeter surveys to be performed at yet - unexplored wavelengths .
all these models can vary from one another in the choice of their ir sed library , their decomposition of the overall galaxy population in different sub - groups ( e.g. , quiescent galaxies , starbursts , active nuclei , ... ) or the parameterization characterizing how the luminosity function is allowed to vary with lookback time . in figure
[ fig : compare_pheno ] we present the comparison between our cosmos 24@xmath1 m source redshift distributions and the predictions from the phenomenological models of @xcite , ( * ? ? ?
* , in prep . ) and @xcite .
assuming the luminosity - dependent sed library from @xcite , le borgne et al .
( 2009 ) have used a non - parametric `` count inversion '' technique with no specific assumption on the evolution of the luminosity function to fit the most - recently updated results from _ spitzer _
ir surveys . on the other hand the model presented by @xcite did not make use of any _ spitzer _ data but was constrained with 15@xmath1 m observations carried out with the _ infrared space observatory_. it assumes four distinct populations of sources ( quiescent galaxies , star - forming sources , type 1 and type 2 agns ) , each of them evolving independently from the others . initially these populations were characterized by a single template sed @xcite , but their scenario was recently updated to account for new procedures fitting individual galaxies with agn and starburst seds simultaneously from the optical up to the mid - ir wavelengths ( gruppioni et al .
, in prep . )
finally , @xcite assume a population of luminous starbursts characterized by luminosity - dependent seds as well as a population of quiescent cold galaxies evolving independently from the star - forming sources .
their model was updated from a previous version developed before the launch of _ spitzer _ to account for the new constraints imposed at 24@xmath1 m by the mips cosmological surveys .
the comparisons between these three models and our cosmos data are illustrated for 24@xmath1 m flux limits of 0.08 , 0.15 and 0.3mjy .
the possible range of observed redshift distributions shown in the figure was calculated as described in sect.[sec : z_distrib ] by taking into account the combined effect of flux and redshift uncertainties as well as the poisson noise and the cosmic variance .
we did not include any comparison below @[email protected] as the small comoving volume sampled by our survey at these low redshifts would imply large statistical uncertainties . at @xmath5@xmath31 and faint fluxes ( e.g. , @[email protected] )
we note that the model of le borgne et al .
( 2009 ) shows a rather good agreement with our observations , while the models from @xcite and @xcite show a significant overestimate of the number of 24@xmath1 m sources around @xmath5@xmath02 . as we discussed in sect.[sec : z_distrib ] the bimodal distribution predicted by these two models is mainly produced by @xmath8-correction effects boosting the detectability of starburst sources when their 7.7@xmath1 m aromatic feature is redshifted into the mips 24@xmath1 m bandpass at 1.5@xmath7@[email protected] .
the predicted bimodality is clearly not as prominent in the data .
this disagreement could reflect either the presence of too strong features in the seds assumed by the models or a too rapid evolution of the source space density at the faintest fluxes typically probed by our survey . on the other hand ,
the break observed at @[email protected] in the redshift distribution of the cosmos galaxy population at bright fluxes seems to be better reproduced by the model of gruppioni et al .
than in the predictions from le borgne et al .
( where the break is barely apparent ) or lagache et al .
( where the break is too prominent due to the strong excess of sources expected at @xmath5@xmath02 ) . at @xmath5@xmath01 @xcite predict however too many sources at @[email protected] while the model of le borgne et al . may also overestimate the luminosity function at 0.4@xmath7@[email protected] .
because of ( i ) the unprecedented accuracy of the photometric redshifts used in this work , ( ii ) our very high rate of redshift identifications obtained down to @xmath2=80@xmath1jy , and ( iii ) the large area covered by our survey , our analysis represents the most statistically significant characterization of the redshift distributions of the 24@xmath1 m source number counts achieved so far .
we found that the distribution of the 24@xmath1 m selected galaxies is characterized by a prominent peak at @xmath5@xmath01 and a relatively large fraction of sources at 1@xmath7@xmath5@xmath73 , which is globally consistent with the distributions of mips-24@xmath1 m sources previously determined with other surveys @xcite .
however , specific differences between our results and these other studies can be noticed .
for instance , the redshift distribution obtained by @xcite shows a gradual and rapid decline from @xmath5@xmath01 up to @xmath5@xmath03 , while we observe a break at @[email protected] as well as a larger fraction of sources at @xmath5@xmath31 .
similarly , our distributions exhibit smoother variations than the redshift distribution determined by @xcite in the field of the _ chandra deep field south _ ( cdfs ) .
all these deviations from our results could originate either from less accurate photometric redshifts that were used in some of these other studies , or from the effect of cosmic variance over small fields of view like the cdfs .
the large number of 24@xmath1 m sources at 0.5@xmath7@[email protected] and our analysis of the history of the 24@xmath1 m background light also shows that the cib must originate predominantly from moderately - luminous galaxies at intermediate redshifts rather than highly obscured and luminous dusty sources at @xmath5@xmath32 .
this result has already been infered from previously - published studies ( e.g. , * ? ? ?
* ; * ? ? ?
* ) but our data provide for the first time a large enough statistics to explore with good accuracy the build - up of the mid - ir background as a function of cosmic time .
the characterization of the far - ir properties of the 24@xmath1 m sources with the already - existing mips-70/160@xmath1 m and mambo-1.2 mm observations of cosmos ( frayer et al . , submitted ; * ? ? ? * ) or with the future coverage of this field with forthcoming facilities like _ herschel _ will provide an even more detailed picture of the history of the ebl over the whole range of ir wavelengths . yet , in spite of this globally coherent picture that has emerged from mid - ir surveys our comparison between the cosmos 24@xmath1 m data and the model predictions of @xcite reveal that the evolution and the properties of luminous galaxies in the distant universe are still far from being properly reproduced by the semi - analytical simulations of cosmic history . although the rapid evolution of the ir luminosity function up to @xmath5@xmath02 is globally well predicted by the model of @xcite , discrepancies in the expected source density per bin of ir luminosity and in the optical / near - ir properties of 24@xmath1 m sources are clearly apparent .
interestingly we found that the number of 24@xmath1 m sources with bright near - ir counterparts ( i.e. , @[email protected] vega ) in the simulations is clearly underestimated at all redshifts beyond @[email protected] .
it implies that the predicted source density of massive galaxies with strong on - going activity of star formation such as those recently discovered at high redshift with _
( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) is too low .
besides we note an overestimate of simulated sources with 20.5mag@xmath35@[email protected] ( vega ) at @xmath5@xmath01 , while their counterparts in the @xmath78band seem to be more representative of the optical properties of 24@xmath1 m sources at this redshift .
in addition to the overall underestimate of the predicted source density at this redshift , we thus conclude that a large fraction of simulated galaxies at @xmath5@xmath01 are too blue with respect to the typical 24@xmath1 m detected sources .
this trend actually recalls the discrepancy that we also observed at @xmath5@xmath02 where the overestimate of predicted sources originates from blue galaxies with @xmath32band luminosities much fainter than the bulk of 24@xmath1 m sources found at this redshift .
the origin of this excess of blue galaxies with a 24@xmath1 m detection above our sensitivity limit is not straight forward to determine .
assuming that the star formation rate ( sfr ) of individual sources is correctly predicted by the simulations , it may be related to an incorrect description of their spectral energy distributions . as we mentioned in sect.[sec : z_distrib ] the broad aromatic features commonly observed at 7.7@xmath1 m , 11.3@xmath1 m and 12.7@xmath1 m in the spectra of starburst galaxies can enhance their detectability when these lines are redshifted into the mips-24@xmath1 m bandpass .
an overestimate of the strength of the 7.7@xmath1 m feature could thus result in too many 24@xmath1 m sources predicted at @xmath5@xmath02 .
furthermore the @xmath78band emission of galaxies at such high redshifts relates to their rest - frame uv properties , which are obviously very sensitive to dust extinction given our selection performed at mid - ir wavelengths
. a wrong treatment of the obscuration affecting these 24@xmath1 m sources would thus bias their predicted characteristics at optical wavelengths as well as the estimate of their reprocessed dust emission in the mid - infrared . on the other hand , this excess could also originate from an overestimate of the galaxy star formation rate itself . in the simulations of @xcite galaxies with on - going bursts of star formation account for @xmath050% and @xmath390% the 24@xmath1 m source population at @xmath5@xmath01 and @xmath5@xmath02 respectively ( the other sources being classified as quiescent galaxies , see their figure a4 ) .
their mid - ir emission directly scales with their star - forming activity , which could lead to a too large number of predicted sources if their sfr is statistically too high .
indeed , star formation processes critically depend on fundamental mechanisms and physical parameters that may still not be fully constrained in the simulations , like for instance the critical density above which stars can form depending on the galaxy internal dynamics and the various energy feedback ( agns , supernovae , ... ) .
finally , it is worth noting that the observed optical and near - ir colors may also depend on the initial mass function ( imf ) assumed in the simulations . in the model of @xcite a normal neighborhood
imf leads to a too little evolution of the mid - ir luminosity function compared to the data , and it also underestimates the strong evolution observed in the faint submillimeter number counts @xcite .
a key ingredient characterizing their scenario is the use of a top - heavy imf for the bursts of star formation triggered by galaxy mergers , which seems to better reproduce the evolution of the galaxy luminosity function observed in the mid - ir . under this assumption
though , the larger fraction of blue and massive stars produced in bursts should lead to bluer stellar populations and this could be one of the reasons explaining the excess of blue galaxies in their predictions . a much more detailed comparison between the intrinsic properties ( mass , sfr , seds ) of _ individual _ simulated sources and those characterizing the observed 24@xmath1 m galaxy population will be necessary to identify the critical parameters that need to be better constrained for improving model predictions in the infrared .
given the large number of ingredients to control , the determination of proper predictions of galaxy evolution at ir wavelengths with current semi - analytical models remains obviously a challenging task .
reproducing the evolution of the ir luminosity function requires not only an accurate description of the evolution of the bolometric luminosity function for star - forming galaxies but also an accurate treatment of dust extinction as a function of galaxy properties and a reliable representation of galaxy seds in the infrared . furthermore ,
dust - obscured star - forming galaxies are not the only contributers to the number counts at ir wavelengths .
quasars and dusty agns should also be taken into account , especially at the bright end of the luminosity function where their contribution is far from being negligible ( e.g. , * ? ? ?
given the coeval growth of bulges and super - massive black holes ( e.g. , * ? ? ?
* ) composite " sources experiencing both nuclear and star - forming activity may actually represent a large fraction of the mid - ir selected galaxy population at high redshift .
this large variety of physical properties characterizing ir - luminous galaxies added to the complexity of their spectral signatures in the mid - ir can thus make the predictions from semi - analytical simulations particularly difficult at ir wavelengths .
this contribution of nuclear accretion to cosmic history has not been implemented in the model of @xcite .
therefore the feedback action of agns , which has been recently discussed in the framework of other semi - analytical models as an efficient process to regulate the star - forming activity in galaxies , is not taken into account in their simulations . instead , the decrease of gas cooling for the formation of massive spheroids is controlled in their model by the action of superwinds driven by supernova explosions .
these winds produce a feedback effect qualitatively similar to that of agns , but the underlying physical mechanisms are somehow different and they result in different predictions for the evolution of the galaxy luminosity function with redshift @xcite .
infrared simulations of distant sources accounting for the role of agn feedback were actually presented by @xcite using the model of @xcite .
however their scenario did not directly include the merging of galaxies and dark matter haloes , and contrary to our results their 24@xmath1 m source redshift distribution at @[email protected] was predicted to peak at @xmath5@xmath02 .
given the current @xmath19cdm paradigm as well as the role that nuclear accretion played in the evolution of dusty high redshift galaxies , one possible approach to progress in our understanding of the deep ir universe with semi - analytical models would be therefore to explore the impact of scenarios including both the hierarchical assembly of dark matter haloes and the contribution of agns .
we performed our very first analysis of the deep 24@xmath1 m observations carried out with the _
spitzer space telescope _ over the 2deg@xmath18 of the cosmos field .
we detected almost @xmath030000 sources down to a 24@xmath1 m flux of 80@xmath1jy . using the multi - wavelength ancillary data available in cosmos we identified the optical / near - ir counterparts and the photometric redshifts for @xmath095% of these 24@xmath1 m detections .
the main results that we derived can be summarized as follows : \1 . the redshift distribution of the 24@xmath1m selected galaxies shows a prominent peak at @xmath5@xmath01 but more than @xmath050% of the sample
is located at 1@xmath7@xmath5@xmath73 .
below @xmath5@xmath01 almost all 24@xmath1 m sources are brighter than 24mag ( ab ) in the @xmath6band , while a non - negligible of the mips sources at higher redshifts are associated with very faint optical counterparts ( i.e. , 24mag@xmath7@[email protected] ) .
\2 . below a fixed optical magnitude limit
the fraction of @xmath6band sources with 24@xmath1 m detection strongly increases up to @xmath5@xmath02 .
this reflects the strong evolution that star - forming galaxies have undergone with lookback time .
however this rising trend shows a clear break at @[email protected] , which seems to be more pronounced when higher 24@xmath1 m flux selections are considered ( e.g. , @[email protected] ) .
this feature is also visible in the redshift distributions of 24@xmath1 m sources .
it probably originates from @xmath8-correction effects implied by the presence of the strong aromatic features and/or the silicate absorption commonly observed in the spectra of luminous dusty galaxies .
\3 . using extrapolations of the number counts at faint fluxes we resolved the build - up of the mid - infrared background across cosmic ages .
we find that @xmath050% and @xmath080% of the 24@xmath1 m background intensity originate from galaxies at @xmath5@xmath71 and @xmath5@xmath72 respectively , with a contribution of faint sources reaching @xmath020@xmath5310% below our detection limit .
as already suggested by other previously - published studies we thus infer that the cosmic infrared background is mostly produced by moderately - luminous dusty galaxies at 0.5@xmath7@[email protected] rather than very luminous and obscured sources at @xmath5@xmath32 .
the comparison between our results and semi - analytical predictions reveal substantial discrepancies in the simulated redshift distributions as well as in the @xmath10 colors of the 24@xmath1 m source counterparts .
in particular we found a strong excess of blue sources predicted at @xmath5@xmath02 and the @xmath11band fluxes of simulated galaxies appear to be systematically too low at @[email protected] .
this could reflect an underestimate of the density of high redshift massive sources with strong on - going star formation possibly related to fundamental physical parameters and/or processes that are still not fully constrained in the simulations of galaxy formation .
it could also be due to an inaccurate treatement of dust extinction in the models , which would imply discrepancies in the predictions of the galaxy seds at rest - frame uv , optical and mid - ir wavelengths .
comparisons with some of the most recent backward evolution scenarios reproduce reasonably well the flux / redshift distribution of 24@xmath1 m sources up to @xmath5@xmath03 .
nonetheless none of them is able to accurately match our results at all redshifts over the full range of 24@xmath1 m fluxes probed by our survey .
it is a pleasure to acknowledge the contribution from all our colleagues of the cosmos collaboration .
more information on the cosmos survey is available at http://www.astro.caltech.edu/cosmos .
this work is based on observations made with the _
spitzer space telescope _ , a facility operated by nasa / jpl .
financial supports were provided by nasa through contracts # 1289085 , # 1310136 , # 1282612 and # 1298231 issued by the jet propulsion laboratory .
our research project was also supplemented with phenomenological model predictions from published papers or future publications and that damien le borgne , carlotta gruppioni , francesca pozzi , guilaine lagache and herv dole have been willing to share with us ; they are greatly acknowledged for their contribution .
we are grateful to david elbaz and casey papovich for insightful comments on our results , and we also appreciated the hospitality of the aspen center for physics where part of our work was prepared .
we finally acknowledge the support and the contribution of the _ spitzer science center _ to our research program and we warmly thank the referee for a careful reading of the manuscript .
support for elf s work was provided by nasa / caltech through the _ spitzer space telescope _
fellowship program ( # 1080367 ) .
part of this work was also supported by the grant asi / cofis / wp3110 i/026/07/0 . | we present the first results obtained from the identification of @xmath030000 sources in the _ spitzer_/24@xmath1 m observations of the cosmos field at @xmath2@xmath380@xmath1jy . using accurate photometric redshifts
( @[email protected] at @xmath5@xmath02 for 24@xmath1 m sources with @xmath6@xmath725mag ab ) and simple extrapolations of the number counts at faint fluxes we resolve with unprecedented detail the build - up of the mid - infrared background across cosmic ages .
we find that @xmath050% and @xmath080% of the 24@xmath1 m background intensity originate from galaxies at @xmath5@xmath71 and @xmath5@xmath72 respectively , supporting the scenario where highly obscured sources at very high redshifts ( @xmath5@xmath32 ) contribute only marginally to the cosmic infrared background .
assuming flux - limited selections at optical wavelengths , we also find that the fraction of @xmath6band sources with 24@xmath1 m detection strongly increases up to @xmath5@xmath02 as a consequence of the rapid evolution that star - forming galaxies have undergone with lookback time .
nonetheless this rising trend shows a clear break at @[email protected] , probably due to @xmath8-correction effects implied by the complexity of spectral energy distributions in the mid - infrared .
finally , we compare our results with the predictions from different models of galaxy formation .
we note that semi - analytical formalisms currently fail to reproduce the redshift distributions observed at 24@xmath1 m .
furthermore the simulated galaxies at @xmath2@xmath980@xmath1jy exhibit @xmath10 colors much bluer than observed and the predicted @xmath11band fluxes are systematically underestimated at @[email protected] .
unless these discrepancies mainly result from an incorrect treatment of extinction in the models they may reflect an underestimate of the predicted density of high redshift massive sources with strong on - going star formation , which would point to more fundamental processes and/or parameters ( e.g. , initial mass function , critical density to form stars , feedback , ... ) that are still not fully controlled in the simulations .
the most recent backward evolution scenarios reproduce reasonably well the flux / redshift distribution of 24@xmath1 m sources up to @xmath5@xmath03 , although none of them is able to exactly match our results at all redshifts . |
observations of ca ii h and k line intensity in quiet internetwork areas show a pattern that is roughly the inverse of the underlying granulation called `` reversed granulation '' .
the anticorrelation is excellent , especially if a 3-minute delay for propagation is taken into account .
reversed granulation has long been known to be a result of overturning convection (; ) .
reversed granulation is a hydrodynamic phenomenon that does not require the presence of magnetic field .
purely hydrodynamic simulations that reproduce reversed granulation well also serve to illustrate this point .
reversed granulation is perhaps the most conspicuous feature observed in high - resolution ca ii h and k filtergrams of quiet - sun internetwork .
its ubiquitous nature makes detection of other features , such as internetwork bright points , difficult . here
, we report on small - scale structures that at first glance appear similar to reversed granulation in hinode observations of internetwork quiet sun , but with sharper edges and faster brightness changes .
we investigate image sequences taken with the solar optical telescope on board the hinode spacecraft @xcite in its blue continuum and ca ii h passbands on march 25 , 2007 .
the observations studied here have a high cadence of 6.4 s necessary to resolve the fast apparent motions of the reversed granulation . we apply a high - pass filter separately on each pixel in the ca ii h sequence , keeping only frequencies above 14 mhz .
we also filter the ca ii h cube separately using a `` cone filter '' in @xmath1-@xmath2-space to isolate apparent supersonic motions .
we then investigate the blue continuum , the ca ii h , and the two filtered data cubes interactively using the `` cube slicer '' xslice ( available in solarsoft ) that dissects the data in @xmath3@xmath4 , @xmath3@xmath5 , and @xmath4@xmath5 slices .
the filtered data show conspicuous features above intergranular lanes that are typically arc - like in shape with a horizontal extent of several arcsec .
they are co - spatial with reversed granulation , and display similar general properties , but appear to have sharper edges and show fast brightness change .
their horizontal motions track the underlying intergranular lane closely , with no evident time delay .
we find no obvious recurrent behavior .
an example is shown in fig .
[ fig : fig1 ]
. we will refer to these features as `` enhanced reversed granulation '' ( erg ) .
investigation of the ca ii h sequence suggests that erg is associated with @xmath0-mode oscillations .
the emission from the feature is suddenly increased as the @xmath0-mode maximum moves through it , then returns back to levels similar to the reversed granulation when the @xmath0-mode moves off .
the leftmost column in fig .
[ fig : fig1 ] shows the @xmath0-mode pattern derived by filtering the ca ii h cube in @xmath1-@xmath2-space for apparent supersonic motions .
it shows a maximum at the time that the erg shows its maximum brightness .
erg seems to happen preferentially over wide intergranular lanes .
the lanes also frequently show broken - up granulation in the vicinity .
both are typically considered indications of the presence of magnetic field . while the example in fig .
[ fig : fig1 ] does not show pronounced broadening of the intergranular lane or broken - up granulation , many others identified in the sequence do show these effects .
the response erg in the filtered ca ii h data is caused by the relatively sharp edge , horizontal motion , and rapid change in intensity .
the combination of those properties results in step - like change in intensity for each pixel that ensures an excess of power at high temporal frequencies .
we have investigated several similar sequences , and find that they appear ubiquitously in internetwork areas of quiet sun .
none are detected in the network , possibly due to confusion caused by the presence of many fibrils , or possibly because these features do not exist there .
these features may be the result of interactions of @xmath0-mode oscillations with internetwork magnetic field .
if so , they may form a conduit to channel energy of @xmath0-mode oscillations into the quiet chromosphere .
it is then tempting to assume that these events correspond to low - lying loops , due to the horizontal extent of the feature .
however , the lack of recurrent behavior suggests that they are not associated with more stable magnetic structures . also , an initial investigation of hinode sequences that include both ca ii h imaging as well as magnetograms did not reveal conspicuous signatures of co - spatial magnetic field .
photospheric foot points manifested as strong vertical field would be difficult to miss even in magnetograms with low sensitivity .
it is more difficult to rule out a connection with the recently - discovered `` transient horizontal magnetic fields '' .
photospheric thmfs are located inside granules , whereas the features discussed here are always located above intergranular lanes .
however , no thmfs have been detected in the chromosphere , perhaps because of a lack of sensitivity of the measurements , or because they do not exist there . in any case
, there is currently insufficient evidence to make a statement on this point .
we have extracted the @xmath0-mode pattern from the ca ii h sequence itself
. it would be better to investigate the relation of these features to @xmath0-mode oscillations using sequences that include dopplergrams .
such sequences exist , but suffer from slower cadence that limit their usefulness .
alternatively , one can attempt to extract the @xmath0-mode pattern from the photospheric blue - continuum image sequence , in hopes of at least getting an independent measurement , and compare the positions of maximum brightness ( i.e. , maximum compression ) to the locations of erg .
we will pursue both approaches in the near future .
we thank mats carlsson for discussions and co - aligned data cubes .
_ hinode _ is a japanese mission developed and launched by isas / jaxa , with naoj as domestic partner and nasa and stfc ( uk ) as international partners .
it is operated by these agencies in co - operation with esa and nsc ( norway ) . | we investigate features that are observed in ca ii h sequences from hinode in places where reversed granulation seems to interact with @xmath0-modes .
these features appear ubiquitously in the quiet sun .
they are co - spatial with reversed granulation , and display similar general properties , but have sharper edges and show fast brightness changes .
they also appear predominantly above wide intergranular lanes , indicating a potential connection with magnetism .
we report on the appearance and dynamics of these features using high - resolution , high - cadence observations from hinode , and we discuss their possible origin . |
raman spectroscopy , since its discovery in 1928 , is a widely used technique to investigate the vibrational and electronic properties of materials @xcite .
the crystalline , amorphous or nano - crystalline nature of any material can be ascertained by studying the first order raman spectrum .
the physics behind raman scattering in semiconductors or crystals is based on the inelastic interaction of incidence photons with lattice vibrations i.e. phonons , that are sensitive to internal or external perturbations
. in crystalline materials , the raman scattering is limited to near zone centered phonons ( zcp ) .
but in the case of nano - structures the phonons other than the zone centered ones also contribute due to confinement or localization of phonons in a finite dimension , resulted in a change in the line - shape of the first order raman spectrum .
the raman spectrum provides a fast and convenient method to analyze the vibrational properties of crystalline and amorphous semiconductors .
since nano - structures are quasi - crystalline , their raman spectra are expected to be intermediate between the spectra of the corresponding crystalline and amorphous materials .
many research investigations have been carried out in order to understand the vibrational properties in semiconductor nano - structures @xcite , confined in one or more dimensions .
all the results , show a shift in the peak position of the first - order raman line of crystalline semiconductors towards lower energy accompanied by a spectral broadening and phonon softening .
although it is applicable to all the materials , in the present study , we have done a detail analysis of first - order raman line - shape of semiconductor material especially silicon .
raman scattering study on si nano - structures has been studied extensively @xcite .
raman spectrum from the c - si is generally symmetric having a lorentzian line - shape appeared at 521 @xmath0 with a natural line - width of @xmath1 4 @xmath0 .
any deviation from symmetry or change in its full width at half maximum ( fwhm ) can be explained in terms of quantum confinement , heating related effects and electron - phonon interactions etc .
more than one effect can contribute to the raman line - shape asymmetry or broadening in the si nano - structures .
it is important to explain its behavior as a response to different perturbations present in the system . by providing an appropriate theoretical explanation , it is possible to investigate many processes taking place at the nano - scale level in the si nano - structures . in the si nano - structures ,
frequency of the optical phonons is expected to change due to finite size of the material .
thus , the raman spectra from the si nano - structures are also modified .
this is mainly because , the translational symmetry of the crystalline materials is broken at grain boundaries , which results in the appearance of specific surface and interface vibrational contributions @xcite .
this is more pronounced in reduced dimensional systems .
these properties change the frequency of the phonons in the nanostructures , which is reflected in raman scattering .
different models are presented to depict the raman line - shape resulting from the nano - structures .
one of the well accepted model is phonon confinement model ( pcm ) as proposed by richter et al .
@xcite , which is further modified by campbell et al.@xcite .
intensity of the first - order raman scattering , i ( @xmath2 ) , can be written as @xcite @xmath3 ^ 2~+~(\gamma/2)^2 } d^nq,\ ] ] where @xmath4 is expressed in units of @xmath5 , @xmath6 being the lattice constant , 0.543 nm .
the parameter @xmath7 stands for the average size ( diameter ) of the nano - crystals . @xmath8
being the linewidth of the si optical phonon in bulk c - si ( @xmath1 4 @xmath0 ) .
following tubino _
@xcite , the dispersion @xmath2(q ) of the optical phonon in a spherical si nanocrystal can be taken as @xmath9 where @xmath10 @xmath11 and @xmath12 @xmath11 . in the present study
, we have discussed in depth , the evolution of asymmetric raman line - shape in nano - materials starting from a symmetric raman line - shape from its bulk counterpart .
understanding the each term in the asymmetric raman line - shape ( due to phonon confinement effect ) will enable the researchers to explore the possibility of pcm beyond elemental semiconductor nano - materials like si , ge and apply the same to other compound semiconductors like @xmath13 , @xmath14 , etc .
as discussed earlier , raman line - shape is a manifestation of raman active phonon mode of a semiconductor obtained after scattering of light @xcite .
it is an established fact that a lorentzian raman line - shape is observed from a bulk crystalline semiconductor , which is symmetric @xcite in nature whereas that from its nano counterpart is asymmetric @xcite .
peak of the raman line - shape from a bulk semiconductor corresponds to the zone - center optic phonon .
figure 1(a ) shows the longitudinal optic ( lo ) branch of phonon dispersion relation for silicon ( si ) .
the discrete points in the figure shows the data reported in literature for the same @xcite .
these data points are represented theoretically by eq . 2 @xcite .
figure 1(b ) shows a typical raman line - shape from bulk si represented by eq .
@xmath15 ^ 2+(\gamma/2)^2 } , \ ] ] where , @xmath16 is the zone center optical phonon frequency corresponding to q = 0 in figure 1a and ` @xmath8 ' is the fwhm of the raman peak . for bulk
si , the values of @xmath16 = 521 @xmath0 and @xmath8 = 4 @xmath0 , are used to calculate the raman line - shape as shown in figure 1b .
raman line - shape from nano - structured si as given in eq .
1 was derived by richter et al @xcite and later modified by campbell et al @xcite .
all the terms in eq.1 represents a concept and is present with an underlying physics which will be discussed now . at nano - scale ,
phonons that participate in raman scattering are confined within the boundary of nano - structure .
due to this quantum confinement effect breakdown of raman selection rule takes place and phonons , away from the zone center ( @xmath17 ) , also take part in raman scattering @xcite . as a result , non zero raman intensities , @xmath18 ,
are also observed at lower frequencies ( in wave - number units ) away from @xmath16 .
the final raman line - shape will be summation of all the intensities corresponding to wave - numbers at non - zero wave - vectors .
after inclusion of this fact in eq .
3 , raman line - shape function can be written as eq . 4 as follows : @xmath19 ^ 2+(\gamma/2)^2 } , \ ] ] where summation over different values of ` @xmath4 ' ( varying from 0 to 1 ) represents the contributions from all phonons . since eq .
4 involves a summation , variation in ` @xmath4 ' ( 0 to 1 ) can be done in different steps .
raman line - shape in eq .
4 is plotted in figure 2 for three different step sizes @xmath20 0.1 , 0.05 and 0.02 .
it is important to mention here that the plots in figure 2 are not actual raman line - shapes .
spikes in figure 2 correspond to the contribution in line - shape from individual phonon corresponding to a particular value of ` @xmath4 ' .
for example , the peak marked with asterisk ( * ) , centered at @xmath1 414 @xmath0 , represents the contribution of phonon represented by @xmath21 . for smaller step sizes , the line - shape is continuous instead of spiky as can be seen in figure 2(c ) .
most visible feature of line - shape in figure 2(c ) is , its asymmetry as compared to the line - shape in figure 1(b ) .
a careful observation of evolution from figure 2(a ) - 2(c ) reveals that the asymmetric line - shape arises if all phonons of a branch in phonon dispersion relation take part in the scattering .
in addition to this it is also clear from figure 2 that spikes are prominent in the lower wave - number region ( 400 - 475 @xmath0 ) as compared to the higher wave - number region ( 475 - 521 @xmath0 ) .
this observation is directly related to the slope of the phonon dispersion relation in figure 1(a ) .
the phonon dispersion relation is nearly flat near the zone center ( @xmath22 ) whereas it is monotonically decreasing near the zone edge ( @xmath21 ) making the dispersion relation nonlinear .
due to this nonlinearity , contributions from the phonons in the vicinity of zone center add together to result in higher intensity near 521 @xmath0 ( corresponding to @xmath23 phonon frequency ) .
line - shapes in figure 2 have been plotted by assuming that all the phonons from q = 0 to 1 contribute equally to the raman intensity and the phonons are discrete ( as represented by steps in the value of q ) .
the actual raman line - shape from nano - structures can now be derived by taking into account the actual contribution from all the phonons and by considering that the phonon spectrum is continuous rather than discrete @xcite . as discussed above , the actual asymmetric raman line - shape ( as observed from nano - structures )
will be derived here for the case of si nano - structures .
the contribution of the whole phonon spectrum in raman line - shape equation can be incorporated in eq . 4 by integrating over the full dispersion relation rather than doing summation . as a result
, the resultant line - shape equation can be written as eq .
5 as follows .
@xmath24 ^ 2+(\gamma/2)^2 } dq , \ ] ] equation 5 is plotted in figure 3(a ) no spikes are seen in the asymmetric plot due to integration over all the possible values of ` @xmath4 ' . almost flat intensity in the 420 - 480 @xmath0 region is due to the hypothesis that all phonons contribute equally or having equal probability to participate in the scattering .
the actual contribution of a phonon depends on the extent of confinement inside a particular nano - crystal .
this is expresses quantitatively by introducing a `` weighting function '' @xcite to care of the actual contribution of a particular phonon .
the nature of this weighting function depends on the dimensionality of confinement . for simplicity ,
an exponential weighting function has been suggested by campbell et al @xcite for all practical purposes .
after introduction of the size dependence , the raman line - shape is represented by eq .
6 as follows : @xmath25 ^ 2+(\gamma/2)^2 } dq\ ] ] where ` @xmath7 ' is the size of nano - structure and ` @xmath6 ' is the lattice parameter of semiconductor crystal .
6 is plotted by taking the value of @xmath7 = 2.5 nm and @xmath6 = 0.543 nm ( lattice constant of si ) in figure 3(b ) alongside 3(a ) for comparison . by incorporating the size factor ,
the line - shape reflects the true asymmetric behavior as observed in the raman spectra from nano - structures .
incorporation of the exponential term in eq . 6 also takes care of the size dependence of the raman line - shape , which is in consonance with the observed size dependent raman scattering @xcite .
equation 6 can be understood as the basic raman line - shape for nano - structured semiconductors in particular and nano - structured materials in general .
modifications can be done eq . 6 to obtain various special raman line - shapes reported to investigate the effects of stress , temperature , electron - phonon ( fano ) effect , quantum confinement effect etc .
details about these functions are available in the literature .
temperature dependence of phonon dispersion relation can directly be used if one wants to investigates the temperature dependent quantum confinement effects .
in addition , degree of confinement can also be considered by using the appropriate integral ( surface or volume integral ) and will result in eq . 1 from where the discussion of this paper started .
if in a sample all the nano - structures are not of same dimension , a distribution function can successfully be introduced not only to take care of the distribution effect but also to know the exact size distribution available in the sample by theoretical fitting of the experimental data with the modified raman line - shape .
an asymmetric raman line - shape is observed from semiconductor nano - structures in contrast with a symmetric one from its bulk counterpart .
this asymmetric line - shape is used to get information about the shape , size and distribution of nano - structures by theoretically fitting with a theoretical raman line - shape function .
physical significance of different terms in the theoretical raman line - shape can be explained for better consonance between theoretical raman line - shape and experimental raman scattering data .
better understanding of theoretical background makes the use of raman line - shape function more versatile and can be used for other semiconductors and materials .
if information about the phonon dispersion , size dependence etc is known a raman line - shape can be derived theoretically from the basic lorentzian line - shape .
effects of other perturbations ( like stress , temperature etc ) can also be easily incorporated to get a new raman line - shape .
help from mr ashish kumar yadav is highly acknowledged .
one of the author ( gs ) would also like to acknowledge dst , govt . of india , for the funding under dst fast track scheme for young scientists , project no
. sr / ftp / ps-007/2012 .
raman c ( 1928 ) a new radiation .
indian j phys 02:387 .
raman c v , krishnan k s ( 1928 ) a new type of secondary radiation .
nature 121:501502 .
doi : 10.1038/121501c0 richter h , wang zp , ley l ( 1981 ) the one phonon raman spectrum in microcrystalline silicon
. solid state commun 39:625629 .
doi : 10.1016/0038 - 1098(81)90337 - 9 .
campbell ih , fauchet pm ( 1986 ) the effects of microcrystal size and shape on the one phonon raman spectra of crystalline semiconductors . solid state commun 58:739741 .
doi : 10.1016/0038 - 1098(86)90513 - 2 .
sahu g , kumar r and mahapatra d p , 2013 _ silicon _ * 5 * , .
kumar r , mavi hs , shukla ak ( 2010 ) spectroscopic investigation of quantum confinement effects in ion implanted silicon - on - sapphire films .
silicon 2:2531 .
doi : 10.1007/s12633 - 009 - 9033-z kumar r , shukla ak , mavi hs , vankar vd ( 2008 ) size - dependent fano interaction in the laser - etched silicon nanostructures .
nanoscale res lett 3:105108 .
doi : 10.1007/s11671 - 008 - 9120-x sahu g , joseph b , lenka h p , kuiri p k , pradhan a and mahapatra d p , 2007 _ nanotechnology _ * 18 * , 495702 .
sahu g and mahapatra d p , 2011 _ mrs proceedings spring meeting _ , * 1354 * ( 2011 ) : doi : 10.1557/opl.2011.1212 .
gouadec g , colomban p ( 2007 ) raman spectroscopy of nanomaterials : how spectra relate to disorder , particle size and mechanical properties .
prog cryst growth ch 53:156 .
doi : 10.1016/j.pcrysgrow.2007.01.001 tubino r , piseri l , zerbi g ( 1972 ) lattice dynamics and spectroscopic properties by a valence force potential of diamondlike crystals : c , si , ge , and sn .
j chem phys 56:10221039 .
doi : doi:10.1063/1.1677264 .
bergman l , nemanich rj ( 1996 ) raman spectroscopy for characterization of hard , wide - bandgap semiconductors : diamond , gan , gaaln , aln , bn .
annu rev mater sci 26:551579 .
doi : 10.1146/annurev.ms.26.080196.003003 mavi hs , islam ss , kumar r , shukla ak ( 2006 ) spectroscopic investigation of porous gaas prepared by laser - induced etching .
j non - cryst solids 352:22362242 .
doi : 10.1016/j.jnoncrysol.2006.02.046 mavi hs , prusty s , kumar m , et al .
( 2006 ) formation of si and ge quantum structures by laser - induced etching .
phys status solidi a - appl mat 203:24442450 .
doi : 10.1002/pssa.200521027 pivac b , furi k , desnica d , et al .
( 1999 ) raman line profile in polycrystalline silicon .
j appl phys 86:4383 .
doi : doi:10.1063/1.371374 li b , yu d , zhang s - l ( 1999 ) raman spectral study of silicon nanowires .
phys rev b 59:16451648 .
doi : 10.1103/physrevb.59.1645 wang r , zhou g , liu y , et al .
( 2000 ) raman spectral study of silicon nanowires : high - order scattering and phonon confinement effects .
phys rev b 61:1682716832 .
doi : 10.1103/physrevb.61.16827 prusty s , mavi hs , shukla ak ( 2005 ) optical nonlinearity in silicon nanoparticles : effect of size and probing intensity .
phys rev b 71:113313 .
doi : 10.1103/physrevb.71.113313 konstantinovi mj , bersier s , wang x , et al .
( 2002 ) raman scattering in cluster - deposited nanogranular silicon films .
phys rev b 66:161311 .
doi : 10.1103/physrevb.66.161311 piscanec s , ferrari ac , cantoro m , et al . ( 2003 ) raman spectrum of silicon nanowires .
mater sci eng c 23:931934 .
doi : 10.1016/j.msec.2003.09.084 piscanec s , cantoro m , ferrari ac , et al .
( 2003 ) raman spectroscopy of silicon nanowires .
phys rev b 68:241312 .
doi : 10.1103/physrevb.68.241312 brockhouse bn ( 1959 ) lattice vibrations in silicon and germanium .
doi:10.1103/physrevlett.2.256 kumar r , mavi hs , shukla ak , vankar vd ( 2007 ) photoexcited fano interaction in laser - etched silicon nanostructures .
j appl phys 101:064315 .
doi : 10.1063/1.2713367 kumar r , shukla ak ( 2009 ) quantum interference in the raman scattering from the silicon nanostructures .
phys lett a 373:28822886 .
doi : 10.1016/j.physleta.2009.06.005 shukla ak , kumar r , kumar v ( 2010 ) electronic raman scattering in the laser - etched silicon nanostructures .
j appl phys 107:014306 .
doi : 10.1063/1.3271586 adu kw , xiong q , gutierrez hr , chen g , eklund pc ( 2006 ) raman scattering as a probe of phonon confinement and surface optical modes in semiconducting nanowires .
phys . a 85:287 - 297 .
doi : 10.1007/s00339 - 006 - 3716 - 8 adu kw , gutierrez hr , kim uj , sumanasekera gu and eklund pc ( 2005 ) confined phonons in si nanowires .
nanoletters 5:409414 .
doi : 10.1021/nl048625 . | a step - by - step evolution of an asymmetric raman line - shape function from a lorentzian line - shape is presented here for low dimensional semiconductors .
the evolution reported here is based on the phonon confinement model which is successfully used in literature to explain the asymmetric raman line - shape from semiconductor nano - structures .
physical significance of different terms in the theoretical asymmetric raman line - shape has been explained here .
better understanding of theoretical reasoning behind each term allows one to use the theoretical raman line - shape without going into details of theory from first principle .
this will enable one to empirically derive a theoretical raman line - shape function for any material if information about its phonon dispersion , size dependence etc is known . |
there is clear evidence for a nearly harrison zeldovich and gaussian spectrum of perturbations from cosmic microwave background and large scale structure observations .
one of the most pressing questions in early
universe cosmology is to understand the mechanism underlying the generation of this primordial spectrum .
inflation @xcite is one explanation , but it is important to understand what other scenarios predict a spectrum of curvature perturbations consistent with observations .
this question has motivated much research into alternative theories ; for instance the pre
big bang scenario of @xcite@xmath0@xcite , string gas cosmology @xcite@xmath0@xcite and ekpyrotic cosmology @xcite@xmath0@xcite .
production of a scale invariant spectrum of fluctuations by itself is not enough it is also highly desirable for the background solution to be an _
attractor_. technically , this is achieved by demanding that the curvature perturbation on uniform - density hypersurfaces , @xmath1 , goes to a constant in the long wavelength limit @xmath2 . in this limit , @xmath3 is interpreted as a constant perturbation of the scale factor , which may therefore be absorbed locally by a spatial diffeomorphism @xcite . for models involving a single , canonical scalar field ( _ i.e. _ with unit sound speed , @xmath4 ) minimally coupled to einstein gravity
, it has been shown recently that there are only _ three _ independent cosmological solutions which produce a scale invariant spectrum of curvature perturbations on an attractor background @xcite , assuming vacuum initial conditions .
see @xcite@xmath0@xcite for related work .
the most well known of these solutions is of course inflation @xcite , which relies on exponential expansion of the background with @xmath5 .
more recently , the adiabatic ekpyrotic @xcite scenario has been proposed , in which a scale invariant spectrum is produced by a rapidly evolving equation of state @xmath6 on a slowly contracting background . the third solution can be viewed as a variant of adiabatic ekpyrosis , where curvature perturbations are again sourced by a rapidly changing equation of state , but this time on a slowly expanding background @xcite . at the level of the two - point function , these three scenarios yield indistinguishable power spectra .
however , the degeneracy is broken at the three point level .
the non inflationary solutions have strongly scale dependent non
gaussianities @xcite , which can be traced to the rapid growth of the equation of state parameter . in these models , @xmath7 grows rapidly at small scales and perturbative control is lost when @xmath8 .
this difficulty can be avoided by suitably modifying the potential so that @xmath1 becomes much smaller on small scales .
but this in turn restricts the range of scale invariant and gaussian modes to about 5 decades ( @xmath9 ) or @xmath10 e - folds in @xmath11space @xcite . as a result , with @xmath4 and attractor background ,
inflation is the unique single field mechanism capable of producing many decades of scale invariant and gaussian perturbations .
( of course , as a theory of the early universe inflation must still surmount some foundational issues , such as the measure problem and low - entropy initial conditions @xcite .
here we leave aside these critical questions and note that inflation , viewed as a mechanism for generating density perturbations , remains weakly coupled over a large range of modes . ) in this paper we generalize the analysis to the case of time - varying sound speed , @xmath12 , as obtained , for instance , with non - canonical scalar fields . with einstein gravity plus a single degree of freedom ,
the sound speed is the only remaining knob at our disposal . as shown in @xcite , allowing for @xmath12 greatly broadens the realm of allowed cosmologies that yield a scale invariant power spectrum . in particular , any cosmology with constant equation of state can be made scale invariant by suitably choosing the evolution of the sound speed . in this work
we show that non - gaussianities impose stringent constraints on the allowed cosmologies .
our analysis is very general and applies to _ arbitrary _ time - dependent @xmath13 and @xmath12 , with the only restriction that the null energy condition be satisfied : @xmath14 .
we begin by reviewing how the time dependence of the sound speed results in an effective cosmological background for the curvature perturbation , as was first shown in @xcite . in this effective background , which depends both on the evolution of the scale factor and the sound speed , @xmath1 propagates at the speed of light .
we derive a consistency equation that the scale factor and the sound speed must satisfy in order to have scale invariance at the two point level . in the spirit of @xcite ,
given an evolution for the scale factor , solving this equation gives a suitable evolution for the sound speed for which @xmath1 has a scale invariant two point function on an attractor background .
this shows that a time dependent sound speed vastly increases the degeneracy at the two point level . as in the canonical case ,
this degeneracy is generically broken by the three point function .
in particular , if the three point function is strongly scale dependent , we generically expect the theory to become strongly coupled either in the infrared ( ir ) or in the ultraviolet ( uv ) . to avoid such perturbative breakdown , we demand that certain contributions to the three point function be scale invariant .
this turns out to be extremely restrictive : we show that slow
roll inflation is the _
unique _ cosmology with this property .
conversely , if the three point function is not scale invariant , then non gaussianities will increase rapidly with scale , resulting in a finite range ( @xmath15 modes ) of perturbations consistent with observations , as in the canonical case .
this is a remarkable fact ; it is extremely surprising , in light of the vast degeneracy afforded by a variable sound speed , that slow
roll inflation should be the unique possibility .
there is , navely , a way to avoid our argument ; since all the vertices in the cubic action are proportional to @xmath16 , it may be possible to have a strongly scale dependent three
point function while remaining weakly coupled by taking @xmath17 .
there has recently been much interesting work done related to this idea .
a concrete example of this limit is given by the scenario considered in @xcite@xmath0@xcite , based on earlier ideas in @xcite@xmath0@xcite .
in these scenarios , the sound speed is assumed @xmath18 in the very early universe and decreases in time .
it should be noted that theories with superluminal propagation can not be uv completed by a theory with an analytic s
matrix , for example local quantum field theory or perturbative string theory @xcite .
however , for the sake of generality , we remain agnostic about the nature of any uv completion of the theories we consider .
we review the particular superluminal mechanism where @xmath19 and @xmath20 are constant and show that such cosmologies are also tightly constrained by strong coupling considerations . again in these cases , only a finite range of scale invariant modes can be generated without hitting strong coupling . in order to understand this constraint , recall that the @xmath1 power spectrum is given by p_~o()~10 ^ -10 .
it is immediately clear that at early times , when @xmath17 , there is a tension between keeping @xmath21 and maintaining the observed amplitude for density perturbations . for @xmath22 ,
this keeps us from taking @xmath23 to be too large , which bounds the range of scale invariant modes to about 8 decades ( @xmath24 ) .
this tension may be alleviated by taking @xmath19 to be exponentially large , but this causes the amplitude of non gaussianities to grow . because @xmath23 is decreasing in time , demanding that the theory remain weakly coupled imposes a bound of about 12 decades ( @xmath25 ) on mode production . though this range is significantly larger than the @xmath26 viable modes in the canonical case , we still have to account for the coincidence that these overlap with the observable window .
our results have some overlap with that of baumann _ et al . _
@xcite , who reached similar conclusions .
however , our analysis is much more general in the variable sound speed case , baumann _ et al _ considered only sub
luminal @xmath23 and nearly constant @xmath19 .
in contrast , we allow _ arbitrary _ evolution for @xmath19 and @xmath23 including the superluminal regime in section [ threepoint ] and analyze in detail the constant @xmath19 superluminal scenarios in section [ loophole ] .
we begin by considering how a varying sound speed affects the classification of scale invariance at the two point level recently undertaken in @xcite .
we make no assumptions about the underlying dynamics , only that they may be well modeled by a perfect fluid .
perturbing around a friedmann robertson walker ( frw ) background in co moving gauge , where @xmath27 and @xmath28 , the quadratic action for @xmath1 is given by @xcite s_2=m_pl^2d^3xd z^2 , where @xmath29 , and @xmath30 denotes conformal time , @xmath31 .
this action is familiar from canonical single field models , except for the sound speed factor multiplying the spatial gradient term and appearing in the measure factor . in order to eliminate this complication , following
@xcite we define the sound horizon time coordinate by @xmath32 .
( note that when @xmath33 , the variable @xmath34 measures the size of the sound horizon . ) additionally , we define qz = . in terms of these new variables , the quadratic @xmath1 action takes the familiar form s_2=m_pl^2d^3xdy
q^2 , where primes indicate derivatives with respect to @xmath34 .
the virtue of this change of variables is manifest @xmath1 now propagates luminally , but in _ effective _ cosmological background defined both by the scale factor and the sound speed .
the mode functions of the canonically normalized field , @xmath35 , obey the familiar - looking equation of motion v_k+(k^2-)v_k=0 .
[ mode ] this has exactly the same form as the corresponding equation in the canonical case , but with @xmath36 replacing @xmath37 , a manifestation of the curvature perturbation evolving in an effective geometry .
assuming the usual adiabatic vacuum , it is well known that ( [ mode ] ) will yield a scale invariant spectrum of perturbations provided that = .
[ qeq ] note that modes freeze out when @xmath38 , which corresponds to sound - horizon crossing in the constant @xmath23 case , hence we take @xmath39 .
the solution for the mode functions is then v_k(y ) = ( 1-)e^iky , [ vk ] which describes a scale invariant spectrum , @xmath40 , in the limit @xmath41
. equation ( [ qeq ] ) has two solutions , @xmath42 and @xmath43 , but only the former describes a background which is a dynamical attractor . to see this , note that in the long wavelength ( @xmath44 ) limit we have the following expression for the power spectrum of the solution ( [ vk ] ) p_= k^3_k^2 ~ , which is indeed independent of @xmath11 . when @xmath45 , @xmath46 outside the horizon , indicating perturbative stability @xcite .
the other solution @xmath43 , however , implies that @xmath1 grows outside the horizon , @xmath47 , signaling that the background is unstable .
since we are interested in attractor backgrounds , we henceforth ignore the @xmath43 solution .
recalling the definition of @xmath36 , the condition for scale invariance in an attractor background may therefore be succinctly expressed as q^2== , [ inv ] where @xmath48 is an arbitrary ( positive ) constant .
it is important to note that @xmath49 and @xmath19 are not independent degrees of freedom , but are related by @xmath50 .
changing time variables to @xmath34 , this relation becomes = = ( ) . using the condition ( [ inv ] ) for scale invariance
, we can rewrite this to obtain the master equation a()= .
[ diffeq ] this equation ensures a scale invariant spectrum on an attractor background .
as noted in @xcite , in the case where @xmath33 , this equation may be recast as a particular instance of the generalized emden fowler equation . for completeness
, we review the results of @xcite . in the case of constant sound speed ( without loss of generality
, we may take @xmath4 ) , there are three distinct scale invariant solutions : * _ inflation _ is a solution where the scale factor grows as @xmath51 and the equation of state parameter is constant @xmath52 @xcite . to check that this is in fact scale invariant
, we note that @xmath53 , where @xmath54 because @xmath23 is constant . * _ adiabatic ekpyrosis _ is a solution where the equation of state parameter varies rapidly , @xmath55 , while the background remains nearly static , @xmath56 .
again , we can check that this gives a scale invariant spectrum @xmath57 .
in fact , this corresponds to two distinct solutions , one where the background is slowly _ contracting _ @xcite and one where the background is slowly _ expanding _
it is important to note that in these scenarios modes freeze out on sub
hubble scales and are subsequently pushed outside the horizon during a contracting ekpyrotic phase with constant @xmath58 .
returning to the general case , given any evolution for @xmath49 we can find an evolution of @xmath23 that will make the spectrum of perturbations scale invariant by solving ( [ diffeq ] ) .
alternatively , specifying a relation between the evolution of @xmath23 and @xmath49 is sufficient to determine the evolution . as a result
, we see that there is an enormous amount of degeneracy at the two point level .
an excellent illustration of this degeneracy is the case of cosmologies with constant @xmath19 . with constant @xmath23 , as reviewed above inflation is the only solution that has constant @xmath19 .
but for more general sound speed , there is a power - law evolution for @xmath23 that yields a scale invariant spectrum for arbitrary positive values of @xmath19 . indeed ,
constancy of @xmath19 and @xmath59 is sufficient to deduce the scaling solutions a~(-y)^ , c_s~(-y)^ .
[ acspowerlaw ] inserting these expressions into ( [ diffeq ] ) , we find that the solution is scale invariant for @xmath60 , in agreement with @xcite .
non - gaussianities offer a powerful tool for differentiating between the different cosmologies with degenerate power spectra .
since the precise form of the cubic action depends the underlying physics , we must choose to parameterize the microphysics in some way .
a convenient and quite general choice is to consider a non
canonical scalar field @xmath61 , described by a @xmath62 lagrangian s = d^4 x , [ action ] where @xmath63 .
it is straightforward to show that the energy density and sound speed are related to the choice of @xmath64 by @xcite = 2xp_,x - p ; c^2_s= .
the cubic action for @xmath1 is derived in @xcite . making the transformation to the sound horizon time variable @xmath32
, the action takes the form , up to a field redefinition , @xmath65~ , \label{s3pt } \end{split}\end{aligned}\ ] ] where @xmath66 , and = x^2p_,xx+x^3p_,xxx ; = xp_,x+2xp_,xx= .
although we focus on a particular class of microphysical models , namely @xmath62 theories , the model dependence of the action is encoded only in @xmath67 .
all other vertices in the cubic action are functions of the scale factor and the sound speed .
for example , for a dbi action , the @xmath68 term in ( [ s3pt ] ) vanishes identically @xcite .
since the form of this first term will not be material to our arguments , our analysis even at the cubic level is rather general , but there may be some potential model
dependent effects from the @xmath68 vertex which we have not considered . to estimate non gaussianities , a useful approximation is the _ horizon crossing approximation _ , whereby @xmath69 is estimated by f_nl~._ky=1 .
[ fnl ] here @xmath70 and @xmath71 are terms in the quadratic and cubic lagrangians , respectively . since temporal and spatial gradients are comparable at horizon crossing ( @xmath72 ) , we may trade them freely in ( [ fnl ] ) .
the horizon - crossing approximation generally offers a good estimate of @xmath69 since modes are in their ground state at early times when they are far inside the horizon and become constant outside the horizon .
we therefore expect non - gaussianities to peak around horizon crossing .
contributions peak at late times , well after horizon crossing @xcite .
although @xmath1 goes to a constant outside the horizon , the rapid growth of the vertex , @xmath73 , overwhelms the suppression from @xmath1 derivatives becoming small .
thus , the horizon - crossing approximation is a conservative estimate of non - gaussianities . ] at a classical level , perturbations are highly non - gaussian for @xmath74 , corresponding to @xmath75 , and classical perturbation theory breaks down . at a quantum level , the right hand side of ( [ fnl ] ) also offers an estimate for the magnitude of loop corrections to the two - point function @xcite .
thus , classical and quantum perturbation theory break down , and the theory becomes strongly coupled , whenever ~1 , or @xmath8 .
this is the same strong coupling criterion used in @xcite .
in particular , if @xmath69 is strongly scale dependent , then the growth of non - gaussianities will generically lead to a breakdown of perturbation theory either in the ir or in the uv .
( as mentioned earlier , an exceptional case is the limit @xmath17 , which will be treated separately in section [ loophole ] . )
this expectation is borne out by the analysis of the canonical case @xcite . to avoid strong coupling , therefore , we demand that @xmath69 be approximately scale invariant . amongst the terms in the cubic action
is the vertex s_3
d^3xdy .
[ s3particular ] evaluating this vertex at horizon crossing , we find that its non - gaussian contribution is f_nl^^3 ~._k y=1 ~()^2 .
[ fnleps3 ] substituting the condition ( [ inv ] ) for scale invariance at the two point level , @xmath76 , this reduces to f_nl^^3 ~_k y=1 .
now , in order for the full three point function to be scale invariant , a necessary condition is that the contribution from this vertex be scale invariant , barring miraculous cancellations .
this implies that the scale factor must be growing as a ~ , [ yinf ] which corresponds to an effective de sitter geometry .
remarkably , simply demanding scale invariance of the two- and three - point correlation functions , without any consideration of the independent dynamics of @xmath49 and @xmath23 , has led us to focus on backgrounds that are effectively de sitter , albeit in terms of the @xmath34 variable .
thus the question becomes is it possible to have inflation without inflation ? by this , we mean , is there an evolution where the modes see an effective de sitter space in terms of the @xmath34 variable but for which the true geometry is far from de sitter ?
unfortunately , the answer appears to be no , as we now argue .
with @xmath77 , ( [ inv ] ) immediately implies that @xmath78 , where @xmath79 is an arbitrary ( positive ) constant .
this is all we need to solve ( [ diffeq ] ) , with the result c_s(y ) = ; ( y ) = ; [ zero ] where @xmath80 .
both @xmath19 and @xmath23 start out infinite and decrease rapidly to zero . by construction ,
this solution is scale invariant at the two point level and the aforementioned three
point vertex is also scale invariant . at first sight
, we might expect that this solution is far from de sitter because @xmath58 initially .
however , because @xmath19 is decreasing so rapidly , by the time @xmath81 ; @xmath19 is already less than unity , indicating an inflationary spacetime . as such
, this solution is only a small deformation away from the de sitter geometry , specifically only about one e
fold of evolution is non inflationary .
it is worth noting that while the particular vertex ( [ s3particular ] ) we have considered yields a time - independent ( and therefore scale invariant ) non - gaussian contribution once ( [ yinf ] ) is imposed , some of the other vertices vary logarithmically with @xmath34 , such as the @xmath82 vertex .
but this slow growth as a function of scale is of course acceptable from the viewpoint of avoiding large non gaussianities and the breakdown of perturbation theory .
the analysis thus far strongly suggests that if we want scale invariance at the level of the three - point function , we are forced to consider an evolution that is very close to de sitter that is , the cosmological background must inflate .
the intuition for this statement is that every term in the cubic action is multiplied by powers of @xmath19 . in the case of slow - roll inflation ,
we take @xmath83 and therefore get a nearly gaussian spectrum .
note that there is another factor common to every term , namely @xmath16 .
one might therefore suspect that a nearly gaussian spectrum could also result from taking @xmath17 , even with a strongly scale dependent @xmath69 .
this will be treated separately in section [ loophole ] , where we will show that other considerations should greatly constrain this case as well .
one may wonder whether the assumption of scale invariant three - point function is not too stringent .
a possible loophole would be to have non - gaussianities peak in the ir , where the onset of strong coupling would lie on unobservably large scales .
however it is clear from the above analysis that this would require a scale factor that expands more rapidly than @xmath84 , which is super inflationary in terms of the @xmath34 variable .
we have studied this case by generalizing the above analysis to power - law dependencies for @xmath69 , and found that inflation ( in terms of cosmic time ) remains the only solution .
given the above arguments , it seems as though it is impossible to have small non - gaussianities over a broad range of modes without invoking a period of inflation .
however , there is an illuminating exceptional case which we now treat separately . as we noted in passing earlier
, every term in the cubic action has a factor of @xmath16 multiplying it . in analogy with slow
roll inflation , we may imagine controlling the size of the three - point function by making this factor very small , corresponding to superluminal @xmath23 . in principle , this should allow for the production of modes with an extremely gaussian spectrum over a broad range in @xmath11space @xcite . in practice , however , we will find that only a finite range of scale invariant and gaussian modes can be generated by such mechanisms .
superluminal scenarios have been proposed recently @xcite@xmath0@xcite , building on previous work on decaying sound speed cosmologies @xcite@xmath0@xcite .
these models take advantage of having an extremely large speed of sound at early times which suppresses non gaussianities on large scales . by virtue of the fact that the sound horizon at early times is much larger than the co moving hubble radius ,
the horizon problem is addressed .
the large sound speed decays as time goes on , causing the co moving sound horizon to shrink in time .
see fig .
[ horizons ] . in the same manner that a shrinking hubble horizon causes modes to freeze out in inflation , modes freeze out when they exit the sound horizon in this case . additionally , the flatness and horizon problems are effectively decoupled in this scenario .
of course , these models manifestly rely on superluminality , which , as mentioned before , makes them unsuitable for uv completion within a theory that is local in the usual sense @xcite .
current models in the literature @xcite are cast in terms of constant @xmath19 cosmologies mentioned earlier and studied extensively in @xcite . in these scenarios , we assume scaling solutions for @xmath49 and @xmath23 given by ( [ acspowerlaw ] ) : variable evolves , the comoving sound horizon decreases and the comoving hubble horizon grows until the two meet at the coincidence scale denoted by the dashed line.,width=240 ] a=()^ ; c_s=()^ , [ acsgg1 ] where @xmath85 is the time when @xmath86 .
using the freedom to rescale spatial coordinates , we have set @xmath87 at this time also .
the hubble parameter is given by h=()^ .
[ hcsgg1 ] the contribution to @xmath69 given by ( [ fnleps3 ] ) scales as f_nl^^3 ~()^2 ~_k|y| = 1 ~k^4 , [ fnlcsgg1 ] which thus peaks on small scales . despite this strong scale dependence
, the range of modes accessible to observations can be highly gaussian by having @xmath17 , which amounts to making the coefficient in ( [ fnlcsgg1 ] ) very small .
first note that the hubble horizon at mode freeze - out and the amplitude of the power spectrum have a simple relationship .
indeed , the power spectrum in the constant @xmath19 case is given by @xcite p_=_k|y| = 1 = ( ky_1)^ , [ pcsgg1 ] which confirms the statement made earlier that exact scale invariance requires @xmath60 .
we will henceforth assume that @xmath60 .
it then follows from ( [ pcsgg1 ] ) that the hubble parameter and the power spectrum are related by h^2 = m_pl^2 c_s p _ , [ h2interm ] where everything is evaluated at freeze - out .
staring at this expression , it is immediately clear that in the limit @xmath17 a tension arises between matching the observed normalization of @xmath88 and keeping @xmath89 . to be completely general , suppose that the phase of decaying sound speed lasts from some initial time , @xmath90 , to some final time @xmath91 . without loss of generality
, we will assume that @xmath92 there is no gain in having the decaying @xmath23 phase lasts any longer , since the modes generated when @xmath93 will never be outside the hubble horizon .
thus the range @xmath94 of scale invariant modes is given by @xmath95 .
substituting this into ( [ hcsgg1 ] ) , we find that the ratio of the hubble parameter at the beginning and end of the decaying @xmath23 phase satisfies = ( ) ^ = n^ .
[ hratio ] meanwhile , from ( [ h2interm ] ) we have 10 ^ -5 , where we have used the observational constraint @xmath96 .
we therefore obtain 10 ^ -5 n^ . for consistency of classical gravity , we demand that the hubble parameter be sub
planckian at the onset the decaying sound speed phase , @xmath97 .
this gives the following upper bound on @xmath94 : n ( ) ^ .
[ bound1 ] remarkably , this bound on the range of scale invariant modes follows merely by requiring the correct amplitude for perturbations and that the physics be sub
so far we have said nothing about strong coupling issues .
it turns out that the bound can be made more precise by considering non - gaussianities . as before ,
consider the contribution to @xmath69 given by ( [ fnlcsgg1 ] ) .
demanding that this be consistent with perturbation theory requires f_nl^^3 ~()^210 ^ -5 1 , which implies the lower bound @xmath98 .
inserting this into ( [ bound1 ] ) , the bound on @xmath94 becomes n ^ .
[ bound2 ] of course we are interested in the case @xmath99 , for otherwise the universe would be inflating .
the bound ( [ bound2 ] ) is weakest for @xmath100 , which is sensible , since in this regime @xmath23 has to be just barely superluminal in order to suppress non gaussianities . in this case , ( [ bound2 ] ) enforces @xmath101 for the range of scale invariant modes in @xmath11space .
while this is sufficient to account for microwave background and large scale structure observations , it does represent a relatively narrow range within which the observable window must lie . to summarize
, it appears that strong coupling considerations also greatly constrain scenarios based on superluminality .
the bound on the allowed range of scale invariant modes is weaker than in the canonical case ( @xmath102 versus @xmath103 , respectively ) , but is nevertheless restrictive .
the scenario must somehow explain the coincidence of why the observational window happens to fall within the limited scale invariant range .
although here we have focused on the particular realization ( [ acsgg1 ] ) , we expect that similar arguments will bound mode production in other realizations of the limit @xmath17 .
a fundamental objective of early universe cosmology is to explain the origin of the observed spectrum of density perturbations .
the primordial perturbations inferred from observations are nearly scale invariant and highly gaussian .
any mechanism that purports to explain their origin should be capable of generating perturbations with the desired properties over many decades in @xmath11space .
the present paper concludes a body of work , initiated in @xcite , aimed at identifying _ all _ possible cosmologies capable of generating scale invariant and gaussian perturbations .
the analysis assumes a single degree of freedom ( purely adiabatic perturbations ) , and a background which represents a dynamical attractor . in the case of @xmath4 , it was shown in @xcite that only inflation can generate suitable perturbations over many decades in @xmath11-space .
two other phases , corresponding to adiabatic ekpyrosis , can produce at most @xmath104 modes ( or 12 e - folds ) of scale invariant and gaussian modes , before hitting strong coupling .
et al . _
@xcite also considered a time - dependent sound speed , but their analysis was restricted to sub
luminal @xmath23 and nearly constant @xmath19 .
in this paper , we have generalized the analysis by allowing for _ arbitrary _ time - dependent @xmath13 and @xmath12 , with the only restriction that the null energy condition be satisfied : @xmath14 . at the two - point level
, any cosmological background can in principle yield a scale invariant spectrum , provided that the sound speed evolves in a suitable way . to avoid strong coupling we have argued that the three - point function should generically also be nearly scale invariant . by demanding that a particular contribution to @xmath69 be scale invariant , we have found that inflation remains the unique solution .
consequently , non inflationary solutions have strongly scale dependent non gaussianities , and therefore the range of modes consistent with observations is subject to a bound similar to the canonical case .
one exceptional case is the limit of superluminal sound speed , which has attracted considerable attention recently @xcite@xmath0@xcite . for concreteness ,
we have focused on superluminal scenarios where both @xmath19 and @xmath20 are constant in time . in this case , non - gaussianities are strongly scale dependent , but strong coupling is navely avoided by making @xmath17 . while this does alleviate the constraints , we have shown that the combined requirements of sub planckian hubble parameter and controlled non - gaussianities imply a bound of at most @xmath105 modes ( or 28 e - folds ) with nearly scale invariant and gaussian statistics . this bound is weaker than in the canonical case , but is nevertheless quite restrictive . although we have not considered it in detail here , we expect arguments similar to those in section [ loophole ] to bound mode production in more general superluminal scenarios .
we conclude that , under the assumptions of single degree freedom , attractor background and adiabatic vacuum initial conditions , inflation is the only mechanism capable of generating suitable perturbations over a wide range of scales .
alternative mechanisms must inevitably rely either on an instability of the background , non - vacuum initial state , and/or multiple degrees of freedom .
_ note added : _ as this paper was being completed , reference @xcite appeared , which reaches conclusions similar to ours using different arguments .
specifically , the authors find that that in order to produce more than 7 e folds of scale invariant modes in a monotonically expanding universe without inflation barring super
planckian energy densities a superluminal speed of sound must be invoked . however , this result relies on the assumption of an expanding background , while our arguments do not require this constraint . _
acknowledgments : _ it is our pleasure to thank bryan chen , kurt hinterbichler , mitchell lerner , godfrey miller and leonardo senatore for helpful discussions .
this work was supported in part by funds from the university of pennsylvania , nsf grant phy-0930521 , and the alfred p. sloan foundation .
a. h. guth , `` the inflationary universe : a possible solution to the horizon and flatness problems , '' phys .
d * 23 * , 347 ( 1981 ) ; a. albrecht and p. j. steinhardt , `` cosmology for grand unified theories with radiatively induced symmetry breaking , '' phys .
lett . * 48 * , 1220 ( 1982 ) ; a. d. linde , `` a new inflationary universe scenario : a possible solution of the horizon , flatness , homogeneity , isotropy and primordial monopole problems , '' phys .
b * 108 * , 389 ( 1982 ) .
m. gasperini and g. veneziano , `` pre - big bang in string cosmology , '' astropart .
* 1 * , 317 ( 1993 ) [ arxiv : hep - th/9211021 ] . m. gasperini and g. veneziano , `` the pre - big bang scenario in string cosmology , '' phys
. rept . * 373 * , 1 ( 2003 ) [ arxiv : hep - th/0207130 ] .
m. gasperini and g. veneziano , `` string theory and pre - big bang cosmology , '' arxiv : hep - th/0703055 .
r. h. brandenberger and c. vafa , `` superstrings in the early universe , '' nucl .
b * 316 * , 391 ( 1989 ) .
a. nayeri , r. h. brandenberger and c. vafa , `` producing a scale - invariant spectrum of perturbations in a hagedorn phase of string cosmology , '' phys . rev .
lett . * 97 * , 021302 ( 2006 ) [ arxiv : hep - th/0511140 ] .
r. h. brandenberger , a. nayeri , s. p. patil and c. vafa , `` tensor modes from a primordial hagedorn phase of string cosmology , '' phys .
lett . * 98 * , 231302 ( 2007 ) [ arxiv : hep - th/0604126 ] .
r. h. brandenberger , a. nayeri , s. p. patil and c. vafa , `` string gas cosmology and structure formation , '' int .
j. mod .
a * 22 * , 3621 ( 2007 ) [ arxiv : hep - th/0608121 ] .
r. h. brandenberger _
_ , `` more on the spectrum of perturbations in string gas cosmology , '' jcap * 0611 * , 009 ( 2006 ) [ arxiv : hep - th/0608186 ] .
t. battefeld and s. watson , `` string gas cosmology , '' rev .
* 78 * , 435 ( 2006 ) [ arxiv : hep - th/0510022 ] .
j. khoury , b. a. ovrut , p. j. steinhardt and n. turok , `` the ekpyrotic universe : colliding branes and the origin of the hot big bang , '' phys .
d * 64 * , 123522 ( 2001 ) [ arxiv : hep - th/0103239 ] .
r. y. donagi , j. khoury , b. a. ovrut , p. j. steinhardt and n. turok , `` visible branes with negative tension in heterotic m - theory , '' jhep * 0111 * ( 2001 ) 041 [ arxiv : hep - th/0105199 ] . j. khoury , b. a. ovrut , n. seiberg , p. j. steinhardt and n. turok , `` from big crunch to big bang , '' phys .
d * 65 * , 086007 ( 2002 ) [ arxiv : hep - th/0108187 ] .
j. khoury , b. a. ovrut , p. j. steinhardt and n. turok , `` density perturbations in the ekpyrotic scenario , '' phys .
d * 66 * , 046005 ( 2002 ) [ arxiv : hep - th/0109050 ] .
d. h. lyth , `` the primordial curvature perturbation in the ekpyrotic universe , '' phys .
b * 524 * , 1 ( 2002 ) [ arxiv : hep - ph/0106153 ] .
r. brandenberger and f. finelli , `` on the spectrum of fluctuations in an effective field theory of the ekpyrotic universe , '' jhep * 0111 * , 056 ( 2001 ) [ arxiv : hep - th/0109004 ] .
p. j. steinhardt and n. turok , `` cosmic evolution in a cyclic universe , ''
d * 65 * , 126003 ( 2002 ) [ arxiv : hep - th/0111098 ] .
a. notari and a. riotto , `` isocurvature perturbations in the ekpyrotic universe , '' nucl .
b * 644 * , 371 ( 2002 ) [ arxiv : hep - th/0205019 ] .
f. finelli , `` assisted contraction , '' phys .
b * 545 * , 1 ( 2002 ) [ arxiv : hep - th/0206112 ] .
s. tsujikawa , r. brandenberger and f. finelli , `` on the construction of nonsingular pre - big - bang and ekpyrotic cosmologies and the resulting density perturbations , '' phys .
d * 66 * , 083513 ( 2002 ) [ arxiv : hep - th/0207228 ] .
s. gratton , j. khoury , p. j. steinhardt and n. turok , `` conditions for generating scale - invariant density perturbations , '' phys .
d * 69 * , 103505 ( 2004 ) [ arxiv : astro - ph/0301395 ] .
a. j. tolley , n. turok and p. j. steinhardt , `` cosmological perturbations in a big crunch / big bang space - time , '' phys .
d * 69 * , 106005 ( 2004 ) [ arxiv : hep - th/0306109 ]
. b. craps and b. a. ovrut , `` global fluctuation spectra in big crunch / big bang string vacua , '' phys .
d * 69 * , 066001 ( 2004 ) [ arxiv : hep - th/0308057 ] .
j. khoury , p. j. steinhardt and n. turok , `` great expectations : inflation versus cyclic predictions for spectral tilt , '' phys .
* 91 * , 161301 ( 2003 ) [ arxiv : astro - ph/0302012 ] .
j. khoury , p. j. steinhardt and n. turok , `` designing cyclic universe models , '' phys .
lett . * 92 * , 031302 ( 2004 ) [ arxiv : hep - th/0307132 ] .
j. khoury , `` a briefing on the ekpyrotic / cyclic universe , '' arxiv : astro - ph/0401579 .
p. creminelli , a. nicolis and m. zaldarriaga , `` perturbations in bouncing cosmologies : dynamical attractor vs scale invariance , '' phys .
d * 71 * , 063505 ( 2005 ) [ arxiv : hep - th/0411270 ] .
j. l. lehners , p. mcfadden , n. turok and p. j. steinhardt , `` generating ekpyrotic curvature perturbations before the big bang , '' phys .
d * 76 * , 103501 ( 2007 ) [ arxiv : hep - th/0702153 ] .
e. i. buchbinder , j. khoury and b. a. ovrut , `` new ekpyrotic cosmology , '' phys .
d * 76 * , 123503 ( 2007 ) [ arxiv : hep - th/0702154 ] .
e. i. buchbinder , j. khoury and b. a. ovrut , `` on the initial conditions in new ekpyrotic cosmology , '' jhep * 0711 * , 076 ( 2007 ) [ arxiv:0706.3903 [ hep - th ] ] .
e. i. buchbinder , j. khoury and b. a. ovrut , `` non - gaussianities in new ekpyrotic cosmology , '' phys .
lett . * 100 * , 171302 ( 2008 ) [ arxiv:0710.5172 [ hep - th ] ] .
p. creminelli and l. senatore , `` a smooth bouncing cosmology with scale invariant spectrum , '' jcap * 0711 * , 010 ( 2007 ) [ arxiv : hep - th/0702165 ] .
k. koyama and d. wands , `` ekpyrotic collapse with multiple fields , '' jcap * 0704 * , 008 ( 2007 ) [ arxiv : hep - th/0703040 ]
. k. koyama , s. mizuno and d. wands , `` curvature perturbations from ekpyrotic collapse with multiple fields , '' class .
* 24 * , 3919 ( 2007 ) [ arxiv:0704.1152 [ hep - th ] ] .
j. l. lehners and p. j. steinhardt , `` non - gaussian density fluctuations from entropically generated curvature perturbations in ekpyrotic models , '' phys .
d * 77 * , 063533 ( 2008 ) [ erratum - ibid .
d * 79 * , 129903 ( 2009 ) ] [ arxiv:0712.3779 [ hep - th ] ] .
j. l. lehners and p. j. steinhardt , `` intuitive understanding of non - gaussianity in ekpyrotic and cyclic models , '' phys .
d * 78 * , 023506 ( 2008 ) [ erratum - ibid .
d * 79 * , 129902 ( 2009 ) ] [ arxiv:0804.1293 [ hep - th ] ] .
j. l. lehners and p. j. steinhardt , `` non - gaussianity generated by the entropic mechanism in bouncing cosmologies made simple , '' phys .
d * 80 * , 103520 ( 2009 ) [ arxiv:0909.2558 [ hep - th ] ] .
j. khoury and p. j. steinhardt , `` adiabatic ekpyrosis : scale invariant curvature perturbations from a single scalar field in a contracting universe , '' phys .
lett . * 104 * , 091301 ( 2010 ) [ arxiv:0910.2230 [ hep - th ] ] .
a. linde , v. mukhanov and a. vikman , `` on adiabatic perturbations in the ekpyrotic scenario , '' jcap * 1002 * , 006 ( 2010 ) [ arxiv:0912.0944 [ hep - th ] ] .
j. l. lehners , `` ekpyrotic and cyclic cosmology , '' phys .
rept . * 465 * , 223 ( 2008 ) [ arxiv:0806.1245 [ astro - ph ] ] .
j. l. lehners , `` ekpyrotic non - gaussianity : a review , '' adv .
astron . *
2010 * , 903907 ( 2010 ) [ arxiv:1001.3125 [ hep - th ] ] .
r. brandenberger , `` matter bounce in horava - lifshitz cosmology , '' phys .
rev . * d80 * , 043516 ( 2009 ) .
[ arxiv:0904.2835 [ hep - th ] ] .
w. h. kinney and a. m. dizgah , `` flow in cyclic cosmology , '' phys .
d * 82 * , 083506 ( 2010 ) [ arxiv:1007.0753 [ astro-ph.co ] ] .
g. w. gibbons , n. turok , `` the measure problem in cosmology , '' phys .
rev . * d77 * , 063516 ( 2008 ) . [ hep - th/0609095 ] . | for a single scalar field with unit sound speed minimally coupled to einstein gravity , there are exactly three distinct cosmological solutions which produce a scale invariant spectrum of curvature perturbations in a dynamical attractor background , assuming vacuum initial conditions : slow - roll inflation ; a slowly contracting adiabatic ekpyrotic phase , described by a rapidly - varying equation of state ; and an adiabatic ekpyrotic phase on a slowly expanding background . of these three ,
only inflation remains weakly coupled over a wide range of modes , while the other scenarios can produce at most 12 e - folds of scale invariant and gaussian modes . in this paper , we investigate how allowing the speed of sound of fluctuations to evolve in time affects this classification . while in the presence of a variable sound speed there are many more scenarios which are scale invariant at the level of the two - point function , they generically suffer from strong coupling problems similar to those in the canonical case .
there is , however , an exceptional case with superluminal sound speed , which suppresses non - gaussianities and somewhat alleviates strong coupling issues .
we focus on a particular realization of this limit and show these scenarios are constrained and only able to produce at most 28 e - folds of scale invariant and gaussian perturbations .
a similar bound should hold more generally the condition results from the combined requirements of matching the observed amplitude of curvature perturbations , demanding that the hubble parameter remain sub - planckian and keeping non - gaussianities under control .
we therefore conclude that inflation remains the unique cosmological scenario , assuming a single degree of freedom on an attractor background , capable of producing arbitrarily many scale invariant modes while remaining weakly coupled .
alternative mechanisms must inevitably be unstable or rely on multiple degrees of freedom . |
quantum field theories of gravity are nonrenormalizable by power counting feynman diagrams .
this leads to the widely held belief that all unitary gravity field theories must be ultraviolet divergent at some loop order .
indeed , no known symmetry is powerful enough to render such theories ultraviolet finite . on the other hand , recent years
have made it abundantly clear that scattering amplitudes contain hidden symmetries and new structures beyond those expected from lagrangians .
while these are not yet fully understood , they can have profound consequences on ultraviolet properties . in this paper
, we identify a new class of multiloop ultraviolet cancellations that go beyond the ones established by standard - symmetry arguments .
we call these _ enhanced ultraviolet cancellations_. these are defined as cancellations that can not be displayed term - by - term in any local covariant diagrammatic formalism . by such a formalism
we mean that the poles in the diagram integrands are simply the standard feynman propagator ones . using the maximal cut conditions , as defined in ref .
@xcite , we can identify terms unique to a given diagram which we can then power count . to illustrate enhanced cancellations , we use previous three- and four - loop calculations in @xmath1 supergravity @xcite , as well as new calculations in @xmath0 supergravity performed here .
the study of the ultraviolet properties of gravity theories has a rich history , starting with the seminal work of t hooft and veltman @xcite .
they showed that pure einstein gravity is finite at one loop , but divergent with the addition of matter , a point on which other papers elaborated as well @xcite . goroff and sagnotti later showed that at two loops , pure einstein gravity diverges @xcite . with the addition of supersymmetry
, the ultraviolet behavior tends to improve : pure ungauged supergravities are known to have no divergences prior to three loops @xcite .
however , the consensus reached from studies in the 1980s was that all supergravity theories would likely diverge at the third loop order ( see , for example , ref .
@xcite ) , though one can raise the loop order with additional assumptions @xcite .
the complexity of gravity theories makes it difficult to explicitly check these expectations .
this situation was ameliorated by the advent of the unitarity method @xcite , which makes it possible to directly determine ultraviolet properties of gravity theories at high loop orders .
more recently , a new constraint on gauge - theory and gravity amplitudes has been introduced the duality between color and kinematics found by carrasco , johansson and one of the authors ( bcj ) @xcite allowing additional new nontrivial computations to be carried out . for maximally supersymmetric supergravity ( in @xmath4 this is @xmath5 supergravity ) @xcite ,
explicit calculations show that four - point amplitudes are finite at three loops for dimensions @xmath6
@xcite and at four loops for dimensions @xmath7 @xcite . in @xmath4 ,
these ultraviolet cancellations were subsequently understood to follow from supersymmetry and the @xmath8 duality symmetry of @xmath5 supergravity @xcite . a purely supersymmetric explanation has also been developed by bjrnsson and green @xcite using a field - theory version of the berkovits pure spinor formalism @xcite .
the current consensus based on symmetry considerations is that a @xmath9 counterterm is valid under all standard symmetries , leading to the expectations of a seven - loop divergence in @xmath4 and a five - loop divergence in @xmath10 .
while technical difficulties have prevented the @xmath5 supergravity expectations from being confronted by calculation , there is now evidence that implies even better behavior in this case than that suggested by standard - symmetry arguments : similar argumentation in half - maximal supergravity leads to predictions of valid counterterms in cases where no divergences exist . in particular , at three loops , half - maximal @xmath1 supergravity @xcite is ultraviolet finite in four dimensions @xcite , while similar supersymmetry and duality - symmetry considerations suggest that it should diverge @xcite .
( see ref .
@xcite for string - theory arguments for finiteness . ) in addition , half - maximal pure supergravity in @xmath11 is ultraviolet finite at two loops , again contrary to symmetry considerations @xcite .
an important question is whether it is possible that the observed three - loop finiteness of @xmath1 supergravity in @xmath4 can be explained using only arguments based on supersymmetry and duality symmetry .
an attempt to find such an explanation relied on the assumption of the existence of an appropriate non - lorentz covariant off - shell 16-supercharge superspace @xcite .
not surprisingly , the assumption carries other consequences as well : in particular , it predicts additional finiteness conditions when matter multiplets are added @xcite that directly contradict subsequent explicit calculations @xcite .
three - loop finiteness of pure @xmath1 supergravity therefore remains unexplained by standard - symmetry arguments .
nevertheless , it remains a key problem to understand the extent to which such arguments can restrict divergences . to carry out further probes of the ultraviolet properties of supergravity theories , together with smirnov and smirnov , we recently computed the four - loop four - point divergence of @xmath1 supergravity in @xmath4 @xcite , finding that the theory does diverge at four loops .
naively , this might suggest that all supergravity theories should diverge at some sufficiently high loop order .
however , when one looks at the details of the divergence , a rather different picture emerges : the divergence appears to be tied to the duality - symmetry anomaly of @xmath1 supergravity found by marcus @xcite .
the consequences of the anomaly on the amplitudes of @xmath1 supergravity have been described in some detail in ref .
the role of the anomaly implies that divergences of this type should not exist in @xmath12 supergravity , since these theories have no such analogous anomalies . in this paper
, we identify a subset of terms in @xmath5 supergravity that are ultraviolet divergent in four dimensions at seven loops , reproducing the analysis of bjrnsson and green @xcite from a different perspective .
to identify irreducible terms with poor power counting , we use maximal cuts .
the expectation of a seven - loop divergence is also consistent with other standard - symmetry arguments @xcite .
a key question is whether there are nontrivial enhanced cancellations between the divergent terms that then make the amplitude as a whole ultraviolet finite .
unfortunately , the high required loop order makes it unfeasible at present to test for the existence of enhanced cancellations in @xmath5 supergravity . instead
, here we demonstrate the presence of enhanced cancellations in @xmath1 and @xmath0 supergravities , since they are easier to work with , because the required loop order is lower .
as we demonstrate in this paper , enhanced cancellations are responsible for making the four - point @xmath1 supergravity amplitudes finite at three loops and @xmath0 supergravity amplitudes finite at four loops .
we also demonstrate three - loop cancellations in four - point @xmath0 supergravity amplitudes beyond those needed for finiteness .
such cancellations are reminiscent of the types of nontrivial cancellations noticed in certain unitarity cuts @xcite .
the surprising aspect is that no covariant local diagrammatic representation can make these results manifest .
what might be behind enhanced ultraviolet cancellations ? in a previous paper with huang @xcite , using the duality between color and kinematics , we explicitly tied the enhanced cancellations at two loops in half - maximal supergravity in @xmath11 to corresponding cancellations in pure yang - mills theory that prevent forbidden color factors from appearing in divergences .
a key feature is that the ultraviolet cancellations occur between the planar and nonplanar sectors of the theory .
this case is particularly simple to analyze in detail because the supergravity amplitudes are simple linear combinations of yang - mills amplitudes even after integration .
unfortunately , the situation beyond two loops is much more complex because different sets of integrals appear in the supergravity case than in the corresponding gauge - theory case . to carry out our computations
, we use the same techniques as those used for three and four loops @xcite in @xmath1 supergravity .
our computations make use of the many advances in constructing integrands , including the unitarity method @xcite and the duality between color and kinematics @xcite . while nonplanar integrands can not be uniquely defined , they can still be integrated to obtain unique results .
once the integrands are constructed , a mass is introduced as an infrared regulator
. we then series expand in small external momenta ( or equivalently large loop momenta ) to focus on ultraviolet singularities @xcite . at four loops the resulting vacuum integrals
are nontrivial .
to deal with them we use fire5 @xcite , which implements the laporta algorithm @xcite , to reduce the integrals to a basis set .
the basis integrals are known since they are identical to those used in the evaluation of the four - loop qcd @xmath13 function @xcite .
this paper is organized as follows . in
, we summarize the methods used to carry out the calculations .
in , we review the results of standard - symmetry power counting and show that power counting maximal cuts gives identical results .
then in , we exhibit the enhanced cancellations responsible for ultraviolet finiteness of @xmath1 supergravity at three loops @xcite .
we also present a new three - loop calculation in @xmath0 supergravity , pointing out that it too exhibits enhanced cancellations beyond those needed for ultraviolet finiteness . in
, we demonstrate that the four - point amplitudes of @xmath0 supergravity are all ultraviolet finite and again display enhanced cancellations .
we present our conclusions in .
the duality between color and kinematics and the associated gravity double - copy property @xcite make it simple to construct supergravity amplitudes once corresponding gauge - theory amplitudes are arranged into a form that makes the duality manifest .
( for a review of this duality and its applications , see ref .
@xcite . ) at loop level the duality remains a conjecture , but we rely only on explicitly constructed forms of @xmath1 super - yang - mills amplitudes where the duality is manifest @xcite . the duality between color and kinematics is usually formulated via graphs with only cubic vertices .
any @xmath14-loop @xmath15-point gauge - theory amplitude with all particles in the color - adjoint representation can be written in terms of such graphs as @xmath16 the sum labeled by @xmath17 runs over the set of distinct non - isomorphic graphs , while the sum over @xmath18 is over all @xmath19 permutations of external legs .
the symmetry factor @xmath20 removes over - counts arising from automorphisms of the diagrams .
the product in the denominator runs over all feynman propagators of graph @xmath17 , and the integrals are over @xmath14 independent @xmath21-dimensional loop momenta .
the color factor @xmath22 of graph @xmath17 is given by dressing every three - vertex with a group - theory structure constant , @xmath23 , while @xmath24 is the kinematic numerator of graph @xmath17 depending on momenta , polarizations and spinors .
so far this representation involves nothing more than absorbing contact - term contributions into graphs with only cubic vertices by multiplying and dividing by appropriate propagators .
the nontrivial part is the all - loop - order conjecture that there exists a form of gauge - theory amplitudes where kinematic numerators satisfy the same algebraic relations as color factors .
these are known as bcj representations of amplitudes . for the theories we discuss in this paper ,
this amounts to imposing the same jacobi identities on the kinematic numerators as those satisfied by adjoint - representation color factors : @xmath25 where the indices @xmath26 denote the diagram to which the color factors and numerators belong .
the basic jacobi identity is illustrated in embedded in an arbitrary diagram .
the numerator factors are also required to have the same antisymmetry properties as color factors . in general , the duality relations ( [ bcjduality ] )
work only after appropriate nontrivial rearrangements of the amplitudes .
remarkably , we can obtain corresponding gravity loop integrands simply by replacing color factors in a gauge - theory amplitude by kinematic numerators of a second gauge - theory amplitude where the duality is manifest @xcite : @xmath27 putting in the appropriate gravitational coupling gives us the double - copy form of gravity amplitudes , @xmath28 only one of the two sets of numerators @xmath24 or @xmath29 needs to satisfy the duality relation ( [ bcjduality ] ) @xcite .
we note that at tree level ( @xmath30 ) , the duality encodes the kawai - lewellen - tye @xcite relations between gauge - theory and gravity amplitudes , as well as nontrivial relations between color - ordered gauge - theory partial amplitudes @xcite . in this paper
we construct the three- and four - loop four - point @xmath0 supergravity integrands using the procedure presented above .
we do so by starting with an @xmath31 super - yang - mills integrand and then replacing the color factors with the bcj forms of kinematic numerators of @xmath1 super - yang - mills theory given in refs .
similar constructions of less - than - maximal supergravity amplitudes are found in refs .
we express @xmath32 where `` sugra '' and `` sym '' are shorthands for , respectively , supergravity and super - yang - mills theory .
we further decompose the @xmath0 amplitudes into a direct sum , @xmath33 where `` @xmath34 '' refers to ordinary pure nonsupersymmetric yang - mills theory .
the first term in the direct sum is the pure @xmath1 supergravity amplitude , while the second term is the difference between the @xmath0 and @xmath1 supergravity amplitudes .
the second term comes from taking the diagrams of pure @xmath31 super - yang - mills , subtracting out the pure - gluon part , and then replacing the color factors with the bcj numerators of @xmath1 super - yang - mills theory . on the @xmath31 super - yang - mills side
, this amounts to separating the pure gluon contributions from the contributions including also gluinos . following refs .
@xcite , we use ordinary feynman diagrams for the @xmath35 and @xmath31 super - yang - mills integrands .
while this might seem to be a poor starting point given the complexity of such diagrams , the bcj construction ( [ doublecopy ] ) ensures that only a small fraction of diagrams actually contribute .
whenever an @xmath1 super - yang - mills diagram vanishes , we do not need to evaluate corresponding diagrams in @xmath35 or @xmath31 super - yang - mills theory .
at one , two , three and four loops , the bcj - satisfying representations of the four - point amplitudes of @xmath1 super - yang - mills theory have only , respectively , 1 , 2 , 12 and 85 nonvanishing diagrams ( up to permutations of external legs ) . this is already a remarkable simplification , allowing the calculation to proceed .
the decomposition in eq .
results in an integrand where the @xmath1 supersymmetry cancellations are manifest , but the @xmath31 ones are not . while we do not do so here
, one could simplify the @xmath0 supergravity integrand to make cancellations from all supersymmetries manifest .
this could be accomplished by using the unitarity method to systematically move terms between diagrams , subject to maintaining the unitarity cuts . however , as we shall see below , no covariant local representation exists either in @xmath1 or @xmath0 supergravity that makes manifest the complete set of ultraviolet cancellations that we find .
because of this , there is no obvious way to avoid direct integration to see the cancellations .
we consequently call such cancellations enhanced .
once we have an integrand , the next step is to extract the ultraviolet divergences .
the procedure that we use has been described in some detail in ref .
@xcite , so here we only briefly summarize it . to deal with potential ultraviolet divergences we use dimensional reduction @xcite . rather than evaluate integrals with their full momentum dependence ,
it is much simpler to series expand the integrands prior to integration in order to pick up only the desired ultraviolet divergences @xcite .
this procedure introduces new unphysical infrared singularities beyond the standard ones , so one needs an infrared cutoff to separate the ultraviolet divergences from the infrared ones .
an especially good choice for regulating infrared singularities is to introduce a uniform mass into all feynman propagators prior to expanding in external momenta @xcite .
for the cases we study in this paper , where there are no lower - loop divergences , the subdivergences should all cancel amongst themselves with the use of this regulator .
the uniform mass regulator therefore greatly simplifies the computation since we do not need to compute subdivergences .
we have , however , performed extensive checks confirming that they cancel as expected .
we note that if the mass regulator were introduced later in the calculation , for example after the expansion in external momenta and tensor integral simplifications , it would ruin the cancellations of subdivergences between different integrals .
one would then need to include all subdivergence subtractions to properly remove the regulator dependence , greatly complicating the calculation .
the procedure results in a large number of vacuum integrals . at three loops , evaluating the integrals is straightforward @xcite , but at four loops it is a more serious challenge . to deal with this , we use the fire5 program @xcite , which is a highly - efficient implementation of integration - by - parts relations @xcite using the laporta algorithm @xcite .
it allows us to write down any given integral as a linear combination of a small number of so - called master integrals . in our four - loop calculation , the reduction to master integrals is especially nontrivial due to the high powers of numerator loop momenta that occur in gravity .
earlier related calculations already determined the four - loop vacuum master integrals @xcite ; we use the master - integral basis and values given in ref . @xcite .
.selected valid counterterms based on supersymmetry and duality - symmetry considerations @xcite .
@xmath36 is the number of supercharges . [
cols="<,^,^",options="header " , ] the restrictions supersymmetry and duality symmetry impose on counterterms have been studied in great detail over the years .
the most recent power - counting predictions based on symmetry considerations are collected in . in @xmath4 , apparently valid counterterms exist at loop orders @xmath37 in @xmath5 supergravity @xcite , @xmath38 in @xmath1 supergravity @xcite , and @xmath39 in @xmath0 supergravity @xcite . by increasing the space - time dimensions , one can also lower the loop order at which a potential counterterm can correspond to a divergence .
for example , in @xmath10 , maximal 32-supercharge supergravity has a valid five - loop counterterm @xcite .
similarly , half - maximal 16-supercharge supergravity in @xmath11 has an apparently valid two - loop counterterm @xcite . as explained in ref .
@xcite , the counterterms listed in can not be written as full - superspace integrals , but they do appear to be valid under all known standard - symmetry considerations .
see also refs .
@xcite for an attempt to put tighter restrictions on the counterterms and the associated difficulties with doing so .
bjrnsson and green @xcite constructed a first - quantized pure spinor formalism useful for power counting .
their formalism exposes all supersymmetry cancellations and gives an identical power count as other recent methods , including those that account for duality symmetry @xcite .
their results imply that unless there are some extra nonstandard cancellations beyond those implied by supersymmetry , @xmath5 supergravity will diverge at five loops in @xmath10 and at seven loops in @xmath4 , corresponding to the first and fourth rows of . we know that through four loops in @xmath5 supergravity , such symmetry - based predictions match the explicitly computed critical dimensions where a divergence first appears @xcite .
a key question is whether this pattern continues or whether there are cancellations beyond the well - understood ones . while it is not currently feasible to answer this for @xmath5 supergravity , we can answer it for @xmath1 and @xmath0 supergravity . from previous work @xcite
, we already know that the three - loop @xmath40 counterterm of @xmath1 supergravity in @xmath4 listed on the second row of does _ not _ result in a three - loop divergence .
similarly , half - maximal 16-supercharge supergravity at two loops in @xmath11 is free of divergences @xcite . as we discuss below in some detail for the three - loop @xmath1 case , these cancellations are a nontrivial manifestation of enhanced cancellations . however , one may worry that these two cases are special and not representative of a general pattern .
in particular , @xmath1 supergravity in @xmath4 has a @xmath41 anomaly @xcite that would not occur in theories with higher supersymmetry .
the two - loop @xmath11 case also has some special features : at two loops the bcj kinematic numerators of maximally supersymmetric yang - mills four - point amplitudes are independent of loop momenta , implying that half - maximal supergravity amplitudes are simple linear combinations of the corresponding pure yang - mills ones . to go beyond these special cases , here
we study the case of @xmath0 supergravity in @xmath4 to show that there is no divergence associated with the counterterm listed on the third line of .
this case is not entangled with any known anomaly .
furthermore , unlike the two - loop case , the kinematic numerators do depend on loop momenta , so the gravity integrals are different from the corresponding yang - mills ones . in order to describe the phenomenon of enhanced cancellations
, we turn to power counting using maximal cuts .
the terms selected by a maximal cut are a gauge - invariant set that are unique to a diagram . using maximal cuts
, we can incorporate all supersymmetric cancellations into supergravity power counts using the known power counts of super - yang - mills theories . because all supersymmetric cancellations are accounted for ,
this gives us a power - counting method equivalent to the one of bjrnsson and green @xcite .
the maximal cut of a given diagram is obtained by replacing all propagators with on - shell conditions .
while the cut conditions set various terms to zero , they do allow us to identify terms with poor behavior , in some cases worse behavior than that of the full amplitude .
once we have selected terms using the maximal cuts , we promote them back to feynman integrals , making sure that the obtained representation has the _ minimum power count _ consistent with the cut .
if any term is then found whose ultraviolet behavior is worse than that of the amplitude as a whole , then by definition , we have enhanced cancellations .
of course , some care is required to be sure that we are using a form that has minimum power count but is also consistent with the cut . to be clear
, we are defining enhanced cancellations entirely by their integrand power - counting properties and not by cancellations that appear only after integration .
we do the power counting in @xmath21 dimensions , viewing @xmath21 as arbitrarily large , to not include hidden relations in the cut solutions that might lead to extra cancellations . in this way we maintain @xmath21-dimensional covariance .
we will show that the maximal cuts give power counts equivalent to the potential counterterms in .
this should not be too surprising given that the maximal cuts are a gauge - invariant subset built from objects that respect all standard symmetries of the amplitudes .
as with other power counts , the maximal cuts do not make enhanced cancellations visible because they do not account for nontrivial cancellations between diagrams .
indeed , at a sufficiently high loop order , the amplitudes of _ every _ supergravity theory necessarily have divergences in individual terms selected by the maximal cuts . to see a better behavior in unitarity cuts , one needs to instead look at cuts that collect together many diagrams so as to allow cancellations between them .
the cuts analyzed in ref .
@xcite suggesting improved all - loop behavior of the amplitudes are examples of this .
.5 cm as a warm - up , we first consider maximal cuts in @xmath5 supergravity .
we consider the diagram in ( a ) as a simple first example . in @xmath5 supergravity
, a kinematic numerator consistent with the maximal cuts is given by @xcite @xmath42 where @xmath43 is the four - point gravity tree amplitude , and @xmath44 and @xmath45 are the standard four - point mandelstam invariants .
the maximal - cut conditions have no effect on this numerator since it is independent of loop momentum .
counting the three @xmath21-dimensional loop integrals , no powers of loop momentum in the numerator , and ten propagators gives us the power count , @xmath46 where @xmath47 is an ultraviolet cutoff .
the critical dimension where an ultraviolet divergence first occurs is thus @xmath48 for the maximal - cut terms of this diagram .
however , the @xmath5 three - loop amplitude also contains worse - behaved terms .
we consider instead the diagram in ( b ) . in @xmath5 supergravity
, a kinematic numerator consistent with the unitarity cuts of this diagram is given in ref .
@xcite : @xmath49 where the momenta correspond to the labels in the diagram .
applying the maximal - cut conditions , we set @xmath50 and obtain the minimal power - counting form , @xmath51 after promoting this back to the numerator of a full three - loop feynman integral , we count the powers of loop momenta .
counting the three @xmath21-dimensional loop integrals , two powers of loop momentum in the numerator , and ten propagators gives us an overall power count for the diagram of @xmath52 the critical dimension of this contribution is thus @xmath53 , which matches the critical dimension obtained from explicit divergence calculations @xcite . in this case then , there are no enhanced cancellations .
.5 cm -.2 cm next we consider the maximal cut of the five - loop diagram in ( a ) .
the simplest numerator consistent with the diagram s maximal cut in @xmath5 supergravity is @xmath54 where the momenta follow the labels of ( a ) .
the numerator follows from the rung rule @xcite a rule devised to give the correct iterated two - particle cuts after dropping terms that vanish with the on - shell conditions @xmath55 .
promoting the maximal - cut terms back to numerators of a feynman integral , we have five @xmath21-dimensional loop integrals and sixteen propagators . together with the numerator ( [ fiveloopnumerator ] )
, we then have a power count , @xmath56 this gives a critical dimension @xmath57 , matching the bjrnsson and green analysis @xcite .
after stepping through the diagrams , this turns out to be the worst - behaved contribution . if it were to turn out that the critical dimension of the full amplitude is greater than @xmath58 , then by definition there would be enhanced ultraviolet cancellations .
similarly , we go through the same exercise for the seven - loop diagram shown in ( b ) . in this case ,
the simplest form of the numerator consistent with the maximal cuts is @xmath59 together with seven @xmath21-dimensional loop integrations and 22 propagators , we obtain the power count , @xmath60 thus , the critical dimension is @xmath61 , again in agreement with other power - counting methods @xcite . while it is not yet technically feasible to directly study the enhanced cancellations in @xmath5 supergravity five- and seven - loop amplitudes , we are able to study them in @xmath1 and @xmath0 supergravities .
we therefore turn to power counting in these theories .
consider @xmath1 supergravity at three loops .
as explained in ref .
@xcite , the bcj construction of the integrand is in terms of the 12 diagrams displayed in . to be concrete ,
we examine diagram ( a ) in for @xmath1 supergravity .
the maximal - cut conditions on the kinematic invariants are @xmath62 applying these and taking the explicit expression for the numerator of ( a ) in @xmath1 supergravity obtained by the double - copy procedure , we obtain @xmath63 where we kept only those term with the largest powers of loop momenta .
the momentum labels are the ones shown in the figure and @xmath64 is an @xmath1 super - yang - mills tree amplitude depending only on the external states and momenta .
the @xmath65 are polarization vectors of gluons .
as discussed in , the pure @xmath1 supergravity states are just the direct product of states of the two gauge theories .
the displayed term in is irreducible in that its power count can not be lowered by imposing the maximal - cut conditions ( [ maxcuts ] ) . since the term ( [ diagramanumerator ] ) is uniquely assigned to the diagram , it is a lower bound on the power count of the diagram . after including the three @xmath21-dimensional loop integrals , eight powers of numerator loop momentum and ten propagators
, we obtain a power - counting for this diagram , @xmath66 thus , in @xmath4 this diagram has divergent terms . as a direct confirmation of this power count
, we integrated the irreducible numerator in after putting back the propagators .
indeed , it is ultraviolet divergent as indicated from the power count .
this power count agrees with the one based on standard - symmetry arguments @xcite . on the other hand ,
explicit calculations show that the three - loop four - point @xmath1 supergravity amplitude is finite @xcite .
given that there are divergent terms in ( a ) that can only cancel against terms that were set to zero by the maximal cut conditions or terms from other diagrams , the finiteness of the amplitude as a whole is a prime example of an enhanced cancellation .
the maximal - cut constraints can sometimes lower the power count of diagrams below their true critical dimension .
for example , for the diagram in ( b ) , under the maximal - cut conditions , all @xmath67 can be made to be no worse than linear in loop momenta . by choosing the minimal resulting power count , this results in an integrand that is ultraviolet finite , even after including an extra power of loop momentum from the @xmath1 super - yang - mills side of the double copy .
another point is that after integration , it may be possible to combine terms even from a single diagram to get a finite result . in particular
, one can imagine taking the numerator of and combining it with a judiciously chosen set of terms that vanish on the cuts to cancel the ultraviolet divergences .
however , this is not relevant for enhanced cancellations which are defined in terms of power counting individual terms at the integrand level .
if even a single term in the integrand of a single diagram has a worse power count compared to the actual behavior of the full amplitude and the power count can not be lowered by maximal - cut conditions , then we have identified enhanced cancellations .
we also note that , by using spinor - helicity , we can set the integrated divergence resulting from the diagram in ( a ) to zero ( i.e. , the integration results in terms containing @xmath68 that can be set to zero by special reference momentum choices ) .
indeed , we shall do so later to simplify various tables . of course
, this does not change the fact that there is a term in the integrand that has a worse power count than the amplitude as a whole that can not be set to zero .
the counting for @xmath0 supersymmetry is similar except that one should subtract two powers of loop momenta from each numerator because of additional @xmath31 supersymmetry cancellations described in ref .
@xcite . taking this into account , for the diagram ( a )
, we obtain a maximal - cut power count in @xmath0 supergravity of @xmath69 thus , the critical dimension from the maximal cut of this diagram is @xmath70 , so we expect there to be no obstruction to finding a covariant representation of @xmath0 supergravity that is manifestly ultraviolet finite in @xmath4 at this loop order .
nevertheless , in we will show that on top of the supersymmetric cancellations , there are additional enhanced cancellations beyond those needed for ultraviolet finiteness .
supergravity should diverge in four dimensions , contrary to the behavior of the four - point amplitude as a whole . ]
if we repeat the same exercise at four loops for @xmath1 supergravity using similar power counting on , for example , the diagram in , we have the behavior , @xmath71 here we count four @xmath21-dimensional loop integrals , 13 propagators , 10 powers of numerator loop momenta from the yang - mills vertices and 2 additional powers of loop momenta from the @xmath1 super - yang - mills numerator .
this means that terms in the diagram in have a critical dimension of @xmath72 and that in @xmath73 it is quadratically divergent by power counting .
if we increase the supersymmetry to @xmath0 supergravity , as noted above , the extra @xmath31 supersymmetry decreases the maximal - cut power count by two powers of loop momentum so that @xmath74 which corresponds to a critical dimension of @xmath61 . therefore , based on the maximal - cut power counting , we would expect @xmath0 supergravity to be logarithmically divergent at four loops .
this is consistent with the standard - symmetry power count of ref .
@xcite , leading to an expected counterterm on the third line of . in we
show that because of enhanced cancellations , the @xmath0 four - loop four - point amplitude is , in fact , ultraviolet finite , contrary to these power counts .
+ + as a warm up to our four - loop calculation , we first present the corresponding three - loop calculation in @xmath0 supergravity .
we follow the same techniques summarized in sect .
[ methodssection ] and described in some detail in ref .
in contrast to @xmath1 supergravity , in this case we should be able to construct a covariant integrand that is manifestly ultraviolet finite , bypassing the need for loop integration to demonstrate that it is ultraviolet finite .
however , we do not do so here . instead , we proceed the same way as at four loops by first computing the @xmath1 supergravity divergences and then adding in the extra contributions needed in @xmath0 supergravity .
this allows us to observe enhanced cancellations .
in fact , we are able to show finiteness with the enhanced cancellations alone , even without accounting for cancellations arising from the extra supersymmetry in the @xmath0 theory compared to the @xmath1 theory .
thus , the cancellations are stronger than those required to demonstrate finiteness . in the calculation
, we leave two state - counting parameters to make it simple to switch between various supergravity theories .
the first parameter is @xmath75 , which is obtained from contractions of the metric @xmath76 from the lorentz algebra , while the second is @xmath77 , which counts the number of majorana fermions added to the pure yang - mills side of the double copy . by choosing @xmath78 and @xmath79
, we obtain pure @xmath1 supergravity . by setting the parameters to @xmath78 and @xmath80
, we obtain @xmath0 supergravity .
we can also obtain results for @xmath1 supergravity with @xmath81 matter multiplets by choosing @xmath82 and @xmath83 , where @xmath81 is the number of internal matter vector multiplets @xcite .
our @xmath4 divergence - calculation results are summarized in with the results corresponding to each graph in fig .
[ threeloopfigure ] used to organize these calculations .
the 12 diagrams correspond to the nonvanishing ones of @xmath1 super - yang - mills theory in the bcj representation @xcite .
contains all contributions that do not depend on the parameter @xmath84 , and contains all the pieces that do depend on @xmath77 . in the calculation
, we take the external states to be gluons on the @xmath35 or @xmath31 side of the double copy , keeping the polarizations vectors formal . however , to simplify the tables we apply four - dimensional spinor - helicity ( see ref .
@xcite for a recent review ) on the polarization vectors and specify the external - gluon states to be @xmath85 .
we also make convenient choices of reference momenta : @xmath86 and @xmath87 .
this choice makes the divergences in diagrams ( a)(d ) vanish for both tables .
it also results in terms containing a factor of @xmath88 , from two fermion loops , to vanish individually in all diagrams in table [ fermunsubtable ] ( instead of in the sum over diagrams ) .
we have not included subdivergence subtractions in the tables , but we have explicitly confirmed that , with the use of the uniform mass regulator , the subdivergences cancel as expected , given that there are no lower - loop divergences . in the tables ,
the @xmath89 are the standard riemann zeta constants .
the transcendental constant @xmath90 appearing in the tables is @xmath91 where @xmath92 is the clausen function .
as the tables illustrate , when one sums over all diagrams , the result is finite for any choice of the state - counting parameters .
although we made special helicity choices for the tables , our calculation is based on using formal polarization states and is therefore valid for any external state that is a direct product of a gluon state and an @xmath1 super - yang - mills state .
this corresponds to a subset of the @xmath0 supergravity states .
nevertheless , the result also extends to _ any _ @xmath0 state because the @xmath31 super - yang - mills supersymmetry identities @xcite are powerful enough to relate all four - point amplitudes to the gluonic ones .
( a discussion of these identities at two loops is given in ref .
it is interesting that for _ any _ values of the state - counting parameters , the divergences vanish .
summarizing , not only is @xmath1 supergravity ultraviolet finite at three loops , but the extra pieces needed to obtain @xmath0 supergravity using the decomposition ( [ neqonesusydecomp ] ) are finite by themselves : @xmath93 the independent vanishings in show that the extra supersymmetry of the @xmath0 theory compared to @xmath1 theory is not needed to make @xmath0 supergravity finite .
while this is no surprise given the three - loop finiteness of @xmath1 supergravity , it does explicitly demonstrate that ultraviolet cancellations exist in subpieces for which there is no power - counting argument .
thus , the @xmath0 case is another explicit example of enhanced ultraviolet cancellations that go beyond the ones that have been understood by any standard - symmetry considerations .
we now consider four loops .
we first summarize the calculation of the @xmath1 supergravity four - point divergence presented in ref .
@xcite , giving a few additional intermediate results .
we then turn to the corresponding calculation in @xmath0 supergravity , showing that the divergence vanishes .
@xmath94{\hspace{-0.1cm}\includegraphics[clip , scale=0.32]{diags / diag1.eps } } \hspace{0.4 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.32]{diags / diag2.eps } } \hspace{0.4 cm } \subfigure[]{\hspace{-0.15cm}\includegraphics[clip , scale=0.32]{diags / diag3.eps } } \hspace{0.65 cm } \subfigure[]{\hspace{-0.25cm}\includegraphics[clip , scale=0.32]{diags / diag4.eps } } \hspace{0.7 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag5.eps } } \hspace{0.8 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag6.eps } } \nonumber\\ & \hskip .1 cm \subfigure[]{\hspace{-0.2cm}\includegraphics[clip , scale=0.32]{diags / diag7.eps } } \hspace{0.8 cm } \subfigure[]{\hspace{-0.25cm}\includegraphics[clip , scale=0.32]{diags / diag8.eps } } \hspace{0.8 cm } \subfigure[]{\hspace{-0.25cm}\includegraphics[clip , scale=0.32]{diags / diag9.eps } } \hspace{0.7 cm } \subfigure[]{\hspace{-0.25cm}\includegraphics[clip , scale=0.32]{diags / diag10.eps } } \hspace{0.6 cm } \subfigure[]{\hspace{-0.25cm}\includegraphics[clip , scale=0.32]{diags / diag11.eps } } \hspace{0.6 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.32]{diags / diag12.eps } } \nonumber\\ & \hspace{0.2 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag13.eps } } \hspace{0.8 cm } \subfigure[]{\hspace{-0.15cm}\includegraphics[clip , scale=0.32]{diags / diag14.eps } } \hspace{0.8 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag15.eps } } \hspace{0.8 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag16.eps } } \hspace{0.8 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag17.eps } } \hspace{0.7 cm } \subfigure[]{\hspace{-0.2cm}\includegraphics[clip , scale=0.32]{diags / diag18.eps } } \nonumber\\ & \hspace{0.1 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.32]{diags / diag19.eps } } \hspace{1.cm } \subfigure[]{\hspace{-0.2cm}\includegraphics[clip , scale=0.32]{diags / diag20.eps } } \hspace{1.1 cm } \subfigure[]{\hspace{-0.15cm}\includegraphics[clip , scale=0.32]{diags / diag21.eps } } \hspace{0.8 cm } \subfigure[]{\hspace{-0.2cm}\includegraphics[clip , scale=0.32]{diags / diag22.eps } } \hspace{0.8 cm } \subfigure[]{\hspace{-0.2cm}\includegraphics[clip , scale=0.32]{diags / diag23.eps } } \hspace{0.7 cm } \subfigure[]{\hspace{-0.2cm}\includegraphics[clip , scale=0.32]{diags / diag24.eps } } \nonumber\\ & \hspace{0.1 cm } \subfigure[]{\hspace{-0.2cm}\includegraphics[clip , scale=0.32]{diags / diag25.eps } } \hspace{0.8 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.32]{diags / diag26.eps } } \hspace{0.8 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.32]{diags / diag27.eps } } \hspace{0.6 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.32]{diags / diag28.eps } } \hspace{0.7 cm } \subfigure[]{\hspace{-0.2cm}\includegraphics[clip , scale=0.32]{diags / diag29.eps } } \hspace{0.6 cm } \subfigure[]{\hspace{-0.2cm}\includegraphics[clip , scale=0.32]{diags / diag30.eps } } \nonumber\\ & \hspace{0.1 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.32]{diags / diag31.eps } } \hspace{0.9 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.32]{diags / diag32.eps } } \hspace{1.cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag33.eps } } \hspace{1.0 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.32]{diags / diag34.eps } } \hspace{0.8 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag35.eps } } \hspace{0.8 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag36.eps } } \nonumber \\ & \hspace{0.1 cm } \subfigure[]{\hspace{-0.2cm}\includegraphics[clip , scale=0.32]{diags / diag37.eps } } \hspace{0.8 cm } \subfigure[]{\hspace{-0.2cm}\includegraphics[clip , scale=0.32]{diags / diag38.eps } } \hspace{0.8 cm } \subfigure[]{\hspace{-0.2cm}\includegraphics[clip , scale=0.31]{diags / diag39.eps } } \hspace{0.7 cm } \subfigure[]{\hspace{-0.2cm}\includegraphics[clip , scale=0.30]{diags / diag40.eps } } \hspace{0.9 cm } \subfigure[]{\hspace{-0.2cm}\includegraphics[clip , scale=0.30]{diags / diag41.eps } } \hspace{0.7 cm } \subfigure[]{\hspace{-0.2cm}\includegraphics[clip , scale=0.30]{diags / diag42.eps } } \nonumber\end{aligned}\ ] ] @xmath95{\hspace{-0.1cm}\includegraphics[clip , scale=0.30]{diags / diag43.eps } } \hspace{0.8cm}\vspace{10 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.30]{diags / diag44.eps } } \hspace{0.8 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.31]{diags / diag45.eps } } \hspace{0.7 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.30]{diags / diag46.eps } } \hspace{0.8 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.30]{diags / diag47.eps } } \hspace{0.7 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.30]{diags / diag48.eps } } \nonumber\\ & \hspace{0.2 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.30]{diags / diag49.eps } } \hspace{0.7 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.30]{diags / diag50.eps } } \hspace{0.6 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.33]{diags / diag51.eps } } \hspace{0.6 cm } \subfigure[]{\hspace{-0.2cm}\includegraphics[clip , scale=0.32]{diags / diag52.eps } } \hspace{0.7 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag53.eps } } \hspace{0.5 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag54.eps } } \nonumber\\ & \hspace{0.15 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag55.eps } } \hspace{0.55 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag56.eps } } \hspace{0.7 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag57.eps } } \hspace{0.6 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag58.eps } } \hspace{0.6 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag59.eps } } \hspace{0.6 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag60.eps } } \nonumber\\ & \hspace{0.15 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag61.eps } } \hspace{0.5 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag62.eps } } \hspace{0.7 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag63.eps } } \hspace{0.6 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag64.eps } } \hspace{0.6 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag65.eps } } \hspace{0.55 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag66.eps } } \nonumber\\ & \hspace{0.2 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag67.eps } } \hspace{0.55 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag68.eps } } \hspace{0.6 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag69.eps } } \hspace{0.6 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag70.eps } } \hspace{0.7 cm } \subfigure[]{\hspace{-0.3cm}\includegraphics[clip , scale=0.32]{diags / diag71.eps } } \hspace{0.7 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.32]{diags / diag72.eps } } \nonumber\\ & \hspace{0.2 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.32]{diags / diag73.eps } } \hspace{0.6 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.32]{diags / diag74.eps } } \hspace{0.5 cm } \subfigure[]{\hspace{-0.2cm}\includegraphics[clip , scale=0.32]{diags / diag75.eps } } \hspace{0.5 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.32]{diags / diag76.eps } } \hspace{0.6 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.32]{diags / diag77.eps } } \hspace{0.6 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.32]{diags / diag78.eps } } \nonumber\\ & \hspace{1.7 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.32]{diags / diag79.eps } } \hspace{0.4 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.32]{diags / diag80.eps } } \hspace{0.5 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.32]{diags / diag81.eps } } \hspace{0.4 cm } \subfigure[]{\hspace{-0.1cm}\includegraphics[clip , scale=0.32]{diags / diag82.eps } } \nonumber\end{aligned}\ ] ] the calculation of the @xmath1 supergravity divergence starts from pure yang - mills feynman diagrams , keeping only those diagrams with color factors that match the 82 @xmath1 super - yang - mills diagrams displayed in .
the color factors are then replaced the bcj forms of the @xmath1 super - yang - mills numerators given in ref .
there are three additional diagrams , displayed in . in pure yang - mills they contain ultraviolet divergences ( canceled by infrared divergences ) , but in @xmath96 supergravity an extra power of zero in the form of an on - shell massless external momentum squared in the @xmath1 super - yang - mills numerators sets such potential ultraviolet contributions to zero . after feeding the integrand so constructed through the integration procedure summarized in
, we find that pure @xmath1 supergravity in @xmath4 is divergent @xcite : @xmath97 where @xmath98 is the dimensional - regularization parameter , and @xmath99 with @xmath100 the sum runs over all 24 permutations of the external legs .
@xmath101 is the linearized field - strength tensor given in terms of polarization vectors for leg @xmath17 as @xmath102 this form makes explicit the fact that each state of pure @xmath1 supergravity corresponds to a direct product of a color - stripped state of @xmath1 super - yang - mills theory and of pure nonsupersymmetric yang - mills theory . by taking linear combinations , the divergences can be separated into distinct helicity classifications : @xmath103 each of the obtained operators are nonvanishing only for the indicated helicity configurations and their parity conjugates and relabelings .
the helicity labels refer to those of the polarization vectors used in on the pure yang - mills side and not the supergravity states , which are obtained by a direct product of these states with those of @xmath1 super - yang - mills theory . for explicit helicity states in @xmath4 , we have @xmath104 using four - dimensional spinor - helicity notation . as explained in ref .
@xcite , the appearance of the divergences in the three independent helicity configurations in is unexpected and points to the source of the divergence being the marcus @xmath41 duality - symmetry anomaly @xcite . without the anomaly , the @xmath105 and @xmath106 helicity sectors would vanish .
@xcite explains how the anomaly leads to poor ultraviolet behavior even in the @xmath85 sector . in
we have collected together groups of diagrams in order to display the nontrivial cancellations between diagrams .
the first three entries correspond to the sums over diagrams 130 , 3160 and 6182 , while the final one gives the sum over all diagrams .
the final sum displays an enormous cancellation between the diagrams to yield a remarkably simple result
. we do not include subdivergences which automatically cancel amongst themselves .
in the table , the @xmath89 are the standard riemann zeta constants .
the value of @xmath107 is already defined in , while the other constants are @xcite @xmath108 as noted earlier , @xmath92 is the clausen function .
these transcendental constants arise from our use of an infrared mass regulator and , as expected , cancel from the final ultraviolet divergence @xcite .
next we turn to @xmath0 supergravity in @xmath4 .
as discussed in , we obtain @xmath0 supergravity from @xmath1 supergravity by adding in the contributions from the bcj construction based on the direct product of @xmath1 super - yang - mills with additional contributions from adding a single majorana fermion to the pure gluon theory .
the calculation is somewhat more complicated than the pure @xmath1 supergravity calculation because of the long fermion traces that appear at four loops . because @xmath0 supergravity has no duality - symmetry anomaly
, we expect it to be ultraviolet finite at four loops . indeed
, additional @xmath31 supersymmetry identities are sufficient to show amplitudes and any associated potential ultraviolet divergences vanish in the @xmath105 and @xmath106 helicity sectors in .
only the @xmath85 sector gives nonvanishing amplitudes and therefore needs checking , although we have calculated the other two sectors as well .
we have only computed the case where the external @xmath0 supergravity states are those obtained from a direct product of @xmath1 super - yang - mills and pure yang - mills states (
i.e. the subset of states that are also in the pure @xmath1 supergravity spectrum ) .
however , as mentioned in , @xmath31 and @xmath1 supersymmetry identities in the direct product @xcite allow us to express any of the four - point amplitudes in terms of one of them , so ruling out divergences in this sector rules out _ all _ four - point divergences .
we find that the four - loop four - point amplitudes of @xmath0 supergravity are finite .
this may be unsurprising given that the additional supersymmetry compared to the @xmath1 supergravity case should improve the ultraviolet properties .
however , the fact remains that , at present , there is no standard - symmetry explanation for the finiteness .
in addition , as we explain in , no covariant diagrammatic formalism can display the cancellations manifestly , so the vanishing of the divergence is another example of enhanced cancellations . in we give the extra contributions to the potential divergence coming from the additional states that are present in @xmath0 supergravity compared to @xmath1 supergravity . as can be seen from the final entry in the table ,
the contribution is equal and opposite to the contribution that comes solely from the @xmath1 supergravity states given in . therefore the total divergence vanishes : @xmath109 the nontrivial way the cancellations occur in the sum of the entries in suggests that there should be a better way to see them . while it may be simple to state the obvious , as already explained in , finding a formalism that makes these cancellations manifest is nontrivial , given that no covariant diagrammatic representation exists that does so
besides the information given in , in accompanying mathematica attachments @xcite we give the divergences for each diagram for pure @xmath1 supergravity as well as for the additional contributions needed for @xmath0 supergravity . as in the tables
, we do not include subdivergences in these files since they automatically cancel amongst themselves in our calculation .
we note that while our calculation proves that there are no four - loop four - point divergences in @xmath0 supergravity , it does not rule out five - point @xmath110-type divergences .
it would of course be interesting to study these as well in the future .
in this paper , we described the phenomenon of _ enhanced ultraviolet cancellations _ in supergravity theories . by definition , when all local covariant diagrammatic representations of an amplitude contain terms that have a worse power count than the amplitude as a whole , we have enhanced cancellations . to illustrate this phenomenon , we first discussed @xmath1 supergravity in four dimensions at three loops . by power counting maximal cuts , we identified terms in the four - point amplitude that are divergent at three loops , in agreement with supersymmetry and duality - symmetry arguments @xcite , when , in fact , the amplitude is three - loop finite in @xmath4 @xcite .
the theory does diverge at four loops @xcite , but it appears to be due to a rigid @xmath41 duality - symmetry anomaly @xcite .
such anomalies are not present in @xmath111 supergravity theories , suggesting that these theories can not have similar divergences .
this motivated us to study the four - loop four - point amplitudes of @xmath0 supergravity . again in this case , power counting maximal cuts identifies divergent terms in @xmath4 , consistent with standard - symmetry considerations @xcite .
however , explicit calculations performed in this paper shows this amplitude is ultraviolet finite , again illustrating enhanced ultraviolet cancellations .
if similar enhanced ultraviolet cancellations hold in @xmath5 supergravity , then this theory will be finite at seven loops as well , contradicting predictions based on power counting @xcite .
the underlying reason for enhanced ultraviolet cancellations is not fully understood .
there are some indications that the duality between color and kinematics @xcite is responsible .
an explicit study shows that this duality is responsible for improved ultraviolet behavior in the relatively simple two - loop case of half - maximal supergravity in @xmath11 @xcite .
an important challenge is to push this understanding to higher loop orders .
another important question is whether there might be an explanation for enhanced cancellations based on supersymmetry or duality symmetry .
such an explanation would have to be novel , given that enhanced ultraviolet cancellations are nonstandard .
the potential three - loop @xmath40 counterterm of @xmath1 supergravity and four - loop counterterm @xmath112 counterterm of @xmath0 supergravity can not be written as full superspace integrals @xcite .
an interesting open question is whether this plays any role in the vanishing of the associated divergences . while we do not yet know if perturbatively ultraviolet - finite unitary field theories of gravity exist , based on the results of this paper it is clearly premature to conclude otherwise .
more generally , nontrivial multiloop enhanced cancellations in gravity theories are a new and surprising phenomenon , contrary to expectations based on standard - symmetry considerations which suggest viable counterterms .
the existence of these cancellations gives us confidence that further nontrivial surprises await us as we probe supergravity theories to ever higher loop orders using modern tools .
we thank guillaume bossard , john joseph carrasco , paolo di vecchia , sergio ferrara , enrico herrmann , ian jack , henrik johansson , tim jones , sean litsey , josh nohle , radu roiban , james stankowicz , kelly stelle and jaroslav trnka for helpful discussions .
we especially thank alexander and volodya smirnov for their assistance with the integrals .
this material is based upon work supported by the department of energy under award number de - sc0009937 .
we also thank the danish council for independent research . we gratefully acknowledge mani bhaumik for his generous support .
we also thank academic technology services at ucla for computer support .
z. bern , j. j. m. carrasco , h. johansson and d. a. kosower , phys .
d * 76 * , 125020 ( 2007 ) [ arxiv:0705.1864 [ hep - th ] ] .
z. bern , s. davies , t. dennen and y .- t .
huang , phys .
lett . * 108 * , 201301 ( 2012 ) [ arxiv:1202.3423 [ hep - th ] ] .
z. bern , s. davies , t. dennen , a. v. smirnov and v. a. smirnov , phys . rev .
lett . * 111 * , no .
23 , 231302 ( 2013 ) [ arxiv:1309.2498 [ hep - th ] ] .
g. t hooft and m. j. g. veltman , annales poincare phys .
a * 20 * , 69 ( 1974 ) .
s. deser and p. van nieuwenhuizen , phys .
d * 10 * , 401 ( 1974 ) ; + s. deser , h. s. tsao and p. van nieuwenhuizen , phys . rev .
d * 10 * , 3337 ( 1974 ) ; + m. fischler , phys . rev .
d * 20 * , 396 ( 1979 ) .
m. h. goroff and a. sagnotti , nucl .
b * 266 * , 709 ( 1986 ) .
m. t. grisaru , phys .
b * 66 * , 75 ( 1977 ) ; + e. tomboulis , phys .
b * 67 * , 417 ( 1977 ) .
p. s. howe and k. s. stelle , int .
j. mod .
a * 4 * , 1871 ( 1989 ) .
m. t. grisaru and w. siegel , nucl .
b * 201 * , 292 ( 1982 ) [ erratum - ibid .
b * 206 * , 496 ( 1982 ) ] .
z. bern , l. j. dixon , d. c. dunbar and d. a. kosower , nucl .
b * 425 * , 217 ( 1994 ) [ hep - ph/9403226 ] ; nucl .
b * 435 * , 59 ( 1995 ) [ hep - ph/9409265 ] .
z. bern , l. j. dixon , d. c. dunbar , m. perelstein and j. s. rozowsky , nucl .
b * 530 * , 401 ( 1998 ) [ hep - th/9802162 ] .
z. bern , j. j. m. carrasco and h. johansson , phys .
d * 78 * , 085011 ( 2008 ) [ arxiv:0805.3993 [ hep - ph ] ] .
z. bern , j. j. m. carrasco and h. johansson , phys .
* 105 * , 061602 ( 2010 ) [ arxiv:1004.0476 [ hep - th ] ] .
e. cremmer and b. julia , nucl .
b * 159 * , 141 ( 1979 ) .
z. bern , j. j. carrasco , l. j. dixon , h. johansson , d. a. kosower and r. roiban , phys .
lett . * 98 * , 161303 ( 2007 ) [ hep - th/0702112 ] .
z. bern , j. j. m. carrasco , l. j. dixon , h. johansson and r. roiban , phys .
d * 78 * , 105019 ( 2008 ) [ arxiv:0808.4112 [ hep - th ] ] .
z. bern , j. j. carrasco , l. j. dixon , h. johansson and r. roiban , phys .
lett . *
103 * , 081301 ( 2009 ) [ arxiv:0905.2326 [ hep - th ] ] .
m. b. green , j. g. russo and p. vanhove , jhep * 1006 * , 075 ( 2010 ) [ arxiv:1002.3805 [ hep - th ] ] ; + g. bossard , p. s. howe and k. s. stelle , jhep * 1101 * , 020 ( 2011 ) [ arxiv:1009.0743 [ hep - th ] ] ; + n. beisert , h. elvang , d. z. freedman , m. kiermaier , a. morales and s. stieberger , phys . lett .
b * 694 * , 265 ( 2010 ) [ arxiv:1009.1643 [ hep - th ] ] .
g. bossard , p. s. howe , k. s. stelle and p. vanhove , class .
grav . * 28 * , 215005 ( 2011 ) [ arxiv:1105.6087 [ hep - th ] ] .
j. bjrnsson and m. b. green , jhep * 1008 * , 132 ( 2010 ) [ arxiv:1004.2692 [ hep - th ] ] ; + j. bjrnsson , jhep * 1101 * , 002 ( 2011 ) [ arxiv:1009.5906 [ hep - th ] ] .
n. berkovits , jhep * 0004 * , 018 ( 2000 ) [ hep - th/0001035 ] .
e. cremmer , j. scherk and s. ferrara , phys .
b * 74 * , 61 ( 1978 ) .
p. tourkine and p. vanhove , class .
* 29 * , 115006 ( 2012 ) [ arxiv:1202.3692 [ hep - th ] ] .
z. bern , s. davies , t. dennen and y .- t .
huang , phys .
d * 86 * , 105014 ( 2012 ) [ arxiv:1209.2472 [ hep - th ] ] . g. bossard , p. s. howe and k. s. stelle , phys .
b * 719 * , 424 ( 2013 ) [ arxiv:1212.0841 [ hep - th ] ] .
g. bossard , p. s. howe and k. s. stelle , jhep * 1307 * , 117 ( 2013 ) [ arxiv:1304.7753 [ hep - th ] ] .
z. bern , s. davies and t. dennen , arxiv:1305.4876 [ hep - th ] .
n. marcus , phys .
b * 157 * , 383 ( 1985 ) .
j. j. m. carrasco , r. kallosh , r. roiban and a. a. tseytlin , arxiv:1303.6219 [ hep - th ] .
z. bern , l. j. dixon and r. roiban , phys .
b * 644 * , 265 ( 2007 ) [ hep - th/0611086 ] .
z. bern , j. s. rozowsky and b. yan , phys .
b * 401 * , 273 ( 1997 ) [ hep - ph/9702424 ] .
a. a. vladimirov , theor .
phys . *
43 * , 417 ( 1980 ) [ teor . mat . fiz . *
43 * , 210 ( 1980 ) ] ; + n. marcus and a. sagnotti , nuovo cim . a * 87 * , 1 ( 1985 ) .
k. g. chetyrkin and f. v. tkachov , nucl .
b * 192 * , 159 ( 1981 ) .
a. v. smirnov , jhep * 0810 * , 107 ( 2008 ) [ arxiv:0807.3243 [ hep - ph ] ] ; + a. v. smirnov and v. a. smirnov , comput .
commun .
* 184 * , 2820 ( 2013 ) [ arxiv:1302.5885 [ hep - ph ] ] ; + a. v. smirnov , arxiv:1408.2372 [ hep - ph ] .
s. laporta , phys .
b * 549 * , 115 ( 2002 ) [ hep - ph/0210336 ] .
t. van ritbergen , j. a. m. vermaseren and s. a. larin , phys .
b * 400 * , 379 ( 1997 ) [ hep - ph/9701390 ] .
m. czakon , nucl .
b * 710 * , 485 ( 2005 ) [ hep - ph/0411261 ] .
j. j. m. carrasco and h. johansson , j. phys .
a * 44 * , 454004 ( 2011 ) [ arxiv:1103.3298 [ hep - th ] ] . z. bern , j. j. m. carrasco , l. j. dixon , h. johansson and r. roiban , phys . rev .
d * 85 * , 105014 ( 2012 ) ) [ arxiv:1201.5366 [ hep - th ] ] .
z. bern , t. dennen , y .- t . huang and m. kiermaier ,
d * 82 * , 065003 ( 2010 ) [ arxiv:1004.0693 [ hep - th ] ] .
h. kawai , d. c. lewellen and s. h. h. tye , nucl .
b * 269 * , 1 ( 1986 ) .
z. bern , c. boucher - veronneau and h. johansson , phys .
d * 84 * , 105035 ( 2011 ) [ arxiv:1107.1935 [ hep - th ] ] ; + c. boucher - veronneau and l. j. dixon , jhep * 1112 * , 046 ( 2011 ) [ arxiv:1110.1132 [ hep - th ] ] .
w. siegel , phys .
b * 84 * , 193 ( 1979 ) ; + i. jack , d. r. t. jones and k. l. roberts , z. phys .
c * 62 * , 161 ( 1994 ) [ hep - ph/9310301 ] .
m. misiak and m. munz , phys .
b * 344 * , 308 ( 1995 ) [ hep - ph/9409454 ] .
k. g. chetyrkin , m. misiak and m. munz , nucl .
b * 518 * , 473 ( 1998 ) [ hep - ph/9711266 ] .
s. laporta , int .
j. mod .
5087 ( 2000 ) [ hep - ph/0102033 ] . m. b. green , j. h. schwarz and l. brink , nucl .
b * 198 * , 474 ( 1982 ) .
h. elvang and y .- t .
huang , arxiv:1308.1697 [ hep - th ] .
m. t. grisaru , h. n. pendleton and p. van nieuwenhuizen , phys .
d * 15 * , 996 ( 1977 ) ; + m. t. grisaru and h. n. pendleton , nucl .
b * 124 * , 81 ( 1977 ) ; + s. j. parke and t. r. taylor , phys .
b * 157 * , 81 ( 1985 ) [ erratum - ibid .
* 174b * , 465 ( 1986 ) ] .
z. bern , a. de freitas and l. j. dixon , jhep * 0306 * , 028 ( 2003 ) [ erratum - ibid . * 1404 * , 112 ( 2014 ) ] [ hep - ph/0304168 ] ; + a. de freitas and z. bern , jhep * 0409 * , 039 ( 2004 ) [ hep - ph/0409007 , hep - ph/0409007 ] .
see the ancillary file of the arxiv version of this manuscript . | .5 cm we show that the four - loop four - point amplitudes of @xmath0 supergravity are ultraviolet finite in four dimensions , contrary to expectations based on supersymmetry and duality - symmetry arguments .
we explain why the diagrams of any covariant local formalism can not manifestly exhibit the necessary cancellations for finiteness but instead require a new type of ultraviolet cancellation that we call an `` enhanced cancellation '' .
we also show that the three - loop four - point amplitudes in @xmath1 and @xmath0 supergravity theories display enhanced cancellations . to construct the loop integrand
, we use the duality between color and kinematics .
we apply standard methods for extracting ultraviolet divergences in conjunction with the fire5 integral reduction program to arrive at the four - loop results .
ucla/14/tep/106 @xmath2 @xmath3 |
deeply inelastic electron - proton scattering at very high energies has shown that the cross section of the virtual photon at high virtuality @xmath11 increases faster with energy than at low virtuality .
small size objects experience a stronger energy dependent cross section than large size objects .
perturbative @xmath2 has been partially successful in explaining this physics . with suitable starting distributions next to leading order dglap - evolution
can reproduce the experimental data .
nnlo analysis @xcite shows a slight improvement .
but at small @xmath9 the dglap- summation of @xmath12 terms and neglect of @xmath13 terms is theoretically unsatisfactory .
therefore there has been a considerable effort @xcite to include the leading @xmath13 contributions .
it has been suggested @xcite that the relevant collective variables at high energies are the gluon phase factors ( wilson lines ) collected by the low @xmath14 partons in the photon on near light like trajectories .
we have followed this promising method in ref .
@xcite relating high energy scattering to the behavior of wilson line correlations in a @xmath0 dimensional hamiltonian near the light cone @xcite . in this ref .
@xcite the dynamics of wilson lines in @xmath2 with two colors ( @xmath15 ) has been considered . in the present paper
we address the realistic case of su(@xmath16 ) with three colors ( @xmath17 ) .
we propose that with increasing energy the effective dynamics reaches a critical point which is characterized by the symmetry properties of @xmath3 .
universality tells us that @xmath3 wilson lines have the same correlation functions as potts spins in the center group @xmath6 in three dimensions .
since the correlation functions of wilson lines influence the structure of the photon wave function at high energies or small @xmath9 , the @xmath6 symmetry determines the strong increase of the @xmath18 cross section at high energies .
the discovery of asymptotic freedom @xcite gives a fixed point of @xmath3 at @xmath19 , which has opened up the possibility of perturbative calculations in @xmath2 . at infinitely high energies the longitudinal momentum transfers are minute
, therefore @xmath2 at high energies reduces to a @xmath0 dimensional theory , when the longitudinal space variables have been integrated out . only transverse dynamics matters .
the resulting effective hamiltonian has a fixed point at @xmath20 .
lattice simulations for @xmath21 @xcite have found an effective coupling at this fixed point which is not small .
these lattice simulations still have to be extended to @xmath3 .
therefore in the following paper we will use mostly symmetry arguments to characterize the physics near this critical point .
this critical point influences the physics whenever a large number of low @xmath9 partons is involved . beyond virtual photon - proton scattering
other testing grounds for critical point dynamics will be rhic and lhc . as framework we use the formalism of a near light cone hamiltonian .
the light cone formulation of @xmath2 has been a useful tool in perturbative calculations and extended beyond .
since the vacuum in the strictly light cone formulation is simple , the hamiltonian must include all the complicated structures of @xmath2 like condensates .
therefore we prefer a theoretical form developed in refs . @xcite where the light cone is reached in a limiting procedure and quantization is always on spacelike surfaces .
we start by choosing the following @xmath22 coordinates which smoothly interpolate between the lorentz and light front coordinates .
the parameter @xmath23 specifies how near to the light cone the coordinate system is : x^t = x^+ & = & 12 \ { ( 1 + ) x^0 + ( 1 - ) x^3 } , + x^- & = & 12 ( x^0-x^3 ) .
[ coor ] for high energy photon proton scattering at small @xmath24 , with @xmath25 as c.m .
energy squared we define two light like vectors using the photon vector @xmath26 and the proton vector @xmath27 . e_1 & = & q- p + e_2 & = & p. for finite energies the vector of the photon @xmath28 which points in the direction of @xmath29 can be calculated as a linear combination of the light like vector @xmath30 with a small amount of @xmath31 admixed e _ & = & q + xp - p + & = & e_1-e_2 .
one sees that the mixing @xmath23 is related to the bjorken variable @xmath9 and vanishes in the limit of infinite energies as @xmath32 .
therefore it is natural to formulate high energy scattering in near light cone coordinates . in vacuum
it is appropriate to use the hamiltonian with periodic boundary conditions . in this work we only study the effective quark density function of the photon , i.e. a vacuum problem without theoretically modeling the proton target .
the outline of the paper is as follows : in section ii we give the near light cone hamiltonian and its reduction to @xmath33 dimensions for small @xmath9 .
section iii is devoted to the approach of the critical point .
section iv gives the behavior of the correlations of wilson lines near the critical point .
in section v this correlation length is used to model the effective quark density in the photon and calculate the structure function @xmath34 of the proton . a comparison to the hera - data is given in the same section .
section vi contains the conclusions .
for small @xmath9 the eikonal phases acquired by the quarks / antiquarks are the relevant collective variables .
the light cone hamiltonian on the finite light like @xmath35 interval of length @xmath36 has wilson line or polyakov operators similarly to @xmath2 formulated on a finite interval in imaginary time at finite temperature p ( _ ) = 1n_c ( ig _ 0^l dx^-a_-(_,x^- ) ) .
[ polyakov ] the dynamics of these polyakov operators is determined by the near light cone hamiltonian @xmath37 which has been derived in reference @xcite . in the weyl gauge @xmath38
, the gauss - law constraint can be resolved for @xmath39 and one obtains a hamiltonian which depends on the zero mode momentum @xmath40 .
the subsequent modified light front gauge @xmath41 eliminates @xmath35 dependent fields @xmath42 , but allows fields @xmath43 which are functions of the transverse coordinates only .
the final @xmath2 hamiltonian near the light cone has the following form @xcite : @xmath44 with @xmath45 \right]^2 + \frac{1}{\eta^2 } \mbox{tr } \left [ \pi_{\bot}-\left(\partial_-a_{\bot}- ig[a_-,a_{\bot}]\right)\right]^2 \nonumber \\ & + & \frac{1}{\eta^2 } \mbox{tr } \left [ \frac{1}{l } e_{\bot } - \nabla_{\bot } a_-\right]^2 + \frac{1}{2 l^{2 } } p_{-}^ { \ , \dagger}(x_{\bot } ) p_{-}(x_{\bot } ) \nonumber \\ & + & \frac{1}{l^{2}}\int_{0}^{l } dz^{- } \int_{0}^{l } dy^{- } \sum_{p , q , n}\,^{'}\frac { g_{\bot qp}(x_{\bot } , z^{- } ) g_{\bot pq}(x_{\bot},y^{-})}{\left [ \frac{2\pi n}{l } + g(a_{-q}(x_{\bot})- a_{-p}(x_{\bot } ) ) \right]^{2 } } e^{i2\pi n(z^{-}-y^{-})/l } \nonumber\\ & - & i\frac{2}{\eta^2}\psi^{\dagger}_-(\partial_- -ig a_-)\psi_- - i\frac{1}{\eta}\psi^{\dagger}(\alpha_{\bot}-ig a_{\bot})\psi + m\frac{1}{\eta}\psi^{\dagger}\beta\psi , \label{hamil}\end{aligned}\ ] ] the term @xmath46 depends on an external source . e_^c_0=g _ dy^- dy _ + ( f^c_0 a b a_^a(y_,y^- ) _
^b(y_,y^- ) + + ^c_0(y_,y^-)^c_0/2 ) . in the @xmath47 limit ,
those pieces @xmath48 of the hamiltonian @xmath49 dominate which are most singular at @xmath50 and do not couple to the three dimensional gauge fields @xmath51 and @xmath52 .
this reduced hamiltonian has collective variables @xmath53 with the color indices @xmath54 a_-^c_0= _
0^l dx^-a_-^c_0(_,x^- ) which determine the wilson lines and live in a @xmath0 dimensional space .
the wilson line operators @xmath55 can always be parameterized in terms of the diagonal color matrices @xmath56 and @xmath57 by suitable gauge transformations .
we consider the dynamics of the fields @xmath58 in vacuum , i.e. without the source term @xmath46 .
a possible external source , e.g. , a gaussian distributed random color charge @xcite , can be taken into account via this term which shifts the zero mode fields @xmath59 to fluctuate around the classical fields @xmath60 .
we recall that the @xmath22 coordinates correspond to the physics in a fast moving frame and factorize the reduced energy from the lorentz boost factor @xmath61 and a transverse lattice cut off @xmath62 h_red & = & 2 a dx^- dx _ h^ + & = & dx^- dx _ _ c^0=3,8 ( tr(e_^c_0- _ a_-^c_0)^2+p_-^c_0 p_-^c_0 -^_- g a_-^c_0 _ - )
. then we redefine modified zero mode fields and their conjugate momenta on a transverse lattice at positions @xmath63 : ^c_0 ( b_)&= & g l a_-^c_0(b _ ) , + & = & a^2 . their dynamics on the lattice is determined by the lattice hamiltonian @xmath64 ( cf . also ref .
@xcite for the case of @xmath21 ) , h_lat & = & _ b , c_0[-g^2_eff 1j j + & & + _ ( ( ^c_0()-^c_0(+))^2 ) -<^_-_->^c_0 ( ) ^c_0 ] . [ hhat ]
the summation goes over color indices @xmath54 and all @xmath65 -dimensional lattice sites .
the geometry of the @xmath3 manifold enters via the jacobian @xcite : j=^2(^3)^2((^3-^8 ) ) ^2((^3+^8 ) ) .
because of dimensional reduction the effective coupling constant depends on the length @xmath36 of the interval in @xmath66 direction and the @xmath2-coupling @xmath67 multiplied by the parameter @xmath23 characterizing the nearness to the lightcone [ geff ] g^2_eff = .
the phase factors @xmath68 are defined in the fundamental domain ( cf .
1 ) @xmath69 the jacobian vanishes on the boundaries of the fundamental domain .
the coupling to the quarks is introduced by the external field ^c_0=dx^- _ -^c_0 _ -^c_0 .
it was shown in perturbation theory @xcite that the product of `` bad '' light cone components @xmath70 has a zero mode contribution . in structure functions
such a singular piece is generated at bjorken @xmath71 . in @xmath2 on the light cone @xcite the negative energy states obey a constraint equation which allows a constant @xmath35 independent solution .
therefore we assume that the fermion negative energy states develop an expectation value @xmath72 .
further work is still needed to show how such a nonvanishing zero mode density @xmath72 is realized near the light cone . in the continuum limit of the transverse lattice theory
the lattice size @xmath62 goes to zero and/or the extension @xmath36 of the lattice to infinity .
this limit combined with the light cone limit @xmath73 leads to an indefinite behavior of the effective coupling constant ( cf .
( [ geff ] ) ) . the critical behavior of the zero - mode theory resolves this ambiguity . in ref .
@xcite we have done a finite size scaling ( fss ) analysis for @xmath21 @xmath2 obtaining a second order transition as a function of the coupling @xmath74 between a phase with massive excitations at strong coupling and a phase with massless excitations at weak coupling . in the strong coupling domain of @xmath74 the energy of the rotators
@xmath75 is dominated by the electric energy @xmath76 which corresponds to the laplacian in the group manifold .
each site has an energy spectrum with a gap @xmath77 in @xmath21 or @xmath78 in @xmath3 @xcite . with decreasing @xmath74
the magnetic coupling @xmath79 becomes stronger .
a larger nearest neighbor coupling leads to a coherently aligned ground state which has massless excitations .
consequently the mass gap vanishes at a sufficiently small @xmath80 .
the resulting critical @xmath21 theory is in the same universality class as the @xmath81 theory or the ising model in @xmath82-dimensions , which has been checked in the lattice simulations @xcite with the available numerical accuracy . barring any unusual effects arising from the coupling to the three dimensional gluon fields
, we think that the full theory may shift the exact value of the critical coupling , but does not change the critical exponents which are determined by the symmetry of the problem .
the abelian collective fields in @xmath21 have reflection symmetry around @xmath83 which corresponds to the up - down symmetry of the ising spins .
the reduced @xmath3 hamiltonian has rather different symmetry properties than the @xmath21 hamiltonian
. a closer look at the fundamental domain in @xmath3 ( cf .
1 ) shows three regions separated by dashed lines .
an element from the center group @xmath6 can be mapped from one region to another by large gauge transformations . in @xmath21
these large gauge transformations are reflections around @xmath83 on the @xmath84 axis . in previous work @xcite
we found that the universality class of the zero mode @xmath21 theory is the same as @xmath81 theory , therefore we think that the universality class of the reduced hamiltonian in @xmath3 is the three state potts model @xmath6 . in each subregion of the fundamental domain the zero mode variables
@xmath68 are represented by one spin orientation .
the relevant center group @xmath6 has a weak first order transition whose critical line ends in a second order point in the presence of an external field .
we conjecture that this external field is provided by the fermion zero mode density near the light cone . to match the hamiltonian lattice with scattering we consider the same physical picture as in ref .
the lattice constant @xmath62 is chosen to coincide with the photon resolution @xmath85 .
the longitudinal extension @xmath36 must be larger than the color coherence length of the @xmath86 state in the photon - proton c.m .
the photon and proton move on almost light like trajectories and the coherence length grows with @xmath9 as : x^-= .
we demand therefore that @xmath87 . in the limit of small @xmath9 the parameter @xmath23 decreases as @xmath88 . assuming that a stable fixed point in the effective running coupling @xmath89 exists , we get .
since @xmath90 , we can choose a positive constant @xmath91 such that the interval @xmath36 exceeds the color coherence length of the @xmath92 state in the photon , namely [ e17 ] = ( n_0 + ) .
in @xmath6 theory the critical behavior depends on two parameters , the effective field @xmath93 and the coupling @xmath94 : using equation ( [ e17 ] ) one finds that the external field @xmath95 converges in the limit @xmath47 to @xmath96 : h^*= .
the limiting external field @xmath96 is independent of the short distance cutoff @xmath62 under the assumption that the light cone density of the quarks @xmath97 behaves in the same way with @xmath62 as the perturbative quark density in the photon namely @xmath98 .
it is specific to high energy scattering that the external field and the coupling are approaching their critical values in a similar way for @xmath47 h=|h - h^*| h_0 , + = - g^ * 2_eff t_0
. it would be important to find another physical situation where the experimental conditions regulate both quantities independently .
the earliest work on critical behavior of pure @xmath3 gauge theory @xcite has been done in the context of the finite temperature phase transition .
recent studies @xcite have established common features of @xmath3 @xmath2 and the three state potts model @xmath6 with a first order phase transition line which ends in a critical point where the transition is second order . in refs .
@xcite the universality class has been identified as the 3-dimensional ising model in three dimensions which has the critical exponents : @xmath99 in order to assess the divergence of the correlation length , we rescale the lattice constant by a factor @xmath100 and the external field and coupling with powers of @xmath100 according to their respective anomalous dimensions = b f(b^_1,b^_2 h ) . when the external field @xmath101 and @xmath102 is sufficiently close to zero , the critical behavior of the correlation length is calculable by choosing @xmath103 , then one obtains & = & h^-1/_2 f(b^_1,1 ) + & = & h^-1/_2 f_h(h^-_1/_2 ) .
[ e22 ] the function @xmath104 .
note the power @xmath105 governing the low @xmath9 behavior of the correlation length is considerably smaller than the power @xmath106 which determines the power for a small ratio @xmath107 . in the latter case one
gets with the function @xmath108 the behavior & = & ^-1/_1
f(1,h ^-_2/ _ 1 ) + & = & ^-1/_1f _ ( h ^-_2/ _ 1 ) . in the previous work on @xmath21 we considered a second order transition in the absence of an external field . since the universality class was also of the ising type
a critical behavior of the correlation length @xmath109 resulted .
such a power gave a qualitative description of the scarce data on the longitudinal structure function @xcite , but definitely can not describe the f@xmath110- data as shown in ref.@xcite .
in the mathematical framework of the @xmath3 critical theory exposed here , it is clear that the growth of the correlation function with @xmath111 is realized , since for small @xmath9 the argument @xmath112 of the scaling function @xmath113 is small : @xmath114 .
the argument of @xmath115 , however , @xmath116 is large for small @xmath9
. therefore we can well approximate the argument of @xmath113 by @xmath117 for high energy scattering .
recently the scaling function @xmath113 has been calculated in lattice simulations @xcite and found to become constant at @xmath117 .
substituting @xmath118 in eq .
( [ e22 ] ) one finds that the correlation length @xmath119 increases with @xmath1 as ( ) ^- f_h(0 ) . near the critical point
the wilson lines experience long range correlations which means that dipoles in the photon wave function are correlated over large distances . for finite correlation length there
exists an intermediate range where @xmath120 for which the correlation function of wilson lines is power behaved : [ e17p ] < p(x _ ) p(0 ) > where @xmath121 in ising like systems .
this region is responsible for the well known effect of critical opalescence in the gas liquid transition . for larger distances @xmath122
the correlation function decreases exponentially [ e18 ] < p(x _ ) p(0 ) > e ^-x_/.
the photon is ideal to investigate high energy cross sections of hadronic objects with a variable size . in the course of @xmath9-evolution
the photon wave function develops many dipoles which in general diffuse into distance scales beyond the original size @xmath123 .
this increase in parton density and/or size of the photon wave function is generally believed to be the origin of the increasing high energy cross section . in this work
, we do not follow the development of the photon dipole state in detail , we only give a qualitative description of the effective photon size as a function of @xmath9 using the results of the @xmath0 dimensional critical @xmath2 @xmath3 theory as a guiding principle .
we parameterize the longitudinal and transverse photon probability densities as : _
^t&=&_f e_f^2 ^2 [ z^2 + ( 1-z)^2]f_t(x _ ) , + _ ^l&=&_f e_f^2 4 q^2 z^2 ( 1-z)^2 f_l(x _ ) , + & = & .
the perturbative scale for the photon quark density is given by @xmath124 .
we modify the photon wave function depending on the relation of the correlation length @xmath119 of the wilson loops to the perturbative scale @xmath124 .
we set = ( ) ^- . for the reference bjorken parameter @xmath125
the correlation length is fixed at the perturbative scale .
the critical exponent @xmath126 determines the wilson line correlations for @xmath127 . if the transverse size of the dipole is smaller than the perturbative length scale @xmath128 we use the perturbative quark densities @xmath129 and @xmath130
this is the uninteresting case .
for @xmath131 we modify the perturbative photon density using the correlation functions of the critical theory , eqs .
( [ e17p],[e18 ] ) , f_t / l(x _ ) & = & k_1/0(x_)^2 x _ .
[ eftl ] the above parameterizations extends the photon density into the scaling region using the scaling index @xmath121 . for distances beyond the scaling region
the density decays exponentially .
the connections are made in such a way that the density is continuous at the respective boundaries of each region .
the photon density combined with an energy independent dipole - proton cross section determines the structure function @xmath34 and the photon- proton cross section : f_2(x , q^2 ) & = & ( _ p^t , tot+ _ p^l , tot ) , + _
p^t / l , tot&=&d^2 x__0 ^1 dz _
^t / l(x_,z ) _ dip(x _ ) first , we use the golec - biernat - wsthoff @xcite dipole - proton cross section at fixed @xmath125 with @xmath132 to calculate the proton structure function , _
gbw(x_,r_0 ) = _ 0 ( 1-e^- ) , + r_0 = ( ) ^0.145 .
the numerical values entering the above formulas are taken over from the original reference @xcite .
note , the value @xmath133 at @xmath125 is independent of @xmath9 . in fig . 2 we present the proton structure function @xmath34 measured at hera @xcite as a function of @xmath9 for various @xmath11 . the solid theoretical curve describes the data rather accurately .
one must be cautious not to overestimate the agreement , since due to the fixed beam energy the values of @xmath11 and @xmath9 are correlated in electron - proton scattering . for intermediate values
@xmath134 the theory slightly overestimates the data .
if one approximates the gbw - dipole proton cross section by a simple quadratic function at small distances @xmath135 and a constant function for @xmath136 @xcite , [ sgbw ] ( r ) _ 0 ( ( 2 r_0- r ) + ( r-2r_0 ) ) , one can estimate the photon proton cross section rather simply .
small errors are introduced if one further neglects the exponentially suppressed part of the photon density in the integral over large transverse distances and sets the anomalous dimension @xmath137
. then one can integrate the dominant transverse cross section in @xmath34 up to the correlation length @xmath119 analytically . _
t= e_f^2 q^2 _ 0 + _ 0 ^ 1 dz(z^2+(1-z)^2)_0^ r dr + ( 1- ) redefining the integration variable as @xmath138 one obtains the @xmath139 cross section as if the original gbw cross section @xmath140 with running @xmath141 would have been used . _
t= e_f^2 q^2 _ 0 + _ 0 ^ 1 dz(z^2+(1-z)^2)_0^ r dr + ( 1-^2r^2 ) where r(x)&=&r_0()^ , + & = & 0.2 .
therefore the agreement with the data can be understood analytically .
the critical theory gives the phenomenologically obtained power dependence of @xmath142 with @xmath9 .
the favored @xmath9 -dependence of gbw is in the range @xmath143 and @xmath144 . without a model for the proton source , it is not possible to obtain the absolute length @xmath145 .
the structure function @xcite scales as a function of @xmath146 as can be easily derived for the simplified dipole cross section given in eq .
( [ sgbw ] ) . in order to test the sensitivity of our model to the assumed form of the dipole cross section we also tried the rather different dipole cross section of forshaw et al .
@xcite with @xmath147 which parameterizes the weak energy dependence of the cross section for large dipoles separately from the strong energy dependence of the cross section for small dipoles .
( s , r)&=&_soft(s , r)+_hard(s , r ) , + _ soft(s , r)&=&g_1(r)s^0.06 , + _ hard(s , r)&= & g_2(r)s^0.38 . in the expression for @xmath148
we fix @xmath125 .
the @xmath11 dependence in the cross section remains .
these cross sections for different values of @xmath11 look rather different from the gbw cross section and its simplified form ( cf .
fig . 3 ) . using the same modified photon wave function inside the scaling region we find the theoretical ( dashed ) curves in fig . 2 . in spite of the rather different shape of the dipole proton cross section the agreement of the theory with the data is again rather good . in our opinion
this represents a strong indication that the theory does not rely on a specific choice of the dipole proton cross section .
any reasonable choice of dipole - proton cross section at fixed @xmath149 should do .
the presented calculation of the proton structure function suggests the discovery of a new critical point in @xmath2 .
besides the fixed point of asymptotically vanishing coupling at infinite momentum transfer @xmath19 we propose that @xmath2 has another critical point at infinite energies @xmath150 .
it is characterized by the exponents of the center group @xmath6 of @xmath3 and determines the correlation functions of wilson lines near the light cone . a correlation length which increases as @xmath151 with the critical exponent @xmath5 given by the 3-state potts model @xmath6 in three dimensions gives the correct effective photon wave function to explain the growing photon proton cross sections observed in hera @xcite .
this result is independent of the specific form of the dipole proton cross section which is used as an input at the scale @xmath125 . for the golec - biernat - wuesthoff @xcite and the forshaw et al .
@xcite cross sections good agreement can be achieved using the effective photon dipole density which has been modified according to the critical scaling greens function . in this paper
we have made a further step extending the picture of a dilute parton gas into a liquid like phase at small @xmath9 . at small @xmath9 a critically diverging correlation length in transverse space
leads to increasing cross sections in qualitative agreement with the observed growth of structure functions for high @xmath11 deep inelastic scattering . in various studies @xcite of high energy scattering the conformal invariance of the resulting effective theory has been pointed out . in @xmath0 dimensions near the critical point
the effective theory is also conformal invariant , whether the greens functions in transverse space show this symmetry is not clear . clearly on the theoretical side
it is necessary to take into account the target at @xmath152 in the calculation . as a next step in a lattice calculation ,
we plan to include an external source representing the proton target and calculate the wilson line correlation functions with twisted boundary conditions .
averaging over a stochastic source distribution is also possible @xcite .
after such a calculation one does not have to rely anymore on parameterized dipole proton cross sections which make the extraction of the critical dynamics difficult .
the agreement so far between the reduced @xmath2 hamiltonian near the light cone with experiment gives a strong motivation to pursue the investigation of the full hamiltonian in @xmath3 .
* acknowledgments * : we thank our colleagues f. wegner , r. khn , c. ewerz , m. thies and f. lenz for helpful discussion on the critical system and its possible realizations in the high energy world .
we are grateful to f. karsch and especially j. engels to communicate results on the scaling function in @xmath6 prior to publication .
a. d. martin , r. g. roberts , w. j. stirling and r. s. thorne , arxiv : hep - ph/0201127 .
i. i. balitsky and l. n. lipatov , sov .
j. nucl .
* 28 * , 822 ( 1978 ) [ yad .
* 28 * , 1597 ( 1978 ) ] ; e. a. kuraev , l. n. lipatov and v. s. fadin , sov .
jetp * 45 * , 199 ( 1977 ) [ zh . eksp
. teor .
72 * , 377 ( 1977 ) ] .
i. balitsky , nucl .
b * 463 * , 99 ( 1996 ) [ arxiv : hep - ph/9509348 ] .
v. s. fadin and l. n. lipatov , phys .
b * 429 * , 127 ( 1998 ) [ arxiv : hep - ph/9802290 ] .
m. ciafaloni and g. camici , phys .
b * 430 * , 349 ( 1998 ) [ arxiv : hep - ph/9803389 ] .
g. altarelli , r. d. ball and s. forte , arxiv : hep - ph/0104246 . l. v. gribov , e. m. levin and m. g. ryskin , phys .
* 100 * , 1 ( 1983 ) ; y. v. kovchegov and a. h. mueller , phys .
b * 439 * , 428 ( 1998 ) [ arxiv : hep - ph/9805208 ] .
e. m. ilgenfritz , y. p. ivanov and h. j. pirner , phys .
d * 62 * , 054006 ( 2000 ) [ arxiv : hep - ph/0003005 ] .
d. j. gross and f. wilczek , phys .
d * 9 * , 980 ( 1974 ) .
h. d. politzer , phys . rept . * 14 * , 129 ( 1974 ) .
e. v. prokhvatilov and v. a. franke , sov .
j. nucl . phys .
* 49 * , 688 ( 1989 ) [ yad .
fiz . * 49 * , 1109 ( 1989 ) ] .
m. burkardt and y. koike , arxiv : hep - ph/0111343 . s. j. brodsky , h. c. pauli and s. s. pinsky , phys . rept .
* 301 * , 299 ( 1998 ) [ arxiv : hep - ph/9705477 ] .
m. burkardt , adv .
* 23 * , 1 ( 1996 ) [ arxiv : hep - ph/9505259 ] .
b. svetitsky and l. g. yaffe , nucl . phys .
b * 210 * , 423 ( 1982 ) .
f. karsch and s. stickan , phys .
b * 488 * , 319 ( 2000 ) [ arxiv : hep - lat/0007019 ] .
f. karsch , c. schmidt and s. stickan , arxiv : hep - lat/0111059 .
hans j. pirner and feng yuan , nucl .
suppl . * 108 * , 313 ( 2002 ) .
j. engels , private communication .
k. golec - biernat and m. wusthoff , phys .
d * 59 * , 014017 ( 1999 ) [ arxiv : hep - ph/9807513 ] ; phys .
d * 60 * , 114023 ( 1999 ) [ arxiv : hep - ph/9903358 ] .
i. abt _ et al .
_ [ h1 collaboration ] , nucl .
b * 407 * , 515 ( 1993 ) ; t. ahmed _ et al .
_ [ h1 collaboration ] , nucl . phys .
b * 439 * , 471 ( 1995 ) [ arxiv : hep - ex/9503001 ] ; s. aid _ et al .
_ [ h1 collaboration ] , nucl .
b * 470 * , 3 ( 1996 ) [ arxiv : hep - ex/9603004 ] ; c. adloff _ et al .
_ [ h1 collaboration ] , nucl .
b * 497 * , 3 ( 1997 ) [ arxiv : hep - ex/9703012 ] .
m. derrick _ et al . _
[ zeus collaboration ] , z. phys .
c * 72 * , 399 ( 1996 ) [ arxiv : hep - ex/9607002 ] ; phys .
b * 316 * , 412 ( 1993 ) ; z. phys .
c * 65 * , 379 ( 1995 ) ; z. phys .
c * 69 * , 607 ( 1996 ) [ arxiv : hep - ex/9510009 ] ; j. breitweg _ et al .
_ [ zeus collaboration ] , phys .
b * 407 * , 432 ( 1997 ) [ arxiv : hep - ex/9707025 ] .
k. golec - biernat , arxiv : hep - ph/0109010 . | we discuss deep inelastic scattering at high energies as a critical phenomenon in @xmath0 space - time dimensions . in the limit of bjorken @xmath1 , @xmath2 @xmath3 with quark fields
becomes a critical theory with a diverging correlation length @xmath4 where the exponent @xmath5 is obtained from the center group @xmath6 of @xmath3 .
we conjecture that the dipole wave function of the virtual photon for transverse sizes @xmath7 obeys correlation scaling @xmath8 before exponentially decaying for distances larger than the correlation length . using this behavior combined with different @xmath9 -independent dipole proton cross sections we calculate the proton structure function and compare with the experimental data .
we take the good agreement with the measured proton structure function f@xmath10 as an indication that at high energies dimensional reduction to an effective three dimensional theory with a critical point occurs . |
of recent years the proton spin puzzle has attracted much attention . as reported in ref .
@xcite , the quark spin is found to be responsible for only a small amount of the proton spin .
the gluon spin contribution is also estimated from the global analysis of the parton distribution function@xcite , but has yet large uncertainties .
the experiment of polarized proton collision at rhic in progress is expected to give some clue to the gluon spin content .
such being the situation , it is desirable to study the possibility that the orbital angular momentum may contribute to some extent to the proton spin and the orbital motion may induce intrinsic deformations of the baryons .
non - spherical components in baryons have been discussed in the context of the color magnetic interaction@xcite , the electromagnetic transition@xcite and the generalized parton distribution@xcite .
deformed baryon states are investigated also in soliton models@xcite .
since the topological soliton is quantized to the fixed ( iso-)spin states by rotation , the intrinsic deformations directly correlate with orbital motions in this description . considering that the shape of the charge density is probed via the photon
, we study the skyrme model coupled with the electromagnetic field .
note that there are two kinds of terms which are brought about by the coupling of the gauge field ; non - anomalous terms through covariant derivatives and anomalous terms through the gauged wess - zumino ( wz ) term .
the gauge fields are minimally incorporated through the covariant derivatives . using variation of an action obtained by the minimal replacement , the authors of ref .
@xcite computed configurations of the gauge field as well as the soliton and estimated the magnetic moment .
they found that the dipole magnetic field is generated around the soliton and the magnetic moment has reasonable value .
however the topological current is not taken into account in their variational equations .
the topological baryon current is naturally incorporated in the maxwell equation if one considers the gauged wz term .
the gauge fields affect the system not only through the covariant derivatives but also through the gauged wz term , which is not obtained by the minimal replacement of derivatives in the wz term .
the gauged wz term is designed to account for the non - conservation of the axial current by the anomaly and to describe correctly anomalous processes like @xmath0 in the pion sector .
this term possibly influences the system also in the soliton sector .
above all , the gauged wz term is indispensable to assure the gauge invariance of conserved currents like the baryon current and the electric current as well .
actually , the gauged wz term provides an anomalous contribution@xcite of the magnetic field to the isospin charge of the soliton .
although the electromagnetic field is usually treated perturbatively owing to the small coupling constant , it is worth estimating the anomalous contribution of the gauge field as the wz term is essential for the anomalous baryon current .
it is possible that the gauged wz term exerts significant influences on the physical quantities of the isospin fixed solitons . in this work ,
we adopt an action including the gauged wz term and construct spin polarized skyrmions with a proper electric charge imposing boundary conditions upon the chiral field and the gauge field .
we calculate the configuration of these fields and study the properties of the spin polarized skyrmions , particularly characteristic spatial structure of the electric charge , current and the magnetic field . through these results ,
we discuss the intrinsic deformations of the charge density as well as the significance of the effects of the electromagnetic fields from the gauged wz term in the soliton model .
we present a model written in terms of the chiral field and the gauge fields . to make it manifest how the gauge fields are incorporated in the system , we consider a chirally gauged model first in a general context and
then restrict ourselves to a model coupled with the electromagnetic field only .
the action of the chiral field consists of the non - linear sigma term , the skyrme term with pion mass term and the wess - zumino term as @xmath1 ^ 2 + \frac{1}{8}m_\pi^2f_\pi^2({\rm tr}u-2 ) \right ) + \vargamma_{\rm wz } \ .\end{aligned}\ ] ] the wess - zumino term , @xmath2 , is expressed compactly by the use of differential forms as @xmath3 with the abbreviation p.c . for parity conjugate : @xmath4 .
the covariant derivative for the chiral field is defined as @xmath5 .
this action except the wess - zumino term and the pion mass term has the gauge symmetry and is invariant under the gauge transformation : @xmath6 the wess - zumino term is constructed so as to reproduce the chiral anomaly attended by this transformation and hence can not be obtained by the minimal replacement of differential operators which makes an action gauge - invariant .
the noether current@xcite associated with the vector symmetry , _ i.e. _ @xmath7 , generated by @xmath8 is @xmath9[ud^\mu u^\dag , ud^\nu u^\dag ] + ( { \rm p.c . } ) \right ) \nonumber \\ & - \frac{in_{\rm c}}{48\pi^2 } \epsilon^{\mu\nu\lambda\rho } { \rm tr}\ iq [ ( udu^\dag)_{\nu\lambda\rho}^3- \{u d_\nu u^\dag , f_{{\rm r}\lambda \rho}+ \frac{1}{2 } u f_{{\rm l}\lambda\rho}u^\dag \ } - ( { \rm p.c . } ) ] , \label{cur}\end{aligned}\ ] ] where @xmath10 , @xmath11 are the field strengths for the gauge fields @xmath12 , @xmath13 respectively .
if we set @xmath14 to get the baryon current , only the last term the contribution from the wz term survives and we find the baryon number is equivalent to the winding number .
therefore we can identify a baryon with a topological soliton in this model .
now we restrict ourselves to study of the su@xmath15 chiral field and the u@xmath16 gauge field , @xmath17 .
the electric charge matrix satisfies the gell - mann nishijima relation : @xmath18 .
a soliton of a unit baryon number is realized in the hedgehog form : @xmath19 with the boundary conditions @xmath20 .
the electric potential is assumed to be spherical @xmath21 for simplicity , and the angular dependence of the vector potential is fixed as @xmath22 to match the ampre law .
here , @xmath23 and @xmath24 are the polar and azimuthal angle respectively .
all these do not depend on time and such a static configuration is always taken by the gauge transformation ( [ g - tr ] ) .
even a ( time - dependent ) collective rotation of the chiral field can be described by taking @xmath25 as the rotation matrix , if one lets @xmath26 take a constant value in proportion to the angular velocity .
the total lagrangian density including the kinetic term of the electromagnetic field is written in terms of @xmath27 and @xmath28 as @xmath29 + \frac{s_f^2}{2\epsilon^2}\left[\left(f^{\prime 2}+\frac{s_f^2}{r^2}\right ) \right(v^2\sin^2\theta -\frac{(1+h\sin^2\theta)^2}{r^2}\left ) -\frac{f^{\prime 2}}{r^2 } \right ] \\ & -\frac{1}{4}m_\pi^2f_\pi^2(1-c_f ) + \frac{1}{4\pi^2r^2}\left[s_f^2 f^{\prime 2}v + \frac{1}{2}s_fc_fhv'\sin^2\theta -v\left(f'h\cos^2\theta+ \frac{1}{2}c_fs_f h'\sin^2\theta\right)\right ] \\ & + \frac{1}{2e^2}\left(v^{\prime 2}-4\frac{h^2}{r^4}\cos^2\theta -\frac{h^{\prime 2}}{r^2}\sin^2\theta\right ) \ , \end{aligned}\ ] ] where @xmath30 and @xmath31 .
after the angular integration and scaling @xmath32 by @xmath33 to be dimensionless , we are led to the variational equations ; @xmath34 the gauss law @xmath35 with the angular averaged charge density , @xmath36,\end{aligned}\ ] ] and an equation deduced from the ampre law @xmath37 , @xmath38,\end{aligned}\ ] ] with the fine structure constant @xmath39 .
note that the magnetic field @xmath40 ( or @xmath41 in r.h.s of eq .
( [ eqv ] ) ) is instrumental in producing partly the electric charge @xmath42 as a source of the electric field @xmath43 and that @xmath43 contributes the electric current @xmath44 as a source of @xmath40 .
this complementarity comes from the gauged wz term in the action which provides the anomalous couplings between electromagnetic field and the chiral field like @xmath45 .
the term brings to @xmath46 an anomalous current of the dual field strength as @xmath47 with the non - vanishing pion field @xmath48 in the soliton sector .
in addition to this , here we emphasize that the topological term of the baryon density is correctly incorporated in the charge density due to the gauged wz term , as one can see the first term in r. h. s. of eq .
( [ eqv ] ) .
we solve the coupled equations ( [ eqf])-([eqh ] ) imposing boundary conditions for the fields , @xmath49 where @xmath50 is the charge number of the soliton ( to be set at 1 or 0 ) and @xmath51 is a modified pion mass@xcite caused by ` rotation ' of the soliton .
the angular velocity is turned to an asymptotic constant @xmath52 by a gauge transformation as mentioned previously and thus non - zero @xmath52 corresponds to a ` rotation ' of the soliton . in practical computing
, we impose a boundary condition on @xmath53 instead of @xmath54 without any tuning of @xmath52 and confirm finite values , actually .
however , the rotation matrix does not cover the whole su(2 ) group because the generator @xmath55 of u@xmath56(1 ) gauge group includes only the @xmath57 component of the isospin matrix .
accordingly , @xmath58 can not be excluded from nucleon states without the casimir projection as performed in ref .
@xcite . in this sense ,
our treatment is not equivalent to the collective quantization of rotation fully , but to one - dimensional cranking .
nonetheless , the @xmath57 component of isospin ( as the electric charge minus one half of the baryon number ) is assured of a quantized value @xmath59 for the charged and neutral soliton ( @xmath60 ) only by the boundary conditions for @xmath61 and @xmath62 , though this does not mean a topological quantization of isospin in contrast to that of the baryon number . since the asymptotic value @xmath52 affects the boundary condition of @xmath63 , the self - consistent approach we took to solve the differential equations corresponds to the variation after projection .
furthermore , the @xmath57-component of the spin takes @xmath64 for the charged and neutral solitons , since the spatial rotation means the inverse iso - rotation for the hedgehog configuration and the photon contribution to the spin is negligible which we confirm numerically .
therefore we regard the soliton as a spin polarized nucleon with possible admixture of the @xmath58 component .
we solve the differential equations with the boundary conditions for proton ( @xmath65 ) and neutron ( @xmath66 ) using the relaxation method .
three sets of parameters @xmath67 and @xmath68 are chosen while @xmath69 is fixed at the physical pion mass .
set i : @xmath70 is determined so that the charge radius of the proton is reproduced with physical pion decay constant .
set ii : @xmath71 and set iii : @xmath72 are taken from refs .
@xcite and @xcite , respectively .
we find that for set i , @xmath73 such that @xmath74 decreases by a few % compared with that of set ii .
while this means @xmath63 for set i is squeezed , @xmath63 for set iii , on the contrary , swells compared to set ii .
although the boundary condition constrains the fields to have the different asymptotic behavior , profiles of @xmath63 with sets ii and iii are almost the same as that obtained in refs .
@xcite and @xcite respectively .
the configurations of the electromagnetic field are also determined simultaneously , as shown in fig .
[ fig : vh ] . .
bold lines are for the charged soliton and thin lines are for the neutral one .
the electric potentials ( normalized by @xmath33 ) with the asymptotic constants subtracted are shown by solid lines and @xmath75 s which determine the magnetic field are by dashed lines.,width=292 ] the electric potentials display correctly the coulombic behavior for large @xmath76 , and the smeared peak near the origin means finite structures of soliton s charge unlike a point particle .
the function @xmath28 gives the magnetic field as @xmath77 and the radial distance of their peak determines the size of the dipole magnetic field . the spatial distribution of the fields discussed in detail later .
physical quantities with these parameter sets are shown in table [ tab ] .
.static properties of the nucleon described by the soliton with several sets of parameters .
@xmath78 , @xmath79 and @xmath80 are the mean square charge , baryon number and magnetic radii . [ cols="<,^,^,^,^,^,^,^,^,^ " , ] for sets i and iii , the mass of solitons deviates from the experimental values , but these receive corrections of meson loop and higher chiral - order terms@xcite .
mass differences between proton and neutron are of proper magnitude for all sets of parameters but show the opposite sign .
this is because the spherical ansats for the electric field is oversimple to give the correct sign , though we take the electric field into account even for the neutron .
in fact , the energy of the electric field contributes about 0.2% to the proton mass which is reasonable for the mass difference , while it almost vanishes for the neutron case .
this setting is also reflected in the charge radius defined as @xmath81 with the charge density , @xmath82 which is obtained with assignment @xmath83 in eq .
( [ cur ] ) and scaling to the dimensionless variables . for the proton
, the positive charge density is piled up and the charge radius is well reproduced comparatively . for the neutron ,
however , the positive and negative charge densities compete with each other and the charge radius of neutron is overestimated several times as much .
the main contribution to the charge density , ( [ j0 ] ) , comes from the first term of the baryon - number density near the origin and from the last term of the isospin charge for the medium range .
since the latter has a remarkable polar - angle dependence , the spatial distribution of the electric - charge density presents a deformed or a toric structure with a core according to the isospin , as shown in fig .
[ fig : ch ] .
the baryon number density depends also on the polar angle through the gauge field , @xmath84 but it is almost spherical because the small fine structure constant suppresses the magnitude of field @xmath41 . on the other hand ,
the isospin charge is mainly contributed from the non - linear sigma term proportional to @xmath85 as is seen in eq .
( [ j0 ] ) .
this term affects the electric charge with the opposite sign for charged and neutral solitons , especially in the region such that @xmath86 and @xmath87 . as seen in fig .
[ fig : chzen ] , the solid curves of the charged soliton change according to the polar angle and smoothly merge into the dashed lines of the neutral one at the polar region .
although the angular - averaged density of the electric charge agrees with that obtained in ref .
@xcite , the @xmath23 dependence makes an oblate shape for the charged soliton and a prolate positive - charge surrounded by a toric negative - charge for the neutral soliton .
the deformed charge density is interpreted as a consequence of the centrifugal force accompanying the rotation of soliton .
the radius of the negative - charge torus is about @xmath88 fm with the parameter set ii , which is governed by the distance @xmath73 such that @xmath89 .
since such a deformed structure can not be realized for the nucleon with spin @xmath90 , the density distribution is interpreted to stand for an intrinsic deformation of the nucleon and also to reflect the admixed @xmath58 component .
plane ( left ) and the neutral soliton in the @xmath91 space ( right ) calculated with the parameter set ii .
note that the soliton spin is polarized along the @xmath57 axis .
gray and black dots are positive and negative charges .
the charge density of the charged soliton is deformed to an oblate shape and that of the neutral soliton has a saturnian structure.,title="fig:",width=264,height=264 ] plane ( left ) and the neutral soliton in the @xmath91 space ( right ) calculated with the parameter set ii .
note that the soliton spin is polarized along the @xmath57 axis .
gray and black dots are positive and negative charges .
the charge density of the charged soliton is deformed to an oblate shape and that of the neutral soliton has a saturnian structure.,title="fig:",width=283,height=264 ] rl plane show the magnetic field , which is axial symmetric .
dots in the negative @xmath92 region represent the current density running around the @xmath57 axis clockwise.,width=302 ] plane show the magnetic field , which is axial symmetric .
dots in the negative @xmath92 region represent the current density running around the @xmath57 axis clockwise.,width=302 ] the magnetic moment @xmath93 is evaluated from the electric current as @xmath94 with @xmath95\ , \label{ji}\end{aligned}\ ] ] which is equivalent to the estimation from asymptotic behavior of the magnetic field , @xmath96 , owing to the ampre law .
our results in unit of the nuclear magneton are comparable to the experimental values including their sign . the reasonable estimation of the magnetic moment results from the dipole magnetic field generated by the current of soliton
actually , the magnetic field expressed by eq .
( [ eb ] ) has a dipole structure induced by the circular current eq .
( [ ji ] ) , as shown in fig .
[ fig : bj ] .
these field configurations are the same for both of the charged and neutral cases , while the spin of the neutral ( charged ) soliton is polarized ( anti-)parallel to the @xmath57 axis .
the radius of the circular current is also controlled by the scale @xmath73 satisfying @xmath89 .
the spatial distribution of the magnetic field is understood also on an analogy with the meissner effect in the super conductor .
the coefficient of the vector potential in the current , @xmath97 , plays a role of a square ` mass ' of the gauge field , because this factor can be read as the coefficient of @xmath98 in the action . in case of the hedgehog configuration ,
the large ` mass ' region forms a torus whose core , as one can see from the factor @xmath99 , is located at the circle around the @xmath57-axis such that @xmath100 and @xmath101 .
the configuration of the magnetic field is determined so as to avoid the large ` meissner mass ' region to save the total energy . as a consequence of this fact , the magnetic field coils around a genus of the torus .
the axial coupling tends to be one half of experiments , which agrees with refs .
the pion - nucleon sigma term evaluated from the symmetry breaking term results in the adequate value as shown in table [ tab ] . in any case , contributions of the electromagnetic field via the gauged wess - zumino term are controlled by the fine structure constant and hence amount to 1% of the physical quantities , at most .
therefore the complementary dynamics of the electromagnetic field is not conspicuous in the present situation .
however , this anomalous contribution might be influential in high dense matter .
we have studied the effects of the electromagnetic field coupling to the soliton configuration via the gauged wess - zumino term .
we pointed out that the gauged wess - zumino term provides the maxwell equation with anomalous terms in the presence of the soliton .
unless this term is considered , the topological baryon current can not correctly be incorporated in the charge density of variational equations .
furthermore non - vanishing pion fields make the anomalous coupling @xmath102 effective in the soliton sector , and the electric and magnetic fields contribute complementarily to the source current of each other .
such a current from the dual field strength is interesting theoretically , but in reality these effects are suppressed by the small coupling constant .
because the anomalous contribution of the electric field is small , the spatial distribution of the electric current is mainly determined by the non - anomalous term of the isospin current . from this , it is concluded that the electric current has the toric configuration , and accordingly that the magnetic field has a poloidal structure coiling around the circular current according to the ampre law .
this dipole magnetic field is understood as a consequence of the meissner - like effect .
actually , the meissner ` mass ' of the gauge field is estimated from the chiral angle @xmath61 as @xmath103 corresponding to the condensate of the cooper pairs .
contrast to the superconductor , the large mass region does not extend uniformly but forms a finite torus , and hence the magnetic flux quantization can not be achieved in our case . as discussed above ,
the symmetric axis of the torus of the field configuration is to be interpreted physically as a spin - quantization axis of the soliton .
this leads us to a reasonable consequence that the magnetic moments evaluated from the asymptotic form of the poloidal magnetic fields are ( anti-)aligned with the spin of the charged ( neutral ) soliton .
the theoretical value of the magnetic moment derived from the variational solution is of the same order as the experimental value of the nucleon .
we have found that the electric charge distributions are distinctive for their polar angle dependence .
the equatorial swelling of the isospin charge makes the difference while the spherical core of baryon number density is common to charged and neutral solitons .
our results show that the charged soliton has the oblate shape of the charge density and the neutral soliton has the toric negative charge surrounding the prolate shape of positive charge .
these distributions are consistent with the pion cloud rounding the nucleon or may imply a u - d diquark core accompanied by the other valence quark . in any case , the characteristic distributions of the charge density suggest the intrinsic deformation of baryons and the isospin - dependent quadrupole moment , which might be observed in the @xmath104 transition . as in the dynamo theory for the planetary magnetic field ,
a poloidal magnetic field may induce a toroidal magnetic field through the rotation of the soliton , and _ vice versa _
, even though the physical scale is quite different .
it is our future work to take account of this effect by extending the variational space of the magnetic field .
the non - radial electric field is also important to estimate the mass of the proton and the neutron and to discuss a correlation between the orbital motion and the intrinsic deformation .
one of the authors ( m. o ) would like to acknowledge enlightening discussions with k. yazaki .
he is indebted to o. morimatsu , h. fujii and h. enyo for useful comments .
this work was supported in part by the japan society for the promotion of science for young scientists .
99 european muon collaboration , phys .
* b 206 * ( 1988 ) , 364 ; nucl . phys .
* 328 * ( 1989 ) , 1 m. hirai , s.kumano and n. saito , phys .
* d 69 * ( 2004 ) , 054021 .
n. isgur , g. karl and r. koniuk , phys .
* d 25 * ( 1982 ) , 2394 .
a. wirzba and w. weise , phys .
* b 188 * ( 1987 ) , 6 ; a. hosaka , m. takayama and h. toki , nucl .
phys . * a 678 * ( 2000 ) , 147 .
a. v. belitsky , x. ji and f. yuan , phys .
* d 69 * ( 2004 ) , 074014 . c. hajduk , b. schwesinger , phys .
b145 * ( 1984 ) , 171 ; j. wambach , h. w. wyld and h. m. sommerman , phys .
* b 186 * ( 1987 ) , 272 . j. p. blaizot and g. ripka , phys . rev .
* d 38 * ( 1988 ) , 1556 .
f. leblond and l. marleau , phys . rev .
* d 58 * ( 1998 ) , 054002 .
b. m. a. g. piette and d. h. tchrakian , phys .
rev . * d 62 * ( 2000 ) , 025020 .
y. brihaye , n. k. pak and p. rossi , phys .
b 149 * ( 1984 ) , 191 . n. dorey , j. hughes and m. p. mattis , phys .
rev . * d 50 * ( 1994 ) , 5816 ; m. schleif , r. wunsch , t. meissner , int .
j. mod .
* e7 * ( 1998 ) , 121 . g. adkins , c. nappi and e. witten , nucl .
* b 228 * ( 1983 ) , 552 .
g. adkins and c. nappi , nucl .
* b 233 * ( 1984 ) , 109 . b. moussallam , ann . of phys .
* 225 * ( 1993 ) , 264 ; g. holzwarth and h. walliser , nucl .
phys . * a 587 * ( 1995 ) , 721 . | in soliton models expressed in terms of the non - linear chiral field , the electric current has an anomalous gauge - field contribution as the baryon current does .
we study the spin polarized skyrmions coupled with the electromagnetic field via the gauged wess - zumino term and calculate configurations of the skyrmion and the gauge field with boundary conditions to ensure the physical charge number for baryons .
although the electromagnetic field via the gauged wess - zumino term affects physical quantities in small amounts , we find that the magnetic field forms a dipole structure owing to a circular electric current around the spin quantization axis of the soliton .
this is understood on an analogy with the meissner effect in the super conductor .
the electric charge distributions turn out to have characteristic structures depending on the total charge , which suggests the intrinsic deformation of baryons due to orbital motions of the constituents . |
elements in the solar system are formed from a variety of nucleosynthesis processes in stars , such as the slow and rapid neutron - capture processes ( the s - process and r - process respectively ) .
the s - process signature of the solar system abundances can be split into three components ; the weak - component , which involves nuclei with @xmath0 , the main - component , which involves nuclei having atomic mass between @xmath3 , and the strong component , which accounts for the production of the solar @xmath4pb @xcite .
the s - process site is attributed to massive stars for the weak component , and to agb stars having initial mass between 1 and 3@xmath1 at solar - like metallicity for the main component and at low metallicity for the strong component @xcite .
stellar evolution models of massive stars can be used to determine information on the conditions within stellar interiors and calculate the s - process yields relevant for the weak - component .
any changes to the input physics , such as improved laboratory nuclear reaction rates , can affect the evolution of the simulated star and consequently affect the s - process yields .
therefore , different nuclear reaction rates can be tested for their astrophysical impact . in this work
, we will show how variations in the rate of the @xmath2c + @xmath2c reaction affect the evolution of a 25@xmath1 star and the consequential s - process yields of the star .
the motivation for this work originates from nuclear physics experiments and theory ( see for instance spillane et al .
2007 @xcite and gasques et al .
2007 @xcite ) , which are probing low enough energies to investigate the reaction within the gamow window for @xmath2c+@xmath2c fusion .
carbon - core burning occurs at @xmath5 gk .
the dominating reactions are @xmath2c(@xmath2c,@xmath6ne and @xmath2c(@xmath2c , p)@xmath7na which , combined with the efficient @xmath7na(p,@xmath6ne reaction , leaves @xmath8o and @xmath9ne as the dominant isotopes in the core ( the @xmath8o being a remnant from the previous helium burning stage ) .
in addition to core burning , there are typically multiple stages of shell burning later in the evolution of the star , where the number of shells differs depending on the initial mass of the star .
the main neutron source for the s - process in massive stars is @xmath10ne by the reaction @xmath10ne@xmath11n@xmath12 mg .
@xmath10ne is produced in helium burning from the @xmath13n@xmath14f@xmath15o@xmath16ne reaction chain during the supergiant phase .
@xmath10ne is partially burnt at the end of helium burning with a neutron density , @xmath17 @xmath18 ( see , for example , raiteri et al . 1991 @xcite ) .
the @xmath10ne left in the co - core is then burnt during carbon - shell burning with higher neutron densities ( @xmath19@xmath18 ) , with the @xmath2c(@xmath2c,@xmath6ne reaction providing the @xmath20 particles @xcite .
the carbon - shell burning is sensitive to the profile of @xmath2c after core - carbon burning , which is in turn sensitive to the still uncertain @xmath2c@xmath21o rate and the choice of convection physics in the models @xcite@xcite@xcite .
in addition , the yields are sensitive to changes in reactions involving @xmath10ne or its formation @xcite@xcite .
most of the s - process - rich material ejected by the supernova event of a 25@xmath1 star is formed by carbon - shell burning , affecting the weak s - component @xcite@xcite .
stellar models of a 25@xmath1 star with metallicity @xmath22 were generated using the geneva stellar evolution code , with the nuclear network post - processed using the nugrid ppn tool @xcite .
the post - processing was computed using the khaos cluster at keele university .
variations in the carbon burning rate were chosen in relation to the ` standard ' caughlan and fowler rate ( cf88 from now on ) .
models were generated with the cf88 rate ( referred to as model c12s in this work ) , the cf88 rate multiplied by 10 ( c12t10 ) and divided by 10 ( c12d10 ) .
the ratio of the p- and @xmath20-channels was chosen to be 35%:65% @xcite . the additional uncertainty associated with this choice of ratio will be investigated in a forth - coming paper .
models at solar metallicity using different @xmath2c burning rates . here ,
( a ) is model c12t10 , ( b ) is model c12s and ( c ) is model c12d10 .
the shaded regions correspond to convection zones present in the star , with the type of burning indicated by labels . ]
differences in the convective history can be seen in each of the different models .
[ fig : kips ] shows structure evolution diagrams for the models c12t10 , c12s and c12d10 . in each case the convection zones for carbon shell burning at @xmath23(time in years until core collapse ) @xmath24 1 to -2 differ in size and duration . for model c12t10 ,
the duration of the second carbon - shell burning process is much longer and the size of the convection zones is larger .
a large fraction of overlap between the first shell and the second shell is observed and isotopes produced in the first shell by the s - process will be mixed into the second .
the presence of overlapping carbon shells was previously noted by chieffi et .
for model c12d10 , the second carbon - shell burning episode occurs later in time than for the other two models , which is due to the star contracting further before it can reach a temperature high enough so that carbon burning is activated in the shell .
[ fig : ab_c12d10 ] shows the relative abundances of stable isotopes in the second carbon shell of the c12d10 model with respect to the c12s model .
[ fig : ab_c12t10 ] shows the equivalent plot for the c12t10 model with respect to the c12s model . in fig .
[ fig : ab_c12d10 ] , a clear signature of a higher neutron density is shown , e.g. higher production of r - only species @xmath25zn , @xmath26ge and lower production of @xmath27sr due to the @xmath28se branching .
the reason is that the star in model c12d10 contracts further , resulting in carbon shells that burn at a higher temperature than the c12s model .
therefore the @xmath10ne neutron source is burned at a higher temperature , increasing the neutron density .
the s - process efficiency is quite similar in model c12s and c12d10 ( see , for example , the similar production of neutron magic isotope @xmath29sr ) . concerning model c12t10 , in fig . [
fig : ab_c12t10 ] , a general increase in the s - process efficiency for isotopes with @xmath0 is shown , compared to model c12s .
the overlap between the first and second convective shells causes the initial distribution of isotopes at the start of the second carbon - shell burning to be affected by the products of the first shell . in the first shell ,
neutrons are also efficiently produced by an additional neutron source , @xmath30o(@xmath20,n)@xmath31ne , where @xmath30o is mostly produced by @xmath8o(p,@xmath32)@xmath30f(@xmath33)@xmath30o . in this paper
, it has been shown that changes to the carbon burning rate by a factor of 10 in stellar models of a 25@xmath1 star significantly affect the s - process yields .
the overlap of convective carbon shells active at different burning temperatures also has important implications for the s - process yields .
further analysis will be conducted in a forth - coming paper .
10 url # 1#1urlprefix[2][]#2 f , beer h and wisshak k 1989 _ reports on progress in physics _ * 52 * 9451013 f , diehl s , fryer c l , hirschi r , hungerford a , magkotsios g , pignatari m , rockefeller g , timmes f x , young p and bennett m e 2008 _ proceedings of the 10th symposium on nuclei in the cosmos ( nic x ) .
july 27 - august 1 , 2008 mackinac island , michigan , usa .
available online at http://pos.sissa.it/cgi-bin/reader/conf.cgi?confid=53_ e f , rosales p , martinez - quiroz e , murillo g , fernndez m , berdejo h , lizcano d , gmez - camacho a , policroniades r , varela a , moreno e , chvez e , ortz m e , huerta a , belyaeva t and wiescher m 2006 _ physical review c _ * 73 * 064601 | the slow neutron capture process in massive stars ( the weak s - process ) produces most of the s - only isotopes in the mass region @xmath0 .
the nuclear reaction rates used in simulations of this process have a profound effect on the final s - process yields .
we generated 1d stellar models of a 25@xmath1 star varying the @xmath2c + @xmath2c rate by a factor of 10 and calculated full nucleosynthesis using the post - processing code ppn .
increasing or decreasing the rate by a factor of 10 affects the convective history and nucleosynthesis , and consequently the final yields . |
spiral arm structures are the beautiful structures that have fascinated astronomers for a long time .
one of the long - time mysteries of the spiral arm was the so - called `` winding - dilemma '' . from the observations of the rotation curve of disc galaxies
, it is known that the stars in the inner region of the disc rotate faster , i.e. the angular speed is higher , than the stars in the outer region .
therefore , if the spiral arm is a material arm , i.e. the spiral arm is moving at the same speed as the stars , the spiral arm should wind up quickly ( e.g. * ? ? ?
spiral density wave theory described in @xcite solved the issue by considering the spiral pattern as a density wave .
then , the density wave can be a rigidly rotating feature with a constant pattern speed , irrespective of the stellar rotation speed , and consequently long - lived .
recently , thanks to the powerful computational facilities , the resolution of three - dimensional n - body simulations has improved significantly , and the artificial heating from which the previous low - resolution simulations suffered is minimized ( e.g. * ? ? ?
* ; * ? ? ?
such high - resolution simulations allow us to study further the spiral arm theory .
however , even with such high - resolution simulations , so far no single n - body simulation reproduces a long - standing spiral arm feature such as that is suggested in @xcite @xcite .
recent studies show that the spiral arms in the numerical simulations are transient and recurrent ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
for example , @xcite demonstrated that the spiral arm was rotating with the same speed as the stars , i.e. co - rotating ( see also * ? ? ? * ) , and therefore winding .
still in each snapshot , the spiral arms are always apparent , and the spiral arms are constantly forming and disrupting , i.e. recurrent , with a lifetime of about 100 myr . although the co - rotating spiral arm leads to the winding - dilemma , @xcite demonstrated that the spiral arms were disrupted before they wound up completely , and the pitch angle of the spiral arms correlated with the shear rate of the disc , as observed ( e.g. * ? ? ? * ) .
interestingly , the winding nature of the spiral arm seen in n - body simulations can naturally explain the observed scatter in the correlation between the pitch angle and the shear rate ( see * ? ? ?
* for more thorough discussion ) .
@xcite demonstrated that the spiral arms in n - body simulations were forming with a similar mechanism to the so - called swing amplification theory suggested by @xcite and @xcite ( see also * ? ? ?
* ; * ? ? ?
* ) . however , while the swing amplification is considered to happen at a single co - rotation radius , where the spiral arm pattern speed is consistent with the rotation speed of the stars , in the numerical simulations the co - rotation resonance occurs at all radii , and the swing amplification is happening ( or propagating ) at every radius .
this is one of the current explanations for the transient and winding spiral arm features . in others , for example
, the transient features of the spiral arm can be interpreted as overlapping multiple - wave modes ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , where each mode appears around the co - rotation radius and is relatively long lived . still , there is no clear explanation of the origin and nature of the spiral arms , which remains as a challenge for the galactic astronomer .
interestingly , there is also observational evidence against long - lived spiral arms .
for example , @xcite analysed the pattern speed of the spiral arm as a function of radius for ngc 1068 using their generalized version of the method from @xcite , so - called tremane - weinberg method .
they showed that the pattern speed of the spiral arm decreases with radius , and therefore the lifetime of the spiral arm must be short .
similar pattern speeds were also observed with the same technique in other galaxies ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . however , the accuracy and validity of the tremaine - weinberg method are still needed to be tested against the future observations ( e.g. * ? ? ?
* ; * ? ? ?
another observational test is the so - called `` offset '' method .
if there is a long - lived rigidly rotating spiral arm , one can define the co - rotation radius .
since the angular velocity of gas and stars is observed to be faster in the inner region , the gas and stars in the region inside ( outside ) the co - rotation radius will move faster ( slower ) than the spiral arm .
the gas component piles up in the spiral arms , experiencing a shock that induces star formation @xcite . in this scenario ,
the youngest stars born from the molecular clouds in the spiral arms would be found slightly ahead of the arm traced by the molecular gas , if located within the co - rotation radius , and behind the arm outside of the co - rotation radius .
therefore , if we observe the tracers of the different stages of star formation , such as hi , co and h@xmath4 , an offset among them as a function of radius would be expected .
by combining h@xmath4 imaging and @xmath5 near - ultraviolet ( nuv ) data , @xcite distinguished the regions with ongoing star formation and the regions with star formation a few hundred million years ago in the grand - design spiral galaxy , m100 .
contrary to the expectation from the density - wave theory , no offset was found between these two regions .
the same conclusion was reached in @xcite , although some studies claimed to find a significant offset for some galaxies ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the milky way is a ( barred ) spiral galaxy ( e.g * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) which we can observe in great detail .
for example , the detailed map of hi and co observations provide the global position and kinematics of the spiral arm in the gaseous phase ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ?
* ; * ? ? ?
star clusters provide us with reasonable photometric distances , and young star clusters can be used to trace the spiral arm and also measure the pattern speed of the spiral arm using a similar technique to the offset method mentioned above ( e.g. * ? ? ?
* ; * ? ? ?
the influence of the spiral arms on the stellar motion has also been measured and compared with the models ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the maser sources associated with high - mass star forming regions are recognized as a unique source to trace the spiral arm structures , because very long baseline interferometry ( vlbi ) observations allow their parallaxes and proper motions to be measured with great accuracy , @xmath6 @xmath7as .
for example , recently the bar and spiral structure legacy ( bessel ) survey and japanese vlbi exploration of radio astrometry ( vera ) provided the parallaxes and proper motion measurements for over 100 maser sources @xcite .
in the future , the european space agency s _ gaia _ satellite ( launched in december 2013 ) will produce accurate measurements of the parallax and proper motion for about a half billion disc stars ( e.g. * ? ? ?
to study the nature of the spiral arms from these observations , we need to compare the observational data with the theoretical prediction from different scenarios of the origins of spiral arms .
although there are many successful studies for reconstructing the galactic bar structure and the pattern speed by comparing the observational data with the theoretical model prediction , it is more complicated for the spiral arms ( e.g. see * ? ? ?
* for a review ) .
@xcite assumed the rigid rotation of the spiral arm , and made predictions of the radial velocity distribution of stars in the cases of different number of arms , pitch angles and pattern speed .
@xcite studied how the rigidly - rotating spiral structures affect the stellar kinematics and the distribution of radial and rotational velocities of the stars ( see also * ? ? ? * ) .
@xcite studied the vertex deviation map from both rigidly - rotating spiral arms and transient spiral arms .
@xcite discussed the origin of the large peculiar velocities observed for the maser sources at that time @xcite .
they argued that such large peculiar velocities are difficult to be explained with the density wave theory , but using n - body / smoothed particle hydrodynamics ( sph ) simulations , they showed that the observed large peculiar velocities of the maser sources can be reproduced by the transient and recurrent spiral arms .
therefore , more predictions for the observable signatures from the transient and recurrent spiral arms in the numerical simulations should be valuable for future and on - going observational surveys . in this paper , we study the kinematic structure of both the star and gas components around a spiral arm in a simulated barred galaxy similar in size to the milky way .
the simulated galaxy has a transient , recurrent and co - rotating spiral arm similar to that seen in @xcite .
we target a spiral arm similar to the perseus arm .
we then make a prediction of the observational signatures of the kinematics of the stars and gas around the perseus arm , if it is also a transient , recurrent and co - rotating spiral arm .
section [ meth - sec ] describes briefly the numerical simulation code and numerical models .
section [ res - sec ] presents the results .
a summary of this study is presented in section [ sum - sec ] .
we use our original n - body / sph @xcite code , gcd+ , which can be used in studies of galaxy formation and evolution in both a cosmological and isolated setting @xcite .
gcd+ incorporates self - gravity , hydrodynamics , radiative cooling , star formation , sne feedback , and metal enrichment , including metal diffusion @xcite .
we have implemented a modern scheme of sph suggested by @xcite , including their artificial viscosity switch @xcite and artificial thermal conductivity to resolve the kelvin - helmholtz instability ( see also * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
following @xcite , we integrate the entropy equation instead of the energy equation .
as suggested by @xcite , we have added the individual time step limiter , which is crucial for correctly resolving the expansion bubbles induced by sne feedback ( see also * ? ? ?
* ; * ? ? ?
we also implement the fast scheme @xcite which allows the use of different timesteps for integrating hydrodynamics and gravity .
the code also includes adaptive softening length for the gas particles @xcite .
our recent updates and performance in various test problems are presented in @xcite and @xcite .
radiative cooling and heating are calculated with cloudy ( v08.00 : * ? ? ?
* ) following @xcite .
we tabulate cooling and heating rates and the mean molecular weight as a function of redshift , metallicity , density and temperature adopting the 2005 version of the @xcite uv background radiation .
the details of the recipe of star formation and stellar feedback are described in @xcite .
we adopt the star formation threshold density , @xmath8 @xmath9
. the main free parameters of our feedback model include energy per sn , @xmath10 , and stellar wind energy per massive star , @xmath11 , and we adopt @xmath12 erg and @xmath13 erg s@xmath14 for this paper .
the other parameters are the same as @xcite .
we simulate the evolution of a barred disc galaxy similar in size to the milky way .
we initially set up an isolated disc galaxy which consists of gas and stellar discs , with no bulge component , in a static dark matter halo potential @xcite , to save computational costs .
a live dark matter halo can respond to the disc particles by exchanging angular momentum .
this becomes important for long term evolution of a bar ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . however , the effect of the assumed static dark matter halo is expected to be small for transient spiral arms which we focus on in this paper . furthermore , a live dark matter halo is often modeled with particles more massive than disc particles , which may introduce some scattering and heating ( e.g. * ? ? ?
therefore , in the interest of computational speed and a more controlled experiment , we use a static dark matter halo .
we use the standard navarro - frenk - white ( nfw ) dark matter halo density profile @xcite , assuming a @xmath15-dominated cold dark matter ( @xmath15cdm ) cosmological model with cosmological parameters of @xmath16 , @xmath17 , and @xmath18 : @xmath19 where @xmath20 and @xmath21 where @xmath22 is the characteristic density of the profile @xcite , @xmath23 is the distance from the centre of the halo and @xmath24 is the scale radius . the total halo mass is set to be @xmath25 and the concentration parameter is set at @xmath26 . the halo mass is roughly consistent with ( or slightly higher than ) the recent measured mass of the milky way ( e.g. * ? ? ? * ) .
the stellar disc is assumed to follow an exponential surface density profile : @xmath27 where @xmath28 is the stellar disc mass , @xmath29 is the scale length and @xmath30 is the scale height .
following the observational estimates of the milky way , we adopt @xmath31 m@xmath32 , @xmath33 kpc and @xmath34 kpc .
for simplicity , we set up only a thin disc component , and ignore the thick disc contribution .
the gaseous disc is set up following the method described in @xcite .
the radial surface density profile is assumed to follow an exponential law like the stellar disc with a scale length , @xmath35 kpc .
the initial vertical distribution of the gas is iteratively calculated to reach hydrostatic equilibrium assuming the equation of state calculated from our assumed cooling and heating function .
the total gas mass is @xmath36 m@xmath37 .
we use 1,000,000 gas particles and 4,000,000 star particles for the initial condition .
this leads to 10,000 m@xmath32 for each particle , and means that both the star and gas particles have the same mass resolution in our simulations .
this is required from our modelling of star formation , feedback and metal diffusion .
we apply the spline softening and variable softening length suggested by @xcite for sph particles .
we set the minimum softening length at 158 pc ( the equivalent plummer softening length is about 53 pc ) , which corresponds to the required softening for gas to resolve @xmath38 @xmath9 with solar metallicity .
three line of sight directions ( @xmath39 and 150 deg ) are highlighted with the dotted lines .
the galaxy is rotating clockwise . ]
fourier mode normalised to the @xmath40 mode as a function of radius . ]
we ran a simulation for 1 gyr from the initial conditions described above .
we do not include any continuous external inflow of gas for the simplicity of the numerical setup .
we have chosen a snapshot at @xmath41 gyr , because it shows a spiral arm in a similar location to the perseus arm located in the second quadrant of the milky way .
the snapshot is shown in fig .
[ snap - fig ] .
we assumed the location of the observer at @xmath42 kpc in the simulated galaxy .
the mean circular velocity at the radius of 8 kpc is @xmath43 km s@xmath14 . in this paper , we focus on the spiral arm highlighted in fig .
[ snap - fig ] .
the pitch angle of the spiral arm is 39 deg , which is much larger than the typically estimated pitch angle for the milky way spiral arms , @xmath44 deg ( e.g. * ? ? ?
[ ampm2-fig ] shows the amplitude of the @xmath45 fourier mode normalised to the @xmath40 mode calculated with equation ( 12 ) of @xcite .
the amplitude is similar to the observed spiral galaxies ( e.g. * ? ? ?
we focus on the spiral arm in the outer region , @xmath46 kpc , because the amplitude of the spiral arm is higher and the spiral arm is clearer in the outer region .
note that the simulation used is not intended to reproduce the whole structure of the milky way .
the aim of this study is to qualitatively discuss the kinematical signatures of the co - rotating spiral arm in general .
fig [ omgpr - fig ] shows the measured pattern speed of the spiral arm .
the pattern speed is measured in the same way as @xcite by tracing the density peak . though the pattern speed shows some scatter .
the pattern speed looks flat , but with a slight trend of decreasing pattern speed with the radius .
the pattern speed is similar to the circular speed in the radial range between 8 and 12 kpc .
this means that the spiral arm is co - rotating with the stars .
we have confirmed that the spiral arm slowly winds up , and it is disrupted at @xmath47 gyr , i.e. transient .
[ omgpr - fig ] also shows the pattern speed of the bar .
the pattern speed of the bar is measured with spectrogram analysis of the @xmath45 mode as described in @xcite , because the density structure of the bar keeps changing and it is difficult to define the bar position angle for each snapshot ( see also * ? ? ?
* ; * ? ? ?
using a gravitational field method ( e.g. * ? ? ?
* ; * ? ? ?
* ) described in @xcite , we obtain the bar strength of @xmath48 .
table 1 of @xcite summarises the current estimates of the strength of the bar in the milky way , which is between @xmath49 and 0.83 .
the bar strength of our simulated galaxy is consistent with the lowest estimate for the bar strength of the milky way .
@xcite showed numerical simulations with a bar of @xmath50 and 0.11 .
@xcite demonstrated that although the stronger bar causes a flatter pattern speed of the spiral arm , compared with simulations without a bar , the pattern speed of the spiral arm is still similar to the rotation velocity of stars in the spiral galaxy with the bar of this level of strength .
@xcite also suggested that the strong influence of the bar can induce more rigidly rotating spiral arm , which could be tested with the galactic disc surveys , such as _ gaia _ and apogee .
we now show the kinematic features expected from the co - rotating spiral arm .
we focus on the properties of the star and gas particles within the lines of sight highlighted by the white dotted lines in fig .
[ snap - fig ] .
these lines of sight correspond to the galactic latitude of @xmath51 and galactic longitudes of @xmath39 and 150 deg .
the upper panels of fig .
[ disvrot - fig ] demonstrate the rotation velocity of the stars as a function of their distance from the observer .
the sample of star particles are selected within a square angular range of @xmath52 and @xmath53 deg from the three lines of sight highlighted in fig .
[ snap - fig ] .
the vertical line in each panel shows the position of the spiral arm , whose galactocentric radius corresponds to @xmath54 and 10.9 kpc . at 11 kpc .
because the difference is small , we use @xmath43 km s@xmath14 irrespective of the galactocentric radius . ] note that the left ( right ) hand side of the vertical line , i.e. closer to ( further from ) the observer , is the trailing ( leading ) side of the spiral arm .
the colours of the dots indicate the angular momentum change of the star particles during the @xmath55 myr evolution from the time step that we focus on . the short time window of @xmath5640 myr is chosen to highlight the angular momentum change at the snapshot .
the stars in the trailing ( leading ) side of the spiral arm tend to gain ( lose ) angular momentum .
this trend is much clearer in the edge of the arm , compared to around the centre of the arm where there are stars gaining ( losing ) angular momentum at the slightly leading ( trailing ) side of the spiral arm .
@xcite showed that the angular momentum loss and gain was due to the tangential force from the spiral arm , and the stars at the leading ( trailing ) side lost ( gained ) angular momentum at each time .
the tangential force is highest at the edge of the spiral arm , and changes the sign at the centre of the spiral arm . as a results , the radial migration due to the spiral arm is more effective at the edge of the spiral arm than around the centre of the spiral arm .
also note that , the position of the stars in fig . [ disvrot - fig ] is at the snapshot we chose , but the angular momentum gain and loss is measured within the time bin of @xmath5640 myr .
stars around the centre of the spiral arm can easily move to the other side of the spiral arm within 40 myr , which blurs the trend of the loss and gain of angular momentum with respect to the side of the spiral arm .
radial migration ( change of the angular momentum ) is happening at all radii , because the spiral arm is co - rotating with the stars , which is consistent with @xcite .
a large range of rotation velocities are seen around the spiral arm .
[ vrothis - fig ] compares the distribution of the rotation velocity in the trailing side and the leading side , where we selected the star particles in the distance range between 1 and 2 kpc away from the location of the spiral arm ( the solid line in fig .
[ snap - fig ] ) at each line of sight . as shown in @xcite , in general , stars in the trailing ( leading ) side rotate slower ( faster ) , because they tend to be apo - centre ( peri - centre ) phase .
still , there are some stars which rotate very fast ( slow ) , but are in the trailing ( leading ) side of the spiral arm . especially in the leading side , the rotation velocity distribution shows ( at least ) two peaks .
although one of the peaks is faster than the peak of the rotation distribution in the trailing side and close to the circular velocity , the other peak is slower than the peak in the trailing side .
the slowly rotating stars in the leading side are also visible in fig .
[ disvrot - fig ] .
following @xcite , we identified stars with v@xmath57165 km s@xmath14 in the leading side ( d@xmath588 kpc ) at @xmath59 deg and tracked their evolution .
they were located at an inner radius very close to the arm , but slightly in the trailing side , when the spiral arm was forming .
they were at the peri - centre phase then , and were accelerated by the spiral arm formation .
consequently , they overshoot the spiral arm , and at the time of the snapshot they are at the apo - centre phase and are decelerated by the arm , which leads to their slow rotation .
we also tracked the stars rotating faster than 250 km s@xmath14 in the trailing side ( d@xmath583 kpc ) of the arm at @xmath59 deg .
they were rotating at a larger radius before the spiral arm formed .
they were slightly in the leading side of the arm , when the spiral arm starts forming .
then , they were passed by the spiral arm , i.e. moved from the leading side to the trailing side , just before the time step selected .
they are close to the peri - centre phase at the selected time step .
these particles will pass the spiral arm , i.e. they will move from the trailing side to the leading side , after this time step .
therefore , the particles are orbiting around the spiral arm .
the particles that have this type of the orbit are called `` non - migrators '' in @xcite .
the lower panels of fig .
[ disvrot - fig ] show the rotation velocity of the gas component as a function of distance from the observer in the same line - of - slight as the upper panels . because the gas component does not have the epicyclic motion which leads to the variation in the rotation velocities for stars
, there is a clear trend seen around the spiral arm that the gas rotates slower in the trailing side and faster in the leading side , especially at @xmath60 and 150 deg .
( as seen in fig .
[ snap - fig ] at @xmath59 deg , there is a bubble induced by sne , which disturbs the kinematics of the gas component . )
we consider this to be the result of the radial migration of the gas .
for example , the gas in the trailing side gains angular momentum and moves to larger radii .
because the spiral arm is trailing , if the migrating gas is accelerated up to the same speed as the gas at the new radius , it will pass the centre of the arm .
to stay in the trailing side at the new outer radius , it should be accelerated to a slightly slower rotation velocity than that of the spiral arm .
the opposite is required in the leading side , and only the gas rotating faster can stay in the leading side of the arm .
these velocity features in stars and gas are due to the co - rotation of the spiral arm .
these features are seen at different galactic longitude samples , because the spiral arm is co - rotating at all of the radial range on which we focus .
these features , therefore , would be good observable indicators to examine if the milky way spiral arms are co - rotating arms .
recently , @xcite published their parallax and proper motion measurements for over 100 of the maser sources .
we converted their proper motion to the rotation velocity , assuming that the sun s rotation velocity with respect to the circular velocity is @xmath61 km s@xmath14 and the galactocentric distance of the sun is @xmath62 kpc @xcite .
we also scaled the rotation velocity to adjust from the measured circular velocity at the solar radius of @xmath63 km s@xmath14 @xcite to the circular velocity of the simulated galaxy , @xmath43 km s@xmath14 .
then , we compare the rotation velocity of the maser sources with the gas component of the simulation in fig .
[ snap - fig ] .
interestingly , the rotation velocity of the maser sources ( indicated with open stars ) show a similar variation in the rotation velocity to our gas components . in the middle panel of @xmath60 deg sample ,
the maser sources show slower ( faster ) rotation velocity in the trailing ( leading ) side , which is consistent with the prediction of the simulation which has a co - rotating spiral arm .
unfortunately , there are insufficient maser sources to test against the simulated co - rotating spiral arm at different radii .
however , this shows a possible test of the different explanations for spiral arms . fig .
[ disvr - fig ] shows similar results to fig .
[ disvrot - fig ] , but plots the radial ( rather than rotational ) velocity of stars and gas as a function of the distance from the observer in the simulated galaxy .
there are a large variation of galactocentric radial velocities , @xmath64 , for star particles observed around the spiral arm , because their epicycle motion is modified owing to the radial migration @xcite . especially in the gas component , the gas particles in the trailing side of the spiral arm are moving outward , and the gas in the leading side of the spiral arm are moving inward .
this clear sign of the radial migration is observed at different @xmath65 and therefore different radii , because of the co - rotating nature of the simulated spiral arm .
again , although the radial velocity of the maser sources show larger scatter in radial velocity , they show a tentative agreement with the simulation results at @xmath60 deg . deg . ]
although currently the the distances for the milky way stars are not measured accurately at the distance which we are interested in , @xmath662 kpc , figs .
[ disvrot - fig ] and [ disvr - fig ] indicate that there is a large variation in the velocity for the stars around the spiral arm if the spiral arm is co - rotating .
therefore , it is interesting to compare the heliocentric radial velocity distribution of observed stars in the milky way , which is independent of the errors in the distance , with the simulated star particles .
[ vlosl90-fig ] provides a comparison between our simulated data and apogee dr10 data @xcite .
we have chosen the @xmath67 deg data from apogee dr10 data , and have applied the same selection as the hq giant sample of @xcite which is shown in table 1 of their paper .
the solid histogram in fig .
[ vlosl90-fig ] shows the distribution of the radial velocity from the local standard of the rest , @xmath2 , for the hq giant sample at @xmath67 deg , where we assume @xmath61 km s@xmath14 @xcite .
the dotted line is the results for the star particles within @xmath68 deg and @xmath69 deg in our simulation . to increase the sample of star particles
, we applied a larger selection area for our simulation data than the observed area in apogee data .
however , for this crude comparison , we do not think that this is a serious issue . because we do not know the distance of the apogee stars , for simplicity we have selected the star particles whose distance from the observer is less than 3.5 kpc , simply assuming that stars further than that are not observed by the apogee survey due to the strong extinction in the spiral arm . from fig .
[ disvrot - fig ] , we expect the variety of @xmath2 around the co - rotating spiral arm , and it is not surprising to see the large range of @xmath2 in the simulation data .
interestingly , the apogee data show a significant number of stars with @xmath70 km s@xmath14 , as in the simulations .
these high @xmath2 stars are induced by the co - rotation resonance of the spiral arm in our simulation , and this could be another observational signature of the co - rotating spiral arm .
however , @xcite demonstrated that the heliocentric radial velocity distribution at @xmath3 deg and the other apogee fields can be described well with an axisymmetric model , i.e. the width of the velocity distribution can be explained purely by the velocity dispersion .
therefore , we can not make any conclusions from this comparison .
still , it is encouraging to see that our simulated galaxy has a velocity distribution consistent with the apogee data . from figs .
[ disvrot - fig ] and [ vrothis - fig ] , we expect that the velocity distribution would be different between the leading and trailing side of the spiral arm .
therefore , to extract the effect of the spiral arm , the distance information would be crucial .
taking into account stellar populations and the dust extinction may allow meaningful constraints on the distance of the observed stars ( e.g. * ? ? ? * ) and help to identify the location of the resonances by examining the apogee data at the different galactic longitudes .
we observed our n - body / sph simulation of a milky way - sized disc galaxy in a similar way to how we can observe the milky way , with particular interest in the stellar and gas motion around the spiral arm .
as our first study , we have focused on the three lines of sight on the disc plane , @xmath0 and ( 150,0 ) deg , and analysed the rotation and radial velocity of stars and gas as a function of the distance from our assumed location of the observer .
similarly to the recent literature based on n - body simulations , our simulated galaxy shows a co - rotating spiral arm , i.e. the spiral arm is rotating with the same speed as the circular velocity , at the lines of sight selected .
we show that the stars around the spiral arm show a large variation in both radial and rotational velocities owing to the co - rotating spiral arm .
if the spiral arm is indeed a co - rotating spiral arm , we should observe a similar variation of the rotation and radial velocities around the spiral arm at every galactocentric radius .
an accurate measurement of the distance , proper motion and radial velocity of the stars is required , and the _ gaia _ data will be a critical test for the co - rotating spiral arm .
we show that the stars behind the spiral arm always gain angular momentum , while the stars at the front of the spiral arm lose angular momentum .
the stars tend to rotate slower ( faster ) behind ( at the front of ) the spiral arm and move outward ( inward ) . because of the epicycle motion of the stars
, we also see the stars with high ( low ) rotation velocity behind ( at the front of ) the spiral arms .
we find that these stars came from the outer ( inner ) region and decelerated ( accelerated ) at the front of ( behind ) the arm .
then , they are passed by ( passed ) the spiral arm and are observed at their peri - centre ( apo - centre ) phase .
these are consistent with @xcite , and indicate a variety of orbits owing to the co - rotating spiral arm .
these variety of orbits are likely to be closely related to the formation and disruption of the spiral arm .
however , we need further investigation to reach a firm conclusion . still , this study indicates that numerical simulations provides useful tests in the study of the nature of spiral arms .
we have also analysed the rotation and radial velocity of the cold gas component .
this is much simpler than the stars .
we found a clear trend in the gas component which rotates slower ( faster ) and moves outward ( inward ) behind ( at the front of ) the spiral arm .
we have compared the results with the observed data of the maser sources from @xcite .
interestingly , the data show similar trend in @xmath71 around the expected position of the spiral arm around @xmath1 deg .
more data from the accurate astrometric measurement of the maser sources will provide additional constraints on the nature of the spiral arm . although we have observed three lines of sight for one snapshot of the numerical simulation , the observed gas and stellar motions are naturally expected features from our previous studies @xcite , and therefore the observed trend should be common features in the co - rotating spiral arms in n - body / sph simulations .
we have analyzed several snapshots at different timesteps of the simulation , and confirmed that similar kinematic trends are always observed .
encouraged by the success of this study , we are currently working to make a quantitative prediction for the upcoming _ gaia _ data for the co - rotating spiral arm , by taking into account the stellar population , dust extinction and expected _ gaia _ errors by improving the methods in @xcite and @xcite .
the entire topic will be further illuminated by another set of theoretical models , e.g. the analytical model of the stellar kinematics from the spiral arm , and also the numerical simulations with a fixed spiral arm potential and a constant pattern speed including both gas and stars ( e.g. * ? ? ?
* ; * ? ? ?
we thank an anonymous referee for their constructive comments and helpful suggestions which have improved the manuscript .
we also thank jo bovy and the anonymous referee for their critical comments on our over - interpretation of fig .
[ vlosl90-fig ] in the first version of our draft .
the calculations for this paper were performed on the ucl facility legion , the iridis hpc facility provided by the centre for innovation and the dirac facilities ( through the cosmos consortium ) jointly funded by stfc and the large facilities capital fund of bis .
we also acknowledge prace for awarding us access to resource cartesius based in netherlands at surfsara .
funding for sdss - iii has been provided by the alfred p. sloan foundation , the participating institutions , the national science foundation , and the u.s .
department of energy office of science .
the sdss - iii web site is http://www.sdss3.org/. sdss - iii is managed by the astrophysical research consortium for the participating institutions of the sdss - iii collaboration including the university of arizona , the brazilian participation group , brookhaven national laboratory , carnegie mellon university , university of florida , the french participation group , the german participation group , harvard university , the instituto de astrofisica de canarias , the michigan state / notre dame / jina participation group , johns hopkins university , lawrence berkeley national laboratory , max planck institute for astrophysics , max planck institute for extraterrestrial physics , new mexico state university , new york university , ohio state university , pennsylvania state university , university of portsmouth , princeton university , the spanish participation group , university of tokyo , university of utah , vanderbilt university , university of virginia , university of washington , and yale university .
f. , chiappini c. , santiago b. x. , rocha - pinto h. j. , girardi l. , da costa l. n. , maia m. a. g. , steinmetz m. , minchev i. , schultheis m. , boeche c. , miglio a. , montalbn j. , schneider d. p. , beers t. c. , cunha k. , allende prieto c. , balbinot e. , bizyaev d. , brauer d. e. , brinkmann j. , frinchaboy p. m. , garca prez a. e. , hayden m. r. , hearty f. r. , holtzman j. , johnson j. a. , kinemuchi k. , majewski s. r. , malanushenko e. , malanushenko v. , nidever d. l. , oconnell r. w. , pan k. , robin a. c. , schiavon r. p. , shetrone m. , skrutskie m. f. , smith v. v. , stassun k. , zasowski g. , 2014 , a&a , 564 , a115 r. a. , churchwell e. , babler b. l. , indebetouw r. , meade m. r. , whitney b. a. , watson c. , wolfire m. g. , wolff m. j. , ignace r. , bania t. m. , bracker s. , clemens d. p. , chomiuk l. , cohen m. , dickey j. m. , jackson j. m. , kobulnicky h. a. , mercer e. p. , mathis j. s. , stolovy s. r. , uzpen b. , 2005 , apj , 630 , l149 j. , allende prieto c. , beers t. c. , bizyaev d. , da costa l. n. , cunha k. , ebelke g. l. , eisenstein d. j. , frinchaboy p. m. , garca prez a. e. , girardi l. , hearty f. r. , hogg d. w. , holtzman j. , maia m. a. g. , majewski s. r. , malanushenko e. , malanushenko v. , mszros s. , nidever d. l. , oconnell r. w. , odonnell c. , oravetz a. , pan k. , rocha - pinto h. j. , schiavon r. p. , schneider d. p. , schultheis m. , skrutskie m. , smith v. v. , weinberg d. h. , wilson j. c. , zasowski g. , 2012 , apj , 759 , 131 j. , nidever d. l. , rix h .- w .
, girardi l. , zasowski g. , chojnowski s. d. , holtzman j. , epstein c. , frinchaboy p. m. , hayden m. r. , rodrigues t. s. , majewski s. r. , johnson j. a. , pinsonneault m. h. , stello d. , allende prieto c. , andrews b. , basu s. , beers t. c. , bizyaev d. , burton a. , chaplin w. j. , cunha k. , elsworth y. , garca r. a. , garca - herandez d. a. , garca prez a. e. , hearty f. r. , hekker s. , kallinger t. , kinemuchi k. , koesterke l. , mszros s. , mosser b. , oconnell r. w. , oravetz d. , pan k. , robin a. c. , schiavon r. p. , schneider d. p. , schultheis m. , serenelli a. , shetrone m. , silva aguirre v. , simmons a. , skrutskie m. , smith v. v. , stassun k. , weinberg d. h. , wilson j. c. , zamora o. , 2014 , arxiv e - prints , 1405.1032 m. j. , menten k. m. , brunthaler a. , zheng x. w. , dame t. m. , xu y. , wu y. , zhang b. , sanna a. , sato m. , hachisuka k. , choi y. k. , immer k. , moscadelli l. , rygl k. l. j. , bartkiewicz a. , 2014 , apj , 783 , 130 a. c. , luri x. , reyl c. , isasi y. , grux e. , blanco - cuaresma s. , arenou f. , babusiaux c. , belcheva m. , drimmel r. , jordi c. , krone - martins a. , masana e. , mauduit j. c. , mignard f. , mowlavi n. , rocca - volmerange b. , sartoretti p. , slezak e. , sozzetti a. , 2012 , a&a , 543 , a100
a. , famaey b. , binney j. , burnett b. , faure c. , minchev i. , williams m. e. k. , bienaym o. , bland - hawthorn j. , boeche c. , gibson b. k. , grebel e. k. , helmi a. , just a. , munari u. , navarro j. f. , parker q. a. , reid w. a. , seabroke g. , siviero a. , steinmetz m. , zwitter t. , 2012 , mnras , 425 , 2335 | we have observed a snapshot of our n - body / smoothed particle hydrodynamics simulation of a milky way - sized barred spiral galaxy in a similar way to how we can observe the milky way .
the simulated galaxy shows a co - rotating spiral arm , i.e. the spiral arm rotates with the same speed as the circular speed .
we observed the rotation and radial velocities of the gas and stars as a function of the distance from our assumed location of the observer at the three lines of sight on the disc plane , @xmath0 and ( 150,0 ) deg .
we find that the stars tend to rotate slower ( faster ) behind ( at the front of ) the spiral arm and move outward ( inward ) , because of the radial migration .
however , because of their epicycle motion , we see a variation of rotation and radial velocities around the spiral arm . on the other hand
, the cold gas component shows a clearer trend of rotating slower ( faster ) and moving outward ( inward ) behind ( at the front of ) the spiral arm , because of the radial migration .
we have compared the results with the velocity of the maser sources from @xcite , and find that the observational data show a similar trend in the rotation velocity around the expected position of the spiral arm at @xmath1 deg .
we also compared the distribution of the radial velocity from the local standard of the rest , @xmath2 , with the apogee data at @xmath3 deg as an example .
[ firstpage ] galaxy : disc galaxy : kinematics and dynamics methods : numerical |
the standard model ( sm ) like higgs boson has discovered at the lhc experiment , and its mass is obtained by the atlas and cms combined experiments as @xmath2 @xcite . in the sm ,
this value of the higgs mass leads the vacuum to be unstable , that is , the higgs quartic coupling @xmath3 becomes zero below , but close to , the planck scale ( @xmath4gev ) @xcite .
the vacuum stability problem may suggest appearance of new physics below the planck scale .
if new particles appear beyond the sm , runnings of the gauge couplings become larger compared to the sm case .
then , the vacuum can be stable up to the planck scale , since runnings of @xmath3 also becomes larger .
furthermore , the change of the running gauge couplings can realize the gauge coupling unification ( gcu ) at a high energy scale ( see ref .
@xcite for a general discussion ) .
in addition to the vacuum instability , the hierarchy problem of the higgs mass would be appeared in the sm .
in fact , the quadratic divergence of the higgs mass term can be always multiplicatively subtracted at some energy scale from the bardeen s argument @xcite .
since a renormalization group equation ( rge ) of the higgs mass term is proportional to itself in the sm , once it is zero at a high energy scale , e.g. , the planck scale , it continues to be zero at lower energy scales .
however , if there are heavy particles coupling with the higgs doublet , the higgs mass term gives logarithmic correction as @xmath5 , where @xmath6 and @xmath7 are mass of the heavy particle and renormalization scale , respectively .
therefore , the hierarchy problem can be solved if no large intermediate scales exist between the electroweak and the planck scales . in this paper
, we will consider that the planck scale as a fundamental scale , in which the quadratic divergences are assumed to be completely removed out . to solve the hierarchy problem
, we do not consider any intermediate scale except for the tev scale . under this context , we will investigate possibilities for the realization of the gcu at the planck scale , and discuss the vacuum stability .
this paper is based on our previous work @xcite .
we investigate possibilities for the realization of gcu at some high energy scales . solving the rges
, we can see the behavior of the gauge couplings in an arbitrary high energy scale .
the one - loop level rges of the gauge couplings @xmath8 are given by @xmath9 , where @xmath10 , 2 , and 3 , and the coefficients of @xmath11 , @xmath12 , and @xmath13 gauge couplings are given by ( @xmath14 , @xmath15 , @xmath16)=(41/6 , @xmath17 , @xmath18 ) in the sm . for simplicity ,
we only consider a gut normalization factor of 3/5 as in @xmath19 gut , i.e. @xmath20 . once particle contents in the model are fixed , contributions to @xmath21 are systematically calculated as in table [ bi ] @xcite . in this table , fermions included vector - like to avoid the gauge anomalies .
.contributions to @xmath21 from fermions ( left ) and scalars ( right ) .
@xmath11 hypercharge `` @xmath22 '' can take different values for different representations , and an electric charge is given by @xmath23 with isospin @xmath24 .
@xmath25 is given by @xmath26 . [
cols="^,^ " , ] [ combination_f ] .
the solid and dashed lines correspond to the extended sm and the sm , respectively .
three vertical lines represent 0.5tev , 3tev , 10tev , @xmath27 , and @xmath28 , respectively.,title="fig:",width=264 ] .
the solid and dashed lines correspond to the extended sm and the sm , respectively .
three vertical lines represent 0.5tev , 3tev , 10tev , @xmath27 , and @xmath28 , respectively.,title="fig:",width=264 ] in fig.[running_f ] we show runnings of the gauge , the top yukawa , and the higgs quartic couplings , which correspond to the first one of table [ combination_f ] . here
, we assume extra fermions do not strongly couple with the higgs doublet , and then significantly change runnings of the top yukawa and the higgs quartic couplings .
since @xmath29-functions of the gauge couplings change several times , our previous naive analyses are modified , that the gcu can be realized around @xmath28 by extra fermions satisfying @xmath30 .
to realize the gcu , all the gauge couplings are large compared to those in the sm , which lead the smaller @xmath31 .
then , @xmath3 becomes larger and remains in positive value up to the planck scale , since both the smaller @xmath31 and the larger @xmath32 make @xmath33 become larger .
thus , when the gcu is realized at the planck scale , we expect the vacuum can become stable .
we have investigated possibilities of the gcu at the planck scale in the extended sm which includes extra particles around the tev scale . we have found that the gcu at the planck scale can be realized when extra particles include some relevant scalars , while it can not be realized ( up to one - loop level ) when all extra particles are fermions and their masses are the same . on the other hand , when extra fermions have different masses , the gcu around @xmath28 can be realized . in this case , the vacuum can become stable because of the change of the running gauge couplings .
we thank n. haba , r. takahashi and h. ishida for useful discussion and fruitful collaborations .
the works of y.y . is supported by research fellowships of the japan society for the promotion of science for young scientists , grants no . 26@xmath342428 .
s. chatrchyan _ et al . _
[ cms collaboration ] , phys .
b * 718 * , 348 ( 2012 ) [ arxiv:1210.1797 [ hep - ex ] ] .
s. chatrchyan _ et al . _
[ cms collaboration ] , phys .
b * 729 * , 149 ( 2014 ) [ arxiv:1311.7667 [ hep - ex ] ] .
g. aad _ et al . _ [ atlas collaboration ] , jhep * 11 * , 104 ( 2014 ) [ arxiv:1409.5500 [ hep - ex ] ] . | to solve the hierarchy problem of the higgs mass , it may be suggested that there are no an intermediate scale up to the planck scale except for the tev scale . for this motivation ,
we investigate possibilities of gauge coupling unification ( gcu ) at the planck scale ( @xmath0 ) by adding extra particles with the tev scale mass into the standard model .
we find that the gcu at the planck scale can be realized by extra particles including some relevant scalars , while it can not be realized only by extra fermions with the same masses .
on the other hand , when extra fermions have different masses , the gcu can be realized around @xmath1 . by this extension ,
the vacuum can become stable up to the planck scale . |
0 by `` active '' mass one means the mass considered as source of a gravitational field .
this concept can be distinguished from that of the `` passive '' mass measuring the response to a gravitational field and of an `` inertial '' mass describing resistance against acceleration by any force , gravitational or other .
these distinctions are useful for considering non - newtonian or non - einsteinian theories of gravity , or to assess which aspects of mass are tested in experiments . in newton s and einstein s theories these distinctions are not necessary , in agreement with available physical and astronomical evidence and post - newtonian approximations .
joseph louis lagrange in his 1773 `` memoir on the secular equation of the moon ''
@xcite introduced the function @xmath0 through the equation @xmath1 he recognized that it is easier to first compute @xmath0 and then to get the accelerations by differentiation than to calculate the accelerations directly .
the function @xmath2 denotes here the density of the active mass . the active mass
is then given by @xmath3 lagrange did not use vector notation and did not exhibit newton s gravitational constant @xmath4 because astronomers set it equal to one as we do in this paper .
we recognize now lagrange s function @xmath0 as the negative of the gravitational potential @xmath5 .
the use of the negative of the potential was customary before the principle of the conservation of energy began to dominate physics and astronomy .
lagrange did not supply a name for his function @xmath0 .
carl friedrich gauss called it the potential in his 1839 paper on `` general theorems about forces of attraction and repulsion depending on the inverse square of the distance '' @xcite .
gauss had not been aware of george green s 1828 article `` an essay on the application of mathematical analysis on the theory of electricity and magnetism '' @xcite that had named @xmath0 the `` potential function '' .
green had published his paper privately and had rarely referred to it in his later papers .
it was only several years after green s death in 1841 that william thomson ( the later lord kelvin ) discovered green s paper and arranged for the publication of its results .
the acceleration vector @xmath6 for a massive particle was given by lagrange in an inertial system by @xmath7 this law may be read as saying that the inertial mass may be identified with the passive gravitational mass . in his 1782 `` theory of the attraction of spheroids and the figure of the earth '' @xcite pierre simon laplace introduced for the function @xmath0 the equation @xmath8 that is now named the laplace equation .
laplace did not write it , as we just did , in cartesian coordinates @xmath9 but used spherical polar coordinates .
he published the form in cartesian coordinates in his `` memoir about the theory of saturn s ring '' @xcite in 1787 .
it was only 26 years later that simon - denis poisson pointed out in his `` remarks about an equation that occurs in the theory of attraction of spheroids '' that the laplace equation does not hold within the substance of the attracting body @xcite and has there to be replaced by @xmath10 this equation is now known as the poisson equation . a proof was first given by gauss in 1839 @xcite . if we write the equations in terms of the potential @xmath5 we have @xmath11 the relation between the gravitational potential and the density of the active mass in an inertial system of newton s theory .
if @xmath5 is required to vanish at infinity , ( [ i.6 ] ) implies ( [ i.1 ] ) .
0 in general relativity @xcite the ten components @xmath12 of the metric tensor @xmath13 replace the potential of lagrange
. moreover , not just the mass - density , but all ten components of the energy - momentum - stress tensor @xmath14 become contributing sources to the gravitational field . for neutral matter @xmath14 is given by @xmath15 with energy density @xmath16 , pressure tensor @xmath17 and four - velocity @xmath18 subject to the normalisation @xmath19 and @xmath20 for a perfect fluid @xmath21 instead of poisson s equation relating the gravitational potential to the active mass density @xmath16 we have now the einstein field equations @xmath22 connecting the @xmath12 and their derivatives up to the second order to the components of @xmath14 .
the riemann tensor is defined through the interchange of order in the second covariant derivatives of a covariant vector field @xmath23 @xmath24 and the ricci tensor by @xmath25 the ricci scalar @xmath26 is obtained through contraction from the ricci tensor @xmath27 . here
we use units where the speed of light `` c '' is put equal to 1 . for a static gravitational field in which the matter is at rest we are able to find an analog of lagrange s potential and its poisson equation . in section [ s : a1 ]
we give an invariant derivation of the necessary developments . here
we give just the results .
a static gravitational field can be described by the metric @xmath28 with no @xmath29 terms .
latin indices run here from 1 to 3 .
the components of the metric tensor do not depend on the time coordinate @xmath30 and the four- velocity @xmath18 of the matter is given by @xmath31 we have then for the static potential @xmath32 with @xmath33 @xmath34 the spatial metric differs from the flat metric only by terms of order @xmath5 , on the left - hand side of the relativistic poisson equation we have the laplace operator on the spaces t = const . for the function
@xmath32 that we now identify as the static relativistic potential .
this expression in riemannian geometry was derived by eugenio beltrami in his 1868 paper `` on the general theory of differential parameters '' and became known as the second beltrami parameter @xcite . the cosmological term in ( [ gr.6 ] )
can be put on the right - hand side of the equation as @xmath35 interpreted as an energy - momentum tensor of a perfect fluid @xmath36u_\mu u_\nu -p(\lambda)g_{\mu\nu}\ ] ] we have @xmath37 the contribution to the active mass density ( `` dark energy '' ) becomes @xmath38 on the right - hand side we have poisson s @xmath39 term .
apart from einstein s lambda term that acts now if positive as a negative mass density contributing to the active mass , we have two changes in comparison with poisson s equation .
one is the factor @xmath32 on the right hand side of the equation . since the discovery of energy conservation the potential had been defined as the work to displace a unit mass ( charge ) from infinity and it was usually normalized to vanish at spatial infinity . in einstein
s theory of gravitation the component @xmath40 for a finite mass distribution in a static gravitational field is usually taken as @xmath41 or 1 at infinity .
the reason is that now one includes the rest energy of a particle as part of the potential energy and assumes that the metric of an isolated system becomes euclidean at large distances .
we have then for @xmath32 at large distances @xmath42 from the center of mass @xmath43 where @xmath44 is the total active mass .
thus , for weak fields the factor @xmath32 in the relativistic poisson equation differs from 1 only slightly .
the most surprising correction to the relativistic poisson equation is the term @xmath45 that is actually @xmath46 .
this term was first noted explicitly as a consequence of einstein s field equations by tullio levi - civit in 1917 in his series of papers on `` einstein s static '' @xcite .
it is this term that we shall discuss in other sections of this paper . finally , @xmath47 in the formula for the active mass is the proper volume element .
the justification for identifying @xmath32 with the relativistic global potential rests on its definition as the specific potential energy .
we shall show in section [ s : eps ] that the test particle of unit mass resting at @xmath48 has potential energy ( apart from its rest - energy ) @xmath49 the poisson equation in newtonian gravity is written down in a global inertial system .
the closest analog to such a system for einstein s static gravitational fields ( without cosmological constant ) is based on the coordinates adapted to the time - like hypersurface - orthogonal killing vector .
it also defines a rest system through the divergence - free time - like unit vectors orthogonal to the hypersurfaces t = const . reaching out to spatial infinity where the space - time metric is assumed to become minkowskian .
these coordinates , it has to be stressed , are _ not _ local inertial systems .
the acceleration of a particle at rest is given by @xmath50
0 richard chase tolman in his 1930 paper `` on the use of the energy - momentum principle in general relativity ''
@xcite derived a formula for the total energy of a fluid sphere in a quasi - static state for einstein s theory of gravitation .
he wrote with a proper volume element @xmath51 @xmath52 suppressing the cosmological term , where @xmath16 is the relativistic energy density . in statistical mechanics of an ideal gas consisting of particles with mass @xmath53 , and momentum @xmath54 , velocity @xmath55 , and number density @xmath56 ,
the pressure @xmath57 is given by daniel bernoulli s formula @xcite from his 1738 `` hydrodynamics , commentaries about forces and motions in fluids '' @xmath58 where the bar indicates averaging . in ( [ tp.2 ] ) @xmath57 is the kinetic contribution to the pressure . in general
@xmath57 , or @xmath59 contains contributions from short - range interactions , which have to be added to ( [ tp.2 ] ) .
attractive interactions contribute negative @xmath57 .
this formula , as written above , holds also for a relativistic ideal gas as discussed by franz juettner in his paper on `` maxwell s law of velocity distribution in the relative theory''@xcite . in the high energy limit of photons with energy @xmath60
this gives @xmath61 since tolman was particularly interested in the `` gravitational mass of disordered radiation , '' the @xmath45-term in the active mass density caught his attention .
this was no longer one of the tiny effects of general relativity .
tolman s observation led to the following paradox : while matter at rest in a spherical container of negligible matter might exhibit a total mass @xmath44 in the far field of the container , transformation of the mass inside the container into disordered radiation would then double the total mass violating the conservation of @xmath44 . after the discovery of the positron in 1932 and its annihilation with the electron such complete transformations of matter into radiation
were no longer impossible .
0 the authors of this paper were members of a seminar on relativity that regularly met at hamburg university in the 1950 s .
when we learned about the @xmath45-term in the poisson equation the following test was suggested : since nucleons move in atomic nuclei with about two tenths of the speed of light , the @xmath45-term might significantly contribute to the active mass density of all nuclei except hydrogen and the neutron .
a simple calculation for the pressure in an ideal fermi gas of nucleons at zero temperature ( see section [ s : b1 ] ) gave a pressure contribution to the active mass density of 4.3% for nuclear matter .
a ball of hydrogen should , therefore , have an active mass about 4% smaller then a ball of lead of the same inertial and passive gravitational mass that could be checked by weighing them on scales .
however , the only way such an effect might be seen in the laboratory was by the cavendish experiment for the determination of the gravitational constant where the active mass came into play . while it would be forbidding to work with a ball of ultra - cold solid hydrogen one might consider a material with a high hydrogen content that was solid at room temperature like polyethylene of formula @xmath62 or lithium hydride of formula @xmath63 . for lithium hydride
we have @xmath64 and for polyethylene @xmath65 if one were to use balls of these materials for a determination of the gravitational constant through the cavendish experiment one should thus get lower values for the gravitational constant @xmath4 . in the fifties
the value of @xmath4 was uncertain by about 0.1% @xcite and so the effect might just be measurable .
when one of us ( els ) talked in 1958 to robert dicke of princeton university about a possible experimental test dicke had his doubts whether it could be done since machining homogeneous spheres in those materials might be forbidding .
however , a year later it became clear that there should be no different value for @xmath4 with balls of hydrogen .
dieter brill from princeton who had joined the hamburg seminar alerted us to a paper in the physical review by charles misner and peter putnam @xcite about active mass ( about peter putnam who died in 1987 see john wheeler s memoir @xcite . )
misner and putnam showed , assuming gravity to be negligible , that the 3p - term for a gas in a container was canceled by negative contributions to the mass from the stresses in the walls of the container that kept the gas together .
it had not been clear to us that negative _ surface _ contributions to the energy would exactly cancel the positive _ 3p - volume _ contribution to the total mass when we had the model of a bubble in mind .
0 we imagined a gas of constant density @xmath16 and pressure @xmath57 enclosed in a spherical two - dimensional shell of radius @xmath42 with surface mass density @xmath66 and surface tension @xmath67 . is filled with a gas of pressure @xmath57 .
the bubble is kept in equilibrium by a surface tension @xmath66 with dimension force by length . ]
the surface tension should be just strong enough for keeping the bubble in equilibrium .
the gravitational binding energy of the bubble was supposed to be negligible compared to its mass . for finding the relation of surface tension to pressure for equilibrium we imagined a plane cut through the bubble removing the southern hemisphere . to keep the northern hemisphere in equilibrium one now had to balance the upward pressure over the equatorial disc of area @xmath68 against the surface tension pulling down along the equator over the length @xmath69 .
this gives the relation ) is removed and replaced by its forces on the upper hemisphere . ]
@xmath70 the total active mass @xmath44 would then be obtained by the surface contribution @xmath71 and by the volume contribution @xmath72 @xmath73 leaving us with half the volume contribution of the @xmath45-term to the active mass .
something was wrong .
it was only last summer in a nostalgic moment when we talked about this problem again that we saw the solution : if the active mass density of a 3-dimensional distribution has to be complemented by a @xmath45-term , then that of a 2-dimensional shell needs a @xmath74-term and a 1-dimensional disk a @xmath57-term ( corresponding to the trace of a 2- or 1-dimensional isotropic stress tensor ) if they were stresses instead of pressures they would come in with a negative sign .
now all was clear : the surface contribution to the active mass of the bubble was @xmath75 and we obtain now instead ( [ sb.2 ] ) @xmath76 this simple remark settled also the case of the active mass of a circular disk .
0 we consider a circular disk of radius @xmath42 with mass density @xmath66 and pressure @xmath57 carries a surface pressure @xmath57 with dimension force / length .
the pressure is balanced by the tension @xmath77 with dimension force along its perimeter . ] the disk is kept in equilibrium by a one - dimensional string around its circumference of linear mass density @xmath78 and stress @xmath77 . to find the relation between the pressure @xmath57 and the stress @xmath77 we imagine a linear cut through the center of the disk removing the lower half ) has been removed and replaced by the forces acting on the upper part . ] the pressure over the diameter @xmath79 must now be balanced against the stress @xmath77 at the left and the right end of the semicircle .
this gives @xmath80 the active mass @xmath44 of the disk is then obtained by taking the active mass density @xmath81 over the area @xmath68 and adding the active mass density @xmath82 of the bounding string along the circumference @xmath69 .
this gives @xmath83 the promised result .
0 in newtonian theory there was always the understanding , because of `` actio = reactio '' , that the active and passive masses were equal .
if this were not the case a system of two passive unit masses but different active masses would show an acceleration of its center of mass violating newton s law _ actio = reactio_. in einstein s theory of gravitation the equality of active and passive
mass is not so obvious since einstein s cosmological constant can be seen as giving rise to self - acceleration for the center of mass of a double star . while we have no reason to doubt the equality of the three kinds of masses we believe that tests involving the active mass are certainly desirable
this is especially true for situations where the gravitational binding energy significantly contributes to the mass . in general relativity
the fundamental variables are `` local '' ones , such as @xmath84 , @xmath85 , @xmath16 , @xmath86 , @xmath57 neither the active nor the passive or the inertial mass of an ( extended ) body have been exactly defined so far , these concepts belong to perturbative theories in general relativity .
0 a stationary gravitational field is characterized by the existence of a time - like killing field that generates an infinitesimal transformation that leaves the metric unchanged .
the field is called static if the killing vector field is hypersurface - orthogonal ; this means the covariant vector is a product of a gradient by a scalar function .
a killing vector field @xmath23 fulfills the equation @xmath87 if we write down equation ( [ gr.5 ] ) three times with cyclic permutation of the indices @xmath88 @xmath89 @xmath90 adding the first two equations and subtracting the third , we have because of ( [ a1.1 ] ) and the cyclic symmetry of the riemann tensor @xmath91 that @xmath92 these equations are known as the integrability conditions for the killing field .
if we define @xmath93 we obtain from ( [ a1.6 ] ) by anti - symmetrization and contraction maxwell s equations for a field tensor @xmath94 and a four - current @xmath95 @xmath96 since @xmath97 the killing vector for a stationary gravitational field plays the rle of an electromagnetic four - potential .
0 we choose coordinates in such a way that @xmath98 that can be done for any contravariant vector field in a finite region .
this specialization still allows gauge transformation with an arbitrary function @xmath99 @xmath100 the killing equation ( [ a1.1 ] ) can be written @xmath101 this gives with our normalization ( [ a2.1 ] ) @xmath102 that is independence of the coordinate @xmath103 . since we requested that the killing vector @xmath104 be time - like we have that @xmath103 is a distinguished time coordinate with @xmath105 using now the condition that the killing vector is hypersurface - orthogonal ( static field ) @xmath106 we have after contraction with @xmath104 , @xmath107 this gives after division by @xmath108 @xmath109 we have , therefore , that @xmath110 is a gradient of a scalar function @xmath111 @xmath112 we can write @xmath113 comparison with ( [ a2.2 ] ) shows that we can choose the gauge transformation @xmath114 such that @xmath115 by dropping the bar on @xmath116 we have then for the metric the form ( [ gr.8 ] ) @xmath117 the purely spatial coordinate transformations are still free .
the time - like hypersurface - orthogonal killing vector is , in general , unique up to a constant factor .
the square of this factor multiplies @xmath40 .
the length of the killing vector is @xmath32 .
a state of rest is then desribed by a time - like unit vector @xmath118 we study now the second set of maxwell s equations ( [ a1.8 ] ) in adapted coordinates .
we have because of time independence @xmath119 since by ( [ a1.7 ] ) @xmath120 the only non - vanishing covariant components are @xmath121 entered into ( [ a3.2 ] ) this gives @xmath122 with @xmath123 we obtain @xmath124 the four - current @xmath95 from ( [ a1.7 ] ) becomes with the einstein field equations ( [ gr.4 ] ) @xmath125 or @xmath126 here @xmath127 is the trace ( contraction ) of the tensor @xmath128 .
the second set of maxwell s equations state that @xmath129 saying that the density of momentum and energy flux vanish in this static situation .
calling the energy density @xmath16 and the trace of the pressure tensor @xmath45 @xmath130 we have @xmath131 with ( [ a3.7 ] ) this gives the relativistic poisson equation for a static gravitational field . @xmath132
0 a particle of constant mass @xmath53 and four - velocity @xmath133 moving on a geodesic in a static gravitational field obeys @xmath134 the energy integral for unit mass is given by @xmath135 since @xmath136 the first term on the right hand side vanishes because of the geodesic equation ( [ eps.1 ] ) while the second term is zero due to the killing equation ( [ a1.1 ] ) . in terms of the components of the local rest frame we have
@xmath137 where @xmath138 is the local velocity in terms of the speed of light .
0 a particle at rest is characterised by its four - velocity @xmath139 the acceleration of the particle is given by @xmath140 since @xmath40 is independent of time the first term on the right - hand side vanishes .
we obtain then with ( [ apr.1 ] ) @xmath141 or @xmath142
0 the pressure for such a gas at zero temperature was derived by enrico fermi in his 1926 paper `` on the quantization of the ideal mono - atomic gas '' @xcite . for @xmath143 identical particles in volume
@xmath0 we have a number density @xmath144 and according to ( [ tp.1 ] ) @xmath145 for a non - relativistic motion this gives with a mass @xmath53 @xmath146 where @xmath147 is their average kinetic energy .
for an ideal fermi gas at zero temperature in a cubic box of side length @xmath148 for a statistical weight @xmath149 we can fit @xmath149 particles in a cell of phase space of size @xmath150 while fermi took @xmath151 , the case of general @xmath149 was discussed by wolfgang pauli in his 1927 paper `` on gas degeneration and paramagnetism '' @xcite .
we have thus for @xmath152 particles @xmath153 and for the total number @xmath143 of particles in terms of the maximum fermi momentum @xmath154 @xmath155 we obtain the average kinetic energy by averaging the expression @xmath156 and obtain @xmath157 this gives for the pressure from ( [ b1.2 ] ) @xmath158 the ratio of @xmath46 to the mass density @xmath16 becomes @xmath159 the equation ( [ b1.5 ] ) allows us to express the fermi momentum in terms of the particle density and we obtain thus @xmath160 substituting this expression into ( [ b1.8 ] ) gives @xmath161 or in terms of dirac s @xmath162 @xmath163
0 werner karl heisenberg considered an atomic nucleus as a fermi gas of nucleons in his 1933 paper @xcite `` on the structure of atomic nuclei iii '' .
we want to calculate the term @xmath164 for an ideal fermi gas of nucleons at zero temperature.the interactions among nucleons that keep them confined are simply taken into account by a potential well in which they move freely . for a rough estimate we disregard the difference between protons and neutrons and take the statistical weight @xmath149 to be @xmath165 j.l .
lagrange , `` sur lquation sculaire de la lune '' .
mmoires de lacadmie royale des sciences , _
paris_. 1773 .
oevres de lagrange ( ed .
serret ) vol . 6 , 349 .
_ paris _ : gautier- villars ( publisher ) .
gauss , allgemeine lehrstze in beziehung auf die i m verkehrtem verhltnisse des quadrats der entfernung wirkenden anziehungs- und abstossungskrfte .
gauss werke , vol . 5 , _ gttingen .
_ 1877 , page 200 .
g. green , `` an essay on the application of mathematical analysis on the theory of electricity and magnetism . '' _ nottingham _ 1828 .
reprinted in `` the mathematical papers of the late george green '' cambridge 1871 .
reprinted again by chelsea , _ bronx , ny . _ 1981 .
laplace , `` thorie des attractions des sphrodes et de la figure des plantes '' .
mmoires de lacadmie royale des sciences .
_ 1782 , 135 .
oevres compltes vol .
_ paris _ : gauthier - villars ( publisher ) .
laplace , `` mmoire sur la thorie de lanneau de saturne '' .
mmoires de lacadmie royale des sciences .
_ 1787 , 252 .
oevres compltes vol .
_ paris _ : gauthier - villars ( publisher ) .
e. beltrami , `` sulla teoria generale di parametri differenziali '' .
memorie dellaccademia delle scienze dellistituto di bologna serie ii , tomo viii(1896 p. 557 - 590 . )
opere matematiche vol . 2 , 98 .
_ milano _ : hoepli .
t. levi - civit , `` statica einsteiniana '' rendiconti della r. accademia dei lincei , ser .
opere matematiche , vol.4 , 72 .
bologna 1960 .
see also `` the absolute differential calculus '' .
dover , 1977 .
d. bernoulli , `` hydrodynamica , sive de viribus et motibus fluidorum commentarii .
x. '' _ argentorati(strasbourg ) _ 1738 .
hydrodynamics by daniel bernoulli ( translated from the latin by thomas carmody and helmut kobus .
preface by hunter rouse ) .
dover publications .
_ new york_. e. fermi , zur quantelung des idealen einatomigen gases .
zeitschrift fr physik vol .
36 , 902 - 912 . reprinted in collected papers , vol .
i , 186 - 195 ; edited by e. amaldi et al .
_ chicago_. the university of chicago press . 1962 .
jrgen ehlers + max - planck - institut fr gravitationsphysik , golm , germany + istvn ozsvth # + department of mathematics , the university of texas at dallas , + richardson , texas 75083 - 0688 + engelbert
l. schucking + department of physics , new york university , 4 washington place , + new york 10003 + # electronic mail : [email protected] | after a historical introduction to poisson s equation for newtonian gravity , its analog for static gravitational fields in einstein s theory is reviewed .
it appears that the pressure contribution to the active mass density in einstein s theory might also be noticeable at the newtonian level .
a form of its surprising appearance , first noticed by richard chase tolman , was discussed half a century ago in the hamburg relativity seminar and is resolved here . |
geometric frustration of interacting spin systems is the driving force of a variety of fascinating phenomena in low - dimensional magnetism @xcite . in this context
the term _ frustration _ describes a situation where in the ground state of a classical spin system not all interactions can be saturated simultaneously . a typical picture for such a situation is a triangle of antiferromagnetically coupled spins , where classically the spins are not in the typical up - down - up configuration , but assume a ground state that is characterized by a relative angle of @xmath4 between neighboring spins .
this special classical ground state characterizes among others several frustrated spin systems which are built of corner - sharing triangles , among them the giant keplerate molecule \{@xmath0 } which is a perfect icosidodecahedron @xcite and the kagome lattice antiferromagnet .
the pyrochlore lattice , which consists of corner - sharing tetrahedra and thus has a different structure , nevertheless shares several important properties with the above systems .
research in this field is naturally focused on the low - energy , i.e. low - temperature , low - field behavior .
a key observation is that the quantum spin systems possess many or even infinitely many singlet states below the first triplet state and the classical counterpart systems display a non - trivial ground state degeneracy @xcite .
another important observation concerns a plateau in the magnetization curve for @xmath5 , which for the systems made of corner - sharing triangles is at @xmath6 , see e.g. @xcite , whereas for the pyrochlore it is at @xmath7 , see e.g. @xcite .
in this article we focus on special properties of these systems which arise at low temperatures but high magnetic fields
. we will show that it is possible to construct exact many body states which are product states of independent one - magnon states .
these states become ground states in high magnetic fields . in a wider perspective such an arrangement of independent single - particle objects can be understood as condensation of bosons @xcite .
the linear scaling of the minimal total energy with the number of these objects explains their unusual high - field behavior which is expressed in magnetization jumps @xcite , non - zero @xmath3-entropy @xcite , and an enhanced magnetocaloric effect @xcite .
in the following we assume that the spin systems under consideration are modeled by an isotropic heisenberg hamiltonian augmented with a zeeman term , i.e. , @xmath8 @xmath9 are the individual spin operators at sites @xmath10 and @xmath11 is the @xmath12-component of the total spin .
@xmath13 are the matrix elements of the symmetric coupling matrix .
we will consider only antiferromagnetic couplings .
@xmath3-magnetization curve of the kagome chain , fig .
[ f-1-a ] with @xmath14 .
the magnetization jump of @xmath15 is marked by an arrow . ]
@xmath3-magnetization curve of the kagome chain , fig .
[ f-1-a ] with @xmath14 .
the magnetization jump of @xmath15 is marked by an arrow . ]
one of the early unexpected results was that for the icosidodecahedron ( e.g. \{@xmath0 } ) the minimal energies @xmath16 in each hilbert subspace of @xmath17 of total magnetic quantum number @xmath18 scale linearly with @xmath18 close to the saturation @xcite .
figure [ f-1-a ] shows as another example the minimal energy levels of a special kagome chain ( introduced in @xcite ) where the seven highest levels are on a straight line . in an applied magnetic field
this leads to a simultaneous crossing of the lowest zeeman levels at the saturation field , which gives rise to an unusual magnetization jump , that for the kagome chain is shown in fig .
[ f-1-b ] .
independent localized magnons on the kagome lattice antiferromagnet .
@xmath19 independent localized magnons can be placed on the kagome lattice .
] independent localized magnons on the kagome lattice antiferromagnet .
@xmath19 independent localized magnons can be placed on the kagome lattice . ]
it turns out that the linear dependence of @xmath16 on @xmath18 can be understood in terms of localized independent magnons .
figure [ f-2-a ] shows the structure of a localized magnon on a part of a model lattice as realized for instance in the kagome chain of fig .
[ f-1-a ] .
the one - magnon state is given by @xmath20 it can be shown that this state is an eigenstate of lowest energy in one - magnon space @xmath21 . on an extended lattice such as the kagome chain , fig .
[ f-1-a ] , or the two - dimensional kagome lattice , fig .
[ f-2-b ] , many of these objects can be placed in such a way that they do not interact @xcite .
also in this case it can be shown that these states are eigenstates of lowest energy in their respective @xmath1-magnon space .
an immediate consequence of the independence of the localized one - magnon states and thus of the linear dependence of the minimal energies @xmath16 on @xmath18 is the magnetization jump at the saturation field @xmath22 as well as the high degeneracy of levels at this field value . in an infinite lattice such as the kagome or pyrochlore lattice both quantities are macroscopic @xcite , although it turns out that it is rather involved to evaluate the exact degeneracy at the saturation field due to possible relations between the @xmath1-magnon product states @xcite .
the ground - state degeneracy at @xmath22 is related to a finite @xmath3 entropy per site , i.e. @xmath23 ( for the kagome chain one obtains @xmath24 @xcite ) . in the context of magnetocalorics
such a residual entropy gives rise to large adiabatic cooling rates @xmath25 in the vicinity of the saturation field @xcite . in a real compound ,
e.g. \{@xmath0 } , the perfect degeneracy at @xmath22 would be lifted by residual , e.g. dipolar or dzyaloshinskii - moriya interactions .
consequently , the magnetization jump would be smeared out , compare ref .
@xcite for \{@xmath0}. nevertheless , the low - lying density of states remains large at @xmath22 , which would still be clearly visible in magnetocaloric investigations .
it is clear that the energetically degenerate independent localized one - magnon states on translationally symmetric lattices can be superimposed to form eigenstates of the translation operator with the same energy which leads to a flat band .
figure [ f-3-a ] shows the three energy bands of the kagome chain introduced in fig .
[ f-1-a ] .
we observe that one third of all one - magnon states form independent magnons or equally well belong to the flat band .
therefore a jump of one third of the saturation magnetization occurs at @xmath22 for @xmath26 , see fig .
[ f-1-b ] .
the emergence of flat bands has been already noted in the context of _ line graphs _ and _ flat - band ferromagnetism _ , see e.g. @xcite .
a connection can be made by replacing the heisenberg model by a hubbard model @xcite .
the roles of the magnetic exchange @xmath27 and the applied magnetic field are then played by the hopping integral @xmath28 the chemical potential @xmath29 , respectively .
if the flat band is the lowest band , then non - interacting localized excitations can be constructed for such fermionic systems in a manner very similar to the heisenberg model , and these fermionic systems exhibit similar thermodynamic properties , but now as function of temperature and chemical potential .
an example is given by fig .
[ f-3-b ] where the isentropes for free spinless fermions @xcite on the kagome chain ( fig .
[ f-1-a ] ) are displayed . at @xmath30
the ground state is degenerate with an extensive @xmath3 entropy @xmath31 .
the value @xmath32 corresponds to the saturation field @xmath22 of the heisenberg model , and a very similar behavior would be observed in its vicinity , e.g. the slopes of the isentropes correspond to the adiabatic cooling rates @xcite .
however , at lower values of @xmath29 ( @xmath33 ) there are qualitative differences : free spinless fermions ( whose dispersion is essentially given in fig . [ f-3-a ] ) have no band gaps , while the magnetization curve of the @xmath26 heisenberg model presumably has plateaus ( corresponding to gaps ) at least at one third and two thirds of the saturation magnetization ( see fig . [ f-1-b ] ) .
it may be remarked here that flat - bands on partial line graphs do not need to be ground - state bands @xcite .
lines of constant entropy for free spinless fermions on the kagome chain .
the value of the entropy per site @xmath34 is indicated next to each line . ]
lines of constant entropy for free spinless fermions on the kagome chain .
the value of the entropy per site @xmath34 is indicated next to each line . ] | strongly frustrated antiferromagnets such as the magnetic molecule \{@xmath0 } , the kagome , or the pyrochlore lattice exhibit a variety of fascinating properties like low - lying singlets , magnetization plateaus as well as magnetization jumps . during recent years exact
many - body eigenstates could be constructed for several of these spin systems .
these states become ground states in high magnetic fields , and they also lead to exotic behavior . a key concept to an understanding of these properties is provided by independent localized magnons .
the energy eigenvalue of these @xmath1-magnon states scales linearly with the number @xmath1 of independent magnons and thus with the total magnetic quantum number @xmath2 . in an applied field
this results in a giant magnetization jump which constitutes a new macroscopic quantum effect .
it will be demonstrated that this behavior is accompanied by a massive degeneracy , an extensive @xmath3-entropy , and thus a large magnetocaloric effect at the saturation field .
the connection to flat band ferromagnetism will be outlined . |
recently , deep neural networks ( dnns ) have been successfully applied to solve complicated problems in pattern recognition , computer vision , and speech recognition , cf . @xcite . despite its great success ,
designing and training a deep neural network is still among the greatest challenges in the field , cf .
@xcite . in this work ,
we focus on the study of the feedforward neural networks ( fnns ) .
one major reason for the difficulty in designing and training
an fnn is that their performance is highly dependent on various factors in a very complicated way .
besides the determinative factor being the architecture of an fnn @xcite , the work in @xcite demonstrates the impact of different activation functions to the performance of an fnn .
moreover , the choice of error functions is also shown to be influential , cf .
@xcite . even with a well designed fnn architecture , training the fnn efficiently is as challenging as the design of the network .
the most popular fnn training algorithm is the well known backpropagation ( bp ) algorithm , cf . @xcite .
although the bp algorithm shares a great convenience of being very simple , early works argue that problems with the bp algorithm is essentially its nature of being a gradient descent algorithm , cf . @xcite .
since an overall cost function for training an fnn is often in a large scale and highly non - convex , the bp algorithm suffers from two major drawbacks , namely , ( i ) existence of undesired local minima , and ( ii ) slow convergence speed . one major instrument in such a study is via an error / loss surface analysis , cf .
however , even for a very simple problem , such as the classic xor problem , the error surface analysis is very complicated and the results are still hard to conclude .
specifically , works in @xcite demonstrate such a difficulty of characterising the error surface of an fnn . on the other hand
, although the bp algorithm is suspected to be sensitive to initialisations , e.g. @xcite , recent results reported in @xcite suggest that modern fnn learning algorithms can overcome the problem of local optima quite conveniently .
such an observation could be explained by the pioneering works in @xcite , which developed some conditions on the structure of fnns to eliminate undesired local minima .
unfortunately , these local minima free conditions seem to be still limited for an application in analysing dnns . in order to deal with slow convergence speed of the bp algorithm ,
various modified bp algorithms have been developed , such as momentum based bp algorithm @xcite , conjugate gradient algorithm @xcite , and bfgs algorithm @xcite .
although newton s method is always of great interest for training an fnn , a complete implementation of newton s fnn algorithm is often computationally prohibitive .
instead , certain approximations of the hessian matrix have been already proposed in the community to deal with the issue , such as a diagonal approximation structure @xcite , or a block diagonal approximation structure @xcite . however , without a true evaluation of the hessian , performance of these heuristic approximations is hardly convincing .
although existing works @xcite have characterised the hessian matrix by applying the partial derivatives , unfortunately these results fail to provide further information of the hessian matrix , due to the limitations of partial derivatives . therefore , in this work , we propose to employ the techniques from multivariable differential calculus to analyse fnns from a perspective of smooth optimisation .
the paper is organised as follows . in section [ sec:02 ] , by revisiting the classic bp algorithm , we introduce the concepts and techniques from multivariable differential calculus that are crucial in our analysis and development . in section [ sec:03 ] ,
we characterise the critical point conditions of the general fnn learning cost function , and derive several conditions or design principles for eliminating local minima . with a further analysis of the hessian matrix of the cost function at the global minima , we characterise the isolatedness properties of the global minima in section [ sec:04 ] .
an approximate newton s algorithm is developed by leveraging a simple structure of the true hessian at the global minima . in section [ sec:05 ] , we study the classic xor problem .
some classic results on xor networks are reviewed in the developed framework of this paper .
section [ sec:06 ] studies a challenging benchmark of the four - region classification problem with numerical experiments .
finally , conclusions and outlooks are given in section [ sec:07 ] .
by revisiting the classic bp algorithm in this section , we aim to introduce the mathematical notations and concepts used in our development and analysis .
we refer to @xcite for further readings on the technique .
let us denote by @xmath0 the number of layers in an fnn structure , and by @xmath1 the number of processing units in the @xmath2-th layer with @xmath3 .
specifically , by letting @xmath4 , we refer to the input layer .
let @xmath5 be an elementary activation function , which is often constructed to be non - constant , continuous , and monotonically increasing .
popular choices of activation function include the sigmoid function or @xmath6 function .
we denote by @xmath7 and @xmath8 the first derivative and second derivative of the activation function . for the @xmath9-th unit in an fnn architecture , referring to the @xmath10-th unit in the @xmath2-th layer , we can define each unit mapping as @xmath11 where @xmath12 denotes the output from the @xmath13-th layer , @xmath14 is the weight vector associated with the @xmath9-th unit , and @xmath15 is a scalar bias associated with each unit . for the sake of convenience in technical representation of our analysis , we drop the scalar bias @xmath16 in our further development . straightforwardly , we can define the @xmath2-th layer evaluation mapping by stacking all unit mappings in the layer as @xmath17^{\top}\!\!\ ! , \end{split}\ ] ] with @xmath18 \in \mathbb{r}^{n_{l-1 } \times n_{l}}$ ] being the @xmath2-th weight matrix .
specifically , let us denote by @xmath19 the input , then we define @xmath20 iteratively .
by composing all the layer - wise mappings together , we end up with the overall network mapping as @xmath21 where @xmath22 . for a specific learning task ,
one often deploys a suitable error function @xmath23 . in this work ,
we only consider the supervised learning setting with a dataset with @xmath24 samples , denoted by @xmath25 . finally ,
for a given input @xmath26 , we define the overall _ fnn learning cost function _ as @xmath27 it is trivial to see that , if the error function @xmath28 is differentiable in @xmath29 , then the cost function @xmath30 is differentiable in both fnn weights @xmath31 and the input data @xmath32 . for the supervised learning setting , i.e. , given dataset @xmath25 , the overall fnn learning cost function can be formed as @xmath33 for the convenience of presentations in the rest of the paper
, we will drop the summation , only focus on the analysis of sampled cost function @xmath30 . in order to develop a gradient descent algorithm to minimise the cost function @xmath30 .
we need to define the derivative of all processing units , i.e. , for a given @xmath9-th unit , we have @xmath34 and similarly , for the @xmath2-th layer @xmath35^{\top}\!\!\!. \end{split}\ ] ] for simplicity , we denote @xmath36 . due to its layer - wise structure of the fnn , by fixing a parameter @xmath37
, we can define a truncated learning cost function , referred to as the @xmath2-th _ tail error function _ , as @xmath38 by applying the chain rule of multivariable derivative , we compute the first derivation of @xmath30 with respect to @xmath37 in direction @xmath39 as @xmath40 where @xmath41 and @xmath42 refer to the derivative of @xmath43 with respect to the first and second argument .
explicitly , the first derivative of @xmath43 , i.e. , @xmath44 @xmath45 in direction @xmath46 , is computed as @xmath47 where the operator @xmath48 puts a vector into a diagonal matrix form , and the derivative of @xmath43 with respect to the second parameter in direction @xmath49 as @xmath50 let us denote @xmath51 for all @xmath3 . then , it is straightforward to derive the gradient of @xmath30 with respect the @xmath2-th weight matrix @xmath52 as @xmath53 which is a rank - one matrix update . by exploring the layer - wise structure of the fnn ,
the corresponding vector @xmath54 , referred to as the _ @xmath2-th error vector _ , can be computed iteratively backwards from the output layer @xmath0 , i.e. @xmath55 with @xmath56 . with such a backward mechanism in computing the gradient @xmath57 , we recover the classic bp algorithm , as presented in algorithm [ algo:01 ] .
such a strategy can be easily adapted to both the batch learning and the online learning setting . in
what follows , we will focus on an analysis in the batch learning setting .
with the gradient computed explicitly as in , the critical points of the fnn learning cost @xmath30 are characterised by simply setting it to zero , namely , @xmath58 . more explicitly , for a given sample @xmath59
, we have the following equation system in @xmath22 as @xmath60 obviously , it is hardly possible to completely characterise all critical points of @xmath30 .
however , the major goal for training an fnn is to minimise the error function @xmath61 . in other words , it is more sensible to characterise the gradient of the error function @xmath28 with respect to the output of the fnn , i.e. , @xmath62 , instead of the gradient of the overall fnn learning cost @xmath63 . in order to guarantee uniqueness of optimal solutions in minimising the error function @xmath61 ,
it is reasonable to ensure that the function @xmath28 is strictly convex .
with such a construction , training an fnn has no problems in ending at local minima of the error function .
[ prin : error ] the error function @xmath23 needs to be strictly convex .
nows , we rewrite the critical point conditions of the overall learning cost @xmath30 in eq . , in terms of @xmath64 .
similar to the iterative construction of the error vector @xmath54 as in eq .
, we construct a sequence of matrices as , for all @xmath65 , @xmath66 with @xmath67 . for a given @xmath68-th sample @xmath59 , the equation system as in eq .
can be rewritten as @xmath69 \nabla_{e}(\phi_{l}^{(i ) } ) = 0,\ ] ] where @xmath70 denotes the kronecker product of matrices .
let us denote the total number of variables in an fnn by @xmath71 then , by collecting all equation systems from all the samples @xmath72 , we end up with the following gigantic equation system in @xmath73 as @xmath74 } _ { = : \mathbf{p } \in \mathbb{r}^{n_{net } \times ( t \cdot n_{l } ) } } \ !
\left[\!\!\ ! \begin{array}{c }
\nabla_{e}(\phi_{l}^{(1 ) } ) \\ \vdots \\
\nabla_{e}(\phi_{l}^{(t ) } ) \end{array}\!\!\!\right ] \!= 0.\ ] ] obviously , the trivial solution of @xmath75 for all @xmath72 is the only solution , if and only if the rank of matrix @xmath76 is equal to @xmath77 .
thus , we just proved the following theorem .
let the error function @xmath23 be strictly convex .
then the fnn learning cost function @xmath78 is free of local minima , if and only if , @xmath79 given the number of rows of @xmath76 being @xmath80 , we conclude our second principle of fnn design as follows .
[ prin : number ] the total number of variables in an fnn , i.e. , @xmath80 , needs to be greater than or equal to @xmath77 .
now , let us have a closer look at the matrix @xmath76 to identify potential strategies to ensure the rank condition to be satisfied as much as possible . by collecting the entries of @xmath81 s and @xmath82 s
, we construct the following two partitioned matrices @xmath83\ ] ] and @xmath84.\ ] ] the two matrices certainly share the same partition structure by construction .
then the matrix @xmath76 is simply the tracy - singh product of @xmath85 and @xmath86 as defined in eq .
, i.e. , the pairwise kronecker product for each pair of partitions in @xmath85 and @xmath86 , denoted by @xmath87 by the properties of the tracy - singh product @xcite , it follows directly @xmath88 since the rank of @xmath86 is determined by the data and the fnn architecture , from a design perspective , it is therefore more sensible to guarantee the rank of @xmath85 . recall the construction of partitions @xmath81 as in eq .
, i.e. , products of @xmath89 s and @xmath37 s , it is thus reasonable to make these matrix terms as of full rank as possible .
we can then conclude the following two principles of designing an fnn .
[ prin : weight ] weight matrix @xmath90 needs to be of full rank for all @xmath3 .
[ prin : act ] the derivative of activation function @xmath91 needs to be nonzero for all @xmath92 , i.e. , @xmath93 .
it is trivial to verify that most of popular activation functions , such as sigmoid , @xmath6 , logistic , softplus , and softsign , are monotonically increasing , hence satisfy principle [ prin : act ] . in the rest of this section
, we consider an fnn architecture with only one hidden layer , i.e. , @xmath94 , and derive a classic result in the literature .
we prove the following well known , but still arguable , results as shown in @xcite .
[ prop:01 ] given an fnn architecture with only one hidden layer , including a dummy unit .
assume that the training dataset consists of @xmath24 unique samples .
then an fnn learning cost function is free of local minima , if the following four conditions are fulfilled : 1 .
the error function @xmath28 is strictly convex ; + 2 .
there are @xmath95 units in the hidden layer ; + 3 . with any weight matrix @xmath96 in the hidden layer having no identical columns , @xmath24 unique samples produce a basis in the output space of the hidden layer ; + 4
. the derivatives of activation functions employed in the output layer are nonzero . without loss of generality
, we can bring the dummy hidden unit to be real , which represents the associated scaler bias and receives a constant input of value @xmath97 from the input layer , i.e. , @xmath98 .
then the critical point conditions at the output layer @xmath99 reads as @xmath100 } _ { : = p_{2 } \in \mathbb{r}^{(t \cdot n_{2 } ) \times ( t \cdot n_{2 } ) } } \left[\!\!\ ! \begin{array}{c } \nabla_{e}(\phi_{2}^{(1 ) } ) \\ \vdots \\
\nabla_{e}(\phi_{2}^{(t ) } ) \end{array}\!\!\!\right ] \!= 0,\ ] ] with @xmath101 equations ,
i.e. , @xmath102 is a square matrix .
it is straightforward to conclude that condition ( 4 ) ensures @xmath103\big ) = n_{2}.\ ] ] with condition ( 3 ) , the matrix @xmath102 is of full rank .
hence , the result follows .
certainly , fnn architectures used in real applications often consist of more hidden layers .
nevertheless , if we fix the @xmath2-th hidden layer with @xmath104 , then locally the previous @xmath13-th layer can be treated as the input layer , while the following @xmath105-th layer can be considered as the output layer .
the learning cost function in the output @xmath106 is nothing but the @xmath105-th tail error function as defined in eq . .
the results in preposition [ prop:01 ] can be used for choosing the number of units in the hidden layers . for a given @xmath2-th hidden layer with @xmath104
, we assume that there are @xmath107 unique patterns after forwarding thought the previous @xmath108 layers
. then we need at least @xmath109 units in this layer to ensure a potential exemption of local minima .
this design principle is very heuristic , since knowledge of the number of patterns in the hidden layers are generally unknown .
when pooling is used in a feedforward convolutive neural network , it might be interesting to see the role of this principle .
we treat this aspect as one potential research direction in the future .
the information from the hessian matrix is critically important for designing efficient numerical algorithms .
specifically , definiteness of the hessian matrix is the indicator to the isolatedness of the critical points , which will affect significantly the convergence speed of the algorithms .
the hessian form of the fnn learning cost function @xmath30 is a bilinear operator @xmath110 , computed by the second derivative of @xmath30 .
its computation is rather straightforward but very tedious . for the sake of readability
, we only present one component of the second derivative with respect to two specific indices @xmath10 and @xmath2 , i.e. , @xmath111 where @xmath112 is the diagonal matrix with the second derivative of the activation functions on the diagonal .
let @xmath113 be the desired output of the fnn , reached by an global minima @xmath114 . then all gradients @xmath115 of the error function are equal to zero . as a sequel , the second summand in the last equation in eq .
therefore , the hessian @xmath116 evaluated at @xmath114 is given as @xmath117 where @xmath118 is the hessian form of the error function @xmath28 . for a given sample @xmath59 ,
we construct the following two identically partitioned matrices , i.e. , @xmath119 \mathsf{h}_{e}(\phi_{l}^{*(i ) } ) \left[\!\!\ ! \begin{array}{c }
\psi_{l}^{(i ) } \\ \vdots \\
\psi_{1}^{(i ) } \end{array}\!\!\!\right]^{\top}\ ] ] and @xmath120 \left[\!\!\ ! \begin{array}{c }
\phi_{l-1}^{*(i ) } \\ \vdots \\
\phi_{0}^{*(i ) } \end{array}\!\!\!\right]^{\top}.\ ] ] then , the hessian of @xmath30 at a global minima @xmath114 can be represented in a matrix form as @xmath121 if the activation functions used at the output layer follow principle [ prin : act ] , then the rank of matrix @xmath122 as defined in eq .
is determined by the rank of the hessian @xmath123 .
if the error function @xmath28 is chosen according to principle [ prin : error ] , then it needs to be further assume have a non - degenerate hessian , i.e. , @xmath28 being a morse function , cf .
@xcite consequently , we have @xmath124 hence , we introduce the next design principle on the choice of error function , in addition to principle [ prin : error ]
. 1.a[strong choice of error function ] [ prin : error2 ] the error function @xmath23 needs to be strictly convex and morse , i.e. , the hessian @xmath125 is non - degenerate for all @xmath126 .
since both matrices @xmath127 and @xmath122 are positive semi - definite , the hessian matrix @xmath128 is simply a sum of @xmath24 rank-@xmath129 positive semi - definite matrices .
we can conclude the following result . given an fnn architecture satisfying principle [ prin : error ] , [ prin : number ] , [ prin : act ] , [ prin : error2 ] , and a training dataset consisting of @xmath24 unique samples .
if a global minimum @xmath114 of the fnn learning cost is reachable , then the rank of the hessian matrix of @xmath30 is bounded from above by @xmath130 it is important to notice that the hessian @xmath128 is neither diagonal nor block diagonal , which demotivates the existing approximate strategies of the hessian in @xcite . with our explicit characterisation of the hessian at global minima , we propose to approximate the hessian of @xmath30 at arbitrary point @xmath31 with the structure as shown in eq . .
the corresponding approximate newton s algorithm is presented in algorithm [ algo : newtonbp ] .
it is important to notice that , in step 6 , a scaled identity matrix with @xmath131 is added to the approximate hessian in order to ensure the existence of the inverse .
moreover , the computation of the newton update in step 6 can be efficiently carried out by implementing a block cholesky decomposition .
the xor problem is a classic and most well known demonstrative scenario for fnn training .
two specific network structures , i.e. , the @xmath132-@xmath132-@xmath97 xor network and the @xmath132-@xmath133-@xmath97 xor network , have been extensively studied in the literature . in this section
, we apply our developed analysis to investigate some classic results on the xor networks .
we confine ourselves to the network structure with only one hidden layer , while the structure of the input and output layers are fixed .
namely , we have @xmath94 , @xmath134 , and @xmath135 .
all processing units are chosen to be the sigmoid function .
we define the input patterns as @xmath136 = \left [ \begin{matrix } 0 & 0 & 1 & 1 \\ 0 & 1 & 0 & 1 \end{matrix } \right ] \in \mathbb{r}^{2 \times 4},\ ] ] and correspondingly , the output @xmath137 = [ 0 , 1 , 1 , 0 ] \in \mathbb{r}^{1 \times 4}.\ ] ] there are @xmath138 unique samples
. then the fnn learning cost function is defined as @xmath139 where @xmath140 with @xmath141 and @xmath142 .
firstly , let us have a look at the @xmath132-@xmath132-@xmath97 xor network , i.e. , @xmath143 .
we then investigate the rank of the following matrix @xmath144 \psi_{1}^{(1 ) } \otimes \phi_{0}^{(1 ) } & \!\!\ldots\!\ ! & \psi_{1}^{(4 ) } \otimes \phi_{0}^{(4 ) } \end{array}\!\!\!\right].\ ] ] by considering the biases associated with each unit , the total number of fnn variables is @xmath145 since @xmath135 , it is easily to see that @xmath146 we can construct two matrices @xmath147 \psi_{1}^{(i ) } \end{array}\!\!\!\right ] \mathsf{h}_{e}(\phi_{2}^{*(i ) } ) \left[\!\!\ ! \begin{array}{c }
\psi_{2}^{(i ) } \\[1 mm ] \psi_{1}^{(i ) } \end{array}\!\!\!\right]^{\top}\ ] ] and @xmath148 \phi_{0}^{*(i ) } \end{array}\!\!\!\right ] \left[\!\!\ ! \begin{array}{c }
\phi_{1}^{*(i ) } \\[1 mm ] \phi_{0}^{*(i ) } \end{array}\!\!\!\right]^{\top}.\ ] ] since the hessian @xmath149 is a scalar , it is trivial to see that the hessian of @xmath150 is a sum of four rank - one matrices , i.e. , @xmath151 , hence degenerate .
we then conclude the classic results as in @xcite .
let a @xmath132-@xmath132-@xmath97 xor network architecture satisfy all four principles [ prin : error]-[prin : act ] .
then the network has no local minima and global minima have a degenerate hessian .
similar analysis on the @xmath132-@xmath133-@xmath97 xor network leads to the following results . let a @xmath132-@xmath133-@xmath97 xor network architecture satisfy three principles [ prin : error ] , [ prin : number ] , [ prin : act ] .
then the network has no local minima and global minima have a degenerate hessian .
it is clear then that any xor network with an architecture @xmath132-@xmath152-@xmath97 for @xmath153 has only degenerate global minima .
in our numerical experiments , we investigate performance of our proposed approximate newton s algorithm on the four regions classification benchmark , as originally proposed in @xcite . in @xmath154 around the origin , we have a square area @xmath155 , and three concentric circles with their radiuses being @xmath97 , @xmath132 , and @xmath133 .
four regions / classes are interlocked , nonconvex , as shown in figure [ fig:1a ] .
we draw randomly @xmath156 samples in the box for training , and specify the corresponding output to be the @xmath68-th basis vector in @xmath157 .
we deploy an fnn architecture with two hidden layers , i.e. , @xmath158 . in both hidden layer
, there are @xmath159 units each .
hence , we have @xmath134 , @xmath160 , and @xmath161 .
all activation functions are chosen to be sigmoid . finally , the standard least squares error is used as the error function , which is morse and strictly convex .
our experiments indicates that the approximate newton s algorithm is capable of handling vanishing gradient , and reach much lower value in the error function .
more specifically , the convergence speed of the proposed approximate newton s fnn algorithm is much faster than the classic bp algorithm , shown in figure [ fig : converge ] . by random sampling
, the learned fnn produces quite accurate results as shown in figure [ fig:1c ] , expect the points on close to the boundary . in order to copy with such an issue , we construct a framed dataset , consisting @xmath162 samples on the two sides along the boundary with the other @xmath163 random samples shown in figure [ fig:1b ] .
we test the same test sample with the fnn learned from the framed dataset , leading to a much more accurate classification shown in figure [ fig:1d ] .
in this work , we developed a smooth optimisation perspective on the challenge of designing and training an fnn architecture in the supervised learning setting . by characterising the critical point conditions of the overall learning cost function , we derive some mild conditions to ensure an fnn to be free of local minima .
classic results on this matter in both xor scenario and general setting are also reviewed .
our analysis also identifies the hessian structure of the cost function at the global minima , which leads to an approximate newton s fnn algorithm .
x. glorot and y. bengio , `` understanding the difficulty of training deep feedforward neural networks , '' in _ proceedings of the thirteenth international conference on artificial intelligence and statistics ( aistats-10 ) _ , vol . 9 , 2010 , pp . 249256 .
s. sun , w. chen , l. wang , x. liu , and t .- y .
liu , `` on the depth of deep neural networks : a theoretical view , '' in _ proceedings of the @xmath164 aaai conference on artificial intelligence _
, 2016 , pp . 20662072 .
t. falas and a. g. stafylopatis , `` the impact of the error function selection in neural network - based classifiers , '' in _ proceedings of the international joint conference on neural networks ( ijcnn ) _ , vol . 3 , 1999 , pp .
17991804 .
r. s. sutton , `` two problems with backpropagation and other steepest - descent learning procedures for networks , '' in _ proceedings of the @xmath165-th annual conference of the cognitive science society _ , 1986 , pp .
823831 .
i. g. sprinkhuizen - kuyper and e. j. w. boers , `` the local minima of the error surface of the 2 - 2 - 1 xor network , '' _ annals of mathematics and artificial intelligence _ , vol . 25 , no . 1 ,
pp . 107136 , 1999 .
a. choromanska , m. henaff , m. mathieu , g. b. arous , and y. lecun , `` the loss surfaces of multilayer networks , '' in _ proceedings of the @xmath166 international conference on artificial intelligence and statistics ( aistats ) _ , 2015 , pp .
192204 .
i. j. goodfellow , o. vinyals , and a. m. saxe , `` qualitatively characterizing neural network optimization problems , '' published at the @xmath167 international conference on learning representations ( iclr ) .
arxiv:1412.6544 . , 2015 . | despite the recent great success of deep neural networks in various applications , designing and training a deep neural network is still among the greatest challenges in the field . in this work
, we present a smooth optimisation perspective on designing and training multilayer feedforward neural networks ( fnns ) in the supervised learning setting . by characterising the critical point conditions of an fnn based optimisation problem
, we identify the conditions to eliminate local optima of the corresponding cost function . moreover ,
by studying the hessian structure of the cost function at the global minima , we develop an approximate newton fnn algorithm , which is capable of alleviating the vanishing gradient problem .
finally , our results are numerically verified on two classic benchmarks , i.e. , the xor problem and the four region classification problem .
designing and training feedforward neural networks : a smooth optimisation perspective + hao shen + e - mail : [email protected] + department of electrical and computer engineering + technische universitt mnchen , germany + * * + forward neural networks ( fnns ) , smooth optimisation , critical point analysis , hessian matrix , approximate newton s method . |
cosmological tests using high - redshift type ia supernovae ( sne ia ) rule out an @xmath2=1 universe @xcite and provide evidence for an acceleration of the cosmic expansion @xcite , attributed to a nonzero cosmological constant ( @xmath3 ) .
the statistical weight of these results is high , so the current focus of our high - z supernova search team @xcite is on systematic effects , most of which can be convincingly tested using sne at @xmath40.8 , regardless of their precise nature .
we expect dimming from systematic effects to grow larger with increasing redshift . however , the effect of a nonzero @xmath3 on apparent magnitude can decrease with increasing redshift , due to the different redshift dependence of @xmath5 and @xmath6 ( see figure 4 of filippenko & riess 2000 ) . obtaining data for sne at @xmath40.8
is therefore critical for testing systematic bias versus real cosmological effects and the presence of a nonzero @xmath6 .
in addition to photometry of high-@xmath0 sne ia needed to determine a luminosity distance for the hubble diagram , we obtain spectra of our sne to measure the redshift , determine the supernova type , and look for spectral differences between the low-@xmath0 and high-@xmath0 sn ia populations that may indicate evolution of sne with time . here
we present spectra of two high-@xmath0 sne ia and address uncertainties in using the spectra to determine redshifts and the sn type .
details of the observing program , finder charts , and light curves will be presented by tonry et al .
sn 1999ff @xcite was observed with lris ( oke et al .
1995 ) on the keck - ii 10-m telescope on 1999 november 8 ut with 0.7@xmath7 seeing .
use of a 1@xmath7 slit and the 150 line mm@xmath8 grating resulted in a spectral resolution of @xmath120 .
the total exposure time was 2200 s , divided into three exposures dithered 10@xmath7 along the slit to reduce the effects of fringing .
the slit was oriented at a position angle ( pa ) of 156@xmath9 to include the galaxy nucleus .
sn 1999fv @xcite was observed with the same setup on 1999 november 10 ut .
the seeing was 0.9@xmath7 , and a slit width of 0.7@xmath7 resulted in a spectral resolution of @xmath116 .
the slit was oriented to pa=252@xmath9 to include a star .
the total exposure time was 6000 s , taken in four dithered frames .
sn 1999fv was also observed on 1999 november 11 ut with the 400 line mm@xmath8 grating , with a 1.0@xmath7 slit at the same pa and 1@xmath7 seeing , in four dithered exposures of 1000 s each .
standard ccd processing and optimal spectral extraction were done with iraf , and we used our own idl routines to calibrate the wavelengths and flux of the spectra and to correct for telluric absorption bands .
the sn 1999fv data from both nights were combined , after binning to the same resolution . only the first two exposures from november 11 were used , as the last two were overwhelmingly dominated by sky noise .
rest - frame spectra of sn 1999ff and sn 1999fv are shown in figure 1 .
the observed wavelength range for sn 1999ff is 40507700 , corresponding to 27855290 in the rest frame , while for sn 1999fv the observed wavelength range is 62009150 , corresponding to 28304180 in the rest frame .
sn 1999ff is located @xmath12@xmath10 from the center of a large , bright elliptical galaxy with @xmath11=19.1 mag , whereas the sn was @xmath123 mag . to eliminate contamination from the host galaxy , we scaled and subtracted the host galaxy spectrum from the sn spectrum .
the high-@xmath0 spectra have been smoothed with a savitsky - golay filter @xcite , a polynomial fit that preserves line features better than boxcar smoothing .
the spectra of sn 1999ff and sn 1999fv were heavily smoothed with a 70 and 180 filter width , respectively , in the rest frame . for sn 1999ff
we estimate the age based on its spectral features @xcite to be 5 @xmath122 days .
photometric data of sn 1999ff will be presented by @xcite and will determine the true epoch of maximum brightness .
sn 1999fv was significantly fainter and we were unable to get a reliable spectral age for it , in part due to the lack of available early - time spectra of low-@xmath0 sne ia that extend blueward of 4000 .
the redshift of the host galaxy of sn 1999ff is @xmath0=0.455 @xmath120.001 , as determined by balmer absorption lines in its spectrum , and we adopt this redshift for the sn .
the host galaxy of sn 1999fv was not visible in our images , and to calculate the redshift of this sn we cross - correlated the high-@xmath0 spectrum with several low-@xmath0 sn ia spectra near maximum light ( as defined by the @xmath13-band maximum ) , the ages of which are determined from their light curves .
correlating against a sample of 31 spectra with ages between @xmath148 days to + 5 days relative to maximum , the redshift of sn 1999fv is @xmath0=1.17 - 1.22 for the unsmoothed spectrum , somewhat lower than the initial estimate of tonry et al .
( 1999 ; @xmath0=1.23 ) made at the telescope . using the same technique on the spectrum of sn 1999ff , correlating against a sample of 18 spectra with ages between 7 and 3 days , we find @[email protected] , consistent with the host galaxy redshift .
we also include galaxy and m - star spectra among our templates for cross - correlation , which did not fit the high-@xmath0 spectra well .
the redshift of sn 1999fv has the quoted uncertainty ( 1.171.22 ) due to the low signal - to - noise ( s / n ) ratio of the spectrum , and to its high redshift placing key sn features near bright sky lines at observed wavelengths @xmath15 . in order to check the validity of our redshift estimates , we performed blind tests with fake spectra added to the 2d frames .
we redshifted each of 14 low-@xmath0 sn ia spectra by an arbitrary amount between @xmath0=0.5 and @xmath0=1.3 , scaled the flux level down to the signal of sn 1999fv in each 1000 s exposure , and added poisson noise .
we added the spectra to the 2d frames ( which are dominated by sky emission ) , convolving the signal with a gaussian across several pixels perpendicular to the dispersion , imitating seeing effects .
each spectrum was then extracted , and we verified that its s / n ratio as a agreed with that of sn 1999fv . the spectrum was smoothed heavily with a savitsky - golay filter to get an initial redshift estimate by eye , and subsequently cross - correlated with a database of low - z spectra to find a quantitative redshift estimate and uncertainty .
we obtained good results for 12 of the 14 spectra , all of which matched the input redshift to within the error bars . for one spectrum
we were unable to estimate the redshift , as the signal was dominated by sky lines and sn features were not clearly visible .
for another spectrum we obtained two equally likely redshifts , one of which was the correct input redshift . in neither of these two cases
were we led to believe an incorrect redshift .
the redshift range for sn 1999fv of z=1.171.22 is roughly a 95@xmath16-confidence level , based on the tests done .
it is important to note that the difference in luminosity distance at z=1.17 and z=1.22 is 0.11 mag ( 5@xmath16 ) which is much smaller than our distance uncertainty .
this redshift uncertainty therefore has negligible impact on our ability to constrain cosmology .
the largest concern with using sne ia in cosmological tests is a possible photometric difference between the low-@xmath0 and high-@xmath0 samples .
spectral comparisons of the distant and nearby sne could show subtle evolutionary effects , if present .
if the spectra do not show differences , this does not prove that the peak luminosities of the sne are identical , but it does build confidence that the two samples are similar .
spectra of normal , nearby sne ia at the same phase are quite homogeneous @xcite . as a qualitative comparison with our high-@xmath0 spectra , we present in figure 1 spectra of four low-@xmath0 sne ia near the same epoch : sn 1989b at both 7 and 1 days relative to maximum @xcite , sn 1992a at 1 day @xcite , and sn 1981b at maximum @xcite .
the sn 1989b spectra have been dereddened by @xmath17
@xmath18)=0.32 mag @xcite .
the 1 day sn 1989b spectrum is a composite of a ctio spectrum redward of 3300 and an _ international ultraviolet explorer _ ( _ iue _ ) spectrum from 4 days blueward of 3300 .
the sn 1992a spectrum is a composite of a ctio spectrum redward of 3600 and a _ hubble space telescope _ ( @xmath19 ) spectrum blueward of 3600 , taken 5 days after maximum @xcite . for comparison
we also show sne ii , ib , and ic spectra : sn iib 1993j @xcite , sn ib 1998dt at maximum @xcite , and sn ic 1994i at 4 days , dereddened by @xmath17
@xmath18)=0.45 mag ( filippenko et al .
1995 ; see filippenko 1997 for a discussion of sn subclasses ) .
the sn 1998dt spectrum has been smoothed with a 30 savitsky - golay filter .
we also show an _ iue _ spectrum of sn ib 1983n at 4 days from 2800 to 3300 @xcite at the blue end of the sn ib 1998dt spectrum to compare with the high-@xmath0 spectra .
the sn iib 1993j spectrum , taken 16 days past discovery and 3 days relative to maximum , resembles a relatively normal sn ii .
the sn 1999ff spectrum exhibits many of the features seen in the low-@xmath0 sn ia spectra , most prominently the deep h&k absorption ( `` a '' ) , the @xmath204130 dip blueshifted to 4000 ( `` b '' ) , @xmath204555 and/or @xmath204481 ( `` d '' ) , and @xmath205051 ( `` i '' ) .
there are several weaker features seen in all of the low-@xmath0 sn ia spectra that are also apparent in the spectrum of sn 1999ff , including a shoulder blueward of @xmath204555 ( `` c '' ) , @xmath204560 ( `` e '' ) , @xmath204816 ( `` f '' ) , and / blends ( `` g '' and `` h '' ) .
all line identifications ( but not the specific letter labels ) are from jeffery et al .
( 1992 ) , kirshner et al .
( 1993 ) , and mazzali et al .
unfortunately , the sn 1999ff spectrum does not extend to rest wavelength 6150 , where the @xmath206355 feature is prominent in sne ia and much less apparent in sne ic / ib .
this line is the usual diagnostic for distinguishing between sne ia and ib / ic .
sn 1999ff does not , however , resemble the spectra of sn ib 1998dt or sn ic 1994i , in which the and features `` b '' and `` e '' are either very weak or absent . the sn ib and ic spectra also have a broad peak centered at 5250 that is not seen in sn 1999ff or the sn ia spectra . a comparison with the sn iib 1993j spectrum shows that while sn 1993j does have h&k ( `` a '' ) and absorption at 3980 due to blueshifted h@xmath21 which results in a double - peak centered just blueward of 4000 , the resemblance with sn 1999ff redward of 4100
is poor .
very early sn ii spectra are blue and nearly featureless and would be easily classified ; they begin to resemble the sn iib 1993j spectrum shown here a few weeks after discovery . due to its high redshift
, the spectrum of sn 1999fv does not have broad rest - frame coverage and does not extend redward of 4200 .
sn 1999fv does display the h&k feature ( `` a '' ) as well as @xmath204130 ( `` b '' ) .
in the uv region of the spectrum , there are perhaps hints of small rises at 3150 and just blueward of 3000 , as seen in the _ iue _ spectra of sn ia 1992a and sn ic 1983n .
sn 1999fv is bluer than the low-@xmath0 spectra below 3700 . our discovery image for this object does not reveal an obvious host galaxy , and it is unlikely that the blue color is due to host galaxy contamination as sn 1999fv was at least 1 to 1.5 mag brighter than the host .
the sn ib 1998dt and sn ic 1994i spectra provide a worse match to sn 1999fv around 4000 than do the normal sne ia .
while the spectrum of sn 1999fv has a limited rest - frame wavelength range , the presence of the @xmath204130 ( `` b '' ) feature is very important , as it results in a broad double - peak in the spectrum centered at @xmath14000 that is not seen in sn ib / ic spectra @xcite , which show a small dip at this wavelength but lack a strong peak just redward at @xmath14100 .
thus , even for @xmath41 sne with noisy spectra , it is possible to distinguish features that differentiate sne ia from ib and ic events near 4000 .
the sn iib 1993j spectrum has a double - peak centered near 4000 , but the widths of the features blueward of 4200 are significantly narrower than in the sn ia spectra and the sn 1999fv spectrum .
in general , substantial contamination of the high-@xmath0 sne ia sample with sne ic , ib , and ii seems unlikely .
first , typical sne ic , ib , and iib are much fainter than sne ia ; on average , the @xmath13-band peak brightness of sne ib and ic is @xmath11.5 mag lower , while sne ii are @xmath11.8 mag lower ( miller & branch 1990 ; richmond et al .
. however , clocchiatti et al .
( 2000 ) have found a very luminous sn ic , and if such objects are common at high redshift they might contaminate the sample .
second , nearby sne ic , ib , and iib are rarer than sne ia , with a relative rate of 0.25 to 0.5 @xcite , though the relative rates are likely to increase at high redshift . additional photometric data ( light - curve shapes ) are useful in ruling out contamination from sne ic , ib , iib , and ii .
we do not believe that there is confusion with very luminous , peculiar sne ic , also known as `` hypernovae , '' in the high-@xmath0 sample , as their spectra are quite dissimilar from those of our high-@xmath0 objects @xcite . although the analysis techniques of hamuy et al .
( 1996 ) and riess et al .
( 1998 ) , in which light - curve shapes are used to derive luminosity corrections , can be applied to sne ia having a range of luminosities and light - curve shapes , it is of interest to determine whether the proportion of peculiar high-@xmath0 sne ia differs from that at low-@xmath0 . in figure 2
we compare our spectra of sn 1999ff and 1999fv to the two archetypical , very peculiar sne ia : overluminous sn 1991 t at 5 days @xcite and underluminous sn 1991bg at maximum @xcite .
our two high-@xmath0 sne ia do not appear to be these kinds of peculiar events , as seen by the lack of correlation of their spectral features with those of sn 1991 t and sn 1991bg .
sn 1991 t lacks h&k as well as @xmath224130 , 5051 , and exhibits strong lines that are not seen in sn 1999ff and sn 1999fv .
sn 1991bg , on the other hand , shows prominent h&k absorption but lacks the usual lines and has several features not seen in the high-@xmath0 spectra or in normal sne ia . as a further comparison with peculiar sne ia
, we show in figure 2 a spectrum of sn 1999aa taken at 1 day .
this object is similar to sn 1991 t but also exhibits h&k absorption .
several recent low-@xmath0 sne ia of this nature have now been classified @xcite .
while our two high-@xmath0 sne ia are not peculiar in the sense of sn 1991t - like and sn 1991bg - like events , they do share some features with the spectrum of sn 1999aa .
however , they agree best with the normal sne ia shown in figure 1 . we conclude that sn 1999ff is a relatively normal sn ia , and we do not see any compelling evidence for sn 1999fv being a very peculiar sn ia , though wider wavelength coverage and higher s / n ratios would permit a more definitive comparison .
in addition , more uv sn ia spectra are needed to better understand the nature of peculiar and normal sne ia at these wavelengths .
we present spectra of two high-@xmath0 sne ia discovered during our late-1999 high-@xmath0 sn search campaign @xcite .
exposure of a sn at @xmath0 @xmath10.5 yields a high - quality spectrum with detailed features from which a spectral age can be derived .
a 130-min .
exposure of a sn at @xmath0 @xmath11.2 results in a moderate - quality spectrum with prominent , broad features .
the redshift of sn 1999ff is 0.455 , as given by stellar absorption lines in the host galaxy , and from spectral features we estimate the age to be 5 @xmath23 days . for sn 1999fv
we use the spectrum to determine a redshift of 1.171.22 .
we find that the double - peak in the spectrum centered at @xmath14000 and the strength of the @xmath204130 line can be used as a diagnostic for distinguishing sne ia from sne ib and ic when the usual @xmath206355 feature is not available , as is often the case for high-@xmath0 sne . however , moderate - quality spectra are needed to determine the sn type , especially for @xmath41 sne .
we conclude that sn 1999ff is most consistent with normal sne ia , and that sn 1999fv is a sn ia and is not peculiar in the sense of sn 1991 t and sn 1991bg - like objects .
this research was supported by nsf grants ast-9417213 and ast-9987438 to a.v.f .
, by an nsf graduate research fellowship to a.l.c . , and by nasa grants go-7505 and g0 - 8177 to the high - z sn search team from the space telescope science institute , which is operated by the association of universities for research in astronomy , inc . , under nasa contract nas5 - 26555
. the keck observatory is operated by the california institute of technology , the university of california , and nasa .
we are grateful to the keck staff for their assistance with the observations .
we thank ryan chornock and weidong li for useful discussions and the referee for valuable comments .
branch , d. , fisher , a. , & nugent , p. 1993
, , 106 , 6 branch , d. , lacy , c. h. , mccall , m. l. , sutherland , p. g. , uomoto , a. , wheeler , j. c. , & wills , b. j. 1983 , , 270 , 123 cappellaro , e. , evans , r. , & turatto , m. 1999 , , 351 , 459 cappellaro , e. , turatto , m. , tsvetkov , d. yu . , bartunov , o. s. , pollas , c. , evans , r. , & hamuy , m. 1997 , , 322 , 431 cappellaro , e. , turatto , m. , & fernley , j. 1995 , esa - sp 1189 clocchiatti , a. et al .
2000 , , 529 , 661 filippenko , a. v. 1997 , , 35 , 309 filippenko , a. v. , matheson , t. , & ho , l. c. 1993 , , 415 , l103 filippenko , a. v. , & riess , a. g. 2000 , type ia supernovae : observations and theory , ed . j. niemeyer and j. truran ( cambridge : cambridge univ . press ) , 1 filippenko , a. v. et al . 1992a , , 384 , l15 filippenko , a. v. et al .
1992b , , 104 , 1543 filippenko , a. v. et al .
1995 , , 450 , l11 garnavich , p. et al .
1998 , , 493 , l53 hamuy , m. , phillips , m. m. , schommer , r. a. , & suntzeff , n. b. 1996 , , 112 , 2391 iwamoto , k. et al .
1998 , , 395 , 672 iwamoto , k. et al .
2000 , , 534 , 660 jeffery , d. j. , leibundgut , b. , kirshner , r. p. , benetti , s. , branch , d. , & sonneborn , g. 1992 , , 397 , 304 jha , s. , garnavich , p. , challis , p. , kirshner , r. , & calkins , m. 1998 , iau circ .
7011 kirshner , r. p. et al .
1993 , , 415 , 589 li , w. , filippenko , a. v. , treffers , r. r. , riess , a. g. , hu , j. , & qiu , y. 2000 , apj , in press ( astro - ph/0006292 ) li , w. et al . 2001 , in preparation mazzali , p. a. , lucy , l. b. , danziger , i. j. , gouiffes , c. , cappellaro , e. , & turatto , m. 1993 , , 269 , 423 miller , d. l. , & branch , d. 1990 , , 100 , 530 oke , j. b. et al .
1995 , , 107 , 375 perlmutter , s. et al .
1998 , , 391 , 51 perlmutter , s. et al .
1999 , , 517 , 565 press , w. h. , teukolsky , s. a. , vetterling , w. t. , & flannery , b. p. 1992
, numerical recipes , 2ed ( cambridge univ . press ) richmond , m. w. , filippenko , a. v. , & galisky , j. 1998 , , 110 , 553 riess , a. g. et al .
1997 , , 114 , 722 riess , a. g. et al .
1998 , , 116 , 1009 schmidt , b. p. et al .
1998 , , 507 , 46 tonry , j. et al .
1999 , iau circ .
7312 tonry , j. et al .
2001 , in preparation wells , l. a. et al .
1994 , , 108 , 2233 | we present optical spectra , obtained with the keck 10-m telescope , of two high - redshift type ia supernovae ( sne ia ) discovered by the high - z supernova search team : sn 1999ff at @xmath0=0.455 and sn 1999fv at @xmath0 @xmath11.2 , the highest - redshift published sn ia spectrum .
both sne were at maximum light when the spectra were taken .
we compare our high-@xmath0 spectra with low-@xmath0 normal and peculiar sne ia as well as with sne ic , ib , and ii .
there are no significant differences between sn 1999ff and normal sne ia at low redshift .
sn 1999fv appears to be a sn ia and does not resemble the most peculiar nearby sne ia . |
although our model arrives at a number of conclusions that are developed by combining three different physical theories , the assumptions and approximations used in each are well developed in the condensed matter physics and biophysics literature . here , we summarize the main physical assumptions and review the mathematical approximations used : @xmath324 _ steady - state and equilibrium assumptions _ : ribosome diffusion and motion along the mrna
are treated within steady state , while the configurational distribution of the mrna polymer is not directly coupled to ribosome diffusion or motion , and is considered in thermodynamic equilibrium .
the inverse `` harmonic distance '' @xmath93 is determined from _ equilibrium _ mrna configurational distributions , but parametrically influence the nonequilibrium steady - state processes of diffusion and the tasep .
equilibration times of unentangled polymers and diffusion times over the length of the mrna are on the order of milliseconds to seconds , while the relaxation to steady - states in the tasep occur over seconds to on the order of a couple minutes .
thus , on experimental time scales longer than these , transients in the ribosome throughput have dissipated , and the steady - state and equilibrium assumptions are appropriate .
one might be tempted to formulate the specific mechanisms in terms of the common notions of reactions being kinetically or thermodynamically controlled . in this biochemical terminology ,
the tasep is kinetically controlled , since the ribosomes take irreversible steps as each amino acid is added during elongation .
the mrna configurations , computed under equilibrium conditions , are by definition thermodynamically controlled .
however , each of the proposed mechanisms is a simple , single , independent process , the notion of kinetic control _ versus _ thermodynamic control is irrelevant . within each mechanism , there are no alternate `` reaction paths '' or outcomes for kinetic or thermodynamic control to apply . however , it is possible that the mrna conformations and the binding protein - mediated loop formation does not reach equilibrium on the time scale of measurements of ribosome throughput .
this possibility is also discussed in the experimental consequences and proposed measurements section .
@xmath324 _ gaussian chain polymer model for mrna _ : unlike trna , the coding regions of mrna is relatively devoid of secondary structure .
the single - stranded mrna is treated using standard statistical physics of polymers that assumes nonintersecting random walks of step size defined by the polymer persistence length .
for single - stranded mrna without adsorbed proteins , the persistence length @xmath325 bases .
when loaded with large ribosomes , we assume that the persistence length is on the order of the ribosome size and that it is approximately uniform along the chain . although the ribosome loading might varying slightly along the chain , this variation occurs only near the ends and does not appreciably affect the equilibrium end - to - end distributions .
although we treat only phantom ( nonintersecting ) polymers , effects due to the binding of finite - sized pabs and cap proteins are explicitly treated when computing the end - to - end distribution functions in the small distance regime where steric exclusion of the end proteins are important .
@xmath324 _ single component `` ribosomes '' _ : the assembly of ribosomes before or during adsorption onto the initiation site can be modeled as an effectively single , rate - limiting component that undergoes standard diffusion in the bulk solution . including more chemical details
will not qualitatively alter our results , since in diffusive steady - state , all species concentrations would be spatially distributed as @xmath93 and parametrically affect the tasep in the same qualitative manner .
@xmath324 _ equal particle and step sizes _ : ribosomes moving along mrna are treated with a discrete tasep where the step size is exactly equal to the particle diameter .
however , ribosomes are large and occlude @xmath34 codons so that they move one particle diameter only after about @xmath326 steps ( amino acid transfers ) . nonetheless , the qualitative behavior of the currents for different @xmath35 remain unchanged . for the sake of simplicity and clear analytic expressions ( eqs .
[ kstar ] , [ bstar ] , and [ jexactb ] ) , we have restricted our analysis to @xmath327 .
exact large @xmath14 asymptotic expressions for the steady state current for general @xmath35 are given in appendix d. @xmath324 _ uniform elongation step rates in the tasep _ : the analytic solutions represented by eqs .
[ jexact ] , [ sexact ] , and [ asemeqn ] are based on uniform elongation rates @xmath12 along the mrna .
it is known that @xmath12 can vary by factors of 2 - 10 @xcite , depending on the codon in question and the availability of the associated trna . as a first step ,
we have simply assumed a scenario in which the elongation rates do not vary appreciably over the coding region .
more elaborate models that include specified elongation rates @xmath328 across the mrna chain would require extensive simulations for each realization of @xmath329 .
@xmath324 _ bulk diffusion limited adsorption _ : ribosomes , or the relevant rate - limiting component of a ribosome , diffuses in bulk and directly attaches to the initiation site .
capture of the ribosome by the initiation end of the mrna may occur in a two - step process of nonspecific adsorption from bulk , followed by linear diffusion along a segment of the mrna , before ultimately interacting specifically with the initiation site @xcite .
although studied in the context of linear diffusion and search along dna @xcite , direct evidence for such scanning mechanisms in the initiation of mrna translation has been hard to obtain @xcite .
for example , secondary structure in the form of small mrna knots near the 5 region must be melted before efficient ribosome scanning can occur @xcite .
nevertheless , one - dimensional diffusion of ribosomes along the mrna near the initiation sight _ is _ implicitly included in our model .
the conjectured scanning mechanisms suggest that ribosomes scan locally near the start codon @xcite .
thus , if ribosome recycling via diffusion through the bulk is rate - limiting , the scanning region near the initiation where the linear diffusion occurs can be considered as binding region of larger _ effective _ capture radius @xmath75 .
consider identical particles that are driven through a long one - dimensional lattice of @xmath330 sites .
the lattice is discretized into steps of unit length ( a step size corresponding to a codon step ) , while the particles are of integer size @xmath331 . for each particle to move a distance roughly equal to its diameter
, @xmath35 consecutive steps must be taken .
thus , we expect that effectively , the mean current would be approximately described by equations [ jexact ] or [ asemeqn ] but with @xmath12 replaced by @xmath332
. a mean field model for the asymmetric exclusion process containing particles that occupy @xmath35 substrate lattice sites ( mrna codons ) has been solved .
the analysis is beyond the scope of this paper , but the resulting steady - state currents follow the same qualitative phase diagram " ( fig .
[ phase0 ] ) as the tasep with particles of size @xmath327 .
that is , for large entrance and exit rates , there is a maximal current phase ( iii ) , bounded by low ( i ) and high ( ii ) density phases .
the effects of increasing the particle size to @xmath333 only quantitatively changes the values of the currents in each of these phases , and can be straightforwardly integrated into the present study .
the general ( for all particle sizes @xmath35 ) result for the steady - state currents in the infinite chain length limit are @xmath334 \mbox{(ii ) } & \displaystyle \b
< { p\over 2 } , \b<\a & \:\ , \displaystyle j \equiv
j_{r } = { \b(1-\b / p)\over 1+(q-1)\b / p } \\[13pt ] \mbox{(iii ) } & \displaystyle \a,\b\geq { p\over 2 } & \:\ , \displaystyle j\equiv j_{max}= { p\over ( \sqrt{q}+1)^{2}}. \end{array } \label{asemeqnq}\ ] ] these results have been verified to be exact ( to within numerical precision ) by extensive monte - carlo simulations .
note that for large @xmath35 , the maximal current @xmath335 is that given by eq . [ asemeqn ] but with @xmath336 .
these results only serve to quantitatively shift the phase boundaries between the different current regimes and decrease the magnitude of the currents .
for example , if @xmath337 , the phase boundary between the low density and the maximal current regime occurs at @xmath338 , respectively , rather than at 0.5 .
for the sake of simplicity and manageable algebraic expressions , we have in this study only considered the @xmath327 case .
our analysis should be applied to the mrna translation problem with the understanding that @xmath12 in eq .
[ asemeqn ] and subsequent equations is roughly the rate for a ribosome to move its molecular size , not the rate for an individual trna transfer .
if , however , the above expressions were used , then @xmath12 in expressions [ asemeqnq ] would be identified with the typical single amino acid transfer rate .
is situated at the initiation site ( not drawn to scale ) . @xmath284 and
@xmath102 correspond to the number of bases of the utr s which are assumed to be relatively protein - free and have short persistence length @xmath339 . here , the persistence lengths in the coding regions ( thick curve , described by the tasep ) is @xmath340.,height=249 ] consider the probability distribution @xmath121 of the initiation - termination separation _ in the absence _ of loop formation . since the ribosome can be much larger than the typical persistence length in the noncoding region of single - stranded mrna , @xmath341 . for @xmath342 , @xmath343 , unless the noncoding regions are very long , with @xmath344 . for shorter noncoding regions ,
the expression for @xmath345 must be evaluated more carefully , particularly for small @xmath346 , in order to compute @xmath347 correctly .
assume the termination site starts a random walk from any position on the sphere .
details of the different segments of mrna are shown in figure [ fig3 ] .
the problem maps to that of heat diffusion from a sphere of size @xmath75 with reflecting boundary conditions and an instantaneous uniform temperature source on the surface .
the probability that the initiation site ( that is linked to the termination site via @xmath158 persistence lengths ) is within @xmath346 of the sphere can also be described by the temperature near a sphere with an exterior instantaneous source of temperature .
the diffusion equation for the probability distribution @xmath348 obeys @xmath349 where the thermal conductivity is associated with the squared persistence length , @xmath350 , and time corresponds to the length @xmath351 .
the initial and boundary conditions corresponding to a chain that originates from the surface of the otherwise impenetrable ribosome particle are @xmath352 where we have assumed spherical symmetry .
following carslaw and jaeger ( 1959 ) , we define @xmath353 to reduce ( [ heat ] ) to @xmath354 , with boundary conditions @xmath355 the solution for @xmath356 is found using laplace transforms , and is @xmath357 \displaystyle \hspace{1 cm } { e^{r / a-1}e^{\kappa t / a^{2}}\over 4\pi a^{2 } } \mbox{erfc}\left[{r - a \over 2\sqrt{\kappa t } } + { \sqrt{\kappa t}\over a}\right ] .
\end{array}\ ] ] the probability distribution is thus @xmath358 \:\hspace{1 cm } \displaystyle e^{n\ell^{2}/(6a^{2 } ) } \mbox{erfc}\left[\sqrt{{3\over 2n}}{r - a \over\ell}+\sqrt{{n\over
6}}{\ell\over a}\right ] .
\end{array } \label{wmn}\ ] ] note that _ if _
@xmath359 , as it is for @xmath360 , equation ( [ wmn ] ) would be approximately @xmath361 \ : & \displaystyle \bigg[1 - 36{a^{2}\over l^{2 } } \left(1-{x^{2}\over l^{2}}\right)+o\left({a^4\over l^{4}}\right)\bigg ] , \end{array}\ ] ] which reduces to end - to - end probability distribution for a gaussian random chain .
however , since @xmath362 , we need to use the full expression eq .
[ wmn ] for the loop contribution ( eq .
[ wloop ] ) in the calculation of @xmath363 and @xmath93 . for the wlc
, an approximate probability distribution function can be reconstructed from commonly used phenomenological force - extension relationships .
if the force - extension interpolation given by marko and siggia @xcite is shifted to take into account the finite - sized origin , @xmath364.\ ] ] the initiation - termination distance distributions can be estimated using @xmath365 \over \int_{a}^{n\ell}\mbox{d}r \exp\left[-\int_{n\ell}^{r } f(z ) dz\right]}.\ ] ] this end - to - end probability distribution from both fjc and wlc models are plotted in figs .
[ prob]@xmath123 .
the wlc model gives qualitatively similar distributions to those of the fjc model , provided the contour length is appropriately reduced .
furthermore , the wlc and fjc models provide qualitatively similar averages @xmath366 if the @xmath14 used in the wlc is sufficiently reduced . upon using eqs .
[ wloop ] and [ weff ] , one can compute the effective end - to - end distribution of a chain with segments of different persistence length and with attached loop binding proteins , as shown in fig .
[ prob]@xmath22 .
fjc and wlc models for @xmath121 for @xmath367 .
the wlc distribution approximates that of the fjc if the effective number of persistence lengths @xmath14 is reduced .
this reduction compensates for the stiffness of the chain that tends to give more weight at larger distances .
@xmath3 fjc and wlc distributions for @xmath368 .
note the heuristic cutoff applied to the wlc model at @xmath369 .
as expected , for equal @xmath14 , the wlc model gives a typically larger separation and hence smaller @xmath166 ; however , @xmath370 for @xmath371 in all cases .
@xmath10 the effective end - to - end distance distribution @xmath50 constructed from @xmath372 via equations [ ploop0 ] and [ wloop].,height=489 ]
asymptotic expressions for the steady - state current given by derrida
_ et al . _
@xcite are valid only far from the phase boundaries . however ,
in our present model , we are interested in how a change in the mrna length @xmath14 allows the system to cross over from one behavior to another . for the sake of completeness
, we derive limiting forms for the current @xmath373 near phase boundaries .
an asymptotic expansion in the rates about @xmath374 is taken first , with @xmath14 fixed . from the exact expression eq .
[ sexact ] given by @xcite , we find the following asymptotic expansion clemens , m. j. 1996 .
protein kinases that phosphorylate eif2 and eif2b , and their role in eukaryotic cell translation control .
_ in _ translational control
. j. w. b. hershey , n. sonenberg , and m. b. mathews , editors .
cold spring harbor press , cold spring harbor , ny . derrida , b. and m. r. evans .
1997 . the asymmetric exclusion model : exact results through a matrix approach in nonequilibrium statistical mechanics in one dimension
. cambridge university press , cambridge , uk .
duncan , r. , s. c. milburn , and j. w. b. hershey .
1987 . regulated
phosphorylation and low abundance of hela cell initiation factor eif4f suggest a role in translational control .
_ j. biol .
_ , 262:380 - 388 .
dunn , t. m. _ et al_. 1984 .
an operator at -280 base pairs that is required for repression of arabad operon promoter : addition of dna helical turns between the operator and promoter cyclically hinders repression , _ pnas _ , 81:5017 - 5020 .
jackson , r. j. 1996 . a comparative view of initiation site selection mechanisms .
translational control .
j. w. b. hershey , n. sonenberg , and m. b. mathews , editors .
cold spring harbor press , cold spring harbor , ny . kruger , m. k. , s. pedersen , t. g. hagervall , and m. a. sorensen .
the modification of the wobble base of trna(glu ) modulates the translation rate of glutamic acid codons _ in vivo_. _
_ , 284:621 - 631 .
marko , j. f. , and e. d. siggia .
stretching dna .
_ macromolecules _ , 28:8759 - 8770 .
zacharias , m. and p. j. hagerman . 1996 .
the influence of symmetric internal loops on the flexibility of rna .
_ 96:276 - 289 .
martnez - salas , e. , r. ramos , e. lafuente , and s. lpez de quinto .
functional interactions in internal translation initiation directed by viral and cellular ires elements .
virology _ , 82:973 - 984 .
mathews , m. b. , n. sonenberg , and j. w. b. hershey .
origins and targets of translational control .
_ in _ translational control
. j. w. b. hershey , n. sonenberg , and m. b. mathews , editors .
cold spring harbor press , cold spring harbor , ny .
michel , y. m. , d. poncet , m. piron , k. m. kean , and a. m. borman .
cap - poly(a ) synergy in mammalian cell - free extracts : investigation of the requirements for poly(a)-mediated stimulation of translation initiation .
_ j. biol .
_ 275:32268 - 32276 .
mills , j. b. , cooper , j. p. and p. j. hagerman . 1994 .
electrophoretic evidence that single - stranded regions of one or more nucleotides dramatically increase the flexibility of dna .
_ biochemistry_. 33:1797 - 1803 .
wickens , m. , j. kimble , and s. strickland . 1996 .
translational control of developmental decisions .
_ in _ translational control .
j. w. b. hershey , n. sonenberg , and m. b. mathews , editors .
cold spring harbor press , cold spring harbor , ny . | @xmath0 * abstract * we explore and quantify the physical and biochemical mechanisms that may be relevant in the regulation of translation .
after elongation and detachment from the 3 termination site of mrna , parts of the ribosome machinery can diffuse back to the initiation site , especially if it is held nearby , enhancing overall translation rates .
the elongation steps of the mrna - bound ribosomes are modeled using exact and asymptotic results of the totally asymmetric exclusion process ( tasep ) @xcite .
since the ribosome injection rates of the tasep depend on the local concentrations at the initiation site , a source of ribosomes emanating from the termination end can feed back to the initiation site , leading to a self - consistent set of equations for the steady - state ribosome throughput .
additional mrna binding factors can also promote loop formation , or cyclization , bringing the initiation and termination sites into close proximity .
the probability distribution of the distance between the initiation and termination sites is described using simple noninteracting polymer models .
we find that the initiation , or initial ribosome adsorption binding required for maximal throughput can vary dramatically depending on certain values of the bulk ribosome concentration and diffusion constant .
if cooperative interactions among the loop - promoting proteins and the initiation / termination sites are considered , the throughput can be further regulated in a nonmonotonic manner .
potential experiments to test the hypothesized physical mechanisms are discussed .
keywords : polymers , asymmetric exclusion process , protein production * introduction * the rate of protein production needs to be constantly regulated for all life processes . genetic expression , protein production , and post - translational modification , as well as transport and activation ,
are all processes that can regulate the amount of active protein / enzymes in a cell .
although much recent research has focused on the biochemical steps regulating the switching of genes and rates of transcription , translational control mechanisms , post - translational processing , and macromolecular transport are also important .
for example , during embryogenesis , nuclear material is highly condensed , transcriptional regulation is inactive , and translational control is important @xcite . in other instances ,
transcriptional regulation is accompanied by long lag times , particularly with long genes .
translational regulation is also the only means by which rna viruses express themselves .
protein production , as with other cellular processes , requires the assembly of numerous specific enzymes and cofactors for initiation .
this assembly occurs in free solution and on the 5 initiation site of mrna .
translation involves unidirectional motion of the ribosome complex along the mrna strand as amino - acid - carrying trna successively transfer amino acids to the growing polypeptide chain .
images of mrna caught in the act of translation often show numerous ribosome complexes attached to the single - stranded nucleotide ( fig .
[ afm]@xmath1 ) .
the multiple occupancy is presumably a consequence of very active translation , when many copies of protein are desired . under certain conditions , the local concentration of trna , ribosomes , initiation factors , etc .
, will control protein production .
one possible _
physical _
feedback mechanism underlying all the other biochemical regulation processes utilizes local concentration variations of the components of translation machinery .
moreover , there is ample biochemical evidence that the 5 and 3 ends of eukaryotic mrna interact with each other , aided by proteins that bind to the poly(a ) tail and/or regions near the initiation site @xcite , particularly if the 5 initiation terminus is capped .
the presence of both a poly(a ) tail and a 5 cap have been found to synergistically enhance translation rates in a number of eukaryotic systems @xcite .
numerous proteins that initiate translation , such as eukaryotic initiation factor eif4 , have been identified to bind to the cap and initiate ribosomal binding @xcite .
a different set of proteins , poly(a ) binding proteins ( pab ) such as pab1p , are found to bind to the poly(a ) tail .
the proteins on the 5 cap and the poly(a ) tail are also known to form a complex ( cap - eif4e - eif4g - pab1p - poly(a ) tail ) which can increase translation rates @xcite .
_ in vitro
_ solutions of capped , poly(a)-tailed mrna , trna , and ribosomes fail to display synergy @xcite , indicating that additional factors are required for cooperative interactions between the cap and the poly(a ) tail .
however , _ in vitro _ systems that include caps , poly(a ) tails , eif s , and pab s reveal circularized mrna structures in electron micrograph ( em ) ( fig .
1@xmath1 ) and atomic force microscopy ( afm ) ( fig .
1@xmath2 ) images . in this way
, it is thought that various components of the translation machinery can be recycled after termination without completely reentering the enzyme pool in the cytoplasm .
an electron micrograph of polysomes on mrna . @xmath3 an afm micrograph of circularization of mrna mediated by loop forming proteins . from wells _
et al . _ , ( 1998 ) .
these images are of double stranded rna of approximate length 2 - 4@xmath4 the dsrna persistence length .
single stranded end segments with loop binding factors comprise the ends.,height=566 ] even in uncapped mrna , there is evidence that certain sequences in the terminal 3 untranslated region ( utr ) can enhance translation to levels comparable to those seen in capped mrnas @xcite . additionally , there are indications that proteins near the termination end can , upon contact , directly activate @xcite or inactivate @xcite ribosome entry at the 5 initiation site .
loops also appear to be a common motif in _ dna _ structures @xcite and appear to take part in transcriptional regulation @xcite .
double stranded dna has a much longer persistence length than single - stranded nucleic acids ( such as mrna ) and is much less likely to form loops without accompanying binding proteins or specific sequences .
direct evidence for rna `` circularization '' is shown in figure [ afm]@xmath2 , which shows loop formation of relatively short double - stranded mrna in the presence of loop - binding factors at their ends @xcite .
it is reasonable to expect that the more flexible single - stranded mrna decorated with ribosomes can form similar loops . besides the afm - imaged loop of double stranded rna shown in fig .
[ afm]@xmath2 , there is also substantial evidence , particularly in viral mrnas , that base pairing between uncapped 5 regions and non - polyadenylated 3 regions forms closed loops of many kilobases @xcite .
this loop formation by direct base pairing , or `` kissing , '' is a very plausible mechanism by which the 3 utr recruits ribosomes and delivers them to the 5 initiation site @xcite . in this paper , we model the proposed cyclization , or `` circularization '' @xcite and ribosome recycling mechanisms .
cooperative interactions of the initiation and termination sites with eif s and pab proteins will also be considered within a number of reasonable assumptions .
since translation employs an immense diversity of mechanisms and proteins that vary greatly across organisms @xcite , we will only develop an initial , qualitative physical picture of cytoplasmic mrna translation consistent with the ingredients mentioned above .
three different coupled effects are considered in turn : ( @xmath5 ) a totally asymmetric exclusion process ( tasep ) describing the unidirectional stochastic motion of the ribosome along the mrna , ( @xmath6 ) the diffusion and adsorption / desorption kinetics from the mrna initiation / termination sites , and ( @xmath7 ) the polymer physics associated with how the termination and initiation sites are spatially distributed relative to each other . the ribosome density along the mrna , as well as the time - averaged throughput of ribosomes , the ribosome `` current , '' are described by solutions of the tasep .
the parameters in the tasep are the internal hopping rates and the injection and extraction rates at the initiation and termination sites , respectively .
since ribosome components that diffuse in bulk must adsorb on the initiation site , the injection rate used in the tasep will be proportional to the local concentration of the rate - limiting ribosome .
ribosomes that reach the termination site desorb and reenter the pool of diffusing ribosomes .
the distance between the termination end and the initiation site , when ribosomes are released , can thus influence the absorption rate and hence the overall translation rate .
the initiation - termination end - to - end distance distribution can be estimated with basic polymer physics .
the end - to - end distance distribution can include effects such as specific binding of poly(a ) associated proteins with the 5 cap , thereby forming a loop , bringing the initiation and termination sites into close proximity .
although our model applies only to cytoplasmic mrna translation , many of its components can also be adapted to treat mrna adsorption on endoplasmic reticulum ( er ) and er - assisted translation .
* physical models * we now consider the physical processes necessary to describe the above - mentioned translation processes . at the relevant time scales
, we will see that fluctuations in these physical mechanisms are uncorrelated with each other .
this allows us to consider simple steady - states where time or ensemble averages of the tasep , ribosome diffusion in the cytoplasm , and the mrna chain conformations are uncorrelated and can be taken independently of each other .
a simplifying schematic of the basic ingredients of mrna translation is given in fig .
[ fig1 ] . an mrna chain loaded with ribosomes ( green ) , in various stages of protein ( black ) production .
ribosomal components as well as other components such as trna exist at a uniform background concentration .
the initiation and termination sites are additional sinks ( @xmath8 ) and sources ( @xmath9 ) , respectively , of ribosomes . @xmath3 binding factors ( yellow and dark grey ) can increase the probability of loop formation or `` circularization , '' which brings the poly(a ) tail ( red ) in better proximity to the initiation site , enhancing ribosome recycling .
@xmath10 schematic of the associated tasep with injection ( @xmath11 ) , internal hop ( @xmath12 ) , and desorption ( @xmath13 ) rates labelled.,height=460 ] * the asymmetric exclusion process * the tasep is one of a very small number of interacting nonequilibrium models with known exact solutions .
asymmetric exclusion models have been used to effectively model qualitative features of diverse phenomena including ion transport @xcite , traffic flow @xcite , and the kinetics of biopolymerization @xcite .
briefly , the model consists of a 1d lattice of @xmath14 sites , each of approximately the molecular size of a ribosome unit .
each variable @xmath15 represents the ribosome occupation at site @xmath5 of the coding region of mrna .
each site can be occupied by at most one ribosome and the mean occupation @xmath16 at each site @xmath17 .
the probability in time @xmath18 that an individual ribosome moves forward to the next site ( toward the 3 end ) is @xmath19 , provided the adjacent site immediately in front is unoccupied .
backward moves are not allowed , since ribosomes are strongly driven motors that move unidirectionally from 5 to 3. the entrance and exit rates at the initiation @xmath20 and termination @xmath21 sites are denoted @xmath11 and @xmath13 , respectively ( cf . fig [ fig1]@xmath22 ) . the exact steady state solutions to this kinetic model , including the average density @xmath23 , and the mean particle ( ribosome ) current
have been found by derrida and evans @xcite , using a matrix product ansatz , and by schtz and domany @xcite , using an iteration method .
an exact representation for the steady - state current across an @xmath14-site chain is @xcite @xmath24 where @xmath25 in the @xmath26 limit , the 1d tasep ( eq . [ jexact ] ) admits three nonequilibrium steady - state phases , representing different regimes of the steady state current @xmath27 : @xmath28 \mbox{(ii ) } & \displaystyle \b < { p\over 2 } , \b<\a & \:\ , \displaystyle j \equiv j_{r } = \b(1-{\b\over p } ) & \:\ , \s_{{n\over 2 } } = 1-{\b\over p } \\[13pt ] \mbox{(iii ) } & \displaystyle \a,\b\geq { p\over 2 } & \:\ , \displaystyle j\equiv j_{max}={p\over 4 } & \:\ , \s_{{n\over 2 } } = { 1\over 2}. \end{array } \label{asemeqn}\ ] ] the phases ( i ) , ( ii ) , and ( iii ) defined by eqs . [ asemeqn ]
are denoted as the maximal current , low density , and high density phases , respectively , and are delineated in fig .
[ phase0 ] by the dotted phase boundaries .
qualitatively , when @xmath13 is small , and injection rates are faster than extraction rates ( @xmath29 ) , the rate - limiting process is the exit step at @xmath9 .
therefore , the high occupancy phase ( ii ) has a low current which is a function of only the slow step @xmath13 . in the opposite limit of fast desorption at @xmath9 , and slow injection at @xmath8 ( small @xmath11 )
, the chain is always nearly empty , and has a small current @xmath27 that depends only upon the rate limiting step @xmath11 . for large @xmath30 ,
the system attains maximal current @xmath31 where the effective rate - limiting steps are internal hopping rates @xmath12 . in this phase ,
the constant current @xmath32 is independent of further increases in @xmath11 or @xmath13 .
the ribosomal currents given by eqs . [ asemeqn ] and the associated phase diagram in fig .
[ phase0 ] are valid only in the @xmath26 limit . nonetheless , the @xmath33 phase diagram is qualitatively accurate for the currents expected at large but finite @xmath14 . )
limit nonequilibrium phase diagram of the standard tasep . the maximal current ( ii ) , low density ( i ) , and high density ( ii ) phases and their corresponding steady state currents are indicated . in this and
subsequent phase diagrams , solid curves correspond to phase boundaries across which the slope of the steady - state currents ( with respect to the parameters ) is discontinuous . across the dashed phase boundaries , the currents and their first derivatives are continuous.,height=288 ]
there may appear to be a microphysical inaccuracy : the tasep defined above corresponds to individual movements with step length equal to the ribosome size .
however , ribosomes typically occlude @xmath34 codons , so that it takes @xmath34 microscopic steps for the ribosome to move the distance of its own size @xcite .
an accurate approximation for the throughput @xmath27 ( eq .
1 ) is to assume that each step between two sites defined in our model consists of @xmath34 actual trna transfers .
the effective rate @xmath12 is thus the average trna transfer rate reduced by a factor of @xmath34 . with this consideration ,
the tasep completely determines the steady state ribosome throughput as long as the effective rate @xmath12 is appropriately defined .
therefore , we will treat the mrna translation problem using steps sizes equal to the ribosome size , with the understanding that for appropriately rescaled transition rates , our results will be qualitatively correct .
the exact currents of a tasep , where the particle diameters are @xmath35 times the step size , is given in appendix b @xcite .
explicit monte - carlo simulations have also been performed on large particle@xmath36 small step size dynamics to confirm the accuracy of the results @xcite .
what remains is to determine the self - consistent dependence of the model parameters , in particular @xmath11 and @xmath13 , on the local ribosome concentration ( which in turn depends on the mean current @xmath27 ) , diffusion rates , circularization , etc .
for example , the injection rate @xmath11 at the initiation site will be proportional to a microscopic binding rate @xmath37 times the local ribosome concentration . * steady - state release , diffusion , and capture * the complete mrna translation machinery is extremely complicated , since it is comprised of many auxiliary rna and protein cofactors , as well as a collection of active mrna chains .
since there are many active mrna chains in the cytoplasm , each mrna chain feels the sinks ( initiation sites ) and sources ( termination sites ) of all the other mrna chains .
however , these other randomly distributed chains , each with their own initiation and termination sites , contribute an averaged background ribosome concentration .
thus , it is only the termination site ( ribosome source ) associated with the initiation site _ on the same _ mrna chain that resupplies the initiation site in a correlated manner .
we thus consider a single `` isolated '' mrna chain and for the sake of simplicity , assume that a single component , say phosphorylated elongation initiation factor eif4f or eif2 , say @xcite , is key to a rate limiting step .
we will generically call this component the `` ribosome . ''
consider a source of newly - detached ribosomes ( emanating from the 3 termination site ) at position @xmath38 away from the 5 initiation site .
the probability of finding this particle within the volume element d@xmath38 about @xmath38 obeys the linear diffusion equation with the termination site acting as a source , @xmath39 where @xmath40 is the bulk ribosome diffusion constant , @xmath41 is the instantaneous rate of ribosome release from the termination end , and @xmath42 is the probability that the termination site is within the positions @xmath38 and @xmath43 from the initiation site .
although eq .
[ diff ] can be solved exactly for all times , the tasep result ( eq . [ jexact ] ) is appropriate only in the steady - state , so we must consider that limit for all processes .
the typical mrna passage time of a single ribosome is on the order of one minute . the bulk diffusion constant of the 10 - 15 nm radius ( @xmath44 nm ) ribosome unit is @xmath45@xmath46/s .
a ribosome molecule will diffuse the length of a 1 kb pair mrna strand in @xmath47s .
therefore , with each release of a ribosome from the termination site , the probability density appears as a pulse which passes through the initiation site over a time scale shorter than it takes for a ribosome to stochastically hop a few lengths of its size along the mrna chain .
therefore , an _
upper _ bound on the amount of correlation between concentration fluctuations and @xmath48 can be found by considering the _ equal _ time two - point correlation in the maximal current phase @xmath49 @xcite .
two - point correlations in other current regimes are smaller , and decay exponentially with @xmath14 @xcite .
therefore , we can neglect the correlation of the current @xmath41 with the occupancy @xmath48 at the initiation site .
moreover , the end - to - end distribution @xmath50 arises from the statistics of the mrna polymer configurations and is also assumed independent of both @xmath41 and @xmath48 .
the steady state ribosome distribution can thus be found by setting @xmath51 on the left - hand - side of eq .
[ diff ] and taking the time , or ensemble , average of the remaining poisson equation to obtain @xmath52 where @xmath53 is the steady - state current of ribosomes emanating from the termination end of the mrna re - entering the bulk ribosome pool , and @xmath54 is the ensemble average of @xmath55 .
the boundary condition for @xmath56 at the initiation site will depend on the occupancy of that site .
when it is empty , there is a flux due to the microscopic adsorption step onto the first site .
when @xmath57 , the bulk ribosome probability distribution will obey perfectly reflecting boundary conditions .
since the probability _ at _ @xmath58 , @xmath59 depends on the occupation @xmath48 , @xmath60
. the mean concentration at @xmath58 must be found by averaging the currents in the two states , @xmath57 , and @xmath61 .
when the initiation site is empty , @xmath62 since the steady state current @xmath63 when the initiation site is full , the averaged steady - state current is @xmath64 where @xmath65 is the fraction of time that the initiation site is unoccupied , ready to absorb a ribosome from the bulk .
this probability is not directly dependent on the distribution @xmath66 , but will depend on the time - averaged local concentration @xmath56 , which in turn depends on @xmath50 only through the distance of the source site at @xmath9 .
the solution to eq .
[ laplace ] , obeying the boundary conditions eq .
[ bc1 ] and @xmath67 , is @xmath68 \:\hspace{2cm}\displaystyle + { j\over d}\int d\r ' g(\r-\r')w_{\it eff}(\r ' ) , \end{array } \label{cr}\ ] ] where @xmath38 is distance measured from the initiation site , and @xmath69 \:\hspace{1cm}\displaystyle \sum_{\ell=0,m=-\ell}^{\infty , m=+\ell } \left[{ka^{\ell}-4\pi a^{2 } d\ell a^{\ell-1 } \over ka^{-\ell-1}+ 4\pi a^{2 } d(\ell+1)a^{-\ell-2}}\right ] \\[13pt ] \ :
\displaystyle \hspace{1.6 cm } \times { r_{<}^{-\ell-1}\over ( 2\ell+1)r_{>}^{\ell+1}}y^{*}_{\ell m}(\omega)y_{\ell m}(\omega ' ) \label{green } \end{array}\ ] ] is the associated green function . in eq .
[ green ] , @xmath70 is the smaller(larger ) of @xmath71 and @xmath72 are the spherical harmonic functions of the solid angle @xmath73 defined by the vector @xmath38 @xcite .
the first two terms in eq .
[ cr ] arise from the uniform concentration @xmath74 at infinity and the effects of a sink of radius @xmath75 at the initiation site .
the sink decreases the effective concentration to a level below that of @xmath74 .
the last term proportional to @xmath27 increases the local concentration and is the result of the source ( termination site ) some finite distance away from the initiation site .
if @xmath76 , and ribosomes do not bind even when the initiation site is empty , the current @xmath27 must vanish , and @xmath77 , as expected .
however , one can not simply consider the limit @xmath78 in eq .
( [ cr ] ) because @xmath37 and @xmath79 are related through @xmath27 , the current determined by the tasep in the rest of the chain .
this can be seen by considering the limit @xmath78 .
if the rest of the tasep contains the rate - limiting step to ribosome throughput , making @xmath27 very small , it will effectively block clearance of the initiation site , since all sites of the chain will be nearly occupied . in this case ,
@xmath80 and @xmath81 is small ( despite a large @xmath37 ) , and @xmath82 , as expected .
however , if the rest of the chain is not rate - limiting , and if clearance of the initiation site can occur fast enough , @xmath83 and @xmath81 can be large . in this case , @xmath84 .
the tasep current @xmath27 will eventually be balanced with @xmath85 .
note that @xmath27 is determined by eq .
[ jexact ] which in turn _ depends _ on the entry rate @xmath11 ( in other words , @xmath86 ) .
thus , steady - state currents need to be _ self - consistently _ determined , since @xmath87 and @xmath79 are not parameters , but dynamical variables that will in turn be determined by setting @xmath85 .
the analysis which uses eq .
[ jexact ] to find self - consistent explicit expressions for @xmath27 will be presented in the results and discussion .
since the averaged bulk concentration profile is spherically symmetric about the initiation site , only the @xmath88 terms in the expression for @xmath89 survive and @xmath90 \ : & \displaystyle = { 4\pi a^{2 } d kc_{\infty } \over ka+4\pi a^{2}d } + { 4\pi a^{2 } d kj\over 4\pi d(ka+4\pi a^{2 } d)}\int_{r'>a } \!\!\!\!\!\!\!\ ! d\r ' { w_{\it eff}(\r ' ) \over r ' } \\[13pt ] \ : & \displaystyle = { k a \over k+4\pi a d}\left[4\pi dc_{\infty}+ { j\over r}\right ] , \label{phi } \end{array}\ ] ] where @xmath91 the surface concentration at the sink surface @xmath75 is reduced from the `` bulk '' value by a factor of @xmath92 , due to adsorption and diffusional depletion .
however , part of this initiation site concentration is also replenished at a rate proportional to the flux @xmath27 , due to the presence of a nearby termination ( source ) site .
the effects of this replenishment are measured by the mean inverse separation @xmath93 .
the `` harmonic distance '' @xmath94 defines the effective distance felt by diffusing ribosomes as they make their way from the termination end back to the initiation site .
this particular @xmath95 scaling is a consequence of the solution to poisson s equation ( eq .
5 ) in three dimensions , and is related to the capture probability of diffusing ligands , as analyzed by berg and purcell @xcite . equation [ phi ] contains two unknowns , @xmath87 and @xmath96 . we can use the explicit solution eq .
[ jexact ] if we identify the injection rate @xmath11 of the tasep with the unoccupied initiation site current @xmath97 .
equation [ jexact ] then relates @xmath86 to @xmath96 .
a second equation can be used by noticing that the flux itself must be balanced . upon using @xmath98 in eq .
[ phi ] , a second relationship between @xmath86 and @xmath79 can be found .
substitution of the solution for @xmath86 ( in terms of experimentally known or controlled parameters @xmath99 ) into eq .
[ jexact ] determines the self - consistent , steady - state ribosome current .
this analysis , using the three different explicit forms of eq .
[ jexact ] ( in the long chain limit ) is presented in the results and discussion . * end - to - end distribution @xmath50
* we now find @xmath66 in order to compute @xmath94 and obtain @xmath87 . in some cases ,
the mrna chain may be anchored to cellular scaffolding or er membranes such that the initiation - termination separation is fixed . if one is interested in steady - state protein production over a period which allows little change in initiation - termination distance , @xmath100 , and @xmath101 . in other cases ,
the mrna may be free to explore numerous conformations on the protein production time scale .
although it is possible that long mrna strands may contain secondary structure , we will assume that ribosomes , as they move along the mrna , melt out these structures .
although there is evidence that mrna can contain small , local loops @xcite , it is less likely that they have larger - scale tertiary structure .
thus , we will estimate @xmath50 and @xmath94 with simple polymer models . and @xmath102 persistence lengths @xmath103 are solid black , while the the neglected short ends are dashed curves .
the loop binding factors are of typical size @xmath104 .
@xmath105 nonlooped conformations in which the initiation - termination site distribution function is governed by @xmath106 . @xmath3 the initiation - termination distribution function in looped configurations is denoted @xmath107 .
@xmath108 is weighted more strongly at small @xmath109 relative to @xmath106 . for stronger attraction between loop binding factors the probability of loop formation increases , decreasing the effective distance @xmath94 that ribosomes must diffuse to be recycled back to the initiation site.,height=355 ] as shown in fig .
[ fig1 ] , the mrna is comprised of three segments divided between two qualitatively distinct regions .
typical coding regions are @xmath110 base pairs , corresponding to @xmath111 . at low ribosome densities ,
the uncovered mrna base pairs will be rather flexible , and the effective persistence length @xmath112 will be a local average between @xmath75 and the 2 - 4 nucleotide persistence length @xmath103 of uncovered mrna .
large reductions in the persistence length of dsdna containing segments of single - stranded regions have also been observed by mills _
et al .
_ @xcite .
more sophisticated theories for variable persistence lengths can be straightforwardly incorporated ; however , for simplicity , we approximate the persistence length in the coding region to be a uniform constant on the order of @xmath113 , the individual ribosome exclusion size
. the contour length of the coding region is thus @xmath114 with @xmath115 .
the untranslated regions , or utrs between the initiation site and the binding factor ( dark gray ) , and between the termination site and the loop - binding factor ( yellow ) , with persistence lengths @xmath103 , have contour lengths of @xmath116 and @xmath117 , respectively .
typical @xmath118 are on the order of 100 bases so that @xmath119 . however , extremely long noncoding segments of order 1kbp can exist @xcite where @xmath120 .
in what follows we will also neglect all the excluded volume effects of the remaining short ends of the mrna chain . as demonstrated by wells _
et al . _
@xcite in figure [ afm]@xmath2 , mrna can form loops in the presence of binding proteins . therefore
, we expect that @xmath66 ( and hence @xmath93 ) will be a linear combination of @xmath121 and @xmath122 , the initiation - termination probability distributions in open and looped mrna configurations , respectively .
these configurations are shown in figs .
[ fig2]@xmath123 . for simplicity ,
we will use probability distributions associated with noninteracting ( phantom ) chains and approximate the distributions @xmath124 with both a freely jointed chain ( fjc ) and worm like chain ( wlc ) models with appropriate persistence lengths @xmath112 .
the finite - sized , short distance behavior of the @xmath125 will be important for accurately computing @xmath126 .
as we will see , @xmath122 can be constructed from the more fundamental quantity @xmath121 @xcite .
since we are eventually interested in either ribosome transport from termination to initiation or in activation / deactivation of initiation or release sites due to direct contact with the end proteins , we compute in appendix c the distance distribution @xmath121 in the state where site @xmath9 is occupied and site @xmath8 is unoccupied . using the @xmath121 computed in appendix c
, we can thus consider the contributions of looped configurations to the effective end - to - end distance distribution .
the binding energy between the 5-cap and poly(a ) tail proteins , @xmath127 ( in units of @xmath128 ) , determines the probability that the chain is looped : @xmath129 \ : & = \displaystyle { e^{u_{0 } } \over e^{u_{0 } } + \omega_{0}(\mbox{open})/\omega_{0}(\mbox{loop } ) } , \end{array}\ ] ] where the free energies of a closed and open mrna chain are @xmath130 and @xmath131 , respectively . since the ratio of the number of configurations under looped and open chain conditions is the ratio of probabilities of loop formation in the _ absence _ of head - tail interactions ( @xmath132 ) , @xmath133 , and @xmath134 the probability that the ends of a noninteracting ( in the absence of loop binding proteins ) chain intersects itself within the interaction volume defined by a thin spherical shell of thickness @xmath135 ( the binding interaction range ) is approximately @xmath136 \ : & \quad\times w_{\ve}(\r_{m+n+n}\vert\mbox{open})\mbox{d}\r_{1}\mbox{d}\r_{2 } \\[14pt ] \ : & \displaystyle \approx \sqrt{{6\over \pi}}\left({d\over l_{t}}\right)^{2}\left({\delta \over l_{t}}\right)\left[1+o(d / l_{t})\right ] , \end{array } \label{ploop0}\ ] ] where @xmath104 is the typical size of the loop binding factors and @xmath137 .
we have assumed the total radius of gyration @xmath138 , and used a gaussian chain as a qualitative approximation for the distributions used in the calculation of @xmath139 .
the conditional probability distribution @xmath122 for a looped chain is @xmath140 where @xmath141 denotes the single segment , open chain probability distributions in the two segments with persistence lengths @xmath142 . for @xmath143 , the loop distribution given by eq .
[ wloop ] is qualitatively similar to the distribution function @xmath144 of the short segment of persistence length @xmath103 .
^{-1}$ ] over which recycled ribosomes must diffuse .
@xmath105 the dependence of @xmath145 as a function of loop binding energy @xmath146 is shown for @xmath147 persistence lengths of coding mrna . for large binding energies
@xmath146 , the initiation and termination sites are brought closer together .
the crossover between the end - to - end distribution function of a free chain to that of a loop occurs near @xmath148 .
increasing the length of the short noncoding ends of the mrna predominantly increases the typical distance @xmath94 in the large @xmath146 , looped regime . @xmath3
the @xmath14-dependence of @xmath145 with the ratio of noncoding persistence lengths to coding persistence lengths @xmath149 .
the @xmath14-dependence manifests itself primarily in the low @xmath146 , open chain regime . @xmath10
the @xmath14 dependence of @xmath145 for various @xmath146.,height=499 ] using eqs .
[ ploop ] , [ ploop0 ] , [ wloop ] , and [ wmn ] we construct the effective initiation - termination distance distribution @xmath150 @xmath151 is plotted in appendix c ( fig .
[ prob ] ) for various @xmath146 .
qualitatively similar loop probability distributions have also been computed by liverpool and edwards @xcite within the wlc model but without finite - sized molecules at the ends . here and
in all subsequent analyses , we use the typical parameters @xmath152 , and @xmath153 . as @xmath146
is increased , the distance distribution function switches over from @xmath106 to @xmath122 .
the statistics of @xmath106 and @xmath122 are governed by @xmath154 and @xmath155 , respectively .
the loop forming factors , since they are close to the initiation and termination sites ( @xmath156 ) , enhance the probability that the ends are close to each other , particularly when the binding energy @xmath146 is large .
the harmonic distance , @xmath94 , determined using @xmath50 is shown in figs .
[ reff]@xmath123 as functions of loop binding energy @xmath146 .
the result given by the last line in eq .
[ ploop0 ] , when used in eqs .
[ ploop ] and [ weff ] qualitatively describes a crossover in @xmath50 from @xmath121 to @xmath122 behavior at @xmath157 + o(d^{2}\delta / l_{t}^{3}).\ ] ] in fig .
[ reff]@xmath1 , @xmath145 is shown with @xmath147 , but at various noncoding lengths @xmath158 . in the large binding strength limit
, @xmath145 depends only on the short distance @xmath159 .
when loops rarely form , the typical separation between initiation and termination sites can only depend on @xmath160 which is the only quantity varied in fig .
[ reff]@xmath2 .
notice that the exact fjc solution ( appendix c ) , or truncated wlc solution for @xmath161 ensures that @xmath162 for all values of @xmath163 , and @xmath146 .
the dependence of @xmath145 on @xmath14 is shown in fig .
[ reff]@xmath22 for various @xmath146 .
when @xmath146 is small , the initiation - termination harmonic distance @xmath94 is controlled by @xmath160 and increases as @xmath164 . for larger @xmath146 , the chain is partially bound into a loop where the distance is controlled by the much shorter @xmath165 .
the harmonic distance @xmath94 remains small unless @xmath14 becomes extremely large so that entropy can dominate and the loop ends can unbind .
we now couple our mathematical models by incorporating the @xmath50-weighted inverse harmonic distance @xmath166 into the local , effective concentration @xmath167 given by eq .
[ phi ] .
the effective injection rates @xmath168 that control the translation rate within the steady - state tasep are then self - consistently determined .
* results and discussion * here , we compute the possible currents @xmath27 and the parameter space in which each are valid .
we will use the exact solution eq .
[ jexact ] , or its three asymptotic forms ( eqs .
[ asemeqn ] ) , as well as @xmath98 in eq .
[ phi ] , to find all relevant quantities and parameter phase boundaries .
substitution of @xmath98 into eq .
[ phi ] and solving for @xmath79 , we find @xmath169 upon multiplying eq .
[ y ] by @xmath86 , we find @xmath170 \equiv j(kc(a),\b , p ) = \left\ { \begin{array}{l}kc(a)(1-kc(a)/p ) \\[11pt ] \b(1-\b / p ) \\[11pt ] p/4 \end{array } \right . \end{array } \label{kc}\ ] ] to find @xmath87 in terms of known parameters , we use the explicit solutions of the tasep for the current @xmath171 ( eq . [ jexact ] or [ asemeqn ] ) as indicated on the right - hand - side of eq .
[ kc ] .
the exact solution eq .
[ jexact ] yields an @xmath172 order equation in @xmath86 which we solve numerically .
only one of the @xmath172 roots of eq .
[ kc ] is real , yields occupations between zero and one , and is the physically relevant .
the self - consistent solutions for @xmath86 are used to evaluate @xmath171 , which are plotted in figures [ jexactfig]@xmath123 . as expected ,
shorter chains yield slightly higher current .
larger @xmath40 also increases the current and makes the approximate maximal current phase obtainable at smaller @xmath173 .
asymptotic limits for the current near phase boundaries and at large @xmath14 are given in appendix d. .
the self - consistent currents were found by numerically finding the roots to the polynomial in @xmath27 obtained by substituting the last line of expression 19 into the exact equation 1 .
@xmath105 steady - state currents as a function of the injection rate @xmath173 for @xmath174 and various @xmath175 . for @xmath176 , @xmath177 and @xmath178
are compared .
@xmath3 @xmath27 as a function of length @xmath14 for @xmath179 and @xmath180.,width=307 ] the numerical solutions depicted in fig .
[ jexact ] show , that for even modest @xmath181 , the currents are accurately described by their asymptotic expressions eq .
[ asemeqn ] .
therefore , we can very accurately solve for @xmath86 and steady - state ribosome currents by separately considering each phase and its associated asymptotic form of @xmath27 .
first assume that the detachment rate @xmath182 and consider the maximal current ( phase iii in the tasep ) where @xmath32 .
this occurs when both @xmath183 . to determine the parameter regime in which @xmath32 holds , we solve for @xmath87 and determine for what range of parameters @xmath184 . using @xmath32 in eq .
[ kc ] , we find @xmath185 the criterion for maximal current , @xmath186 , is thus @xmath187 upon solving eq .
[ k1 ] for @xmath37 , we find the minimum @xmath188 required to achieve maximal current @xmath32 : @xmath189 note that for large enough @xmath190 the critical value @xmath191 can diverge .
the divergence is more likely or larger @xmath94 and occurs when there is simply not enough ribosome nearby to provide a large enough `` on '' rate @xmath192 to achieve maximal current .
even when the source ( termination end ) is held at the initiation site ( @xmath193 ) , there is the possibility that @xmath191 , and maximal current , are never attained .
this behavior arises because even for ribosomes released at an infinitely absorbing spherical initiation surface , there is a probability of escape @xcite .
next , let us consider small @xmath194 and large @xmath168 .
the mrna has a high ribosome occupancy and a steady - state current @xmath195 .
this regime ( phase ii ) is termination rate - limited and occurs for @xmath196 and @xmath197 . upon using @xmath198 in eq .
[ kc ] , @xmath199 the only physical range of @xmath13 that satisfies inequality [ betakc ] is @xmath200 \ : \displaystyle \hspace{1cm}\times \left[\sqrt{1 + { 4(r / a)\bar{d}kc_{\infty } \over p((\bar{d}+1)r / a-1)^{2}}}-1\right ] , \end{array } \label{bstar}\ ] ] where @xmath201 . equation [ bstar ] defines the phase boundary between the high - density , exit rate - limited phase ( ii ) and the low - density , initiation rate - limited phase ( i ) .
this phase boundary is plotted as a function of @xmath173 for fixed @xmath202 in figs .
[ phase1]@xmath2 . in the limit @xmath203 , the phase boundary straightens as in the standard tasep and is approximately @xmath204.\ ] ] finally , when @xmath205 , but the entrance rate @xmath86 is low ( @xmath206 ) , a low density phase with @xmath207 exists .
the phase boundary delineating the low density phase ( i ) is defined by @xmath208 and @xmath209 . upon using the current @xmath210 in eq .
[ kc ] , we find @xmath211 , and the current in the initiation rate - limited phase ( i ) : @xmath212 \:\displaystyle \hspace{1cm}\times \left[\sqrt{1+{4(r / a)\bar{d}kc_{\infty}\over p((\bar{d}+1)r / a-1)^{2}}}-1\right]-4\pi dr c_{\infty}. \end{array } \label{jexactb}\ ] ] in the limit @xmath213 , @xmath214 \ : \quad \displaystyle { ( 4\pi a d)^{2}(k+4\pi a d)k c_{\infty } \over ( k(1-a / r ) + 4\pi a d)^{3}}\left({kc_{\infty } \over p}\right ) + o(p^{-2 } ) , \end{array } \label{jlasymp}\ ] ] which reduces to the result one would expect from infinitely fast initiation site clearance . summarizing , the large-@xmath14 steady - state ribosome currents ( given by eq . [ jexact ] ) in terms of ribosome concentrations and kinetic `` on '' rates are @xmath215 \mbox{(ii ) } & \displaystyle \b < { p\over 2 } , \b <
kc(a ) & \:\ , \displaystyle j \equiv j_{r } = \b(1-{\b\over p } ) \\[13pt ] \mbox{(iii ) } & \displaystyle k > k^{*},\b\geq { p\over 2 } & \:\ , \displaystyle j\equiv j_{max}={p\over 4}. \end{array } \label{asemeqn2}\ ] ] where @xmath86 in phase ( i ) is expressed in terms of known parameters according to eq .
[ kc ] .
the mean occupations of the initiation and termination sites , in each regime , can now be readily found . at the first site , @xmath216 , where we use @xmath217 , or @xmath218 ( currents associated with each phase ) , and @xmath86 found from eqs . [ kc ] , [ betakc ] , or [ kcmax ] .
similarly , the occupation at the last site is @xmath219 .
all of our results can be expressed in terms of three of the four nondimensional parameters : @xmath220 , @xmath173 , @xmath221 , and @xmath145 .
we shall present our results in terms of the relevant nondimensional parameters appropriate for the discussion at hand . )
mrnas . @xmath105 the minimum binding rate ( eq . [ kstar ] ) required to support the maximal current phase assuming that @xmath222 .
this value depends on the bulk ribosome concentration @xmath74 and the distance @xmath94 between the initiation and termination sites . @xmath3 the modified phase diagrams as functions of @xmath173 for @xmath223 and various @xmath145 . @xmath10
modified phase diagrams as functions of @xmath173 for fixed @xmath220 and @xmath224.,width=316 ] figure [ phase1]@xmath1 shows the critical value @xmath191 , above which an @xmath225 tasep is in the maximal current phase ( provided @xmath226 ) .
when @xmath74 is small and @xmath12 is large , there is not enough ribosome in the cytoplasm to feed the initiation fast enough compared to the clearance rate @xmath12 .
therefore the maximal current ( @xmath32 ) arises only when the binding is efficient and @xmath227 is large . for smaller @xmath94 ( termination site close to the initiation site ) ,
smaller values of @xmath228 can still support maximal current . from eq .
[ kstar ] , we see that when @xmath229 , the critical value @xmath191 diverges and the maximal current can never be reached .
there is simply not enough ribosomes or the diffusion is too slow for there to be sufficient concentration at the initiation site to support the maximal current phase .
if the diffusion constant @xmath40 and @xmath74 are chosen such that , for example , @xmath221 is small , the critical values @xmath191 vary considerably with @xmath145 , as shown by the green points ( @xmath230 ) in fig .
[ phase1]@xmath1 .
the effects of depletion arise suddenly , with onset only at values of @xmath231 . for large @xmath145
, values of @xmath232 will render the critical @xmath191 values very sensitive to @xmath94 .
if the initiation site has an interaction size of @xmath233 nm , and @xmath234/s ( 20 - 30 codons / s ) @xcite , a diffusion constant of @xmath235 @xmath46/s requires an effective concentration of @xmath236 m for the phase diagram to be sensitive to diffusional depletion and @xmath94 .
although typical total cytoplasmic ribosome concentrations are @xmath237 m , many components must assemble in order to activate a translation - viable ribosome .
for example , eif4f exists at 0.01 - 0.2 times the total ribosome concentration @xcite .
furthermore , this already low abundance of eif often needs to be further phosphorylated to be active .
thus , the effective concentrations @xmath74 ( and even diffusion constants ) appropriate for our model may very well be low enough to fall within the range for the phase boundaries to be extremely sensitive to diffusional effects .
figures [ phase1]@xmath238 show the steady - state phase diagrams as functions of @xmath239 and effective binding rate @xmath173 . in these phase diagrams , as in the unperturbed ones defined by eq .
[ asemeqn ] , the upper left region corresponds to a low density phase , the lower right region corresponds to a high density phase , and the upper right region describes a half - occupied ( except near the ends @xmath240 ) , maximal current phase .
the current @xmath27 is constant throughout the maximal current phase and is not changed if @xmath173 or @xmath13 is increased beyond @xmath241 and @xmath242 , respectively . the phase diagram is modified by ribosome diffusion and depletion near the initiation site .
the unmodified phase boundary between phases ( i ) and ( ii ) of the tasep ( eq . [ asemeqn ] ) would simply be defined by the straight line segment @xmath243 .
the main effects of diffusional depletion ( by the initiation sink ) and replenishment ( by the termination source ) on the standard phase diagram fig .
[ phase0 ] is to shift the low density - maximal current phase boundary to larger effective injection rates @xmath173 and bend the low density - high density phase boundaries accordingly .
figure [ phase1]@xmath2 depicts the phase boundaries defined by eqs . [ kstar ] and [ bstar ] for fixed @xmath244 , and fixed @xmath245 as indicated by the green points in fig .
[ phase1]@xmath1 . in this example
@xmath246 for @xmath247 , respectively .
note that for @xmath248 that @xmath191 diverges and the maximal current phase is never attained . if @xmath249 , then there will be a finite value of @xmath145 such that @xmath191 diverges .
if , instead , @xmath250 is held fixed , the phase boundaries are nearly straight , as shown in fig .
[ phase1]@xmath22 . here , we fixed @xmath224 , and plotted the phase diagrams for @xmath251 .
the corresponding values of @xmath173 above which the maximal current phase is attained are @xmath252 and @xmath253 , respectively .
our results up to this point are contingent on the fact that measurements are averaged over time scales such that the tasep and the diffusion processes have reached steady - state , and the mrna chain distribution has thermally equilibrated .
the possibility exists that the chain conformations are not in thermodynamic equilibrium while the tasep and the bulk ribosome diffusion has reached steady - state for a given chain conformation .
thus , although not relevant within each of the three well - defined physical processes , the issue of kinetic _ versus _ thermodynamic control of ribosome throughput arises when one considers measurements over time scales that are insufficient to allow equilibration of the mrna chain .
the consequences of this are discussed in the following section .
* experimental consequences and proposed measurements * the basic physical mechanisms described in our model for mrna translation suggest a number of experimental tests .
however , it must be emphasized that the model is meant to provide qualitative guidelines most useful for studying trends and how they depend on _ physical _ parameters .
translation occurring _ in vivo _ involve too many molecular species and biochemical processes to be quantitatively modeled , especially in the absence of significantly more detailed experimental findings .
nonetheless , our proposed mechanisms can be probed with carefully designed , simplified , _ in vitro _ experiments . here , we discuss in detail the basic expected phenomena and their regimes of validity . first note from figure [ jexactfig ] and from appendix d that the exact currents for a finite number of codons @xmath14 very rapidly approach the asymptotic values given by eq .
[ asemeqn ] as @xmath14 increases . even when @xmath14 is only @xmath254 , the steady - state ribosome currents are only a few percent off the exact @xmath33 results . in other words , the exact solution eq .
[ jexact ] is a very good approximation to eq .
[ asemeqn ] for @xmath181 .
therefore , as a mental guide , it is typically sufficient to consider the currents @xmath27 corresponding to an infinite chain ( @xmath33 ) given by eq .
[ asemeqn ] , but nonetheless consider a finite initiation - termination separation ( measured by the harmonic distance @xmath94 ) . * polysomal density variations * although we have focused on the steady - state current , the particle ( ribosome ) densities in each of the three current regimes are different and may be detected . in the tasep model ,
the ribosome density profiles along the mrna chain vary only near the initiation and termination ends . in the interior of the mrna
, the density is relatively uniform and are given by the last column in eqs .
[ asemeqn ] . in the exit - rate limited phase ( small @xmath255 ) , where @xmath198 , the midpoint density @xmath256 is high , while in the low injection rate case
, @xmath257 , and @xmath258 is low .
the typical density in the maximal current regime is @xmath259 .
these densities are also approximately correct when one explicitly treats large ribosomes that occlude many codon `` lattice sites . ''
therefore , we might expect that one may be able to predict in which current regime translating mrna exists if ribosome densities can be estimated from images taken with _
e.g. _ afm or em techniques .
for example , in figure [ afm]@xmath1 , the high density of ribosomes suggests that the system is in phase ( ii ) where the steady - state current @xmath260 is a function only of the detachment rate @xmath13 . *
kinetic binding rate and ribosome concentration dependences * figure [ phase1]@xmath22 shows the minimum effective attachment rate @xmath241 necessary for a large system to be in the maximal current regime ( where the ribosome current @xmath261 ) as a function of the effective ribosome diffusion constant .
an additional requirement is that the effective detachment rate @xmath262 . the value of @xmath191 can be tuned perhaps by substitution of the codons comprising the initiation sites , or by other physical means .
although ribosome diffusion constants are difficult to vary over a wide range ( by modifying the solution viscosity ) , the critical @xmath191 is a very sensitive function of @xmath40 , particularly for small @xmath40 .
it is thus possible that slightly increasing the ribosome diffusivity can dramatically decrease the @xmath191 necessary for the system to be in the maximal current regime . as mentioned
, changing the mrna length @xmath14 does not significantly affect the overall steady - state current along the chain ( beyond about @xmath263 ) but it _ can _ change the statistics of the initiation - termination separation by changing @xmath94 .
increasing the harmonic separation @xmath94 has qualitatively the same effect as decreasing the ribosome diffusivity , since terminated ribosomes now have further to diffuse back to the initiation site . for @xmath264
the maximal current regime is never reached .
this can be easily seen from equation [ kstar ] .
thus , rather than tuning the ribosome diffusivity , decreasing @xmath74 may preclude the system from entering the maximal current phase if eq .
[ dcrit ] is satisfied .
there is simply not enough ribosome available for sufficient initiation to be achieved so that the maximal current phase arises .
phase diagrams for @xmath224 .
@xmath105 phase diagram for fixed @xmath265 with trajectories @xmath266 corresponding to increasing kinetic `` on '' rate @xmath37 .
@xmath3 phase diagram when @xmath267 is fixed , and trajectories @xmath268 correspond to increasing bulk ribosome concentration @xmath74 .
trajectory @xmath269 traverses the ( i)-(iii ) phase boundary for @xmath270 ( thick curves ) but not for @xmath271 ( thin curves ) .
trajectory @xmath272 on the other hand , traverses the ( i)-(ii ) phase boundaries for both @xmath273.,width=316 ] when inequality [ dcrit ] is not satisfied , the maximal current phase can exist . in figure [ bbar_idiot]@xmath1 , we replot the phase diagram corresponding to @xmath224 shown in fig . [ phase1 ] .
fixing the parameter @xmath274 allows @xmath37 to be the only free parameter .
this kinetic `` on '' rate @xmath37 can be tuned by varying ribosome recruitment proteins such as eif4e .
if @xmath226 , @xmath275 and @xmath12 are held constant , increasing @xmath37 from a sufficiently small value allows one to traverse the trajectory @xmath276 .
the steady - state ribosome current starts in the low density phase ( i ) with current given by eq .
[ jexactb ] . as @xmath37 is increased , the steady - state current increases until it continuously crosses over into the maximal current regime ( iii ) , where the ribosome throughput is given by @xmath32 .
further increasing @xmath37 when inside the maximal current phase ( iii ) will no longer affect the steady - state ribosome current . if , however , @xmath277 , the current behavior abruptly crosses over ( along trajectory @xmath278 ) from that given by eq .
[ jexactb ] to @xmath279 corresponding to the high ribosome density phase ( ii ) . in this phase
the detachment step is rate - limiting , and further increases in @xmath37 will no longer affect the throughput . if @xmath37 is held fixed and the ribosome concentration is independently varied instead , it is more instructive to plot the phase diagram for fixed @xmath280 and @xmath145 , as shown in fig .
[ bbar_idiot]@xmath2 . here
, we choose the representative values @xmath224 and @xmath281 and motivate parameter trajectories obtained by varying only @xmath74 . for @xmath226 , increasing the bulk ribosome concentration traces out the trajectory @xmath269 continuously from the low density phase ( i ) ( eq .
[ jexactb ] ) to the maximal current ( @xmath32 ) phase . further increasing the concentration well into the maximal current phase
will no longer affect the throughput .
similarly , if @xmath282 , increasing @xmath74 can shift the behavior from that of the low density phase to that of the high density , exit rate - limited phase .
alternatively , one may vary @xmath12 , the mean elongation rate of individual ribosomes , by controlling the trna concentration in solution . for example , decreasing available trna will move the system from the lower left to upper right in fig .
[ bbar_idiot]@xmath2 , eventually reaching a steady - state current @xmath32 . despite the apparent fundamental importance of the kinetic binding , or `` on '' rate in translation ,
there are no systematic and independent measurements of @xmath37 in the literature .
the required independent estimates of @xmath37 may be achieved by perhaps combined kinetic and affinity measurements of the association of a minimal set of components , including only the ribosomes and a portion of the 5 initiation codons and cofactors . for the off rate @xmath194 ,
similar ideas can be employed .
the trna or ribosome release factor concentrations for the last codon can also be adjusted to tune the off rate @xmath13 . * codon and utr length dependences * in experiments where it is possible to vary the number of codons @xmath14 , the typical harmonic distance @xmath94 can also be tuned .
the phase diagrams in figs .
[ phase0 ] , [ phase1 ] , and [ bbar_idiot ] all correspond to different regimes of eq .
[ jexact ] in the large @xmath14 limit . in practice eq .
[ jexact ] , is no longer sensitive to @xmath14 for @xmath181 ; however , the harmonic distance @xmath94 between initiation and termination sites continues to increase as @xmath164 , affecting the local concentration @xmath87 , and thus the effective _
parameter _
@xmath283 in eq .
[ jexact ] . as shown in fig .
[ phase1]@xmath22 , increasing @xmath145 shifts the phase boundaries to the right , making the maximal current phase ( iii ) harder to attain unless @xmath37 or @xmath74 is concomitantly increased .
however , due to the @xmath164 dependence , this effect would be relatively weak for all but enormous values of @xmath14 .
hence we have chosen the qualitatively reasonable value @xmath224 in figs .
[ bbar_idiot]@xmath123 .
although there may be a weak increase in @xmath145 as one increases the mrna length , the effects of increasing the coded sections ( @xmath14 ) or the noncoded sections ( the untranslated regions @xmath284,@xmath102 ) , can be different depending on @xmath146 . for large @xmath146
, looped configurations dominate and the distance between initiation and termination sites will be more sensitive to @xmath158 , the shortest distance between them ( cf . fig . [
fig2]@xmath2 ) . the effect of lengthening @xmath158 on @xmath145 in the high @xmath146 regime is clearly shown in fig .
[ reff]@xmath1 . for small @xmath146 ,
open configurations dominate and the short segments @xmath284 and @xmath102 at the two ends do very little to affect @xmath145 relative to @xmath14 .
thus , although length dependences are expected to be weak , increasing the codon length @xmath14 would more likely increase @xmath145 ( and hence decrease throughput @xmath27 ) in the small @xmath146 , or repulsive limit .
conversely , increasing @xmath284,@xmath102 would more likely increase @xmath145 when @xmath146 is large and loops dominate the mrna conformations .
* initiation - termination cooperative effects * we have so far considered only the effects of the binding energy @xmath146 on loop formation , @xmath93 , and the resulting local ribosome concentration at the initiation site . however , evidence suggest that contact between elongation factor proteins and/or poly(a ) tail proteins can enhance or suppress the kinetic binding rates @xmath37 through direct molecular contact and cooperativity @xcite .
there is the possibility that in looped states , pab s can interact with _ initiation _ machinery and modify @xmath37 , and/or elongation factors can assist or hinder detachment of ribosomes at _ termination_. modification of @xmath37 and/or @xmath194 through direct contact between proteins associated near the initiation and termination sites may be an additional mechanism by which translation rates can span the regimes shown in figs .
[ phase1]@xmath238 and [ bbar_idiot ] .
qualitatively , the experimental finding that contact between the mrna ends affect the initiation or possibly termination processes can be modeled by assuming _ effective _
`` on '' or `` off '' rates @xmath285 = k_{0}(1-p_{\it loop } ) + k_{1}p_{\it loop } \\[13pt ] \beta_{\it eff}[u_{0 } ] = \beta_{0}(1-p_{\it loop } ) + \beta_{1}p_{\it loop } , \label{keff } \end{array}\ ] ] where @xmath286 and @xmath287 are the binding and `` off '' rates when the mrna is open and looped , respectively .
as @xmath146 is varied , both the intrinsic rates as well as the sink - source separation @xmath94 are modified . using expression [ keff ] for @xmath37 and @xmath13 in equations [ kstar ] and [ bstar ]
, the dependence of @xmath27 on the binding energy @xmath146 can be mapped .
a number of qualitatively different scenarios are possible . if @xmath288 but @xmath289 , the current is a monotonically increasing function of @xmath146 because the binding rate increases and the ribosome source ( 3 terminus ) is brought closer .
both of these effects monotonically increase the steady - state current . however , if for fixed @xmath194 , @xmath290 , then these two effects can partially balance each other and there is the possibility of a maximum in @xmath291 .
a maximum occurs when initially , as @xmath146 is increased , the decrement in @xmath292 can not keep up with the enhancement in local ribosome concentration due to the increasing likelihood of loop formation ( _ i.e. _ the shifting of the high current phase boundary to lower @xmath292 ) .
however , if @xmath293 is sufficiently small , @xmath292 eventually diminishes , such that one arrives at the low density , low current regime .
these effects are illustrated in the sequence of figures [ control]@xmath294 .
the steady - state current , self - consistently calculated from eqs .
[ jexact ] , [ kc ] , and [ keff ] , has a possible maximum and is shown as a function of @xmath146 in figure [ control]@xmath40 . here
, we have chosen @xmath295 , and @xmath296 .
only certain sets of parameters permit a maximum .
small values of @xmath228 and large @xmath14 result in the largest maxima . for large values of @xmath228 , diffusion
is fast , local ribosome concentrations are not significantly depleted by the initiation site , and the high current regime is already pushed to low values of @xmath173 . therefore , increasing @xmath146 and decreasing @xmath94 does not further drive the high current regime towards significantly lower @xmath173 . for
essentially the same reason , smaller @xmath14 enhance ribosome recycling , increasing the current at low @xmath146 , thereby rendering the maximum in @xmath27 to lower values of @xmath146 . as illustrated in the exampled given in figure [ control]@xmath40 , increases of @xmath297 above the background currentare possible as @xmath146 is varied .
thus , we see that the two processes , direct molecular catalysis of initiation and termination , and ribosome diffusional depletion , balance each other and may provide delicate control mechanisms during later stages of gene regulation . ) as a function of @xmath146 when the ribosome `` on '' rate @xmath37 can be modified by direct interactions with elongation factor and pab proteins . the gaussian chain approximation is used with persistence length @xmath298 .
@xmath299 show hypothetical , qualitative trajectories in the presence of a changing phase diagram .
as @xmath146 is increased , @xmath94 decreases . with
@xmath300 fixed , the phase boundaries shown in @xmath294 correspond to @xmath301 , respectively .
in addition , if @xmath302 , the effective binding rate @xmath303 also decreases with increasing @xmath146 , resulting in the trajectories indicated by the dot .
@xmath304 currents for @xmath305 and @xmath306 and @xmath147 .
the weak maximum appears only for small @xmath228.,width=316 ] * kinetic _ vs. _ thermodynamic control * finally , we point out that our analysis has been confined to the steady - state ( for the bulk ribosome diffusion and individual ribosome movement along the mrna ) and thermodynamic equilibrium ( for the statistics of the polymer statistics ) .
since it is possible for diffusion and ribosome elongation along the mrna to reach steady - state before the mrna chain reaches conformational equilibrium ( in the presence of loop - forming proteins ) , a possibility exists for `` kinetic _ versus _ thermodynamic control '' for the measured ribosome throughput .
although the loop - binding energy @xmath146 determines the equilibrium distribution of open and closed mrna conformations via @xmath307 , the kinetics of loop opening and closing are determined by energy _ activation _
barriers of the loop binding proteins .
for example , if the activation energy for creating a looped state is high , the mrna may sample only unlooped conformations on time scales of the `` steady - state '' ( with respect to the tasep and diffusion ) . in this scenario ,
the effect of the loop binding protein does not arise and the harmonic distance @xmath308 would appear to be that associated with an open chain ( @xmath309 in fig .
[ reff]@xmath123 ) .
conversely , if the the mrna chain happens to be in a looped conformation and the free energy barrier for dissociation of the loop is large , the measured current may be that corresponding to only a closed mrna loop ( mimicking the case @xmath310 ) .
this is likely to occur if the measurement time @xmath311 , where @xmath312 is the spontaneous dissociation time ( or the kramers escape time ) and @xmath313 is the activation barrier energy/@xmath314 .
the activation energy @xmath313 depends on the specific molecular details of the loop - forming proteins ; however , measurements using fluorescence quenching can be used to independently determine the distribution of times the mrna chain is looped or unlooped @xcite .
only when @xmath146 or @xmath313 are large does ribosome recycling get significantly enhanced by loop formation .
transient measurements , as well as fluctuations of the measured throughput , is beyond the scope of the paper .
* summary * we have constructed a simple model and road map for the possible physical effects at play during translation . the model incorporates driven diffusive motion which obeys exclusion statistics for ribosomes along mrna .
the initiation and termination sites are considered as sinks and sources of ribosome concentration , described by the steady - state diffusion equation ( laplace s equation ) .
the average conformations of the mrna chain define the typical initiation - termination distance which determines the how terminated ribosomes directly diffuse back to the initiation site and affect the local concentration there .
this local concentration is a parameter ( the injection rate ) in the exclusion process , but also depends on the overall ribosome throughput ( the strength of the sink and source ) .
thus , the current @xmath27 needs to be solved self - consistently .
direct cooperative enhancement of kinetic binding and `` off '' rates were also incorporated .
although it is thought that the rate - limiting step is binding and initiation of ribosomes at the initiation site @xcite , the fact that polysomes have been found to exist in both high and low ribosome occupancy states suggests that under physiological conditions , steady - state ribosome fluxes can span the regimes defined by the phase diagrams depicted figs .
[ phase0 ] and [ phase1]@xmath238 . at high occupancy ,
the rate limiting step is the off rate @xmath194 which controls the steady - state flux ( cf . phase ( ii ) in fig .
[ phase0 ] ) .
ribosome depletion by the sink and replenishment by the source can drastically affect the constant @xmath315 phase diagram , as shown in figures [ phase1 ] .
the critical values of @xmath241 that define the the left boundary of the maximal current phase ( in the @xmath26 limit ) is most sensitive to the dimensionless parameter @xmath228 when @xmath316 . for sufficiently small @xmath228 ,
the effective injection rate can not reach @xmath242 and the maximal current phase can not be attained . when @xmath317 , the explicit currents were computed from eq .
[ jexact ] and plotted in figure [ jexactfig ] .
given the possibility of cooperative interactions in looped mrna configurations , we have also found a maximum in ribosome throughput as a function of loop - binding energy @xmath146 .
many molecular and chemical details have been neglected .
as mentioned , we have ignored the fact that numerous components must assemble before initiation and have modelled only an `` effective '' rate - limiting component .
the surface concentration parameter @xmath87 in our model would be an effective concentration reflecting the local density of ribosomes capable of initiation .
proposed mechanisms of ribosome scanning @xcite , whereby ribosomes attach to segments of mrna and undergo one - dimensional diffusion before encountering the initiation site , can be adequately modeled with the present approach if one assumes that the rate - limiting step is initial adsorption onto an mrna segment .
furthermore , we have assumed that the ribosomes do not detach from the mrna until they reach the termination site and that their forward hopping rates are uniform across the whole coding region . finally , in our simple polymer model
, we have neglected both self - avoidance ( of both chain - chain and chain - ribosome exclusion ) and the fact that the effective persistence length may varying along the mrna , depending on the local ribosome density . despite these simplifying assumptions
, we find that _
qualitatively _ , subtle control mechanisms can come into play , depending on biologically reasonable physical parameters .
although there are numerous experiments probing translation , both _ in vivo _ and _ in vitro _
, many different systems and physical conditions are employed , rendering quantitative comparison with measurements difficult .
nonetheless , our model suggests new measurements that can be used to qualitatively probe the various physical hypotheses and exhibit our predicted physical trends .
for example , the effective @xmath74 can be varied in a number of ways to test with the predicted current regimes .
occupancy along the mrna can also be correlated with the high , low , and intermediate density phases .
additionally , the noncoded regions between the elongation factors and the initiation site , and the termination site and the poly(a ) tail - bound pab can be varied to test possible cooperative interactions defined by eq .
[ keff ] . since the loop formation probability @xmath318 depends on the total statistical length @xmath319 , which is dominated by the length of the coding region ( @xmath320 ) , varying @xmath284 and @xmath102 would affect , through the likelihood of molecular contact in the looped states , only @xmath292 and @xmath321 , respectively .
the actual probability of loop formation @xmath318 , and hence @xmath94 , would not be significantly affected .
chemical modification of the elongation factors or the pab s _ would _ affect @xmath146 , and hence @xmath322 , and @xmath94 through @xmath323 .
using micromanipulation techniques @xcite , it might also be possible to fix the initiation - termination distance _ in vitro_. numerous extensions to the presented models can be straightforwardly incorporated to more precisely model the chemical and microphysical processes .
codon and trna concentration - dependent variations in the internal transition rates @xmath12 @xcite , as well as random detachment processes , can be implemented using simple lattice simulations .
sites along the mrna chain at which ribosomes pause can be treated as `` defects '' in a tasep and the whole process can be treated with mean - field theory @xcite .
multiple coding regions in prokaryotic translation ( shine - dalgarno sequences ) can be modeled as a sequence of initiation ( sinks ) and termination ( sources ) sites .
similarly , cap - independent initiation at internal ribosome entry sites @xcite ( ires ) can also be treated as sinks within our basic model .
translation of er - associated mrna further involve ribosomes that attach the mrna at certain points on the er membrane . in this case
, one expects that density of cytoplasmic and er - bound ribosomes to have a strong effect on localization of mrna to er and overall translation rates .
one can also consider cases where the protein product itself is a ribosome product necessary for its self - translation ; this processes would result in initially autocatalytic protein production .
although these more complicated and interesting extensions have not been considered here , the simple models we have presented represent a first step towards the rich problem of identifying and quantifying the physical and biological mechanisms that control late stages of expression .
the author thanks s. bump , t. chang , d. herschlag , g. lakatos , e. landaw , j. rudnick , and m. suchard for vital discussions and helpful suggestions .
the author especially wishes to thank an anonymous reviewer for correcting an error in the original formulation of the adsorption boundary conditions .
this work was supported by the national science foundation through grants dms-9804370 and dms-0206733 . |
materials bearing local resonators are known to support unique dynamic phenomena , such as negative , highly anisotropic , and extreme effective properties @xcite .
systems exhibiting these phenomena are commonly referred to as `` locally resonant metamaterials , '' and are often described using linear dynamical models .
one well - known model is a chain of `` mass - in - mass '' unit cells , which consists of a lattice of interconnected lumped masses , each with coupled local resonators @xcite . by incorporating nonlinearity into locally resonant metamaterials , their dynamics become more complex . for example , metamaterials for both acoustic @xcite and electromagnetic @xcite waves have demonstrated numerous nonlinear phenomena , including tunability , harmonic generation , and the existence of nonlinear localized modes .
a promising means to create nonlinear acoustic metamaterials is provided by granular media , which consist of closely packed systems of particles that interact elastically .
granular media have been shown to support a wide range of nonlinear dynamic phenomena not encountered in conventional materials @xcite . in granular materials , the microstructural geometric nonlinearity that stems from the shape of particles in contact ( commonly modeled using hertzian contact mechanics @xcite ) results in an effective macroscopic nonlinear material response .
previous works on granular media have demonstrated numerous nonlinear effects , including solitary waves , shocks , discrete breathers , tunable band gaps , frequency conversion , and non - reciprocal wave propagation @xcite .
recent theoretical and experimental works have combined the concepts of locally resonant metamaterials with granular media .
relevant contexts include , but are not limited to , frequency shifting , harmonic generation , and localized band gap modes @xcite , traveling waves , including ones with non - vanishing tails @xcite , wave interaction @xcite , and localized and extended modes @xcite , as well as temporally periodic breathing states @xcite ( to which we will return in what follows ) . in each of these examples ,
the granular media provided a nonlinear intersite coupling , while the local resonators were linear .
less attention has been paid to cases in which granular particles play the role of nonlinear local resonators . in this work ,
we consider a one - dimensional , mass - in - mass system with linear intersite coupling and nonlinear local resonators that follow the hertzian contact model .
one motivation for considering this model is its relevance in describing a granular metamaterial consisting of a monolayer of microscale spheres adhered to a substrate , wherein surface localized elastic waves , such as rayleigh surface acoustic waves ( saws ) and lamb modes , have been shown to hybridize with the contact resonances of the microspheres in thick @xcite and thin @xcite substrates , respectively . within this context
, we imagine the portion of the substrate through which the localized elastic wave is traveling as a linearly coupled chain that is locally coupled to an array of nonlinear resonators representing the microspheres .
systems similar to the one - dimensional , linearly coupled chain with nonlinear local resonators considered here have been previously explored .
for example , amplitude - dependent band - gaps have been studied in a one - dimensional linear chain with local resonators containing a cubic nonlinearity @xcite .
other relevant works have also considered linear chains with nonlinear coupling to a rigid foundation @xcite , or a nonlinear local attachment @xcite , demonstrating heavily enriched dynamics caused by small nonlinear perturbations .
the structure of interest in the present work is the discrete breather ( db ) .
discrete breathers are solutions that are periodically oscillating in time and exponentially localized in space @xcite that have been studied theoretically and experimentally in many settings , involving a wide array of physical mechanisms @xcite .
more recently , dbs have been demonstrated in theoretical @xcite and experimental @xcite studies of ordered granular chains without local resonators , and theoretically in the presence of linear local resonators @xcite , as well as in nonlinear , locally resonant magnetic metamaterials @xcite and systems of electromechanical resonators @xcite .
we use our model to describe a locally resonant granular metamaterial for rayleigh saws , consisting of a monolayer of microspheres adhered to a thick elastic substrate .
the two independent model parameters are fit to an experimental system used in past work @xcite , so as to provide realistic parameter values to the model wherever possible .
we predict the existence of dbs in the extreme limits of vanishing and strong intersite coupling , and numerically compute a family of dbs connecting the two regimes .
furthermore , we examine the frequency - energy dependence of the dbs along the relevant branch of solutions .
we study the formation and evolution of the discrete breathers via direct numerical simulations , including within contexts that may be relevant to future experimental studies .
our motivating physical scenario is shown in the schematic of fig .
[ fig_ms_model](a ) , which describes sagittally - polarized , plane saws traveling along the surface of a thick substrate .
rayleigh saws are surface localized elastic waves that travel along a solid surface ( represented as an elastic half space in theoretical descriptions ) , and have both in- and out - of - plane ( with respect to the sample surface plane ) displacement components @xcite .
previous studies on monolayers of microspheres adhered to thick substrates have shown that rayleigh saws in the substrate hybridize with , and excite , microsphere contact resonances having translational out - of - plane @xcite , and coupled , in - plane , translational and rotational motion @xcite .
the hybridization with each of these resonances leads to classic `` avoided crossing '' phenomena @xcite characteristic of locally resonant metamaterials and mass - in - mass chains . for the analysis herein
, we focus on the avoided crossing with the contact resonance having solely out - of - plane motion @xcite .
because a plane saw is confined to the surface of the medium , it can be considered as traveling in one dimension , and as such , we represent the portion of the substrate through which the saw is traveling as an infinite lattice of lumped masses @xmath0 connected by springs with linear stiffness @xmath1 . because the contact - based modes of the microspheres @xcite have frequencies much lower than the intrinsic spheroidal vibrational frequencies of the isolated spheres @xcite ( e.g. for the microspheres studied in ref .
@xcite , the out - of - plane contact resonance was measured to be @xmath2 mhz , while the spheroidal resonance was predicted to be @xmath3 ghz ) , we model the microspheres as point masses ( of mass @xmath4 ) connected to the main chain by nonlinear springs modeling hertzian contact with a static adhesive load . the resulting discrete model of our locally resonant granular metamaterial
is shown in fig .
[ fig_ms_model](b ) . as can seen in fig .
[ fig_ms_model](b ) , the chain elements are both drawn such that their motion is in the horizontal direction .
we note that this depiction simply represents the coupling between a substrate ( or chain ) and a resonator , each having a single degree of freedom with the same , albeit arbitrary , direction of motion . within the context of the previously described physical scenario , this degree of freedom represents out - of - plane motion of the substrate and the microsphere , as the saws propagate along the sample surface indexed by @xmath5 . in dimensionless form , the associated equations of motion of this system read @xmath6_+^{3/2 } - 1 \right ) & = 0 \\
\ddot{v}_j - \frac{2}{3 } \left([u_j - v_j + 1]_+^{3/2 } - 1 \right ) & = 0 , \label{eq : eom2 } \end{aligned}\ ] ] where @xmath7 and @xmath8 are , respectively , the displacements from equilibrium of the main chain and resonators , @xmath9 , @xmath10 characterizes the relative strength of the elastic and hertzian terms , where the hertzian coefficient @xmath11 depends on the geometry and material properties of the particles in contact @xcite , and @xmath12 is the static overlap induced by the adhesive force at equilibrium .
the dimensionless time variable @xmath13 is defined in terms of the physical time @xmath14 by @xmath15 , where @xmath16 is the resonant frequency of the local oscillator on the elastic halfspace , and the displacements @xmath7 and @xmath8 are normalized to @xmath12 .
the subscript @xmath17_+$ ] indicates that the contact force vanishes for resonators that lose contact with the main chain , i.e. when the relative displacement @xmath18 exceeds the static overlap .
the hamiltonian ( energy ) corresponding to eqs . and is @xmath19^{5/2}_+ - ( u_j - v_j ) \right ) - \frac{4}{15 } \right ] .
\label{ham } \end{aligned}\ ] ] upon linearization , this system is identical to the one - dimensional mass - in - mass chain discussed in @xcite .
its dispersion relation is given by @xmath20\left(\frac{\omega}{\omega_0^{hs}}\right)^2 \nonumber \\ & & + 2k(1-\cos{(kd ) } ) = 0 , \label{eq : dr } \end{aligned}\ ] ] where @xmath21 is the bloch wave number , @xmath22 is the angular frequency , and @xmath23 is the unit cell width , taken from the physical system as the width of the granular particles .
the resulting model , as can be inferred from eqs . ( [ eq : eom1])-([eq : eom2 ] ) , possesses two effective lumped parameters , namely @xmath24 and @xmath25 .
we now attempt to fit these discrete model parameters to describe the granular metamaterial of ref .
@xcite , using the material and geometric properties specified therein .
we are intending for the model to provide an adequate representation of dynamical evolution for wavelengths significantly larger than the sphere diameter , such that the dispersion relations of the continuous and discrete models are in close agreement , and effects found at the brillouin zone boundary @xcite are avoided @xcite .
the dispersion relations for a continuous substrate ( from eq .
( 2 ) of ref .
@xcite ) and the discrete model of eqs . and
are superimposed in fig .
[ fig_dr_fitted ] .
the values of @xmath24 and @xmath25 used to plot the curves for the discrete model shown in fig .
[ fig_dr_fitted ] are chosen to match the dispersion curves in the small wavenumber regime .
more specifically , we first choose the ratio @xmath26 such that the long - wavelength sound speed of the discrete lattice , given by @xmath27 @xcite , matches the speed of rayleigh waves in the substrate for the model from ref .
this can be seen graphically in fig .
[ fig_dr_fitted ] , as the lower branches of the two dispersion relations have equal slopes at the origin .
second , making use of the analytical expression for the dispersion relation of the continuous system in ref .
@xcite , we select @xmath28 , such that the dispersion relations coincide at the intersection with the line of slope @xmath29 , where @xmath29 is the transverse sound speed of the substrate material . using these two criteria , we find the approximate fitted values @xmath30 and @xmath31 .
the physical significance of these parameter values , @xmath24 and @xmath25 , is as follows .
for large mass ratios ( @xmath32 ) , waves in the main chain ( corresponding to rayleigh saws in the substrate ) are only perturbed at frequencies very close to the local resonance ; this is confirmed in fig . [
fig_dr_fitted ] , by the relatively narrow band gap encompassing @xmath33 .
the large stiffness ratio ( @xmath34 ) indicates strong coupling between lattice sites , compared to the coupling between the main chain and the resonators .
intuitively , parameters much greater than unity are indeed expected for this system .
this is because the spheres are much smaller and less massive than the region of the substrate beneath them that is influenced by rayleigh waves , whose displacements decay exponentially from the surface with a characteristic decay length on the order of one wavelength @xcite .
similarly , the effective stiffness for the region of bulk material of the substrate influenced by the rayleigh wave can be thought to have a significantly greater effective stiffness than the relatively soft microsphere - substrate hertzian contact .
while the fitted constants depend on material and geometric properties , a simple estimate can be used to show that @xmath35 and @xmath28 are generally larger than unity when considering long waves in realistic materials , as described in appendices a and b. ) and rayleigh ( @xmath36 ) wave sound speeds of the substrate . the black dashed line denotes the linear natural frequency of the hertzian local resonators . the discrete model uses the fitted parameters @xmath37 and @xmath38 . ] both of the dispersion relations shown in fig .
[ fig_dr_fitted ] are split , as a result of the hybridization with the local resonance , into two branches : the lower ( `` acoustic '' ) branch , in which the vertical motions of the substrate surface and the spheres are in phase , and the upper ( `` optical '' ) branch , in which the motions are out of phase .
the two branches of the discrete model are separated by a band gap of width @xmath39 , given by @xmath40 for the parameters values @xmath24 and @xmath25 used herein , this band gap results in an upper cutoff frequency of the acoustic branch @xmath41% below @xmath42 and a lower cutoff frequency of the optical branch @xmath43% above @xmath42 . in the continuous model , for phase velocities greater than the sound speed of transverse bulk waves ( @xmath29 ) in the substrate , the optical branch terminates ; this is because the modes above that phase velocity are so - called `` leaky '' modes .
such leaky modes have complex frequency , which represents the radiation of energy into the bulk @xcite . despite this , the one - dimensional , discrete model used in this work was chosen over continuous and/or higher - dimensional models
because it captures many of the important features of the dispersion relation ( linear dispersion at long wavelengths and a band gap created by a local resonance ) , and facilitates theoretical prediction of discrete breathers , and their numerical computation .
we start our analysis by considering the so - called anti - continuum ( ac ) limit of vanishing coupling .
this approach , pioneered by mackay and aubry in @xcite , is based on the limit @xmath44 , corresponding to uncoupled oscillators .
while this limit is of limited physical relevance for our considerations herein , it is a particularly useful mathematical tool as a starting point for considering different breather - type configurations .
this enables a natural starting point to seed continuation algorithms , which are then continued in the parameter @xmath25 in order to identify solutions for different values of the relevant parameter . in the ac
limit our system has the form @xmath45_+^{3/2 } - 1\right ) \\
\ddot{v}_j & = \frac{2}{3}\left([u_j - v_j + 1]_+^{3/2}-1\right)\label{eqn : motion2}. \end{aligned}\ ] ] we obtain a single oscillator by defining @xmath46 , where @xmath47_+^{3/2 } - 1\right ) , \qquad \omega_0 ^ 2 = \frac{1+m}{m}. \end{aligned}\ ] ] in addition to the trivial solution @xmath48 , there are non - trivial solutions of eq . which are the level curves of the energy @xmath49_+^{5/2 } - \frac{5}{2}z\right)$ ] .
to construct a solution along the infinite lattice , each node is given as @xmath50 or the periodic function ( say with frequency @xmath51 ) satisfying .
in this paper , we consider the simplest such configuration , namely the one consisting of zeros at every node with the exception of one ( see fig . [ nlsbreather](a ) ) .
due to ref .
@xcite we know this solution will persist for nonzero @xmath25 as long as the so - called non - resonance condition @xmath52 is satisfied , where @xmath53 is the frequency of solutions of the linearized equations at @xmath54 . while any such value of @xmath51 will yield a persistent breather solution , we chose @xmath55 .
note that the two branches of the dispersion curves given by eq .
bifurcate from @xmath56 and @xmath57 as the coupling @xmath25 becomes nonzero .
thus , by choosing @xmath55 we are able to construct a band - gap breather .
the numerical continuation ( see sec .
[ sectioncont ] ) suggests that the solution constructed in the ac limit persists to the opposite limit @xmath58 . in this limit ,
other analytical techniques are available for the analysis of the solutions , which we explore next . for the purposes of the analysis , we consider small amplitude solutions ( i.e. @xmath59 ) .
thus , it is reasonable to expand the nonlinearity in a taylor series @xmath60^{3/2 } = 1 + \frac{3}{2 } x + \frac{3}{8 } x^2 - \frac{3}{48 } x^3 + h.o.t.\ ] ] in addition , if we formally consider @xmath61 where @xmath62 is the lattice spacing and we let @xmath63 , then eqs . and become @xmath64 we approximate solutions of the above set of equations with the ansatz , @xmath65 where @xmath66 , @xmath67 and @xmath68 with @xmath69 being some small , positive parameter .
substitution of the ansatz into eqs . and and equating the various orders of @xmath70 yields a hierarchy of solvability conditions .
the particular choice of ansatz is well known to yield a nonlinear schrdinger ( nls ) equation for the envelope function @xmath71 in the theory of nonlinear waves @xcite . for our system , in order to derive the nls equation we need several higher order terms : @xmath72 where each @xmath73 and the @xmath74 are real or complex coefficients . in particular , we will use the following , @xmath75 these relations are obtained through the solvability conditions , which can be found in appendix [ nls_append ] .
we highlight here that at @xmath76 the solvability condition is the dispersion relation , @xmath77(\omega/\omega_0^{hs})^2 + k^2=0 , \label{contdisp}\ ] ] where @xmath78 is the bloch wave number and @xmath79 is the angular frequency .
the connection between this dispersion relation and the one in eq . can be seen by taylor expanding the cosine terms of eq . .
the nonlinear schrdinger ( nls ) equation appears at @xmath80 , @xmath81 closed form analytical solutions of the nls equation can be found via the inverse scattering transform @xcite .
one well known solution is the so - called bright soliton , and is given by @xmath82 where @xmath83 and @xmath84 are @xmath70 independent coefficients that depend on the coefficients of the nls equation ( see appendix [ nls_append ] ) and @xmath85 is an arbitrary parameter .
such solutions arise when the coefficient of the dispersion term and that of the nonlinearity have the same sign , which is the case for the parameter values chosen here .
thus , at first order , we have the following approximation @xmath86 which is traveling plane wave that is modulated by a small amplitude , long wavelength and slowly varying localized function . for @xmath87 ,
this approximation reduces to @xmath88 which represents a standing breather with frequency @xmath89 , see fig .
[ nlsbreather](b ) . here
@xmath90 represents the lower cutoff of the optical branch of the dispersion ( see eq . with @xmath87 ) .
since @xmath70 is a small parameter , the breather frequency @xmath51 is near the lower cutoff of the optical branch , but within the gap . note the amplitude and width of the breather are both @xmath91 .
hence smaller amplitude and wider breathers are found closer to the optical branch band edge .
recalling the results of the previous section on the ac limit , we were able to construct a breather solution with frequency @xmath92 .
if we define @xmath93 , then our nls approximation ( which is valid for large @xmath25 ) will have , to first order , the same frequency , as the ac limit breather ( which is relevant for small @xmath25 ) . in the next section , we will perform parametric continuation in @xmath25 in order to connect the approximations in the two opposing limits @xmath54 and @xmath94 .
the analysis above allows us to describe breather solutions of eqs . and in the limits @xmath95 and @xmath58 .
in order to connect these two pictures , we identify periodic orbits and explore their parametric continuation @xcite .
our seed solution ( initial guess ) will be the ac limit solution with a single excited site , ( see fig .
[ nlsbreather](a ) ) . for the example considered in this section , we use the fitted value @xmath96 from section [ sec : param_fit ] , and we choose the breather frequency @xmath97 .
note that in this case @xmath98 . for a fixed breather period @xmath99 , we use the fact that @xmath100 and @xmath101 to construct the poincar map @xmath102;t ) & = \begin{bmatrix } \textbf{u}(0;u_0,v_0 ) \\
\textbf{v}(0;u_0,v_0 ) \end{bmatrix } - \begin{bmatrix } \textbf{u}(t ; u_0,v_0 ) \\
\textbf{v}(t;u_0,v_0 ) \end{bmatrix } \end{aligned}\ ] ] where @xmath103}$ ] and @xmath104}$ ] is the solution to eqs . and with initial condition @xmath105 and @xmath106 .
therefore , a periodic solution with period @xmath107 of eqs . and will be a root to ( [ eqn : poin ] ) .
a newton - raphson algorithm is used to approximate the roots of @xmath108 @xcite .
the jacobian is @xmath109 , where @xmath110 is the monodromy matrix .
the eigenvalues of @xmath110 are the floquet multipliers of the periodic solution .
the breather is considered ( spectrally ) stable if all floquet multipliers lie on the unit circle . since the system is hamiltonian , any floquet multiplier lying off the unit circle signals instability @xcite .
the solution for a frequency ( @xmath111 ) is found for @xmath25 close to @xmath56 ( here @xmath112 ) .
parameter continuation in @xmath25 is performed and is plotted against the hamiltonian energy of eq .
, as is shown in fig .
[ breathers_contk ] . in this way
, we are able to trace out a branch of solutions that emanates from the ac limit and approaches the continuum limit solution , which is well described by the nls approximation of eq .
( see black dashed - line of fig .
[ breathers_contk ] ) . here ,
we terminate the continuation at @xmath113 , which corresponds to the stiffness parameter extracted by fitting the discrete model to the locally resonant half space model from ref .
@xcite ( see sec . [
sec : param_fit ] ) .
the numerical computations were performed on a system of 201 unit cells .
this solution was spectrally stable for all values of @xmath25 considered .
the continuation in @xmath25 of the previous section was terminated at the parameter value @xmath31 ( note @xmath30 and @xmath114 ) .
we now fix @xmath96 and @xmath113 and vary the breather frequency @xmath51 using a pseudo - arclength continuation procedure @xcite .
this allows us to visualize the energy - frequency dependence of the breathers , as may be studied in an experiment with varied excitation amplitude . as a seed for the continuation , we use the eigenmode of the linearized system nearest the lowest optical band edge , which is a time - periodic solution of the full nonlinear equations of motion under conditions of vanishing amplitude
we have included a check in our computations to detect a loss of contact between the main chain and resonators , and continue the solution branch until this point . as shown in fig .
[ breathers_contw ] , the continuation reveals a family of dbs that extends from the linear eigenmode at vanishing amplitude and traverses the band gap ( and into the passband ) . in fig .
[ floquet_contw](a ) , we show the maximum magnitudes of the floquet multipliers of the branch .
this family of dbs exhibits behavior similar to those found in previous studies of diatomic granular chains @xcite , where the dbs are linearly stable for frequencies very close to the lower cutoff frequency of the optical branch . as the frequency decreases ,
the breather profiles become progressively more localized in space , as can be seen in fig .
[ breathers_contw](b - d ) .
the floquet multipliers corresponding to the modes shown in fig .
[ breathers_contw](b - d ) are shown in fig .
[ floquet_contw](b - d ) , with small deviations from the unit circle emphasized in the latter panel .
below the band gap , interactions with the acoustic band generate oscillating tails , as is shown in fig .
[ breathers_contw](e ) , and floquet multipliers depart from the unit circle along the real axis , as shown in fig .
[ floquet_contw](e ) .
we note that boundary effects are significant in the presence of oscillating tails , so the finite - length dbs do not accurately approximate the case of an infinite lattice . as an aside
, we also point out that contrary , e.g. , to what is the case in the work of @xcite for a granular crystal , here the dependence of the energy ( hamiltonian ) on the frequency is monotonic , hence in accordance with the recent criterion of @xcite , no instability arises from changes of monotonicity in this dependence .
( b - e ) breather displacement profiles corresponding to the points labeled @xmath115 - @xmath116 , respectively , in ( a ) .
the main chain displacements @xmath117 are shown as black points , and those of the local resonators @xmath118 are shown in red . ]
- @xmath116 , respectively , in ( a ) , corresponding to the same points in fig .
[ breathers_contw ] . in ( d ) and ( e ) , axes limits are chosen to emphasize the deviation from the unit circle , which is shown as a visual aid ( red dashed lines ) . ]
of the simulated lattice for high and low amplitude excitations , using eigenmode and db profiles as initial shapes .
side panels contain spatial profiles of @xmath119 at the final time step , normalized to the maximum value .
( a ) eigenmode shape with low amplitude ( approximate periodic solution ) .
( b ) db shape rescaled to low amplitude .
( c ) eigenmode shape rescaled to high amplitude .
( d ) db shape with high amplitude ( exact periodic solution ) . ] to explore the dynamics of dbs in our model , we simulate a lattice with 201 unit cells via direct numerical integration of the equations of motion given by eqs .
( [ eq : eom1 ] ) and ( [ eq : eom2 ] ) .
we consider initial conditions in two shapes : the profile of the db with frequency @xmath120 and maximum floquet multiplier magnitude @xmath121 ( as is shown in fig .
[ nlsbreather](b ) and denoted by the star in fig . [ breathers_contw](a ) ) , as well as the profile of the seeding eigenmode used in sec .
[ sectioncont ] .
b. for each of these shapes , we scale the amplitude in two ways . in the case of the db shape , we consider the exact breather shape computed via continuation ( high ) , and then consider a rescaled db shape , such that the initial displacement @xmath122 of the local resonator at the central lattice site is equal to one - hundredth of that of the exact solution ( low ) .
similarly , we consider the shape of the seeding eigenmode scaled such that the initial displacement @xmath122 of the local resonator at the central lattice site is matched to the low - amplitude , rescaled db shape ( low ) , and then finally consider a rescaled eigenmode shape , such that the initial displacement @xmath122 of the local resonator at the central lattice site is equal to that of the exact db solution ( low ) .
thus , there are four sets of initial conditions : the db shape with high amplitude ( fig .
[ fig_hist](d ) ) , which results in an exact periodic solution of eqs . and
; the eigenmode shape with low amplitude ( fig . [ fig_hist](a ) ) , which closely approximates a periodic solution ; the db shape rescaled to low amplitude ( fig .
[ fig_hist](b ) ) , which is not a true periodic solution ; and the eigenmode shape rescaled to high amplitude ( fig .
[ fig_hist](c ) ) , which also is not a periodic solution .
the duration of all simulations is @xmath123 , where @xmath124 is the period of the exact db solution .
spatiotemporal plots of the relative displacements @xmath119 of the simulated lattice , using the low and high amplitude db profiles as initial conditions ( i.e. the rescaled db and exact solution ) are shown in fig . [ fig_hist](b ) and [ fig_hist](d ) , respectively , and the corresponding cases using the eigenmode shape ( i.e. the approximate periodic solution and corresponding rescaled profile ) are shown in fig . [ fig_hist](a ) and fig . [ fig_hist](c ) . as shown in the right column of fig .
[ fig_hist ] , the breather shape spreads out from the central lattice sites at low amplitude , but remains highly localized when initiated with the energy of the exact solution .
conversely , as shown in the left column of fig .
[ fig_hist ] , the eigenmode shape shows no noticeable distortion at low amplitude , but self - localizes and eventually breaks up at high amplitude . in this break - up , many smaller dbs are formed , a process arguably reminiscent of multiple filamentation in nonlinear optics @xcite , which also move in space .
thus , in future experiments on this system ( e.g. using photoacoustic techniques , as in @xcite ) , dbs could be detected by impulsively exciting a large spot on the substrate surface , and observing the formation of smaller , highly localized wave packets .
however , we note that we expect that losses , such as may be caused by dissipation and radiation into the substrate , may play an important role in this case , and should be considered in future models .
in this work , we have demonstrated the existence of discrete breathers in a mass - in - mass chain that models a locally resonant , microscale granular metamaterial composed of microspheres adhered to a substrate .
this chain consists of a linearly - coupled main chain ( representing the substrate ) with nonlinear local resonators that follow the hertzian contact law ( representing the microspheres ) . after fitting the two independent model parameters to a system studied in previous works ,
we analyze the resulting energy trapping , in the form of discrete breathers , theoretically , in the anti - continuum and continuum limits of intersite coupling , as well as numerically , by using the intersite coupling stiffness and oscillation frequency as continuation parameters .
finally , we simulate the formation and filamentation in the form of discrete breathers of single - humped wavepackets using direct numerical integration of the equations of motion .
the simulations suggest that discrete breathers may be observed in experiments on microscale granular metamaterials by generating a long - wavelength saw at high amplitude , and detecting its breakup into many small dbs .
we expect this work to aid in future studies of nonlinear , microscale granular systems , as well as to the more general class of media composed of a linear material with local nonlinear resonant attachments . indeed
, the dynamics and interactions of discrete breathers ( as well as their potential filamentation and dispersion ) in one- to three - dimensional analogs of such systems , may yield numerous topics that could be both theoretically intriguing , as well as experimentally accessible for further study .
also , it should be borne in mind that here we only explored the focusing variants of the relevant models and their bright solitons . yet , in line with recent explorations in granular crystals @xcite and in systems with resonators @xcite , the self - defocusing case may also be interesting and equally accessible in different parametric regimes . n.b . and s.p.w .
acknowledge support from the national science foundation ( grant no .
cmmi-1333858 ) and the army research office ( grant no . w911nf-15 - 1 - 0030 ) .
p.g.k . acknowledges support from the army research office ( grant no .
w911nf-15 - 1 - 0604 ) and the air force office of scientific research ( grant no .
fa9550 - 12 - 10332 ) .
the local resonator mass is taken to be the mass of a single microsphere , given by @xmath125 , where @xmath126 is the density of the microsphere material .
we estimate the main chain mass as that of a rectangular region of the substrate beneath a single microsphere . since rayleigh saws have a characteristic decay length on the order of the wavelength @xmath127 @xcite , this region has mass @xmath128 , where @xmath129 is the density of the substrate material .
the mass ratio is then given by @xmath130 where the quantity in parentheses is expected to be of order @xmath131 for most material combinations , and the quantity @xmath132 is at least of order @xmath133 , since we consider long wavelengths .
therefore , we estimate @xmath134 .
the local resonator stiffness is taken from hertzian contact theory @xcite , linearized about the static overlap distance @xmath135 due to adhesion , using the same model as ref .
this stiffness is given by @xmath136 .
here , @xmath137 , with @xmath138^{-1 } $ ] , where @xmath139 and @xmath140 are the young s modulus and poisson s ratio , respectively , with subscripts corresponding to the substrate and sphere materials @xcite .
the static overlap is @xmath141 , where @xmath142 is the force due to adhesion as per the derjaguin - muller - toporov ( dmt ) adhesive elastic contact model @xcite , and @xmath143 is the work of adhesion between the sphere and substrate materials . the main chain stiffness is estimated in a similar manner to the mass @xmath144 , using an element of the substrate with length @xmath23 in the direction of saw propagation , and cross - sectional area @xmath145 .
the estimated stiffness is then @xmath146 .
after algebraic manipulation , the stiffness ratio can be written as @xmath147 for simplicity , we have chosen to use identical sphere and substrate materials ( so that @xmath148 and @xmath149 ) ; this approximation does not affect the generality of this rough estimate .
the first term in parentheses is of order @xmath150 for realistic values of poisson s ratio , i.e. @xmath151 , and the second term in parentheses is of order @xmath152 , for realistic values @xmath153 and @xmath154 , with @xmath155 , using si units . since quantity @xmath156 , we estimate @xmath157 .
define the residuals as @xmath158 substituting @xmath159 and @xmath160 of eqs . and , we obtain the following residuals organized by orders of @xmath69 : @xmath183 + c.c
. + \textrm{h.o.t}$ ] + if we set each order of @xmath69 to @xmath56 , we can define the coefficients , @xmath74 and parameters , @xmath79 and @xmath184 , such that @xmath185 , which should yield an accurate approximate solution to our original equations of motion .
we now list the heiarchy of solvability conditions : @xmath203 @xmath204 + finally , with these coefficients , we are left with two nls equations at @xmath80 of both @xmath205 and @xmath206 .
+ in @xmath205 , we have : @xmath207 and in @xmath206 , we have : @xmath208 in order for both nls equations to be satisfied , the coefficients in each must match .
thus , we require that @xmath209 so that we obtain a single nls equation .
this equation has the solution @xmath210 with @xmath211 and @xmath212 a free parameter .
m. hiraiwa , m. abi ghanem , s. p. wallen , a. khanolkar , a. a. maznev , and n. boechler , `` complex contact - based dynamics of microsphere monolayers revealed by resonant attenuation of surface acoustic waves , '' _ phys . rev
( accepted ) _ , available at arxiv:1510.05975 ( 2016 ) .
g. theocharis , n. boechler , p. g. kevrekidis , s. job , m. a. porter , and c. daraio , `` intrinsic energy localization through discrete gap breathers in one - dimensional diatomic granular crystals , '' _ phys .
e _ * 82 * , 056604 ( 2010 ) .
y. starosvetsky , m. a. hasan , a. f. vakakis , and l. i. manevitch , `` strongly nonlinear beat phenomena and energy exchanges in weakly coupled granular chains on elastic foundations , '' _
siam j. appl . math _ * 72 * , 1 ( 2012 ) . c. chong , f. li , j. yang , m. o. williams , i. g. kevrekidis , p. g. kevrekidis , and c. daraio , `` damped - driven granular chains : an ideal playground for dark breathers and multibreathers '' , phys .
e * 89 * , 032924 ( 2014 ) .
m. a. hasan , s. cho , k. remick , a. f. vakakis , d. m. mcfarland , and w. m. kriven , `` experimental study of nonlinear acoustic bands and propagating breathers in ordered granular media embedded in matrix , '' _ granul . matter _
* 17 * , 1 ( 2015 ) .
m. peeters , r. vigui , g. srandour , g.kerschen , and j. c.golinval , `` nonlinear normal modes , part ii : toward a practical computation using numerical continuation techniques , '' _ mech .
signal pr . _
* 23 * , 1 ( 2009 ) .
v. m. muller , b. v. derjaguin , and y. p. toporov , `` on two methods of calculation of the force of sticking of an elastic sphere to a rigid plane , '' _ colloids and surfaces _ * 7 * , 251 ( 1983 ) ; m. d. pashley , `` further consideration of the dmt model for elastic contact , '' _ colloids and surfaces _ * 12 * , 69 ( 1984 ) .
c. chong , f. li , j. yang , m. o. williams , i. g. kevrekidis , p. g. kevrekidis , and c. daraio , `` damped - driven granular chains : an ideal playground for dark breathers and multibreathers , '' phys .
e * 89 * , 032924 ( 2014 ) . | we report on the existence of discrete breathers in a one - dimensional , mass - in - mass chain with linear intersite coupling and nonlinear hertzian local resonators , which is motivated by recent studies of the dynamics of microspheres adhered to elastic substrates . after predicting theoretically the existence of the discrete breathers in the continuum and anticontinuum limits of intersite coupling
, we use numerical continuation to compute a family of breathers interpolating between the two regimes in a finite chain , where the displacement profiles of the breathers are localized around one or two lattice sites .
we then analyze the frequency - amplitude dependence of the breathers by performing numerical continuation on a linear eigenmode ( vanishing amplitude ) solution of the system near the upper band gap edge .
finally , we use direct numerical integration of the equations of motion to demonstrate the formation and evolution of the identified localized modes , including within settings that may be relevant to future experimental studies . |
nuclear resonant scattering ( nrs ) of synchrotron radiation ( sr ) has become an established method for the study of nuclear hyperfine interaction during the last two decades @xcite .
the spectrum is conventionally recorded as the time response of the nuclear ensemble following a short resonant synchrotron pulse , which simultaneously excites all resonant transitions between hyperfine - split nuclear sublevels .
the observed beating frequencies are characteristic for the hyperfine fields in the specimen . as an alternative to nuclear resonant forward scattering of sr in time domain ,
a heterodyne detection scheme was suggested @xcite .
the two scatterers , viz .
the one under investigation and as a reference sample a single - line mssbauer absorber , are mounted on a mssbauer drive .
the heterodyne spectrum is the full time integral of the delayed counts , plotted as a function of the doppler velocity of the reference sample .
an advantage of this experimental setup is the similarity of the spectra to those in the conventional energy - domain mssbauer spectroscopy @xcite .
although conventional mssbauer spectroscopy delivers similar information on hyperfine interactions , the special properties of sr like high collimation , high degree of polarization and high brilliance increase the number of possible applications of nrs of sr .
furthermore , the heterodyne setup allows for dense bunch modes of the synchrotron ( with bunch separation time much shorter than the nuclear lifetime ) , which are not suitable for time differential nrs experiments .
undistorted time integration of the nuclear response can only be performed if the large non - resonant intensity contribution is extinguished .
experimentally , this can be achieved by using radiation from a nuclear monochromator @xcite or by applying a polarizer / analizer setup @xcite .
an alternative approach , namely , `` stroboscopic detection '' , is based on appropriate time gating @xcite , i.e. , integration of the delayed time response in a periodic time window .
the period @xmath2 of the observation time window after the sr pulse is chosen so that @xmath3 falls within the frequency range of the hyperfine interactions in the investigated specimen .
this leads to new type of periodic resonances at certain doppler velocities that are shifted from the mssbauer resonances by @xmath4 , with @xmath5 being planck s constant , and @xmath6 an integer number indicating the stroboscopic order @xcite .
the period @xmath2 should be selected according to the hyperfine spectral range , the synchrotron bunch period and the detector dead time @xcite .
so far the theory of stroboscopic detection scheme has only been developed and discussed in detail for forward scattering geometry .
the several applications of nrs in surface and thin - film magnetism that make use of the grazing incidence geometry @xcite call for computer programs that easily allow to fit data obtained by stroboscopic detection as well . in this geometry ,
the interferences of the sr plane waves , scattered from the surface and interfaces of a stratified sample , provide information on the value , direction and topology of the internal fields in the sample with nanometer depth resolution .
recently , interesting experiments have been performed using stroboscopic detection in the grazing incidence case @xcite , which demonstrates the potential of this method .
grazing - incident nrs of sr , often called synchrotron mssbauer reflectometry ( smr ) @xcite , has been established in both time and angular regime @xcite , as time differential ( td ) and time integral ( ti ) smr , respectively . in the forward scattering channel ,
the prompt electronic scattering homogeneously contributes to the stroboscopic spectrum and does not affect the spectral shape . for other scattering channels , including grazing incidence reflection ,
the interference of the electronic and nuclear scattering provides further information .
the stroboscopic smr line shape may considerably differ from the forward mssbauer spectrum , calling for a specialized computer code .
the dynamical theory of x - ray scattering gives a self - consistent description of the radiation field in all scattering channels of the system of scatterers , taking all orders of multiple scattering into account .
theories that expand the coherent elastic scattering to the case of sharp nuclear resonances @xcite have been applied to various scattering geometries .
the simplest cases are the one - beam cases , such as forward and off - bragg scattering , and the two - beam cases , the bragg - laue scattering @xcite and the grazing incidence scattering @xcite . in the grazing incidence limit ,
an optical model was derived from the dynamical theory @xcite , which has been implemented in numerical calculations @xcite .
the reflectivity formulae given by and are suitable for fast numerical calculations in order to actually fit the experimental data @xcite and , as has been shown @xcite , this optical method is equivalent to that of the other approaches in the literature @xcite .
the aim of the present paper is to develop the concept of the heterodyne / stroboscopic detection and to establish the formula that can be applied to any scattering channel , like forward scattering , bragg , off - bragg and grazing incidence scattering .
this paper is organized as follows . in the second section , the heterodyne / stroboscopic intensity formula for the propagation of @xmath7photons in a medium containing both electronic and resonant nuclear scatterers
is derived .
the equivalence to the previously discussed calculations for the forward channel @xcite are shown and the important specific case of the stroboscopic grazing incidence reflection are outlined . in the third section ,
features of the grazing incidence case are demonstrated by least - squares fitted experimental stroboscopic smr spectra on isotope - periodic @xmath0 $ ] and antiferromagnetically ordered @xmath1 $ ] multilayer films .
the setup of a heterodyne / stroboscopic nrs of sr experiment includes two scatterers , the investigated specimen and an additional reference sample @xcite , the latter being mounted on a mssbauer drive ( in forward scattering geometry ) .
the mssbauer drive provides a doppler - shift @xmath8 of the nuclear energy levels , with @xmath9 @xmath10 and @xmath11 being the velocity of light , the velocity of the drive and the energy of the mssbauer transition , respectively .
the polarization dependence of the nuclear scatterers is described adopting the notation of , by @xmath12 transmissivity and reflectivity matrices commonly called _ scattering matrices_.
the scattering of the synchrotron photons on the specimen and the reference sample is described by the _
total scattering matrix _ @xmath13 , @xmath14 a product of the scattering matrices of the reference sample @xmath15 and of the investigated specimen @xmath16 @xcite in the energy domain .
the index @xmath17 specifies the open scattering channel @xcite .
the scattering matrix @xmath18 of the reference sample depends on the doppler - shifted energy @xmath19 , where the channel index @xmath20 is omitted for forward scattering .
note that both the electrons and the resonant mssbauer nuclei scatter the @xmath7photons coherently so the scattering matrices have a resonant nuclear and nearly energy - independent electronic contribution . at energies being far from the mssbauer resonances @xmath21 on a hyperfine scale , the individual scattering matrices @xmath22 , and thus their product @xmath23 in eq .
( [ tottrans ] ) , approach the non - resonant electronic contribution : @xmath24 since the reference is mounted in forward geometry , its scattering matrix @xmath25 is the matrix exponential @xmath26 , \label{blumeexp}\ ] ] where @xmath27 is the index of refraction , @xmath28 the thickness and @xmath29 the vacuum wave number of the incident radiation @xcite .
the index of refraction is related to the susceptibility matrix @xmath30 @xcite through @xmath31 where @xmath32 is the unit matrix and @xmath33 , with @xmath34 and @xmath35 being the density of the scattering centers and the @xmath12 coherent forward scattering amplitude @xcite , respectively .
the susceptibility is the sum of the electronic and the nuclear susceptibilities , @xmath36 with eqs .
( [ blumeexp ] ) , ( [ khidef ] ) and ( [ khisum ] ) , the transmissivity of the reference is expressed as a product of electronic and nuclear transmissivities , @xmath37 where @xmath38 @xmath39 is determined from the respective theory of wave propagation of channel @xmath20 ( forward , bragg - laue , grazing incidence , etc .
scattering ) , i.e. , from the dynamical theory @xcite . in forward scattering , due to the exponential expression in ( [ blumeexp ] ) ,
the total transmissivity @xmath40 is proportional to @xmath41 .
therefore , in this special case , the electronic scattering is a simple multiplicative factor , which does not affect the spectral shape . the intensity @xmath42 allowing for a general polarization state of the incident beam , the @xmath12 polarization density matrix @xmath43 ( ) , is given by @xmath44 .
\label{blumeint}\ ] ] the beating time response to a single short polychromatic photon bunch of sr is obtained by the fourier transform of the energy domain scattering matrices @xcite , @xmath45 \exp \left ( -i\frac{e}{\hbar } t\right ) , \label{timetrans}\ ] ] where , by subtracting the constant @xmath46 , the dirac delta like prompt @xmath47 and the delayed @xmath48 time responses are separated @xcite .
we note that eq .
( [ timetrans ] ) is valid only for delayed times @xmath49 after the sr bunch @xmath47 , but @xmath50 for @xmath51 ! in the same way as for eq .
( [ blumeint ] ) , the delayed intensity in time domain becomes @xmath52 .
\label{blumeinttim}\ ] ] for a heterodyne / stroboscopic nrs of sr experiment a time window function is introduced , which can be described by boxcar functions , namely , @xmath53 for @xmath54 and @xmath55 otherwise , with a time interval @xmath56 between the synchrotron bunches and an integer number @xmath6 . the periodic time window function is expanded in fourier series , @xmath57where @xmath58 is the angular frequency of the sr bunches @xcite .
the total delayed photon rate @xmath59 of one bunch is @xmath60the integral of the intensity @xmath61 times @xmath62 .
since there is no coherence between photons generated by different electron bunches , the integral of the contribution of one bunch reveals the correct contribution of multiple bunches with periodicity of @xmath56 . combining eqs .
( [ timetrans ] ) , ( [ blumeinttim ] ) , ( [ stroboint ] ) and ( [ fourwin ] ) , the delayed count rate can be written as @xmath63where @xmath64 \left [ t_{\tau } \left ( e , e_{v}\right ) -t_{\tau , \infty } \right ] \rho \right\ } \label{defalfa}\]]and @xmath65since the time window @xmath66 and the intensity @xmath67 are real functions , @xmath68 and @xmath69 hold , and eq .
( [ finstrobo ] ) can be rewritten as @xmath70the result in eqs .
( [ finstrobo])([epsdef ] ) is a direct generalization of the intensity formula ( [ blumeint ] ) to the heterodyne / stroboscopic nrs of sr for any observed channel @xmath17 in the applied experimental geometry .
this expression has already been derived for the case of forward scattering @xcite .
the @xmath71 term was called the `` heterodyne spectrum '' @xcite , while the @xmath72 terms were called `` stroboscopic resonances '' of order @xmath6 @xcite .
nevertheless , the stroboscopic resonances are not restricted to the forward scattering case .
they also appear in other experimental geometries , including , as we show below , in the grazing incidence scattering geometry . in
what follows , stroboscopic smr spectra will be discussed . in terms of the dynamical theory ,
grazing incidence is a two - beam case .
the @xmath73 transmission and the @xmath74 reflection channels are open @xcite , the latter being observed in smr .
close to the electronic total reflection , the reflected intensity is high .
therefore , smr is an experimentally fairly instructive special case .
the reflection from the surface of the specimen is a multiple coherent scattering process of the ( sr ) photons on atomic electrons and resonant mssbauer nuclei @xcite . like in the forward case
, this scattering is independent of the atomic positions in the reflecting medium , such that the scattering is described by its index of refraction @xmath75 @xcite .
henceforth , in compliance with the literature @xcite , in the general theory , the scattering matrix @xmath76 will be replaced by the @xmath12 reflectivity matrix @xmath77 , where @xmath78 is the angle of incidence .
this takes into account the interferences of the reflected radiation from the surfaces and interfaces between the layers with different refraction index .
the methods of calculating the reflectivity matrix can be found in the literature @xcite .
accordingly , the total scattering matrix of the specimen and the reference from eq .
( [ tottrans ] ) is @xmath79 similarly , for energies being far from the mssbauer resonances , eq .
( [ eltottrans ] ) reads @xmath80 inserting @xmath81 and @xmath82 into eq .
( [ finstrobo ] ) , the delayed count rate @xmath83 of the heterodyne / stroboscopic spectrum for grazing incidence ( stroboscopic smr intensity ) on the specimen is calculated .
combining eqs .
( [ sepel ] ) , ( [ expnukl ] ) and ( [ defalfa ] ) reveal @xmath84\right .
\notag \\ & & \qquad \times \left.\left [ r\left ( e\right ) \tilde{t}\left ( e - e_v\right ) -r_{\mathrm{el}}\right ] \rho \right\ } , \label{dlform}\end{aligned}\ ] ] where @xmath85 is the electronic absorption of the reference sample . for the sake of simplicity ,
the indices on the right hand side have been omitted , so that @xmath86@xmath87 , @xmath88 and @xmath89 .
note that all reflectivities are those of the specimen , and all transmissivities are those of the reference sample . with the relevant angular parameter @xmath78 for grazing incidence , eq .
( [ finstrobo ] ) reads @xmath90 the observed nuclear , as well as stroboscopic , resonances can be interpreted in a straightforward manner using eq .
( [ dlform ] ) . indeed , far from the resonances , @xmath91 and @xmath92 , and the differences in the square brackets in ( [ dlform ] ) vanish .
we expect a significant contribution to the energy integral only if at least one energy argument of each bracket is close to resonance , i.e. , either [ co1 ] @xmath93 or [ co2 ] @xmath94 are fulfilled , where @xmath95 is the energy of the _ _ i__@xmath96 mssbauer resonance of the specimen . the @xmath97 term of the sum in eq .
( [ simpstrobo ] ) contributes considerably if the doppler velocity is near to the corresponding shifted mssbauer resonance . in this case : [ fincond ] @xmath98 here , @xmath99 is a small deviation ( of the order of the resonance line width)from the energy @xmath100 or @xmath101 , ensuring the appearance of stroboscopic resonances also in grazing incidence geometry . in the case of @xmath71 ,
all four conditions of eqs .
( [ co1 ] ) and ( [ co2 ] ) may be true simultaneously .
this means that , for @xmath71 , nuclear scattering in both samples , i.e. , `` the radiative coupling of the samples '' @xcite , also contributes .
hence , the dynamical line broadening ( coherent speed - up ) is the most effective in the heterodyne spectrum (= baseline and resonances of stroboscopic order 0 ) . in order to perform computer simulations of stroboscopic spectra , eqs .
( defalfa ) and ( [ dlform ] ) were calculated for the forward scattering and smr cases , respectively .
( [ defalfa ] ) , ( [ dlform ] ) and ( riet2 ) were implemented in the evaluation program effi @xcite .
this program allows for least - square fitting of stroboscopic spectra .
moreover , they can be fitted _ simultaneously _ with other types of spectra of the same specimen , such as forward scattering , grazing incidence , conventional mssbauer and other spectra of the implemented theory @xcite .
this way , the fit constraints on the common parameters become very general , as already described @xcite .
in order to test the feasibility of this new reflectometric scheme , we investigated two film specimens , a @xmath102 isotopic and a @xmath103fe / cr antiferromagnetic multilayer , in grazing incidence reflection geometry , using the @xmath104 mssbauer transition of @xmath103fe nuclei .
the experiments were performed at the bl09xu nuclear resonance beam line of spring-8 @xcite .
the experimental setup is shown in fig .
[ expsetup ] .
the synchrotron was operated in the 203-bunch mode , corresponding to a bunch separation time of @xmath105 .
the sr was monochromated by a si(422)/si(12 2 2 ) double channel - cut high resolution monochromator with @xmath106 resolution .
it was incident on the k@xmath107[@xmath103fe(cn)@xmath108 single line pelleted reference sample of effective thickness 11 , and on the multilayer specimen downstream mounted in grazing incidence geometry ( fig .
[ expsetup ] ) .
the mssbauer drive was operated in constant acceleration mode , with a maximum velocity of @xmath109 .
this maximum was calibrated by fitting the velocity separation of the stroboscopic orders in a forward scattering stroboscopic spectrum of a single line @xmath110-enriched stainless steel absorber @xcite .
the delayed radiation was detected using three @xmath111 dead time hamamatsu avalanche photo diodes ( apd ) in series .
to record the delayed intensity , a two - dimensional data acquisition system was used .
each count was indexed according to the time elapsed after the synchrotron pulse ( 1024 channels ) , as well as to the velocity of the reference ( 1024 channels ) .
these stroboscopic smr data were time integrated using appropriate time windows of @xmath112 period and @xmath113 length @xcite . since the energy is measured in mm / s , the shift of the first stroboscopic order , eq .
( [ epsdef ] ) , can be rewritten as @xmath114 = 1000\frac{\lambda ~\left [ \mathrm{nm}\right ] } { t_p~\left [ \mathrm{ns}\right ] } .
\label{stroboshift}\ ] ] with the wavelength @xmath115 for the mssbauer transition of @xmath103fe , the separation between the neighbouring stroboscopic orders can be calculated to be @xmath116 .
note that this is the range of the hyperfine splitting in case of @xmath117 ( outer line separation is @xmath118 at room temperature ) , and the stroboscopic orders would only slightly overlap in case of a sample of low effective thickness in forward scattering .
however , in case of grazing incidence near the critical angle of total external reflection due to the enhanced nuclear and electronic multiple scattering , the mssbauer lines become extremely broad and a strong overlap of the stroboscopic orders is expected .
this interference and partial overlap are manifested in rather complex resonance line shapes and an intriguing angular dependence of the delayed intensity in the various stroboscopic orders .
both multilayers were prepared under ultra - high vacuum conditions by molecular beam epitaxy at the imbl facility in iks leuven .
the@xmath119}_{10}$ ] was prepared at room temperature onto a zerodur glass substrate .
the first layer and all other @xmath110-layers were 95.5% isotopically enriched , and were grown from a knudsen cell .
the natural fe layers , which have a @xmath110-concentration of 2.17% , were grown from an electron gun source .
the nominal layer thickness was @xmath120 throughout the multilayer stack for both @xmath121 and @xmath110 .
conversion electron mssbauer spectra showed a pure @xmath117 spectrum .
this spectrum was compared to a transmission mssbauer spectroscopy spectrum of a natural iron calibration specimen , which was provided by amersham .
both hyperfine magnetic fields were fitted to be identical within the experimental error of 0.04% , and no sign of any second phase contamination was found .
preparation and characterization of the mgo(001)/[@xmath103fe / cr]@xmath122 multilayer sample has been described earlier @xcite .
the layering was verified as epitaxial and periodic , with thicknesses of @xmath123 for the @xmath103fe layer , and @xmath124 for the cr layer .
squid magnetometry showed dominantly antiferromagnetic coupling between neighboring fe layers . according to previous studies on this multilayer @xcite ,
the magnetizations in fe align to the [ 100 ] and [ 010 ] perpendicular easy directions in remanence , respectively corresponding to the [ 110 ] and [ @xmath12510 ] directions of the mgo substrate .
the layer magnetizations were aligned antiparallel in the consecutive fe layers by applying a magnetic field ( 1.6 t ) above the saturation value ( 0.96 t ) in the fe[010 ] easy direction of magnetization , and then releasing the field to remanence .
this alignment is global , the antiferromagnetic domains were only different in the layer sequence of the parallel / antiparallel orientations @xcite . since in a @xmath102 isotope - periodic multilayer the hyperfine field of @xmath110
is that of @xmath117 throughout the sample , this multilayer is particularly suitable for studying the modification of the resonance line shapes due to interference between nuclear and electronic scattering @xcite .
[ fefe ] shows results for the multilayer saturated in a transversal magnetic field of 50 mt .
panel _ a _ and _ b _ give the prompt electronic and delayed tismr curves , respectively .
the stroboscopic smr spectra at the angles indicated by the arrows are given in panel _
c _ to _ e_. the peak in the delayed reflectivity at the total reflection angle in panel _
b _ is a special feature of smr described earlier @xcite . in panels
_ c _ to _ e _ , the four resonance lines of the @xmath127 and @xmath128 stroboscopic orders ( right and left sides , respectively ) partially overlap with the @xmath129 order in the central part of the spectrum .
the delicate interplay between electronic and nuclear scattering is demonstrated by the considerable difference between the stroboscopic smr spectra _
c _ to _ e _ in fig .
[ fefe ] , which are taken at only slightly different grazing angles .
in contrast to the symmetric forward scattering spectra @xcite , the stroboscopic smr spectra are asymmetric due to the interference between the electronic and nuclear scattering .
they also display both `` absorption - like '' and `` dispersion - like '' resonance line shape contributions . in case of decreased nuclear scattering strength and of the same electronic reflectivity ( cf .
panels _ d _ and _ e _ in fig .
[ fefe ] ) , the signal to baseline ratio of the central part ( heterodyne spectrum ) decreases as compared to the signal to baseline ratio of stroboscopic orders @xmath130 in the spectrum wings .
the full lines are simultaneous least squares fits , using the theory outlined above and the computer code effi @xcite .
the interference between nuclear and electronic scattering makes it possible to fit the layer structure in this isotope - periodic multilayer .
the fitted value of the total thickness of pure @xmath117 is @xmath131 comprised of nine times @xmath132 of @xmath121 and @xmath133 of @xmath134 with @xmath135 common roughness at the interfaces . in order to achieve the simultaneous fit , displayed by the full line in figure [ fefe ]
, we had to assume that half a bilayer on top and bottom ( @xmath121 and @xmath110 , respectively ) was modified .
the transversal hyperfine magnetic field was fixed to @xmath136 in the nine @xmath110/@xmath121 bi - layers in the middle of the multilayer , which is the room temperature value for @xmath117 .
[ fecraf ] and fig .
[ fecrfm ] display similar sets of spectra of an @xmath103fe / cr antiferromagnetically coupled epitaxial multilayer on mgo(001 ) .
the dots are the experimental data points , while the continuous lines are simultaneous fits to a model structure of @xmath137 @xmath138 { } ^{}_{20}\:\!$ ] , based on the respective theory .
non - resonant reflectivity , tismr and stroboscopic smr spectra were recorded first with the fe layer magnetizations parallel / antiparallel ( fig .
fecraf ) to the @xmath139vector of the sr beam .
the stroboscopic spectra were taken at the angles of total reflection ( c ) , at the antiferromagnetic ( d ) and at the structural bragg peak ( e ) positions .
after this , a magnetic field of 20 mt was applied to the multilayer in longitudinal direction .
this is known to flop the magnetizations to the perpendicular fe(010 ) easy axis of the magnetization @xcite .
non - resonant reflectivity , tismr and stroboscopic smr spectra at the same angular positions were again collected ( fig .
[ fecrfm ] ) . the major difference between figs .
[ fecraf ] and [ fecrfm ] is the presence , respectively absence , of the af bragg peak in the delayed reflectivity curve b. this antiferromagnetic alignment , i.e. , the longitudinal hyperfine field of alternating sign in consecutive fe layers , is justified by the simultaneous fit in fig .
[ fecraf ] . in fig .
fecrfm , the fitted fe magnetizations are perpendicular to the wave vector of the sr . indeed , the scattering amplitudes depend on the angle of the wave vector and the direction of the hyperfine magnetic field . in the case of perpendicular orientation
, this angle is 90 degrees for consecutive layer magnetizations and no af contrast can be observed . in case of parallel / anti parallel orientations , however , the angles with respect to the wave vector of sr are 0 and 180 degrees , respectively .
therefore , the hyperfine contrast is present and the af bragg peak is visible in panel b of fig .
[ fecraf ] .
the count rate at the baseline of a stroboscopic smr spectrum , measured at a certain grazing angle @xmath78 , is closely related to the tismr spectrum at this angle .
therefore , the respective experimental count rates of the stroboscopic smr spectrum at the af bragg peak positions ( panel d ) differ by almost two orders of magnitude .
spectrum 3d is also the only spectrum for which no considerable enhanced dynamic broadening can be observed .
note that , in panels d , the zeroth order resonances are considerably enhanced with respect to the @xmath140order stroboscopic resonances .
this can be explained by an enhanced radiative coupling of the samples . since the radiative coupling does not contribute to the @xmath140order stroboscopic resonances
, it only influences the baseline and the central resonances . at the multilayer bragg reflections ( panel e ) , and at the total reflection peak ( panel c )
, the suppression of the higher stroboscopic orders is much smaller , which means that the radiative coupling term is not dominating here .
these spectra also show a left / right asymmetry due to the variation of the phase of the total scattering amplitude with energy .
this latter allows for phase determination of the scattering amplitude from a set of stroboscopic smr spectra , which work will be published later .
in summary , the concept of heterodyne / stroboscopic detection of nuclear resonance scattering was outlined for a general scattering channel , with special emphasis on the grazing incidence reflection case . in any non - forward scattering channel , the electronic scattering influences the nrs spectral shape , while in forward scattering , this is a mere multiplicative factor .
the interplay between electronic and nuclear scattering , as a function of the scattering angle , facilitates the determination of the electronic and nuclear scattering amplitudes .
the code of the present theory has been merged into the effi program @xcite , and was used in simultaneous data fitting of x - ray reflectivity , time integral reflectivity and stroboscopic smr spectra . similar to time differential smr , stroboscopic smr spectra have been shown to be sensitive to the direction of the hyperfine fields of the individual layers .
therefore , it is possible to apply this method to the study of magnetic multilayers and thin films . the experiments on @xmath141 _ { 20}$ ] and @xmath0 _ { 10}$ ] multilayers demonstrated that stroboscopic detection of synchrotron mssbauer reflectometry of @xmath103fe - containing thin films is feasible in dense bunch modes , which are not necessarily suitable for time differential nuclear resonance scattering experiments on @xmath103fe .
the authors gratefully acknowledge the beam time supplied free of charge by the japan synchrotron radiation institute ( jasri ) for experiment no : 2002b239-nd3-np .
our gratitude goes to dr .
baron ( spring-8 , jasri ) for kindly supplying the fast hammamatsu apd detectors and dr .
johan dekoster ( iks leuven ) for preparing the multilayer samples for the experiment
. support by the flemish - hungarian inter - governmental project bil14/2002 , the dynasync framework six project of the european commission ( contract number : nmp4-ct-2003 - 001516 ) the fund for scientific research
flanders ( g.0224.02 ) , the inter - university attraction pole ( iuap p5/1 ) and the concerted action of the kuleuven ( goa/2004/02 ) is gratefully acknowledged .
l. dek and r. callens thank the deutscher akademischer austauschdienst ( daad ) and the fwo - flanders , respectively , for financial support . _
{ 10}$ ] isotopic multilayer at grazing angles indicated by the arrows .
vertical dotted lines in panels c ) to e ) indicate the center of the zero and @xmath130 order stroboscopic bands separated by @xmath142 for the applied observation window period . ] _ { 20}$ ] antiferromagnetic multilayer at various angles indicated by arrows in b ) .
the consecutive fe layer magnetizations are aligned parallel / antiparallel with to the sr beam .
vertical dotted lines in panels c ) to e ) indicate the center of the zero and @xmath130 order stroboscopic bands separated by @xmath143 for the applied observation window period . ] _ { 20}/\mathrm{mgo}$ ] antiferromagnetic multilayer at various angles indicated by arrows .
the consecutive fe layer magnetizations are aligned perpendicular to the sr beam .
vertical dotted lines in panels c ) to e ) indicate the center of the zero and @xmath130 order stroboscopic bands separated by @xmath142 for the applied observation window period . ] | the theory of heterodyne / stroboscopic detection of nuclear resonance scattering is developed , starting from the total scattering matrix as a product of the matrix of the reference sample and the sample under study .
this general approach holds for any dynamical scattering channel .
the forward channel , which is discussed in detail in the literature , reveals the speciality that electronic scattering causes only an energy independent diminution of the intensity .
for all other channels , complex resonance line shapes in the heterodyne / stroboscopic spectra as a result of interference of electronic and nuclear scattering is encountered .
the grazing incidence case is evaluated and described in detail .
experimetal data of classical grazing incidence reflection and their stroboscopic detection on @xmath0 _ { 10}$ ] and antiferromagnetic @xmath1 _ { 20}$ ] multilayers are fitted simultaneously .
bottyn callens coussement major nasu serdons spiering yoda the theory of heterodyne / stroboscopic detection of nuclear resonance scattering is developed for various dynamical scattering channels .
the grazing incidence case is discussed in detail and is experimentally demonstrated on magnetic multilayers . |
on 2010 june 15 ( hjd@xmath6 5362.967 @xmath7 hjd@xmath8 ) , the microlensing observations in astrophysics ( moa ) collaboration @xcite detected a new microlensing event moa-2010-blg-310 at ( r.a .
, decl . ) = ( 18@[email protected]@xmath1198 , -25@xmath12 ) ( j2000.0 ) , ( l , b ) = ( 5.17 , -2.96 ) , along our line of sight toward the galactic bulge .
moa announced the event through its email alert system and made the data available in real - time . within a day
, this event was identified as likely to reach high magnification .
because of moa s real - time alert system , the event was identified sufficiently far in advance to allow intensive follow up observations over the peak .
the observational data were acquired from multiple observatories , including members of the moa , ogle , @xmath13fun , planet , robonet @xcite , and mindstep collaborations . in total , sixteen observatories monitored the event for more than one night , and thus their data were used in the following analysis . among these , there is the moa survey telescope ( 1.8 m , moa - red - band ] , new zealand ) and the b&c telescope ( 60 cm , @xmath14 , @xmath15 , new zealand ) ; eight of the observatories are from @xmath13fun : auckland ( ao , 0.4 m , wratten # 12 , new zealand ) , bronberg ( 0.36 m , unfiltered , south africa ) , ctio smarts ( 1.3 m , @xmath14 , @xmath15 , @xmath16 , chile ) , farm cove ( fco , 0.36 m , unfiltered , new zealand ) , kumeu ( 0.36 m , unfiltered , new zealand ) , molehill ( mao , 0.3 m , unfiltered , new zealand ) , vintage lane ( vlo , 0.4 m , unfiltered , new zealand ) , and wise ( 0.46 m , unfiltered , israel ) ; three are from planet : kuiper telescope on mt .
bigelow ( 1.55 m , @xmath15 , arizona ) , canopus ( 1.0 m , @xmath15 , australia ) , and saao ( 1.0 m , @xmath15 , south africa ) ; one is from robonet : liverpool ( 2.0 m , @xmath15 , canaries ) ; and one is from mindstep : la silla ( 1.5 m , @xmath15 , chile ) .
the event also fell in the footprint of the ogle iv survey ( 1.3 m , @xmath15 , chile ) , which was in the commissioning phase in 2010 .
the observatory and filter information is summarized in table [ tab : data ] .
in particular , observations from the moa survey telescope , moa b&c , planet canopus , @xmath13fun bronberg , and @xmath13fun vlo provided nearly complete coverage over the event peak between hjd@xmath17 and hjd@xmath18 . the moa and b&c data were reduced with the standard moa pipeline @xcite .
the data from the @xmath13fun observatories were reduced using the standard dophot reduction @xcite , with the exception of bronberg and vlo data , which were reduced using difference image analysis ( dia ; * ? ? ?
* ; * ? ? ?
data from the planet and robonet collaborations were reduced using pysis2 pipeline @xcite .
data from mindstep were also initially reduced using the pysis2 pipeline .
the ogle data were reduced using the standard ogle pipeline @xcite .
both the moa and mindstep / la silla data were reduced in real - time , and as such the initial reductions were sub - optimal .
in fact , the original moa data over the peak were unusable because they were corrupted .
after the initial analysis , both the mindstep and moa data were rereduced using optimized parameters .
to determine the intrinsic source color , we construct a color - magnitude diagram ( cmd ) of the field of view containing the lensing event ( figure [ fig : cmd ] ) based on @xmath14- and @xmath15-band images from ctio smarts andicam camera .
the field stars in the cmd are determined from three @xmath14-band images and multiple @xmath15-band images .
four @xmath14-band images were taken ; however , only three of the images are of sufficient quality to contribute to the cmd .
we visually checked each of the three images to make sure that there were no obvious defects such as cosmic - ray events in the images . from the fit to the light curve , we find the instrumental magnitude of the source is @xmath19 with an uncertainty of 0.05 mag due to differences between the planet and point lens models as well as the error parameterizations . in the top right - hand corner of figure [ fig : cmd ] , the red dot at @xmath20 marks the centroid of the red clump .
the intrinsic color and magnitude of the clump are @xmath21 @xcite . using the offset between the intrinsic magnitude of the clump and
the observed , instrumental magnitude , we can then calibrate the magnitude of the source to find @xmath22 .
the color of the source is normally estimated from @xmath14- and @xmath15-band images using the standard technique in @xcite . however , with only one highly - magnified @xmath14-band image , this method is unreliable , so we use an alternative technique to determine the instrumental @xmath23 color by converting from the instrumental @xmath24 color . using the simultaneous ctio @xmath15- and @xmath16-band observations , we measure the instrumental @xmath24 color of the source by linear regression of @xmath16 on @xmath15 flux at various magnifications during the event .
we then construct a @xmath25 instrumental color - color diagram from stars in the field ( bottom panel of figure [ fig : cmd ] ) .
the stars are chosen to be all stars seen in all three bands with instrumental magnitude brighter than @xmath26 ( note that the field of view for @xmath16-band observations is @xmath27 compared to @xmath28 for the optical bands ) .
the field stars form a well defined track , which enables us to estimate the @xmath23 source color from the observed @xmath24 source color .
this yields @xmath29 .
note that this method would not work for red stars , @xmath30 , because for these red stars , the @xmath25 relation differs between giants and dwarfs @xcite .
however , the observed color is well blueward of this bifurcation .
there is also a spectrum of the source taken at hjd@xmath31 @xcite .
the `` spectroscopic '' @xmath32 reported in that work is 0.77 , in good agreement with the value calculated here .
a casual inspection of the light curve does not show any deviations from a point lens , so we begin by fitting a point lens model to the data .
a point lens model is characterized by three basic parameters : the time of the peak , @xmath33 , the impact parameter between the source and the lens stars , @xmath34 , and the einstein timescale , @xmath35 .
since @xmath34 is small , the finite size of the source can be important . to include this effect
, we introduce the source size in einstein radii , @xmath36 , as a parameter in the model .
additionally , we include limb - darkening of the source .
the temperature of this slightly evolved source was determined from the spectrum to be @xmath37 5460 k by @xcite .
using @xcite we found the limb - darkening coefficients to be @xmath38 , @xmath39 , and @xmath40 , assuming a microturbulent velocity = 1 km s@xmath41 , log @xmath42 = 4.0 , solar metallicity , and @xmath43k , which is the closest grid point given the @xcite measurements .
we then convert @xmath44 , @xmath45 , and @xmath46 to the form introduced by @xcite @xmath47 to obtain @xmath48 , @xmath49 , and @xmath50 . because the various data sets are not on a common flux scale , there are also two flux parameters for each data set , @xmath51 and @xmath52 , such that @xmath53 where @xmath54 is the predicted magnification of the model at time , @xmath55 , and includes the appropriate limb - darkening for data set @xmath56 .
the source flux is given by @xmath57 , and @xmath58 is the flux of all other stars , including any light from the lens , blended into the psf ( i.e. , the `` blend '' ) .
as is frequently the case for microlensing data , the initial point lens fit reveals that the errors calculated for each data point by the photometry packages underestimate the true errors . additionally , there are be outliers in the data that are clearly seen to be spurious by comparison to other data taken simultaneously from a different site
simply taking the error bars at face value would lead to biases in the modeling . because the level of systematics varies between different data sets , underestimated error bars
can give undue weight to a particular set of data .
additionally , if the errors are underestimated , the relative @xmath59 between two models will be overestimated , making the constraints seem stronger than they actually are . to resolve these issues ,
we rescale the error bars using an error renormalization factor ( or factors ) and eliminate outliers .
we begin by fitting the data to a point lens to find the error renormalization factors .
we remove the outliers according to the procedure described below based on the renormalized errors , refit , and recalculate the error renormalization factors .
we repeat this process until no further outliers are found .
to first order , we can compensate for the underestimated error bars by rescaling them by a single factor , @xmath60 .
the rescaling factor is chosen for each data set , @xmath56 , such that @xmath61 , producing a @xmath62 per degree of freedom of @xmath63 .
we will refer to this simple scheme for renormalizing the errors as `` 1-parameter errors '' .
alternatively , we can use a more complex method to renormalize the errors , which we will call `` 2-parameter errors '' .
this method was also used in @xcite and @xcite . to renormalize the errors
, we rank order the data by magnification and calculate two factors , @xmath60 and @xmath64 , such that @xmath65 where @xmath66 is the original error bar and @xmath67 is the new error bar and the calculations are done in magnitudes .
the index @xmath56 refers to a particular data set , and @xmath68 refers to a particular point within that data set .
the additional term , @xmath69 , enforces a minimum uncertainty in magnitudes , because at high magnification , the flux is large , so the formal errors on the measurement can be unrealistically small .
the error factors , @xmath60 and @xmath64 , are chosen so that @xmath70 for points sorted by magnification increases in a uniform , linear fashion and @xmath63 @xcite . not all data sets will require an @xmath64 term ; it is only necessary in cases for which @xmath70 has a break because the formal errors are too small when the event is bright .
note that the @xmath64 factor will be primarily affected by the points taken when the event is bright , whereas @xmath60 is affected by all points , so if there are many more points at the baseline of the event , these will dominate the calculation of @xmath60 . to remove outliers in the data we begin by eliminating any points taken at airmass @xmath71 or during twilight .
additional outliers are defined as any point more than @xmath72 from the expected value , where @xmath73 is determined by the number of data points , such that fewer than one point is expected to be more than @xmath72 from the expected value assuming a gaussian error distribution .
the normal procedure is to compare the data to the `` expected value '' from a point lens fit .
we do this for data in the wings of the event , @xmath74hjd@xmath75 or @xmath74hjd@xmath76 , when we do not expect to see any real signals .
however , the peak of the event , @xmath77hjd@xmath78 , is when we would expect to see a signal from a planet if one exists .
a planetary signal would necessarily deviate from the expectation for a point lens , so in this region instead of comparing to a point lens model , we use the following procedure to identify outliers : 1 .
for each point , we determine whether there are points from any other data set within 0.01 days .
if there are more than two points from a given comparison dataset in this range , we keep only the point(s ) immediately before and/or after the time of the point in question ( i.e. , a maximum of 2 points ) .
the point is not compared to other points from the same data set.[step3 ] 2 .
if there are matches to at least two other data sets , we proceed to ( [ step6 ] ) to determine whether or not the point is an outlier .
otherwise , we treat the point as good .
we then determine the mean of the collected points , @xmath13 , including the point in question , by maximizing a likelihood function for data with outliers @xcite : @xmath79 where @xmath80 and @xmath81 is the datum and @xmath82 is its renormalized error bar .
if the flux is changing too rapidly for the points to be described well by a mean , we fit a line to the data .
[ step6 ] 4 .
we compare the point to the predicted value , using the likelihood function to see if it is an outlier : + \(a ) if there is only one likelihood maximum , we calculate @xmath83 . if @xmath84 , the point is rejected as an outlier .
+ \(b ) if there is more than one maximum , and the point in question falls in the range spanned by the maxima , we assume the point is good .
if it falls outside the range , we calculate @xmath83 using the nearest maximum to determine @xmath13 .
if @xmath84 , we flag the point as an outlier .
note that for this procedure , we use the point lens model values of @xmath85 and @xmath86 to place all of the data sets on a common flux scale .
this procedure is more complicated than usual , but because we compare the points only to other data , rather than some unknown model , it provides an objective means to determine whether a point is an outlier without destroying real signals corroborated by other data .
we also visually inspect each set of points to confirm that the algorithm works as expected .
because finite source effects are significant in this event , one might expect slight differences among data sets due to the different filters , so as part of this visual inspection we also checked that this did not play a significant role
. for both 1-parameter and 2-parameter errors , table [ tab : data ] lists the error normalization factors and number of points for each observatory that survived the rejection process . which of these error parameterizations correctly describes the data depends on the nature of the underlying errors .
in principle , the noise properties of the data are fully described by a covariance matrix of all data points , but we are unable to calculate such a matrix .
instead , we have two different error parameterizations .
the @xmath60 factor is calculated primarily based on baseline data for which statistical errors dominate . in contrast
, the @xmath64 factor is heavily influenced by points at the peak of the event when systematic errors are important .
thus , 2-parameter errors better reflect the systematic errors , whereas 1-parameter errors better reflect the statistical errors .
correlated errors often have a major impact on our ability to determine whether or not the planetary signal in this event is real . we know that correlations in the microlensing data exist , but there has not been a systematic investigation of this in the microlensing literature .
correlated errors ( red noise ) are generally thought of as reducing the sensitivity to signals , because successive points are not independent , giving related information .
but in fact , sharp , short timescale signals are not degraded by correlated noise and may still be robustly detected .
consider the case of a short - timescale signal superimposed on a long - timescale correlation .
then a model may reproduce the short - timescale signal leading to an improvement in @xmath62 without actually passing through the data because of the overall offset caused by the correlations .
now suppose that the correlated , red noise has a larger - amplitude than the white noise ( i.e. , statistically uncorrelated noise ) .
if we set the error bars by the large - amplitude deviations , which is correct for long timescales for which the data are uncorrelated , the significance of the short timescale jump will be diluted , possibly to the point of being considered statistically insignificant .
however , if the timescale of the signal is much shorter than the correlation length of the red noise , the significance of the signal should actually be judged against the white noise , since on that timescale , the red noise will only contribute a constant offset . in this case , we expect the planetary signal to be quite short , so if the systematic noise is dominated by correlated errors , the noise should be better described by the 1-parameter errors . because the source in this event crosses the position of the lens and there are no obvious deviations due to a planet
, we expect that any planetary signals will be due to very small caustics , which are detectable only at the limb - crossing times ( @xmath87 ) when the caustic passes onto and off - of the face of the source
. therefore , the timescale of such a perturbation will be very short , equal to @xmath88 times the caustic size @xmath89 , which is @xmath90 minutes .
in contrast , observed correlations in the microlensing data are typically on longer timescales , @xmath91hour@xmath92 ( based on our experience with microlensing data which are usually sampled with a frequency of @xmath93 minutes ) .
hence , the timescale of the signal is likely to be less than the timescale of the correlated noise . however , there are other sources of systematic errors that are unrelated to correlated noise such as flat - fielding errors .
if such errors dominate over correlated noise , then the 2-parameter errors are a better description of the error bars over the peak . because the systematic errors , correlated and uncorrelated , have not been studied in detail , we are unable to determine which is the dominant effect .
hence , we are also unable to determine which error prescription better describes our data .
we will begin by analyzing the light curve using 1-parameter errors .
we will then discuss how the situation changes for 2-parameter errors .
the best - fit point lens model and the uncertainties in the parameters are given in table [ tab : models1 ] .
this model is shown in figure [ fig : lc ] , and the residuals to this fit are shown in figure [ fig : res ] .
these exhibit no obvious deviations .
these fits confirm that finite source effects are important , since @xmath36 is well measured and larger than the impact parameter , @xmath34 .
we also fit a point lens model that includes the microlens parallax effect , which arises either from the orbital motion of the earth during the event or from the difference in sightlines from two or more observatories separated on the surface of the earth .
microlens parallax enters as a vector quantity : @xmath94 .
the addition of parallax can break the degeneracy between solutions with @xmath95 and @xmath96 , so we fit both cases .
parallax does improve the fit beyond what is expected simply from adding 2 more free parameters and shows a preference for @xmath95 .
however , we shall see in the next section that a planetary model without parallax produces an even better fit and adding parallax in addition to the planet gives only a small additional improvement .
we search for 2-body models over a broad range of mass ratios , from @xmath97 to @xmath98 .
for each value of @xmath99 , we chose a range for the projected separation between the two bodies , @xmath100 , for which the resulting caustic is smaller than @xmath36 and @xmath101 . for each combination of @xmath100 and @xmath99
, we allow the angle of the source trajectory , @xmath102 , to vary , seeding each run with values of @xmath102 from 0 to 360 degrees in steps of 5 degrees . for our models
, we use the map - making method of @xcite when the source is within two source radii of the position of the center of magnification . outside this time range
, we use the hexadecapole or quadrupole approximations for the magnification @xcite .
we used a markov chain monte carlo to find the best - fit parameters and uncertainties for each @xmath100 , @xmath99 combination .
the grid search reveals an overall improvement in @xmath62 relative to the point lens model .
we find three @xmath62 minima for different angles for the source trajectory .
for central caustics with planetary mass ratios , the caustic is roughly triangular in shape with a fourth cusp where the short side of the triangle intersects the binary axis ; the three trajectories roughly correspond to the three major cusps of the caustic .
an example caustic is shown in figure [ fig : traj ] along with the trajectories corresponding to the three minima .
the angles of the three trajectories are approximately @xmath103 and 235 degrees , and we will refer to them as trajectories `` a '' , `` b '' , and `` c '' , respectively . we then refine our grid of @xmath100 and @xmath99 around each of these three minima .
we repeat these fits accounting for various microlensing degeneracies .
first , we fit without parallax and assuming @xmath104 .
then , we add parallax and fit both with @xmath95 and @xmath96 to see if this degeneracy is broken .
finally , we fit 2-body lens models with @xmath105 and no parallax , since there is a well known microlensing degeneracy that takes @xmath106 .
the best - fit solution has @xmath107 and @xmath108 .
this reflects an improvement in @xmath62 of @xmath109 over the point lens solution .
there is no preference for @xmath104 over @xmath105 , but trajectory c is preferred by @xmath110 over trajectories a and b. the parameters and their uncertainties for the planet fits are given in table [ tab : models1 ] .
the mass ratio between the lens star and its companion is firmly in the planetary regime : @xmath2 .
furthermore , planetary mass ratios are clearly preferred over `` stellar '' mass ratios ( @xmath111 ) , which are disfavored by more than @xmath112 .
parallax further improves the fit by only @xmath113 and has little effect on the other parameters . to compare the point lens and planetary models , in figure [ fig : cumchi2 ] , we plot the `` @xmath62 residuals '' : the difference between the cumulative @xmath62 distribution and the expected value @xmath114 , i.e. each point
is expected to contribute @xmath115 .
for both the point lens and the planet fits , the distribution rises gradually over the peak of the event .
this is expected since 1-parameter errors do not account for correlated noise . however , in the @xmath62 residuals for the point lens , there is a jump seen at the time of the first limb - crossing .
this jump is even more pronounced when looking at the difference between the planet and point lens models .
the jump is caused by moa data at the time of the limb - crossing that do not fit the point lens well , thereby causing an excess increase in @xmath62 .
this is exactly the time when we expect to see planetary signals .
finally , given the extreme finite source effects in this event , we might be concerned that uncertainties in the limb - darkening coefficients due to uncertainties in the source properties could influence our conclusions .
the planetary signal has two components : a limb - crossing signal and an asymmetry .
the limb - darkening could influence the first signal , but not the second . to check that the limb - darkening coefficients do not significantly influence our results ,
we repeat the point lens fits allowing the limb - darkening coefficients to be free parameters . in all cases ( no parallax , parallax , 1-parameter or 2-parameter errors )
, the improvement to the fit from free limb - darkeing is @xmath116 , much smaller than the planetary signal .
furthermore , the value of @xmath117 decreases by @xmath118 , which is excluded by the measured source parameters .
thus , we conclude that our treatment of the limb - darkening is reasonable .
although @xmath119 appears to be significant , we are hesitant to claim a detection of a planet . the planetary signal is more or less equally divided between the jump at the first limb - crossing and a more gradual rise after the second limb - crossing ( see the third panel of fig . [
fig : cumchi2 ] showing the difference between the point lens and planet models ) .
one could argue that the gradual rise , due to a slight asymmetry in the planet light curve , could be influenced by large - scale correlations in the data . comparing figures [ fig : res ] and [ fig : cumchi2 ] shows that most of the signal at the first limb - crossing comes from only a few points .
a careful examination of the residuals in figure [ fig : res ] shows that while the residuals to the planet model are smaller than for the point lens model , they are not zero , and the didactic residuals do not go neatly through the difference between the models as they do for moa-2008-blg-310 @xcite .
hence , the evidence for the planet is not compelling .
if we repeat the analysis using 2-parameter errors , we find a similar planetary solution , although the exact values of the parameters are slightly different compared to other nights at baseline .
if the data from this night are removed , the remaining data predict a different solution with the planetary caustic crossing 18 days earlier . because this solution is pathological , we do not consider it further.[fn : awide ] ] .
the total signal from the planet is significantly degraded for 2-parameter errors , with only @xmath120 between the best - fit planet and point lens models .
table [ tab : models2 ] gives parameters for the complete set of point lens and planet fits for 2-parameter errors .
the residuals and error bars over peak may be compared to 1-parameter errors in figure [ fig : res ] .
we also show the @xmath62 residuals for 2-parameter errors in the bottom set of panels in figure [ fig : cumchi2 ] .
they are more or less flat over the peak , showing that they track the data well in this region .
the offset from zero is caused by systematics elsewhere in the light curve .
the difference plot ( bottom - most panel ) shows that the planet fit is still an improvement over the point lens fit , but the signal from the planet at the first limb - crossing is much weaker .
this is a natural consequence of 2-parameter errors , since the data at the peak of the event , where the planetary signal is seen , have much larger renormalized error bars than for 1-parameter errorshjd@xmath121 , when the main planetary signal is observed . ] .
regardless of the error renormalization , this planetary signal is smaller than the @xmath59 of any securely detected high - magnification microlensing planet .
previously , the smallest @xmath59 ever reported for a high - magnification event was for moa-2008-blg-310 with @xmath122 @xcite . @xcite discuss moa-2011-blg-293 , an event for which the authors argue the planet could have been detected from survey data alone with @xmath123 .
however , although the planet is clearly detectable at this level , it is unclear with what confidence the authors would have claimed the detection of the planet in the absence of the additional followup data , which increases the significance of the detection to @xmath124 . at an even lower level , @xcite find that a planetary companion to the lens improves the fit to macho-98-blg-35 at @xmath125 , but they do not claim a detection . as previously mentioned , @xcite suggest the minimum `` detectable '' planet will have @xmath126 .
however , this threshold has not been rigorously investigated ; the minimum @xmath59 could be smaller .
because of the tenuous nature of the planetary signal , we do not claim to detect a planet in this event , but since including a planet in the fits significantly improves the @xmath62 , we will refer to this as a `` candidate '' planet . finally , it is interesting to note that even though the @xmath59 for the planetary model is too small to be considered detectable , for both 1-parameter and 2-parameter errors the parameters of the planet ( @xmath100 and @xmath99 ) are well defined ( see tables [ tab : models1 ] and [ tab : models2 ] ) .
central caustics can be degenerate , especially when they are much smaller than the source size , so we might expect a wide range of possible mass ratios in this case since the limb - crossings are not well - resolved @xcite .
however , perhaps we should not be surprised that the planet parameters are well - constrained : both @xmath127 and @xmath128 are formally highly significant , which would plausibly lead to reasonable constraints on the parameters . in this case , because we believe that the signal could be caused by systematics , by the same token , the constraints on the parameters may be over - strong .
we conjecture that the limb - crossing signal does not constrain @xmath99 and that this constraint actually comes from the asymmetry of the light curve , since small , central caustics due to planets are asymmetric whereas those due to binaries are not .
because the source size , @xmath36 , is well measured , we can determine the size of the einstein ring , @xmath129 , and the relative proper motion between the source and the lens , @xmath130 from the following relations : @xmath131 keeping the limb - darkening parameters fixed , we find the best fit for the normalized source size to be @xmath132 for the planetary fits ; the value is comparable for the point lens fits .
we convert the @xmath23 color to @xmath133 using @xcite and obtain the surface brightness by adopting the relation derived by @xcite . combining the dereddened @xmath15 magnitude with this surface brightness yields the angular source size @xmath134 .
the error on @xmath135 combines the uncertainties from three sources : the uncertainty in flux ( @xmath136 ) , the uncertainty from converting @xmath23 color to the surface brightness , and the uncertainty from the @xcite estimate of @xmath137 .
the uncertainty of @xmath136 is 3% , which is obtained directly from the modeling output .
we estimate the uncertainty from the other factors ( @xmath138 ) to be 7% .
the fractional error in @xmath135 is given by @xmath139^{1/2}$ ] = 7% , which is also the fractional error of the proper motion @xmath13 and the einstein ring radius @xmath140 @xcite .
thus , we find @xmath141 mas and @xmath142 mas yr@xmath41 .
this microlensing system is close in projection to the globular cluster ngc 6553 .
the cluster is at ( r.a .
, decl . ) = ( 18@[email protected]@xmath1160 , -25@xmath145 ) ( j2000.0 ) , with a distance of 6.0 kpc from the sun and 2.2 kpc from the galactic center @xcite . the half - light radius @xmath146 of ngc 6553 is @xmath147 @xcite , which puts the microlensing event 6.5 half - light radii ( @xmath148 ) away from the cluster center . by plotting the density of excess stars over the background , we find that about 6% of the stars at this distance are in the cluster ( figure [ fig : clusterdensity ] ) .
whether or not the lens star is a member of the cluster can be constrained by calculating the proper motion of the lens star .
@xcite found a relative proper motion of ngc 6553 with respect to the bulge of @xmath149 mas yr@xmath41 .
the typical motion of bulge stars is about @xmath150 corresponding to about @xmath151 given an estimated distance of 7.7 kpc toward the source along this line of sight ( @xmath152 ; * ? ? ?
therefore the expected amplitude of the lens - source relative proper motion if the lens is a cluster member is @xmath153 , which is consistent with the measured value in section ( [ sec : temu ] ) . combining the measurement of the stellar density with the proper motion information , we find the probability that the lens is a cluster member is considerably higher than the nominal value based only on stellar density .
first , of order half the stars in the field are behind the source , whereas the lens must be in front of the source .
second , the lens - source relative proper motion is consistent with what would be expected for a cluster lens at much better than @xmath154 , which is true for only about 2/3 of events seen toward the bulge .
combining these two effects , we estimate a roughly 18% probability that the lens is a cluster member .
one way to resolve this membership issue is by measuring the true proper motion of the lens as was done for a microlensing event in m22 for which the lens was confirmed to be a member of the globular cluster @xcite .
for this event , we have calculated the relative proper motion of the lens and the source to be @xmath155 .
this is consistent with the expected value if the lens were a member of the cluster .
about 10 years after the event , the separation between the lens and the source star will be large enough to be measured with @xmath156 .
based on this followup observation , one will be able to clarify whether the lens is a member of the cluster by measuring the vector proper motion . if it is a member of the cluster , its mass may be estimated from @xmath157 where @xmath158 and @xmath159 is the source - lens relative parallax .
in addition , the metallicity of the lens could be inferred from the metallicity of the globular cluster .
we have found a candidate planet signal in moa-2010-blg-311 . the evidence in support of the planet is 1 .
the planet substantially improves the fit to the data , 2 .
in addition to a general improvement to the light curve , the planet produces a signal when we most expect it , i.e. the time of the first limb - crossing , 3 . the solution has a well - defined mass ratio and projected separation for the planet ( excepting the well known @xmath160 degeneracy ) .
the magnitude of the signal depends on whether the error bars are renormalized using 1-parameter ( @xmath128 ) or 2-parameter error factors ( @xmath127 ) .
we conservatively adopt @xmath127 as the magnitude of the signal , but note that if correlated errors are the dominant source of systematic uncertainty , @xmath128 should be adopted instead ( see section [ sec : errors ] ) .
regardless , this signal is too small to claim as a secure detection . examining the residuals to the light curve and
the @xmath62 residuals shows that the planetary signal is dispersed over many points at the peak of the light curve .
it comes from an overall asymmetry near the peak plus a few points at the limb - crossing .
because the signal is the sum of multiple cases of low - amplitude deviations , it is plausible that the microlensing model could be fitting systematics in the data , which is why the planet signal is not reliable . combined with other studies , this event suggests that central - caustic ( high - magnification ) events and planetary - caustic events require different detection thresholds .
the detection threshold suggested in @xcite of @xmath126 was made in the context of high - magnification events , and our experience so far is consistent with this .
a planet was clearly detectable in moa-2008-blg-310 with @xmath122 @xcite .
however , @xcite are uncertain if a planet would be claimed with @xmath123 for moa-2011-blg-293 . here , @xmath127 is definitely insufficient to detect a planet .
hence , the detection threshold for planets in high - magnification events is around or just below @xmath123 .
in contrast , the planetary caustic crossing in ogle-2005-blg-390 produced a clear signal of @xmath161 @xcite , and the planet would most likely be detectable if the error bars were 50% larger ( @xmath162 ) and might even be considered reliable if the error bars were twice as large .
we suspect the reason for the different detection thresholds is that the information about the microlens parameters and the planetary parameters comes from different parts of the light curve . for planetary - caustic events ,
the planet signal is a perturbation separated from the main peak .
thus , the microlens parameters can be determined from the peak data independently from the planetary parameters , which are measured from the separate , planetary perturbation .
in contrast , for high - magnification events , the planetary perturbation occurs at the peak of the event , so the microlens and planetary parameters must be determined from the same data . the detection threshold for planetary - caustic events will have to be investigated in more detail .
if it is truly lower than for high - magnification events , this is good news for second generation microlensing surveys since that is how most planets will be found in such surveys . at the same time
, it points to the continued need for followup data of high - magnification events since these seem to have a higher threshold for detection , requiring more data to confidently claim a planet .
this is an important consideration because high - magnification events can yield much more detailed information about the planets .
additionally , @xcite show that the magnitude of the planetary signal should decrease as the ratio between the caustic size and the source size ( @xmath163 ) decreases for the same photometric precision ( see also * ? ? ?
there are several cases of events for which @xmath164 : in this case , we have @xmath165 and @xmath128 ; the brown dwarf in moa-2009-blg-411l has @xmath166 and @xmath167 @xcite ; moa-2007-blg-400 has @xmath168 and @xmath169 @xcite ; and in the case of moa-2008-blg-310 , the value is @xmath170 with @xmath122 @xcite .
this sequence is imperfect , but the photometry in the four cases is far from uniform , and it seems that in general the trend suggested by @xcite holds in practice . finally , we note that a large fraction of the planet signal comes from the moa data , but in the original , real - time moa data , this signal would not have been detectable since the data were corrupted over the peak of the event .
it is only after the data quality was improved by rereductions , which in turn were undertaken only because the event became the subject of a paper , that we recovered the planet candidate .
this points to the importance of rereductions of the data when searching for small signals . in the current system of analyzing only the events with the most obvious signals ,
this is not much of a concern .
however , if current or future microlensing surveys are systematically analyzed to find signals of all sizes , this will become important .
color of the source .
the @xmath23 color of the source is then derived from the intersection of this line with the @xmath23@xmath24 relation established by the field stars ( green squares ) , producing the black dot in the bottom panel .
the position of the source in the cmd ( large black dot ) is given by this @xmath23 measurement and the source flux from the fits to the light curve .
the large red dot shows the centroid of the red clump .
[ fig : cmd ] ] , the error bars tend to be larger for 2-parameter errors than for 1-parameter errors
. careful inspection of the residuals to the point lens and planet models shows improvement around the times of the limb - crossings of the source star and also on the falling side of the light curve . in the middle panel of each set ,
the red lines show the difference between the best planetary and best point lens models .
the didactic residuals in the middle panel are the sum of the red line and the residuals to the best planet model ( bottom panel ) .
the dashed lines indicate the limb - crossing times .
the dotted lines in the top panels of each set indicate the time ranges shown in the bottom panels .
note that the scales for the top and bottom residuals panels are not the same .
the colors of the points are the same as in fig .
[ fig : lc ] .
the bronberg data have been binned for clarity ( as in fig .
[ fig : lc])[fig : res],title="fig : " ] , the error bars tend to be larger for 2-parameter errors than for 1-parameter errors . careful inspection of the residuals to the point lens and planet models shows improvement around the times of the limb - crossings of the source star and also on the falling side of the light curve . in the middle panel of each set ,
the red lines show the difference between the best planetary and best point lens models .
the didactic residuals in the middle panel are the sum of the red line and the residuals to the best planet model ( bottom panel ) .
the dashed lines indicate the limb - crossing times .
the dotted lines in the top panels of each set indicate the time ranges shown in the bottom panels .
note that the scales for the top and bottom residuals panels are not the same .
the colors of the points are the same as in fig .
[ fig : lc ] .
the bronberg data have been binned for clarity ( as in fig .
[ fig : lc])[fig : res],title="fig : " ] .
trajectory c ( black ) is preferred by @xmath110 over trajectories a ( green ) and b ( blue ) for models with 1-parameter errors and without parallax .
the source is shown to scale as the large circle ; the lines with arrows indicate the trajectories of the center of the source .
the abscissa in these plots is parallel to the binary axis with the lens star close to the origin and planet to the right .
the scale is such that 1.0 equals the einstein radius.[fig : traj ] ] distribution and the expected distribution ( @xmath171 ) , i.e. the @xmath62 residuals , for both 1-parameter ( top ) and 2-parameter ( bottom ) errors .
the top two panels of each set show the distributions for the point lens model ( a ) and the planetary models ( b ) .
the bottom panels ( c ) of each set show the difference between the top and middle panels .
the distributions for each data set are plotted separately and are shown over the same time range as the bottom panel of fig .
[ fig : lc ] ; data sets without points in this time range are not shown .
the colors are as in fig .
[ fig : lc ] . the thick black line shows the total distribution for all data .
the limb - crossing times are indicated by the dashed lines .
note the jump in the moa data ( _ light green _ ) at the time of the first limb - crossing ( hjd@xmath172 ) .
the signal is much more pronounced for 1-parameter errors than for 2-parameter errors .
note that the vertical scales in the two sets of panels are different .
[ fig : cumchi2],title="fig : " ] distribution and the expected distribution ( @xmath171 ) , i.e. the @xmath62 residuals , for both 1-parameter ( top ) and 2-parameter ( bottom ) errors .
the top two panels of each set show the distributions for the point lens model ( a ) and the planetary models ( b ) .
the bottom panels ( c ) of each set show the difference between the top and middle panels .
the distributions for each data set are plotted separately and are shown over the same time range as the bottom panel of fig .
[ fig : lc ] ; data sets without points in this time range are not shown .
the colors are as in fig .
[ fig : lc ] .
the thick black line shows the total distribution for all data .
the limb - crossing times are indicated by the dashed lines .
note the jump in the moa data ( _ light green _ ) at the time of the first limb - crossing ( hjd@xmath172 ) .
the signal is much more pronounced for 1-parameter errors than for 2-parameter errors .
note that the vertical scales in the two sets of panels are different .
[ fig : cumchi2],title="fig : " ] .
the microlensing event is at @xmath173 ( dashed line ) .
the dip at @xmath174 is caused by a quasi - circular dust lane , which may be seen in the image . even though the density of stars drops quickly as a function of radius ,
cluster members still comprise about 6% of the stars at @xmath175 .
[ fig : clusterdensity],title="fig:",height=288 ] .
the microlensing event is at @xmath173 ( dashed line ) .
the dip at @xmath174 is caused by a quasi - circular dust lane , which may be seen in the image . even though the density of stars drops quickly as a function of radius ,
cluster members still comprise about 6% of the stars at @xmath175 .
[ fig : clusterdensity],title="fig:",height=288 ] mt .
bigelow & i & 1.63 & 44 & 1.63 & 0.0 & 44 + molehill & unfiltered & 0.72 & 69 & 0.72 & 0.0 & 69 + kumeu & wratten # 12 & 1.19 & 188 & 1.18 & 0.0 & 188 + farm cove & unfiltered & 1.27 & 52 & 1.26 & 0.0 & 52 + auckland & wratten # 12 & 0.98 & 84 & 1.00 & 0.0 & 84 + vintage lane & unfiltered & 4.87 & 112 & 3.05 & 0.004 & 112 + b&c & i & 4.07 & 132 & 1.01 & 0.025 & 136 + & v & 1.15 & 53 & 0.66 & 0.03 & 55 + moa & moa - red & 1.68 & 4452 & 1.55 & 0.003 &
4434 + canopus & i & 3.02 & 28 & 2.87 & 0.0 & 29 + wise & unfiltered & 0.52 & 70 & 0.53 & 0.0 & 70 + bronberg & unfiltered & 1.26 & 727 & 1.27 & 0.0 & 727 + saao & i & 2.60 & 128 & 2.21 & 0.0015 & 127 + liverpool & sdss - i & 1.85 & 120 & 1.04 & 0.007 & 119 + la silla & i & 10.02 & 169 & 3.79 & 0.004 & 174 + ctio & i & 1.33 & 22 & 1.34 & 0.0 & 22 + & v & 0.50 & 3 & 0.50 & 0.0 & 3 + & h&&74 & & & 74 + ogle & i & 1.36 & 429 & 1.32 & 0.008 & 429 + point lens & 136.44&0.19615(4 ) & 0.00152(3 ) & 20.34(42 ) & 0.00260(5 ) & & & & & + pl , parallax & 69.61 & 0.1978(2 ) & 0.00167(4 ) & 19.37(41 ) & 0.00275(6 ) & & & & 3.16(41 ) & -1.34(20 ) + pl , parallax , @xmath176 & 123.89&0.19613(9 ) & -0.00158(4 ) & 20.23(42 ) & 0.00262(5 ) & & & & -1.07(36 ) & -0.98(27 ) + a & 35.26&0.19615(4 ) & 0.00167(3 ) & 19.04(34 ) & 0.00279(5 ) & 347.7(6 ) & -0.12(1 ) & -4.46(8 ) & & + a , parallax & -7.02 & 0.1976(2 ) & 0.00172(4 ) & 18.92(40 ) & 0.00283(6 ) & 347.4(3 ) & -0.08(1 ) & -4.9(1 ) & 2.82(44 ) & -1.29(22 ) + a , parallax , @xmath176 & -3.98&0.19626(9 ) & -0.00184(5 ) & 18.51(38 ) & 0.00289(6 ) & -346.(1 ) & -0.17(2 ) & -4.17(8 ) & -2.11(46 ) & -2.43(40 ) + a , wide & 33.07&0.19615(4 ) & 0.00167(4 ) & 19.06(39 ) & 0.00279(6 ) & 348.1(6 ) & 0.12(1 ) & -4.44(8 ) & & + b & 49.06&0.19614(4 ) & 0.00159(4 ) & 19.71(43 ) & 0.00269(6 ) & 118(1 ) & -0.43(4 ) & -3.5(1 ) & & + b , parallax & 21.45 & 0.1973(2 ) & 0.00166(4 ) & 19.40(41 ) & 0.00275(6 ) & 119(2 ) & -0.40(5 ) & -3.6(2 ) & 2.21(45 ) & -1.05(21 ) + b , parallax , @xmath176 & 31.33&0.19617(9 ) & -0.00169(4 ) & 19.46(41 ) & 0.00274(6 ) & -115(1)&-0.40(3 ) & -3.5(1 ) & -1.45(41 ) & -1.50(35 ) + b , wide & 49.11&0.19615(4 ) & 0.00159(3 ) & 19.76(38 ) & 0.00268(5 ) & 118(1 ) & 0.43(4 ) & -3.5(1 ) & & + c & 0.00&0.19613(4 ) & 0.00159(3 ) & 19.68(41 ) & 0.00270(6 ) & 236.4(7)&-0.26(4 ) & -3.7(1 ) & & + c , parallax & -10.65 & 0.1968(3 ) & 0.00164(4 ) & 19.34(39 ) & 0.00275(6 ) & 235(1 ) & -0.4(1 ) & -3.4(3 ) & 0.89(51 ) & -0.78(24 ) + c , parallax , @xmath176 & -12.63&0.19630(9 ) & -0.00171(4 ) & 19.20(38 ) & 0.00278(6 ) & -232(1 ) & -0.51(8 ) & -3.0(2 ) & -1.11(46 ) & -1.55(41 ) +
c , wide & 0.00&0.19613(4 ) & 0.00159(3 ) & 19.73(39 ) & 0.00269(5 ) & 236.4(7)&0.26(4 ) & -3.7(1 ) & & + point lens & 81.12&0.19618(5 ) & 0.00152(3 ) & 20.51(44 ) & 0.00259(6 ) & & & & & + pl , parallax & 25.43 & 0.1978(2 ) & 0.00164(4 ) & 19.80(43 ) & 0.00270(6 ) & & & & 3.20(43 ) & -1.14(21 ) + pl , parallax , @xmath176 & 72.23 & 0.1961(1 ) & -0.00160(4 ) & 20.23(44 ) & 0.00263(6 ) & & & & -1.26(42 ) & -0.96(32 ) + a & 6.37&0.19611(5 ) & 0.00160(4 ) & 19.82(43 ) & 0.00269(6 ) & 354(2 ) & -0.49(5 ) & -3.1(1 ) & & + a , parallax & -12.98 & 0.1972(3 ) & 0.00168(5 ) & 19.25(46 ) & 0.00278(7 ) & 349(3 ) & -0.3(2 ) & -3.7(6 ) & 1.89(70 ) & -1.16(25 ) + a , parallax , @xmath176 & -8.07 & 0.1963(1 ) & -0.00172(5 ) & 19.27(42 ) & 0.00278(6 ) & -348(2 ) & -0.46(7 ) & -3.1(2 ) & -0.98(52 ) & -1.59(48 ) + a , wide & -1.59&0.19614(5 ) & 0.00157(3 ) & 20.05(37 ) & 0.00266(5 ) & 356(3 ) & 0.539(4 ) & -3.01(3 ) & & + b & 13.25&0.19613(5 ) & 0.00160(3 ) & 19.79(41 ) & 0.00270(6 ) & 114(2 ) & -0.51(6 ) & -3.1(2 ) & & + b , parallax & -6.42 & 0.1972(2 ) & 0.00165(4 ) & 19.54(41 ) & 0.00274(6 ) & 112(3 ) & -0.5(1 ) & -3.1(3 ) & 1.97(48 ) & -1.02(23 ) + b , parallax , @xmath176 & 4.22 & 0.1962(1 ) & -0.00167(5 ) & 19.54(41 ) & 0.00274(6 ) & -109(3 ) & -0.50(6 ) & -3.1(2 ) & -0.99(47 ) & -1.24(43 ) + b , wide & 13.10&0.19613(5 ) & 0.00159(4 ) & 19.82(42 ) & 0.00269(6 ) & 113(2 ) & 0.51(6 ) & -3.1(2 ) & & + c & 0.00&0.19614(5 ) & 0.00158(3 ) & 19.94(40 ) & 0.00268(5 ) & 234(1 ) & -0.40(7 ) & -3.3(2 ) & & + c , parallax & -14.90 & 0.1970(3 ) & 0.00164(4 ) & 19.51(41 ) & 0.00274(6 ) & 231(2 ) & -0.56(8 ) & -2.9(2 ) & 1.26(54 ) & -0.96(24 ) + c , parallax , @xmath176 & -10.15 & 0.1963(1 ) & -0.00166(5 ) & 19.49(42 ) & 0.00274(6 ) & -230(2 ) & -0.51(7 ) & -2.9(2 ) & -0.44(51 ) & -1.04(46 ) + c , wide & -0.02&0.19613(5 ) & 0.00158(3 ) & 19.97(42 ) & 0.00267(6 ) & 234(1 ) & 0.40(7 ) & -3.3(2 ) & & + | we analyze moa-2010-blg-311 , a high magnification ( @xmath0 ) microlensing event with complete data coverage over the peak , making it very sensitive to planetary signals .
we fit this event with both a point lens and a 2-body lens model and find that the 2-body lens model is a better fit but with only @xmath1 .
the preferred mass ratio between the lens star and its companion is @xmath2 , placing the candidate companion in the planetary regime . despite the formal significance of the planet ,
we show that because of systematics in the data the evidence for a planetary companion to the lens is too tenuous to claim a secure detection .
when combined with analyses of other high - magnification events , this event helps empirically define the threshold for reliable planet detection in high - magnification events , which remains an open question .
high - magnification events , events in which the maximum magnification of the source , @xmath3 , is greater than 100 , have been a major focus of microlensing observations and analysis .
because the impact parameter between the source and the lens is very small in such cases , @xmath4 , it is likely to probe a central caustic produced by a planetary companion to the lens star .
furthermore , such events can often be predicted in advance of the peak , allowing intensive observations of the event at the time when it is most sensitive to planets .
consequently , a substantial amount of effort has been put into identifying , observing , and analyzing such events
. observed high - magnification events are classified into two groups for further analysis : events with signals obvious to the eye and events without .
only events in the first category are systematically fit with 2(or more)-body models .
the other events are only analyzed to determine their detection efficiencies . as a result , no planets have been found at or close to the detection threshold , and furthermore this detection threshold is not well understood .
@xcite suggest a detection threshold in the range of @xmath5350700 is required to both detect the signal and constrain it to be planetary , but they note that the exact value is unknown . with the advent of second - generation microlensing surveys , which will be able to detect planets as part of a controlled experiment with a fixed observing cadence , it is important to study the reliability of signals close to the detection threshold , since a systematic analysis of all events in such a survey will yield signals of all magnitudes , some of which will be real and some of which will be spurious . in this paper , we present the analysis of a high - magnification microlensing event , moa-2010-blg-311 , which has a planetary signal slightly too small to claim as a detection .
we summarize the data properties in section [ sec : data ] and present the color - magnitude diagram ( cmd ) in section [ sec : cmd ] . in section [ sec : models ] , we fit the light curve with both point lens and 2-body models .
we then discuss why a planetary detection can not be claimed in section [ sec : reliability ] .
we calculate the einstein ring size and relative proper motion in section [ sec : temu ] and discuss the possibility that the lens is a member of the cluster ngc 6553 in section [ sec : cluster ] .
we give our conclusions in section [ sec : discuss ] . |
the zero temperature phase diagram ( i.e. quantum phase diagram ) of a model gives important information on the low temperature behaviors of the system @xcite .
the anisotropic antiferromagnetic heisenberg ( xxz ) spin-1/2 model shows different quantum phases with respect to a symmetry - breaking transverse field ( non - commuting magnetic field ) @xcite .
the non - commuting field imposes quantum fluctuations into the ground state which can induce new phases .
@xmath5 is a quasi - one dimensional spin-1/2 xy - like antiferromagnet with weak inter - chain couplings ( @xmath6 ) which can be studied in terms of xxz chain with anisotropy parameter @xmath7 @xcite . the scaling behavior and quantum phase diagram of the xxz model in the presence of a transverse magnetic field ( @xmath8 )
have been investigated @xcite .
recently , the quantum spin models received much attentions from the quantum information point of views .
these are prototype models to implement and examine the idea of quantum computations which requires the quantum correlations measured by entanglement .
hence , disentangled ground states have to be avoided for such implementations . on the other hand ,
the zero entanglement property of a ground state provides an exact form for it in terms of product of single particle states .
the investigation in this direction resumed recently following the seminal work of j. kurman and his collaboratores@xcite .
several efforts have been devoted to this direction which is important for condensed matter researchers , i.e finding an exact ( factorized ) ground state even at particular values of the coupling constants @xcite .
the factorized ( exact ) ground state is an accurate starting point to investigate the quantum nature of a phase close to the factorizing point in addition to some exact knowledge which gives at the factorized point .
this property is implemented in this article to initiate a spin wave theory to describe the thermodynamic properties of the xxz model in the presence of a transverse magnetic field .
the u(1 ) symmetry of the xxz model is lost upon adding the transverse magnetic field .
initially , a perpendicular antiferromagnetic order is stabilized by promoting a spin - flop phase ( which has a partial moment projection along the field direction ) . at the factorizing field ( @xmath1 ) in the spin - flop phase
the ground state is known exactly as a direct product of single spin states and the staggered magnetization along @xmath9-direction is close to its maximum value . in our model
the factorizing field is @xmath10 , where quantum fluctuations are uncorrelated and the ground state is the classical one ( @xmath11 is the scale of energy and @xmath12 is the anisotropy parameter ) . for slightly larger magnetic field very close to the critical one ( @xmath13 )
the antiferromagnetic order becomes unstable and the staggered magnetization falls rapidly to vanish at the critical point . for @xmath14
the spins become aligned in @xmath15 direction and a fully polarized phase will be appeared .
the region between the factorizing and the critical fields ( @xmath13 ) is the main issue of our study which is not well understood so far .
the zero temperature properties of the intermediate region ( @xmath13 ) induces its signature into the thermodynamic functions of the model .
we have found that this region is characterized by two energy scales and its fingerprint will appear as a double peak in the specific heat .
moreover , the existence of more than an energy scale in the model can be related to a spontaneous symmetry breaking ( ssb ) @xcite .
each broken phase is described by an order parameter which can be zero in the disordered phase . for the aforementioned model
the symmetry breaking occurs where the staggered magnetization becomes zero .
the thermodynamic properties of the spin 1/2 xxz chain in a transverse magnetic field have been studied using the low temperature lanczos method @xcite . however , in this paper we will focus on the intermediate values of @xmath0 and more precisely on the region where a double peak appears in the specific heat of the model .
we implement the exact factorized ground state at @xmath4 ( where the entanglement vanishes ) to build up a spin wave theory appropriate to describe the model in the intermediate values of the magnetic field . in the linear spin wave approximation
we calculate the specific heat and thermal entanglement of the model .
moreover , the spin wave theory gives two different types of quasi particles which are representing the two energy scales .
we have also studied both the zero and finite temperature properties of the model on a finite chain using the low temperature lanczos method @xcite .
our numerical results are in agreement with the spin wave theory counterparts . in the next section
, we will briefly review the zero temperature properties of the model from the quantum information point of view .
we will provide the spin wave theory in sec.[spwt ] where the quantum property , specific heat and thermal entanglement are obtained .
finally , the numerical lanczos results will be presented in addition to discussions on the mentioned topics .
the anisotropic spin-1/2 heisenberg model in the presence of a transverse magnetic field is described by the following hamiltonian , @xmath16 where @xmath17 s are spin-1/2 operators , @xmath12 is the anisotropy parameter , @xmath0 is proportional to the transverse magnetic field and the exchange antiferromagnetic coupling @xmath11 which defines the scale of energy is set to one .
this model has several quantum phases with respect to the transverse magnetic field and anisotropy parameter ( @xmath12 ) .
the @xmath18-component spin structure factor at momentum @xmath19 is defined by @xmath20 and its increasing behavior in term of system size shows the magnetic order at the specified momentum .
we have plotted in fig.([mx ] ) the magnetization along @xmath15 direction ( @xmath21 ) and the @xmath9-component of spin structure factor at momentum @xmath22 versus @xmath0 .
the value of @xmath7 has been fixed to fit the case of real material @xmath5 .
the numerical data has been obtained by zero temperature lanczos method on a finite chain of length @xmath23 with periodic boundary condition .
( color online ) results of numerical zero temperature lanczos method for the magnetization along @xmath15-direction ( right vertical axis ) and the @xmath9-component spin structure factor at momentum @xmath24 ( left vertical axis ) versus the transverse magnetic field @xmath0 for a chain with @xmath23 and @xmath12=0.25 .
the magnetization in @xmath25-direction is zero.,width=377 ] the magnetization curve can be distinguished in three parts : ( 1 ) magnetization for @xmath3 , ( 2 ) magnetization for @xmath13 , and ( 3 ) the paramagnetic phase for @xmath14 . at zero magnetic field , there is no order in the system and all order parameters are zero i.e @xmath26 .
the nonzero value of @xmath0 starts to align the spins in @xmath15 direction and induces a small ferromagnetic order in @xmath15 direction .
the magnetization , @xmath27 , increases monotonically by increasing the magnetic field . for @xmath28
the quantum effects are considerable and the magnetization changes parabolically versus @xmath0 .
increasing @xmath0 , suppresses the quantum correlations and they goes to zero around @xmath1 . for @xmath13 , the magnetization is increased linearly versus @xmath0 .
magnetization increases up to the critical point ( @xmath2 ) and saturates at infinite field .
although the staggered magnetization along y - direction becomes zero at the finite critical field , the magnetization in x - direction will be fully saturated only for the isotropic case ( @xmath29 ) @xcite . in other words , the full saturation in x - direction for @xmath30 will happen for @xmath31 . for @xmath14
all of spins align almost completely ( for @xmath30 ) in the @xmath15 direction and we have a fully polarized paramagnetic phase .
we would like to draw your attentions to the region @xmath13 , where the model behaves surprisingly .
let us first study the spin chain through the entanglement ( @xmath32 ) of two spins @xcite which have no classical counterpart .
the entanglement is defined by the following relation , @xmath33 where , @xmath34 is the concurrence @xcite which is implemented instead of the pairwise zero temperature entanglement between two spins at sites @xmath35 and @xmath36 . for @xmath37 model
, @xmath38 is zero thus the concurrence takes the form @xcite , @xmath39 where @xmath40 using quantum monte carlo simulation , t. roscilde and collaborators @xcite have shown that unlike the standard magnetic order parameters ( fig.[mx ] ) the pairwise entanglement , plays an essential role at the factorizing field . at the factorizing field ( also called classical field , @xmath41 ) @xcite the ground state takes a product form @xcite and its entanglement is zero . in fig.([e - s ] ) , we have plotted @xmath42 and @xmath43 versus the transverse field @xmath0 for xxz spin-1/2 chain with @xmath7 by using exact diagonalization lanczos method . at the classical field ( @xmath44 ) ,
@xmath45 . at this point
the ground state of the model is factorized ( disentangled ) where the neel order is roughly maximized in @xmath9 direction .
( color online ) entanglement estimators of the xxz spin-1/2 chain versus transverse field @xmath0 with parameter @xmath7 .
@xmath46 is depicted by ( black ) circles and @xmath47 by ( red ) squares . at the classical field ( @xmath44 ) , @xmath48.,width=377 ]
let us describe the behaviors of @xmath49 and @xmath50 from a spontaneous symmetry breaking ( ssb ) point of view .
it is found in ref .
@xcite that the competition between the two functions @xmath49 and @xmath50 demonstrates the appearance of a ssb . in other words , for our model , a ssb can be occurred when @xmath51 . at this condition
the magnetic field is greater than the factorizing value ( @xmath52 ) .
moreover , ssb is usually recognized to happen at the position where the order parameter becomes zero . in our model two standard order parameters , @xmath27 and @xmath53 ( staggered magnetization along @xmath9 direction ) can represent the quantum phases of the model .
we have plotted in fig.([magnet ] ) the magnetization along @xmath15 direction ( @xmath27 ) and the @xmath9-component of spin structure factor ( @xmath54 ) at momentum @xmath24 for @xmath7 xxz spin-1/2 chain .
@xmath27 is nonzero in the whole range of magnetic fields however , @xmath55 which is used to show the antiferromagnetic order of the system has a different behavior .
as it is observed from our lanczos data , @xmath55 increases by @xmath0 up to @xmath1 .
this behavior is also found in the @xmath56 curve which has been obtained by density matrix renormalization group ( dmrg ) @xcite .
thus , no ssb occurs for @xmath28 . by
increasing the magnetic field for @xmath52 , @xmath55 ( or equivalently @xmath53 ) decreases rapidly and falls to zero at the critical point @xmath2 . in this respect ,
a symmetry breaking can occur only for @xmath52 . in this region , as mentioned before the slop of @xmath27 with respect to @xmath0 is different from the corresponding one for @xmath28 .
as discussed in the previous section the entanglement is zero at the factorizing field ( @xmath1 ) . it allow s us to write the many body ground state as the direct product of the single spin states on odd and even sublattices @xcite
@xmath57 ) which defines the rotation of the up - spin eigen - state of @xmath58 to the specific direction defined by the factorized state .
let us label the odd ( even ) sublattice by @xmath59 .
thus , ( @xmath60 ) represents a - sublattice while ( @xmath61 ) is the corresponding one for b - sublattice .
it has been shown @xcite that one can consider @xmath62 without loss of generality .
moreover , the magnitude of remaining polar angles are equal @xmath63 in the case of homogeneous spin model ( like here ) and is given by @xmath64 for the antiferromagnetic model the factorized ground state is defined by @xmath65 . before starting the spin wave approach we implement a unitary transformation on the hamiltonian .
all spins on the a - sublattice are rotated with angle @xmath66 counterclockwise around y - direction and clockwise for spins on b - sublattice .
the rotated hamiltonian ( @xmath67 ) is the result of rotations on all lattice points , @xmath68 and @xmath69 .
the single spin rotation operator is @xmath70 . in the rotated spin model
, the hamiltonian will have a factorized ground state which is a ferromagnet at the factorizing field .
the rotated spin hamiltonian is written in terms of boson operators @xmath71 with the following holstein - primakoff transformation , @xmath72 where @xmath73 are the rotated spin operators and @xmath74 is the unitary single spin rotation operator .
the bosonic hamiltonian in the linear spin wave approximation , i.e. @xmath75 , is given by @xmath76 + \sum_{l=0}^n\bigg([\frac{h}{h_f}(1+\delta)-\delta](a_l^{\dagger}a_l+b_l^{\dagger}b_l)\\ & & -\frac{\delta}{2}[a_l(b_l^{\dagger}+b_{l+1}^{\dagger})+h.c ] + \frac{1}{2}(\frac{h}{h_f}-1)\sqrt{1-\delta^2}(a_l+b_l+h.c)\bigg ) .
\label{hswt}\end{aligned}\ ] ] to diagonalize the bosonic model we first implement the fourier transformation and then apply a rotation to the boson operators ( @xmath77 ) @xmath78 @xmath79 the diagonalized hamiltonian in terms of two sets of quasi - particle operators is given by @xmath80 where the excitation spectrums have the following forms @xmath81 in the above calculations , a translation @xmath82 has been performed in the diagonalization procedure of the hamiltonian ( [ hswt ] ) .
the hamiltonian is now represented in terms of two quasi - particles ( bosons ) , each defines an energy scale .
the two energy scales which are the excitation energies of each bosons lead to two different dynamics for the system .
the consequence of two types of dynamics will be seen in the structure of specific heat which will be studied . however , the existence of two different quasi - particle energies ( @xmath83 ) is the first sign of two dynamics in the model . to get the finite temperature properties of the model , we assume @xmath84 , for the spin wave distribution functions , where @xmath85 is the probability of parallel ( @xmath86 ) and perpendicular ( @xmath87 ) normal modes appearing in the @xmath88-momentum state which satisfies @xmath89 for all @xmath88 s
. the substitutions of @xmath90 and @xmath91 ( where @xmath92 represents the thermal average ) in the spin - wave hamiltonian ( [ diagonalh ] ) gives the free energy , @xmath93 the number of bosons are controlled by the following constraint which is the magnetization in x - direction , @xmath94 the free energy is minimized with respect to @xmath85s under the constraint of ( [ cons ] ) which is applied by a lagrange multiplier ( @xmath95 ) via the boson s occupation number @xmath96 the constraint ( [ cons ] ) is applied by the values of @xmath97 which have been obtained by the numerical lanczos method .
we have plotted in fig.([c - swt ] ) the specific heat of @xmath7 xxz spin-1/2 chain versus temperature and different values of magnetic field .
a double peak is observed in the specific heat which is the signature of the existence of two comparable energy scales .
more precisely , the specific heat for @xmath98 have a narrow peak at low temperature ( @xmath99 ) and a broaden one at higher @xmath100 .
we will discuss more on this point in the next section .
( color online ) the spin wave results for specific heat of the xxz model versus @xmath100 for different values of the transverse magnetic field @xmath0 and anisotropy parameter @xmath7 .
( black ) circle is for @xmath101 , ( red ) square : @xmath102 , ( green ) gradient : @xmath103 and ( blue ) diamond : @xmath104 . a narrow peak at low temperature followed by a broaden one for higher temperature
are observed for @xmath105.,width=377 ] the established spin wave theory close to the factorizing point is used to obtain the thermal behaviors of the correlation functions in eqs.([concurrence ] ) and consequently to calculate the concurrence .
this method can describe the thermal entanglement of two spins in linear spin wave approximation . in this approach
, one can find the following expression for @xmath49 and @xmath50 @xmath106 as it is seen from these functions , @xmath49 is always less than zero thus the concurrence is @xmath107 .
we have plotted in fig.([cijs ] ) the thermal entanglement of the @xmath7 xxz spin-1/2 chain versus transverse field at temperature @xmath108 . by increasing the magnetic field ,
quantum fluctuations are decreased and the thermal entanglement is declined . at the factorizing point ,
quantum fluctuations become approximately uncorrelated and the thermal entanglement is very close to zero ( @xmath109 ) .
it means that the thermal entanglement is mainly originated from the ground state and the excited states have very tiny contribution to the entanglement
. thermal entanglement of the xxz model versus @xmath0 and anisotropy parameter @xmath7 at @xmath108 which has been obtained by spin wave approximation .
because of the tiny contribution of excited states to the entanglement , its value at @xmath1 is @xmath109.,width=377 ]
we have studied the effects of transverse magnetic field on the zero and finite temperature properties of @xmath7 xxz spin-1/2 chain . we have focused our attentions on the intermediate region of @xmath13 where the model behaves more interestingly .
m. kenzelmann and his collaborators @xcite have investigated experimentally the effects of a transverse magnetic field on the quasi - one dimensional spin-1/2 antiferromagnet @xmath5 , using single - crystal neutron diffraction . due to the weak inter - chain couplings in @xmath5 ( @xmath6 )
@xcite where @xmath11 is the coupling within a chain , it is proposed in ref .
@xcite that the bulk material has a spin liquid phase at the interval @xmath13 . in a spin liquid phase
all order parameters should be zero and there should be no long range order in the system .
the existence of very weak coupling between magnetic chains in this material makes it feasible to be described by a one dimensional spin-1/2 xxz model with the anisotropy parameter @xmath7 .
however , for the 1d xxz model in the region @xmath13 both magnetization and staggered magnetization are nonzero .
thus the ground state of the model could not be a spin liquid phase .
although the behaviors of the magnetic orders for this region are clearly known , the lacuna of a perfect study of the properties of the system at the intermediate region is still felt . in this respect , we have devoted our attentions to survey the magnetic and thermodynamic behavior of the system at the intermediate region of the transverse field .
we have implemented the low temperature lanczos method ( ltlm ) @xcite to compute the thermodynamic behaviors of the model for a chain of finite length .
ltlm has been used since it is accurate for low temperatures and specially the thermodynamic averages reach the ground state expectation values as temperature approaches zero .
the specific heat versus temperature for different values of the magnetic field on a chain of @xmath23 and with @xmath7 has been plotted in fig .
( [ cv ] ) .
the sampling is taken over @xmath110 lanczos numerical data which have been obtained by different initial random states .
the finite size effect can be ignored since the number of sampling ( @xmath111 ) is rather high .
the result of the spin wave theory ( fig.[c - swt ] ) and ltlm ( fig.[cv ] ) are in mutual agreement ; moreover , both figures show the presence of double peak in the specific heat for @xmath105 which is an evidence for the existence of two scales of energy or equivalently two dynamics in the system . the numerical lanczos results for the specific heat of xxz model versus @xmath100 for different values of transverse magnetic field and @xmath7 .
at the intermediate region of @xmath0 , a double peak structure is observed which indicates the existence of two energy scales in the system.,width=377 ] the specific heat versus @xmath100 for all values of the magnetic fields has schottky like peak at low temperatures which is a remarkable feature of the antiferromagnetic behavior . for small values of the magnetic field ( @xmath112 )
the schottky anomaly is the only one which justifies the existence of a single dynamics in the model .
further increasing of the magnetic field ( being close to the factorizing point ) a broaden peak is emerged in the specific heat data versus @xmath100 .
this bump is the result of the paramagnetic order .
two different orders are in competition and its signature is specially observable for magnetic fields in the region @xmath113 where the magnitude of the two types of ordering becomes comparable . the antiferromagnetic order in y - direction is the result of exchange coupling in the broken u(1 ) symmetry phase . the effect of transverse magnetic field as a paramagnetic order in x - direction shows its presence when it is enough large to define a new scale of energy . around the factorizing field
the two orders manifest their influence on the model as a narrow and broaden peak of the specific heat which happens for @xmath113 .
this is the region where the spontaneous symmetry breaking is started to happen . at the critical point ( @xmath2 )
the antiferromagnetic order vanishes and paramagnetic order is the only representative of the model .
a broaden peak in the specific heat versus @xmath100 is significant for @xmath14 which justifies the paramagnetic order . employing the specific heat data
, one can also scan the behavior of the energy gap ( @xmath114 ) . the existence of energy gap in the model is clearly observed by the exponential decay of the specific heat at very low temperatures ( @xmath115 ) .
@xmath116 of xxz model versus @xmath117 for different values of @xmath52 and @xmath7 . at low temperatures , the slop of the curve
is decreased from @xmath118 to @xmath104 and then starts to increase by increasing of @xmath0 . , width=377 ] @xmath116 of xxz model versus @xmath117 for different values of @xmath52 and @xmath7 . at low temperatures , the slop of the curve
is decreased from @xmath118 to @xmath104 and then starts to increase by increasing of @xmath0 . , width=377 ] at enough low temperatures ( @xmath119 ) the specific heat and energy gap are related by @xcite @xmath120 thus , the slop of ( @xmath116 ) curves versus @xmath121 in log scale gives good information about the energy gap . in figs.([ctb],[cta ] ) we have plotted ( @xmath116 ) versus @xmath121 for different values of transverse field .
fig.([ctb ] ) shows the gap behavior close to the factorizing point , @xmath28 . while its behavior for @xmath52 is presented in fig.([cta ] ) . a turning point of different plots in fig.([cta ] ) for different transverse magnetic field is the result of different behavior for @xmath122 ( decreasing gap ) and @xmath14 ( linear increasing paramagnetic gap ) .
our results fit very well with figs .
( 5.14 and 5.15 ) of ref .
@xcite . motivated by neutron - scattering results @xcite , t. radu @xcite investigated the effects of non - commuting field on the ground state of @xmath123 .
to be more precise in comparison , let us refer also to fig.13 of chapter.5 of ref .
@xcite where the experimental data of the specific heat have been shown .
the experimental data ( specially in the inset of fig.13 of chapter.5 of ref.@xcite ) display a double peak in the specific heat versus temperature which is for the intermediate range of magnetic field ( less than the critical one ) .
it is also worth to point out the behaviors of the system from the internal energy points of view .
the internal energy and specific heat are related by relation @xmath124 .
the double peak structure of the specific heat presages that at the intermediate values of the transverse field , there are two temperatures where the internal energy of the system gets its maximum variation . in other words , at these temperatures the maximum amount of states contribute to the response functions of the systems .
thus , one can also conclude that the appearance of the two energy scales in the system is appropriate with the number of contributed states .
knowledge on these properties could be important in the study of the magneto - caloric effects of the xxz model in the transverse field which is a work in progress .
we would like to thank m. rezai for his helpful comments .
this work was supported in part by the center of excellence in complex systems and condensed matter ( www.cscm.ir ) .
99 radu t 2005 _ thermodynamic characterization of heavy fermion systems and low dimensional quantum magnets near a quantum critical point _ , ph .
d. thesis , max planck institute for chemical physics of solids | we have investigated the zero and finite temperature behaviors of the anisotropic antiferromagnetic heisenberg xxz spin-1/2 chain in the presence of a transverse magnetic field ( @xmath0 ) .
the attention is concentrated on an interval of magnetic field between the factorizing field ( @xmath1 ) and the critical one ( @xmath2 ) .
the model presents a spin - flop phase for @xmath3 with an energy scale which is defined by the long range antiferromagnetic order while it undergoes an entanglement phase transition at @xmath4 .
the entanglement estimators clearly show that the entanglement is lost exactly at @xmath4 which justifies different quantum correlations on both sides of the factorizing field . as a consequence of zero entanglement ( at @xmath4 ) the ground state is known exactly as a product of single particle states which is the starting point for initiating a spin wave theory .
the linear spin wave theory is implemented to obtain the specific heat and thermal entanglement of the model in the interested region .
a double peak structure is found in the specific heat around @xmath4 which manifests the existence of two energy scales in the system as a result of two competing orders before the critical point .
these results are confirmed by the low temperature lanczos data which we have computed . |
graphitic surfaces have attracted considerable interest from several scientific groups . at first , a few decades ago and from a fundamental point of view , atomic resolved stm images were obtained from the highly oriented pyrolitic graphite ( hopg ) surface . as at atomic scale
such stm images show triangular and/or honeycomb lattices , even under different experimental conditions @xcite , several efforts have been done to understand this behavior @xcite , but only recently a conclusive explanation has been reached @xcite . in the nineties
the research on graphitic surfaces shifted the attention to the stm observation of superstructures @xcite , and thereafter to the interaction between such superstructures and grain boundaries in graphite @xcite .
these fields has been recently reactivated by the possibility of getting single and multi graphene layer islands onto different substrates such as : palladium @xcite , iridium @xcite , ruthenium @xcite , copper @xcite , or platinum @xcite . in particular , the stm observation of superstructures on graphitic surfaces can be explained in the frame of the well known _ moir patterns _ : interference patterns caused by a misalignment between two periodic lattices .
thus , a relative rotation between the top graphene layer and the underlying graphite crystal leads regions with characteristic stacking sequences : _ aa stacking regions _ , having a high concentration of atoms with a neighbor directly below ; and _ ab stacking regions _ , having a high concentration of atoms without a neighbor directly below .
these regions appear forming a lattice whose periodicity @xmath0 depends of the mis - rotation angle @xmath1 by @xmath2 $ ] , where @xmath3 is the graphene lattice constant .
this explanation , first proposed by kuwabara et al .
@xcite , became stronger when xhie et al .
@xcite showed experimental evidence in its favour .
however , a controversy arose about which stacking region appears brighter in the stm images @xcite .
on the other hand , as line profiles for moir patterns present large oscillation amplitudes ( which is known as _
giant corrugation _ ) , it was suggested that superstructures originate from surface deformations induced by stm tip @xcite . in this context , recently we have obtained theoretical constant height mode stm images for twisted bilayer graphene ( @xmath4 ) , which reproduce quite well the experimental observations : moir patterns and _ giant corrugation _
besides , these theoretical results point that : ( a ) current intensity maxima are over _ aa stacking regions _ , as previously reported @xcite ; and ( b ) _ giant corrugation _ has a strong electronic origin , as also was quoted recently @xcite .
in this paper we present experimental evidence of the observation of moir patterns induced by the rotation of both : top surface and subsurface graphene layers .
thus , by identifying border domains and defects , we can discriminate between superstructures resulting from layer rotation on the top surface from those caused by a subsurface layer rotation .
also , we can compare the oscillation amplitude of the associated line profiles .
these experimental results are complemented by theoretical stm images obtained from first - principles calculations performed on trilayer graphene which presents relative rotation among them .
the experiments were performed using an stm from omicron running at atmospheric pressure .
the samples were peeled off using an adhesive tape in air .
the stm images were collected at room temperature in both constant current and constant height mode using ptir tips .
the bias voltages refer to the applied sample bias , which corresponds to the filled - states .
every image was scanned from bottom - to - top .
the images were flattened to correct the sample tilting and filtered with smooth filters . in fig .
[ figure1 ] we show two topographic - stm images with atomic resolution taken on two different regions of the same sample : in ( a ) no moir pattern is observed , while in ( b ) it appears clearly . in the region without the superstructure the standard hopg image is reproduced , where only one lattice site is observed .
the corresponding line profile along @xmath5 , figure [ figure1](c ) , shows an oscillation amplitude around 0.8 . on the other hand , in the region with the superstructure
a triangular pattern superimposed at the atomic resolution is easily identifiable .
this resembles giant atoms ordered in a triangular superlattice , with a corresponding line profile that has an amplitude of oscillation of around 1.3 [ see figure [ figure1](d ) ] .
also , in these oscillations we can observe the small peaks associated to the atomic positions . in a given zone in the region with the moir pattern
, we also identified a defect , a void , which provides the chance to observe the subsurface layer ( see fig .
[ figure2 ] ) . inside this void
we also observe the moir pattern , wich indicates that the subsurface graphene layer is probably rotated with respect to the hopg , while the surface layer is not .
additionally , other experimental evidence comes from comparing line profiles inside the void ( lp1 ) with those and on the top surface ( lp2 ) : the line profile amplitude is higher inside the void . in both cases and are shown in ( c ) and ( d ) respectively .
[ figure1 ] ] ]
in order to probe the idea that a rotated subsurface layer can also produce moir patterns , we calculated stm images for a surface model presenting a rotation in a top surface or in a subsurface layer . specifically , the model consists of a repeating three layer graphene ( tlg ) slab , with enough separation between adjacent slabs to avoid mutual interaction . starting from a bernal stacking sequence ,
three cases were analyzed : no rotation ( nr ) , top surface layer rotation ( sr ) and middle layer rotation ( mr ) .
the rotation angle for the two last cases was @xmath6 .
this particular angle produces a commensurable superstructure composed of 222 carbon atoms , which is tractable from the computational point of view , while the periodicity ( @xmath7 nm ) is representative of experimental data @xcite . for this model
we have performed density functional theory calculations using the siesta _ ab initio _ package @xcite which employ norm - conserving pseudo - potentials and localized atomic orbitals as basis set . double-@xmath8 plus
polarization functions were used under the local density approximation @xcite .
all structures were fully relaxed until the atomic forces are smaller than 0.02 ev / .
we consider super - cells with periodic boundary conditions .
the brillouin zone was sampled using a monkhorst - pack mesh of @xmath9 .
the images were obtained using the code _ stm _ 1.0.1 ( included in the siesta package ) .
this code uses the wave functions generated by siesta and computed on a reference plane and extrapolates the values of these waves into vacuum . such reference plane must be sufficiently close to the surface so that charge density is large and well described .
the images were generated under the tersoff - hamann theory @xcite , which means that the states contributing to the tunneling current lies in the energy window [ @xmath10 .
data visualization was possible using the wsxm 5.0 freeware @xcite .
a gaussian smoothing was performed to obtain the final stm images . ) in the middle layer ( a ) and in the top layer ( b ) .
brilliant ( dark ) regions correspond to high ( low ) tunneling current density .
highlighted diagonal lines indicate the segments along which the respective profiles in ( c ) and ( d ) were obtained .
( @xmath11 v , @xmath12 ) .
[ figure4 ] ] by considering a bias voltage @xmath13 v , and a tip - to - surface distance @xmath14 , we calculated stm images for our tlg model .
figure [ figure4 ] shows these theoretical results for a supercell which presents a rotation in : the middle layer ( a ) , and in the top layer ( b )
. brilliant ( dark ) regions represent high ( low ) tunneling current density . in both images
we can identify a moir with a periodicity of 1.5 nm .
we have plotted the profiles along the highlighted lines in figs . [
figure4](a ) and [ figure4](b ) .
clearly , the oscillation amplitude associated to the surface layer rotation [ see fig . [ figure4](d ) ] is larger than that associated to a subsurface layer rotation [ see fig . [ figure4](c ) ] , thus coinciding with the experimental findings . a fat - band analysis give us additional elements to understand the moir pattern formation .
figure [ figure6 ] shows , from left to right , the contribution to the band structure of the orbitals @xmath15 , @xmath16 and @xmath17 in the case tlg with no rotation ( tlg - nr ) . as the orbitals @xmath15 and @xmath16
do not contribute to the band structure around the fermi level ( indicated by the horizontal red line ) , it is possible to recognize in their corresponding graphics the band structure of tlg reported previously @xcite . in the case of a top surface rotated layer , fig .
[ figure7 ] shows the different orbitals contributions .
again only the @xmath17 orbital contributes near the fermi level , although the band structure suffers an important modification respect the tlg - nr case : near the @xmath18 point and for energies range @xmath19 ev , four saddle points appear in the band structure .
four van hove singularities ( vhs ) then appear in the density of states , which explain the current density maxima detected by the stm tip @xcite . ,
@xmath16 and @xmath17 .
[ figure6 ] ] . from left to right the contribution from orbitals : @xmath15 , @xmath16 and @xmath17 .
[ figure7 ] ] a rotation in the middle layer produces notorious modifications respect the preceding cases . even though the contribution to the band structure around the fermi level comes again from the @xmath17 orbitals ( see fig .
[ figure8 ] ) , at @xmath20 the bands dispersion is linear . on the other hand , at the @xmath18 point and for the same energy window considered before ( @xmath19 ev ) , two saddle points appear in the band structure .
two vhs then appear in the density of states which explain the current intensity maxima revealed in the stm image . .
from left to right the contribution from orbitals : @xmath15 , @xmath16 and @xmath17 .
[ figure8 ] ]
as the states that contribute to the stm images lie in the energy window [ @xmath21 , @xmath20 ] , the following observations become interesting . firstly and as was expected , for @xmath23 only the @xmath17 orbitals contribute to the stm image . secondly , for a tlg array without rotation ( see fig .
[ figure6 ] ) , we reproduce the band structure reported prevously @xcite , providing us a checkpoint for our theoretical calculations .
thirdly , a rotation performed in the surface layer ( see fig . [ figure7 ] ) produces four vhs which contribute to the stm image and which explains the large corrugation amplitude of the corresponding line profile .
fourthly , if the middle layer has experienced the rotation , only two vhs appear and , although the corresponding stm image shows a moir pattern , its associated line profile presents a softer corrugation amplitude .
in fact , as experimental and theoretical results shows ( see figs . [ figure2 ] and
[ figure4 ] ) , the line profile amplitude associated with a rotation in the surface layer is twice the amplitude associated to a rotation in the subsurface layer . summarizing , we experimentally observed the formation of moir patterns on hopg surfaces , which originate from rotation on a graphene layer .
we identified zones where the rotated layer is the surface layer and zones where the rotated layer is a subsurface one .
this is evidenced by comparing the oscillation amplitude , which is lower when a subsurface layer rotates .
the calculation of the corresponding stm images agrees with the experimental observation and gives additional evidence of the important role of vhs to explain the moir patterns .
this work was partially supported by universidad de la frontera , project di11 - 0012 .
thanks to the psd53 project . p.v . acknowledges the financial help from fondecyt , grant : 1100508 and from usm , internal grant .
this research was partially supported by the supercomputing infrastructure of the nlhpc ( ecm-02 ) at cmcc - ufro .
16 g. binning , h. fuchs , ch .
gerber , h. rohrer , e. stoll , and e. tosatti , europhys .
* 1 * , 31 ( 1986 ) .
i . park and c. f. quate , appl .
. lett . * 46 * , 832 ( 1986 ) .
r. sonnenfeld and p. k. hansma , science * 232 * , 211 ( 1986 ) .
s. hembacher , f. j. giessibl , j. mannhart , and c.f .
quate , proc .
* 100 * , 12539 ( 2003 ) .
g. s. khara and j. choi , j. phys . : condens .
matter * 21 * , 195402 ( 2009 ) .
e. cisternas , f. stavale , m. flores , c.a .
achete , and p. vargas , phys .
b * 79 * , 205431 ( 2009 ) .
m. ondrek , p. pou , v. rozsval , c. gonzlez , p. jelnek , and r. prez , phys .
. lett . * 106 * , 176101 ( 2011 ) .
h. s. wong and c. durkan , nanotechnology * 23 * , 185703 ( 2012 ) .
m. kuwabara , d. r. clarke , and d. a. smith , appl . phys
. lett . * 56 * , 2396 ( 1990 ) . z. y. rong and p. kuiper , phys . rev .
b * 48 * , 17427 ( 1993 ) .
j. xhie , k. sattler , m. ge , and n. venkateswaran , phys .
b * 47 * , 15835 ( 1993 ) .
y. gan , w. chu , and l. qiao , surf .
* 539 * , 120 ( 2003 ) . c. oshima , j. phys .
: condens . mater . *
9 * , 1 ( 1997 ) .
a. t. ndiaye , s. bleikamp , p.j .
feibelman , and t. michely , phys .
97 * , 215501 ( 2006 ) .
p. w. sutter , j. i. flege , and e. a. sutter , nat . mater . * 7 * , 406 ( 2008 ) .
x. s. li , w. w. cai , j. a. an et al . ,
science * 324 * , 1312 ( 2009 ) .
g. otero , c. gonzlez , a. pinardi et al .
lett . * 105 * , 216102 ( 2010 ) .
pong and c. durkan , j. phys .
phys . * 38 * , r329 ( 2005 ) e. cisternas and j. d. correa , chem . phys .
* 409 * , 74 ( 2012 ) .
j. m. campanera , g. savini , i. suarez - martinez , and m. i. heggie , phys .
b * 75 * , 235449 ( 2007 ) .
i. brihuega , p. mallet , h. gonzlez - herrero et al .
lett . * 109 * , 196802 ( 2012 ) . j. m. soler , e. artacho , j. d. gale , a. garcia , j. junquera , p. ordejn , and j. d. snches - portal , j. phys .
condens . matter .
* 14 * 2745 ( 2002 ) .
d. m. ceperley and b. j. alder , phys .
* 45 * 566 ( 1980 ) .
j. tersoff and d. r. hamann , phys
lett . * 50 * , 1998 ( 1983 ) .
i. horcas , r. fernandez , j. m. gomez - rodriguez , j. colchero , j. gomez - herrero , and a. m. baro , rev .
. instr . * 78 * 013705 ( 2007 ) .
b. partoens and f. m. peeters , phys . rev .
b * 74 * , 075404 ( 2006 ) .
s. latil , v. meunier , and l. henrard , phys .
b * 76 * , 201402(r ) ( 2007 ) .
t. taychatanapat , k. watanabe , t. taniguchi et al . , nat
. phys . * 7 * , 621 ( 2011 ) .
g. li , a. luican , j.m.b .
lopez dos santos et al . , nat .
* 6 * , 109 ( 2010 ) .
a. luican , g. li , a. reina et al .
lett * 106 * , 126802 ( 2011 ) . | we have observed with stm moir patterns corresponding to the rotation of one graphene layer on hopg surface .
the moir patterns were characterized by rotation angle and extension in the plane .
additionally , by identifying border domains and defects we can discriminate between moir patterns due to rotation on the surface or subsurface layer . for a better understanding of moir patterns formation
we have studied by first principles an array of three graphene layers where the top or the middle layer appears rotated around the stacking axis .
we compare the experimental and theoretical results and we show the strong influence of rotations both in surface and subsurface layers for moir patterns formation in corresponding stm images . |
despite the potential for obscuration by dust , a significant fraction of a disk galaxy s uv emission manages to escape and thus be detected by instruments located beyond the earth s atmosphere . even in disk galaxies of high inclination
, uv imaging can reveal widespread emission ( cf .
fanelli et al . , neff et al . , & oconnell , these proceedings ) .
ultraviolet imaging of nearby disk galaxies reveals the star - forming activity with unprecedented clarity .
unlike imaging at h@xmath0 , uv imaging _ directly _ traces the full range of oba stellar spectral types , thereby sampling the recent - epoch `` population i '' component of each galaxy more completely .
uv imaging also provides a cleaner separation of the hot star component in regions dominated by cooler stars ( e.g. central disks & bulges ) .
moreover , uv imaging is unaffected by line absorption in the atmospheres of b & a - type stars unlike imaging at h@xmath0 and h@xmath1 thereby providing a truer representation , where these populations are concentrated ( see figure 1 ) .
finally , the uv colors of ob / hii regions can be used to derive extinction - free uv luminosities @xcite@xcite , whereas few constraints exist for deriving extinction - free euv ( e.g. lyman continuum ) luminosities based on h - line , radio - continuum , or other indirect indices .
uv images of nearby disk galaxies obtained with the shuttle - borne _ ultraviolet imaging telescope _ ( uit ) reveal a remarkable variety of star - forming morphologies .
these pop i patterns yield important insights to the respective roles of _ tides , waves , and resonances _ in orchestrating starbirth activity in disk galaxies .
a comparison of m33 ( scd ) , m74 ( sc ) , and m81 ( sb ) at uv and visible wavelengths highlights the pop i character of the uv imagery @xcite . _ flatter radial distributions _
are evident in the uv with exponential scalelengths that are 2045% larger .
the effects of reddening , abundance , and imf variations do not fully explain the differences .
the flatter uv profiles most likely indicate that the median radius of star - forming activity has migrated outward over the past several gyrs @xcite .
the uv morphologies also show _ narrower arms _ delineated by a combination of direct starlight from oba associations and indirect ( scattered ) radiation from dust associated with the massive young stars @xcite@xcite .
the narrower uv features indicate that star formation over the past @xmath210 myrs occupies a significantly smaller areal domain than the @xmath211000 myr legacy of star formation that is traced at longer wavelengths . in the giant sci spiral m101 ,
multiple linear arm segments ( _ `` crooked arms '' _ ) can be traced throughout the disk ( see figure 2 ) .
these features , along with a faint spiral arm and _ `` curly tail '' _ feature that links the outermost supergiant hii region with the rest of the galaxy , indicate that _ tidal processes of both external and internal origin _ are directing the current starbirth activity @xcite .
numerical simulations of isolated disk galaxies show that the `` crooked arm '' behavior can arise through the action of `` massive disturbers '' orbiting within the disks .
the outermost supergiant hii region , ngc 5471 may represent one of these massive disturbers .
larger - scale morphological and kinematic anomalies in m101 , including the faint arm and `` curly tail '' feature , require external interactions with companion galaxies .
such interactions can induce the formation of massive condensations at the ends of the tidal tails , perhaps explaining the origin of ngc 5471 at the terminus of m101 s `` curly tail '' @xcite .
similar behavior can be found in other giant sc galaxies with `` companions , '' including m51 , ngc 1232 , ngc2805 , and ngc 4303 . in the `` grand - design '' sci galaxy m74 ( ngc 628 ) ,
reflection of the uv - emitting disk upon itself shows the spiral structure to be more symmetric than is observed at visible wavelengths thus arguing for large - scale dynamics ( e.g. density waves ) governing the current - epoch star formation @xcite .
evidence for _ spatio - temporal sequences _ of molecular - cloud aggregation , massive star formation , cluster evolution , and cloud disruption can be found in m74 , m51 ( oconnell , these proceedings ) , and the _ inner disk _ of m101 @xcite . in m101 , far - uv emission is often found on the outer ( downstream ) side of the inner - disk co arms .
modeling the fuv
co displacements according to density - wave dynamics results in a wave pattern speed and co - rotation radius remarkably similar to those derived from a multi - mode analysis of the optical spiral structure @xcite .
similar downstream displacements between the fuv and h@xmath0 emission is evident in the se arm of m74 , again indicating density - wave dynamics at work . in the sab spiral m94 ( ngc 4736 ) , uv imaging reveals an inner starbursting ring and bi - symmetric outer knots in high contrast against the underlying visible bulge and disk ( see figure 1 ) .
dynamical resonances seem to best explain these transient features .
similar uv rings are evident in the inner disks of ngc 1317 ( sba ) , ngc 1512 ( sbb ) , ngc 3351 ( sbb ) , and m100 ( ngc 4321 ) ( sc ) most of which are also of `` early '' morphological type .
resonances may also explain the dearth of uv emission interior to m81 and m31 s ring - like spiral arms @xcite , whereby star - forming gas migrates outwards from the interior and piles up near the inner lindblad resonances ( kenney , these proceedings ) . because the longer - wavelength emission traces the older stellar populations and is so prominent interior to the ringlike arms , we conclude that the resonant locations and their constructive / inhibiting effects must have evolved with time . | ultraviolet imaging of nearby disk galaxies reveals the star - forming activity in these systems with unprecedented clarity .
uv images recently obtained with the shuttle - borne ultraviolet imaging telescope ( uit ) reveal a remarkable variety of star - forming morphologies .
the respective roles of tides , waves , and resonances in orchestrating the observed patterns of starbirth activity are discussed in terms of the extant uv data . |
the motion of a `` simple '' brownian particle is due to fluctuations of the surrounding medium , i.e. the result of random impacts of the molecules or atoms of the liquid or gas , the particle is immersed in .
this type of motion would be rather considered as _ passive motion _ , simply because the brownian particle does not play an active part in this motion . on the other hand , already in physico - chemical systems a _ self - driven motion _ of particles
can be found @xcite .
for instance , small solid particles floating on a liquid surface may produce a chemical substance at a spatially inhomogeneous rate , which locally changes the surface tension . this way the substance induces a net capillary force acting on the particle , which results in the particle s motion .
recent investigations on _ interacting _ self - driven particles show a broad variety of interesting phenomena , such as phase transitions and the emergence of self - ordered motion @xcite . here , the focus is on collective effects rather than on the origin of the particle s velocity ; i.e. it is usually postulated that the particles move with a certain non - zero velocity . in this paper , we focus on the _ energetic _ aspects of self - driven motion in order to derive conditions for an _ active mode _ of motion . _
active motion _ , as the name suggests , occurs under energy consumption , it is related to processes of energy storage and conversion into kinetic energy .
for instance on the biological level , cells or simple microorganisms are capable of _ active _ , self - driven motion , which in many cases has been successfully described by stochastic equations @xcite . in order to describe both the _ random _ aspects and the _ energetic _ aspects of active motion , we have introduced a model of _ active brownian particles _ @xcite .
these are brownian particles with the ability to take up energy from the environment , to store it in an internal depot and to convert internal energy to perform different activities , such as metabolism , motion , change of the environment , or signal - response behavior .
since the focus in this paper is on the energetic aspects of active brownian particles in a specific potential , possible changes of the environment are neglected here .
a very simple mechanism to take - up the additional energy required for active motion , is the pumping of energy by _ space - dependent friction _ @xcite . in this case
, the friction coefficient @xmath3 becomes a space - dependent function , @xmath4 , which in a certain spatial range can be also negative . inside this area the brownian particle , instead of loosing energy because of dissipative processes ,
is pumped with energy , which in turn increases its velocity . provided a supercritical supply of energy ,
the particle should be able to move with a velocity larger than the average thermal velocity . while such an approach will be able to model the spatially inhomogeneous supply of energy , it has the drawback not to consider processes of storage and conversion of energy .
in fact , with only a space - dependent friction , the brownian particle is instantaneously accelerated or slowed down , whereas e.g. biological systems have the capability to stretch their supply of energy over a certain time interval . in order to develop a more realistic model of active motion ,
we have considered an _ internal energy depot _ for the brownian particles @xcite , which allows to store the taken - up energy in the internal depot , from where it can be converted e.g. into kinetic energy , namely for the acceleration of motion . additionally , the internal dissipation of energy , due to storage and conversion ( or metabolism in a biological context ) can be considered . with these extensions ,
the brownian particle becomes in fact a _ brownian motor _
@xcite , which is fueled somewhere and then uses the stored energy with a certain efficiency @xcite to move forward , also against external forces . provided a supercritical supply of energy , we have found that the motion of active brownian particles in the two - dimensional space can become rather complex @xcite . in this paper
we investigate the one - dimensional motion of an ensemble of brownian particles with internal energy depot in piecewise linear potentials .
one particular example is the ratchet potential , i.e. a periodic potential which lacks the reflection symmetry .
ratchet systems recently attracted much interest with respect to transport phenomena on the microscale , since they provide a mechanism to transfer the undirected motion of brownian particles into a directed motion .
hence , the term _ brownian rectifiers _
@xcite has been established . in order to reveal the microscopic mechanisms resulting in directed movement ,
different physical ratchet models have been proposed @xcite , such as _ forced thermal ratchets _
@xcite , or _
stochastic ratchets _
@xcite , or _
fluctuating ratchets _ @xcite .
the model discussed in this paper , aims to add a new perspective to this problem which is based on the idea of internal energy storage . in sect . [ 2 ] , we discuss the basic features of our model without specifying an external potential , and derive the equations of motion both for the general case and the overdamped limit . in particular , we point to the possible existence of a `` high velocity '' or _ active mode _ of motion for the brownian particle , in addition to the usual passive mode of motion . in sect .
[ 3 ] , the necessary conditions for such an active mode of motion are investigated for the case of a linear potential . for the overdamped limit
, we derive critical parameters of energy conversion , which allows the particle to move also _ against _ the direction of the external force , i.e. to move `` uphill '' the potential gradient .
the results are applied to the motion of brownian particles in a ratchet potential in sect . [ 4 ] . after investigating the deterministic motion of a single particle in sect . [ 4.1 ] , we discuss the energetic conditions for the establishment of a net current for the determinsitic motion of an ensemble of particles in sect .
we show that , dependent on the conversion of internal into kinetic energy , a net current into different directions can be established .
the critical parameters for the energy conversion and the resulting velocity of the net current , which are found by means of computer simulations , are compared with the analytical results obtained in sect . [ 2 ] .
we also investigate the influence of the asymmetry of the ratchet potential on the establishment of the net current .
the motion of simple brownian particles in a space - dependent potential , @xmath5 can be described by the langevin equation : @xmath6 where @xmath3 is the friction coefficient of the particle at position @xmath7 , moving with velocity @xmath8 .
@xmath9 is a stochastic force with strength @xmath10 and a @xmath11-correlated time dependence @xmath12 using the fluctuation - dissipation theorem , we assume that the loss of energy resulting from friction , and the gain of energy resulting from the stochastic force , are compensated in the average , and @xmath10 can be expressed as : @xmath13 where @xmath14 is the temperature and @xmath15 is the boltzmann constant .
in addition to the dynamics described above , the brownian particles considered here are pumped with energy from the environment , which they can store in an internal depot .
further , internal energy can be converted into kinetic energy .
considering also internal dissipation , the resulting balance equation for the internal energy depot , @xmath16 , of a pumped brownian particle is given by : @xmath17 @xmath18 is the space - dependent pump rate of energy and @xmath19 describes the internal dissipation assumed to be proportional to the depot energy .
@xmath20 is the rate of conversion of internal into kinetic energy which should be a function of the actual velocity of the particle .
a simple ansatz for @xmath20 reads : @xmath21 where @xmath22 is the conversion rate of internal into kinetic energy .
the energy conversion results in an additional acceleration of the brownian particle in the direction of movement , expressed by the vector @xmath23 .
hence , the equation of motion for the pumped brownian particles has to consider an additional driving force , @xmath24 . in @xcite , we have postulated a stochastic equation , which is consistent with the langevin eq .
( [ langev - or ] ) : @xmath25 if we restrict ourselves to the one - dimensional case , i.e. the space coordinate is given by @xmath26 and further assume @xmath27 for the mass and @xmath28 for influx of energy , the dynamics for the pumped brownian motion , eqs .
( [ langev - pump ] ) , ( [ dep0 ] ) can be specified as follows : @xmath29 where @xmath30 is a stochastic force with white - noise fluctuations : @xmath31 in order to find a solution for the coupled eqs .
( [ model ] ) , let us now consider a relaxation of the dynamics on different time scales . if we assume that the velocity @xmath32 is changing much faster than the space coordinate @xmath33 , for @xmath32 , eq .
( [ model ] ) , a formal solution can be given : @xmath34 dt ' \label{v - adiab}\end{aligned}\ ] ] this solution however depends on the integrals over @xmath35 , reflecting the influence of the energy depot on the velocity .
if we further assume a fast relaxation of the depot , @xmath35 , compared to the relaxation of the velocity @xmath32 , the corresponding equation of eq .
( [ model ] ) can be solved , and we obtain with the initial condition @xmath36 the following quasistationary value for the energy depot : @xmath37 which yields a possible maximum value of @xmath38 .
the overdamped limit is obtained by considering a fast relaxation of the velocities , in which case the set of eq .
( [ model ] ) can be further reduced to : @xmath39 we note that , due to the dependence of @xmath40 on @xmath41 , eq . ( [ x - overd ] )
is coupled to eq .
( [ e0 ] ) .
thus , the overdamped eq .
( [ x - overd ] ) could be also written in the form : @xmath42 eq .
( [ x - overd-2 ] ) indicates a cubic equation for the velocities in the overdamped limit , i.e. the possible existence of non - trivial solutions for the stationary velocity . for the further discussion ,
we neglect the stochastic term in eq .
( [ x - overd-2 ] ) and denote the stationary values of @xmath32 by @xmath43 .
further , the force resulting from the gradient of the potential , , is introduced .
then , eq . ( [ x - overd-2 ] ) can be rewritten as : @xmath44\ , { \mbox{\boldmath $ v$}}_0 = c { \mbox{\boldmath $ f$}}. \label{v0-cubic}\ ] ] depending on the value of @xmath45 and in particular on the sign of the term @xmath46 , eq .
( [ v0-cubic ] ) has either one or three real solutions for the stationary velocity , @xmath47 .
the always existing solution expresses a direct response to the force in the form : @xmath48 this solution results from the analytic continuation of stokes law , @xmath49 , which is valid for @xmath50 .
we will denote this solution as the `` normal response '' mode of motion , since the velocity @xmath8 has the same direction as the force @xmath45 resulting from the external potential @xmath51 .
as long as the supply of the energy depot is small , we will also name the normal mode as the _ passive mode _ , because the particle is simply driven by the external force .
more interesting is the case of three stationary velocities , @xmath47 , which significantly depends on the ( supercritical ) influence of the energy depot . in this case which will be discussed in detail in the following section , the particle will be able to move in a `` high velocity '' or _ active mode _ of motion .
for the one - dimensional motion , in the active mode only two different directions are possible , i.e. a motion into or _ against _ the direction of the force @xmath45 .
but already in the two - dimensional case there are infinitely different possibilities .
this conclusion is of importance when discussing stochastic influences as will be done in a forthcoming paper @xcite . in the one - dimensional case ,
the influence of noise is rather weak , because of the limited number of possible directions , but in the two- and three - dimensional case , the active motion of the particles is very sensitive to stochastic influences , which may determine the direction of motion in the active mode .
in the following , we restrict the discussion to the one - dimensional , deterministic motion of the particle , corresponding to @xmath52 in eq .
( [ model ] ) .
further , we may assume that the force resulting from the gradient of the potential is constant .
then , we have the two coupled equations for @xmath32 and @xmath35 : @xmath53 the stationary solutions of eq .
( [ model2 ] ) are obtained from @xmath54 and @xmath55 : @xmath56 which lead to the known cubic polynom for the amount of the constant velocity @xmath57 , eq .
( [ v0-cubic ] ) : @xmath58 here @xmath59 is defined as a vector @xmath60 . for a first insight into the possible solutions of eq .
( [ v0-cubic-2 ] ) , we consider the force free motion , @xmath61 .
( [ v0-cubic-2 ] ) has either one or three stationary solutions which read : @xmath62 the corresponding bifurcation diagram is shown in fig .
[ bifurc]a which displays a typical fork bifurcation .
the bifurcation point is given by : @xmath63 above a critical supply of energy , which is expressed in terms of the conversion parameter @xmath64 , we find the occurence of two new solutions for the stationary velocity , corresponding to the active modes of motion .
for @xmath61 , @xmath65 and @xmath66 both have the same amount but different directions . below the critical threshold
the passive mode of motion , @xmath67 , is the only stable solution . at the bifurcation point ,
a change of the stability occurs , consequently above the critical value of @xmath64 the active modes of motion are assumed stable . for the case
@xmath68 , the analysis of eq . ( [ v0-cubic-2 ] ) leads to a different bifurcation diagram shown in fig .
[ bifurc]b . above a critical supply of energy
, we find the appearance of two high velocity or active modes of motion .
one of these active modes has the same direction as the driving force , thus it can be understood as the continuation of the normal solution . as fig .
[ bifurc]b shows , the former passive normal mode , which holds for subcritical energetic conditions , is transformed into an active normal mode , where the particle moves into the same direction , but with a much higher velocity . _
additionally _ , in the active mode a new high - velocity motion _ against _ the direction of the force @xmath45 becomes possible . while the first active mode would be rather considered as a
_ normal response _ to the force @xmath45 , the second active mode appears as unnormal ( or non - trivial ) response , which corresponds to an `` uphill '' motion ( cf .
[ flanke1 ] ) .
it is obvious that the particle s motion `` downhill '' is stable , but the same does not necessarily apply for the possible solution of an `` uphill '' motion .
thus , in addition to eq .
( [ v0-cubic-2 ] ) which provides the _ values _ of the stationary solutions , we need a second condition which guarantees the _ stability _ of these solutions . before this
is carried out , we want to derive a handy expression for the stationary velocities in the case @xmath69 . with the assumption that the term @xmath70 is small , the stationary solutions of eq .
( [ v0-cubic-2 ] ) can be given as : @xmath71 we note that eq .
( [ 3stat - c0 ] ) is a sufficient approximation for the stationary velocities , especially in a certain distance from the bifurcation point , thus it will be also used in sect . [ 4.1 ] . for the stability analysis
, we consider small fluctuations around the stationary values , @xmath57 and @xmath40 : @xmath72 inserting eq .
( [ fluct ] ) into eq .
( [ model2 ] ) , we find after linearization : @xmath73 with the ansatz : @xmath74 we find from eq .
( [ lin - stab ] ) the following relation for @xmath75 : @xmath76 in general , we need to discuss eq .
( [ lamb-12 ] ) for the three possible solutions @xmath57 which result from eq .
( [ v0-cubic-2 ] ) . dependent on whether the @xmath75 for each solution have real or complex positive or negative values , we are able to classify the types of the possible stationary solutions in this case . the results are summarized in table [ tab - bifurc ] .
the phase plots shown in fig .
[ phase]a , b present more details .
further , fig .
[ lambda12 ] shows the real part @xmath77 of eq .
( [ lamb-12 ] ) for the active mode corresponding to the `` uphill '' motion of the particle , which is the most interesting one .
.results for the stability analysis , eq .
( [ lamb-12 ] ) . [ tab - bifurc ] [ cols=">,<,<,^ , < " , ]
we find that below the bifurcation point which is @xmath78 for the given set of parameters , only one stable node exists in the @xmath79 phase space , which corresponds to the passive normal mode .
then , a subcritical bifurcation occurs which leads to 3 stationary solutions : a stable and an instable node , and a saddle point , since all the @xmath75 are real . at @xmath80 , however , the nodes turn into focal points . with respect to the `` uphill motion '' we find in fig .
[ lambda12 ] the occurence of an instable node at @xmath81 , which then becomes an instable focus for @xmath82 .
the respective real parts of @xmath75 are equal in this range , i.e. the @xmath83 are complex .
the stability condition is satisfied only if @xmath84 , which is above a second critical value @xmath85 for the given set of parameters .
that means , for @xmath86 , the instable focal point becomes a stable focus , which is also clearly shown in the phase plots of fig .
[ phase]a , b . in both figures
, we see a stable focal point for positive values of the velocity , @xmath87 , which correspond to the stable motion `` downhill '' , i.e. in the direction of the driving force . for @xmath88 ,
the phase plot for negative values of @xmath87 shows an instable focal point , which turns into a stable focal point for @xmath89 .
thus , we can conclude that for @xmath90 an active mode of motion becomes possible , which also implies the possibility of an `` uphill '' motion of the particle .
however , only for values @xmath91 , we can expect a _
stable _ motion against the direction of the force .
the interesting result of a possible stable `` uphill - motion '' of particles with internal energy depot will be employed in the following section , where we turn to a more sophisticated , piecewise linear potential . for our further investigations
, it will be useful to have a handy expression for the critical supply of energy , @xmath92 , which allows a stable `` uphill '' motion .
this will be derived in the following , with only a few approximations . for the parameters used during the computer simulations discussed later , fig .
[ lambda12 ] and table i indicate that the square root in eq .
( [ lamb-12 ] ) is imaginary , thus the stability of the solutions depends on the condition : @xmath93 if we insert the stationary value @xmath40 , eq . ( [ v0-e0 ] ) , eq .
( [ c - stab ] ) leads to a 4th order inequation for @xmath57 to obtain stability : @xmath94 for a stable stationary motion of the particle , both eq .
( [ v0-cubic-2 ] ) and eq .
( [ v0-stab ] ) have to be satisfied .
the _ critical condition _ for stability just results from the equality in eq .
( [ v0-stab ] ) , which then provides a replacement for the prefactor @xmath95 in eq .
( [ v0-cubic-2 ] ) . if we insert the critical condition into eq .
( [ v0-cubic-2 ] ) , we arrive at a 5th order equation for @xmath57 : @xmath96 in order to simplify the further discussion , we assume that the internal dissipation is negligible , @xmath97 . then , eq . ( [ v0 - 5 ] ) gives the simple nontrivial solution : @xmath98 this expression can be used to eliminate the stationary velocity , @xmath47 , in eq . ( [ v0-stab ] ) . with the assumption @xmath97
, we obtain now from the critical condition , i.e. from the equality in eq .
( [ v0-stab ] ) , a relation between the force , @xmath45 , and the conversion parameter , @xmath64 . combining eq .
( [ v0-stab ] ) and eq .
( [ v0 - 3-f ] ) results in @xmath99 because of @xmath100 , the trivial and the negative solution of eq .
( [ d2 - 23 ] ) can be neglected , and we finally arrive at the following _ critical relation _ for @xmath101 : @xmath102 fig .
[ d - f ] shows @xmath92 as a function of the force , @xmath103 . in the limit of negligible internal dissipation ,
the relation @xmath104 describes how much power has to be supplied by the internal energy depot in order to allow a stable motion of the particle in _ both _ directions , in particular a _ stable uphill motion _ of the particle .
for further investigations of the motion of pumped particles , we specify the potential @xmath51 as a piecewise linear , asymmetric potential ( cf . fig .
[ potent ] ) , which is known as a _ ratchet potential _ : @xmath105 further , we will use the following abbreviations with respect to the potential @xmath106 , eq .
( [ ux ] ) .
the index @xmath107 refers to the two pieces of the potential , @xmath108 , @xmath109 .
the asymmetry parameter @xmath110 should describe the ratio of the two pieces , and @xmath111 is the force resulting from the gradient of the piecewise linear potential .
hence , for the potential @xmath106 , eq .
( [ ux ] ) , the following relations yield : @xmath112 whether or not the particle will be able to leave one of the potential wells described by eq .
( [ ux ] ) , depends in a first approximation on the height of the potential barrier , @xmath113 , and on the kinetic energy of the particle . for particles with internal energy depot ,
the actual velocity depends also on the conversion of internal into kinetic energy , eq .
( [ model ] ) . in agreement with the investigations in the previous section , we can in principle
distinguish between two different types of motion : ( i ) a bound motion , i.e. the particle will not leave the potential well because of the subcritical supply of energy from the depot , but its position might oscillate within the boundaries , ( ii ) an unbound motion , i.e. the particle will be able to leave the potential well because of the supercritical supply of energy and move freely .
both types of motion have analogies to the localized and delocalized states of electrons in solid state physics as will be discussed in more detail in a forthcoming paper . in the following ,
we discuss computer simulations of the _ deterministic _ motion of _ one _ pumped brownian particle in a ratchet potential .
the particle ( mass @xmath27 ) starts its motion outside the potential extrema ; hence , there is an initial force on the particle .
the results for a single particle are shown in fig .
[ single - xve - t ] , where two different sets of parameters are used : * a small internal dissipation , @xmath19 , which means a nearly ideal energy depot , and a large friction coefficient , @xmath3 , resulting in a _ strongly damped motion _ , * an internal dissipation , @xmath19 , 10 times larger , an energy influx , @xmath114 , ten times smaller , and a friction coefficient , @xmath3 , 100 times smaller than in ( i ) , resulting in a _ weakly damped motion _
, we note , that in the computer simulations always the complete set of eq .
( [ model ] ) for the particles is solved , regardless of the possible approximations .
the trajectories @xmath33 in fig .
[ single - xve - t ] indicate a nearly uniform motion of the particle into one direction .
the continuous motion corresponds to a _
delocalized _ state of the particle in the ratchet potential , whereas for subcritical energetic conditions only localized states exist .
the transition from localized to delocalized states will be discussed in more detail in a forthcoming paper @xcite . as shown in fig .
[ single - xve - t]a , the less damped motion of a single particle may result in steady oscillations in the velocity , @xmath32 , and the energy depot , @xmath35 . only if the damping is large enough
, the velocity and the energy depot may reach constant values ( cf .
[ single - xve - t]b ) .
these values are of course different for each piece of the potential , hence the periodical movement through every maximum or minimum of the potential results in jumps both in the velocity and the energy depot , which are followed by oscillations . in the phase space shown in fig .
[ damp - traj]a , the motion of the particle appears as a transition between two stable fix points , each of which describes the stable motion on one flank . in the following , we restrict the discussion to the _ strongly damped case_. the oscillations which occur in @xmath87 and @xmath16 are damped out on a characteristic time scale , @xmath115 .
if we assume that the particle moves on the different pieces of the potential @xmath116 ( cf .
[ potent ] ) during the two characteristic time intervals : @xmath117 and @xmath118 , then the particle is subject to a constant force only as long as @xmath119 or @xmath120 , respectively . for times larger than the characteristic time ,
@xmath121 , the motion of the particle can be described by the equation of the overdamped limit , eq .
( [ x - overd ] ) . if we neglect again stochastic influences , eq .
( [ x - overd ] ) can be rewritten in the form : @xmath122\ ; { \mbox{\boldmath $ v$}}_0 + { \mbox{\boldmath $ f$}}\ ] ] where @xmath123 is the velocity in the overdamped limit and @xmath124 is defined by eq .
( [ f - a ] ) . the stationary value for the internal energy depot , @xmath40 , is given by eq . ( [ v0-e0 ] ) .
if we assume again that the term @xmath70 is small , then the constant velocity can be calculated from eq .
( [ 3stat - c0 ] ) , with @xmath45 specified as @xmath125 or @xmath126 , respectively . for any constant force @xmath103 , there are two possible non - trivial solutions of eq .
( [ 3stat - c0 ] ) : a positive and a negative velocity with different , but constant amount , which depend on the gradient of the potential .
the nontrivial values @xmath127 of eq .
( [ 3stat - c0 ] ) can be compared with the constant values obtained in the simulations , and we find : @xmath128 the eqs .
( [ x - overd ] ) , ( [ v0 - 1 ] ) for the overdamped limit indicate that the dynamics remarkably depends on the sign of the prefactor @xmath129 , which governs the influence of the potential and the stochastic force .
therefore , the prefactor should be discussed in more detail , now .
[ damp - traj]b shows that the prefactor @xmath130 displays a behavior similar to the velocity , fig .
[ single - xve - t]b .
the prefactor jumps between a positive and a negative constant value , which can be approximated by means of the constant , @xmath40 , eq .
( [ v0-e0 ] ) , reached after the oscillations damped out .
it is shown that the jump occurs at the same time when the gradient of the potential changes its sign .
this can be also proved analytically .
using eqs .
( [ e0 ] ) , ( [ 3stat - c0 ] ) , the prefactor @xmath131 in eq .
( [ x - overd ] ) and eq .
( [ v0 - 1 ] ) , respectively , can be rewritten and we find after a short calculation : @xmath132\ ] ] this means that the product of the prefactor and the potential gradient always has the same positive ( or negative ) sign , and the direction of motion for the particle is only determined by the initial condition .
the prefactor @xmath129 describes the balance between dissipation and the energy supply from the internal depot of the particle .
therefore it is expected that the time average of the prefactor should be zero , when averaged over one time period @xmath14 : @xmath133 this can be proven using @xmath134 as discussed above .
the values @xmath135 , @xmath136 and @xmath137 , @xmath138 for the related velocity and energy depot along the different pieces of the potential can be calculated by means of eqs .
( [ 3stat - c0 ] ) , ( [ v0 - 1 ] ) which leads directly to eq .
( [ aver-2-a ] ) .
let us now discuss the _ deterministic _ motion of an _ ensemble _ of @xmath139 pumped brownian particles in a ratchet potential .
for the computer simulations , we have assumed that the start locations of the particles are equally distributed over the first period of the potential , @xmath140 and their initial velocity is zero . in the deterministic case , the _ direction of motion _ and the _ velocity _ at any time @xmath141 are mainly determined by the initial conditions .
hence , particles with an initial position between @xmath142 , which initially feel a force into the negative direction , most likely move with a negative velocity , whereas particles with an initial position between @xmath143 most likely move into the positive direction .
this is also shown in fig .
[ damp - ve - x0]a , where the velocity @xmath87 is plotted versus the initial position of the particles .
oscillations occur only at the minima and maxima of the related potential , indicating a strong sensitivity to the initial condition in these regions .
the distribution of the final velocity is shown in fig .
[ damp - ve - x0]b . from fig .
[ damp - ve - x0]b we see two main currents of particles occuring , one with a positive and one with a negative velocity , which can be approximated by eq .
( [ 3stat - c0 ] ) .
the net current , however , has a positive direction , since most of the particles start with the matching initial condition .
the time dependence of the averages is shown in fig .
[ damp - avest ] .
the long - term oscillations in the average velocity and the average energy depot result from the superposition of the velocities , which are sharply peaked around the two dominating values ( cf . fig .
[ damp - ve - x0]b ) .
the existence of periodic stationary solutions , @xmath144 , requires that the particles are able to escape from the initial period of the potential . for a continuous motion which corresponds to delocalized states
, the particles must be able to move `` uphill '' on one or both flanks of the ratchet potential . in sect .
[ 3 ] , we already investigated the necessary conditions for such a motion for a single flank , and found a critical condition for the conversion rate , @xmath64 , eq .
( [ d2-f ] ) . in order to demonstrate the applicability of eq .
( [ d2-f ] ) for the ratchet potential , we have investigated the dependence of the _ net current _ , expressed by the mean velocity @xmath145 , on the conversion rate , @xmath64 , for the overdamped case .
the results of computer simulations are shown in fig .
[ v - d2 ] . in fig .
[ v - d2 ] , we see the existence of _ two different critical values _ for the parameter @xmath64 , which correspond to the onset of a _ negative net current _ at @xmath146 and a _ positive net current _ at @xmath147 . for values of @xmath148 near zero and less than @xmath146
, there is no net current at all .
this is due to the subcritical supply of energy from the internal depot , which does not allow an uphill motion on any flank of the potential .
consequently , after the initial downhill motion , all particles come to rest in the minima of the ratchet potential , with @xmath149 as the only stationary solution for the velocity . with an increasing value of @xmath64
, we see the occurence of a negative net current at @xmath146 .
that means , the energy depot provides enough energy for the uphill motion along the flank with the lower slope , which , in our example , is the one with @xmath150 ( cf . fig . [ potent ] ) .
if we insert this value for @xmath103 into the critical condition , eq .
( [ d2-f ] ) , a value @xmath151 is obtained , which agrees with the onset of the negative current in the computer simulations , fig .
[ v - d2 ] .
for @xmath152 , a stable motion of the particles up and down the flank with the lower slope is possible , but the same does not necessarily apply for the steeper slope .
hence , particles which start on the lower slope with a positive velocity , can not continue their motion into the positive direction since they are not able to climb up the steeper slope .
consequently , they turn their direction on the steeper slope , then move downhill driven by the force into the negative direction , and continue to move into the negative direction while climbing up the lower slope . therefore ,
for values of the conversion rate between @xmath146 and @xmath147 , we only have an _ unimodal _ distribution of the velocity centered around the negative value : @xmath153 which is independent of @xmath64 if the term @xmath154 in eq .
( [ 3stat - c0 ] ) is negligible , which holds for the considered case . for @xmath155 ,
the energy depot also supplies enough energy for the particles to climb up the steeper slope , consequently a periodic motion of the particles into the positive direction becomes possible , now . in our example
, the steeper slope corresponds to the force @xmath156 ( cf .
[ potent ] ) which yields a critical value @xmath157 , obtained by means of eq .
( [ d2-f ] ) .
this result agrees with the onset of the positive current in the computer simulations , fig .
[ v - d2 ] . for @xmath155
, we have a _ bimodal _ velocity distribution , as also shown in fig .
[ damp - ve - x0]b .
the net current , which results from the average of the two main currents , has a positive direction in the deterministic case , because most of the particles start into a positive direction , as discussed above .
we may simply assume , that the number of particles in each direction is roughly proportional to the length of the flank where they started from , which is also indicated by the velocity distribution , fig .
[ damp - ve - x0]b .
then the mean velocity in the strongly damped case can be approximated by : @xmath158 where @xmath159 and @xmath160 are the stationary velocities on each flank , which , in the limit of an nearly ideal energy depot , can be determined from eq .
( [ 3stat - c0 ] ) . with the negative velocity , @xmath159 , eq .
( [ v - neg ] ) , and the positive velocity , @xmath161 we find from eq .
( [ mean - v ] ) for @xmath155 an average velocity , @xmath162 , which also agrees with the computer simulations , fig .
[ v - d2 ] .
the results of the computer simulations have demonstrated that in the deterministic case for the given special initial condition , i.e. the equal distribution of particles over the ratchet period , the direction of the net current can be adjusted by choosing the appropriate values of the conversion rate , @xmath64 .
the critical values for @xmath64 , on the other hand , depend on the slope of the two flanks of the potential , expressed by the force @xmath103 .
lower slopes also correspond to lower values of the conversion rate , because less power is needed for the uphill motion .
we conclude our results by investigating the influence of the slope on the establishment of a positive or negative net current . with a fixed height of the potential barrier , @xmath113 , and a fixed length @xmath163 ,
the ratio of the two different slopes is described by the asymmetry parameter @xmath164 , eq .
( [ f - a ] ) .
the occurence of a current in the ratchet potential requires the possibility of uphill - motion , which depends on the critical supply of energy , described by eq .
( [ d2-f ] ) . in order to obtain the critical value for the asymmetry of the potential
, we replace the force @xmath103 in eq .
( [ d2-f ] ) by the parameter @xmath110 , eq .
( [ f - a ] ) . in our example , the flank @xmath165 of the potential has the steeper slope , so the critical condition is determined by @xmath166 . as the result , we find : @xmath167^{3/2 } -1\ ] ] @xmath168 gives the critical value for the asymmetry , which may result in a reversal of the net current .
for @xmath169 , the flank @xmath165 is too steep for the particles , therefore only a negative current can occur which corresponds to the unimodal velocity distribution discussed above . for @xmath170 , however , the particles are able to move uphill either flank
. hence , also a positive current can establish and the velocity distribution becomes bimodal , which results in a positive net current . the current reversal from a negative to a positive net current is shown in fig .
[ v - as ] .
dependent on the value of the conversion rate , @xmath64 , we see the switch from the negative to the positive value of the net current at a critical value of the asymmetry parameter @xmath110 . because of the definition of @xmath110 , the results for @xmath171 , are the inverse of the results for @xmath172 .
obviously , for a symmetric ratchet , @xmath173 no net current occurs , because the two main currents compensate . from eq .
( [ a - crit1 ] ) , we obtain @xmath174 for @xmath175 , and @xmath176 for @xmath177 , which both agree with the results of the computer simulations , fig .
[ v - as ] .
further , the results of fig . [
v - as ] shows that the stationary velocities are independent of @xmath64 in the limit of an nearly ideal energy depot , which is also indicated by eq .
( [ 3stat - c0 ] ) .
in this paper , we have investigated the dynamics of brownian particles with an internal energy depot , which can be filled by the take - up of energy from the environment .
this extension which is inspired by the biological features of active motion , has several aspects : ( i ) the spatial or spatial - temporal inhomogeneous distribution of energy from the environment can be considered , ( ii ) the supply of depot energy for different activities , i.e. for accelerated motion or signal - response behavior , can be modeled , ( iii ) in addition to external dissipative processes caused by friction , internal dissipation can be included , ( iv ) the adjustment of the time scale for the relaxation of the energy depot allows to consider delay effects , i.e. for the acceleration of motion . in this paper , we are mainly interested in the question how the internal energy depot will affect the particle s motion in an external potential .
we found that provided some supercritical energetic conditions , the brownian particles are able to move in a `` high velocity '' or active mode of motion characterized by a velocity much larger than the stokes velocity .
the latter one can be denoted as the _ passive _ mode of motion .
in addition to stochastic forces , it is basically determined by the response to the gradient of an external potential .
one of the possible active modes of motion can be understood as a continuation of the passive or stokes mode , the particles motion in the direction of the force being accelerated . _
additionally _ , we also found an active mode which describes a motion of the particle _ against _ the gradient of an external potential .
we have investigated the critical conditions for the deterministic motion of the particle in the overdamped limit for the case of a linear potential .
we found that the `` uphill '' motion is described by two critical conversion rates .
the bifurcation diagram shows that , at a critical value @xmath178 , the possibility of an `` uphill '' motion appears as a new solution for the stationary velocity , i.e. `` uphill '' motion requires the existence of multiple stationary solution for the velocity .
however , this active mode remains unstable as long as the conversion rate is below a second critical value @xmath92 , eq .
( [ d2-f ] ) , which depends on parameters describing the energy balance , such as the friction coefficient , @xmath3 , and the influx of energy into the depot , @xmath114 , and on the gradient of the potential , i.e. the resulting force @xmath45 .
a _ stable _
`` uphill ''
motion occurs only for a supercritical conversion rate .
these results allow us to interpret the deterministic motion of an ensemble of particles in a ratchet potential , i.e. a piecewise linear , periodic asymmetric potential with two different slopes .
initially , the particles are equally distributed over the first period of the potential . in order to produce a current
, the particles must be able to escape from the potential well , which depends on the supply of energy .
we are able to find two different critical conversion rates , which describe the onset of a directed current into the different directions .
as long as the particles are only able to move uphill the flank with the lower slope , we have an unimodal velocity distribution , corresponding to a negative net current .
but if the particles are able to move uphill either flank , we find a bimodal velocity distribution , and a positive net current for the given conditions .
hence , in the deterministic case the direction of the net current can be controlled via a single parameter , @xmath148 , which describes the conversion of internal into kinetic energy .
we are able to calculate the critical parameter , as well as the resulting velocity of the net current from analytical considerations , which agree with the results found in the computer simulations .
alternatively , the critical supply of energy can be also reformulated in terms of a critical asymmetry of the ratchet potential , which restricts the occurence of a ( positive or negative ) net current . to conclude the results obtained , we have shown that an ensemble of pumped brownian particles moving in a ratchet potential can produce a _ directed net current_. in this respect
, our result agrees with the conclusions of other physical ratchet models which have been proposed to reveal the microscopic mechanisms resulting in directed movement . due to
@xcite a _ ratchet system _ which is meant to be `` a system that is able to transport particles in a periodic structure with nonzero macroscopic velocity although on _ average _ no macroscopic force is acting . ''
indeed , we have shown in eq .
( [ aver-2-a ] ) that for the stationary approximation @xmath179 holds , i.e. the force acting on the particle is of zero average with respect to one period , @xmath121 . however , different from other ratchet models , ie .
the rocking ratchet @xcite or the diffusion ratchet @xcite which assume a spatially uniform time - periodic force , the force in our model switches between two constant values , dependent on the moving direction and the flank the particle is moving on . hence , it does not represent a spatially uniform force , and we may conclude that the mechanism of motion which eq .
( [ x - overd ] ) is based on , should be different from the previous mechanisms which originate directed motion in a ratchet potential . in this paper
, we have restricted the discussion to the deterministic motion of brownian particles with an internal energy depot .
we note that the stochastic motion of an ensemble of brownian particles in a ratchet potential will be investigated in detail in a forthcoming paper @xcite , which pays particular attention to the influence of stochastic effects on the establishment of the net current .
additionally , we will also discuss the non - equilibrium distribution functions for brownian particles with internal energy depot @xcite . as a final remark , we note in a more general sense that the introduction of an internal energy depot adds an interesting and new element to the known model of brownian particles , which can be useful in two different respects .
with respect to the physical aspects , the dynamical system now has a _ new degree of freedom _ , which increases the phase space , @xmath180 .
due to the additional energy supply , the system is driven into non - equilibrium .
hence , a qualitative new behavior in the particle s motion can be obtained . with respect to possible biological aspects ,
the extension of the brownian particle model by mechanisms of energy take - up , storage and conversion , should contribute to the development of a _ microscopic theory _ of active biological motion .
the final goal of such a project could be a microscopic image of well known phenomenological models of biological motion , taking into account energy balances that are related to the mechanisms of energy pumping and energy dissipation . | we consider brownian particles with the ability to take up energy from the environment , to store it in an internal depot , and to convert internal energy into kinetic energy of motion .
provided a supercritical supply of energy , these particles are able to move in a `` high velocity '' or active mode , which allows them to move also against the gradient of an external potential .
we investigate the critical energetic conditions of this self - driven motion for the case of a linear potential and a ratchet potential . in the latter case , we are able to find two different critical conversion rates for the internal energy , which describe the onset of a directed net current into the two different directions .
the results of computer simulations are confirmed by analytical expressions for the critical parameters and the average velocity of the net current .
further , we investigate the influence of the asymmetry of the ratchet potential on the net current and estimate a critical value for the asymmetry in order to obtain a positive or negative net current .
* uphill motion of active brownian particles + in piecewise linear potentials * + frank schweitzer@xmath0 , benno tilch@xmath1 , werner ebeling@xmath2 + _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * _ institute of physics , humboldt university berlin , invalidenstrae 110 , d-10115 berlin , germany _ * _ ii .
institute of theoretical physics , university of stuttgart , pfaffenwaldring 57/iii , d-70550
stuttgart , germany _ * present address : _ gmd institute for autonomous intelligent systems , schloss birlinghoven , 53754 sankt augustin , germany , e - mail : [email protected]_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ + 05.40.jc brownian motion + 05.45.-a nonlinear dynamics and nonlinear dynamical systems + 05.60.-k transport processes + 82.20.mj nonequilibrium kinetics + 87.10.+e general theory and mathematical aspects |
the search continues for a theory which describes gravity on quantum scales , and provides us with falsifiable experimental predictions ( for a review , see , for example , @xcite ) . in recent years
, the subject of quantum gravity phenomenology has rapidly grown ( see @xcite for a review ) , complementing theoretical work .
one particular feature of quantum gravity which has attracted much attention in particle and astro - particle physics phenomenology is the breaking of cpt symmetry ( see , for example , @xcite for a review ) .
cpt symmetry is anticipated to be broken in quantum gravity because the prerequisite conditions of quantum coherence , locality or lorentz invariance may no longer hold . in previous work
@xcite we have studied the effect of quantum decoherence on atmospheric and astrophysical neutrino oscillations , and in the present work we turn our attention to lorentz invariance violation .
we focus on atmospheric neutrino oscillations , having considered lorentz invariance violating effects on astrophysical neutrinos in @xcite . as in our previous work @xcite ,
in this paper we take a phenomenological viewpoint .
we examine , independent of any one particular theory of quantum gravity , what we may expect to observe if lorentz symmetry is broken or deformed .
there is already a large body of work relating to lorentz invariance violation and neutrinos , some of which is briefly reviewed in the following section .
furthermore , existing atmospheric neutrino data has been thoroughly analyzed for both lorentz invariance violating effects and other new physics ( see , for example , @xcite ) .
although the analysis of super - kamiokande and k2k data @xcite strongly disfavours lorentz invariance violation as either the leading or sub - leading cause of atmospheric neutrino oscillations , an analysis of high - energy atmospheric neutrino oscillations , as would be observed at a neutrino telescope , would enable a wider range of lorentz invariance violating models to be probed , as well as offering the possibility of further strengthening existing bounds on lorentz invariance violating parameters .
this can be seen in the recent analysis of macro data @xcite using both low ( @xmath0 30 gev ) and high ( @xmath1 130 gev ) energy neutrinos . in this paper
, we consider the sensitivity of the antares neutrino telescope @xcite to lorentz invariance violating effects although we expect a similar analysis to be applicable to other neutrino telescopes , such as icecube @xcite , nestor @xcite and nemo @xcite . whilst neutrino telescopes are primarily designed to observe neutrinos from astrophysical and cosmological origins , atmospheric neutrinos form the primary background to these sources and it is likely that they will provide the first physics results .
the outline of this article is as follows . in section [ sec : theory ] , we briefly discuss some of the theoretical consequences from lorentz invariance violation , in particular modified dispersion relations .
we then model how modified dispersion relations alter atmospheric neutrino oscillations in section [ sec : mdr ] .
we present the results of our simulations for the antares neutrino telescope in section [ sec : sim ] , where we first consider how the observed spectra of events may be altered by lorentz invariance violating effects and then present the results of our sensitivity simulations . in section [ sec : conc ] , we present our conclusions and compare the results obtained here with those from previous analyses of experimental data @xcite .
unless otherwise stated , we use units in which @xmath2 .
in quantum gravity , space - time is expected to take on a foamy nature @xcite .
models of space - time foam have been built in many frameworks , including critical @xcite and non - critical string theory @xcite , and loop quantum gravity @xcite .
interactions with this space - time foam may lead to the breaking of cpt symmetry , leading to the violation of lorentz invariance @xcite .
in addition , some theories of quantum gravity predict that there is a minimum length scale , of order the planck length ( @xmath3 m ) @xcite .
if this is the case , then the existence of a fundamental length scale may also lead to lorentz invariance violation ( lv ) . to illustrate this ,
imagine if we were able to somehow measure this fundamental length scale .
we could measure it in our lab frame and probably find that it was of order the planck length , @xmath3 m. if we now repeated the experiment in a different inertial frame ( moving relative to the first frame ) , we would expect the length to be shorter , as predicted by the usual lorentz contraction .
however , we are considering a fundamental length scale and , therefore , we would also expect to measure the same length in all inertial frames .
there is , therefore , either an error in our naive thought experiment or lorentz symmetry , at least at the planck scale , is broken .
therefore , there exist three possibilities for the fate of lorentz invariance within theories of quantum gravity : 1
. it may be that our naive thought experiment is wrong , or that a fundamental length scale does not exist , and so lorentz invariance remains intact at planck scales ; 2 . it may be that , at the planck scale , lorentz invariance is broken and there is a preferred frame ; 3 .
it may be that lorentz invariance as we know it is deformed , so , for example , the lorentz transformation involves a second quantity which is not observer dependent .
so , perhaps the planck length is invariant in an identical way to the speed of light . in the second and third options , one consequence of such breaking / deforming of lorentz invariance
may be that quantities we think of as being lorentz invariant , such as the standard dispersion relation , @xmath4 , may also have to be modified .
lorentz invariance violation may manifest itself in many different ways and in many different experimental systems @xcite .
our focus here is on neutrinos and modified dispersion relations .
however , modifications of dispersion relations have also been considered for photons @xcite , leading to dispersion relations of the form @xcite : @xmath5 where @xmath6 is the frequency of the photon , @xmath7 its momentum and @xmath8 is the mass scale of the quantum gravitational theory . by considering photons which originate in astrophysical objects such as gamma ray bursts and pulsars , lower bounds of @xmath9 gev @xcite were placed on @xmath8 by considering the delay in arrival times . in the literature
, the effects of lv on neutrino oscillations can be broken down into two distinct categories . in the first category
are those models in which the only modification is to the neutrino oscillation length @xcite , which is the type of lv model we shall consider in this paper . in the second category
are those models in which both the oscillation length and also the mixing angle are altered . in our simulations , we found that including lv effects on the mixing angle as well as the neutrino oscillation length made no difference to our results and so we do not study this type of model in detail .
this type of lv , where both the mixing angle and oscillation length are altered , has been considered for the icecube @xcite and minos @xcite experiments in addition to future neutrino factories @xcite .
our interest in this paper is how lv effects , as described in the previous section , may alter the oscillations of atmospheric neutrinos .
therefore , in the analysis which follows , we restrict ourselves to a simple two neutrino system although it is possible to consider how these effects manifest themselves in a three neutrino system @xcite .
if we allow lv then this may lead to modifications of the dispersion relation @xcite : @xmath10 where @xmath11 is the energy of the neutrino , @xmath12 the neutrino momentum , @xmath13 is the mass eigenvalue of the neutrino and @xmath14 is the planck energy .
the function @xmath15 contains all the novel lv effects . for our analysis here , we find it useful to parameterize this function so the dispersion relation ( [ lvtheory : eqn : moddisp ] ) becomes @xcite @xmath16 where @xmath17 and @xmath18 are parameters of the lorentz invariance violating theory . using ( [ lvtheory : eqn : parmoddisp ] ) and assuming the parameter @xmath17 has a dependence upon the mass eigenstate
, then we find that the hamiltonian in the mass basis , for a two neutrino system , may be written , to first order in @xmath17 , as @xmath19 where @xmath20 and @xmath21 are the mass and lv parameter of the @xmath22th eigenstate . here
we are considering only the case when the hamiltonian is diagonal in the mass basis , as off - diagonal terms will modify the neutrino mixing angle as well as the oscillation length . the probability that a neutrino of flavour @xmath23 oscillates to one of flavour @xmath24 is then given by @xmath25 = \left|\sum_{i=1}^{2}u_{ai}e^{-h_{ii}}u_{bi}^{*}\right|^{2},\ ] ] where the @xmath26 s are the components of the neutrino mixing matrix : @xmath27 with @xmath28 being the mixing angle .
the oscillation probability is therefore @xmath29 = \sin^{2}2\theta\sin^{2}\left[\frac{\delta m^{2}l}{4e}+\frac{\delta\eta e^{\alpha+1}l}{4e_{p}^{\alpha}}\right],\ ] ] where @xmath30 and @xmath31 .
replacing @xmath32 and @xmath33 and absorbing the planck energy into the parameter @xmath34 gives @xmath35 = \sin^{2}2\theta\sin^{2}\left[1.27\frac{\delta m^{2}l}{e}+1.27\times10^{9(n+1)}\delta\eta e^{n}l\right],\ ] ] where @xmath36 is measured in ev@xmath37 , the neutrino energy @xmath11 in gev , the path length @xmath38 in km , and we have set @xmath39 .
if we set @xmath40 in ( [ lvtheory : eqn:2nprobch ] ) , then we recover the standard oscillation probability @xmath41 = \sin^{2}2\theta\sin^{2}\left[1.27\frac{\delta m^{2}l}{e}\right],\ ] ] whereas , if we set @xmath42 , we find a probability in which neutrinos oscillate due to lv effects only : @xmath43 = \sin^{2}2\theta\sin^{2}\left[1.27\times10^{9(n+1)}\delta\eta e^{n}l\right ] .\ ] ] in ( [ lvtheory : eqn:2neutlvham ] ) , we have assumed that the lv parameter @xmath17 has a dependence upon the mass eigenstate .
if this is not the case and the parameter @xmath17 is universal to all mass eigenstates , then , even if the theory violates lorentz invariance , the neutrino oscillation probability is unchanged . in this paper
, we shall consider three models with differing energy dependences , namely : 1 .
model lv1 - with @xmath44 : here the lv parameter @xmath45 in ( [ lvtheory : eqn:2nprobch ] ) is dimensionless ; 2 .
model lv2 - with @xmath46 : then @xmath45 has units of ev@xmath47 ; 3 .
model lv3 - with @xmath48 : in which case @xmath45 has units ev@xmath49
. it may be , in a phenomenology of quantum gravity , that @xmath50 is not an integer but we only consider integer values here ( the results for non - integer @xmath50 can be found by interpolating between the two bounding integer values ) .
models with @xmath51 , @xmath52 and @xmath53 have been examined using super - kamiokande @xcite data , and more complicated models with @xmath54 or @xmath53 have been analyzed using both super - kamiokande and k2k data @xcite .
we do not consider @xmath55 here as @xmath54 models have been severely constrained by these previous analyses , and , since in that case the lv modifications of the oscillation probability are independent of the neutrino energy , we do not expect to be able to improve on their bounds by using higher - energy atmospheric neutrinos .
therefore we only consider values of @xmath50 which are strictly positive , and include @xmath44 for comparison with previous work @xcite . from our work on quantum decoherence effects on atmospheric neutrino oscillations @xcite
, one might anticipate that we will be able to place more stringent bounds on lv effects as @xmath50 increases . in the next section ,
we shall be concerned with finding upper bounds on the lv parameters .
however , we are able to extract crude order of magnitude estimates for these parameters from the oscillation probability ( [ lvtheory : eqn:2nprobch ] ) .
the lv effects only become significant when @xmath56 the peak in the atmospheric flux is of the order of @xmath53 gev , which is the typical energy scale for experiments such as super - kamiokande @xcite , whilst neutrino telescopes , such as antares , are sensitive to much higher energy neutrinos .
table [ tab : estbounds ] shows a comparison of the typical bounds on @xmath34 for low and high energy neutrino experiments . here
, we have taken the peak in the atmospheric flux , @xmath57 gev as the value of the neutrino energy for low energy experiments and @xmath58 gev as a typical value for neutrinos to be detected by a neutrino telescope such as antares . as a comparison , the `` most conservative '' bound for @xmath44 from super - kamiokande data @xcite is @xmath59 at the 90% confidence level .
the authors of @xcite comment that the strongest bounds are coming from upgoing muon events , which peak at about 100 gev , so their results are in line with naive expectations .
the bounds of @xcite are improved by a factor of about 8 in the combined analysis of super - kamiokande and k2k data @xcite .
we would hope that the much greater number of such events that will be observed at a neutrino telescope would enable these bounds to be improved . to see how lorentz invariance violation affects atmospheric neutrino oscillations , it is useful to compare the oscillation probabilities . whilst the neutrino oscillation probabilities themselves are not observables , they are useful for gaining insight into the underlying physics .
firstly , for comparison , we consider the standard oscillation probability which is given by ( [ lvtheory : eqn : standprob ] ) . in all the probabilities presented here , we take @xcite @xmath60 fig .
[ fig : standard1 ] shows the oscillation probability plotted as a function of the path length , @xmath38 , for neutrino energy 200 gev and as a function of the neutrino energy , @xmath11 , for a fixed path length , @xmath61 km ( which is comparable with the diameter of the earth , and used in the oscillation probability plots in @xcite ) . ) .
the left plot shows the oscillation probability as a function of path length @xmath38 ( measured in km ) , for fixed neutrino energy @xmath11 equal to 200 gev .
the right plot show the oscillation probability as a function of energy for path length @xmath61 km.,title="fig:",width=257 ] ) .
the left plot shows the oscillation probability as a function of path length @xmath38 ( measured in km ) , for fixed neutrino energy @xmath11 equal to 200 gev .
the right plot show the oscillation probability as a function of energy for path length @xmath61 km.,title="fig:",width=257 ] figs .
[ fig:2nprobn1 ] , [ lvtheory : fig:2nprobn2 ] and [ lvtheory : fig:2nprobn3 ] show the oscillation probability including lv effects ( [ lvtheory : eqn:2nprobch ] ) for @xmath62 and 3 respectively . to construct these plots
, we used the upper bounds found in section [ sec : sim ] and shown in table [ lvsim : tab : diagbounds ] .
( @xmath63 ) .
the left plot shows the oscillation probability as a function of path length for neutrinos with an energy of @xmath64 gev whilst the right plot shows the oscillation probability as a function of energy for path length @xmath61 km.,title="fig:",width=257 ] ( @xmath63 ) .
the left plot shows the oscillation probability as a function of path length for neutrinos with an energy of @xmath64 gev whilst the right plot shows the oscillation probability as a function of energy for path length @xmath61 km.,title="fig:",width=257 ] but with @xmath46 ( @xmath65).,title="fig:",width=257 ] but with @xmath46 ( @xmath65).,title="fig:",width=257 ] but with @xmath48 ( @xmath66).,title="fig:",width=257 ] but with @xmath48 ( @xmath66).,title="fig:",width=257 ] in each case , we found that for low energy neutrinos , the effects of lv on the neutrino oscillation probability are negligible .
however , for higher energy neutrinos , this is not the case .
if we compare the plots of probability against path length for high energy neutrinos , we see that for standard oscillations , almost no oscillations occur , however , if we include the lv effects , the oscillation probability becomes significant at much shorter path lengths . indeed , as @xmath50 increases
, the probability becomes significant at smaller path lengths .
when we consider the plots in which the probability is a function of energy , we see that for standard oscillations , the probability tends towards zero , where as , in the plots which include lv effects , the probability continues to oscillate with the period of oscillations decreasing for increasing values of @xmath50 .
corresponding plots of atmospheric neutrino survival probabilities can be found in @xcite for the mass and velocity eigenstate mixing lv model .
their plots have some features in common with ours , in particular the existence of oscillations in the probability at high energies , when the survival probability in the standard oscillation scenario has converged to unity . in @xcite , we studied in depth the effect of quantum decoherence on atmospheric neutrino oscillations . in that case , there were also marked differences in the oscillation probability , particularly for high energies and long path lengths .
however , the probabilities with lv are rather different from those when we have quantum decoherence . in the decoherence case ,
the oscillations in the probability are damped ( the rate of damping depending on a number of factors ) , but here we find that the oscillations in the probability continue for ever . while these two possible sources of new physics give very different properties of the oscillation probability , the key test is whether these differences can be observed experimentally .
we turn to this issue in the next section .
the antares neutrino telescope @xcite is sensitive to high energy neutrinos above the energy threshold of 10 gev .
these neutrinos do not form the primary source of neutrinos for antares , instead they form the background to astrophysical and cosmological neutrinos .
our simulation strategy follows that presented in @xcite , so we only briefly outline the key features .
further details are documented in @xcite , and in the analysis of standard atmospheric neutrino oscillations @xcite , on which our simulations are based ( with suitable modifications for the inclusion of lv ) .
atmospheric neutrino events were generated using a monte - carlo ( mc ) production assuming a spectrum of @xmath67 for neutrinos in the energy range @xmath68 .
the angular distribution was assumed to be isotropic .
a total of twenty five years of data were simulated in order that errors from the monte - carlo statistics could be ignored and event weights are used to adapt the monte - carlo flux to a real atmospheric neutrino flux .
the bartol theoretical flux @xcite was used here although we do not anticipate the results presented will change using a different flux , however , there may be a significant change to our results if a different spectral index is assumed .
the simulations presented here are based on three years worth of data taking with antares , corresponding to around 10000 atmospheric neutrino events .
all errors are purely statistical with simple gaussian errors assumed .
we are able to produce spectra in either @xmath11 , @xmath69 or @xmath38 and , by altering the oscillation probability in the neutrino oscillation code , we are able to include lv effects
. the path length @xmath38 is related to the zenith angle @xmath70 ( not to be confused with the mixing angle @xmath71 ) as follows , provided @xmath72 : @xmath73 where @xmath74 km is the radius of the earth , and @xmath75 corresponds to a neutrino travelling vertically upwards .
we are therefore interested in typical path lengths of the order @xmath76 km .
antares is particularly sensitive to the zenith angle @xmath70 and hence the path length @xmath38 . in the sensitivity analysis
, we therefore used the spectra in @xmath69 .
however , the spectra in @xmath38 alone also have some interesting features , which we discuss below . in order to produce sensitivity regions , we use a @xmath77 technique which compares the @xmath77 for oscillations with lv , or lv effects alone , with that for the no - oscillation hypothesis .
we leave the total normalization as a free parameter so that our results are not affected by the normalization of the atmospheric neutrino flux .
sensitivity regions at both the 90 and 99 percent confidence level are produced .
although our analysis is independent of the normalization of the atmospheric neutrino flux , in order to produce spectra based on the number of observed events , some normalization is necessary . for standard oscillations with no lv effects , bins at high @xmath69
can be used to normalize the flux as standard oscillations become negligible here . however , since all our lv models are proportional to positive powers of the neutrino energy , lv effects are significant at high energies but negligible at low energies , and thus low @xmath69 . in this case
, we use the low @xmath69 bins to normalize the flux . in our simulations , we consider cases where @xmath36 is zero , so neutrino oscillations arise due to lv effects only , and where @xmath36 is non - zero , so that the lv effects modify standard neutrino oscillations . in the case when @xmath78 is non - zero , we also allow this parameter to vary in order to check whether the experimental point of best fit lies inside our sensitivity regions . in this case , we have three parameters , @xmath36 , @xmath79 and @xmath80 .
we therefore obtain a three dimensional sensitivity region at 90 and 99% confidence levels . in order to consider the details of the sensitivity region , we plot the boundaries of the projection of the surface bounding this volume onto the coordinate planes , thus obtaining three distinct plots .
in addition to data based on the total number of observed events , the spectra of these events , typically in @xmath69 , is also important .
for example , spectral analysis has been performed for super - kamiokande @xcite , kamland @xcite and k2k @xcite .
in addition to being used in the construction of sensitivity regions , the spectra themselves are also important for bounding the size of possible lv effects and possibly ruling effects out due to the observed shape of the spectra .
as explained above , we used the spectra in @xmath69 in the sensitivity analysis , so we focus on those spectra ( apart from fig .
[ lvsim : fig : lvspectralvn=1 ] ) .
we begin by examining the spectra obtained when we consider the possibility that neutrino oscillations occur as a consequence of lv only ( @xmath42 in eq .
( [ lvtheory : eqn:2nprobch ] ) ) , when the lv effects are proportional to the neutrino energy , @xmath44 .
( upper plots ) and @xmath81 ( bottom plots ) . the solid line represents the mc simulation of standard oscillations with the dotted / dashed lines showing the spectra for oscillations from lv only .
the simulated data points correspond to three - years data taking , assuming standard oscillations only with no lv.,title="fig:",width=257 ] ( upper plots ) and @xmath81 ( bottom plots ) .
the solid line represents the mc simulation of standard oscillations with the dotted / dashed lines showing the spectra for oscillations from lv only .
the simulated data points correspond to three - years data taking , assuming standard oscillations only with no lv.,title="fig:",width=257 ] ( upper plots ) and @xmath81 ( bottom plots ) .
the solid line represents the mc simulation of standard oscillations with the dotted / dashed lines showing the spectra for oscillations from lv only .
the simulated data points correspond to three - years data taking , assuming standard oscillations only with no lv.,title="fig:",width=257 ] ( upper plots ) and @xmath81 ( bottom plots ) .
the solid line represents the mc simulation of standard oscillations with the dotted / dashed lines showing the spectra for oscillations from lv only .
the simulated data points correspond to three - years data taking , assuming standard oscillations only with no lv.,title="fig:",width=257 ] fig .
[ lvsim : fig : lvspectralvn=1 ] shows the expected spectra of events as a function of @xmath82 and @xmath83 .
the solid line corresponds to the mc simulation with standard oscillations whereas the dotted / dashed lines are the mc simulations for lv only induced oscillations . in each case , we have also plotted simulated data points , based on the standard oscillation picture with no lv effects . on the right - hand - side
, we show the ratios of the number of events compared to no oscillations .
the spectra show that , for the larger values of the lv parameter , there is a definite suppression in the number of observed events by a factor of approximately two .
if , however , we consider the case in which the lv parameter takes the smaller value , it is more difficult to distinguish oscillations that arise as a consequence of lv effects from those arising in the standard oscillation picture . if we consider the ratios , we note that there are some features which would allow us to differentiate between the two phenomenologies . in the top right frame
, we note that at low energies , the lv model does not result in an oscillation minimum whereas the standard oscillation model does . secondly ,
although the difference is small , the bottom right frame shows a deviation from the standard picture at high values of the zenith angle corresponding to large path lengths .
we will not consider further spectra just in the path length @xmath81 , but note that for all the models we study and all the energy dependences of the lv parameters , there are slight deviations in the shape of the spectra in @xmath81 compared with the standard oscillation case . in the analysis of super - kamiokande data
@xcite it is the spectra in the path - length @xmath38 which are used to bound the magnitudes of lv parameters , as the inclusion of lv effects again alters the shape of the spectra . for the remainder of this section , we focus on the ratios of the numbers of events in each case compared with the expected number of events if there were no oscillations .
[ lvsim : fig : lvspectran=1 ] shows the ratios in the case of standard oscillations plus lv as a function of @xmath82 .
again , we note that large values of the lv parameters result in a large suppression of the number of expected events . also , for very large values of the lv parameter , we see a considerable flattening in the ratio of the number of events compared with no oscillations .
the situation is much trickier , however , for very small values of the lv parameter . in this case , the number of events expected is not suppressed and we also see an oscillation minimum
. therefore , this makes it very difficult to distinguish between standard oscillations and those which are modified to include lv effects , if the lv parameter is of the order of @xmath84 or smaller .
we first consider the spectra when the lv parameters are proportional to the square of the neutrino energy .
[ lvsim : fig : lvspectralvn=2 ] shows the ratio of number of events with the lv effects only to those expected if there were no oscillations . for large values of the lv parameters , we note that the spectra are very different from standard neutrino oscillations .
firstly , as in the case before , the number of events is significantly suppressed for very large values of the lv parameter . for standard neutrino oscillations ,
we observe an oscillation minimum , however , for large values of the lv parameter , the ratio increases and so we see a peak . even for smaller values of the lv parameter , the ratio is flat and so by considering the spectra it should be possible to distinguish between these two models . fig .
[ lvsim : fig : lvspectran=2 ] shows the spectra in the case where the lv effects modify standard oscillations , then , as in subsection [ subsec : lv1 ] before , it becomes much more difficult to untangle any lv effects from those of standard oscillations for small lv parameter .
but the dotted / dashed lines show the spectra for standard oscillations plus lv.,width=377 ] for large values of the lv parameter , the only distinguishing feature is that the number of events we expect to see is somewhat reduced .
in addition , for large values of the lv parameter , the curve is much flatter than we would expect from standard oscillations .
we find qualitatively similar results when the lv effects are proportional to the cube of the neutrino energy , both in the case where oscillations are due to lv effects only , and when lv effects modify standard neutrino oscillations . for both models lv2 and lv3
, the shape of the spectra in the zenith angle @xmath81 only are slightly different to that shown in for model lv1 in fig .
[ lvsim : fig : lvspectralvn=1 ] .
the ratios of the number of events compared with no oscillations are flatter than in the bottom right - hand plot in fig .
[ lvsim : fig : lvspectralvn=1 ] , particularly for model lv3 .
we do not use these spectra for the following sensitivity analysis , but it may be possible , with real data , to use those spectra as well as those for numbers of events in terms of @xmath85 . in the previous section , we saw how lv effects can modify the number of events seen in the detector and how they would modify the spectra .
we now turn to the discussion of the antares sensitivity regions found from our numerical simulations . as in the case of quantum decoherence @xcite , we are particularly interested in finding upper bounds for the lv parameters . in our first model ,
the lv effects are proportional to the neutrino energy .
we firstly consider the case in which @xmath42 , so that there are no standard oscillations , only those which arise as a consequence of the lv effects .
[ lvsim : fig : lvsenselv1 ] shows the sensitivity curves in this case .
it is clear that the smallest values of the lv parameter may be probed when @xmath80 is close to one , which is the region of parameter space in which we are most interested .
we are able to place an upper bound on @xmath45 of @xmath86 .
we also considered the case of non - zero @xmath36 , when standard atmospheric neutrino oscillations are modified by lv effects . with no lv effects @xcite.,title="fig:",width=257 ] with no lv effects @xcite.,title="fig:",width=257 ] with no lv effects @xcite.,title="fig:",width=257 ] fig .
[ lvsim : fig : lvsensen=1 ] shows the sensitivity contours in this case . since we have varied three parameters ( @xmath87 , @xmath36 and @xmath45 ) , we obtain a sensitivity volume .
projecting the boundary surface of this volume onto the co - ordinate planes gives us the sensitivity contours plotted in fig .
[ lvsim : fig : lvsensen=1 ] . as we see from fig .
[ lvsim : fig : lvsensen=1 ] , the experimental point of best fit of @xmath36 @xcite , denoted by the triangle , lies within the sensitivity region .
however , the region extends down to very small values of @xmath36 and so we find that our results include the previous case where neutrino oscillations arise as a consequence of lv effects only . the top right frame in fig .
[ lvsim : fig : lvsensen=1 ] shows the sensitivity contour for the @xmath45 as a function of the mixing angle , @xmath28 .
again , @xmath88 is in this region , and so standard oscillations with no lv effects are also included , with an upper limit shown in table [ lvsim : tab : diagbounds ] .
this is the same order of magnitude as the upper bound from super - kamiokande and k2k data @xcite so it seems unlikely that antares will be able to significantly improve on current data .
we have performed a similar analysis for the model lv2 , when the lv effects are proportional to the neutrino energy squared .
the sensitivity regions for the cases where oscillations are due to lv effects only , and where standard oscillations are modified by lv effects , are very similar in shape to those shown in figs .
[ lvsim : fig : lvsenselv1 ] and [ lvsim : fig : lvsensen=1 ] respectively . in particular , the experimental best - fit point shown by a triangle in fig .
[ lvsim : fig : lvsensen=1 ] again lies within the sensitivity region when standard oscillations are modified by lv effects . in this case
, the sensitivity region again extends to both @xmath42 and @xmath88 .
the upper bound on @xmath45 from this sensitivity region is given in table [ lvsim : tab : diagbounds ] . the final model we consider has @xmath48 .
if we set @xmath42 , so that oscillations are due to lv effects only , then we find that we are unable to derive any meaningful sensitivity contours , so we do not consider this possibility further . for @xmath89 ,
the sensitivity contours are shown in fig .
[ lvsim : fig : lvsensen=3 ] . with no lv effects @xcite.,title="fig:",width=257 ] with no lv effects @xcite.,title="fig:",width=257 ] with no lv effects @xcite.,title="fig:",width=257 ] the sensitivity contours in this case have a rather simpler shape than in models lv1 and lv2 . again the experimental point of best fit is included in the sensitivity region but in this case , @xmath42 is not included , so it seems that we are not sensitive to the case when oscillations arise from lv effects only for this energy dependence .
this is in direct agreement with the first part of our analysis as we were unable to find any meaningful results for this model when we considered lv effects only .
we note however that @xmath88 is contained in the sensitivity region . the upper bound on @xmath45 in this sensitivity region
is given in table [ lvsim : tab : diagbounds ] .
we have shown that neutrino telescopes , such as antares , will be able to place stringent bounds on lv parameters which modify the dispersion relation for massive neutrinos .
table [ lvsim : tab : diagbounds ] shows the upper bounds for this model for the various energy dependences we have studied .
we found that if the lv effects were proportional to either the neutrino energy or the neutrino energy squared , then we are sensitive to oscillations induced from these effects only , with no contributions from standard oscillations .
however , for the case of @xmath48 , then we can not measure the effects of lv only in atmospheric neutrino oscillations .
for all values of @xmath50 studied , we have been able to produce sensitivity curves for standard atmospheric neutrino oscillations modified by subdominant lv effects .
we should stress that although our simulations are based on the antares neutrino telescope , we expect that other neutrino telescopes , such as amanda and icecube , will also be able to probe these models . from a naturalness perspective
, we would expect any lv effects to be suppressed by the planck mass , @xmath90 gev . since our lv parameters @xmath45 contain powers of the planck mass already , this gives @xmath91 ev@xmath92 for @xmath93 .
these values are also given in table [ lvsim : tab : diagbounds ] for comparison .
we note that for the case @xmath46 , the bounds obtained from our simulations surpass the expected value of the lv parameter by seven orders of magnitude and so antares may be able to observe these effects or rule out this model . for the case
@xmath48 , however , the bound from our simulations is nine orders of magnitude larger than expected and so we can not rule out this model using atmospheric neutrinos . in this case better sensitivity can be obtained using higher - energy , astrophysical neutrinos @xcite .
the situation is somewhat more complicated if @xmath44 as this corresponds to @xmath63 in ( [ lvtheory : eqn : parmoddisp ] ) and so we have no indication as to the value of @xmath34 except to say that it should be small . in all three cases
, we can see from table [ lvsim : tab : diagbounds ] that our estimates based on the values of the lv parameters at which lv effects become comparable to standard neutrino oscillations are very similar to the upper bounds we find from our sensitivity curves .
our results for the case @xmath44 are comparable with the model investigated in @xcite with super - kamiokande and k2k data .
we find that we are not able to significantly improve on their results ; however an analysis of atmospheric neutrino data from antares should be able to confirm their upper bounds .
although in a rather different lv model , our results are also comparable in magnitude to those found by the macro collaboration @xcite . in that paper , the argument of the sine term in the oscillation probability is identical to that considered here and they found an upper bound on @xmath94 in the range @xmath95 , similar to that found here and shown in table [ lvsim : tab : diagbounds ] .
our results are also comparable to those found in @xcite using data from the amanda experiment . there ,
an upper bound of @xmath96 was placed on the lv parameter . in @xcite , we studied in depth the spectra of numbers of events for standard atmospheric neutrino oscillations with quantum decoherence .
those spectra had some of the same key features as the ones presented here in figs .
[ lvsim : fig : lvspectralvn=1]-[lvsim : fig : lvspectran=2 ] , in particular , a suppression in the expected number of atmospheric neutrino events observed by antares .
the shape of the spectra do have some differences , but it remains to be seen whether they will really be distinguishable experimentally . at this point , in the absence of real data , one can only speculate on what might be observed , but one can foresee two possibilities .
firstly , it may be that antares ( or any other neutrino telescope ) does not observe any distinguishable difference in numbers of events compared with the standard atmospheric neutrino oscillation model .
we argued in @xcite that such a null observation would be able to place very stringent bounds on quantum decoherence effects , and we draw the same conclusion here from our event spectra .
secondly , it may be the case that antares or an equivalent neutrino telescope does indeed find evidence for a suppression in the number of events compared with the standard picture .
in this situation it is likely to be very challenging experimentally to disentangle the possible source of new physics .
the authors acknowledge the antares collaboration for supplying a parametrization of the response of a neutrino telescope .
this work was supported by pparc , and dm was supported by a phd studentship from the university of sheffield .
we would like to thank nick mavromatos for many helpful discussions . c. kiefer , _ ann .
phys . _ * 15 * , 129 ( 2005 ) , .
g. amelino - camelia , _ lect .
notes phys . _
* 669 * , 59 ( 2005 ) , .
n. e. mavromatos , _ lect .
notes phys . _
* 669 * , 245 ( 2005 ) , .
d. morgan , e. winstanley , j. brunner , and l. f. thompson , _ astropart .
phys . _ * 25 * , 311 ( 2006 ) , .
d. hooper , d. morgan , and e. winstanley , _ phys .
lett . _ * b609 * , 206 ( 2005 ) , .
d. hooper , d. morgan , and e. winstanley , _ phys .
_ * d72 * , 065009 ( 2005 ) , .
g. l. fogli , e. lisi , a. marrone and g. scioscia , _ phys .
rev . _ * d60 * , 053006 ( 1999 ) , .
m. c. gonzalez - garcia and m. maltoni , _ phys .
_ * d70 * , 033010 ( 2004 ) , .
g. battistoni et al . , _ phys .
* b615 * , 14 ( 2005 ) , ; m. giorgini , _ czech .
_ * 56 * a291 ( 2006 ) , .
e. v. korolkova ( antares collaboration ) , _ nucl .
suppl . _ * 136 * , 69 ( 2004 ) , ; m. spurio ( antares collaboration ) , _ status report of the antares project _ , .
j. ahrens et al .
( icecube collaboration ) , _ nucl .
_ * 118 * , 388 ( 2003 ) , ; a. goldschmidt ( icecube collaboration ) , _ nucl .
suppl . _ * 110 * , 516 ( 2002 ) ; p. desiati ( icecube collaboration ) , _ icecube : toward a km@xmath97 neutrino telescope _ , .
j. a. wheeler , _ superspace and the nature of quantum geometrodynamics _ , in : c. dewitt , j. a. wheeler ( eds . ) , battelle rencontres : 1967 lectures in mathematical physics ( w. a. benjamin , new york , 1968 ) .
g. t. horowitz , _ new j. phys . _ * 7 * , 201 ( 2005 ) , .
t. mohaupt , _ lect .
notes phys . _
* 631 * , 173 ( 2003 ) , ; j. h. schwarz , _ introduction to superstring theory _ in : proceedings of techniques and concepts of high - energy physics , u.s.a .
, edited by h. b. prosper and m. danilov ( nato , 2000 ) , .
a. dhar , _ current science _ * 81 * , 1598 ( 2001 ) ; e. dhoker , _ mod .
lett . _ * a6 * , 745 ( 1991 ) ; e. dhoker and p. s. kurzepa , _ mod .
lett . _ * a5 * , 1411 ( 1990 ) ; j. distler and h. kawai , _ nucl . phys . _
* b321 * , 509 ( 1989 ) ; j. r. ellis , n. e. mavromatos , and d. v. nanopoulos , _ mod .
lett . _ * a12 * , 1759 ( 1997 ) , ; n. e. mavromatos and j. l. miramontes , _ mod .
lett . _ * a4 * , 1847 ( 1989 ) . c. rovelli , _ living rev .
_ * 1 * , 1 ( 1998 ) ,
. o. w. greenberg , _ phys .
lett . _ * 89 * , 231602 ( 2002 ) , . c. rovelli and l. smolin , _ nucl .
phys . _ * b442 * , 593 ( 1995 ) , .
d. mattingly , _
living rev .
* 8 * , 5 ( 2005 ) , .
s. r. coleman and s. l. glashow , _ phys .
_ * d59 * , 116008 ( 1999 ) , .
j. r. ellis , n. e. mavromatos , and d. v. nanopoulos , _ gen .
* 32 * , 127 ( 2000 ) , ; r. gambini and j. pullin , _ phys .
_ * d59 * , 124021 ( 1999 ) , .
j. albert et al .
( magic collaboration ) , _ probing quantum gravity using photons from a mkn 501 flare observed by magic _ , ; j. bolmont , j. l. atteia , a. jacholkowska , f. piron and g. pizzichini , _ study of time lags in hete-2 gamma - ray bursts with redshift : search for astrophysical effects and quantum gravity signature _ , ; j. r. ellis , n. e. mavromatos , and d. v. nanopoulos , _ probing models of quantum space - time foam _ , in : proceedings of beyond the desert 1999 : accelerator , non - accelerator and space approaches into the next millenium , edited by h. v. klapdor - kleingrothaus and i. v. kirvosheina ( iop , 2000 ) , ; j. ellis , n. e. mavromatos , d. v. nanopoulos , and a. s. sakharov , _ astron .
astrophys . _
* 402 * 409 ( 2003 ) , ; j. ellis , n. e. mavromatos , d. v. nanopoulos , a. s. sakharov and e. k. g. sarkisyan , _ astropart .
phys . _ * 25 * 402 ( 2006 ) , ; r. lamon , n. produit and f. steiner , _ study of quantum gravity effects in integral gamma - ray bursts _ , .
g. lambiase , _ mod .
_ * a18 * , 23 ( 2003 ) , .
j. christian , _ phys .
rev . _ * d71 * , 024012 ( 2005 ) , .
j. l. kelley ( icecube collaboration ) , _ testing lorentz invariance using atmospheric neutrinos and amanda - ii _ , in : proceedings of the first workshop on exotic physics with neutrino telescopes , edited by c. de los heros ( uppsala university , 2007 ) , .
m. c. gonzalez - garcia , _ non - standard neutrino oscillations at icecube _ , in : proceedings of the first workshop on exotic physics with neutrino telescopes , edited by c. de los heros ( uppsala university , 2007 ) , ; m. c. gonzalez - garcia , f. halzen and m. maltoni , _ phys .
_ * d71 * , 093010 ( 2005 ) , . g. amelino - camelia , _ on the fate of lorentz symmetry in loop quantum gravity and noncommutative spacetimes _ , . g. amelino - camelia , j. kowalski - glikman , g. mandanici , and a. procaccini , _ int . j. mod . phys . _ * a20 * , 6007 ( 2005 ) , .
y. ashie et al .
( super - kamiokande collaboration ) , _ phys .
_ * 93 * , 101801 ( 2004 ) , . m. c. gonzalez - garcia and c. pena - garay , _ phys .
rev . _ * d68 * , 093003 ( 2003 ) , . f. blondeau and l. moscoso ( for the antares collaboration ) , _ detection of atmospheric neutrino oscillations with a 0.1 km@xmath37 detector : the case for antares _
, now 98 conference , amsterdam , netherlands ( 1998 ) ; j. brunner ( for the antares collaboration ) , _ measurement of neutrino oscillations with neutrino telescopes _
, 15th international conference on particles and nuclei ( panic 99 ) , uppsala , sweden ( 1999 ) ; c. carloganu ( for the antares collaboration ) , _ nucl .
suppl . _ * 100 * 145 ( 2001 ) . | one possible feature of quantum gravity may be the violation of lorentz invariance . in this paper
we consider one particular manifestation of the violation of lorentz invariance , namely modified dispersion relations for massive neutrinos .
we show how such modified dispersion relations may affect atmospheric neutrino oscillations .
we then consider how neutrino telescopes , such as antares , may be able to place bounds on the magnitude of this type of lorentz invariance violation .
lorentz invariance violation , neutrino telescopes 04.60.-m , 11.30.cp , 14.60.pq , 14.60.st , 95.55.vj |
in standard problems of radiation hydrodynamics ( rh ) where radiation contains a large fraction of the energy and momentum density , the boltzmann equation ( be ) must be coupled to the hydrodynamic equations in order to obtain the evolution of the system as well as the spectrum and angular distribution of the radiation field .
however , an algorithm built to solve the be numerically ( a boltzmann solver ) in a non - stationary case is too time - consuming , from a computational point of view , to allow for the extension to more than one dimension of the existing numerical codes @xcite . in many cases , instead of the be , its _ angular moments _ are considered , obtaining then the multigroup ( or multi - frequency ) equations or the even simpler energy averaged equations .
standard rh methods @xcite have found their way into the literature , and one can find several rh codes with different approaches for the radiation transport part , from single - energy - group ( or two temperature ) such as visphot @xcite or titan @xcite , to multigroup radiative transfer , such as stella @xcite .
other authors built codes devoted more specifically to the radiation transport , but at the expense of detailed hydrodynamics .
an example is the code eddington @xcite in which free expansion is assumed .
the hydrodynamics part of most of the existing codes is a one dimensional implicit finite difference scheme , including artificial viscosity terms for problems requiring an accurate treatment of shock waves and discontinuities . during the last decade
, a new subclass of numerical methods , the so - called godunov - type methods , has been gradually substituting the schemes based on numerical viscosity due to their easier extension to multidimensional cases and their greater capabilities in the treatment of strong shocks .
the lack of a radiation transport method fully compatible with hydrodynamical godunov - type schemes is one of the motivations of these paper . turning back to the radiation transport part , another important issue related to the angular moment equations is the closure relation . by taking angular moments of the distribution function the complexity of
be is highly reduced , but the information about the angular dependence of the radiation field is partially lost .
each @xmath0 angular moment equation of the be contains angular moments of higher order , thus any truncated hierarchy of moment equations contains more unknowns than equations and must be supplied with additional equations or _ closure relations _ @xcite . theoretically ,
if the correct closure for a given problem is known , the solution obtained for the first moments of the distribution function by solving the moment equations should be the same as the solution obtained by solving the be .
although the moment equations are much simpler to solve than the be , in many situations an additional simplification is made by neglecting some other terms , leading to the _ diffusion approximation _ ( da ) .
in this approximation , however , the resulting energy flux can be higher than the limit predicted by causality , specially in the regions where the radiation mean free path becomes comparable to the characteristic length .
different extensions of da , like _ flux - limited diffusion _
@xcite or _ artificial opacities _ @xcite
have being used to overcome this problem .
these extensions only partially solve the breakdown of the da in the semi - transparent and transparent layers , and all of them are based on the same idea : to reintroduce some terms which had been dropped in the original assumptions of the da .
the term which is usually kept out of the equations is the time derivative of the fluxes . by neglecting this term ,
the character of the system of equations is changed to parabolic , and causality is therefore violated ( disturbances propagate at infinite velocity ) . in order to develop a method which is valid in all regimes ( from diffusion to free streaming ) while preserving causality , one must solve the full set of equations keeping its hyperbolic character . in this paper
, we address the problem of establishing a well - defined hyperbolic system of equations for the first angular moments of the be in the non - stationary spherically symmetric case .
we will see that , besides the important question of its hyperbolicity , some constraints on the mathematical properties of the closure relations can be derived .
they might help to disregard among some of the closures previously proposed .
moreover , the behaviour of the characteristic fields of the hyperbolic system , which gives information on the velocity of perturbations , can also be used to establish additional constraints to the closures .
a direct consequence of considering the moment equations in hyperbolic form is that it permits the application of powerful numerical techniques that have been developed in recent years for hyperbolic systems of conservation laws ( the equations of classical and relativistic fluid dynamics for perfect fluids , for example ) . among the different numerical techniques , the so - called high resolution shock capturing ( hrsc ) methods
have a number of interesting features such that stability , being conservative , convergence to physical solutions and high accuracy in regions where the solution is smooth .
hrsc methods are based on the resolution of local riemann problems ( an initial value problem with discontinuous data ) at the interfaces of numerical shells , ensuring a consistent treatment of discontinuities and steep gradients @xcite .
their special relativistic extension @xcite has shown its potential in simulations of heavy ion collisions and extra - galactic jets , and different attempts to extend the method to general relativity have been done @xcite .
we refer the interested reader to the recent reviews by mart & mller and ibez & mart for a detailed description of the current status of hrsc techniques in numerical relativistic hydrodynamics .
although hrsc are specially designed for hyperbolic systems without source terms which corresponds to the transparent regime , we will show in numerical experiments that , by appropriate treatment of the sources , accurate solutions can be obtained also in the diffusion regime , where the source terms are dominant .
the structure of the paper is the following : in 2 we summarise the deduction of the angular moment equations of the be in a spherically symmetric case , and for a static background .
we also describe the general form of the collision terms when emission - absorption and iso - energetic scattering processes are included . in
3 we discuss the most common techniques used to solve the transport equations , such as diffusion or flux - limited diffusion , remarking their main features and limitations . in
4 the hyperbolic character of the equations and its implications for the closure relations are discussed . in 5 , the numerical techniques employed to solve the transport equations as a hyperbolic system of conservation laws are discussed .
a number of numerical experiments solving the transport equations in several test problems are displayed in 6 , and the feasibility of the hyperbolic treatment in all kinds of regimes is demonstrated .
finally , main conclusions and advantages of our proposal are summarised in 7 .
we shall start our discussion from the radiative transfer equations in a static medium , deferring to a future work the inclusion of fluid velocity terms in the equations .
although the restriction to zero velocity seems to be specific , there are some astrophysical scenarios where the assumption that matter is at rest is a reasonable approach .
moreover , the inclusion of velocity terms , in some cases of interest does not change the essentials of our conclusions . in a static background ,
the boltzmann equation for massless particles in spherical coordinates can be written as follows . ]
@xmath1 = \left ( \frac { d{\cal i}}{d \tau } \right)_{coll}\ ] ] where @xmath2 is the invariant distribution function , @xmath3 is the cosine of the angle of the particle momentum with respect to the radial direction , @xmath4 the energy of the particle and the right hand side term is the invariant source or collision term coming from the interaction between the radiation field and matter . as stated in the introduction , a common method of solving the be is the method of moments @xcite , which involves taking angular moments of the equation by applying the operator @xmath5 denoting by @xmath6 , @xmath7 and @xmath8 the angular moments of the specific intensity @xmath9 , @xmath10 being the statistical weight ( @xmath11 for photons and @xmath12 for neutrinos ) @xmath13 the equations corresponding to the first two moments ( 0 and 1 ) of the be are : @xmath14 @xmath15 where the moments of the collision term are defined as @xmath16 @xmath17 notice that , in the absence of velocity fields or strong gravitational fields , the different energy groups in a multifrequancy scheme are only coupled through the source terms , rendering the mathematical character of the left hand side of equations ( [ mom0])-([mom1 ] ) formally identical to the energy - averaged problem . the most general form for the collision term including emission , absorption and isoenergetic scattering is the following : @xmath18\right)\ ] ] where @xmath19 is the emissivity , @xmath20 the absorption opacity including final states blocking ( for fermions ) or stimulated emission ( for bosons ) and @xmath21 the scattering opacity related to the scattering reaction rate ( @xmath22 ) through @xmath23 where @xmath24 is the angle between the in - going and outgoing particle and @xmath25 is the azimuthal angle .
the opacities and the emissivity are expressed in units of inverse length . for isotropic scattering the source terms appearing in equations ( [ mom0 ] ) and ( [ mom1 ] )
can be written as @xmath26 @xmath27 where @xmath28 and @xmath29 is the value of @xmath6 in equilibrium with matter . in the above
, we have assumed matter in local thermodynamic equilibrium ( lte ) , and only the radiation field ( photons or neutrinos ) is allowed to deviate from equilibrium .
the conclusions of our study might be affected by large velocity fields or for frequencies close to line discontinuities , which are present in some astrophysical scenarios , and deserve a more careful study @xcite .
one of the approaches most widely used to solve the transport equations numerically is the _ diffusion approximation _ , in which the invariant distribution function is assumed to be nearly isotropic in the comoving frame .
thus an expansion in terms of legendre polynomials to the order @xmath30 is enough to maintain the main features of the radiation field @xmath31 where @xmath32 .
consistent with this assumption , the time derivative of the flux is neglected in equation ( [ mom1 ] ) , and together with equation ( [ q1 ] ) gives the following relation for the flux in terms of the energy gradient @xmath33 as stated before , the da breaks down when the mean free path is large compared to the typical scale of the problem , and the fluxes calculated with the former formula may give non - causal values , in the sense that the flux can be greater than the energy density .
this pathology comes from the fact that we have neglected some terms in the momentum equation in order to obtain a simple formula for the fluxes .
flux limiters have been introduced _
ad hoc _ in the literature to avoid this non - causal behaviour ; see , _
e.g. _ , minerbo , levermore & pomraining or cernohorsky & bludman . in order to illustrate where the problems stem from , we begin by deriving the _ flux - limited diffusion _ equations .
let us define the _ flux factor _ , @xmath34 , and the _ eddington factor _
, @xmath35 : @xmath36 @xmath37 by subtracting equation ( [ mom0 ] ) multiplied by @xmath34 from ( [ mom1 ] ) , and after some algebra ( now keeping all terms ) , one obtains : @xmath38 where the quantities , including the flux and eddington factors , are now energy dependent .
the explicit expression for the different _ opacities _ are @xmath39 @xmath40 the _ physical opacity _ , @xmath41 , or effective albedo only depends on the interaction of radiation with matter and includes an out of equilibrium correction to absorption opacity .
the second term in the denominator , @xmath42 , is the multigroup extension of the so - called _ artificial opacity _
@xcite and takes into account geometrical corrections and deviations from the diffusion limit ( it vanishes in the limit @xmath43 ) .
the term in the numerator of equation ( [ fdif ] ) , @xmath44 , plays the role of a flux limiter .
it has the value @xmath45 in the diffusion limit and goes to zero as the free streaming limit ( @xmath46 , @xmath47 ) is reached . to close the set of equations , a closure relation for @xmath35
is needed , and it is directly related to flux limiter @xmath48 .
we will re - address the closure relation problem in next section . with this description of the transport equations , the simple structure of the diffusion equation and its parabolic character
is maintained when the first term in @xmath42 is neglected ( @xmath49 = 0 ) .
the geometric and velocity field corrections can also be included through artificial opacities as corrections to the physical opacity .
although the flux - limiter helps to avoid non - causal behaviour in the sense @xmath50 , causality is also violated when the time derivative of @xmath34 is neglected , because the character of the equations change to parabolic , which results in an infinite speed of propagation of the information . by keeping this term ,
the diffusion scheme is , in some applications , numerically unstable and further numerical tools are needed to handle these instabilities .
this is the subject of the next section .
the system of equations ( [ mom0 ] ) and ( [ mom1 ] ) can be written in a compact form as follows @xmath51 where the vectors @xmath52 and @xmath53 are @xmath54 @xmath55 and the sources @xmath56 , according to equations ( [ s0]-[s1 ] ) , are given by @xmath57 notice that the sources are free of partial derivative operators .
the above system of equations ( [ system ] ) is said to be a hyperbolic system of conservation laws ( with source terms ) when the jacobian matrix associated to the _ fluxes _ , @xmath58 , has real eigenvalues and there exists a complete set of corresponding eigenvectors . to analyse the jacobian matrix
, we will assume that the eddington factor is only a function of the flux factor @xmath59 .
this assumption is commonly used by a number of authors who have proposed different closure relations or flux - limiters in the context of flux - limited diffusion . in this case , the jacobian matrix is _ = ( ) where @xmath60 .
the eigenvalues of the above matrix are @xmath61 and the right eigenvectors are @xmath62 the mathematical character of the system of equations ( [ system ] ) depends on the value of the discriminant in equation ( [ lambda ] ) , which can be written as follows [ disc ] = ( p-2f)^2 + 4(p -f^2 ) . since @xmath35 and @xmath34 are normalised moments of a non negative weight function , on the unit sphere , there is an extra relation between @xmath34 and @xmath35 , the _ schwarz inequality _ [ schw ]
f^2 p 1 , that ensures that for any closure satisfying ( [ schw ] ) the eigenvalues are always real and hence the hyperbolic character is guaranteed .
in addition , causality imposes that the velocity of propagation of perturbations must be smaller than the speed of light , which gives another constraint on the eigenvalues | _ | 1 , that can be expressed in terms of @xmath34 , @xmath35 and @xmath63 as follows - p .
[ ppb ] this restriction , together with equation ( [ schw ] ) and the condition @xmath64 , defines the region in the @xmath65 $ ] space allowed for causal closures . in many astrophysical problems , radiation diffuses out from a central opaque region to an outer optically thiner region .
in such a scenario , the radiation angular distribution is nearly isotropic in the central region ( @xmath66 ) and forward peaked in the transparent region , far from the opaque centre ( @xmath46 ) .
thus , the closure in that problem must satisfy p(f=0)= p(f=1 ) = 1 .
the first limit comes from the diffusion approximation while the second limit comes from the schwarz inequality ( [ schw ] ) , and both are satisfied by all the closures found in the literature ( see table i ) .
another general property is that , in the diffusion regime the eddington factor has the constant value of @xmath67 , thus another restriction on the derivative can be imposed [ cons2 ] p(f=0 ) = 0 , that leads to the characteristic speeds @xmath68 . in the limit @xmath69 , equations ( [ schw ] ) and ( [ ppb ] ) gives the bounds @xmath70 and the eigenvalues in this limit are @xmath71 and @xmath72 . according to the theory of hyperbolic systems ,
this result can be interpreted as a first wave propagating outwards at the speed of light and a second wave with a characteristic speed that depends on the value of @xmath63 .
although @xmath73 depends on the form of the closure , for @xmath69 the amplitude of this second wave is zero and the value of @xmath63 does not affect the evolution .
however , for @xmath34 close but not equal to one , the choice of @xmath63 can be important , as we will show in section 5.2 .
similar conclusions can be obtained for the case @xmath74 . in table 1
we summarise the main properties of some of the most widely used closures : minerbo , lp @xcite , the linear formula by auer , kershaw s parabolic approach @xcite , the flux - limiters used by bruenn and bowers & wilson , janka s fit to montecarlo simulations of neutrino transport in supernovae @xcite and that corresponding to an isotropically emitting sphere @xcite , named _
vacuum_. there are several comments about table i that we would like to address .
first , one closure relation ( bw ) leads to non - causal values of one eigenvalue in the diffusion regime .
secondly , four closures do not satisfy the condition @xmath75 ( auer , bruenn , bw , vacuum ) .
auer s formula failure is due to an excessive simplification ( it is a linear approach ) while vacuum has been obtained from the assumption of an isotropically radiating sphere , which restricts their validity to @xmath76 .
the closure relations proposed by kershaw , janka , minerbo and lp have the common features of preserving causality leading to reasonable values of the characteristic velocities ( @xmath68 ) in the diffusion regime .
thirdly , it is worthy to point out that all the closures except auer s give @xmath77 when approaching the radially - streaming limit @xmath78 , which makes both eigenvalues be positive .
it ensures that no information can be propagated inwards for sufficiently high values of @xmath34 .
it is clear that these closures can not be used to treat problems in which a small perturbation is propagating inwards from the transparent region , as we will show in next section . in a more general situation ,
the eddington factor depends on both , the flux factor and the occupation density corresponding to the zeroth angular moment of the distribution function , @xmath79 .
the jacobian matrix associated to the fluxes is @xmath80 and now the eigenvalues are @xmath81 where the discriminant is @xmath82 in contrast to ( [ disc ] ) , the sign of the discriminant might be negative for negative values of @xmath83 , and the system would become elliptic instead of hyperbolic . as stated before ,
hyperbolic equations are related to finite speeds of propagation of the perturbations , while elliptic equations are related with boundary problems and do not allow for solutions with propagating signals .
this fact indicates that , in order to preserve hyperbolicity , only closures which give positive values of the discriminant should be considered .
a closure of this kind has been proposed by cernohorsky & bludman using arguments of maximum entropy .
minerbo s , lp s and the vacuum closures can be obtained as limiting cases of cb closure in the low occupation limit ( classical statistics ) , large ocuppation limit for bose - einstein statistics and maximal forward packing limit for fermi - dirac statistics , respectively .
we have checked that cb closure gives real and causal eigenvalues for both , bose - einstein and fermi - dirac statistics , in any regime .
the numerical code employed for the calculations presented in 6 is a finite finite difference scheme in which the left ( l ) and right ( r ) states for the riemann problems set up at each interface have been obtained with a monotonic , piecewise linear reconstruction procedure @xcite .
the numerical fluxes are evaluated following the idea of harten , lax & van leer , that is based on approximate solution of the original riemann problem with a single intermediate state .
the resulting numerical flux is given by ^hll = where @xmath84 and @xmath85 .
the time - advance algorithm is a tvd - preserving , third order runge - kutta that takes into account the influence of the source terms in intermediate steps . the source term in this problem
deserves a special mention .
it acts as a relaxation term which leads to a long - time behaviour governed by the reduced parabolic system described in section 3 .
although high resolution methods for hyperbolic conservation laws usually fail to capture this feature ( unless a very fine grid is used ) , the use of a modified flux @xcite and an implicit linearization of the source term in each runge - kutta step allows us to succeed in the treatment of such stiff relaxation term with coarse grids and time - steps longer than the source time - scale ( @xmath86 ) , as we will show in subsection 6.1 .
although we have constrained the numerical experiments to situations when only scattering processes are permitted , the inclusion of absorption - emission processes can be treated in a similar way .
however it introduces energy and momentum exchange with the matter and makes necessary to solve the equations of radiation transport coupled with the equations of hydrodynamics or the equations of hydrostatic equilibrium with number and heat transfer .
such a problem is out of the scope of this paper , in which we focussed on the radiation part , and will be addressed in forthcoming works .
to check the applicability of hrsc techniques to radiation transport problems , we have performed a number of numerical tests considering several simplified models in which the analytical solution is known . in the remainder of this section
we describe the different tests and display the results .
our intention is to prove that our method can solve radiation transport in all kinds of regimes , from diffusion ( test 1 ) to radially streaming ( test 2 ) , going through the semi - transparent region ( test 5 ) and in situations where strong gradients are developed ( test 4 ) .
ideally , in a multi - group scheme we have as many systems of two equations as energy bins in which the radiation density is splitted . in a radiation hydrodynamics problem
all those pairs of equations are coupled to the matter equations and must be solved together . since our objective is to study the applicability of hrsc techniques for the radiation part , in what follows we assume to have only one energy group or , equivalently , we consider the energy integrated equations .
notice , though , that the conclusions are not affected since , when velocity fields are not taken into account , the equations for different energies are only coupled through the source terms .
a different problem will arise when velocity fields are important or gravitational and doppler red - shift are considered . in those situations
either the anisotropy of the source term ( in the observer frame ) or the advection and aberration terms ( in the comoving frame ) cause the energy redistribution along the photon ( or neutrino ) path and a more careful treatment needs to be done . in the diffusion limit
, the closure is obviously @xmath87 and the time derivative of the flux can be neglected , leading to equation ( [ f1 ] ) . setting the absorption opacity to zero and the scattering opacity to a constant value , the analytical solution of the system formed by ( [ mom0 ] ) and ( [ f1 ] )
, corresponding to an initial value problem consisting in a dirac delta located in @xmath88 at @xmath89 , is the following e(r , t ) = ( ) ^3/2 \ { - } f(r , t ) = e(r , t ) we imposed @xmath90 as inner boundary conditions and outflow boundary conditions in the outermost shell .
we have performed two simulations .
first , we set a value of @xmath91 on a equally spaced grid of 100 shells in the domain @xmath92 $ ] .
it corresponds to a peclet number ( @xmath93 ) of order unity .
the cell _ peclet number _ is a measure of the spatial resolution relative to the relaxation scale and a grid is coarse when @xmath94 .
this first simulation starts at @xmath95 to avoid the numerical problems produced by an infinite value in @xmath89 .
results are shown in figure [ fig1a ] . the energy density ( left panel ) and the flux ( right panel )
are displayed for different times in dimension - less units , @xmath96 from top to bottom .
the solid line corresponds to the analytical solution and the crosses correspond to the numerical solution .
notice that , although we are finding numerical solutions of system ( [ system ] ) , the analytical solution of the diffusion limit is well reproduced by the numerical code , since the opacity is large enough to be close to the diffusion limit . the second simulation has been performed in a grid of 100 shells in the domain @xmath97 $ ] , and using a value of @xmath98 , which corresponds to @xmath99 .
the initial time in this case is 200 and the courant time - step is of the same order ( @xmath100 ) as in the previous simulation , three orders of magnitude larger than the relaxation scale .
results are displayed in fig .
[ fig1b ] .
the agreement between the analytical and the numerical solutions clearly proofs the capability of the code to handle situations with very stiff source terms . in the radially streaming limit ( @xmath101 , @xmath102 ) , and assuming there is no interaction between radiation and matter ( @xmath103 , @xmath104 ) equations ( [ system ] ) are reduced to @xmath105 the solution to this equation is any linear combination of ingoing and outgoing spherical waves , _
i.e. _ , any linear combination of functions of the form @xmath106 .
for the first numerical experiment we choose a gaussian spherical wave propagating outwards , e(r , t ) = \ { - ( r - t)^2 } f(r , t ) = e(r , t ) .
our numerical grid consists on 200 equally spaced shells going from @xmath107 to @xmath108 and the boundary conditions consist in imposing the analytical solution in both boundaries .
we show the results of this test in figure [ fig2a ] . in the figure we displayed the analytical solution ( solid line ) and
the numerical ( crosses ) for the energy density . in order to analyse the effect of the particular choice for the closure
, we repeated the experiment using different values for @xmath109 , which leads to different speeds of propagation of the waves , and no difference was observed .
the final reason is that all the closures displayed in table 1 give @xmath110 .
this is a general property of any closure , since @xmath111 is a positive defined function that goes to zero as @xmath46 and , therefore , its derivative has to be negative in a vicinity of @xmath69 and @xmath112 . thus taking the limit @xmath46 in ( [ lambda ] ) ,
one obtains _ + = = 1 to understand why the value of @xmath113 is not important in this problem , we study the diagonalized system obtained by adding and subtracting the two equations in ( [ system ] ) after taking @xmath102 , @xmath114 in this equations it is clearly seen that the quantities propagated with velocities @xmath115 ( riemann invariants ) are @xmath116 , respectively . since in our particular problem we took @xmath117 ,
the quantity propagated by @xmath113 is identically null and the value of the speed of propagation does not affect the solution .
a different situation is found when we add to the initial conditions a second wave propagating in the opposite direction .
figure [ fig2b ] shows the results from such experiment , in which a second wave with smaller amplitude and opposite direction is located initially at @xmath118 .
the closure used was @xmath102 , @xmath119 , which gives @xmath120 , @xmath121 . in the figure ,
the analytical solution is denoted by the solid lines and the numerical solution by crosses and each panel corresponds to a different time .
it can be seen how the code is able to solve the interaction of the two waves until they completely cross over .
the difference between the analytical and numerical solution near the center for the last panel are due to the boundary conditions ( initially , it was an ingoing wave ) .
to illustrate the effect of a wrong speed of propagation , the experiment is repeated with kershaw s closure , that gives a correct speed of propagation when @xmath122 but a wrong value otherwise .
results are shown in figure [ fig2c ] for the same time steps as in the previous case . at the beginning
both waves seem to propagate correctly but , as they cross each other and the flux factor departs from unity , the numerical result differs from the analytical solution .
a similar behaviour is obtained with other closures .
this simple example manifests the importance of the choice for the closure , since it fixes the speeds of propagation of the waves . in the two previous tests we studied the limiting cases , diffusive and free - streaming regimes .
another academic problem in the limit of vanishing opacities consists in a sphere of radius @xmath123 radiating isotropically into vacuum ( figure [ vac ] ) . given the explicit form of the distribution function at the surface of the sphere , @xmath124
, the general solution can be easily obtained ( t , r , ) = \ { .
where s = r ( - ) x(r , t ) = \ { . where @xmath125 .
we take a linear time dependence @xmath126 , being @xmath127 an arbitrary constant , so that the angular integrals can be done analytically .
the angular moments of the distribution function are , then , given by e(t , r)= f(t , r)= p(t , r)= .
we set up a numerical grid of 200 points , uniformly distributed , from @xmath128 to @xmath129 .
the initial conditions are @xmath130 and @xmath131 with @xmath132 being a small number , different from zero for numerical purposes .
we have taken @xmath133 .
the radiation field is introduced through the inner boundary conditions , @xmath134 and @xmath135 .
outflow is assumed in the outer boundary . in figure [ fig3a ] the evolution of the radiation energy density as a function of the radial coordinate
is shown for three different evolution times .
the test has been repeat for three different closure relations from those described in table 1 : vacuum ( dashed lines ) , kershaw ( dotted lines ) and minerbo ( dashed - dotted lines ) . as in the previous figures the analytical solution is represented by the solid line .
although the appropriate closure for this problem ( vacuum ) gives a better approximation to the real solution , it is remarkable that other closures also give similar results .
the differences between different closures stem from the slightly different propagation speeds of the simple waves in the regime of interest . unlike in the two - wave problem described in previous section , for this specific problem any reasonable closure might be used without major deviations from the real solution . with this test we intend to simulate a more realistic scenario in which , initially , radiation produced in the center of a star diffuses out with constant luminosity .
we begin by setting up a sphere which is in radiative equilibrium at a given luminosity @xmath136 . for @xmath137 , the stationary solution
is given by e(r ) = f(r ) = where @xmath123 is the radius of the outer boundary , at which @xmath69 .
starting with the initial model corresponding to @xmath138 , we let it evolve in three different situations : 1 ) we take @xmath139 as boundary conditions at the inner boundary ( located at @xmath128 ) , and the radiation field should diffuse out continuously ; 2 ) we change the inner boundary condition to @xmath140 , ten times larger than the initial model .
the behaviour is now the opposite , the energy density must increase until the new stationary solution for the new luminosity is reached ; 3 ) we again take @xmath139 at the inner boundary but we switch on a central point source in the middle of the star ( @xmath141 ) .
the radiation field must evolve towards the corresponding stationary solution , which is different in the two parts of the star defined by the source .
inside , the luminosity is zero and the energy density is constant . in the region
external to the energy source , the stationary solution is defined by @xmath142 , where @xmath143 is the energy per unit time injected by the source .
the results for @xmath91 are shown in figures [ fig4a]-[fig4c ] , for cases 1 to 3 , respectively .
all three simulations have been done on the same grid of 200 uniform zones and outflow boundary conditions in the outer edge have been imposed .
left panels display the energy density as a function of radius .
the thick solid line in the figures corresponds to the initial model and the thick dotted line in figure [ fig4b ] represent the stationary solution with @xmath140 . in figure
[ fig4b ] , we can see how the heat wave lasts about 1000 time - units ( the diffusion time @xmath144 ) to arrive to the surface and a new stationary solution is successfully obtained after @xmath145 times the diffusion time scale .
testing a method in the semi - transparent regime becomes a tough issue due to a lack of analytical solutions to compare with . to overcome this problem ,
we have performed monte carlo simulations , which is the closest solution to an exact one , and results in an excellent test of the performance of the method in the semi - transparent regime .
we have used a monte carlo code that simulates the transfer of a radiation field in a spherically symmetric geometry .
we consider a region with inner boundary at @xmath146 and outer boundary at @xmath147 , divided in 200 spherical shells . within each shell
a constant value of the scattering opacity ( elastic and isotropic ) is assumed .
we do not consider emission or absorption processes .
our monte carlo procedure starts by injecting particles at the inner boundary outwardly - directed .
then , we compute the trajectory of each particle as it scatters off matter , according to the standard random walk laws @xcite .
the trajectory ends when the particle escapes by crossing the outer boundary or it re - enters the inner boundary .
once the path of a particle is computed we proceed to calculate the contribution of this particle to the moments of the radiation field .
the contribution of each particle to the energy density at a given numerical shell is calculated by adding the path length between two successive scattering events , within the shell , divided by the volume of the shell .
consequently , the total energy density is obtained just by adding the contribution of each particle multiplied by an arbitrary factor analogously , to obtain the contribution of each particle to the flux , the path length is weighted using the cosine of the angle with respect to the radial direction .
finally , pressure is calculated by means of the same procedure , but using the square of the cosine as weight of the path length . in our numerical experiment
we have taken : @xmath148 and @xmath149 .
the opacity law is an exponentially decreasing function in radius , taking the value @xmath150 at the inner boundary and @xmath151 at the outer boundary .
the corresponding values of the optical depth are such that an important part of the region is semi - transparent .
we have run the monte carlo code using @xmath152 particles in order to obtain the stationary solution .
the set up for our transport code is the following : i ) the above opacity law , ii ) the inner boundary conditions given by the stationary solution ( obtained from monte carlo ) , and iii ) the closure relation .
we have performed two different calculations using kershaw s closure relation and a cubic polynomial fit to monte carlo results , explicitly p = 0.3228 + 0.1902 f-0.0476 f^2 + 0.5131 f^3 starting with arbitrary initial profiles , we let the radiation field evolve until the stationary solution is reached and results can be compared to those obtained from the monte carlo simulations . in figure
[ fig6 ] we show the stationary solutions calculated from monte carlo simulations ( solid line ) and from our code with the two different closures , kershaw s ( dotted line ) and the above fit ( dashed line ) . from the results displayed in the figure we can conclude that the evolutionary code describes the semi - transparent regime accurately .
the differences in the values of the energy density ( left panel ) , normalized to its inner boundary value , amounts less than a few percentage in any case .
the slight discrepancies in the flux factor are mainly due to the closure relation employed , being less than 6% for kershaw s closure and less than 3% for the fit .
the closure , therefore , sets the maximum accuracy that can be obtained , although qualitative results at a level of 10% uncertainty can easily be obtained with any reasonable closure in the semi - transparent regime .
our last experiment intends to illustrate how the numerical method we proposed can succeed in situations when standard flux - limited diffusion techniques fail .
we set up a sphere of radius @xmath153 with a radiation energy density of @xmath154 , embedded in a background where the opacity varies as follows : = \ { . where we have taken @xmath155 . outside the sphere
we take @xmath156 ( in practice , @xmath157 for numerical reasons ) , and initially the flux is zero everywhere .
it tries to mimic a situation in which radiation scatters out in a central star , where the diffusion approximation remains valid , through an opacity - varying atmosphere up to the surface ( @xmath158 ) , where the opacity suddenly vanishes .
then , it crosses a transparent region to find a very opaque layer , where diffusion operates again .
this sort of problem is obviously a tough challenge for diffusion - based codes , since they can not work properly in the transparent region . in figure [ fig6 ]
we show results from the evolution of this problem on a grid of 200 uniform shells and using kershaw s closure . in the left panel
we show the energy density for different times .
it can be seen how a gradient is developed in the diffusive regions @xmath159 while radiative equilibrium @xmath160 in the transparent region between the two opaque layers is rapidly obtained .
notice also that only for the last snapshot ( @xmath161 , dashed curve ) there is a flux of energy in the outer transparent region , because it takes some time to the radiation to diffuse through the opaque layer and develop a gradient that keeps the outgoing flow .
we have studied the mathematical character of the angular moment equations of radiative transfer , considering the implications of using different closure relations .
the role played by the closure and their derivatives is more apparent in the hyperbolic formulation .
after this analysis , we conclude that the hyperbolic character is assured for a given variety of closures widely used in the literature : those given by @xmath59 . for general closures @xmath162 ,
hyperbolicity can not be guaranteed , although the closure proposed by cernohorsky & bludman , based on maximum entropy arguments , leads to real eigenvalues .
additional constraints on the closures are imposed on the base of the behaviour of the eigenvalues of the jacobian matrix since they give the velocity of propagation of perturbations .
causality limitation is written in terms of the closure relation helping us to select , from this point of view , among the different closures reviewed .
it turns out that only one of the closures analysed does not satisfy causality , allowing for velocities of the waves higher than the speed of light . for the other closures ,
although derived without taking into account the mathematical issues addressed in this paper , we found that they are consistent with hyperbolicity and causality . writing the moment equations of the distribution function as a hyperbolic system of conservation laws
allows one to apply numerical techniques specifically designed for such systems , the so - called hrsc methods .
we have applied hrsc methods to solve the transport equations in a static background showing , for a number of test problems , the feasibility of the method . from the results ,
the main conclusion is that the numerical method can be used in any regime , from optically thick to transparent regions , obtaining numerical solutions with high accuracy .
it turns out that the closure relation plays an important role , more evident than in traditional methods for radiation transfer , that can be useful to choose the more appropriate closure for a given problem .
when the radiation transport equations are written as a system of conservation laws , the coupling with hydrodynamics is straightforward @xcite .
this feature make us to be confident in the possibility of applying the method to problems involving radiative flows .
moreover , we are optimistic in the applicability to multidimensional problems , since the extension of the method is straightforward .
this work has been supported by the spanish dgicyt ( grant pb97 - 1432 ) . j.a.p . also acknowledges previous financial support from the ministerio de educacin y cultura .
groth , c. p. t. , gombosi , t. i. , roe , p. , l. , brown , s. l. , 1996 , proceedings of the _ fifth international conference on hyperbolic problems _ , eds .
glimm , j. , graham , m.j . ,
grove , j.w .
, world scientific . | we study the mathematical character of the angular moment equations of radiative transfer in spherical symmetry and conclude that the system is hyperbolic for general forms of the closure relation found in the literature .
hyperbolicity and causality preservation lead to mathematical conditions allowing to establish a useful characterization of the closure relations .
we apply numerical methods specifically designed to solve hyperbolic systems of conservation laws ( the so - called godunov - type methods ) , to calculate numerical solutions of the radiation transport equations in a static background .
the feasibility of the method in any kind of regime , from diffusion to free - streaming , is demonstrated by a number of numerical tests and the effect of the choice of the closure relation on the results is discussed .
radiative transfer methods : numerical |
impurity models , describing small quantum systems coupled to one or multiple baths of bosons or fermions , have seen a lot of activity over the last years , for a variety of reasons : ( i ) impurity models display a rich phenomenology , including local fermi - liquid and non - fermi - liquid behavior,@xcite phase transitions and quantum criticality,@xcite as well as interesting properties far from equilibrium.@xcite ( ii ) impurity models can often be simulated by numerical means more efficiently than lattice models,@xcite such that , on the one hand , high - accuracy numerical results can guide analytical approaches and , on the other hand , analytical concepts can be readily tested numerically .
a particularly interesting branch is non - equilibrium physics where quantum impurity models have served a test bed for methodological developments .
( iii ) impurity models find realizations in diverse settings such as dilute magnetic moments in bulk solids,@xcite electrons in quantum dots coupled to leads,@xcite quantum bits in a dissipative environment,@xcite and charge - transfer processes in organic molecules.@xcite the design of impurity models in cold - atom systems provides further means of manipulating and detecting impurity phenomena.@xcite the spin - boson model ( sbm1 in the following ) is a simple paradigmatic model for quantum dissipative systems.@xcite it describes a two - level system , i.e. , a spin 1/2 , which is coupled to both a bath of harmonic oscillators and a transverse field .
while the field induces tunneling ( i.e. delocalization ) between the two states , the oscillator bath causes friction and impedes tunneling . for gapless baths , characterized by a power - law spectral density @xmath0 with @xmath1 , this competition results in a quantum phase transition between a delocalized and a localized phase which has been studied extensively.@xcite as has been shown both analytically and numerically,@xcite this quantum phase transition obeys the so - called quantum - to - classical correspondence : it is equivalent to the thermal phase transition of a classical ising chain with long - ranged interactions falling off as @xmath2 where @xmath3 is the distance between two classical spins.@xcite in this paper we consider the generalization of the spin - boson model to two baths ( @xmath4 below),@xcite dubbed sbm2 .
it is described by @xmath5 with [ eq : h ] @xmath6 the two - level system ( or quantum spin , with @xmath7 being the vector of pauli matrices ) is coupled both to an external field @xmath8 and , via @xmath9 and @xmath10 , to two independent bosonic baths , whose spectral densities @xmath11 are assumed to be of the same power - law form , @xmath12 where @xmath13 defines the unit of energy used throughout the paper . for a symmetric coupling to identical bath , i.e.@xmath14 , and @xmath15 the model displays a u(1 ) symmetry , corresponding to a rotation of the impurity spin about its @xmath16 axis combined with corresponding bath - mode rotation .
in addition , the model features a separate z@xmath17 symmetry for @xmath18 , corresponding to @xmath19 .
the model sbm2 is governed not only by the competition between the local field , which tends to point the spin in the @xmath8 direction , and the dissipative bath effects , but also by a competition between the two baths , as an oscillator bath which couples to @xmath20 tends to localize the spin in @xmath21 direction . as a result
, the combined dissipative effect of both baths in sbm2 can be smaller than that of one bath alone ( in a sense which will become clear in the course of the paper ) an effect which has been dubbed `` frustration of decoherence''.@xcite in practical realizations of sbm2 , the two baths can be two different sources of dissipation influencing a quantum bit,@xcite or two spin - wave modes which couple to a magnetic impurity in a magnet.@xcite the model sbm2 is of particular theoretical interest because it displays a non - trivial intermediate - coupling ( i.e. critical ) phase , characterized by partial screening of the impurity degree of freedom corresponding to a fractional residual moment , not unlike in the two - channel kondo state.@xcite the existence of this critical phase , originally deduced by perturbative rg arguments,@xcite was recently confirmed numerically.@xcite the latter study , performed using a variational matrix - product - state ( vmps ) approach , also revealed that the critical phase is unstable at large couplings , resulting in a complex phase diagram .
it is the purpose of this paper to study the physics of sbm2 in some detail , extending the results published in ref . , with particular focus on the quantum phase transitions occurring in this model . to this end , we combine vmps calculations with analytical renormalization - group and scaling approaches . our implementation of vmps , including the use of the u(1 ) symmetry and an optimized boson basis , enables highly accurate studies of quantum critical behavior
. we have used vmps to determine quantitative phase diagrams for the u(1)-symmetric version of sbm2 as function of the bath exponent @xmath22 , the dissipation strength @xmath23 , and the transverse field @xmath24 . for @xmath25 and finite @xmath24
, there is always a transition between a delocalized ( de ) and a localized ( lo ) phase , fig .
[ fig : phasediag ] with the lo phase spontaneously breaking the model s u(1 ) symmetry .
there is no localization for @xmath26 ( not shown)@xcite this is qualitatively different from the behavior of the standard single - bath spin - boson model ( sbm1 ) and reflects the frustration of decoherence mentioned above . for @xmath18 the critical ( cr )
phase emerges , existing for @xmath27 and small @xmath23 .
( a ) and @xmath28 ( b ) . for a bath exponent @xmath29 in ( a )
, the sbm2 ground state is either in a delocalized ( de ) or localized ( lo ) phase depending on coupling strength @xmath23 and magnetic field @xmath24 ; the corresponding quantum phase transition is controlled by the critical fixed point qc2 . for larger @xmath30 in ( b ) an additional critical phase ( cr ) emerges at @xmath18 and small couplings .
the quantum phase transition between lo and cr is controlled by a different critical fixed point qc1 . ]
based on numerical and analytical results for the quantum critical behavior , we conclude that the transition between the de and lo phases , controlled by a fixed point labelled qc2 in the body of the paper , is in the universality class of the classical xy chain with @xmath2 interactions , i.e. , obeys a quantum - to - classical correspondence .
in particular , @xmath31 corresponds to the upper - critical dimension for this transition , with mean - field behavior found for @xmath32 .
in contrast , the transition between cr and lo , controlled by a different fixed point qc1 , does not appear to obey a quantum - to - classical correspondence .
its exponents fulfill hyperscaling relations for @xmath18 , but hyperscaling is violated in the presence of a transverse field .
we propose how to construct a critical field theory which should ultimately enable an analytical understanding of this conceptually interesting non - classical transition .
the body of the paper is organized as follows . in section
[ sec : method ] we introduce the employed vmps method .
in particular we discuss both the variational choice of bosonic basis states and the implementation of the u(1 ) symmetry into the algorithm .
section [ sec : phases ] describes the phase diagram of the u(1)-symmetric sbm2 , together with the main characteristics of the stable phases .
the subsequent section [ sec : flow ] analyzes the numerical findings in terms of renormalization - group flow and discusses the resulting quantum critical points .
section [ sec : rg ] is devoted to analytical approaches to the critical phenomena of sbm2 , using the toolbox of field theory and epsilon expansion .
in particular , we highlight that qc2 is expected to follow the quantum - to - classical correspondence while qc1 is not . in section [ sec :
numerical ] we show numerical results for critical properties of sbm2 .
we will extract numerous critical exponents as function of the bath exponent @xmath22 , confirming the analytical expectations .
the concluding section [ sec : concl ] will highlight open problems as well as connections to other impurity and lattice problems .
in addition , the physics of sbm2 with broken u(1 ) symmetry will be quickly discussed .
technical details are relegated to various appendices .
we start by describing the numerical vmps approach which we employed to study sbm2 .
this extends the corresponding presentation in ref . .
in particular the explicit implementation of the u(1 ) symmetry , which we found crucial to obtain accurate critical exponents , is a novel ingredient here .
since both bosonic baths of sbm2 are non - interacting and gapless , it is possible to transfer the concept of energy - scale separation frequently employed in numerical renormalization group ( nrg).@xcite to this end , the spectral functions of the baths are logarithmically discretized .
then the hamiltonian is mapped on a semi - infinite tight binding chain , a so - called wilson chain .
the choice of a logarithmic coarse graining of the spectral function @xmath33 is motivated by the fact that the study of critical behavior requires exponentially small energy scales .
to resolve these scales appropriately , a logarithmic coarse - graining is necessary , since it yields an exponentially enhanced low - energy resolution compared to a linear or power - law discretization . assuming the spectral function @xmath33 of each bosonic bath has a non - zero contribution for energies @xmath340,\omega_c]$ ] , with @xmath35 being an upper cut - off frequency
, we introduce a dimensionless discretization parameter @xmath36 which defines a set of intervals with discretization points,@xcite @xmath37 with @xmath38 an arbitrary shift . averaging over different @xmath16 uniformly distributed in @xmath39
is referred to as @xmath16-averaging*???*. considering a symmetric coupling of the impurity to two identical baths and using @xmath40 for simplicity , the discretized hamiltonian is represented by @xmath41\ , , { \quad \phantom{.}}\end{aligned}\ ] ] with @xmath42 being a discrete bosonic state at energy @xmath43 and coupling strength @xmath44 to the impurity spin . for general @xmath45 one has@xcite @xmath46 employing the improved @xmath16-averaging scheme of itko and pruschke to reduce discretization artifacts,@xcite the explicit expressions for the parameters for general @xmath16 are given by@xcite @xmath47^{\frac{1}{1+s } } \quad & ( m=0)\ , , \nonumber \\
\xi^z_m = & \big[\tfrac{\lambda^{-(s+1)(m+z ) } ( \lambda^{(1+s)}-1 ) } { ( 1+s ) \ln\lambda } \big]^{\frac{1}{1+s } } \sim \omega^z_m \quad & ( m>0)\ , , \nonumber \\\end{aligned}\ ] ] @xmath48 following the standard nrg protocol , the discretized hamiltonian in eq.([eq : hstar ] ) is mapped using an exact unitary transformation onto a semi - infinite tight - binding chain , dubbed wilson chain , with the impurity coupled to the open end only .
the resulting hamiltonian including @xmath49 bosonic sites is given by @xmath50 with [ eq : hchain ] @xmath51\ , , \nonumber \\
\label{eq : hchainbath}\end{aligned}\ ] ] with the operator @xmath52 counting the number of bosons of bath @xmath21 on chain site @xmath53 .
each bosonic site represents a harmonic oscillator at frequency @xmath54 that is coupled to its nearest neighbors by the hopping amplitude @xmath55 . assuming identical baths
, @xmath56 describes the overall coupling between a bath and impurity .
note that the impurity spin now couples to a single bosonic degree of freedom located at @xmath57 , i.e.the first site of a bosonic tight - binding chain ( see also fig .
[ fig : vmpsfig2 ] below ) .
their combined local hamiltonian is given by @xmath58 .
the steps remaining in the nrg procedure would involve an iterative diagonalization by adding one site at a time and a subsequent truncation of the high - energy states of the system , keeping only the @xmath59 lowest lying energy eigenstates .
however , the bosonic nature of the model complicates the nrg approach drastically . employing nrg , it is required to truncate the infinite dimensional local bosonic hilbert spaces on site @xmath53 to manageable number of @xmath60 states .
thus , a priori , nrg is not able to take into account the growing oscillator displacements @xmath61 occurring in the system s localized phase .
this restricts its application to the delocalized phase . already at the phase boundary , in combination with the inherent mass - flow error@xcite , this leads to non - mean - field results for the critical exponents of sbm1 in the regime @xmath32.@xcite to resolve the issue of bosonic state space truncation , guo et al.@xcite proposed a variational matrix - product - state ( vmps ) approach involving an optimized boson basis ( obb ) , that allows an accurate numerical study of the entire phase diagram in the ( generalized ) spin - boson model .
since we heavily used this method for the numerical results presented here for sbm2 , we briefly outline the concept of this powerful approach.@xcite the starting point of the variational procedure is setting up an initially random many - body state @xmath62 of the truncated wilson chain described by @xmath63 [ having @xmath49 sites in total ] in the language of matrix - product states ( mps):@xcite @xmath64@xmath65@xmath66@xmath67@xmath68 are the eigenstates of @xmath9 and the states @xmath69 represent the boson - number eigenstates of the truncated fock space , i.e. @xmath70 with @xmath71 .
combining the state spaces of both chains in eq . and to supersites , @xmath72 should be interpreted as a combined index of the x- and y - chain .
each @xmath73}$ ] forms a @xmath74 matrix with elements @xmath75})_{\alpha \beta}$ ] , except for @xmath76}$ ] and @xmath77}$ ] connecting to local impurity and vacuum states respectively , as indicated in fig .
[ fig : vmpsfig2 ] . using standard mps methods ,
we optimize @xmath62 by iteratively varying one @xmath73}$ ] at a time in order to find an appropriate representation of the ground state of @xmath63 . in eq .
( [ eq : mps1 ] ) using the obb representation with explicit bosonic shift . the first index of @xmath76}$ ] and the last index of @xmath78}$ ] link to the impurity and the right - vacuum state @xmath79 , respectively .
, scaledwidth=45.0% ] the main advantage of vmps is the possibility to change the local basis during the step - wise optimization process , while nrg in principle requires the local basis to be fixed a priori before starting to diagonalize . to exploit this
, we implement the obb using two key features : + 1 . _ effective local basis _ :
a basis transformation @xmath80 is introduced with @xmath81 , which maps the local harmonic oscillator basis @xmath82 onto a smaller effective basis @xmath83 on each site @xmath53 , @xmath84 with @xmath60 and @xmath85 denoting the size of the original and effective basis , respectively . merging @xmath80 into the @xmath86-tensors on each bosonic site , the structure of @xmath73}$ ] in eq .
( [ eq : mps1 ] ) is then given by @xmath87}_{\alpha,\beta } = \sum_{\tilde{n}_k=0}^{\tilde{d}_{k}-1 } \tilde{a}^{[\tilde{n}_k]}_{\alpha,\beta } v_{\tilde{n}_k , n_k}\ , .
\label{eq : mps_obb}\ ] ] nevertheless , from an efficiency point of view , it is desirable to keep the separate structure of @xmath88 and @xmath80 , where @xmath89}$ ] links the effective bosonic basis to the left and right part of the chain , while @xmath80 maps the original to the effective local basis .
the local optimization procedure for each site thus splits into two steps : at first , @xmath80 is updated and in this process the optimal effective local basis set @xmath83 is determined .
then we optimize @xmath89}$ ] using the new local basis states and move to the next site . + note that with the introduction of the obb a second adjustable dimension @xmath90 besides the bond dimension @xmath59 exists . treating @xmath88 and @xmath80 as separate structures ,
both dimensions are fixed before the start of the ground - state optimization .
if a dynamical adjustment of the bond dimensions is required , one has to switch to a two - side optimization procedure or variants of these , which is numerically more expensive.@xcite this is for example necessary when enforcing explicit symmetry conservation . in practice
, this implementation makes an increase of the size of the local basis sets from @xmath91 to @xmath92 possible , while using typically @xmath93 below .
oscillator shifts _ :
moreover , in the localized phase we incorporate an oscillator shift in the hamiltonian to take the oscillator displacement into account .
the oscillator coordinates @xmath94 are shifted by their equilibrium value @xmath95@xcite to be determined self - consistently in a variational setting , such that obb captures the quantum fluctuations near the shifted coordinate @xmath96 .
this is achieved by formulating the shift @xmath97 as unitary transformation acting on the hamiltonian itself .
with @xmath98 the shifted local bosonic operators @xmath99 and @xmath100 are @xmath101 by the application of @xmath102 we automatically shift @xmath103 by @xmath97 , @xmath104 after processing the local optimization procedure , we calculate the mean displacement @xmath95 . by setting @xmath105 and replacing @xmath106 with the displaced @xmath107 , the shift is included exactly on the hamiltonian level , @xmath108 .
afterwards , the optimization of the current site is repeated in the shifted local bosonic basis until @xmath95 converges , before moving to the next site .
+ the implementation of an obb with shifted oscillator modes allows us to simulate an effective local basis that would require a local dimension of @xmath109 in the non - shifted basis , while the actual shifted basis can be kept small , @xmath110 .
in addition , since the variational procedure determines the optimal shift @xmath97 for each site of the wilson chain individually , the exponential growth of @xmath111 with increasing iteration number @xmath53 no longer represents a barrier for the method .
working in the wilson chain setup with an exponentially decreasing energy scale , it is advantageous to replicate the nrg rescaling procedure in the iterative vmps procedure in order to avoid losing numerical accuracy towards higher iterations . therefore , when optimizing @xmath73}$ ] , we rescale the hamiltonian in the local picture by a factor @xmath112 to ensure that optimization can take place on the effective energy scale @xmath113 . employing standard vmps methods
, we determine the convergence of @xmath62 by calculating the variance of the ( unscaled ) energy @xmath114 of the ground state calculated at each site @xmath53 .
the iterative optimization procedure is stopped , once @xmath115 , using double precision accuracy @xmath116 with @xmath117 and thus @xmath118 .
the resulting state @xmath62 is considered to be a reliable approximation of the system s ground state given @xmath63 . when computing systems where the effective energy resolution drops below double precision , the relevance of numerical noise as a perturbation to @xmath63 should be double - checked by additionally studying the energy - flow diagrams .
most results shown in this paper have been obtained using parameters @xmath119 , @xmath120 , @xmath121 , unless noted otherwise .
considering the case with no in - plane magnetic fields , @xmath15 , the system exhibits an abelian u(1 ) symmetry : the hamiltonian is invariant under simultaneous rotation of the impurity spin and the bosonic baths in the @xmath122-plane by an arbitrary angle @xmath123 , leading to a two - fold degeneracy of the resulting ground state .
a rotation of this type is described by a unitary operator @xmath124 , @xmath125 where @xmath126 is the generator of the continuous u(1 ) symmetry , given by @xmath127 with @xmath128 = 0 $ ] . in the form of eq .
( [ eq : u1_generator ] ) , however , the symmetry operation @xmath126 involves a hopping between the two baths in the local bosonic state spaces , which poses a serious impediment for the numerical implementation of the symmetry due to truncation of the bosonic state space .
essentially , the discrete quantum number associated with the symmetry requires a diagonal representation .
hence , it is useful to apply a canonical transformation in order to bring @xmath126 in a diagonal form in the spinor space of @xmath129 .
this leads to @xmath130 note that this transformation also alters the coupling term in the hamiltonian .
in this form , the symmetry sectors are characterized by the @xmath16-component of the impurity spin and the difference in the bosonic occupation number in both baths in contrast to the hopping term of eq .
( [ eq : u1_generator ] ) , allowing an exact symmetry implementation in the vmps procedure in the presence of a truncated bosonic state space.@xcite given a simultaneous eigenstate @xmath131 of @xmath126 and @xmath132 , the application of the generator results in @xmath133 where @xmath134 is the total number of bosons occupying the wilson chain of the individual baths and @xmath135 is the spin component in @xmath16-direction . given any ground state @xmath136
, it follows that one may obtain another ground state via @xmath137 .
noting that the ground state comes with a symmetric distribution of boson numbers @xmath138 , we conclude that @xmath139 should be chosen to be @xmath140 , @xmath141 where @xmath142 is the ground - state energy .
hence the ground state is doubly degenerated .
the expectation value @xmath143 evaluated using the symmetry ground states @xmath144 is zero by symmetry . how to reconstruct the magnetization of the symmetry - broken ground state , which is a linear superposition within @xmath145 , is described in appendix[app2 ] .
it turns out that the u(1 ) symmetry implementation can not be combined with the shifted obb . employing a continuous shift @xmath146 to the bosonic creation and annihilation operators via eqs .
( [ eq : b ] ) leads to additional terms of the form @xmath147 in the symmetry generator .
these linear corrections add non - diagonal elements to @xmath148 precluding an explicit implementation of the u(1 ) symmetry in the way indicated above .
this limits the application of symmetry - enforced vmps effectively to the parameter regime @xmath149 , in which the bosonic state space truncation error does not spoil the calculations of physical quantities such as critical exponents ( see appendix[app1 ] for more details ) . when vmps is applied to a wilson - chain hamiltonian such as eq .
( [ eq : hchain ] ) , it is possible to generate an energy - level flow diagram akin to the ones of nrg . to this end
, we calculate the eigenvalues @xmath150 of the left block hamiltonian @xmath151 in each iteration @xmath152 when sweeping from the left to the right end of the wilson chain truncated to @xmath153 sites
. multiplied with the proper rescaling factor @xmath154 , the spectrum @xmath155 relative to the ground - state energy @xmath156 corresponds to the rescaled eigenspectrum determined in a nrg step .
the energy flow of excited states is not as smooth as using nrg , since our variational procedure focuses on optimizing the global ground state of the system only .
however , it can be systematically improved by incorporating symmetries of the model and keeping more states .
energy flow diagrams contain information about the fixed points of the impurity model , as illustrated in fig .
[ fig : vmps_flow ] for sbm2 , where the upper panels ( a , b ) are generated by enforcing the u(1 ) symmetry while for the center panels
( c , d ) a shifted obb is employed in the vmps procedure .
the flow towards a localized fixed point with a two - fold generated ground state is depicted in the left panels of fig . [
fig : vmps_flow ] . only the usage of obb accounts for the exponential growth of bosonic occupation numbers in the localized phase [ cf .
[ fig : vmps_flow](e ) ] .
the energy flow in ( c ) is distorted when introducing the bosonic shift on the wilson chain , since energy - scale separation is effectively broken due to the exponential growth in local bosonic occupation .
the ground - state degeneracy is conserved , however , when enforcing the symmetry in the vmps optimization ( a ) . in case
the system moves towards a delocalized fixed point with a single ground state at the end of the wilson chain , both methods generate flow diagrams of similar quality [ cf .
[ fig : vmps_flow](b),(d ) ] since no bosonic shift is necessary to appropriately describe the system s ground state .
hence , energy scale separation remains intact in this case . in the particular example of figs .
[ fig : vmps_flow](b ) and ( d ) , the intermediate fixed point visible at earlier iterations corresponds to the critical fixed point qc2 discussed below .
in addition to determining the system s phase or the convergence of the numerical data , flow diagrams can be used to extract information about the effective energy scales characterizing the crossover between fixed points .
for example , the transition from the critical to the de fixed point is governed by the low - energy scale @xmath157 , with @xmath158 for the parameters used in fig .
[ fig : vmps_flow](b ) and ( d ) . in two phases for different values of @xmath23 and @xmath24 .
@xmath53 parameterizes the running energy scale according to @xmath159 . while in the two upper panels the flow
is generated using the symmetry - enforced vmps , the center panels show diagrams generated by employing the shifted obb .
the energy levels flow to a localized fixed point in ( a),(c ) and to a delocalized fixed point in ( b),(d ) with degenerate ( non - degenerate ) ground - state space , respectively .
the degeneracy of each state is indicated by the numbers to the right side of each curve .
the colors in ( a ) and ( b ) decode the symmetry label @xmath139 of each energy level ( black and red for @xmath160 , green and purple for @xmath161 , and blue for @xmath162 ; matching colors are used in panels ( c ) and ( d ) ) . panels ( e ) and
( f ) display the corresponding occupation numbers @xmath163 .
, scaledwidth=50.0% ]
in this section , we describe the phase diagram of the u(1)-symmetric sbm2 , together with the main characteristics of the stable phases .
the most important observables for sbm2 employed in this study are the static magnetization , @xmath164 and the corresponding susceptibility @xmath165 in the case of u(1 ) symmetry , we distinguish @xmath166 and @xmath167 as well as @xmath168 and @xmath169 .
we will also monitor the occupation numbers of the bath modes of the discretized wilson chain , @xmath170 with @xmath4 .
we start with an overview on the stable phases numerically found for sbm2 .
the description is augmented by an assignment of the corresponding rg fixed points ( which are trivial with the exception of the critical phase ) , with their locations specified in terms of renormalized values of the coupling constants @xmath23 and @xmath24 .
an asymptotically free spin is controlled by the free - spin ( f ) fixed point , corresponding to vanishing dissipation , @xmath171 , and @xmath18 .
the ground state is doubly degenerate , and the susceptibility follows @xmath172 for all field directions .
is driven to zero by increasing @xmath24 past the critical value @xmath173 , indicated by the dashed line ( a ) .
correspondingly , the slight kink in the transverse - field response of @xmath169 at the lo de transition in ( b ) indicates the expected higher - order singularity .
we note that the numerics tends to spontaneously favor ordered states with @xmath174 , as these are the least entangled states . ] for large dissipation , the system enters a phase with spontaneously broken u(1 ) symmetry , controlled by the localized ( lo ) fixed point .
lo is located at @xmath175 and @xmath18 .
the bath - oscillator displacements are strongly coupled to the impurity spin , which develops a @xmath176 expectation value in an arbitrary fixed direction in the @xmath122-plane , fig .
[ fig : phasebound](a ) .
this phase is stable for finite ( small ) transverse field @xmath24 in which case the expectation values of the impurity describe a canted spin , fig .
[ fig : phasebound](b ) .
since the symmetry - broken phase exists at @xmath176 only , its associated finite-@xmath177 susceptibility is expected to be curie - like , albeit with a classical prefactor,@xcite @xmath178 . for dominant transverse field ,
the impurity spin is polarized along the z - axis and asymptotically decoupled from the bath .
this situation is controlled by the delocalized ( de ) fixed point , located at @xmath179 and @xmath171 .
the ground state is unique , the in - plane magnetizations @xmath180 and @xmath181 vanish , fig . [ fig : phasebound](a ) , and all susceptibilities are finite . and ( b ) response to finite @xmath182 at two points in cr and lo phase .
the small but finite magnetization in the cr phase in panel ( a ) is caused by finite - size effects as discussed in appendix [ app : finite ] . ]
the non - trivial feature of sbm2 is the existence of a stable critical _
phase_. this is reached for non - zero ( but not too large ) dissipation strength @xmath23 and @xmath18 in a certain range of bath exponents @xmath22 .
it is controlled by an intermediate - coupling fixed point , not unlike the celebrated two - channel kondo fixed point.@xcite in this phase , the expectation value of the impurity moment vanishes , but its temporal correlations decay with a fractional power law .
this translates into non - linear response functions with fractional exponents , as shown in fig .
[ fig : crmag](b ) .
in contrast to assumptions based on early rg work@xcite see also section [ sec : rgf ] below the critical phase is _ not _
stable for all dissipation strengths @xmath23 , fig .
[ fig : crmag](a ) , and does not even exist for bath exponents @xmath183 , with a critical value @xmath184 .
we note that the critical nature of the cr phase implies significant finite - size effects for the magnetization , as discussed in appendix [ app : finite ] . in order to study the critical phenomena of sbm2
, it is necessary to accurately determine the phase boundaries , i.e. , to numerically calculate the critical coupling @xmath185 and the critical transverse field @xmath173 , which define the location of the lo cr and lo de transitions . in our experience ,
the most accurate and efficient way to calculate @xmath185 and @xmath173 is to distinguish the phases by the characteristic behavior of the bosonic occupation numbers @xmath186 on the wilson chain .
the average occupation of boson modes increases towards the end of the wilson chain in the localized phase , while it decreases in both critical and delocalized phases .
moreover , right at the phase boundary ( i.e. at criticality ) the occupation numbers stay almost constant throughout the chain , except for a sharp decay at the end due to choosing a finite @xmath153 for the wilson chain .
this characteristic behavior , illustrated in fig .
[ fig : searchpt ] , can be used to determine the phase boundaries with high accuracy .
we have thus adopted this approach throughout to determine the precise values of @xmath185 and @xmath173 involved in the results described in section [ sec : numerical ] .
the accessible accuracy depends on the length @xmath153 of the wilson chain .
specifically the calculation of @xmath185 or @xmath24 up to @xmath187 decimals requires a minimal chain length@xcite @xmath188 where @xmath189 is the correlation - length exponent .
thus for regions in the phase diagram where @xmath189 becomes larger we have to increase the length of the wilson chain , making calculations numerically more demanding .
in this section , we use the insights gained in section [ sec : phases ] to deduce the qualitative rg flow of sbm2 .
the discussion will primarily be made in the language and coupling constants of the original hamiltonian .
a more complete discussion of rg beta functions is given in section [ sec : rg ] .
we start by rephrasing our numerical findings in rg language while referring to the qualitative rg flow diagrams in fig .
[ fig : rgflow ] . for @xmath18 the model sbm2 displays three phases : f , cr , and lo . for @xmath190 the free - spin phase f is the only stable phase , i.e. , the coupling to the dissipative bath does not qualitatively change the free - spin behavior .
this can be contrasted with the physics of sbm1 , where large dissipation causes localization in the ohmic case @xmath26 this distinction reflects the frustration of decoherence in sbm2 . for @xmath191 ,
f is unstable against any finite @xmath23 , whereas the localized phase lo is stable for sufficiently large @xmath23 .
finally , the critical phase cr only exists for @xmath192 and small values of @xmath23 .
a transverse field @xmath193 destabilizes f for any @xmath22 and drives the system into the de phase .
cr is unstable against any finite @xmath24 as well .
in contrast , lo is stable and hence requires a critical @xmath24 to be destroyed .
this collection allows us to construct the qualitative rg flow diagrams for the ranges of bath exponents @xmath194 , @xmath27 , and @xmath190 , as shown in fig .
[ fig : rgflow ] .
we also note that the system is always localized for @xmath195 provided that @xmath196 .
in addition to the cr fixed point corresponding to the critical phase , there are two further critical fixed points , qc1 and qc2 , which control the quantum phase transitions of sbm2 .
these are described in more detail in the next subsection .
for @xmath18 there are two fixed points at intermediate coupling , namely cr and qc1 , with qc1 controlling the transition between cr and lo .
both intermediate - coupling fixed points are unstable w.r.t .
finite @xmath24 .
both fixed points only exist for @xmath192 , and it is interesting to discuss their location upon variation of the bath exponent @xmath22 : as will be shown analytically in section [ sec : rg ] below , cr moves towards f as @xmath197 whereas qc1 moves towards lo in the same limit , in the fashion characteristic of a lower critical dimension .
in contrast , as @xmath198 , both cr and qc1 approach each other , merging at @xmath199 , and disappear for @xmath183 .
this merging of two intermediate - coupling fixed points leads to rather unusual behavior , with the phase boundary of lo jumping upon variation of @xmath22 across @xmath200 . for finite @xmath24
a transition can be driven between de and lo , which is controlled by qc2 .
qc2 moves towards lo as @xmath201 , again in a manner of a lower critical dimension .
this is consistent with the fact that the localized phase ceases to exist for @xmath202 . in the limit @xmath203 ,
qc2 approaches de , such that de becomes unstable w.r.t .
finite @xmath23 for @xmath204 , reflecting that the system is always localized .
the quantum phase transitions of sbm2 can be characterized by standard critical exponents.@xcite for a transition which can be accessed by varying @xmath23 ( at fixed @xmath24 ) , with the transition point at @xmath205 , the following exponents can be defined from the zero - temperature order parameter @xmath206 and its conjugate field @xmath207 : @xmath208 transitions which occur at finite @xmath24 can also be driven by varying @xmath24 at fixed @xmath23 ; correspondingly , the exponent @xmath209 may be defined via @xmath210 as well .
in contrast , for @xmath18 transitions , @xmath24 takes a role different from @xmath211 , as it reduces the symmetry of the model from u(1)@xmath212z@xmath17 to u(1 ) .
it is useful to introduce an exponent for the non - linear response to @xmath24 according to @xmath213 a correlation - length exponent is defined as usual from the divergence of a correlation length , here equivalent to the vanishing of a crossover energy @xmath214 according to @xmath215 note that there is no separate dynamical exponent for the @xmath216-dimensional impurity model under consideration , formally @xmath217 . for fixed points located at @xmath18 which are unstable towards finite @xmath24
we additionally define @xmath218 the linear - response order - parameter susceptibility diverges at the quantum critical point , in the approach from either smaller @xmath23 or from finite @xmath177 , according to @xmath219 within the quantum - to - classical correspondence , @xmath220 is related to the finite - size scaling of the classical model s susceptibility at criticality .
one may also consider the dynamic version of the order - parameter susceptibility , which follows a power - law behavior at criticality , @xmath221 corresponding to power - law autocorrelations of the impurity spin in time .
the exponent @xmath222 contains the same information as the usually - defined anomalous exponent @xmath223 , with @xmath224 . at the critical points of sbm2 ( and other spin models with long - ranged interactions ) ,
@xmath225 ( equivalently , @xmath226 ) is believed to be an exact relation , see also section [ sec : rg ] . due to the anisotropic nature of the spin fluctuations ,
different power laws arise for the @xmath16-component susceptibility : @xmath227 finally , it is also useful to introduce exponents which describe the location of the de lo phase boundary at small @xmath24 . for @xmath194
this phase boundary is connected to the @xmath228 point , and we define @xmath229 in contrast , for @xmath27 the de lo boundary terminates at the cr lo transition located at @xmath205 and @xmath18 , and we use @xmath230 the exponents introduced above can be related to each other via scaling relations , following textbook strategy.@xcite the standard scaling relations do hold , @xmath231 the exact result @xmath226 then implies @xmath232 for critical points with hyperscaling , additional scaling relations apply , which involve spatial dimensionality @xmath233 : @xmath234 furthermore , hyperscaling implies @xmath235 .
for @xmath236 and using the exact result @xmath226 , the hyperscaling relations can be converted into @xmath237 for critical points of sbm2 with @xmath18 , the scaling hypothesis underlying hyperscaling can be extended to include the dependence on @xmath24 [ in addition to that on @xmath211 , @xmath207 , and @xmath177 ] .
this then yields additional hyperscaling relations : @xmath239 and @xmath240 see appendix [ app : hyper ] for a derivation .
we recall that hyperscaling , which is of general interest because it implies simple and powerful scaling relations which can be applied in analyzing both experimental and numerical data , usually holds for phase transitions below their upper critical dimension .
hyperscaling is spoiled by the existence of dangerously irrelevant variables in the critical theory ; the most important example here is the quartic coupling of a ( classical ) @xmath241 theory in dimensions @xmath242 .
we now describe analytical approaches to the critical - point properties of sbm2 , utilizing the field - theoretic toolbox with renormalization - group and epsilon - expansion techniques .
the free - impurity fixed point f is characterized by a doubly - degenerate impurity at @xmath171 , @xmath18 . tree - level power counting yields the scaling dimensions ( recall that @xmath243 ) : @xmath244 & = 1-s\ , , \\ { \rm dim}[h_z ] & = 1\,.\end{aligned}\ ] ] from the scaling dimensions one can immediately read off the asymptotic behavior of the flow trajectories leaving the f fixed point , @xmath245 .
this also applies to the de lo separatrix in the exponent range @xmath194 , yielding the phase - boundary exponent according to eq . as @xmath246
now we turn to a rg analysis of the flow of @xmath23 at @xmath18 . given that the dissipation is a marginal perturbation at @xmath26 , this is akin to a standard epsilon expansion with @xmath247 , which can give reliable results for small @xmath248 .
straightforward perturbation theory , along the lines of refs . , yields the two - loop beta function:@xcite @xmath249 this beta function indicates the existence of an infrared - stable fixed point at @xmath250 and @xmath18 this is the cr fixed point .
its properties can be obtained in a double expansion in @xmath23 and @xmath248 .
the exact result @xmath251 follows from the diagrammatic structure of the susceptibility@xcite or , alternatively , from a ward identity.@xcite from this we have : @xmath252 as above .
the @xmath16 component correlator requires an explicit computation , with the two - loop result:@xcite @xmath253 the remaining exponents involving the @xmath24 response can be calculated from the hyperscaling relations and , with the result : @xmath254 we point out that the rg flow towards the cr fixed point is rather slow , because the leading irrelevant operator , its prefactor being @xmath255 , has a small scaling dimension of @xmath256 .
therefore , quickly converging numerical results are best obtained using a bare coupling close to @xmath257.@xcite it is interesting to note that the beta function in eq .
displays _ two _ non - trivial fixed points at @xmath258 . while @xmath259 corresponds to the stable cr fixed point of eq .
, the infrared - unstable fixed point at @xmath260 is outside the range of validity of the epsilon expansion .
however , _ if _ we choose to ignore this restriction , the comparison with the numerical results suggests to associate @xmath260 with qc1 .
remarkably , @xmath260 and @xmath259 approach each other upon decreasing @xmath22 from unity , and the criterion @xmath261 yields @xmath262 which is extremely close to the numerical determined value of @xmath263 where cr and qc1 merge . while this can be interpreted as a remarkable success of the epsilon expansion
it predicts not only the existence of the cr phase , but also its disappearance for @xmath183 we note that this epsilon expansion does not provide means to reliably calculate critical properties of qc1 , simply because @xmath260 is never small . as we show below , the presence ( absence ) of hyperscaling in a field at cr ( qc1 ) even indicates a qualitative difference between cr and qc1 which is not apparent from this epsilon expansion .
it is also possible to devise an expansion around the delocalized fixed point de , located at @xmath179 , @xmath171 .
such an expansion has been first used in ref . for sbm1 ,
but the analysis there missed the presence of a dangerously irrelevant operator ( the quartic coupling @xmath264 in eq .
below ) and erroneously assumed hyperscaling , which led to partially incorrect conclusions.@xcite here we correct this approach and apply it to sbm2 . for convenience we assume equal couplings between the impurity and the different oscillator modes , @xmath265 , such that the energy dependence of @xmath266 is contained in the density of states of the oscillator modes @xmath267 , and we have @xmath268 . at de we have a single low - lying impurity level , @xmath269 , while @xmath270 is separated by an energy @xmath24 .
low - energy interaction processes between the impurity and the baths arise in second - order perturbation theory , controlled by the coupling @xmath271 in the low - energy sector this corresponds to projecting out the @xmath270 state the effective theory reads ( assuming from here on @xmath272 or @xmath273 ) : @xmath274 with @xmath275 and @xmath276 .
we have defined @xmath277 and we have omitted higher - order terms in eq . . fig . [
fig : diag1 ] illustrates how the @xmath278 and @xmath264 terms are generated from @xmath132 of the original model sbm2 ; this approach is valid provided that @xmath279 . and @xmath280 impurity states the two states are separated by a gap @xmath24 .
the wiggly line is the local bath boson @xmath281 .
a ) interaction vertex @xmath282 .
b ) bilinear @xmath123 term .
c ) quartic @xmath123 term . ]
the theory @xmath283 can be understood as a theory for the local bosonic fields @xmath281 .
their `` bare '' propagator arises from @xmath284 and is given by @xmath285 at low energies , with @xmath286 for the power - law spectrum in eq . .
the main role of the impurity in @xmath283 is that of an additional mass term ( recall that the impurity spin degree of freedom has been projected out ) .
importantly , @xmath283 in eq . is identical to a local @xmath241 theory for an xy - symmetric order - parameter field @xmath287 with long - ranged interactions @xmath288 in imaginary time .
it displays a critical point which corresponds to a vanishing @xmath123 mass . ignoring the influence of the quartic interaction @xmath264
, this happens at @xmath289 ; this can alternatively be understood by interpreting @xmath278 as the strength of a potential scatterer , where @xmath290 is the condition for a zero - energy pole of the t matrix . for positive mass ,
i.e. , small @xmath291 , @xmath283 is in a disordered phase corresponding to de , whereas negative mass drives the system into an ordered phase with spontaneously broken xy symmetry this can be identified with lo .
consequently , the critical point of @xmath283 corresponds to qc2 . as the @xmath241 theory in question is the low - energy theory of a classical xy chain with long - range interactions ,
we conclude that qc2 obeys a quantum - to - classical correspondence at least if qc2 is located in the small-@xmath291 parameter regime where the above mapping to @xmath283 is valid , i.e. , for small @xmath22 .
the critical properties for the classical xy chain are listed in section [ sec : xyexp ] below ; in particular , mean - field behavior obtains for @xmath32 . an alternative approach to @xmath283 is to analyze the flow of the couplings @xmath278 and @xmath264 near the de fixed point by rg means .
power counting w.r.t .
the @xmath292 limit gives @xmath293 & = -s\,,\\ { \rm dim}[u ] & = 2s-1\,,\end{aligned}\ ] ] i.e. , @xmath278 is marginal at @xmath294 while @xmath264 is irrelevant . near @xmath294
we can follow the flow of @xmath278 which yields at one - loop order : @xmath295 besides the stable de fixed point at @xmath296 ( i.e. @xmath171 ) this flow equation displays an infrared unstable fixed point ( qc2 ) at @xmath297 which controls the transition between the de and lo phases .
corrections from @xmath264 only enter at higher orders in @xmath22 , because the initial values obey @xmath298 .
we note that the value of @xmath299 in eq .
is consistent with @xmath300 from above .
expanding the rg beta function around the fixed point ( [ fp ] ) gives the correlation - length exponent @xmath301 apparently in agreement with the classical mean - field result .
one may employ renormalized perturbation theory to calculate critical exponents in a double expansion in @xmath278 and @xmath22 .
this is , however , complicated by the facts that ( i ) for many observables of the original model sbm2 one needs to restore the impurity hilbert space , i.e. , undo the elimination of the @xmath270 state , and ( ii ) the quartic coupling @xmath264 is dangerously irrelevant and can not be neglected . for selected exponents ,
we have checked that this procedure yields results consistent with eqs . - .
one may ask whether a general mapping of sbm2 to a classical statistical - mechanics model exists .
such a mapping , using a feynman path integral representation , can indeed be formulated for the single - bath spin - boson model ( sbm1 ) and directly leads to an ising chain with both short - ranged and long - ranged @xmath2 interactions.@xcite here we sketch what happens when applying the same procedure to sbm2 . for simplicity ,
we restrict ourselves to @xmath302 .
the hamiltonian may be written as @xmath303 with @xmath304 the feyman path integral for the partition function can be expressed using eigenstates of @xmath9 , @xmath10 , and the oscillators coordinates .
inserting the identities for the spin variables those for the oscillator coordinates are standard and do not lead to any complications
it reads : @xmath305 where @xmath153 is the number of trotter slices , @xmath306 , and @xmath307 . in principle , a classical spin model can now be obtained by integrating out the bath oscillators , which generates long - ranged interactions for the variables coupled to these oscillators . in the case of sbm2 , these are both @xmath9 and @xmath10 , such that one ends up with a representation in terms of sets of ising spins . to rewrite this in terms of a classical spin model requires to express the matrix elements in eq . as exponentials of classical interactions
remarkably , the set of matrix elements @xmath308 can not be expressed as @xmath309 with a classical real hamiltonian function @xmath310 , i.e. , the feynman path - integral representation of sbm2 leads to an ill - defined classical model with negative boltzmann weights.@xcite clearly , this problem can be traced back to the non - commutativity of the two spin components which couple to the oscillator baths in sbm2 .
we recall , however , that the physics of the qc2 fixed point of sbm2 _ can _ be mapped onto that of a classical xy model at least near @xmath294 , see section [ sec : rgd ] above .
assuming that the character of qc2 does not change fundamentally as function of @xmath22 , this implies that a quantum - to - classical correspondence indeed holds for qc2 . as will be shown in section [ sec : numerical ] below , our numerical results for the critical behavior near qc2 are perfectly consistent with this assertion .
here we collect and summarize the available results for critical exponents of the classical xy chain with long - range interactions decaying as @xmath2 these have been discussed in refs . .
the classical model has a thermal phase transition for @xmath25 ; no ordered phase exists for @xmath190 . for @xmath32 and @xmath311 one may utilize the language of a @xmath241 theory .
power counting shows that the quartic interaction is marginal for @xmath31 , such that mean - field behavior attains for @xmath32 , with @xmath312 with hyperscaling being violated . in the non - mean - field regime , @xmath313
, one can obtain exponents in an expansion in @xmath314 , with two - loop results as quoted in ref .
: @xmath315 the latter result is believed to be exact to all orders,@xcite and the constant @xmath316\ ] ] in terms of the digamma function @xmath317 , with @xmath318 .
hyperscaling holds for @xmath319 ; this allows to derive the remaining exponents : the correlation - length exponent @xmath189 follows from the scaling relation @xmath320 , while @xmath209 can be read off from the hyperscaling relation @xmath321 , with the results @xmath322 near @xmath26 rg equations can be derived@xcite using a variant of a method proposed by polyakov @xcite this is similar to an ordered - phase expansion in @xmath323 dimensions for magnets with short - range interactions .
exponents are formally obtained in an expansion in @xmath248 ; the one - loop results read:@xcite @xmath324 the latter result again believed to be exact to all orders . using hyperscaling , we obtain the one - loop result for @xmath209 as @xmath325
taking into account the insights gained in the preceding sections [ sec : flow ] and [ sec : flow ] , we now focus on the numerical results obtained for the critical behavior of sbm2 . to this end
, we employ the vmps methodology as introduced in section [ sec : method ] at the quantum phase transitions qc1 and qc2 , as well as in the cr phase to extract various critical exponents .
our main results are that ( i ) the transition between lo and de , controlled by qc2 , indeed obeys quantum - to - classical correspondence , i.e. , its critical properties are that of a classical xy chain with long - range interactions , and ( ii ) the transition between cr and lo , controlled by qc1 , is of unusual nature , with no quantum - to - classical correspondence and hyperscaling present only at @xmath18 . in the following , we distinguish vmps results obtained using the shifted obb ( denoted by obb ) from the symmetry - enforced approach ( denoted by u(1)@xmath326 ) . in all calculations ,
we work with a fixed wilson discretization parameter @xmath119 , bond dimension @xmath327 and local bosonic dimension @xmath328 while varying the chain length @xmath153 and the effective local dimension @xmath329 , as denoted in the figures . since @xmath60 and @xmath90 are set equal for different sites during a single vmps calculation , we omit the label @xmath53 in the following .
this choice of @xmath59 and @xmath330 ensures that we keep all singular values larger than @xmath331 in our calculation .
we start the discussion with the continuous quantum phase transition between the lo and de phases that is controlled by the critical fixed point qc2 .
as explained in section [ sec : rg ] , this transition should correspond to the thermal transition of the xy chain with long - ranged interactions .
here we show that our numerical results are in excellent agreement with analytical predictions of scaling and epsilon - expansion calculations , listed in section [ sec : xyexp ] , and therefore fully confirm the quantum - to - classical correspondence .
accessible only at finite @xmath24 , we drive the transition between the de and lo phases by varying @xmath24 for fixed @xmath23 .
hence , the critical exponent @xmath209 is defined via @xmath332 at the phase boundary moving into the lo phase .
panels ( a ) and ( d ) of fig .
[ fig : exp_qc2 ] show the corresponding numerical data .
the characteristic power - law behavior of the magnetization for fixed @xmath333 close to the critical point on a log - log scale is displayed in fig .
[ fig : exp_qc2](a ) .
the exponent derived from a linear fit to this data , namely @xmath334 , corresponds to the mean - field prediction in eq .
( [ eq : betamf ] ) within the error bars .
deviations from power - law behavior at small @xmath335 can be attributed to a combination of finite chain length @xmath336 and numerical errors of vmps .
our numerical method generates power - law plots of similar quality for all @xmath337 ; the resulting exponents are collected in fig .
[ fig : exp_qc2](d ) .
these are found to be in excellent agreement with the predictions of the quantum - to - classical mapping . as for the classical xy chain
, the exponent assumes its mean - field value @xmath338 for @xmath339 , while it follows the two - loop rg results in eq .
( [ eq : beta_rg ] ) for @xmath340 . in the limit of @xmath201 , @xmath209 shows the tendency to approach the value 1/2 , consistent with eq .
( [ eq : beta_hyp ] ) .
the growing shifts in the localized phase , in combination with the decreasingly low - energy scale necessary to precisely access the critical point , prevents our numerics to extract accurate results for @xmath209 in the deep subohmic regime ( @xmath341 ) .
the second issue ( decreasingly low - energy scale ) also applies in the limit @xmath201 . in the lo close to the lo de transition .
the smooth behavior in the first iterations reflect the characteristics of the critical fixed point while the bending and jumps in the lines suggest that the system flows to the localized fixed point .
the red bar indicates the characteristic iteration @xmath342 of the transition that is used to calculate the low - energy scale @xmath343 . by tuning @xmath24 close to its critical value , @xmath342 moves towards higher iterations.,scaledwidth=50.0% ] as defined in eq .
( [ eq : delta ] ) , @xmath344 can be extracted from the response at criticality of the order parameter @xmath206 to an external magnetic field . fig . [
fig : exp_qc2](b ) displays the typical power - law scaling of the magnetization at the critical point .
the deviations at small @xmath182 are again related to finite system size and numerical artifacts . determining @xmath344 from power - law fitting over 6 decades for fixed @xmath333 , we find it to be in accordance with the mean - field predictions of the quantum - to - classical mapping , @xmath345 .
although the deep subohmic regime @xmath346 is again not accurately accessible for our vmps approach , the collected results for @xmath337 depicted in fig .
[ fig : exp_qc2](e ) strongly support the validity of quantum - to - classical mapping : for @xmath32 @xmath344 approaches its mean - field value of @xmath345 , while for @xmath149 it clearly follows the hyperscaling relation in eq .
( [ hs5 ] ) .
the definition of the correlation - length exponents @xmath189 in eq .
( [ nu ] ) involves a crossover energy scale @xmath214 that can easily be derived using the vmps energy - flow diagrams introduced in section [ sec : flowdiag ] . to this end
, we determine the site on the wilson chain @xmath347 where the flow starts to significantly deviate from the characteristically smooth flow at the critical point .
this approach is illustrated in fig .
[ fig : nu_flow ] , where two typical energy flows inside the localized phase close to qc2 are displayed . in the beginning the system resides at the critical fixed point ( smooth energy flow ) , then a transition to the localized fixed point occurs .
this crossover is indicated by the red bar corresponding to the iteration @xmath347 .
it is defined by the point where the first excited energy level drops below @xmath348 in rescaled energy units .
the crossover energy scales @xmath214 determined from such an analysis are collected in fig .
[ fig : exp_qc2](c ) for fixed @xmath333 and @xmath349 close to the phase transition .
the power - law scaling of @xmath214 over several orders allows us to extract @xmath189 with high accuracy .
studying the @xmath22-dependence in fig .
[ fig : exp_qc2](f ) , we again find excellent agreement with the classical xy model : @xmath189 closely follows the mean - field prediction [ eq .
( [ eq : numf ] ) ] for @xmath339 ( black dashed line ) , and also agrees with the perturbative rg calculations near @xmath31 [ eq .
( [ eq : nu_rg ] ) ] and @xmath26 [ eq . ( [ eq : nu_rg2 ] ) ] . as a further check
, we analyze the validity of the hyperscaling relation eq .
( [ hs4 ] ) involving both @xmath209 and @xmath189 by usage of our numerical results .
[ fig : hyp](a ) shows that the numerically extracted exponents obey hyperscaling for @xmath149 but clearly violate the respective relation in regime @xmath32 , as expected by quantum - to - classical correspondence . and
@xmath189 at qc1 ( a ) and qc2 ( b ) with the numerical data ( dots ) is compared to the exact results ( dashed line ) .
we find excellent agreement with the theory at qc1 for all values of @xmath350 as well as for @xmath149 at qc2 .
as expected by quantum - to - classical mapping , the numerical data confirm that hyperscaling ( hyp ) fails at qc2 below @xmath32 . ]
next , we consider the second continuous quantum phase transition of sbm2 between lo and cr phase , which is controlled by the critical fixed point qc1 . in this case , the quantum - to - classical correspondence is presumably violated , see section [ sec : qcc ] , and no analytical predictions for the critical exponents are available .
the transition between cr and lo phase can be driven by varying @xmath23 at @xmath18 and @xmath350 .
hence , @xmath209 is defined according to eq .
( [ eq : beta ] ) with the corresponding numerical data displayed in figs .
[ fig : exp_qc1](a ) and ( d ) .
the upper panel ( a ) depicts the scaling of @xmath180 close to @xmath185 for fixed @xmath351 on a log - log scale , where a power - law behavior is apparent over more than 3 decades , with an exponent @xmath352 .
the lower panel ( d ) shows the dependence of @xmath209 on the bath exponent @xmath22 gained from power - law scaling fits with similar quality as fig .
[ fig : exp_qc1](a ) .
we find increasing values of @xmath353 for @xmath354 , while in the limit of @xmath197 our vmps calculations suggest that @xmath209 approaches the value @xmath355 .
furthermore , we are able to show that @xmath209 in combination with @xmath189 satisfies the hyperscaling relation in eq .
( [ hs4 ] ) , as illustrated in fig .
[ fig : hyp](b ) . note that the extraction of @xmath209 is particularly complicated for qc1 , since this transition comes with a large exponent @xmath189 for the correlation length , on which we elaborate below .
this property relates to a low - energy scale required to resolve @xmath185 appropriately eq .
( [ eq : accuarcy ] ) , a precondition to obtain a solid power - law scaling of the order parameter @xmath180 .
such calculations involving large chain lengths ( @xmath356 ) become extremely sensitive to artificial symmetry breaking caused by numerical noise . therefore , the use of the symmetry - enforced vmps is essential in this parameter regime , for performance and accuracy reasons . in particular , the application of obb fails for energy scales significantly below double precision accuracy , since the small `` perturbations '' introduced by a shifted basis grow exponentially for later wilson shells , and hence break the energy - scale separation on the wilson chain .
this should not affect the validity of our results , since a shifted basis is not strictly required for @xmath149 ( see appendix [ app1 ] ) . and @xmath357 in the cr phase ( a ) and at lo cr critical point ( b ) .
the numerical results ( dots ) are in reasonable agreement with the hyperscaling relation ( dashed line ) in the cr phase ( a ) . at qc1 , however , hyperscaling appears to be violated .
note that the dots show @xmath358 , as we find that @xmath359 for all @xmath22 . ] for a transition at @xmath18 , it is possible to extract both exponents @xmath344 and @xmath357 via the order parameter s response to a magnetic field according to eqs .
( [ eq : delta ] ) and , respectively .
focusing first on @xmath344 , figs .
[ fig : exp_qc1](b ) and ( e ) illustrate the results of our vmps calculations .
again , the upper panel ( b ) shows the typical response of the magnetization to an increasing @xmath182 at the critical point for @xmath28 .
the robust power - law scaling over more than six decades allows us to extract @xmath360 with high accuracy .
the data collected from obb calculations with different values of @xmath22 in the lower panel ( e ) indicates that @xmath344 closely follows the hyperscaling relation in eq .
( [ hs5 ] ) .
in contrast to @xmath344 , we find the exponent @xmath357 , corresponding to the @xmath24 response , to be completely independent of the bath exponent @xmath22 , having @xmath359 for all @xmath22 at the lo cr transition ( not shown ) . as described above , the crossover energy scale @xmath214 characterizing the lo
cr transition is obtained by studying energy - flow diagrams close to the phase boundary .
[ fig : exp_qc1](c ) displays the extracted @xmath214 for fixed @xmath351 and @xmath18 , with clear power - law scaling being apparent over several decades .
this allows us to extract @xmath189 with high accuracy by fitting .
[ fig : exp_qc1](f ) shows the @xmath22 dependence of the exponent @xmath189 .
our results suggest that @xmath189 diverges both in the limit @xmath361 and @xmath362 , in a manner reminiscent of the approach to a lower critical dimension .
we have verified that the exponent @xmath189 is identical for both sides of the transition , i.e. , independent of whether qc1 is approached from the lo or from the cr phase .
generally , the computed values of @xmath189 take large values for the entire range of bath exponents .
as previously discussed , this causes our vmps calculations to require large chains ( @xmath356 ) in order to access the ultra - low energy scales needed to accurately determine @xmath185 .
a finite @xmath24 applied at the zero - field critical coupling @xmath364 places the system into the de phase .
the characteristic crossover scale @xmath365 obtained from the energy - flow diagrams determines the critical exponent @xmath363 , which only diverges in the limit @xmath362 but not for @xmath361 , in contrast to @xmath189 .
most importantly , @xmath357 and @xmath363 in combination do not obey the hyperscaling relation , as illustrated in fig .
[ fig : hyp_nup](b ) . hence , our results suggest that qc1 obeys hyperscaling properties only in the absence of a transverse field @xmath24 .
the underlying reason for this exotic critical behavior is not understood .
we note that the values for @xmath363 can be read off from fig .
[ fig : hyp_nup](b ) as @xmath359 for all @xmath22 .
they imply that @xmath366 for @xmath367 while @xmath368 for @xmath369 , i.e. , the role of the leading relevant operator at qc1 changes at @xmath370 , see fig .
[ fig : rgflow](b ) .
we have also investigated the flow _ along _ the separatrix between de and lo at small @xmath24 , in order to verify that qc1 is unstable along this separatrix , which implies that any finite - field transition is controlled by qc2 . to this end
, we first identify the stable energy - flow patterns corresponding qc1 and qc2 by placing the system at criticality for @xmath18 ( qc1 ) and sizeable @xmath24 ( qc2 ) , see fig . [
fig : qc1qc2_flow ] .
second , we study the energy - flow diagrams for parameters sets at criticality and very small @xmath24 . as displayed in fig . [
fig : qc1c2cross_flow ] , we observe a clear flow from qc1 at high energies to qc2 at lower energies , thus confirming the schematic rg flow diagram in fig .
[ fig : rgflow](b ) . .
, scaledwidth=50.0% ] we supplement the analysis of the critical phenomena of sbm2 by briefly elaborating on the properties of the impurity spin in the cr phase .
although an abridged version of the results has already been presented elsewhere,@xcite this section completes the picture and also includes a discussion on the validity of hyperscaling inside the cr phase .
the rg calculations around the free - spin fixed point , presented in section [ sec : rgf ] , predict a non - linear scaling of the magnetization in the cr phase , see eqs .
( [ eq : delta_cr ] ) and .
our numerical data confirm this non - linear response , as illustrated in figs .
[ fig : exp_cr](a ) for @xmath344 and in ( b ) for @xmath357 at different values of @xmath22 , @xmath23 chosen close to the cr fixed point @xmath371 .
we find a clear power - law scaling over several decades .
the extracted values for @xmath344 in fig .
[ fig : exp_cr](d ) are perfectly consistent with the hyperscaling result , while those for @xmath357 in fig .
[ fig : exp_cr](e ) are in good agreement with the perturbative results for @xmath201 .
the small deviations of the numerical data from the rg calculations for larger values of @xmath248 is expected , since the higher - order contributions in eq .
become more important .
, i.e. , close to qc1 . in both cases
, the level energies clearly flow from qc1 at high energies to qc2 at low energies , thus confirming the instability of qc1.,scaledwidth=50.0% ] for selected values of the bath exponent @xmath22 , obtained by u@xmath372 with @xmath373 ( a ) and @xmath374 ( b ) .
( a ) regime of @xmath183 where @xmath375 . ( b ) regime of @xmath27 where @xmath376 .
the power - law fits to determine the phase - boundary exponent @xmath377 are shown by dashed lines . ]
the energy - flow diagrams ( not shown ) confirm that the cr phase is unstable w.r.t .
a finite transverse field @xmath24 , i.e. , applying any finite @xmath24 places the system into the de phase .
the corresponding crossover scale @xmath378 between the cr and de fixed points allows us to extract the correlation - length exponent @xmath363 .
the collected results for different @xmath22 are displayed in fig .
[ fig : exp_cr](c ) and ( f ) , with @xmath363 being in fair agreement with perturbative prediction in eq . .
in contrast to qc1 , where the hyperscaling relation is not met by the numerical data , @xmath363 and @xmath357 obey hyperscaling in the critical phase , as depicted in fig .
[ fig : hyp_nup](a ) .
we have determined the location of the de lo phase boundary for small @xmath24 , in order to extract the expected power - law behavior .
sample results for @xmath183 , where the phase boundary starts at @xmath171 , are shown in fig .
[ fig : psi](a ) ; they are in essentially perfect agreement with the analytical result @xmath379 . for @xmath27
the phase boundary starts at the zero - field cr lo transition at @xmath205 , and thus determining @xmath377 requires an accurate knowledge of @xmath185 and is therefore rather time - consuming .
sample results are in fig .
[ fig : psi](b ) .
a hyperscaling - based guess would be @xmath380 which we find approximately fulfilled for @xmath381 and @xmath382 , but violated for @xmath351 .
( recall that the hyperscaling relation between @xmath363 and @xmath357 is violated as well . )
using the variational matrix - product - state approach , we have numerically determined the phase diagram of the u(1)-symmetric two - bath spin - boson model ( sbm2 ) , which is characterized by the phenomenon of frustration of decoherence .
our detailed study of the quantum phase transitions of sbm2 , using both numerical and analytical techniques , has revealed that the transition between the localized and delocalized phase , accessed at finite transverse field , is in the universality class of the xy spin chain with long - ranged interactions and thus obeys a quantum - to - classical correspondence .
in contrast , the zero - field critical ( intermediate - coupling ) phase and its transition to the localized phase do not have a classical counterpart . our numerical results for the critical exponents can serve as a guide for developing an analytical theory of the latter transition .
given that the relevant critical fixed point ( qc1 ) approaches the localized fixed point ( lo ) as @xmath201 , we believe that an expansion around lo akin to an expansion in @xmath323 dimensions for classical magnets should be able to access the properties of qc1 this task is left for future work .
we recall that the analysis in this paper has been restricted to the model sbm2 with symmetric couplings , i.e. , two identical baths and @xmath383 . for asymmetric couplings , with finite @xmath384 , the behavior of the model is driven towards that of the one - bath model sbm1 .
naturally , the lo phase now displays spontaneous ising order , with the impurity spin localized in direction of the stronger coupled bath .
further , the cr phase is unstable against any finite @xmath385 .
the rich and interesting crossover physics of sbm2 in the presence of small symmetry breaking is beyond the scope of this paper and will be discussed elsewhere .
interesting open questions concern the finite - temperature behavior of sbm2 , specifically the quantum critical finite-@xmath177 susceptibilities and the residual entropy , as well as its equilibrium and non - equilibrium dynamics .
generalizations to three bosonic baths as well as combined fermionic and bosonic baths would be interesting as well .
the former is linked to the problem of impurity spins in quantum critical magnets,@xcite and both occur in self - consistent single - site solutions for certain lattice models , e.g. , in the large-@xmath153-based theory of a gapless spin liquid @xcite and in more general extensions of dynamical mean - field theory.@xcite in the quest for non - trivial quantum critical behavior , we believe that sbm2 presents in a sense the simplest quantum model violating the quantum - to - classical correspondence : it lives in @xmath216 dimensions and is constructed solely from bosonic degrees of freedom .
our analysis reveals that the violation of the quantum - to - classical correspondence is rooted in the non - commutativity of the spin components coupled to the two baths ; this property of a quantum spin can also be re - phrased as a spin berry phase .
we note that quantum phase transitions in quantum impurity models with _
fermionic _ baths frequently behave non - classically , with the pseudogap kondo and anderson models@xcite being well - studied examples . here , the absence of a quantum - to - classical correspondence can be traced back to fermionic `` signs '' , i.e. , exactly integrating out the fermionic bath is only possible at the expense of working with a fermionic impurity , which has no classical analogue .
we thank a. alvermann , s. florens , s. kirchner , k. ingersent , q. si , a. schiller , and t. vojta for helpful discussions .
this research was supported by the deutsche forschungsgemeinschaft through the excellence cluster `` nanosystems initiative munich '' , sfb / tr 12 , sfb 631 , we4819/1 - 1 ( aw ) , for 960 , by the german - israeli foundation through g-1035 - 36.14 , and the nsf through phy05 - 51164 .
here we sketch the use of the scaling hypothesis to deduce hyperscaling relations for qc1 .
the standard homogeneity law for the critical contribution to the free energy implies the scaling form @xmath386 recall that the problem at hand is effectively ( 0 + 1)-dimensional . here ,
@xmath387 and @xmath24 correspond to two operators which drive the system away from criticality , and @xmath388 is a scaling function .
the definitions of the correlation - length exponents in eqs . and lead to the identifications @xmath389 and @xmath390 . taking the derivative of eq .
@xmath182 yields @xmath391 which can be cast into the forms @xmath392 and @xmath393 upon taking the limit @xmath394 in eq .
one deduces the order - parameter exponent as @xmath395 ; similarly eq .
yields @xmath396 or @xmath397 . using @xmath389 then leads to @xmath398 which is consistent with the hyperscaling relations and . taking the second derivative of eq .
@xmath182 yields @xmath399 and facilitates the identification @xmath400 .
together with @xmath397 this yields @xmath401 , consistent with the relations and . in full analogy , taking the derivative of eq .
@xmath24 yields @xmath402 or @xmath403 .
using @xmath390 finally gives @xmath404 which is eq . .
taking the second derivative w.r.t .
@xmath24 yields @xmath405 and then @xmath406 which is eq . .
the hyperscaling relations and can also be applied in the cr phase where @xmath24 corresponds to a relevant operator as well .
note that the nature of the exponent pair @xmath407 is different from that of @xmath408 : @xmath189 parameterizes the scaling dimension of @xmath385 at criticality and @xmath344 the non - linear response to a field conjugate to the order parameter .
in contrast , @xmath363 and @xmath357 correspond to the scaling dimension of and the non - linear response to the _ same _ field , @xmath24 .
as the numerical computations are done for finite wilson chains , it is worth discussing finite - size effects arising from a finite chain length @xmath153 .
the most important effect of finite @xmath153 is to induce a gap @xmath409 in the bath spectrum which scales as @xmath410 .
while this gap has no effect in the de phase , as its fixed point corresponds to @xmath171 , it prevents true spontaneous symmetry breaking in the lo phase .
however , this does not affect our calculations because , with increasing @xmath153 , the finite correlation length induced by @xmath409 increases faster than the system size , such that the finite - size system `` looks '' ordered in the lo phase once @xmath153 is sufficiently large . for different points close to the cr fixed point @xmath371 .
we observe that the magnetization decreases exponentially with system size .
the small value of the decay exponent @xmath411 results in a notable finite - size effects even for large systems [ u@xmath372 with @xmath412 . ] most problematic are finite - size effects in the cr phase . here
@xmath413 in the infinite - system limit , but a bath gap induces a finite residual magnetic moment scaling as@xcite @xmath414 , with @xmath220 defined in eq . .
indeed , our vmps calculations in fig .
[ fig : crmag ] find a small but finite magnetization .
[ fig : app_crfss ] supports that @xmath180 indeed vanishes in infinite - system limit as @xmath415 with the system size @xmath416 on a wilson chain .
given that the exponent @xmath411 governing the decrease of @xmath180 is very small , the order parameter remains finite even for very large systems .
the fit exponent @xmath411 allows us to extract the value of the exponent @xmath220 in the cr phase according to @xmath417 .
the values for @xmath220 obtained in this way are indicated in fig .
[ fig : app_crfss ] and are consistent with the hyperscaling result @xmath418 , see sec .
[ sec : scaling ] .
we note that a direct measurement of @xmath220 at the various critical points is not easily possible using the present numerics , as ( i ) the variational approach is designed for @xmath176 only , and ( ii ) the mass - flow problem@xcite would prevent an accurate approach to critical points using chains of different length .
as numerical artifacts play an increasingly important role close to the critical phase , we found it to be essential to enforce the conservation of the u(1 ) symmetry when trying to access the critical properties of qc1 and qc2 for large bath exponents @xmath419 . since the symmetry incorporation excludes employing a shifted obb - vmps calculation , it is fair to ask whether the bosonic truncation error corrupts the presented results of the critical exponent @xmath209 .
careful analysis revealed a similar situation as in the sbm1,@xcite where the resulting critical exponents are only affected by the truncation error in the regime @xmath32 . in the same fashion ,
hilbert - space truncation in the sbm2 only influences the behavior of critical properties for @xmath32 , as illustrated in fig .
[ fig : app_exp ] . comparing
the scaling of the magnetization to determine @xmath209 and @xmath344 using vmps calculations with and without shifted obb reveals the characteristic difference between @xmath32 and @xmath313 . in fig .
[ fig : app_exp](a ) and ( b ) we observe significant deviations between both types of calculations for @xmath333 .
employing the shifted obb method , the resulting critical exponents are in good agreement with the mean - field predictions @xmath420 and @xmath345 , while vmps calculations without shift lead to considerable deviations from the mean - field values .
in contrast , considering a larger bath exponent @xmath421 , we clearly obtain the same results for both types of vmps calculations , as illustrated in fig . [
fig : app_exp](c ) and ( d ) .
thus for the evaluation of critical exponents we conclude : the shifted obb is only strictly necessary for small bath exponents @xmath32 , whereas for of @xmath313 vmps calculations with and without shifted obb work equally well .
and @xmath344 employing vmps with ( blue ) and without shifted obb ( purple ) .
choosing @xmath422 in panels ( a ) and ( b ) , we observe considerable deviations between both types of vmps calculations where the mean - field prediction for @xmath209 and @xmath344 is only obtained with a shifted obb . in the case of @xmath313 ,
both methods lead to similar results , as illustrated in panels ( c ) and ( d ) for the exponent @xmath421 [ obb with @xmath412 . ]
the ground state of sbm2 in the lo phase with @xmath15 exhibits a continuous degeneracy due to the inherent rotational symmetry , which was elaborated on in section [ sec : u1 ] .
when not enforcing the u(1 ) symmetry , the final ground state of a vmps calculation spontaneously breaks this u(1 ) symmetry , while maximizing magnetization @xmath423 in @xmath220- and @xmath222-direction in the localized phase ( note that this is the least entangled state ) .
in contrast , for a u(1)-symmetric implementation , these expectation values vanish by construction . however , it is possible to attach a well - defined symmetry label ( @xmath424 ) to the numerical ground state .
the two resulting states @xmath425 form an orthonormal pair , which can be used to construct the space of all ( symmetry - broken ) ground states . by symmetry ,
the expectation value @xmath426 evaluated using only one symmetry eigenstate , @xmath427 , gives zero . to reconstruct the magnetization of the `` original '' , symmetry - broken ground state
, we have to calculate the magnetization using non - diagonal elements @xmath428 of the above defined expectation value . in general ,
this can be accomplished in two different ways .
the simple but numerically expensive variant is to use two vmps runs to obtain @xmath429 and @xmath430 separately for the same parameters and explicitly calculated @xmath431 .
alternatively , we may borrow a concept of nrg that allows us to use only a single vmps to determine the magnetization for a system with arbitrary wilson chain length @xmath432 . starting with the right - orthogonalized representation of either @xmath429 or @xmath430
, we construct and diagonalize the left block hamiltonian @xmath151 . after projecting into the subspace of the two lowest - lying energy states @xmath433 , with @xmath434 , @xmath435 } a^{[n_1 ] } \, .. \,a^{[n_{n } ] } \big)_s |n_1,n_2,\, ...
\,,n_{n}\rangle\,,\ ] ] we explicitly determine all matrix elements @xmath436 } \equiv \langle s_k|\sigma_{i = x , y } |s_k'\rangle_n$ ] of the magnetization .
the eigenvalues of the @xmath437 matrix @xmath438}$ ] give the two possible values of the magnetization of the system with chain length @xmath53 in the ground state @xmath439 .
therefore the plain thermal average _ without spontaneous symmetry breaking _ would result in zero magnetization .
independently of how the magnetization is calculated , we face an additional challenge regarding the hilbert space truncation error in the context of explicit symmetry implementation . studying the finite - size scaling of the magnetization in the localized regime
, we expect @xmath180 to saturate at a finite value after an initial decay when moving towards larger systems . as illustrated in fig .
[ fig : app_fss](a ) , our vmps data for @xmath180 indeed saturates as expected when symmetry is not enforced ( solid line ) . however , when employing the symmetry implementation we observe a further decrease ( dashed line ) after the saturation to an intermediate plateau . considering the behavior of the bosonic occupation numbers on the wilson chain in fig .
[ fig : app_fss](b ) , we attribute this effect with the hilbert - space truncation error .
the difference between solid and dashed lines sets in once the bosonic occupation numbers @xmath163 for the symmetry - enforced implementation , where a shifted obb can not be used , begin to saturate ( fig .
[ fig : app_fss](b ) , dashed line ) , whereas those for non - symmetry - enforced implementation , for which a shifted obb can be used , do not yet saturate ( fig .
[ fig : app_fss](b ) , solid line ) . to circumvent this systematic error
, we extract @xmath180 not at the end of the chain but choose an iteration @xmath440 right before @xmath441 saturates ( indicated by the red dashed line in fig . [
fig : app_fss ] ) . at @xmath440
the magnetization from the symmetry - enforced code clearly agrees with a vmps calculation using a shifted bosonic basis . as indicated in the previous section
, this approach is only appropriate for bath exponents @xmath313 . for smaller values of @xmath22
it is absolutely necessary to employ vmps with the shifted obb scheme in order to capture the correct physical properties of the system .
s. m. cronenwett , t. h. oosterkamp , and l. p. kouwenhoven , science * 281 * , 540 ( 1998 ) .
d. goldhaber - gordon , h. shtrikman , d. mahalu , d. abusch - magder , u. meirav , and m. a. kastner , nature * 391 * , 156 ( 1998 ) . y. makhlin , g. schn , and a. shnirman , rev .
phys . * 73 * , 357 ( 2001 ) .
a. h. castro neto , e. novais , l. borda , g. zarand , and i. affleck , phys .
lett . * 91 * , 096401 ( 2003 ) ; e. novais , a. h. castro neto , l. borda , i. affleck , and g. zarand , phys .
b * 72 * , 014417 ( 2005 ) . technically , the rg treatment requires the introduction of a renormalized running coupling .
its initial value is determined by the bare coupling and suitable powers of the ultraviolet cutoff scale , with details depending on the rg scheme . in our case ,
the cutoff is @xmath442 , and we can identify the initial value of the running coupling with the bare coupling . | the spin - boson model , describing a two - level system coupled to a bath of harmonic oscillators , is a generic model for quantum dissipation , with manifold applications .
it has also been studied as a simple example for an impurity quantum phase transition . here
we present a detailed study of a u(1)-symmetric two - bath spin - boson model , where two different components of an su(2 ) spin 1/2 are coupled to separate dissipative baths .
non - trivial physics arises from the competition of the two dissipation channels , resulting in a variety of phases and quantum phase transitions .
we employ a combination of analytical and numerical techniques to determine the properties of both the stable phases and the quantum critical points .
in particular , we find a critical intermediate - coupling phase which is bounded by a continuous quantum phase transition which violates the quantum - to - classical correspondence . |
the linear collider community has set a goal to achieve a relative precision of @xmath1 on luminosity measurement .
presently the forward calorimetry collaboration ( fcal ) @xcite is considering two possible designs for the luminosity detector ( lumical ) .
both designs are based on a tungsten / silicon calorimeter .
they differ in the readout design , pad or strip . here
we report on studies performed to optimize the performance of the pad readout design .
the monte carlo studies , presented in the current note , include simulation of bhabha scattering , beam - beam interactions , beam spread as well as the full simulation of the lumical . for bhabha scattering events we used bhwide @xcite , a monte carlo multi - photon event generator which provides four - momenta of the outgoing electron , positron and photons radiated in the initial and final state .
the program circe @xcite was used to study the distortion of beam energy spectrum due to beamstrahlung .
two different values of a gaussian beam spread , 0.05% and 0.5% of the nominal center of mass energy ( @xmath2 ) , were investigated in the range of beam energy between 50 and 400 gev .
the detector simulation was performed using the brahms @xcite package based on the standard geant 3.21 simulation program @xcite .
the performance of lumical was studied in three stages , * with the basic detector design using single electrons / positrons , * with the basic detector design and a more realistic physics simulation , including simulation of bhabha scattering events , beamstrahlung and beam spread , * varying the detector design for optimization purposes .
[ es ] shows an example of the center of mass energy spectrum of the @xmath0 pair originating from bhabha scattering , including radiative effects , beamstrahlung and a beam spread of 0.05%@xmath2 for nominal 250 gev beam energies . the main contribution to
the tail comes from the initial state radiation in bhabha scattering .
the detector covers polar angles @xmath3 from 27 to 91 mrad with respect to the beam line .
longitudinally , the detector consists of 30 layers composed each of 500 @xmath4 m thick silicon sensors and a tungsten - silicon mixture of 0.34 cm of tungsten and 0.31 cm of silicon and electronics .
the detector , with an inner radius of 8 cm and an outer radius of 28 cm , is subdivided radially into 15 cylinders and azimuthally into 24 sectors .
each layer corresponds to a depth of about one radiation length .
the cell transverse size is approximately one @xmath5 radius .
the total number of cells ( electronic channels ) is equal to 10,800 .
two identical arms , one for the electron side and the second for the positron side , are positioned along the @xmath6 axis ( beam line ) , symmetrically with respect to the interaction point ( ip ) , 3.05 m away from the ip .
for luminosity measurement , the geometric acceptance is the most significant event selection rule .
the strong @xmath3 dependence of the bhabha scattering , @xmath7 , makes the low angle cut crucial .
we used a method , in which only a few layers govern the events selection . in this method
the energy deposited in three layers located in the middle of the detector , close to the shower maximum , is divided into energy deposited in the inner edge cylinders , @xmath8 and outside , @xmath9 .
the variable , _ p _ , defined as @xmath10 is then used to estimate the shower containment .
events with _
p_@xmath110 are rejected as being out of the acceptance region , while events with _ p_@xmath120 are kept .
the behavior of the variable _ p _ as a function of the polar angle of the showering electron is shown in fig . [ selection ] for three different definitions of @xmath8 , summed over one , two , or three cylinders . a given acceptance cut in @xmath3 can be translated into an appropriate number of edge cylinders such that a cut on @xmath13 will reject the right events without necessity for full reconstruction .
the events were preselected using _
p _ only calculated for the electron detector arm .
the bhabha scattering events were further selected by requiring that the showers reconstructed in both arms of the detector be back to back .
two approaches were used to reconstruct the position of the shower , with no attempt to reconstruct clusters of energy .
the position was determined as the weighted average , @xmath14 where @xmath15 is the location of the center of the @xmath16-th pad , @xmath17 is the weight and the sum runs over all pads . the simple , energy weighted average , @xmath18 , is known to be biased , with the bias depending on the size of the pads and the impact point of the shower on the pad . in the second method
@xcite , the weight is assumed to be proportional to the logarithm of the energy deposited , and in addition a cut - off is introduced so that effectively only significant energy deposits contribute , @xmath19\ } , \label{logwei}\ ] ] where @xmath20 . the cut - off , @xmath21 , has to be optimized for best resolution .
the resolution as a function of cut - off is shown in fig .
[ con ] for three different incoming energies .
the optimal cut - off is found to increase with energy . in parallel , for each cut - off value the bias in reconstructing the position was checked .
the point of best resolution turned out to correspond to the least bias . just by tuning the cut - off , the polar angle resolution , @xmath22 , for 250 gev energy electrons was found to improve by factor three with @xmath23 mrad .
in addition , the bias was improved by an order of magnitude .
the polar angle resolution obtained for single particles with the optimal weighting , is shown as a function of the beam energy in fig .
[ thres ] , for the various event configurations .
typically , the best resolution is achieved for the single particle mc sample , while for bhabha scattering with a 0.5% beam spread the resolution is 10% worse . a small residual bias in the polar angle reconstruction , @xmath24 , was observed . for bhabha scattering with a 0.05% beam spread ,
the relative value @xmath25 is of the same order of magnitude as the required luminosity precision .
more statistics is needed to establish , whether this is a genuine effect .
the energy resolution , @xmath26 , as a function of energy is shown in fig .
[ eres ] , again for the various event samples .
the resolution follows the expected behavior of @xmath27 , with @xmath28 varying between [email protected] and [email protected] .
the best resolution is achieved for the sample of bhabha scattering events with a small beam spread .
this is probably due to the requirement that the two showers be back to back , which prevents residual energy leakage .
once the performance of the basic detector design was established , an attempt was made to optimize the depth and the granularity of the calorimeter . for that purpose the calorimeter
was assumed to have 50 active layers .
a sample of 1000 events was generated for this study .
the angular resolution was studied as a function of the depth and granularity , which was improved by increasing the number of cylinders for the same geometry .
the results are shown in fig .
[ dns ] . for
a given depth of the calorimeter , the resolution is improving with increasing number of readout pads .
no improvement is observed beyond a depth of 30 layers .
for 30 active layers , increasing the number of cylinders from the 15 of the basic design to 20 , leads to an angular resolution better than @xmath30 rad .
the number of sectors was also increased .
this improves the resolution of the azimuthal angle , but has no effect on the polar resolution .
increasing the density of the calorimeter would lead to a more compact detector , with less leakage .
however the number of pads would have to be increased to match the angular performance of the basic design .
an attempt was made to achieve a better angular resolution by improving the granularity locally , in the layers containing the maximum of the electromagnetic shower .
if this is done under the constraint of a fixed total number of readout channels , with 20 cylinders in the inner 15 layers and 10 cylinders in the outer layers , the resolution may be improved by 15% .
to achieve sensitivity to systematic effects comparable to the required relative precision on luminosity of @xmath1 , large statistics mc samples are necessary .
this can not be achieved in a conventional manner , by processing events through a full geant simulation . instead , a fast mc was developed , with smearing effects implemented through parameterization of the performance established on smaller samples .
this mc allows detailed studies of various systematic effects , either related to geometry or possible mismatch between mc simulation and detector performance in reality .
the influence of a bias in @xmath3 reconstruction on the luminosity error , obtained with fast mc simulation , is shown in fig .
[ delta - l ] .
sensitivity to shifts of the order of @xmath1 are visible .
the simulation reproduces well the expected analytical result , @xmath31 .
the fast mc was also used to assess whether the presently achieved angular resolution would be sufficient to control the @xmath32 with the required precision . assuming that the bhabha scattering events are selected by requiring that the two electromagnetic showers be back to back within three times the expected resolution , and assuming a precision of 10% on the resolution itself , would lead to an error on the luminosity , @xmath33 . however ,
if the back to back requirement is relaxed to five times the expected resolution , the projected luminosity error , for the same uncertainty on the resolution , is negligible
. therefore , further studies are needed to understand the required quality for selecting bhabha candidate events .
the luminosity detector at the future linear collider is expected to provide measurements with a precision better than @xmath1 . in this study , of a tungsten - silicon calorimeter with pad readout
, we have concentrated mainly on optimizing the reconstruction algorithms .
we have demonstrated that an improvement of a factor of three is possible without changing the granularity .
an attempt was also made to optimize the angular resolution by changing the granularity , either by increasing the total number of readout channels or by improving the granularity at the expected shower maximum location while keeping the total number of channels unchanged .
a fast mc was developed to study systematics effects with a sensitivity compatible with the required precision . | this paper presents the status of monte carlo simulation of one of the luminosity detectors considered for the future @xmath0 international linear collider ( ilc ) .
the detector consists of a tungsten / silicon sandwich calorimeter with pad readout .
the study was performed for bhabha scattering events assuming a zero crossing angle for the beams . |
the delafossite structural series of oxides has recently attracted considerable attention because of the remarkable and varied properties of the compounds in the series .
its general formula is abo@xmath1 , in which a is a noble metal ( pt , pd , ag or cu ) and b a transition metal ( e.g. cr , co , fe , al and ni ) [ ] .
the metal atoms are found in layers with triangular lattices , stacked along the perpendicular direction in various sequences , leading to structures similar to the layered rocksalt structure of the well - known ionic conductor licoo@xmath1 and of nacoo@xmath1 , which superconducts when intercalated by water .
interlayer coupling is weak , so the delafossites are quasi - two dimensional .
the low temperature properties across the series vary considerably with a and b combinations .
known materials include candidate magnetoelectric insulators and thermoelectrics [ , ] , transparent conductors [ ] and band insulators [ ] .
there are also intriguing metals such as agnio@xmath1 and pdcro@xmath1 in which the conduction takes place between mott insulating layers with magnetic order [ ] , and non - magnetic metals with ultra - high conductivity , such as pdcoo@xmath1 .
although they have been less studied than the layered perovskites , the delafossites have considerable potential for new physics and technology ; epitaxial multilayers of delafossites and of delafossite / rocksalt combinations can be imagined .
there is , therefore , a strong motivation to understand the fascinating properties of individual materials in the series . in pdcoo@xmath1 , cobalt has formal valence of 3 + and the @xmath7 configuration , meaning that the states at the fermi level have dominantly pd character [ ] , with a single half - filled band crossing the fermi level . at 300
k the resistivity ( @xmath8 ) is 2.6 @xmath2 cm , lower per carrier than that of cu [ ] . at low temperatures both the shape and
the value of @xmath8 are very unusual .
it is best fitted by @xmath9 ( with @xmath10 a constant and @xmath11 k ) rather than the power law expected for standard phonon scattering . at least as surprising as the form of @xmath8 is its value .
a residual resistivity of only 7.5 n@xmath12 cm has been reported [ ] .
this corresponds to a mean free path ( @xmath13 ) of 20 @xmath14 m , or @xmath15 lattice spacings , an astonishing value in crystals grown from fluxes in hot crucibles that have not been subject to any post - growth purification .
the high value of @xmath13 in pdcoo@xmath1 has important consequences for a number of physical properties , for example the out of plane magnetoresistance which is huge and varies strongly with field angle [ ] , behavior consistent with observation of the long sought ` axial anomaly ' [ , ] .
the subject of this paper , ptcoo@xmath1 , offers an opportunity to extend the study of the intriguing metallic delafossite oxides to a @xmath0 system where spin - orbit coupling ( soc ) is considerably stronger .
although crystal growth [ ] and room temperature resistivity [ ] were reported over forty years ago , no temperature - dependent transport studies were carried out on those crystals , and further crystal growth proved to be extremely challenging .
here we describe crystal growth using a different method to those previously attempted in ptcoo@xmath1 .
this yields high - quality single crystals large enough for spectroscopic measurements . from angle - resolved photoemission ( arpes ) , combined with density - functional theory calculations and precise transport measurements performed on focused ion beam patterned crystals , we show that ptcoo@xmath1 hosts ultra - high conductivity derived from a single half - filled and nearly - free - electron - like pt @xmath0 band .
first , we describe our crystal growth .
shannon _ et .
al _ [ ] reported single crystal growth of pt@xmath16co@xmath17o@xmath1 ( where x and y were @xmath18 ) several decades ago .
stoichiometric ptcoo@xmath1 was grown only under high pressure ( 3000 atm ) , with other growth conditions resulting only in nonstoichiometric crystals .
al _ [ ] were able to grow stoichiometric ptcoo@xmath1 by using a metathetical reaction under vacuum , but the crystal size was limited to @xmath19 m .
here we used a technique similar to that applied in ref .
[ ] to the growth of pdcoo@xmath1 ( see methods ) to realise large stoichiometric ptcoo@xmath1 single crystals . * single - crystal ptcoo@xmath1 samples .
* ( a ) optical microscope image of as - grown crystals of ptcoo@xmath1 .
( b ) le - bail fitting of powder x - ray diffraction pattern along with fitted curve , and the difference curve .
all peaks are labeled with corresponding @xmath20 values .
peaks marked with * correspond to unavoidable unreacted ptcl@xmath1 stuck to the crystal surface .
( c ) sem image of a sample used for transport measurements in which a focused ion beam was used to define a measurement track of well - defined geometry.,width=12 ] fig .
[ f1](a ) shows optical pictures of the as - grown crystals .
they form as triangular or hexagonal plates .
terrace - type lateral growth leads to variation in crystal thickness from one side to another as evident from fig .
[ f1](a ) .
however there were many crystals with uniform thickness without any steps .
due to their brittleness and layered nature , the typical size of these crystals varies from in - plane dimensions of a few @xmath14 m to 0.3 mm , with a maximum thickness of @xmath21 m .
bigger crystals were grown after employing modified techniques with more complex temperature profiles [ ] .
a cluster of such big as - grown crystals is shown in the inset of fig .
[ f1](a ) .
these crystals were used for the arpes measurements discussed below .
single crystals were separated mechanically using a scalpel , and then separated from unreacted coo and from cocl@xmath1 powder by cleaning the product with boiling alcohol .
all crystals were characterized by a scanning electron microscope ( sem ) with an electron probe micro analyzer ( epma ) and confirmed to be single phase and have no inclusions .
le - bail fitting of the powder x - ray diffraction pattern ( xrd ) was performed using space group @xmath22@xmath23@xmath24 ( space group no .
166 ) . the experimental xrd pattern along with the fitted pattern and the difference between the two are shown in fig . 1 ( b ) .
the refined lattice parameters are @xmath25 and @xmath26
. * temperature - dependent transport . * ( a ) the temperature dependent in - plane resistivity of ptcoo@xmath1 in zero applied magnetic field .
the inset shows a magnified view of the low - temperature resistivity revealing an upturn below 16 k. ( b ) the temperature dependence of the hall coefficient ( @xmath27 ) , calculated by taking the field gradient between 7 and 9 t of the data shown in supplementary fig .
s1.,width=10 ] since the resistivity of ptcoo@xmath1 is very low [ ] , it was challenging to obtain accurate absolute values of resistivity due to uncertain geometrical factors . to overcome this , and to enhance the precision with which we could measure even smaller resistances at low temperatures , we made use of focused ion beam techniques to prepare samples with well - defined geometries [ ] .
one example is shown , along with its dimensions ( length between two voltage contacts 40 @xmath14 m , width 8.4 @xmath14 m and thickness 2.6 @xmath14 m ) , in fig . 1(c ) . in fig . [ f2 ]
we show the in - plane resistivity of ptcoo@xmath1 crystals from the growth run described above . in total , approximately ten crystals were studied , and the reproducibility of the temperature dependence was excellent .
the data are from the well - defined microstructure shown in fig .
[ f1 ] , allowing the absolute value of resistivity to be determined with an accuracy of better than 5% . at room temperature , @xmath28 @xmath2 cm , the lowest ever measured in an oxide metal .
this is 20% lower than that of pdcoo@xmath1 [ ] and approximately a factor of 4 [ ] and 12 [ ] lower than in reo@xmath29 and iro@xmath1 , respectively , which are themselves famous examples of good @xmath0 oxide conductors . the resistivity decreases to less than 40 n@xmath12 cm at 16 k , but then rises again by approximately 4% between 16 k and 2 k as shown in the inset of fig . [ f2 ] ( a ) .
the residual resistivity ratio @xmath30 is typically 50 - 60 .
the upturn is observed consistently in both as - grown crystals and those that were microstructured with the focused ion beam , so we are certain that it is not the result of ion - beam induced disorder .
the measured magnetoresistance of ptcoo@xmath1 is positive for all temperatures above 2 k , making it unlikely that the kondo effect is the cause of the upturn
. its origin will be the subject of further investigation . in fig .
[ f2 ] ( b ) we show the measured hall coefficient deduced from measurements of the transverse resistivity of the device shown in fig .
[ f1](c ) ( see also supplementary fig .
s1 ) . below 20 k ,
the data are dominated by the high - field regime in which the fermi surface volume alone is expected to determine the value of the hall coefficient [ ] .
the value in that region is approximately @xmath31 m@xmath32/c , close to the value that we measure for pdcoo@xmath1 under the same conditions ( data not shown ) , and within 10% of the expectation for a single band containing one electron per pt . *
single - band faceted fermi surface . * ( a ) fermi surface of ptcoo@xmath1 measured by arpes integrated over @xmath33 mev . the solid line represents the brillouin zone .
the fermi surface area is 8% smaller than would be expected for a half - filled band , as discussed in the main text , but its shape is in excellent agreement with the fermi surface obtained from gga+so+u calculations ( @xmath34 ev ) , shown in red in ( b ) and scaled to match the experimental area .
the dots represent the fermi momenta extracted from ( a ) by fitting mdcs radially around the measured fermi surface.,width=14 ] our arpes measurements ( fig .
[ f3 ] ) directly reveal just such a fermi surface , comprised of a single electron pocket centred at the @xmath35 point of the brillouin zone .
the sharp spectral features are indicative of quasi - two - dimensional electronic states [ ] .
the near - hexagonal fermi surface , rotated by 30@xmath36 with respect to the brillouin zone , is qualitatively similar to that of the bulk fermi surface of pdcoo@xmath1 [ ] .
we find , however , that the ptcoo@xmath1 fermi surface has sharper corners and slightly concave faces .
this is consistent with previous first - principles calculations [ ] , as well as our own density - functional theory results ( fig .
[ f3](b ) ) .
our calculations predict an additional hole band in the vicinity of @xmath37 at the zone edge ( fig .
[ figdos](a ) ) , whose position depends delicately on the oxygen position ( 0,0,@xmath38 ) as shown in supplementary fig .
s2(b ) . neglecting soc ,
this intersects the fermi level when using either the experimental [ ] or fully - relaxed oxygen positions , creating additional small lens - shaped fermi surface pockets . including soc
, however , lowers these bands by @xmath39 mev , pushing them below the fermi level .
spin - orbit interactions therefore stabilise a robust single - band fermi surface in ptcoo@xmath1 , consistent with our arpes measurements as well as recent de - haas van alphen results [ ] .
( a ) calculated bulk electronic structure ( see methods ) with and without spin - orbit coupling .
the high - symmetry points are labelled on the bulk brillouin zone shown inset .
inclusion of soc pushes the hole bands at , e.g. , the l - point below @xmath37 , while leaving the remaining bands crossing the fermi level almost unchanged .
( b ) partial and ( c ) orbitally - resolved density of states ( including soc ) , revealing the states at e@xmath6 to originate almost exclusively from pt-5d .
, width=10 ] * weakly - interacting quasiparticle dispersion .
* ( a ) electronic structure along the @xmath40 direction . the single band crossing the fermi level can be traced down to more 0.5 ev binding energy with little broadening .
the dashed red line corresponds to a parabolic dispersion with an effective mass @xmath41 .
an mdc at the fermi level ( @xmath42 mev ) is shown by the dots , with a fit to a lorentzian peak indicating a fwhm of @xmath43 @xmath44 .
( b ) the grey and black dots represent the peak positions of the fits to the mdcs along the @xmath45 and @xmath40 direction , respectively . a linear fit to each data set
independently yields a fermi velocity of @xmath46 m / s ( shown by the solid line ) , giving an effective mass of @xmath47 along @xmath40 and @xmath48 along @xmath45.,width=12 ] the fermi surface area we extract from arpes , by fitting momentum distribution curves ( mdcs ) as a function of angle around the brillouin zone , is @xmath49 @xmath50 . the corresponding luttinger count , assuming perfect two - dimensionality , is @xmath51 .
this is comparable to arpes measurements of pdcro@xmath1 [ ] , but is slightly smaller than that expected for a half - filled zone derived from a single conduction electron per pt .
in contrast , de haas - van alphen measurements indicate a luttinger count within 2% of half - filling .
we thus ascribe the deviations in our arpes measurements as due to a possible small difference in surface chemistry from the bulk , sample - to - sample stoichiometry variations , or surface charge transfer . nonetheless , we find almost perfect agreement of the shape of our measured fermi surface with our calculations from density - functional theory scaled to have the same area ( fig .
[ f3](b ) ) .
we therefore conclude that our measurements are indicative of the bulk electronic structure of ptcoo@xmath1 .
we note that , at lower photon energies , weak spectral features emerge giving rise to shoulders on the states reported here and residual spectral weight at the zone centre . from comparison to previous surface electronic structure calculations for pdcoo@xmath1 [ ] and arpes measurements of pdcro@xmath1 [ ] , we attribute these to pt - terminated surface states not visible at the higher photon energies shown here due to photoemission cross - section variations [ ] , although their weak weight precludes an unambiguous assignment between these features .
other cleaves showed stronger surface - derived features but a bulk band in quantitative agreement with the one found here .
[ f5](a ) shows the bulk band dispersion measured along the @xmath40 direction .
consistent with our measured fermi surface , we find a single band in the vicinity of @xmath37 .
it is extremely steep , dispersing by less than 5% of the brillouin zone over the energy range of 0.6 ev below the fermi level shown in fig [ f5](a ) .
indeed , from fits to mdcs , we find negligible deviations from linearity over a range exceeding 0.25 ev below @xmath37 , together with a high fermi velocity of @xmath52 m / s [ @xmath53 ev/@xmath54 .
this is several times larger than observed in other metallic @xmath0 oxides [ , ] , and yields an effective mass , @xmath41 , within 10% of the free - electron mass . with no free parameters
, employing this effective mass within a simple parabolic model well reproduces our measured dispersion over an extended energy range ( fig .
[ f5](a ) ) . the fermi velocity [ @xmath55 m / s ] that we extract along @xmath45 is the same as that along @xmath40 within our experimental resolution ( fig .
[ f5](b ) ) .
this is almost 20% higher than that measured along the @xmath45 direction in pdcoo@xmath1 [ ] . due to the higher @xmath56 along @xmath45 than @xmath40 ,
the effective mass obtained from our measured fermi velocity , @xmath57 , deviates further from the free - electron mass .
the average mass of @xmath58 is in excellent agreement with the value of @xmath59 obtained from quantum oscillations of ptcoo@xmath1 [ ] , and again notably smaller than the values of around @xmath60 obtained for pdcoo@xmath1 [ ] .
the above measurements point to light quasiparticles in this system , in keeping with their extremely high conductivity , with an even wider electronic bandwidth of the state crossing @xmath37 in ptcoo@xmath1 than the pd - derived sister compound . in pdcoo@xmath1 , it has proved controversial whether the high conductivity and low effective masses are mediated by significant pd @xmath61@xmath4 orbital mixing [ , , , , ] . to specifically address this issue ,
we have performed our density functional calculations using a local - orbital rather than plane - wave basis ( see methods ) , which allows a well - justified assignment of orbital characters .
we find that the density of states at the fermi level originates predominantly from pt states , with almost 90% pt - related character .
approximately 85% of these states belong to the pt 5@xmath4 orbitals with only a small admixture of pt @xmath62 and pt 6@xmath63 ( fig .
[ figdos](b ) ) . due to the localized nature of the co @xmath64 shell ,
the strong coulomb repulsion for these states was taken into account explicitly ( see fig .
[ figdos ] ( a ) and methods ) .
this suppresses an unphysical hybridization of the co-@xmath64 states with the pt-@xmath0 states near the fermi level which cause a pronounced rounding of alternating fermi surface corners in contrast to the experiment .
this suppression occurs for @xmath65 ev , after which the fermi surface topology and its orbital character are insensitive to the value of @xmath66 used over our investigated range of up to 6 ev .
the pt @xmath4-orbital derived nature of the fermi surface found here is entirely in keeping with its faceted nature that we observe , while the increase of electronic bandwidth compared to pdcoo@xmath1 can naturally be understood from the spatially more extended nature of pt @xmath0 than pd @xmath67 orbitals .
remarkably , the effective masses we observe here are a factor of 2 - 3 smaller than even those of the strongly
_ s_-_d_-hybridized electron fermi surface sheets in pt metal [ ] . from the bulk fermi surface , we can straightforwardly convert our resistivity measurements shown in fig .
[ f2 ] to a temperature - dependent mean free path , which we find rises from 700 at room temperature to as high as 4 @xmath14 m at 16 k. this is qualitatively consistent with the sharp spectral features we observe in arpes .
we observe narrow line widths of mdcs at the fermi level of better than 0.04 @xmath44 ( fig .
[ f5](a ) ) , very similar to those reported previously for pdcoo@xmath1 [ ] and comparable to our experimental angular resolution .
these remain clearly resolved to high binding energies , with @xmath68 @xmath44 increase in linewidth over the first 350 mev binding energy range .
this indicates a striking resilience of the quasiparticle lifetime away from the fermi level , suggesting weak electron - electron scattering in ptcoo@xmath1 .
this is in contrast to other well - known fermi - liquid oxide metals such as sr@xmath1ruo@xmath69 [ ] and sr@xmath1rho@xmath69 [ ] , which host sharp spectral features at low energy , but rapidly develop substantial broadening away from @xmath37 . additionally , those systems host moderate correlation - induced enhancements of the effective mass .
the lack of both of these features here points to a particularly weak influence of electron - electron interactions on the electronic structure of ptcoo@xmath1 .
this is not the situation in all quasi two - dimensional @xmath0 transition - metal oxides . in the layered iridates , for example ,
electronic correlations are still strong enough to drive a mott - like transition to an insulating state once structural distortions and strong spin - orbit coupling collectively narrow the electronic bandwidth [ , , ] .
in contrast , itineracy is preserved in ptcoo@xmath1 by the isolated single - band nature of the fermi surface . in the absence of band - folding arising from structural distortions , not present in ptcoo@xmath1
, this band remains relatively unaffected by soc , while large hybridization gaps are only opened away from @xmath37 where band crossings between different orbital character states occur ( see , e.g. , around the t point in fig . [ figdos](a ) where a @xmath70 degeneracy is lifted by approximately 600 mev ) .
thus the bandwidth of the state crossing @xmath37 in ptcoo@xmath1 remains large , away from the narrow - band @xmath71 limit of the iridates .
this naturally renders the electronic correlations expected of a @xmath0 system insufficient to drive a mott transition . nonetheless , the extremely good metallicity and lack of spectral broadening away from @xmath37 point not only to large bare bandwidths , but also to suppressed electron - electron interactions compared to other wide - band transition - metal oxides .
indeed , the fact that the fermi velocities shown in fig . [ f5](b ) are so close to the free electron values is itself remarkable , because the highly faceted fermi surface demonstrates the influence of the lattice potential on the states near @xmath37 .
nonetheless , our measurements firmly establish ptcoo@xmath1 as a model half - filled nearly - free - electron oxide metal , with high - quality crystals rendering this system an ideal test - bed for future studies of the nature of electronic correlations and the origins of extremely good metallicity in both pt- and pd - based delafossite oxides .
* single - crystal growth : * powders of reagent grade ptcl@xmath1 ( @xmath72 purity , alfa aesar ) and coo ( @xmath73 purity , alfa aesar ) were ground together for approximately an hour under an inert atmosphere in accordance with the chemical reaction ptcl@xmath1 + 2coo@xmath74 ptcoo@xmath1 + cocl@xmath1 .
the mixed powder was then sealed in a quartz tube under a vacuum of @xmath75 torr .
the sealed quartz tube was heated in a vertical furnace to 800 @xmath76c for 5 hours and cooled down to 740 @xmath76c at a rate of 7.5 @xmath76c / hour and kept at this temperature for 30 hours .
finally , the furnace was cooled from 740 @xmath76c to room temperature at a rate of 90 @xmath76c / hour .
* transport : * transport measurements were performed using standard four - probe a.c . techniques in @xmath77he cryostats ( quantum design ) , with measurement frequencies in the range 50200 hz , magnetic fields of up to 14 t and the use of single - axis rotators .
the hall effect was studied using reversed - field sweeps in the range @xmath78 at a series of fixed temperatures .
* angle - resolved photoemission : * arpes measurements were performed using the i05 beamline of diamond light source , uk .
samples were cleaved _ in - situ _ at the measurement temperature of @xmath79 k. measurements shown in fig . [ f3 ] and fig .
[ f5 ] were performed using 110 ev and 118 ev @xmath63-polarised light , respectively .
a scienta r4000 hemispherical analyser was used for all the measurements .
* density - functional theory calculations : * relativistic density functional ( dft ) electronic structure calculations were performed using the full - potential fplo code [ ] , version fplo14.00 - 47 . for the exchange - correlation potential , within the general gradient approximation ( gga ) ,
the parametrization of perdew - burke - ernzerhof [ ] was chosen .
the spin - orbit coupling was treated non - perturbatively solving the four component kohn - sham - dirac equation [ ] . to obtain precise band structure and fermi surface information , the calculations were carried out on a well converged mesh of 27000 @xmath80-points ( 30x30x30 mesh , 2496 points in the irreducible wedge of the brillouin zone ) . the strong coulomb repulsion in the co-3d shell
was taken into account in a mean field way applying the gga+@xmath81 approximation [ ] in the atomic - limit - flavor ( al ) .
for all calculations , the experimental lattice parameters have been used .
we thank scope - m and philippe gasser at eth zurich for supporting the fib work . this work was supported by the engineering and physical sciences research council ,
uk ( grant number ep / i031014/1 ) .
pdck acknowledges support from the royal society through a university research fellowship .
we thank diamond light source for access to beamline i05 that contributed to the results presented here .
+ + competing interests : the authors declare that they have no competing interests . + + data underpinning this publication can be accessed at http://dx.doi.org/10.17630/7a88d794-d747-4aa2-a099-9335c5b702ec + + author contributions : pk and mps performed the crystal growth ; vs , pk , pjwm , lb , jmr , nn , fa , eh , and pdck performed the experimental measurements ; hr performed the theoretical calculations ; tkk and mh built and maintained the arpes end station and provided experimental support ; apm and pdck were responsible for overall project planning and direction and wrote the paper with input and discussion from all co - authors .
s1 : field - dependent hall effect measurements + fig . s2 : additional density - of - states calculations 10 url # 1`#1`urlprefix[2]#2 [ 2][]#2 r. d. shannon , d. b. rogers , and c. t. prewitt , chemistry of noble metal oxides .
i. syntheses and properties of abo@xmath1 delafossite compounds .
chem . _ * 10 * , 713 ( 1971 ) . c. t. prewitt , r. d. shannon , and d. b. rogers , chemistry of noble metal oxides .
ii . crystal structures of platinum cobalt dioxide , palladium cobalt dioxide , coppper iron dioxide , and silver iron dioxide .
chem . _ * 10 * , 719 ( 1971 ) .
d. b. rogers , r. d. shannon , c. t. prewitt , and j. l. gillson , chemistry of noble metal oxides .
iii . electrical transport properties and crystal chemistry of abo@xmath1 compounds with the delafossite structure . _
chem . _ * 10 * , 723 ( 1971 ) .
m. a. marquardt , n. a. ashmore , and d. p. cann , crystal chemistry and electrical properties of the delafossite structure .
_ thin solid films _ * 496 * , 146 ( 2006 ) .
d. j. singh , electronic and thermoelectric properties of cucoo@xmath1 : density functional calculations .
b _ * 76 * , 085110 ( 2007 ) .
f. wang and a. vishwanath , spin phonon induced collinear order and magnetization plateaus in triangular and kagome antiferromagnets : applications to cufeo@xmath1 .
lett . _ * 100 * , 077201 ( 2008 ) .
h. yanagi , t. hase , s. ibuki , k. ueda , and h. hosono , bipolarity in electrical conduction of transparent oxide semiconductor cuino@xmath1 with delafossite structure .
lett . _ * 78 * , 1583 ( 2001 ) , and references therein .
h. takatsu , h. yoshizawa , s. yonezawa and y. maeno , critical behavior of the metallic triangular - lattice heisenberg antiferromagnet pdcro@xmath1 .
b _ * 79 * , 104424 ( 2009 ) .
h. takatsu , s. yonezawa , s. fujimoto , and y. maeno , unconventional anomalous hall effect in the metallic triangular - lattice magnet pdcro@xmath1 .
rev . lett_. * 105 * , 137201 ( 2010 ) . j. m. ok , y. j. jo , k. kim , t. shishidou , e. s. choi , h. j. noh , t. oguchi , b. i. min , and j. s. kim , quantum oscillations of the metallic triangular - lattice antiferromagnet pdcro@xmath1 .
lett . _ * 111 * , 176405 ( 2014 ) .
h. takatsu , g. nenert , h. kadowaki , h. yoshizawa , m. enderle , s. yonezawa , y. maeno , j. kim , n. tsuji , m. takata , y. zhao , m. green and c. broholm , magnetic structure of the conductive triangular - lattice antiferromagnet pdcro@xmath1 .
b _ * 89 * , 104408 ( 2014 )
. arxiv:1504.08104v1 quantum oscillations and magnetic reconstruction in the delafossite pdcro@xmath1 .
clifford w. hicks , alexandra s. gibbs , lishan zhao , pallavi kushwaha , horst borrmann , andrew p. mackenzie , hiroshi takatsu , shingo yonezawa , yoshiteru maeno , and edward a. yelland .
a. i. coldea , l. seabra , a. mccollam , a. carrington , l. malone , a. f. bangura , d. vignolles , p. g. van rhee , r. d. mcdonald , t. sorgel , m. jansen , n. shannon , and r. coldea , cascade of field - induced magnetic transitions in a frustrated antiferromagnetic metal .
b _ * 90 * , 020401 ( 2014 ) .
v. eyert , r. frsard , and a. maignan , on the metallic conductivity of the delafossites pdcoo@xmath1 and ptcoo@xmath1 .
mater . _ * 20 * , 2370 ( 2008 ) .
r. seshadri , c. felser , k. thieme , and w. tremel , metal - metal bonding and metallic behavior in some abo@xmath1 delafossites .
mater . _ * 10 * , 2189 ( 1998 ) .
k. kim , h. c. choi , and b. i. min , fermi surface and surface electronic structure of delafossite pdcoo@xmath1 .
b _ * 80 * , 035116 ( 2009 ) .
noh , j. jeong , j. jeong , h. sung , k. j. park , j .- y .
kim , s. b. kim , k. kim , and b. i. min , orbital character of the conduction band of delafossite pdcoo@xmath1 studied by polarization - dependent soft x - ray absorption spectroscopy .
b _ * 80 * , 073104 ( 2009 ) . c. w. hicks , a. s. gibbs , a. p. mackenzie , h. takatsu , y. maeno and e. a. yelland , quantum oscillations and high carrier mobility in the delafossite pdcoo@xmath1 .
lett . _ * 109 * , 116401 ( 2012 ) .
h. takatsu , j. j. ishikawa , s. yonezawa , h. yoshino , t. shishidou , t. oguchi , k. murata , y. maeno , extremely large magnetoresistance in the nonmagnetic metal pdcoo@xmath1 .
lett . _ * 111 * , 056601 ( 2013 ) .
arxiv:1412.5168 realization of the axial anomaly in a quasi - two - dimensional metal n. kikugawa , p. goswami , a. kiswandhi , e. s. choi , d. graf , r. e. baumbach , j. s. brooks , k. sugii , y. iida , m. nishio , s. uji , t. terashima , p. m. c. rourke , n. e. hussey , h. takatsu , s. yonezawa , y. maeno , l. balicas arxiv:1503.02069 axial anomaly and longitudinal magnetoresistance of a generic three dimensional metal pallab goswami , j. h. pixley , s. das sarma r. d. shannon , electrically conductive platinum cobalt oxides . _ us patent _ 3,514,414 ( 1970 ) .
m. tanaka , m. hasegawa , h. takei , crystal growth of pdcoo@xmath1 , ptcoo@xmath1 and their solid - solution with delafossite structure .
. growth _ * 173 * , 440 ( 1997 ) .
h. takatsu , s. yonezawa , s. mouri , s. nakatsuji , k. tanaka , and y. maeno , roles of high - frequency optical phonons in the physical properties of the conductive delafossite pdcoo@xmath1 .
_ j. phys .
_ , * 76 * , 104701 ( 2007 ) .
pallavi kushwaha , c. geibel , m. p. schmidt , a. p. mackenzie et al . , to be published .
p. j. w. moll , r. puzniak , f. balakirev , k. rogacki , j. karpinski , n.d .
zhigadlo and b. batlogg , high magnetic - field scales and critical currents in smfeas(o , f ) crystals . _ nature materials _ * 9 * , 628 ( 2010 ) .
t. pearsall and c. lee , electronic transport in reo@xmath29 : dc conductivity and hall effect .
b _ * 10 * , 2190 ( 1974 )
. w. d. ryden , a. w. lawson , and c. c. sartain , temperature dependence of the resistivity of ruo@xmath1 and iro@xmath1 .
_ , * 26a * , 209 ( 1968 ) .
c. m. hurd , the hall effect in metals and alloys , springer us ; boston , ma ( 1972 ) .
j. m. riley , f. mazzola , m. dendzik , m. michiardi , t. takayama , l. bawden , c. granerd , m. leandersson , t. balasubramanian , m. hoesch , t. k. kim , h. takagi , w. meevasana , ph .
hofmann , m. s. bahramy , j. w. wells , and p. d. c. king , direct observation of spin - polarized bulk bands in an inversion - symmetric semiconductor .
_ nature phys . _
* 10 * , 835 ( 2014 ) . h. -j .
noh , j. jeong , j. jeong , e. -j .
cho , s. b. kim , kyoo kim , b. i. min , and h. -d .
kim , anisotropic electric conductivity of delafossite pdcoo@xmath1 studied by angle - resolved photoemission spectroscopy .
lett . _ * 102 * , 256404 ( 2009 ) .
k. p. ong , d. j. singh , and p. wu , unusual transport and strongly anisotropic thermopower in ptcoo@xmath1 and pdcoo@xmath1 .
lett . _ * 104 * , 176601 ( 2001 ) . c. t. prewitt , r. d. shannon , and d. b. rogers , chemistry of noble metal oxides .
crystal structures of ptcoo@xmath1 , pdcoo@xmath1 , cufeo@xmath1 , and agfeo@xmath1
. _ inorganic chemistry _ * 10 * , 719 ( 1971 ) .
f. arnold , e. hassinger et al .
, to be published .
j. a. sobota , k. kim , h. takatsu , m. hashimoto , s .- k .
mo , z. hussain , t. oguchi , t. shishidou , y. maeno , b. i. min , and z .- x .
shen , electronic structure of the metallic antiferromagnet pdcro@xmath1 measured by angle - resolved photoemission spectroscopy .
b _ * 88 * , 125109 ( 2013 ) . h. -j .
noh , j. jeong , b. chang , d. jeong , h. s. moon , e. -j .
cho , j. m. ok , j. s. kim , k. kim , b. i. min , h .- k .
lee , j .- y .
park , h. -d .
kim , and s. lee , direct observation of localized spin antiferromagnetic transition in pdcro@xmath1 by angle - resolved photoemission spectroscopy . _ sci .
report _ * 4 * , 3680 ( 2014 ) .
a. de la torre , e. c. hunter , a. subedi , s. mckeown walker , a. tamai , t. k. kim , m. hoesch , r. s. perry , a. georges , and f. baumberger , coherent quasiparticles with a small fermi surface in lightly doped sr@xmath29ir@xmath1o@xmath82 .
lett . _ * 113 * , 256402 ( 2014 ) .
y. f. nie , p. d. c. king , c. h. kim , m. uchida , h. i. wei , b.d .
faeth , j.p .
ruf , j.p.c .
ruff , l. xie , x. pan , c.j .
fennie , d.g .
schlom , and k. m. shen , interplay of spin - orbit interactions , dimensionality , and octahedral rotations in semimetallic sriro@xmath29 .
* 114 * , 016401 ( 2015 ) .
masayuki tanakaa , masashi hasegawaa , thoru higuchib , takeyo tsukamotob , yasuhisa tezukaa , shik shina , humihiko takeic , origin of the metallic conductivity in pdcoo@xmath1 with delafossite structure .
_ physica b _ * 245 * , 157 ( 1998 ) .
j. b. ketterson , and l. r. windmiller , de haas - van alphen effect in platinum .
b _ , * 2 * , 4813 ( 1970 ) .
a. damascelli , d. h. lu , k. m. shen , n. p. armitage , f. ronning , d. l. feng , c. kim , z .- x .
shen , t. kimura , y. tokura , z. q. mao , and y. maeno , fermi surface , surface states , and surface reconstruction in sr@xmath1ruo@xmath69 .
lett . _ * 85 * , 5194 ( 2000 ) .
f. baumberger , n. j. c. ingle , w. meevasana , k. m. shen , d. h. lu , r. s. perry , a. p. mackenzie , z. hussain , d. j. singh , and z .- x .
shen , fermi surface and quasiparticle excitations of sr@xmath1rho@xmath69 .
lett . _ * 96 * , 246402 ( 2006 ) .
b. j. kim , hosub jin , s. j. moon , j .- y .
park , c. s. leem , jaejun yu , t. w. noh , c. kim , s .-
oh , j .- h . park , v. durairaj , g. cao , and e. rotenberg , novel @xmath83=1/2 mott state induced by relativistic spin - orbit coupling in sr@xmath1iro@xmath69 .
_ , * 101 * , 076402 ( 2008 ) .
p. d. c. king , t. takayama , a. tamai , e. rozbicki , s. mckeown walker , m. shi , l. patthey , r. g. moore , d. lu , k. m. shen , h. takagi , and f. baumberger , spectroscopic indications of polaronic behavior of the strong spin - orbit insulator sr@xmath29ir@xmath1o@xmath82 .
b _ , * 87 * , 241106(r ) ( 2013 ) . k. koepernik , and h. eschrig , full - potential nonorthogonal local - orbital minimum - basis band - structure scheme .
b _ , * 59 * , 1743 ( 1999 ) ; i. opahle , k. koepernik , and h. eschrig , full - potential band - structure calculation of iron pyrite .
b _ , * 60 * , 14035 ( 1999 ) ; http://www.fplo.de j.p .
perdew , k. burke and m. ernzerhof , generalized gradient approximation made simple .
_ , * 77 * , 3865 ( 1996 ) .
h. eschrig , m. richter and i. opahle , relativistic solid state calculations , in : `` relativistic electronic structure theory , ( part ii , applications ) '' ( ed .
p. schwerdtfeger ) , theoretical and computational chemistry , vol.13 , elsevier , 723 , ( 2004 ) h. eschrig , k. koepernik , and i. chaplygin , density functional application to strongly correlated electron systems .
_ j. solid state chemistry _ , * 176 * , 482 ( 2003 ) n. p. ong , geometric interpretation of the weak - field hall conductivity in two - dimensional metals with arbitrary fermi surface .
b _ * 43 * , 193 ( 1991 ) . * supplementary material : nearly - free electrons in a @xmath0 delafossite oxide metal *
field dependence of hall resistivity ( @xmath84 ) for ptcoo@xmath1 for different labeled temperatures for the @xmath85 m thick sample .
the inset highlights the two regions seen as a function of applied magnetic field , using data taken at 80 k as an example.,width=10 ] the hall coefficient shown in the inset of fig . 2 of the main text were obtained from field sweeps performed at a number of temperatures are shown in fig .
we assume that the hall effect in non - magnetic ptcoo@xmath1 or pdcoo@xmath1 has no anomalous component [ 9 ] , and that the signal is dominated by the orbital hall effect .
there is a clear , temperature - dependent separation into low- and high - field regimes with different gradients of the hall resistivity @xmath84 .
below 20 k , the data are dominated by the high - field regime in which the fermi surface volume alone is expected to determine the value of the hall coefficient [ 29 ] . the value in that region is approximately 2.9@xmath86 m@xmath32/c , close to the value that we measure for pdcoo@xmath1 under the same conditions ( data not shown ) , and within @xmath87 of the expectation for a single band containing one electron per pd . our measured mean - free path ( see main text ) enables us to make a quantitative estimate of @xmath88r@xmath89 = @xmath90@xmath91 at each temperature and field ( here the cyclotron radius @xmath92 = @xmath93@xmath94 , the cyclotron frequency @xmath95= @xmath96 = @xmath97@xmath93@xmath98 and @xmath99 = @xmath100@xmath101 where @xmath101 is the relaxation time ) . in fields of 9 t
, we find @xmath88r@xmath89 @xmath102 5 at 15 k , which drops to @xmath88r@xmath89 = 0.1 at 300 k. this indicates , therefore , that the experiment goes well beyond the weak field regime by 9 t at low temperatures , but that weak - field physics applies at all fields at high temperatures .
accordingly , we observe a cross - over from the low - field to the high - field regime ( @xmath88r@xmath89 @xmath103 1 ) at low temperatures , while at elevated temperatures fields in excess of 9 t are required to observe the high - field limit .
we attribute this crossover between weak- and strong - field physics as the origin of two different gradients at intermediate temperatures ( see for example the inset to fig . [ sf1 ] ) .
the hall coefficient data shown in inset to fig . 2 of the main paper were obtained by fitting the gradient of @xmath84 between 7 and 9 t. this ensures that below approximately 50 k , r@xmath104 is obtained from data in the high filed limit that is more directly related to fermi surface volume that data from weak or intermediate fields .
we attribute the temperature dependence seen above 50 k to the crossover to those less reliable regimes in which details of _ k_-dependent scattering strongly effect r@xmath104 [ 49 ] .
supplemental fig . [ sf2 ] shows the dependence of the calculated density of states on the coulomb repulsion @xmath66 in the co 3d shell and the oxygen position within the crystal structure . explicitly taking into account the strong coulomb repulsion @xmath66 suppresses unphysical hybridization of the co-@xmath64 states with the pt-@xmath0 states crossing the fermi level but otherwise has a minor influence on the dos in the vicinity of @xmath37 .
in contrast , it has a significant influence at higher binding energy where the valence bands have significant co orbital character as shown in fig .
4(b ) of the main text .
varying the o - position ( 0,0,@xmath38 ) induces a shift of the energy of a hole band whose band top is slightly below @xmath37 , giving rise to a shift of a step in the density of states visible in the inset of fig .
[ sf2](b ) . in the absence of spin - orbit interactions
, this would cause the hole band to move above @xmath37 for pt - o distances at and above their calculated relaxed value ( @xmath105 ; the experimental value is z=0.114 ) causing a second fermi surface to appear .
however , this is pushed below @xmath37 for any realistic value of @xmath38 when including spin - orbit coupling in the calculation , as observed in fig .
[ sf2](b ) . calculated density of states ( dos ) .
dependence of the dos on ( a ) the coulomb repulsion @xmath66 in the co 3d shell for @xmath106 and ( b ) the o - position ( 0,0,@xmath38 ) for @xmath107 ev .
the insets show magnified views near the fermi energy , @xmath37 , and all calculations include spin - orbit coupling.,width=12 ] | understanding the role of electron correlations in strong spin - orbit transition - metal oxides is key to the realisation of numerous exotic phases including spin - orbit assisted mott insulators , correlated topological solids , and prospective new high - temperature superconductors . to date
, most attention has been focussed on the @xmath0 iridium - based oxides . here , we instead consider the pt - based delafossite oxide ptcoo@xmath1 .
our transport measurements , performed on single - crystal samples etched to well - defined geometries using focussed ion - beam techniques , yield a room - temperature resistivity of only 2.1 @xmath2 cm , establishing ptcoo@xmath1 as the most conductive oxide known . from angle - resolved photoemission and density - functional theory ,
we show that the underlying fermi surface is a single cylinder of nearly hexagonal cross - section , with very weak dispersion along k@xmath3 . despite being predominantly composed of @xmath4-orbital character ,
the conduction band is remarkably steep , with an average effective mass of only 1.14@xmath5 .
moreover , the sharp spectral features observed in photoemission remain well - defined with little additional broadening for over 500 mev below e@xmath6 , pointing to suppressed electron - electron scattering .
together , our findings establish ptcoo@xmath1 as a model nearly - free electron system in a @xmath0 delafossite transition - metal oxide . |
following the successful runi from 1992 to 1996 , the cdf detector has undergone a major upgrade @xcite for the runii which begun in march 2001 .
the approval for the addition of a time - of - flight detector was granted in january 1999 .
the installation of the tof detector was completed in august 2001 and its data has been included in the cdfii readout since then . the primary physics motivation for tof is to complement and enhance the particle identification capability provided by the central drift chamber ( cot ) since it distinguishes @xmath1 and @xmath2 in the momentum region of their cross - over in @xmath16 . with an expected time - of - flight resolution of @xmath0 ,
the tof system will be capable of identifying charged kaons from pions by their flight time difference with at least two standard deviation separation up to kaon momenta of @xmath17 .
such an addition results in an enhancement of the @xmath18 flavor identification power , crucial to improve the statistical precision in cp violation and @xmath19 mixing measurements .
particle identification with tof is performed by measuring the time of arrival of a particle at the scintillator with respect to the collision time @xmath20 .
the particle mass @xmath21 can then be determined from the momentum @xmath22 , the path - length @xmath23 , measured by the tracking system , and the time - of - flight @xmath24 measured by the tof : @xmath25 figure [ fig1 ] shows the time - of - flight difference between @xmath26 , @xmath27 and @xmath28 and the separation power assuming a resolution of @xmath0 .
the tof improves the @xmath26 separation in the momentum region of @xmath29 by @xmath30 or better .
the cdfii tof detector @xcite,@xcite,@xcite consists of 216 bars ( @xmath31 long and with a cross section of about @xmath32 ) of bicron bc-408 scintillator .
the bars are installed at a radius of @xmath33 from the beam line in the @xmath34 of radial space available between the cylindrical outer wall of the central drift chamber ( cot ) and the inner wall of the cryostat of the super - conducting solenoid , covering a pseudo - rapidity region of roughly @xmath35 .
a nineteen - stage fine mesh photomultiplier tube ( pmt ) , hamamatsu r7761 , with a diameter of @xmath36 inches is attached to each bar end , for a total of @xmath37 pmts in the system .
these operate in the @xmath38 solenoidal magnetic field with an average gain reduction of @xmath39 from the nominal gain of @xmath40 .
the optical connection between the pmt and the scintillator is provided by a compound parabolic concentrator .
the pmt in turn is directly attached to a custom designed hv divider base which also connects to a preamplifier card for the readout .
differential signal from the anode and the last dynode of the pmt is fed to the preamplifier , and drives , over @xmath41 of shielded twisted pair cable , front - end electronics that reside in a vme crate mounted outside the detector .
the front - end signal is split into two : one for timing measurement and the other for pulse height measurement .
the timing path enters a leading edge discriminator with an adjustable threshold , whose output serves as the start signal for the time - to - amplitude conversion ( tac ) circuit @xcite .
the tac ramp is terminated by a common stop clock edge , that is synchronized with the @xmath42 bunch crossing , and fanned out to all electronics channels with a design jitter less than @xmath43 .
the voltage output from the tac is sampled by a 12-bit adc , characterized by a least count of roughly @xmath44 over a dynamic range of about @xmath45 .
tac response has shown an excellent stability since the commissioning of the electronics , with a residual variation , after the calibrations , of less than @xmath44 . the primary purpose for measuring the charge of pmt signals is to perform a correction for the dependence of the discriminator on signal amplitude .
the charge measurement is done by a charge sensitive adc @xcite .
the current driver is switched on by a gate of adjustable width , and is initiated by the discriminator output , so that only the charge due to the pulse that fired the discriminator is integrated .
the time at which a pmt pulse is registered , measured with respect to the nominal @xmath42 bunch crossing time , can be modeled by the following expression : @xmath46 where @xmath47 is a constant offset describing the propagation delays in the cables , @xmath20 represents the time at which the @xmath42 interaction occurred , @xmath48 is the time - of - flight of the particle impinging on the bar of scintillator of length @xmath23 , @xmath49 is the effective speed of light in the scintillator , and the last term describes the time - walk effect introduced by the leading edge discriminator .
the @xmath50 coordinate is measured along the length of the bar by the tracking chamber , the positive and negative sign of the term proportional to @xmath50 corresponds to a pmt in the west or east end of the detector , respectively .
the measurement of the parameters in eq .
[ eq2 ] for each channel is an essential step toward reaching the goal of @xmath0 resolution , and it is performed regularly during periods of data taking . when a single charged particle enters a bar , the time difference between signals reaching the east and west ends is essentially a linear function of the track entrance point along the bar , @xmath50 .
a typical distribution of the time difference plotted as a function of @xmath50 is shown in figure [ fig2 ] .
the effective speed of light in the bar can be derived from the fitted slope . from the distribution of the residuals of this fit a first indication of the timing resolution of each pmt , can be obtained .
the resolution averaged over all bars is typically @xmath51 or better .
this is larger by approximately a factor two than the resolution of a mean time measurement , indicating that the tof resolution is not far from our design goals .
it is worth notice here that as systematics effects can cancel in the time difference measurement , this estimate represents only an indication of the capabilities of the detector .
tracks which hit the scintillator at the same @xmath50 position are found to have an adc response that is well described by a landau distribution .
the width of the landau peak divided by the peak position is typically @xmath52 for all channels .
the dependence of the pulse height on the measured time and the parameterization of the time slewing effect have been studied using a sample where each track passes through two adjacent bars of scintillator .
depending on the path length in each bar , a range of adc responses is obtained on the two channels at the same end of the bar . at the same time
because the track entrance point in each bar is similar , the time difference between the two channels depends mainly on time walk effects . due to finite attenuation length in the bars ( @xmath53 ) the pulse height depends on the entrance point of the track , @xmath50 . for this reason the slewing correction introduces a linear dependence in the time difference @xmath54 versus @xmath50 , resulting in a biased effective speed of light determination . after the time slewing correction the width of the speed of light distribution for all the pmts is significantly reduced and observed low - side tail in the distribution disappeared .
after all calibrations and corrections were applied , the timing resolution of a given channel can be estimated by comparing the measured tof for a track with the expected tof under the pion mass hypothesis .
figure [ fig5 ] shows the distribution of the resolution obtained with this procedure .
it shows a mean value of @xmath55 with an rms of @xmath56 , which is close to the design goals . in figure [ fig6 ]
is shown the mass - momentum scatter plot computed using eq .
[ eq1 ] for low momentum tracks . three clear clouds corresponding to pions , kaons and protons can be seen .
actual realistic performance of the entire tof system can be obtained by analyzing the @xmath57 signal in a sample of events recorded by cdfii .
the @xmath58 candidates are selected from all pairs of oppositely charged tracks by assuming the kaon mass , with the only constraint that the two tracks originated from the same vertex .
both tracks are required to have a transverse momentum less than @xmath59 and valid reconstructed tof informations .
figure [ fig7 ] shows the invariant mass distributions obtained before and after applying tof kaons particle identification .
a clear signal peak in the distribution appears in the latter case , corresponding to a factor @xmath60 background reduction and @xmath61 signal efficiency .
the primary purpose of the tof detector is to identify charged pions , kaons and protons . in the context of @xmath62 physics , particle identification is useful both for reconstructing @xmath62 decays and for @xmath18 flavor tagging . in particular
the cdf tof detector was designed to substantially improve the statistical precision in the measurement of cp violation , as well as the sensitivity in measuring @xmath63 oscillations .
+ by the term `` @xmath18 flavor tagging '' we mean the ensemble of procedures and algorithms which act to determine the flavor of the @xmath62 meson when it was produced .
the figure of merit for a @xmath18 flavor tagging method is its `` total tagging effectiveness '' , @xmath64 .
the efficiency @xmath65 accounts for the fact that we can apply a flavor tag only to a fraction @xmath66 of the @xmath62 candidates .
the dilution @xmath67 is related to the fact that sometime the obtained tag is wrong .
if we call @xmath68 the probability that the tag is correct , the dilution is then defined as @xmath69 , so for a perfect tag we will have @xmath70 , and for a random tag @xmath71 . the total tagging effectiveness @xmath64 determines the effective statistical power of the sample , so that , for example , the statistical error on the determination of a cp asymmetry @xmath72 with @xmath73 total candidates is proportional to @xmath74 .
an additional flavor tagging method that would increases @xmath64 by a factor of two would lead to a reduced statistical error on @xmath72 by @xmath75 .
the methods of @xmath18 flavor tagging can be divided in two categories : `` opposite - side '' tagging ( ost ) @xcite and the `` same - side '' tagging ( sst ) @xcite . the former relays on the fact that the dominant production mechanism of @xmath18 quarks in hadron collisions generates @xmath76 pairs . to identify the flavor of a reconstructed @xmath62 meson of interest
, we identify the flavor of the second @xmath62 hadron in the event and infer the flavor at production of the primary one .
the same - side flavor tag exploits instead the correlation between @xmath18 flavor and the charge of the particles accompanying the reconstructed @xmath62 .
in particular , a @xmath77 is expected to be produced in association with a @xmath78 meson , and a @xmath79 with a @xmath80 .
the cdfii tof can be used to enhance both opposite - side and same - side @xmath18 flavor tagging .
the decays of @xmath81 hadrons will contain more likely a @xmath82 than a @xmath83 in the final state due to @xmath84 week decays . therefore identifying a charged kaon from the @xmath62-decay
can be used as an opposite - side @xmath18 flavor tag . in figure [ fig8 ] the montecarlo momentum spectrum of kaons from the opposite - side @xmath62 decay to the @xmath85 signal
more than @xmath86 of the available kaons are in a momentum range where the cdfii tof can give a @xmath30 separation power from charged pions .
the cdfii potential for the opposite - side flavor kaon tag using the tof has been estimated with a montecarlo simulation , assuming the design resolution of @xmath0 .
this gave @xmath87 , which is more than a factor two higher than the typical values obtained in runi @xcite , @xcite .
the tof can be used for same - side tagging as well . in the hadronization process when a @xmath63 meson is produced an @xmath88 pair must be popped from the vacuum during fragmentation . the remaining @xmath49 or @xmath89 quark can join a @xmath90 or @xmath91 quark to form a charged kaon .
the charge of the kaon thus depends on the flavor of the @xmath63 meson at production .
a factor four increase on @xmath64 has been obtained , using montecarlo simulations , for same - side tagging of @xmath63 using tof informations ( from @xmath92 to @xmath93 ) . combining all the @xmath18 flavor tag methods available in runii
, cdfii has estimated a total @xmath64 of @xmath94 for @xmath63 mesons using tof for particle identification @xcite .
this large effectiveness will greatly increase the sensitivity for the measurement of the @xmath4 mixing parameter @xmath95 , one of the unique measurement of tevatron during the runii . in figure [ fig9 ] the luminosity required to achieve a five standard deviation observation of mixing as a function of the mixing parameter @xmath96
the projection @xcite is based on montecarlo simulation of fully hadronic decays of the @xmath4 hadrons , and assumes a fully operational cdf detector and trigger system , work is in progress to determine the sensitivity using runii data . within the standard model various experimental results
indicate that @xmath97 at the @xmath98 confidence level @xcite . therefore ,
if mixing occurs as expected in the standard model , it should be observed even with a few hundreds of @xmath99 of total integrated luminosity .
the new cdfii time - of - flight detector become operational in august 2001 . since then
, it has working reliably without any significant problem .
the first round of calibrations is available and preliminary studies show that the @xmath0 time resolution goal seems reachable .
we expect substantial improvements using the tof in the `` b '' physics program of cdf in coming years .
r.blair et al .
, `` the cdfii detector : technical design report '' , fermilab - pub-96 - 390-e .
c.paus et al . , `` design and performances test of the cdf time - of - flight system '' , nim , vol .
a461 ( 2001 ) 579 - 581 .
c.grozis et al . , `` the time - of - flight detector at cdf '' , nucl.phys.proc.suppl . , vol .
47 ( 2001 ) 344 - 347 .
c.grozis et al . , `` a time - of - flight detector for cdf '' , int.j.mod.phys .
a16s1c ( 2001 ) 1119 - 1121 .
c.chen et al . ,
`` front end electronics for the cdfii time - of - flight system '' , in the proceedings of 2001 ieee nss / mic symposium .
cdf collaboration , f.abe et al .
, phys.rev .
d60 ( 1999 ) 072003 .
cdf collaboration , f.abe et al . ,
phys.rev.lett . 80 ( 1988 ) 2057 and phys.rev .
d59 ( 1999 ) 032001 .
k.anikeev et al . , `` b physics at tevatron : runii and beyond '' , fermilab pub-01/197 , hep - ph/0201071 .
f.parodi , p.rodeau , a.stocchi , nuovo cim .
a112 ( 1999 ) 833 . | the new cdfii detector incorporates a time - of - flight detector ( tof ) , employing plastic scintillator bars and fine - mesh photomultipliers .
since august 2001 the tof system has been fully instrumented and integrated into the cdfii data acquisition system . with a design goal of @xmath0 resolution
the tof system will provide at least two standard deviations separation between @xmath1 and @xmath2 for momenta @xmath3 , complementing low momentum particle identification by means of the specific ionization energy loss measured in the drift chamber .
we describe the design of the tof detector and discuss the current status of its calibration and initial performances .
finally we review the expected impact of the tof detector in the flavor tagging of neutral @xmath4 meson . * * ' '' '' the cdfii time - of - flight detector and impact on beauty flavor tagging + c.grozis@xmath5 , r.kephart@xmath5 , r.stanek@xmath5 , d.h.kim@xmath6 , m.s.kim@xmath6 , y.oh@xmath6 , y.k.kim@xmath7 , g.vermendi@xmath7 , k.anikeev@xmath8 , g.bauer@xmath8 , i.k.furic@xmath8 , a.korn@xmath8 , i.kravchenko@xmath8 , m.mulhearn@xmath8 , ch.paus@xmath8 , s.pavlon@xmath8 , k.sumorok@xmath8 , c.chen@xmath9 , m.jones@xmath9 , w.kononenko@xmath9 , j.kroll@xmath9 , g.m.mayers@xmath9 , m.newcomer@xmath9 , r.g.c.oldeman@xmath9 , d.usynin@xmath9 , r.van berg@xmath9 , g.bellettini@xmath10 , c.cerri@xmath10 , a.menzione@xmath10 , e.vataga@xmath10 , s.dececco@xmath11 , d.depedis@xmath11 , c.dionisi@xmath11 , a.digirolamo@xmath11 , m.rescigno@xmath11 , l.zanello@xmath11 , s.cabrera@xmath12 , j.fernandez@xmath12 , g.gomez@xmath12 , j.piedra@xmath12 , t.rodrigo@xmath12 , a.ruiz@xmath12 , i.vila@xmath12 , r.vilar@xmath12 , m.ahn@xmath13 , b.j.kim@xmath13 , s.b.kim@xmath13 , i.cho@xmath14 , j.lee@xmath14 , i.yu@xmath14 , h.kaneko@xmath15 , a.kazama@xmath15 , s.kim@xmath15 , k.sato@xmath15 , k.sato@xmath15 , f.ukegawa@xmath15 |
results from numerical studies suggest that the two - dimensional ( 2d ) hubbard model exhibits the basic phenomena which are seen in the high-@xmath6 cuprate materials . at half - filling ,
one finds a groundstate with long - range antiferromagnetic order @xcite .
doped away from half - filling there is a pseudogap regime @xcite , and at low temperature a striped phase @xcite as well as d - wave pairing @xcite .
in addition , the various phases appear delicately balanced with respect to changes in parameters .
all of these features remind one of the actual cuprate materials , so that it is of interest to understand the structure of the interaction that leads to @xmath7-wave pairing in this model . here
we will study this interaction for a 2d hubbard model @xcite on a square lattice which contains a one - electron near - neighbor hopping @xmath0 and an onsite coulomb interaction @xmath1 . in this case
the results depend upon just two parameters @xmath8 and the average site occupation @xmath9 .
previous work @xcite has shown that the pairing interaction with @xmath10 and @xmath11 increases with momentum transfer , has a matsubara frequency dependence similar to that of the @xmath12 spin susceptibility and is mediated by a particle - hole @xmath13 exchange channel . here
we extend this work to explore the pairing interaction for larger values of @xmath8 and various dopings .
we are particularly interested in the case in which @xmath1 is of order of the bandwidth @xmath5 . in this case , well developed upper and lower hubbard bands are seen in the single particle density of states @xmath14 versus @xmath15 for @xmath16 , @xmath17 and various values of the temperature @xmath18 for site fillings of a ) @xmath19 , b ) @xmath20.,title="fig:",width=336 ] versus @xmath15 for @xmath16 , @xmath17 and various values of the temperature @xmath18 for site fillings of a ) @xmath19 , b ) @xmath20.,title="fig:",width=336 ] fig .
[ fig : dos ] shows @xmath21 with @xmath22 for various fillings .
these results were obtained from a maximum entropy continuation@xcite of dynamic cluster quantum monte carlo data on a 16 site cluster . at half - filling , fig .
[ fig : dos]a , for @xmath23 one sees broad upper and lower hubbard bands .
then , as the temperature scale drops below the exchange energy scale @xmath24 , two additional structures appear above and below @xmath25 . as preuss _ et .
al _ @xcite first showed , the inner structures arise from the formation of two coherent bands , each of width @xmath26 that form as the antiferromagnetic correlations develop .
this characteristic mott - hubbard - antiferromagnetic four band structure @xcite is clearly seen when @xmath1 becomes of order or larger than the bandwidth . fig .
[ fig : dos]b shows @xmath21 for the doped @xmath27 case . here the chemical potential has moved down into the lower coherent band and the spectral weight in the upper coherent band has essentially vanished .
nevertheless , the remnants of the upper and lower hubbard bands remain .
part of the motivation for this study is to examine the structure of the pairing interaction in this parameter regime . as in our previous work
, the pairing interaction @xmath28 will be calculated using a dynamic cluster quantum monte carlo simulation @xcite . as shown in fig .
[ fig : bse ] , @xmath28 is the irreducible part of the 4-point particle - particle vertex in the zero center of mass and energy channel .
previously , we discussed how one could extract @xmath29 with @xmath30 using a dynamic cluster quantum monte carlo simulation . .10 in and the particle - particle irreducible vertex @xmath31 .
the solid lines are dressed single particle green s functions.,width=336 ] while we studied @xmath31 in our earlier work , here we will focus on the momentum and matsubara frequency dependence of the @xmath7-wave eigenfunction @xmath32 of the homogeneous particle - particle bethe - salpeter equation @xmath33 at low temperatures , @xmath32 has the largest eigenvalue and this eigenvalue goes to one at @xmath34 .
as @xmath18 approaches @xmath6 , the momentum and frequency dependence of @xmath32 reflect the structure of the pairing interaction at @xmath6 , just as the superconducting gap function reflects the @xmath35 and @xmath15 dependence of the pairing interaction in the superconducting state .
thus , while @xmath32 is not a quantity that is directly measurable , it has a @xmath35-dependence related to the momentum dependence of the interaction and a matsubara frequency dependence which decays beyond a characteristic frequency associated with the dynamic character of the interaction .
it also has the great advantage of depending upon one momentum and frequency variable as opposed to the multiple momentum and frequency variables of @xmath36 . in the following section [ sec : dynam - clust - appr ] we review the dynamic cluster approximation and
discuss how one calculates the @xmath7-wave eigenvalue @xmath37 and eigenfunction @xmath32 .
then , in sec . [ sec : struct - pair - inter ]
we investigate how @xmath37 depends upon @xmath1 and @xmath9 . following this
, we examine the @xmath35-dependence of @xmath32 and see how closely it follows the simple @xmath38 dependence .
if it were of this form over the entire brillouin zone , then it would imply a strictly near - neighbor pairing interaction .
then we turn to the @xmath39-dependence which reflects the dynamics of the pairing interaction and study its dependence on @xmath1 . in sec .
[ sec : separ - repr - pair ] , based upon the results for @xmath32 , we construct a simple separable representation of @xmath36 and discuss the strength of the pairing interaction .
[ sec : conclusion ] contains our conclusions .
the dynamical cluster approximation ( dca ) @xcite maps the bulk lattice to a finite size cluster embedded in a self - consistent bath designed to represent the remaining degrees of freedom .
short - range correlations within the cluster are treated explicitly , while the longer - ranged physics is described by a mean - field . by increasing the cluster size
, the dca systematically interpolates between the single - site dynamical mean - field result and the exact result , while remaining an approximation to the thermodynamic limit for finite cluster size .
the essential assumption is that short - range quantities , such as the single - particle self - energy @xmath40 , and its functional derivatives , the two - particle irreducible vertex functions , are well represented as diagrams constructed from a coarse - grained propagator @xmath41 . to define @xmath41 , the brillouin zone in two dimensions
is divided into @xmath42 cells of size @xmath43 . as illustrated in fig .
[ fig : dcabz ] , each cell is represented by the cluster momentum @xmath44 in its center .
the coarse - grained green function @xmath45 is then obtained from an average over the @xmath46 wave - vectors @xmath47 within the cell surrounding @xmath44 , @xmath48 here the self - energy for the bulk lattice @xmath49 has been approximated by the cluster self - energy @xmath50 .
cells each represented by a cluster momentum @xmath44 .
irreducible quantities such as the single - particle self - energy @xmath40 and two - particle irreducible vertex @xmath51 are constructed from coarse - grained propagators @xmath45 that are averaged over the momenta @xmath52 within the cell represented by @xmath44 .
the cluster momenta @xmath44 for a 4-site cluster are shown on the left and those for a 24-site cluster on the right.,title="fig:",width=168 ] cells each represented by a cluster momentum @xmath44 .
irreducible quantities such as the single - particle self - energy @xmath40 and two - particle irreducible vertex @xmath51 are constructed from coarse - grained propagators @xmath45 that are averaged over the momenta @xmath52 within the cell represented by @xmath44 .
the cluster momenta @xmath44 for a 4-site cluster are shown on the left and those for a 24-site cluster on the right.,title="fig:",width=168 ] consequently , the compact feynman diagrams constructed from @xmath53 collapse onto those of an effective cluster problem embedded in a host which accounts for the fluctuations arising from the hybridization between the cluster and the rest of the system . the non - interacting part of the effective cluster action is then defined by the cluster - excluded inverse green s function @xmath54 which accounts for the hybridization between the cluster and the host . given @xmath55 and the interaction on the cluster @xmath56 , one
can then set up a hirsch - fye quantum monte carlo algorithm @xcite to calculate the cluster green function and from it the cluster self - energy @xmath50 which is used in eq .
( [ eq : gbar ] ) to re - calculate the coarse - grained green function @xmath53 @xcite .
this process is then iterated to convergence .
since a determinantal monte carlo method is used , there is also a sign problem for the doped hubbard model .
however , the coupling of the cluster to the self - consistent host significantly reduce the sign problem so that lower temperatures can be reached @xcite .
the dca cluster one- and two - particle green s functions that we calculate have the standard finite temperature definitions [ five ] @xmath57 and @xmath58 here , @xmath59 , where @xmath60 denotes a site in the dca cluster , @xmath61 the imaginary time , @xmath62 is the usual @xmath63-ordering operator , and @xmath64 creates a particle on the cluster with spin @xmath65 .
fourier transforming on both the cluster space and imaginary time variables gives @xmath66 and @xmath67 with @xmath68 .
using @xmath69 and @xmath70 , one can extract the cluster four - point vertex @xmath51 from @xmath71\nonumber \\ & + & \frac{t}{n}\ \delta_{k_1+k_2 , k_3+k_4 } g_c(k_4)\ , g_c(k_3 ) \gamma\ , ( k_4 , k_3 ; k_2 , k_1)\nonumber \\ & \times & g_c(k_2)\ , g_c(k_1)\ , .\end{aligned}\ ] ] then , using @xmath72 and @xmath51 , one can determine the irreducible particle - particle vertex @xmath28 from the bethe - salpeter equation shown in fig .
[ fig : bse ] . using @xmath31 , the eigenvalues and eigenfunctions of the bethe - salpeter equation
may then be calculated from @xmath73 here , the sum over @xmath74 denotes a sum over both momentum @xmath75 and matsubara @xmath76 variables .
we decompose @xmath77 . by assumption ,
irreducible quantities like @xmath28 and @xmath78 do not depend on @xmath79 , allowing us to coarse - grain the green function legs , yielding an equation that depends only on coarse - grained and cluster quantities @xmath80 with @xmath81 . here , we show a number of results for the 4-site cluster shown on the left in fig . [
fig : dcabz ] , which allows us to investigate larger values of @xmath1 and lower temperatures than the 24-site cluster . as discussed in ref .
@xcite , the 4-site cluster does not allow for the effect of pairfield phase - fluctuations .
simulations on the 24-site cluster were used to determine the @xmath35 dependence of @xmath32 .
the temperature dependence of the d - wave eigenvalues @xmath37 calculated using a 4-site cluster for a site filling @xmath27 with @xmath10 , 8 and 12 are shown in fig .
[ fig : lambdad]a . fig .
[ fig : lambdad]b shows @xmath82 versus @xmath8 for @xmath83 .
eigenvalue @xmath37 versus @xmath84 for @xmath85 , @xmath5 and @xmath86 and @xmath20 .
( b ) the @xmath7 eigenvalue @xmath37 versus @xmath8 for @xmath83 and @xmath20.,title="fig:",width=336 ] eigenvalue @xmath37 versus @xmath84 for @xmath85 , @xmath5 and @xmath86 and @xmath20 .
( b ) the @xmath7 eigenvalue @xmath37 versus @xmath8 for @xmath83 and @xmath20.,title="fig:",width=336 ] here one sees that it is favorable to have a coulomb interaction strength @xmath1 of order the bandwidth @xmath5 .
this is consistent with the notion that it is important to have strong short - range antiferromagnetic correlations .
however , because the exchange interaction @xmath24 at strong coupling , the short - range antiferromagnetic correlations decrease when @xmath1 becomes large compared to the bandwidth .
the dependence of @xmath87 on the filling @xmath9 is illustrated in fig .
[ fig : lambdadn ] for @xmath88 .
what one sees is that @xmath87 increases as the system is doped towards half - filling .
however , at half - filling the dominant eigenvalue of the 4-point vertex occurs in the @xmath12 irreducible particle - hole @xmath13 channel as indicated by the dashed line in fig .
[ fig : lambdadn ] , and the groundstate at @xmath89 has long - range antiferromagnetic order .
eigenvalue @xmath37 versus @xmath84 for various band fillings @xmath9 for @xmath88 .
the dashed line represents the the leading eigenvalue @xmath90 in the @xmath12 , @xmath13 particle - hole channel at half - filling.,width=336 ] for @xmath27 and @xmath22 , the momentum dependence of the d - wave eigenvector @xmath91 for the 24-site cluster at @xmath92 is shown in fig .
[ fig : phipitvsk ] .
here , for @xmath93 , @xmath94 and the values of @xmath44 lay along the dashed line shown in fig .
[ fig : dcabz ] .
one clearly sees the d - wave structure of @xmath95 .
eigenvector @xmath96 at @xmath92 , normalized to its value at @xmath97 , versus @xmath44 for @xmath22 , band filling @xmath27 and @xmath93 . in the main figure ,
the @xmath44 points move along the dashed line shown in fig .
the inset shows the behavior of @xmath95 when @xmath44 varies along the @xmath98 axis.,width=336 ] the dependence of @xmath99 for @xmath44 along the @xmath100 axis is shown in the inset of fig .
, one sees that @xmath99 falls off as @xmath44 moves away from the fermi surface towards the zone center .
we have also calculated the projection of @xmath99 on the first and second @xmath7 crystal harmonics @xmath101 with @xmath102 and @xmath103 . in table
[ tab : harmonics ] , we list the values of @xmath104 versus @xmath1 at a filling @xmath27 . here
the sum in eq .
( [ eq:1 ] ) is over the entire brillouin zone and the temperature was adjusted so that the d - wave eigenvalue @xmath82 for each @xmath8 was the same ( @xmath105 ) . if the sum over @xmath44 in eq .
( [ eq:1 ] ) is restricted to values which lay along the dashed line in fig .
[ fig : dcabz ] , this ratio vanishes exactly in the 24-site cluster , since @xmath106 on the momenta @xmath44 along the dashed line .
.the ratio of the second to the first crystal d - wave harmonic projection of @xmath99 for @xmath27 and @xmath105 . [ cols="^,^,^,^",options="header " , ] the matsubara frequency dependence of @xmath107 with @xmath108 is shown in fig .
[ fig : phichi ] for @xmath27 and @xmath10 , 8 and 12 . also shown in each case is the frequency dependence of @xmath109 with @xmath12 .
the decrease of the characteristic exchange energy as @xmath8 increases is seen in figs . [ fig : phichi]a c .
it is clear from these results that the dynamics of the pairing interaction @xmath31 , reflected in the @xmath39 dependence of @xmath32 , is associated with the spin - fluctuation spectrum .
with @xmath108 for ( a ) @xmath10 , ( b ) @xmath22 and ( c ) @xmath110 calculated for @xmath111=4 . here
@xmath112 with @xmath113 and the band filling @xmath20 .
also shown is the frequency dependence of the normalized spin susceptibility @xmath114 $ ] for @xmath115.,title="fig:",width=336 ] with @xmath108 for ( a ) @xmath10 , ( b ) @xmath22 and ( c ) @xmath110 calculated for @xmath111=4 . here
@xmath112 with @xmath113 and the band filling @xmath20 . also shown is the frequency dependence of the normalized spin susceptibility @xmath114 $ ] for @xmath115.,title="fig:",width=336 ] with @xmath108 for ( a ) @xmath10 , ( b ) @xmath22 and ( c ) @xmath110 calculated for @xmath111=4 . here
@xmath112 with @xmath113 and the band filling @xmath20 . also shown is the frequency dependence of the normalized spin susceptibility @xmath114 $ ] for @xmath115.,title="fig:",width=336 ]
with the results obtained for the d - wave eigenvector @xmath116 , one can construct a simple separable representation of the pairing interaction @xmath117 @xmath118 using this separable form for @xmath119 , one finds that eq .
( [ eq : eigvalcg ] ) gives @xmath120 then , using the dressed dca monte carlo single particle green s functions and the dca results for @xmath82 , we can determine the strength @xmath121 of the separable interaction from eq . .
the strength of @xmath121 depends upon both the site occupation @xmath122 and the temperature @xmath18 .
an alternative way of extracting a simple approximation for the pairing interaction is to use the d - wave projected irreducible vertex for @xmath123 @xmath124 with @xmath125 .
[ fig : vd ] shows @xmath126 and @xmath127 for @xmath16 and various values of the site occupation @xmath9 .
both approximations give very similar results .
it is interesting to see that @xmath121 becomes stronger as @xmath9 approaches half - filling .
this is characteristic of the mott - hubbard system and has been previously observed .
a determinantal quantum monte carlo calculation of the d - wave eigenvalue on an @xmath128 lattice at half - filling with @xmath16 found that @xmath82 approached 1 as the temperature was lowered @xcite .
as noted there , in this case the antiferromagnetic eigenvalue remained dominant and on an infinite lattice the groundstate would have long - range antiferromagnetic order at @xmath89 .
a related behavior is seen for the two - leg ladder , where the pair binding energy of a finite ladder is greatest for the first two holes that are added @xcite .
this latent pairing tendency of the half - filled hubbard model is also seen in the magnitude of the probability amplitude for adding two near - neighbor holes in a d - wave state reported by plekhanov _
_ @xcite . versus temperature for different dopings for @xmath16 calculated for @xmath129.,width=336 ] versus @xmath18 for @xmath16 and @xmath130 showing the effect of the opening of the mott - hubbard gap , width=336 ] although @xmath121 increases as @xmath9 goes to 1 , the number of holes that are available for pairing is suppressed as is @xmath6 . a measure of this is given by @xmath131 which is plotted in fig . [ fig : pd0 ] for @xmath16 and @xmath130 .
here one clearly sees that as the temperature is lowered and the mott - hubbard gap opens , @xmath132 is suppressed .
the @xmath133 dependence of @xmath32 reflects a pairing interaction @xmath134 which increases at large momentum transfer @xmath135 , implying a spatially short - range interaction which is repulsive for pair formation on the same site but attractive for singlet pair formation between near - neighbor sites .
the @xmath39 dependence of @xmath32 tells us that the pairing interaction is retarded on a time scale set by @xmath4 .
the strength of the interaction is largest when @xmath1 is of order the bandwidth and increases as the system is doped towards half - filling . of course
, with @xmath16 , as @xmath9 goes to 1 , a mott - hubbard gap opens and there are no holes to pair .
this research was enabled by computational resources of the center for computational sciences at oak ridge national laboratory and conducted at the center for nanophase materials sciences , which is sponsored at oak ridge national laboratory by the division of scientific user facilities , u.s .
department of energy .
this research was supported by nsf dmr-0312680 .
djs would like to thank the stanford applied physics department for their hospitality .
djs and mj acknowledge the center for nanophase materials science at oak ridge national laboratory for support . c. huscroft , m. jarrell , t. maier , s. moukouri , and a.n .
tahvildarzadeh , _ phys .
lett . _ * 86 * , 139 ( 2001 ) ; th .
maier , m. jarrell , th .
pruschke and j. keller , _ eur .
phys . j. b _ * 13 * , 613 ( 2000 ) ; a. macridin , m. jarrell , t. maier , and p.r.c .
kent , cond - mat/0509166 ; | a dynamic cluster quantum monte carlo approximation is used to study the effective pairing interaction of a 2d hubbard model with a near neighbor hopping @xmath0 and an on - site coulomb interaction @xmath1 .
the effective pairing interaction is characterized in terms of the momentum and frequency dependence of the eigenfunction of the leading eigenvalue of the irreducible particle - particle vertex .
the momentum dependence of this eigenfunction is found to vary as @xmath2 over most of the brillouin zone and its frequency dependence is determined by the exchange energy @xmath3 .
this implies that the effective pairing interaction is attractive for singlets formed between near - neighbor sites and retarded on a time scale set by @xmath4 .
the strength of the pairing interaction measured by the size of the d - wave eigenvalue peaks for @xmath1 of order the bandwidth @xmath5 .
it is found to increase as the system is underdoped . |
merely a few years ago the computational power on the order of tflops was an attribute of a supercomputer . today , the latest commodity graphics processing units ( gpu ) are capable of more than 5 tflops and have much lower energy requirements than a set of central processing units ( cpus ) of comparable performance a result achieved by employing a massively parallel architecture with hundreds of cores on a single device . in order to take advantage of the parallel hardware ,
appropriate algorithms have to be developed , and this generally needs to be done on a problem - specific basis .
ideally , the problem naturally decomposes into a large number of ( semi-)independent tasks whose results can be combined in a simple way , as in e.g. numerical monte - carlo solution of a stochastic differential equation @xcite , parameter space studies of dynamical systems , or event simulation and reconstruction in particle physics @xcite . in recent years ,
the lattice boltzmann method ( lbm ) emerged as an interesting alternative to more established methods for fluid flow simulations .
originally developed as an extension of lattice gas automata , nowadays lbm stands on well - established theoretical foundations and serves as a method of choice for many researchers in various fields @xcite , while still retaining its relative simplicity .
perhaps its most important feature in practice is its suitability for parallel architectures
the algorithm is executed on a regular lattice and typically only interactions between nearest neighbor nodes are necessary .
this paper discusses various issues related to the implementation of the lbm on gpus , and presents a concrete and flexible solution , taking a new overall approach to the problem of lb software design .
our _ sailfish _ code has been under development since april 2009 and is publicly available under an open source license . with support for both single and binary fluid simulations , a wide variety of boundary conditions , and calculations on both a single gpu and multiple gpus , it is , to the best of our knowledge , the most comprehensive open source lbm code for gpus .
the text is organized as follows .
first , a short overview of the discussed lattice boltzmann models is presented .
then , the primary design principles of the sailfish project are laid out , and various lb gpu implementations are compared .
this is followed by performance measurements on diverse gpu hardware , both on single machines and in clusters . in the next section ,
the code is validated on 4 standard test cases .
the final section concludes the paper and provides some directions for future work .
in this section , _ all _ models implemented in sailfish as of september 2013 will be briefly overviewed .
we start with the basic concepts and nomenclature of lbm .
we then describe single fluid models : single - relaxation time lbgk and regularized dynamics , multiple relaxation times and the entropic lbm , and multi - component models : the shan - chen model and the free energy model .
finally we give a short overview of our implementation of body forces and various boundary conditions .
we refer the reader to the reviews @xcite for a more detailed discussion of the lbm . in an lb simulation space
is discretized into a regular cartesian grid , where each node represents a small pocket of fluid .
the state of the fluid is encoded by a particle distribution function @xmath0 where @xmath1 spans a set of discrete velocity vectors @xmath2 indicating the allowed directions of mass movement between the nodes of the lattice .
@xmath0 are often also called _
mass fractions_. lb lattices are typically named using the dxqy scheme , where @xmath3 indicates the dimensionality of the lattice , and @xmath4 is the number of discrete velocity vectors .
figure [ fig : lattices ] shows some common lb lattices . macroscopic fluid fields , such as density ( @xmath5 ) or velocity ( @xmath6 ) are defined as moments of the distribution function : @xmath7 in the simplest case the system dynamics is described by : @xmath8 where @xmath9 is a set of equilibrium distributions which are functions of the macroscopic fields ( @xmath5 , @xmath6 ) at node @xmath10 at time @xmath11 , @xmath12 is a relaxation time related to the kinematic viscosity @xmath13 via @xmath14 , and where the positions and time are expressed in the lattice unit system , i.e. the time unit represents a single simulation step and the lattice nodes are separated by 1 distance unit along the primary axes .
this lb model is often referred to as lattice bhatnagar - gross - krook ( lbgk ) , which owes its name to the bgk operator from kinetic theory @xcite , and by which the collision operator on the right hand side of is inspired .
.weights @xmath15 for different lattices implemented in sailfish .
[ cols="<,^,^,^,^ " , ] [ tbl : bgkweights ] when @xmath16 takes the form : @xmath17 where @xmath18 is a set of weights ( see table [ tbl : bgkweights ] ) , the navier - stokes equations can be recovered from in the limit of low mach numbers through chapman - enskog expansion @xcite or expansion in a hermite basis @xcite .
it can be shown that is second - order accurate in space and time in the bulk of the fluid @xcite .
it should be noted that and the right hand side of are fully local , i.e. only information from node @xmath10 is necessary to calculate the system state at the next step of the simulation . when data is exchanged between nodes ,
the process is limited to the nearest neighbors ( left hand side of ) .
this makes the lb method ideally suited for implementation on massively parallel computer architectures .
conceptually , is often decomposed into two steps : relaxation ( evaluation of the collision operator , right hand side ) and propagation ( movement of mass fractions to neighboring nodes , left hand side ) .
while such a decomposition is often suboptimal from the point of view of a software implementation , it still provides a useful mental framework to discuss lb algorithms , and we will use these two names throughout the text . many extended lb models have been proposed to ( a ) to improve numerical stability , or ( b ) simulate more complex systems .
a comprehensive discussion of such models would be prohibitively long , so here we limit ourselves to a short overview of models that are currently implemented within the sailfish framework . in the first group
, we discuss the multiple - relaxation times ( mrt ) model , the entropic lb model ( elbm ) and the smagorinsky large eddy model . in the second group , sailfish implements two multicomponent fluid models : the shan - chen model @xcite , and the free energy model @xcite .
one practical problem with the lbgk approximation is the limited range of viscosities at which the model remains stable , which directly translates into a limited range of reynolds numbers @xmath19 at which fluid flows can be simulated ( @xmath20 being the spatial extent of the domain expressed as a number of nodes ) . with @xmath21
limited to about @xmath22 by the low mach number requirement for most lb models where the speed of sound @xmath23 and the stable range of numerical viscosities at these speeds being about @xmath24 , the highest attainable reynolds number assuming @xmath25 is on the order of @xmath26 .
in practice , the range of reynolds numbers and viscosities is often even narrower due to instabilities introduced by complex geometry or boundary conditions . to address this problem ,
a number of models exhibiting enhanced numerical stability have been developed .
perhaps the simplest modification of the lbgk model aiming at improving its stability is the regularized model @xcite .
the basic idea of this approach is to replace the nonequilibrium part of the distributions @xmath27 with a new first order regularized value @xmath28 derived from the non - equilibrium stress tensor @xmath29 as @xmath30 .
after this regularization step , the collision and propagation proceed as in the standard lbgk scheme . if the particle distributions @xmath0 are considered to form vectors @xmath31 , the collision operator in the lbgk approximation can be interpreted as a diagonal matrix @xmath32 , with @xmath33 acting on the vector difference @xmath34 .
the basis of the distribution vector can be changed using an invertible matrix @xmath35 : @xmath36 . in particular ,
the matrix @xmath35 can be chosen such that @xmath37 is a vector of moments of the particle distributions @xmath0 .
the idea of mrt is to perform the collision in the moment basis , and then change the basis to recover the original form of the particle distributions for the streaming step @xcite .
the mrt collision operator can be written as : @xmath38 where the collision matrix @xmath39 is diagonal with its elements @xmath40 representing the inverses of relaxation times for individual moments .
collision matrix elements corresponding to conserved moments , such as density or momentum are set to 0 .
some of the remaining elements are related to kinematic and bulk viscosities , but the other ones can be tuned to increase the numerical stability of the simulation without changing its physical results .
we refer the reader to the paper of dhumieres et al .
@xcite for a detailed discussion of optimal choices for relaxation times , and the form of the matrix @xmath35 . in the entropic lb model ,
a discrete h - function @xmath41 is defined and the equilibrium distribution function is defined as the extremum of @xmath42 under constraints of mass and momentum conservation : @xmath43 where @xmath44 is the dimensionality of the lattice .
the relaxation process is also modified to include a new dynamically adjusted parameter @xmath45 : @xmath46 with @xmath47 .
@xmath45 is evaluated at every simulation step as the solution of the @xmath48-function monotonicity constraint : @xmath49 this procedure guarantees unconditional stability of the numerical scheme , as the @xmath42 function can be proven to be a lyapunov function for the system .
when @xmath50 as is the case when the system is close to equilibrium , has the same form as .
the corrections resulting from h - theorem compliance can be both positive and negative , temporarily increasing and decreasing the effective local viscosity of the fluid @xcite .
the idea behind the smagorinsky subgrid model is to locally modify the fluid viscosity by adding an eddy viscosity term dependent on the magnitude of the strain rate tensor @xmath51 : @xmath52 where @xmath53 is the baseline viscosity of the fluid and @xmath54 is an eddy viscosity , the form of which depends on the subgrid model used in the simulation . in the smagorinsky model , the relaxation time @xmath55 can be calculated using the momentum flux tensor @xmath56 : @xmath57 where @xmath58 is the smagorinsky constant , which for lb models can be taken to be @xmath59 .
this effectively abandons the single relaxation time approximation by making it a spatially and temporally varying quantity dependent on the local gradients of the fluid velocity @xcite .
the multi - fluid models discussed here are all diffuse interface models , without any explicit interface tracking
the dynamics of the interface emerges naturally from the interactions between the fluid components .
all models in this class use an additional lattice ( whose mass fractions we will call @xmath60 ) to represent the second fluid species and nonlocal interactions between the lattices in the collision process . in the shan - chen model , both lattices
are used to represent fluid components , with the 0-th moment of @xmath0 and @xmath60 representing the density of the first ( a ) and second fluid components ( b ) , respectively .
the equilibrium function and relaxation schemes remain unchanged for both lattices , but an additional coupling term in the form of a body force : @xmath61 is introduced into ( see section [ sec : body_forces ] for a short overview of ways of adding body forces to an lb simulation ) , where @xmath62 is a pseudopotential function , dependent on the density @xmath63 at node @xmath10 and @xmath64 is a coupling constant . a similar term @xmath65 of the form with @xmath66 replaced by @xmath67 and vice versa
is added to the collision operator for the second component .
a commonly used pseudopotential function is @xmath68 .
the velocity of the fluid becomes a weighted average of the first moments of the distribution functions : @xmath69 , where @xmath70 , @xmath71 and @xmath72 , @xmath63 are velocities and densities computed respectively from @xmath0 and @xmath60 using , and @xmath73 , @xmath74 are relaxation times for the two fluid components .
the free energy model is based on a landau free energy functional for a binary fluid @xcite . in the lb realization ,
the 0-th moment of @xmath0 represents the density @xmath5 of both components , while the 0-th moment of @xmath60 is a smoothly varying order parameter @xmath75 , with @xmath76 indicating a pure first component and @xmath77 indicating a pure second component : @xmath78 the macroscopic dynamics of the fluid is described by the system of equations @xmath79 where @xmath35 is a mobility parameter , @xmath80 is the chemical potential and the pressure tensor @xmath81 is defined as : @xmath82 and @xmath83 is the bulk pressure . in the corresponding lb scheme ,
distributions on both lattices are relaxed and streamed using the standard lbgk scheme using the relaxation times @xmath84 and @xmath85 , respectively .
the following equilibrium functions @xcite are used to recover in the macroscopic limit : @xmath86 \rho u_{\alpha } u_{\beta } \right ) + \kappa w_i^{\alpha \beta } \partial_{\alpha } \phi \partial_{\beta } \phi \\
g_i^{\mathrm{eq } } & = w_i \left ( \gamma \mu + e_{i \alpha } u_{\alpha } \phi + \frac{3}{2 } \left [ e_{i \alpha } e_{i \beta } - \frac{\delta_{\alpha \beta}}{3 } \right ] \phi u_{\alpha } u_{\beta } \right ) \label{eq : free_energy_eq}\end{aligned}\ ] ] where @xmath87 is a tunable parameter related to mobility via @xmath88 , and the chemical potential @xmath89 . @xmath90 and @xmath91 are constant parameters related to the surface tension @xmath92 and interface width @xmath93 .
the values of the weights @xmath15 and @xmath94 can be found in the paper by kusumaatmaja and yeomans @xcite . in order to minimize spurious currents at the interface between the two fluids , we use optimized gradient and laplacian stencils @xcite .
sailfish implements various boundary conditions , which taken together make it possible to model a wide range of physical situations : * periodic boundary conditions ( used to simulate a system spatially infinite along a given axis ) , * no - slip ( solid walls ) : half - way bounce - back , full - way bounce - back ( both with configurable wetting when used with the free energy model ) , tamm - moth - smith @xcite , * density / pressure : guo s method , zou - he s method , equilibrium distribution , regularized @xcite , * velocity : zou - he s method , equilibrium distribution , regularized @xcite , * outflow : grad s approximation @xcite , yu s method @xcite , neumann s , nearest - neighbor copy , , , do nothing @xcite .
( b ) streaming , time @xmath95 ( c ) relaxation , time @xmath95 ( arrows ending with disks represent the post - relaxation state ) , ( d ) streaming , time @xmath96.,title="fig : " ] [ fig : bb ] the bounce - back method is a simple idea originating from lattice gas automatons
a particle distribution is streamed to the wall node , and scattered back to the node it came from ( see figure [ fig : bb ] ) . in the full - way bounce - back scheme the wall nodes do not represent any fluid ( , , dry nodes ) and are only used to temporarily store distributions before scattering .
this approach is slightly easier to implement , but has the downside of the distributions needing two time steps to reach back to the originating fluid node ( one step to reach the wall node , then another to be reflected back and streamed to the fluid node ) .
the physical wall is effectively located between the last fluid node and the wall node .
in contrast , in the half - way bounce - back the wall nodes do represent fluid , the wall is located beyond the wall node ( away from the fluid domain ) , and the distributions are reflected in the same time step in which they reach the wall node @xcite . for other types of boundary conditions ,
the problem is more complex , as the mass fractions pointing into the fluid domain are undefined ( there are no nodes streaming to them in these directions ) .
various schemes have been proposed to work around this problem . in the equilibrium approach ,
the missing incoming distributions are assumed to have the same values as those opposite to them ( bounce - back ) , and the post - collision distributions are defined to be @xmath97 with one of @xmath98 being specified as part of the boundary condition , and the other calculated from the local distribution function .
zou s and he s method @xcite uses the idea of the bounce - back of the non - equilibrium parts of the distributions to calculate the missing mass fractions
. the regularized boundary conditions also use the non - equilibrium bounce - back idea , but only for an intermediate step to calculate the 2-nd moment of the non - equilibrium distribution , which is then used to reset all mass fractions at the boundary node via a regularization procedure .
we refer the reader to the excellent paper by latt et al .
@xcite for a detailed overview of these and other boundary conditions .
the grad s approximation method uses a formula dependent on the density , velocity and the pressure tensor @xmath99 to replace all distributions on the boundary node .
the three macroscopic quantities can be taken to be those from the boundary node at the previous time step ( less precise , fully local ) or extrapolated from neighboring fluid nodes ( more precise , nonlocal ) .
other outflow methods , such as the neumann s boundary implementation or yu s method , are also nonlocal and use a simple extrapolation scheme . the simplest methods ( nearest - neighbor copy , do nothing ) do not require extrapolation , but also give less control over the resulting macroscopic fields .
sailfish implements two popular ways of adding a body force @xmath100 to the simulation : guo s method @xcite , and the exact difference method ( edm ) @xcite . in both schemes , the actual fluid velocity @xmath101 can be computed as @xmath102 and a force term @xmath103 is added to the right hand side of .
guo et al .
analyzed a number of popular schemes and concluded that the optimal choice for @xmath103 in an lbgk simulation is @xmath104 \cdot \vec{f}. \label{eq : guo_force}\ ] ] in the edm , the body force term added to the collision operator is equal to the difference of equilibrium distribution functions computed using momentum after and before the action of the force @xmath100 : @xmath105 with @xmath106 .
the advantages of the edm are its lack of any spurious terms in the macroscopic navier - stokes equations and its applicability to any collision operator ( not only lbgk ) .
computational software is traditionally developed in one of two different paradigms .
_ prototype _ code is written with relatively little effort in a high level environment such as matlab or python for exploratory or teaching purposes . due to the relatively low performance of these environments , this type of code is unsuitable for large scale problems .
in contrast , _ production _ code is written in lower level languages such as c++ or fortran in a higher effort process , which can sometimes span many years and involve teams of developers .
the resulting programs are more efficient , but also longer , less readable , and potentially difficult to maintain . in our implementation of the lattice boltzmann method for gpus released under the name _ project sailfish _
, we took a hybrid approach .
we use the python programming language with a template - based code generation module ( using the mako language ) and a computer algebra system ( sympy ) to generate code in cuda c or opencl .
we chose python because it is a very expressive language , with bindings to many system libraries and great support for gpu programming via the pycuda and pyopencl packages @xcite .
in sailfish , python is used for setting up the simulation ( initial conditions , boundary conditions , selecting an lb model ) , for simulation control , communication ( e.g. between compute nodes in a cluster ) and input / output ( loading geometry , saving simulation results ) . we also employ the numpy package to perform matrix operations efficiently . .... from sympy
import rational def bgk_equilibrium(grid ) : out = [ ] for ei , weight in zip(grid.basis , grid.weights ) : out.append(weight * ( s.rho + s.rho * ( 3 * ei.dot(grid.v ) + rational(9 , 2 ) * ( ei.dot(grid.v))**2 - rational(3 , 2 ) * grid.v.dot(grid.v ) ) ) ) return out .... mathematical formulas , such as , , are stored in the form of sympy expressions .
we decided to do this after noticing that computer code for numerical calculations is often very repetitive , e.g. , when evaluated on a d3q19 lattice , would expand to 19 long lines of code , even though the initial formula easily fits in one line ( see listing [ lst : equilibrium ] ) .
this approach makes the code easier to read ( the formulas can also be automatically converted to latex expressions ) , allows for automated consistency checks ( e.g. one can easily , in a single line of code , verify that the 0th moment of @xmath107 is @xmath5 ) , enables easy experimentation with code optimization techniques ( formulas can be reorganized e.g. for speed or precision of floating - point operations ) , and enables code reuse ( e.g. the formula for equilibrium or the bounce - back rule is written down once , but can generate code for various types of lattices , both in 2d and 3d , see listing [ lst : bb ] for an example ) . finally , the mako template language is used to render the formulas into low - level cuda / opencl code .
the generated code is fairly small , contains comments , and is automatically formatted , making it suitable for instructional purposes as well as for one - off , ad - hoc modifications .
this is in stark contrast to large , multimodular codes where the architecture of the system can be difficult to understand without extensive documentation .
sailfish was designed from scratch and optimized for modern massively parallel architectures , such as gpus .
we wanted to achieve high flexibility of the code , ease of use when running or defining new simulations , and not take any performance hits compared to code written directly in cuda c or opencl . in order to achieve
that , we decided to use run - time code generation techniques .
this provides some isolation from low - level hardware details , important in the case of gpus which are still evolving rapidly and changing with every new gpu architecture .
it also makes it possible to generate optimized code on a case - by - case basis and to automatically explore parameter spaces to find optimal solutions , thus saving programmer time and increasing their productivity . with many details , such as the number of compute units , size of on - chip memory , speed of access patterns to on - chip and off - chip memory , host - device latency and bandwidth , memory bandwidth to computational performance ratio directly impacting the performance of the code , experimentation and microbenchmarking
are necessary to find combinations of parameters that work well .
we also considered other metaprogramming approaches to code generation , such as domain - specific languages , and templates in c++ .
we deemed the first solution to have too much overhead , and decided against the latter one since the expanded code can not be saved for inspection and modification , and there were no open source computer algebra libraries providing a level of flexibility and sophistication comparable to sympy . .... $ { device_func } inline void bounce_back(dist * fi ) float t ; % for i in sym.bb_swap_pairs(grid ) : < % a = grid.idx_name[i ] opp_i = grid.idx_opposite[i ] b = grid.idx_name[opp_i ] % >
t = fi->${a } ; fi->${a } = fi->${b } ; fi->${b } = t ; % endfor } ! \vfill\columnbreak ! _ _ device _ _ inline void bounce_back(dist * fi ) { float t ; t = fi->fe ; fi->fe = fi->fw ; fi->fw = t ; t = fi->fn ; fi->fn = fi->fs ; fi->fs = t ; t = fi->fne ; fi->fne = fi->fsw ; fi->fsw = t ; t = fi->fnw ; fi->fnw = fi->fse ; fi->fse = t ; } .... sailfish takes extensive advantage of objected - oriented programming techniques for modularization and code reuse purposes .
each simulation is defined using two classes a simulation class , and a geometry class ( see listing [ lst : ldc_example ] ) .
the simulation domain can be divided into cuboid subdomains , which do not need to fill the parts of the domain that do not contain any fluid ( this makes it possible to handle complex geometries ) .
the geometry class defines initial and boundary conditions for a single subdomain .
the simulation class derives from a base class specific to the lb model used such as ` lbfluidsim ` ( single component simulations ) or ` lbbinaryfluidfreeenergy ` ( binary fluids using the free energy model ) .
the simulation class can optionally also add body forces or define custom code to be run after selected steps of the simulation ( e.g. to display a status update , check whether the steady state has been reached , etc ) .
the base simulation class specifies the details of the used lb model , such as the form of the equilibrium distribution function ( in symbolic form , stored as a sympy expression ) , number and names of macroscopic fields ( density , velocity , order parameter , etc ) , and names of gpu code functions that need to be called at every simulation step . when executed , every sailfish simulation starts a controller process which parses any command line parameters and reads configuration files , decides how many computational nodes and gpus are to be used ( in case of distributed simulations ) , and how the subdomains are going to be assigned to them .
it then starts a master process on every computational node ( see figure [ fig : system_arch ] ) , which in turn starts a subdomain handler process for every subdomain assigned to the computational node .
we use subprocesses instead of threads in order to avoid limitations of the python runtime in this area ( the global interpreter lock preventing true multithreading ) , as well as to simplify gpu programming .
the subdomain handlers instantiate the simulation and geometry classes for their respective subdomains .
the system state is initialized in the form of numpy arrays describing the initial macroscopic fields , and these are then used to initialize the distribution functions on the gpu ( using the equilibrium distribution function ) .
optionally , a self - consistent initialization procedure @xcite can be performed to compute the density field from a known velocity field .
once the system is fully initialized , the simulation starts running , with the subdomain handlers exchanging information in a peer - to - peer fashion as determined by the connectivity of the global geometry . .... from sailfish.subdomain import subdomain2d from sailfish.node_type import ntfullbbwall , ntzouhevelocity from sailfish.controller import lbsimulationcontroller from sailfish.lb_single import lbfluidsim class ldcsubdomain(subdomain2d ) : max_v = 0.1 def boundary_conditions(self , hx , hy ) : wall_map = ( hx = = self.gx-1 ) | ( hx = = 0 ) | ( hy = = 0 ) self.set_node((hy = = self.gy-1 ) & ( hx > 0 ) & ( hx < self.gx-1 ) , ntzouhevelocity((self.max_v , 0.0 ) ) ) self.set_node(wall_map , ntfullbbwall ) def initial_conditions(self , sim , hx , hy ) : sim.rho [ : ] = 1.0 sim.vx[hy = = self.gy-1 ] = self.max_v class ldcsim(lbfluidsim ) : subdomain = ldcsubdomain if _ _ name _ _ = = ' _ _ main _ _ ' : lbsimulationcontroller(ldcsim).run ( ) .... modern gpus are massively parallel computational devices , capable of performing trillions of floating - point operations per second .
we will now briefly present the architecture of cuda - compatible devices as a representative example of the hardware architecture targeted by sailfish .
other devices not supporting cuda but supporting opencl are based on the same core concepts .
cuda devices can be grouped into three generations , called tesla , fermi , and kepler each one offering progressively more advanced features and better performance .
the cuda gpu is organized around the concept of a streaming multiprocessor ( mp ) .
such a multiprocessor consists of several scalar processors ( sps ) , each of which is capable of executing a thread in a simt ( single instruction , multiple threads ) manner .
each mp also has a limited amount of specialized on - chip memory : a set of 32-bit registers , a shared memory block and l1 cache , a constant cache , and a texture cache .
the registers are logically local to the scalar processor , but the other types of memory are shared between all sps in a mp , which allows data sharing between threads .
perhaps the most salient feature of the cuda architecture is the memory hierarchy with 1 - 2 orders of magnitude differences between access times at each successive level .
the slowest kind of memory is the host memory ( ram ) .
while the ram can nowadays be quite large , it is separated from the gpu by the pcie bus , with a maximum theoretical throughput in one direction of 16 gb / s ( pci express 3.0 , x16 link ) .
next in line is the global device memory of the gpu , which is currently limited to several gigabytes and which has a bandwidth of about 100 - 200 gb / s .
global memory accesses are however high - latency operations , taking several hundred clock cycles of the gpu to complete .
the fastest kind of memory currently available on gpus is the shared memory block residing on mps .
it is currently limited in size to just 48 kb ( 16 kb on tesla devices ) , but has a bandwidth of ca 1.3 tb / s and a latency usually no higher than that of a sp register access .
the above description readily suggests an optimization strategy which we will generally follow in the next section and which can be summarized as : move as much data as possible to the fastest kind of memory available and keep it there as long as possible , while minimizing accesses to slower kinds of memory .
when memory accesses are necessary , it also makes sense to try to overlap them with independent computation , which can then be executed in parallel effectively hiding the memory latency . from the programmer s point of view , cuda programs are organized into kernels .
a kernel is a function that is executed multiple times simultaneously on different mps .
each instance of this function is called a thread , and is assigned to a single scalar processor .
threads are then grouped in one- , two- or three - dimensional blocks assigned to multiprocessors in an 1 - 1 manner ( 1 block - 1 mp ) .
the blocks are organized into a one- or two - dimensional grid .
the size and dimensionality of the grid and blocks is determined by the programmer at the time of kernel invocation .
knowledge of the grid position , and the in - block position makes it possible to calculate a thread i d that is unique during the kernel execution . within a single block threads
can synchronize their execution and share information through the on - chip shared memory .
synchronization between blocks is not supported in any way other than serializing the execution of kernels , and through atomic operations on global memory .
the great potential of modern gpus as a hardware platform for simulating both two - dimensional @xcite and three - dimensional flows @xcite with the lattice boltzmann method was quickly realized after the initial release of the cuda programming environment in 2007 .
recently , gpus have also been used to investigate more complex , two - dimensional multicomponent flows @xcite .
there are three main factors that need to be carefully optimized in order to fully utilize the computational power of modern gpus : memory access patterns , register utilization , and overlap of memory transfers and arithmetic operations .
the first factor has received the most attention in the literature , but in this work we show that the remaining two are also important .
the most important data structure used in an lb simulation are the particle distributions @xmath0 .
they are stored in the global gpu memory in a structure of arrays ( soa ) fashion , effectively forming a 4d or 3d array ( for 3d and 2d simulations , respectively ) , with the following index ordering : @xmath108 , where @xmath109 is the number of discrete velocities in the lattice .
an array of structures ( aos ) approach , while elegant from the point of view of object - oriented programming , is completely unsuitable for gpu architectures due to how global reads and writes are performed by the hardware .
global memory accesses are performed by thread warps ( 32 threads ) in transactions of 32 , 64 , or 128 bytes .
on fermi devices , accesses cached in the l1 cache are serviced with 128-byte transactions ( the size of a full l1 cache line ) , while those cached in the l2 cache are serviced with 32-byte transactions . in order to attain good bandwidth utilization
, the memory has to be accessed in contiguous blocks so that all bytes in a transaction are used for meaningful data .
the memory location also has to be naturally aligned , i.e. the first address in the transferred segment must by a multiple of the segment s size . in sailfish
, we run all kernels in 1-dimensional thread blocks spanning the x axis of the subdomain , with a typical block size being 64 , 128 or 192 nodes .
each thread handles a single node of the subdomain . due to the layout of the distributions in global memory , we issue @xmath109 read requests to load a full set of mass fractions in a thread block .
each request results in a fully utilized memory transaction . to ensure natural alignment of transactions ,
the x dimension of the distributions array in global memory is padded with unused nodes so that @xmath3 is a multiple of 32 or 16 , for single and double precision floating point numbers , respectively .
in addition to the distributions array , we also store a node type map ( see section [ sec : boundary ] ) , and macroscopic field arrays in global memory , all following the same memory layout and padding as described above .
a simple lb simulation in sailfish repeatedly calls a single cuda kernel called ` collideandpropagate ` , which implements both the collision and propagation step ( see e.g. ) , as well as any boundary conditions . a high - level summary of this kernel
is presented in algorithm [ alg : cnp ] .
allocate shared memory array @xmath110 $ ] compute global array index @xmath111 for the current node @xmath1 load and decode node type from global memory to @xmath112 load distributions from global memory to @xmath113 compute macroscopic variables @xmath5 and @xmath101 apply boundary conditions save @xmath5 and @xmath101 in global memory arrays compute the equilibrium distribution @xmath107 using @xmath5 and @xmath101 relaxation : @xmath114 write @xmath115 to global memory at @xmath116 @xmath117 \gets f_{k}$ ] write @xmath115 to global memory at @xmath118 synchronize threads within the block write @xmath119 $ ] to global memory at @xmath116 @xmath120 \gets f_{k}$ ] write @xmath115 to global memory at @xmath118 synchronize threads within the block write @xmath119 $ ] to global memory at @xmath116 two basic patterns of accessing the distributions have been described in the literature , commonly called ab and aa @xcite . in the ab access pattern , there are two copies of the distributions in global memory ( a and b ) . the simulation step alternates between reading from a and writing to b , and vice versa . in the aa access pattern ,
only a single copy of the distributions array is stored in global memory .
since there are no guarantees about the order of execution of individual threads , care has to be taken that a thread reads from and writes to exactly the same locations in memory .
this is illustrated in figure [ fig : memory ] .
a third access pattern that is used in practice is the so - called _ indirect addressing_. in this mode , in order to access data for a node , its address has to be read first .
this causes overhead both in storage ( need to store the addresses ) and in memory accesses , but can be very useful for complex geometries where the active nodes are only a tiny fraction of the volume of the bounding box .
indirect addressing can be combined with both aa and ab access patterns . for a dense subdomain using indirect addressing
, the performance can be 5 - 25% lower than when using direct addressing , with c2050 showing better results with the aa access pattern than in ab , and the gtx 680 exhibiting the opposite tendency .
the exact performance is however necessarily geometry - dependent , and as such it is not discussed further here . [
fig : memory ] the propagation step of the lb algorithm shifts the distributions by @xmath121 node in all directions when this happens for the x axis , it results in misaligned reads / writes . to reduce the impact of misaligned writes , sailfish utilizes shared memory to shift data in the x direction within the thread block ( see algorithm [ alg : cnp ] ) .
we performed numerical tests to evaluate the impact of various data access patterns on the overall performance of the simulation .
earlier works @xcite indicate that the cost of unaligned writes is significantly higher than the cost of unaligned reads , and that a propagate - on - read strategy results in up to 15% performance gain .
our experiments confirmed this on older gt200 hardware ( gtx 285 ) , however we failed to replicate this effect on fermi and kepler devices ( tesla c2050 , k10 , k20 ) , where the performance of both implementations was nearly identical ( typically , with a few % loss for propagate - on - read ) .
this is most likely caused by hardware improvements in the fermi and later architectures .
we also compared the performance for the ab and aa access patterns .
on tesla - generation devices ( gtx 285 , tesla c1060 ) , the aa memory layout results in a slightly higher performance , and is therefore clearly preferred . on newer devices
the ab scheme is typically a little faster , but the performance gains over the aa scheme are minor ( @xmath122 ) , and as such the ab scheme is only preferable when ample gpu memory is available .
boundary conditions are handled in sailfish with the help of a _ node type map _ an unsigned 32-bit integer array stored in the global gpu memory .
each entry in this array contains encoded information about the node type ( fluid , unused , ghost , type of boundary condition to use ) , orientation ( vector normal to the boundary , pointing into the fluid ) and a parameter i d .
the parameter i d is an index to a global array of values used by boundary conditions ( e.g. densities , velocities ) .
the encoding scheme uses variable - size bitfields , which are dynamically chosen for every simulation depending on the usage of different boundary conditions in a subdomain ( see figure [ fig : encoding ] ) .
time - dependence is supported for all types of boundary conditions .
when a value changing in time is required , it is typically specified in the form of a sympy expression .
this expression is then transformed into a gpu function and assigned an i d , analogous to the parameter i d for static boundary conditions .
models with more than one distribution function , such as the shan - chen model or the free energy model introduce a nonlocal coupling via one or more macroscopic fields . to minimize the amount of data read from global memory , we split the simulation step into two logical parts , implemented as two gpu kernels . in the first kernel , ` computemacrofields ` , we load the distributions and compute the macroscopic field value for every field which needs to be accessed in a nonlocal manner in the collision kernel .
the collision step is implemented similarly to single fluid models , and the nonlocal quantities ( shan - chen force , gradient and laplacian of the order parameter in the free energy model ) are computed by directly accessing the macroscopic field values in global memory .
we considered three approaches to realizing the collision and propagation part of the algorithm for multifluid models . in the first variant
, we used a single collision kernel , which loaded both distributions into registers and ran the collision and propagation step for both lattices . in the second variant
, we tried to speed - up the calculations of nonlocal values by binding the macroscopic fields to gpu textures and accessing them through these bindings . in the last variant ,
we split the collision process into two separate kernels , one for every lattice .
the second variant yielded minimal speed - ups ( on the order of a few percent ) on old tesla - generation devices , which however did not carry over to fermi and kepler ones .
we abandoned the idea as it introduced unnecessary complexity to the code . the third approach using split kernels proved to be the most efficient one .
we were able to obtain 5.4% ( d2q9 , free energy )
53% ( d3q19 , free energy ) speed - ups as compared to a monolithic kernel , mainly due to the kernels using fewer registers and being able to achieve higher occupancy , which hid the minor overhead introduced by a slightly increased number of global memory accesses . as a side benefit , this approach also resulted in simpler code , so we decided to standardize on it in the sailfish framework .
we also expect it to scale well to models requiring more than 2 lattices . in sailfish ,
the mechanisms that support distributed simulations are very similar to those supporting multiple subdomains on a single gpu or many gpus on one computational node .
every subdomain has a layer of _ ghost _ nodes around it .
these nodes do not participate in the simulation , but are used for data storage for mass fractions leaving the subdomain or for macroscopic fields of neighboring nodes located in remote subdomains .
once mass fractions are streamed into the ghost nodes , we run additional cuda kernels to collect the data leaving the subdomain into a linear buffer in global memory , which is then transferred back to the host and sent to remote nodes using a network link . the subdomain is also split into two areas called _ bulk _ and _ boundary _ , which are simulated via two separate kernel calls .
the boundary area is defined as all nodes belonging to cuda thread blocks where at least one node touches the subdomain boundary .
the simulation step is first performed for nodes in the boundary area , so that communication with remote subdomain runners can start as soon as possible and can overlap in time with the simulation of the bulk of the subdomain . as a further optimization ,
we limit the data sent between computational nodes exactly to the mass fractions that actually need to be sent ( e.g. if two nodes are connected along the x axis , then only the mass fractions corresponding to discrete velocities with a positive x component will be transferred from the left subdomain to the right one ) .
we also make it possible to optionally compress the data before sending it to the remote node , which can be helpful if the simulation is run on multiple machines with slow network links between them , or when only a small fraction of nodes on the subdomain interface plane is active .
many gpus are significantly faster when calculations are done in single precision as opposed to the standard double precision which is typically used in scientific and engineering applications .
the speed - up factor can vary between 10 and 2 , depending on the device model and generation .
the performance gap is smaller in newer devices ( e.g. kepler generation ) .
earlier works @xcite used the lid - driven cavity benchmark to verify that single precision calculations produce satisfactory results . here
, we use the 2d taylor - green decaying vortex flow a benchmark problem with known analytical solution to study the accuracy of our lb implementation in both single ( sp ) and double ( dp ) precision .
the simulations are done using the lbgk relaxation model on a d2q9 lattice .
for single precision , both the standard formulation and the round - off minimizing formulation @xcite ( sro ) are tested .
the taylor - green vortex can be described by the following system of equations : @xmath123 where @xmath124 is a velocity constant and @xmath125 .
can be easily verified to satisfy the incompressible navier - stokes equations .
we performed multiple runs of the simulation on a @xmath126 lattice with periodic boundary conditions in both directions , varying the viscosity @xmath13 but keeping the reynolds number constant at @xmath127 .
each simulation was run until it reached a specific point in physical time , corresponding to @xmath128 iterations at @xmath129 .
we use the l2 norm of the velocity and density field difference to characterize the deviation of the numerical solution from the analytical one : @xmath130 where @xmath131 , @xmath132 is the numerical solution .
the results presented in figure [ fig : precision ] illustrate interesting differences between double precision and single precision calculations . in double precision ,
the error stays constant until @xmath133 and raises quadratically for higher values of velocity , as could be expected from the @xmath134 accuracy of the lbgk model . in single precision however , lower speeds lead to higher error values .
the interplay between model accuracy and numerical round - off leads to a sweet spot around @xmath135 , where the errors are minimized . for higher speeds
, there is no difference between double and single precision .
this result shows that the conventional wisdom that lower speeds always lead to better precision is not true when applied to single precision simulations .
since many practical simulations are run at velocities @xmath22 and higher for reasons of efficiency , it also explains why in many cases no differences are observed between single and double precision results .
the density error shows that the round - off minimization model generates significantly better results than the standard single precision implementation , and as such should be preferred in cases where the accuracy in the density field is important .
all performance figures in the this section are quoted in mlups ( millions of lattice - site updates per second ) .
the tests were run using cuda toolkit 5.0 , pycuda 2013.1.1 on 64-bit linux systems . .
both logical gpus were used for the k10 tests .
left panel : single precision .
right panel : double precision . ]
[ fig : memory_grid ] single fluid models were benchmarked using a lid - driven cavity test case , with a @xmath136 lattice at @xmath127 , using full bounce - back walls and equilibrium velocity boundary conditions for the lid . to illustrate the impact of the entropy - stabilized time - stepping , the elbm solver was also tested at @xmath137 .
the binary fluid models were benchmarked using a simple spinodal decomposition test case , where a uniform mix of both fluid components fills the whole simulation domain ( @xmath138 ) and periodic boundary conditions are enabled in all directions . whenever the domain size was too large to fit on a specific gpu we tested , we reduced the size in the z direction until it fit .
we find our results to be comparable or better to those reported by habich et al .
@xcite , who used hardware very similar to ours ( tesla c2050 and c2070 differ only in memory size , according to nvidia specifications ) .
the performance of single precision simulations can be shown to be limited by global memory bandwidth .
we find that our code runs at @xmath139 of the theoretical bandwidth as reported in nvidia whitepapers , and close to 100% of the real bandwidth measured by sample code from the nvidia sdk .
double precision simulations run at @xmath140 of the theoretical maximum .
they are limited by the double precision instruction throughput on the gpu on fermi - class hardware and by memory bandwidth in kepler hardware .
overall , we find that the memory bandwidth is a reliable indicator of expected performance for single precision simulations across all three generations of nvidia devices ( see also figure [ fig : memory_grid ] , which shows that the simulation performance is inversely proportional to the lattice connectivity @xmath141 in single precision ) . for double precision simulations on fermi hardware
, we have found increasing the l1 cache size to 48 kb , disabling l1 cache for global memory accesses and replacing division operations by equivalent multiplication operations to have a large positive impact on the performance of the code ( @xmath142 speed - up in total ) .
the impact of the l1 cache is understandable if one considers the fact that double precision codes use significantly more registers .
this causes the compiler to spill some of them over to local memory , accesses to which always go via the l1 cache .
the larger the size of the unused part of that cache , the more operations can execute without actually accessing the global memory .
this optimization strategy does not apply to kepler - class devices , where both in single and double precision we found that disabling the preference for l1 cache for the main computational kernel had a positive impact on performance .
we also tested our code on lower end gpus ( mobile versions used in laptops ) with compute capability 3.0 , where we found a significant speed up ( up to 40% ) by using shuffle operations for in - warp propagation , and limiting shared memory for data exchange between warps only .
this optimization strategy does not work with the higher end gpus discussed in this paper , where the performance is limited by global memory bandwidth already .
overall , we unsurprisingly find that more recent gpus perform noticeably better .
the k10 is an interesting option if only single precision is required as it delivers the highest overall performance at a level higher than 1.3 glups per board with d3q19 and simple fluid models . for double precision ,
the k20x card being the most advanced nvidia gpu available on the market when this paper is written , is a clear winner performance - wise . while the computational power provided by a single gpu is impressive
, practical simulations often require large domains , and for these the total size of gpu memory ( a few gbs ) is an important limitation .
the natural solution of this problem is to run a distributed simulation using multiple gpus , which can be physically located in a single computer or multiple computers in a network . in order to measure performance of distributed simulations
, we ran a 3d duct flow test case ( periodic boundary conditions in the streamwise z direction , bounce - back walls at other boundaries ) on the zeus cluster ( part of the plgrid infrastructure ) , consisting of computational nodes with 8 m2090 gpus and interconnected with an infiniband qdr network . the first test we performed measured weak scaling , i.e. code performance as a function of increasing size of the computational domain .
the domain size was @xmath143 , where @xmath144 is the number of gpus .
we used the d3q19 lattice , the aa memory access pattern , a cuda block size of 128 , and single precision .
figure [ fig : weak_scaling ] shows excellent scaling up to 64 gpus , which was the largest job size we were able to run on the cluster .
the 1.5% efficiency loss takes place as soon as more than 1 subdomain is used , and does not degrade noticeably as the domain size is increased .
this small efficiency loss could be further minimized by employing additional optimization techniques , such as using peer - to - peer copies when the gpus are located on the same host . the second test we ran measured strong scaling , i.e. code performance with a constant domain size , as a function of increasing number of gpus .
we used a @xmath145 geometry ( largest domain size that fit within the memory of a single m2090 gpu ) and other settings as in the weak scaling test , and divided the domain into equal - length chunks along the z axis as more gpus were used for the simulation .
the results of this test are presented in figure [ fig : strong_scaling ] .
a slightly worse performance is visible in comparison to the weak scaling test , but even when the simulation is expanded to run on 8 gpus , only a 3.5% efficiency loss can be observed .
it should also be noted , that there is a minimum domain size below which performance quickly degrades due to the overhead of starting kernels on the gpus .
this size can be seen in figure [ fig : min_domain_size ] to be about 14% of the gpu memory or 8.2 m lattice nodes .
cuda gpus provide an alternative hardware implementation of various transcendental functions such as the exponent , logarithm or trigonometric functions .
these functions , known as intrinsic functions , are faster but less precise than their normal counterparts , especially when their arguments do not fall within a narrow range specified for each function .
we analyzed the impact of these functions on the performance and precision of the lb models that can take advantage of them , namely the shan - chen model with a non - linear pseudopotential and the entropic lbm .
with elbm the use of intrinsic functions , together with the fmad ( fused multiply - add ) instruction yields a speed - up of @xmath146% without any noticeable impact on the correctness of the results ( in terms of global metrics such as the total kinetic energy , enstrophy and the kinetic energy spectrum ) . while testing these optimizations , we also found that the ftz ( denormalize to 0 ) option of the cuda compiler causes the elbm simulation to crash . with the proposed optimizations ,
the performance of elbm is at 72% of the lbgk performance , making it a very interesting alternative for some simulations . for the shan - chen model with a nonlinear pseudopotential we saw 17 - 20% speed - ups in 2d and 3d , with relative changes in the density fields smaller than 1.5% after 10000 steps .
unfortunately , the same approach does not yield speed - ups in double precision , as most intrinsic functions are available in single precision only .
in order to validate our implementation of the lbm , we performed simulations for four classical computational fluid dynamics test cases and compared our results with those published in the literature .
the lid - driven cavity geometry consists of a cube cavity with a face length @xmath20 .
the geometric center of the cavity is located at the origin of the coordinate system .
the cube face at @xmath147 moves tangentially in the y - direction with a constant velocity @xmath148 , while all other faces are no - slip walls .
we carried out simulations of this problem at @xmath127 with various lb models ( bgk , mrt , regularized bgk , elbm ) using a @xmath149 d3q19 lattice , full - way bounce - back for no - slip walls , and the regularized velocity boundary condition with @xmath150 for the moving wall .
our results ( both in single and double precision ) agree with those published by albensoeder et al .
@xcite ( see figure [ fig : ldc - comparison ] ) and @xmath151 .
round dots : data from tables 5 and 6 in albensoeder et al .
@xcite solid line : results from sailfish simulations on a @xmath149 lattice , after @xmath152 steps using lbgk , mrt , regularized bgk and elbm in single and double precision .
the results from all sailfish simulations are in agreement to within the width of the line on the plot .
lb results are rescaled using : @xmath153.,scaledwidth=50.0% ] the kida vortex flow is a free decay from the initial conditions @xcite : @xmath154 defined on a cubic domain with face length @xmath155 , with periodic boundary conditions in all directions . to validate our code
, we performed simulations for @xmath156 , @xmath157 , and @xmath158 and compared them with results published in @xcite and @xcite , respectively .
the simulations were run using @xmath159 on a @xmath160 grid ( @xmath161 for @xmath162 ) , using both single and double precision ( with no noticeable difference between them ) .
the @xmath162 and @xmath163 cases were investigated using the lbgk , mrt , regularized lbgk , smagorinsky - les and entropic models . at @xmath164 ,
only simulations using the entropic model and the smagorinsky subgrid model ( with @xmath165 ) remained stable . during the simulation time kinetic energy @xmath166 and enstrophy @xmath167 , where @xmath168 is the volume of the simulation domain ,
were tracked directly on the gpu for optimal efficiency .
vorticity was computed using the first order central difference scheme in the bulk of the fluid and forward / backward differences at subdomain boundaries . for various collision models , ( d ) kinetic energy spectrum for selected collision models and reynolds numbers . for panels ( a)-(c ) time
is rescaled assuming a domain size of @xmath169 and @xmath170 . ] for @xmath171 all four models gave the same results ( figure [ fig : ke - ens ] ) . at @xmath163
some minor differences are visible , particularly in the evolution of enstrophy .
its peak value is slightly underpredicted by both models that locally modify effective viscosity ( smagorinsky , elbm ) ( see figure [ fig : ke - ens](c ) ) . at @xmath164 ,
the differences are more pronounced and we observe that the smagorinsky model underpredicts the absolute value of peak enstrophy .
the kinetic energy spectrum shown on figure [ fig : ke - ens](d ) was computed as @xmath172 , for @xmath173 .
a good agreement is visible in comparison to the kolmogorov scaling @xmath174 , especially for the high reynolds number cases .
all collision models lead to similar spectra , with elbm at @xmath162 predicting a slightly higher value around @xmath175 than lbgk or other models , and with elbm keeping a slightly flatter spectrum for high @xmath176 values at @xmath164 . in all cases
the simulation results show the same features as those discussed in previous papers on this topic @xcite . to verify the binary fluid models we consider a 2d poiseuille flow in the @xmath3-direction .
no - slip walls are imposed at @xmath177 using the half - way bounce - back boundary conditions , and periodic boundary conditions are used in the @xmath3-direction .
a body force @xmath64 drives the flow . in the core flow region ( @xmath178 ) a fluid of viscosity @xmath179 is placed , while in the boundary flow region ( @xmath180 ) the viscosity of the fluid is @xmath53 . the analytical solution for this case
can be expressed as : @xmath181 we run the simulation on a @xmath182 grid with @xmath183 , and @xmath184 .
the simulation starts with @xmath185 in the whole domain and we let the flow reach the stationary state on its own .
the free energy simulation was run with @xmath186 , @xmath187 , and @xmath188 , while for the shan - chen model , @xmath189 was used .
the parameters were chosen to ensure that the interface remains as sharp as possible without destabilizing the simulation .
the exact difference method was used to introduce body forces .
the shan - chen model permits residual mixing between the fluid components , so the effective density and viscosity were calculated as @xmath190 and @xmath191 , respectively .
figure [ fig : poiseuille_comparison ] illustrates the good agreement of the simulation results with . due to mixing between the two fluid components .
the free energy model shows better agreement with the theoretical profile due to a thinner interface . ] in order to verify the binary fluid models also in a dynamic case , we simulate the decay of the amplitude of a capillary wave .
the boundary conditions of the system are the same as in the binary poiseuille flow case , but we now use a larger @xmath192 lattice in order to accommodate higher frequency waves . the region @xmath193 is filled with one component ( a ) and the region @xmath194 is filled with another component ( b ) . for simplicity , we choose both components to have the same viscosities @xmath195 .
the interface between the two fluids is initialized to a sinusoidal curve of wavelength @xmath196 chosen such that an integer number of wavelengths fits exactly in the simulation domain .
the interface is then allowed to relax naturally with no external forces , resulting in a damped capillary wave . at each timestep of the simulation ,
the height of the interface at @xmath197 is recorded . in order to recover the frequency of the wave ,
an exponentially damped sine function is fit to the interface height data .
for the shan - chen model , we used @xmath198 and for the free energy model we used @xmath199 , @xmath200 , and @xmath201 .
for the capillary wave for the free energy and shan - chen models.,scaledwidth=50.0% ] as expected @xcite , the dispersion relation shows a power law form @xmath202 for both the shan - chen and free energy models ( see figure [ fig : capillary ] ) .
in the previous sections we have demonstrated our sailfish code as a flexible framework for implementing lattice boltzmann models on modern graphics processing units . with novel optimization techniques for complex models
it provides a very efficient tool for a wide range of simulations .
we hope that our observations collected while running the presented benchmarks will serve as a guideline in the choice of both lb models and computational hardware for users of sailfish and of other similar codes . for single precision simulations , we advocate a careful choice of parameters and correctness testing via comparisons to similar test cases in double precision .
while all of our benchmark problems did not show any noticeable differences between single and double precision , the taylor - green test case clearly demonstrates that these do exist and can significantly impact the results if the simulation is in the slow velocity regime .
whenever possible , the round - off minimizing model should be used to reduce precision losses without any impact on performance .
while the capabilities of sailfish are already quite extensive , much work remains to be done . among the most important remaining tasks , we mention ongoing efforts to implement a hierarchical lattice , allowing for local grid refinement and a fluid - structure interaction model based on the immersed boundary method .
since the code is freely available under an open source license , we would like to invite the reader to participate in its development and contribute new enhancements according to their interests .
this work was supported by the twing project co - financed by the european social fund , as well as in part by pl - grid infrastructure .
m.j . thanks s. chikatamarla and i. karlin for useful discussions and sharing reference data from their simulations .
the authors would also like to thank nvidia for providing hardware resources for development and benchmarking .
m. januszewski , m. kostur , http://www.sciencedirect.com/science/article/b6tj5-4x7r842-1/2/ff8a1e76fd160f7aba015ca73e39a3ad[accelerating numerical solution of stochastic differential equations with cuda ] , computer physics communications 181 ( 1 ) ( 2010 ) 183 188 .
http://dx.doi.org/doi : 10.1016/j.cpc.2009.09.009 [ ] .
http://www.sciencedirect.com/science/article/b6tj5-4x7r842-1/2/ff8a1e76fd160f7aba015ca73e39a3ad s. chen , g. d. doolen , http://dx.doi.org/10.1146/annurev.fluid.30.1.329[lattice boltzmann method for fluid flows ] , annual review of fluid mechanics 30 ( 1 ) ( 1998 ) 329364 .
http://dx.doi.org/10.1146/annurev.fluid.30.1.329 [ ] .
http://dx.doi.org/10.1146/annurev.fluid.30.1.329 c. k. aidun , j. r. clausen , http://dx.doi.org/10.1146/annurev-fluid-121108-145519[lattice-boltzmann method for complex flows ] , annual review of fluid mechanics 42 ( 1 ) ( 2010 ) 439472 .
http://dx.doi.org/10.1146/annurev-fluid-121108-145519 [ ] .
d. dhumires , m. bouzidi , p. lallemand , http://link.aps.org/doi/10.1103/physreve.63.066702[thirteen-velocity three - dimensional lattice boltzmann model ] , phys .
e 63 ( 2001 ) 066702 . http://dx.doi.org/10.1103/physreve.63.066702 [ ] .
http://link.aps.org/doi/10.1103/physreve.63.066702 p.
l. bhatnagar , e. p.
gross , m. krook , http://dx.doi.org/10.1103/physrev.94.511[a model for collision processes in gases .
i. small amplitude processes in charged and neutral one - component systems ] , physical review online archive ( prola ) 94 ( 3 ) ( 1954 ) 511525 . http://dx.doi.org/10.1103/physrev.94.511 [ ] .
x. shan , h. chen , http://dx.doi.org/10.1103/physreve.47.1815[lattice boltzmann model for simulating flows with multiple phases and components ] , physical review e 47 ( 3 ) ( 1993 ) 18151819 .
j. latt , b. chopard , http://www.sciencedirect.com/science/article/pii/s0378475406001583[lattice boltzmann method with regularized pre - collision distribution functions ] , mathematics and computers in simulation 72 ( 26 ) ( 2006 ) 165 168 , <
ce : title > discrete simulation of fluid dynamics in complex systems</ce : title>. http://dx.doi.org/http://dx.doi.org/10.1016/j.matcom.2006.05.017 [ ] .
s. ansumali , i. v. karlin , h. c. ttinger , http://iopscience.iop.org/0295-5075/63/6/798[minimal entropic kinetic models for hydrodynamics ] , epl ( europhysics letters ) 63 ( 6 ) ( 2007 ) 798 .
s. chikatamarla , s. ansumali , i. karlin , http://link.aps.org/doi/10.1103/physrevlett.97.010201[entropic lattice boltzmann models for hydrodynamics in three dimensions ] , physical review letters 97 . http://dx.doi.org/10.1103/physrevlett.97.010201 [ ] . http://link.aps.org/doi/10.1103/physrevlett.97.010201 h. yu , s. s. girimaji , l .- s .
luo , http://linkinghub.elsevier.com/retrieve/pii/s0021999105001907[dns and les of decaying isotropic turbulence with and without frame rotation using lattice boltzmann method ] , journal of computational physics 209 ( 2 ) ( 2005 ) 599616 . http://dx.doi.org/10.1016/j.jcp.2005.03.022 [ ] . http://linkinghub.elsevier.com/retrieve/pii/s0021999105001907 c. m. pooley , h. kusumaatmaja , j. m. yeomans , http://link.aps.org/doi/10.1103/physreve.78.056709[contact line dynamics in binary lattice boltzmann simulations ] , phys .
e 78 ( 2008 ) 056709 . http://dx.doi.org/10.1103/physreve.78.056709 [ ] .
h. kusumaatmaja , j. m. yeomans , http://pre.aps.org/abstract/pre/v78/i5/e056709[contact line dynamics in binary lattice boltzmann simulations ] , physical review e 78 ( 5 ) ( 2008 ) 056709 .
http://pre.aps.org/abstract/pre/v78/i5/e056709 c. m. pooley , k. furtado , http://link.aps.org/doi/10.1103/physreve.77.046702[eliminating spurious velocities in the free - energy lattice boltzmann method ] , phys .
e 77 ( 2008 ) 046702 . http://dx.doi.org/10.1103/physreve.77.046702 [ ] .
s. chikatamarla , i. karlin , http://linkinghub.elsevier.com/retrieve/pii/s0378437113000113[entropic lattice boltzmann method for turbulent flow simulations : boundary conditions ] , physica a : statistical mechanics and its applicationshttp://dx.doi.org/10.1016/j.physa.2012.12.034 [ ] . http://linkinghub.elsevier.com/retrieve/pii/s0378437113000113 j. latt , b. chopard , o. malaspinas , m. deville , a. michler , http://link.aps.org/doi/10.1103/physreve.77.056703[straight velocity boundaries in the lattice boltzmann method ] , phys .
e 77 ( 2008 ) 056703 . http://dx.doi.org/10.1103/physreve.77.056703 [ ] . http://link.aps.org/doi/10.1103/physreve.77.056703 s. s. chikatamarla , s. ansumali , i. v. karlin , http://stacks.iop.org/0295-5075/74/i=2/a=215[grads approximation for missing data in lattice boltzmann simulations ] , epl ( europhysics letters ) 74 ( 2 ) ( 2006 ) 215 . http://stacks.iop.org/0295-5075/74/i=2/a=215 d. yu , r. mei , w. shyy , improved treatment of the open boundary in the method of lattice boltzmann equation : general description of the method , progress in computational fluid dynamics , an international journal 5 ( 1 ) ( 2005 ) 312 .
m. junk , z. yang , outflow boundary conditions for the lattice boltzmann method , progress in computational fluid dynamics , an international journal 8 3848 . http://dx.doi.org/10.1504/pcfd.2008.018077 [ ] .
a. j. c. ladd , http://dx.doi.org/10.1017/s0022112094001771[numerical simulations of particulate suspensions via a discretized boltzmann equation part i. theoretical foundation ] , journal of fluid mechanics 271 ( 1993 ) 285309 .
http://arxiv.org/abs/comp-gas/9306004 [ ] , http://dx.doi.org/10.1017/s0022112094001771 [ ] . http://dx.doi.org/10.1017/s0022112094001771 q. zou , x. he , http://link.aip.org/link/?phf/9/1591/1[on pressure and velocity boundary conditions for the lattice boltzmann bgk model ] , physics of fluids 9 ( 6 ) ( 1997 ) 15911598 .
http://dx.doi.org/10.1063/1.869307 [ ] .
http://link.aip.org/link/?phf/9/1591/1 a. kupershtokh , d. medvedev , d. karpov , http://www.sciencedirect.com/science/article/pii/s0898122109001011[on equations of state in a lattice boltzmann method ] , computers & mathematics with applications 58 ( 5 ) ( 2009 ) 965 974 .
http://dx.doi.org/10.1016/j.camwa.2009.02.024 [ ] .
m. bernaschi , l. rossi , r. benzi , m. sbragaglia , s. succi , graphics processing unit implementation of lattice boltzmann models for flowing soft systems , phys .
e 80 ( 6 ) ( 2009 ) 066707 . http://dx.doi.org/10.1103/physreve.80.066707 [ ] .
p. bailey , j. myre , s. d. c. walsh , d. j. lilja , m. o. saar , accelerating lattice boltzmann fluid flow simulations using graphics processors , in : parallel processing , 2009 .
icpp09 . international conference on , 2009 , p. 550557 . c. obrecht , f. kuznik , b. tourancheau , j .-
roux , http://www.sciencedirect.com/science/article/pii/s089812211000091x[a new approach to the lattice boltzmann method for graphics processing units ] , computers & mathematics with applications 61 ( 12 ) ( 2011 ) 3628 3638 .
http://dx.doi.org/10.1016/j.camwa.2010.01.054 [ ] .
http://www.sciencedirect.com/science/article/pii/s089812211000091x c. obrecht , f. kuznik , b. tourancheau , j .-
roux , http://dl.acm.org/citation.cfm?id=1964238.1964257[global memory access modelling for efficient implementation of the lattice boltzmann method on graphics processing units ] , in : proceedings of the 9th international conference on high performance computing for computational science , vecpar10 , springer - verlag , berlin , heidelberg , 2011 , pp .
f. kuznik , c. obrecht , g. rusaouen , j .- j .
roux , http://www.sciencedirect.com/science/article/pii/s0898122109006361[lbm based flow simulation using gpu computing processor ] , computers & mathematics with applications 59 ( 7 ) ( 2010 ) 2380 2392 . http://dx.doi.org/10.1016/j.camwa.2009.08.052 [ ] . http://www.sciencedirect.com/science/article/pii/s0898122109006361 c. obrecht , f. kuznik , b. tourancheau , j .- j .
roux , http://www.sciencedirect.com/science/article/pii/s0898122111001064[multi-gpu implementation of the lattice boltzmann method ] , computers & mathematics with applications ( 0 ) ( 2011 ) .
http://dx.doi.org/10.1016/j.camwa.2011.02.020 [ ] .
j. habich , c. feichtinger , h. kstler , g. hager , g. wellein , http://www.sciencedirect.com/science/article/pii/s0045793012000679[performance engineering for the lattice boltzmann method on gpgpus : architectural requirements and performance results ] , computers & fluids ( 0 ) ( 2012 ) . http://dx.doi.org/10.1016/j.compfluid.2012.02.013 [ ] .
s. albensoeder , h. kuhlmann , http://www.sciencedirect.com/science/article/pii/s0021999105000033[accurate three - dimensional lid - driven cavity flow ] , journal of computational physics 206 ( 2 ) ( 2005 ) 536 558 .
http://dx.doi.org/10.1016/j.jcp.2004.12.024 [ ] .
s. kida , http://jpsj.ipap.jp/link?jpsj/54/2132/[three-dimensional periodic flows with high - symmetry ] , journal of the physical society of japan 54 ( 6 ) ( 1985 ) 21322136 .
http://dx.doi.org/10.1143/jpsj.54.2132 [ ] .
s. s. chikatamarla , c. e. frouzakis , i. v. karlin , a. g. tomboulides , k. b. boulouchos , http://dx.doi.org/10.1017/s0022112010002740[lattice boltzmann method for direct numerical simulation of turbulent flows ] , journal of fluid mechanics 656 298308 .
http://dx.doi.org/10.1017/s0022112010002740 [ ] .
http://dx.doi.org/10.1017/s0022112010002740 b. keating , g. vahala , j. yepez , m. soe , l. vahala , http://link.aps.org/doi/10.1103/physreve.75.036712[entropic lattice boltzmann representations required to recover navier - stokes flows ] , phys .
e 75 ( 2007 ) 036712
http://link.aps.org/doi/10.1103/physreve.75.036712 k. langaas , j. m. yeomans , lattice boltzmann simulation of a binary fluid with different phase viscosities and its application to fingering in two dimensions , european physical journal b 15 ( 2000 ) 133141 .
http://dx.doi.org/10.1007/s100510051107 [ ] .
j. chin , e. boek , p. v. coveney , http://dx.doi.org/10.1098/rsta.2001.0953[lattice boltzmann simulation of the flow of binary immiscible fluids with different viscosities using the shan - chen microscopic interaction model ] , mathematical , physical and engineering sciences 360 ( 1792 ) ( 2002 ) 547558 .
http://dx.doi.org/10.1098/rsta.2001.0953 [ ] . | we present sailfish , an open source fluid simulation package implementing the lattice boltzmann method ( lbm ) on modern graphics processing units ( gpus ) using cuda / opencl .
we take a novel approach to gpu code implementation and use run - time code generation techniques and a high level programming language ( python ) to achieve state of the art performance , while allowing easy experimentation with different lbm models and tuning for various types of hardware .
we discuss the general design principles of the code , scaling to multiple gpus in a distributed environment , as well as the gpu implementation and optimization of many different lbm models , both single component ( bgk , mrt , elbm ) and multicomponent ( shan - chen , free energy ) .
the paper also presents results of performance benchmarks spanning the last three nvidia gpu generations ( tesla , fermi , kepler ) , which we hope will be useful for researchers working with this type of hardware and similar codes .
lattice boltzmann , lbm , computational fluid dynamics , graphics processing unit , gpu , cuda * program summary * _ manuscript title : _ sailfish : a flexible multi - gpu implementation of the lattice boltzmann method + _ authors : _ michal januszewski , marcin kostur + _ program title : _ sailfish + _ code repository : _ https://github.com/sailfish-team/sailfish + _ journal reference : _
+ _ catalogue identifier : _
+ _ licensing provisions : lgplv3 _ + _ programming language : _ python ,
cuda c , opencl + _ computer : _ any with an opencl or cuda - compliant gpu + _ operating system : _ no limits ( tested on linux and mac os x ) + _ ram : _ hundreds of megabytes to tens of gigabytes for typical cases .
+ _ keywords : _ lattice boltzmann , lbm , cuda , opencl , gpu , computational fluid dynamics , python .
+ _ classification : _ 12 , 6.5 + _ external routines / libraries : _ pycuda / pyopencl , numpy , mako , zeromq ( for multi - gpu simulations ) + _ nature of problem : _
+ gpu - accelerated simulation of single- and multi - component fluid flows .
+ _ solution method : _ + a wide range of relaxation models ( lbgk , mrt , regularized lb , elbm , shan - chen , free energy , free surface ) and boundary conditions within the lattice boltzmann method framework .
simulations can be run in single or double precision using one or more gpus .
+ _ restrictions : _ + the lattice boltzmann method works for low mach number flows only .
+ _ unusual features : _
+ the actual numerical calculations run exclusively on gpus .
the numerical code is built dynamically at run - time in cuda c or opencl , using templates and symbolic formulas .
the high - level control of the simulation is maintained by a python process .
+ _ running time : _ problem - dependent , typically minutes ( for small cases or short simulations ) to hours ( large cases or long simulations ) + |
the recent advances in fabrication technologies have promoted the microring resonator devices to the status of preferred platform for many photonic signal processing functionalities , such as optical filtering , dispersion compensation , sensing , or group delay control @xcite .
although microring structures can be realized using different materials , most proposals of passive structures have focused on silicon - on - insulator ( soi ) devices to date , due to obvious reasons of convenience and the available enabling technology . in designing photonic structures at the top - level ,
a waveguide coupler is most conveniently treated as an ideal , lumped component characterized by the numerical value of its coupling constant . however , in the standard fabrication process , the coupling constant is determined mainly by the transversal offset , i.e. by the separation between the coupler waveguides . to meet the accuracy typically demanded for the value of the coupling constant , such separation must be set with a precision of around a few nanometers , which is near the current technological limit . to overcome the difficulty described above , an alternative approach was proposed in @xcite .
this technique consists in varying the coupling length rather than the inter - guide separation , which remains constant .
such variation is in turn obtained by changing the longitudinal offset between the two parallel coupled straight waveguides of fixed length .
photolitographic production is then feasible with this technique as the resolution requirements are strongly relaxed .
as it stands , the longitudinal offset technique unavoidably imposes a certain type of phase imbalance between the fields at the two outputs of any coupling section .
thus , relative phase - shifts appear and accumulate throughout the whole structure . in this work ,
we study the control of the relative group delay that can be attained and its use in various photonic devices . a proof of principle of the operation of these designs is obtained by the simulation of their performance using a finite - difference time - domain ( fdtd ) solver for 2d maxwell equations .
applications of the proposed devices in optical beamforming networks and quantum information systems are discussed .
the geometry of a crow implementing the longitudinal offset technique is shown in figure [ fig::crow ] . with reference to figure [ fig::celda ] , @xmath0 is the length of the straight waveguide sections and corresponds to the coupler length for zero lateral displacement , @xmath1 , is the length of each curved waveguide section of radius @xmath2 .
the half ring length is @xmath3 and will be assumed constant .
the longitudinal offset @xmath4 is defined in the figure and will be regarded as positive whenever the displacement is along the direction of the propagation of the field in that section and negative otherwise .
we can trace the effect of the coupling and the propagation phase in figure [ fig::celda ] to write the field relations @xmath5 rewriting the above expressions as @xmath6permits to obtain the one - cell matrix relation @xmath7and the input - output maxtrix relation for the whole chain @xmath8 \left ( \begin{matrix } e_{n+1}^{+ } \\ e_{n+1}^{-}% \end{matrix}% \right ) .
\label{evolucion}\ ] ] if the input signal is placed at @xmath9 , taking into account that @xmath10 , we can obtain @xmath11 , where @xmath12 is the top diagonal element in the matrix relation .
expression evidences that , for the specific case of a crow chain and within the scope of the ideal model ( wavelength - independent coupling constants ) , the offset - induced phase imbalance is harmless as far as the amplitude response is concerned , which was misinterpreted in @xcite , while a contribution to the group delay that is the cummulative effect of the longitudinal offsets appears in the system response .
this effect can be engineered to yield added functionalities to the structure , as described below .
the geometry depicted in figure [ fig::esquema ] shows that there are two sets of input / output ports available , which are respectively marked as @xmath13 and @xmath14 .
when the transmission path is changed from @xmath13 to @xmath14 , there is a sign flip for every value of @xmath4 along the signal path . according to equation ,
this implies a net change in the group delay when the transmission topology is changed .
therefore , there exist two possible transfer functions implemented by the same physical system that are identical except for a global group delay term that can take two opposite values @xmath15 .
figure [ fig::esquema ] displays the schematic of a @xmath16 ring crow .
the transmission path for the _ fast _ @xmath13 input - output ports is marked with dashed arrows and the _ slow _ path relating the @xmath14 input - output ports is marked with solid arrows .
the transfer functions are @xmath17 where @xmath18 is a common term that would correspond to the transfer function of the same system implemented without the longitudinal offsets . [
cols="^ " , ] it should be noted that each input port ( @xmath13 or @xmath14 ) coincides with the reflection path for the converse configuration .
when both transmission paths are simultaneously used , this circumstance will generally call for taking additional measures at the design stage , depending on the specific application .
however , even if the reflected signals can not in general be ignored , in this specific type of crow filter the reflected spectral components correspond mainly to the band - stop transmission of the filter @xcite , the reflection level thus being negligible when the transmitted signal is predominantly localized in the filter pass - band , which avoids the problem .
the reflected signals can also be used to engineer complex architectures , as detailed below .
using a y - branch to combine both inputs , as shown in figure [ fig::dossalidas ] , we can simultaneously obtain two outputs with the same amplitude spectral shaping and a relative delay .
the shape used for the representation of the signals intends to reflect the existing relation in the filtering due to the common term @xmath18 in . cascading
this type of structures would permit to obtain a series of relatively delayed signals generated from the same input with a given filtering .
the filtering in the first stage will substantially mitigate the effect of reflections in subsequent elements . reversing the scheme , as shown in figure [ fig::tdm ] ,
permits to multiplex two signals with a controlled relative time - delay and simultaneously apply the same filtering to both inputs . the relative delay between the two channels
is given by the expression @xmath19 for equal @xmath20 , the above expression reads @xmath21 .
the connection of the two inputs using a 3 db splitter and the outputs with a 3 db coupler , as shown in fig .
[ fig::mz ] , leads to the implementation of a mach - zehnder ( mz ) interferometer and an overall transfer function @xmath22 this type of response can be exploited to double the fsr of an optical filter by adequately tuning of the relative delay .
the position of a spectral null every other transmission band of the crow filter requires the condition @xmath23 or , for constant @xmath20 , with @xmath24 taking into account that @xmath25 , with @xmath26 defined in fig .
[ fig::celda ] , we can write @xmath27 a design based on the above condition is too sensitive to small deviations of the microring parameters in the fabrication process .
a more robust design can be obtained by seeking the desired response only in a limited bandwidth .
if @xmath28 , with @xmath29 a rational number close to @xmath30 , the mz and crow resonance conditions will coincide at a particular band of the crow periodic response and an approximate cancellation will happen in a certain bandwidth that will depend on the value of @xmath29 .
fabrication tolerances in this case will affect the actual response of the filter , but a good approximation to the original design can be obtained in the presence of small deviations from the ideal design parameters . in this case , the value of a constant offset for a given coupler length is given by the expression @xmath31 by adding a phase control to one the output branches , the mz configuration can be used to produce and process quantum superpositions of time - bin states for quantum information applications @xcite . instead of a design based on equation , as in the previous fsr doubling application ,
the offsets are calculated such that produces the required time - bin separation which has to be larger than the pulse width .
this proposal permits an integrated optics implementation with a precise control of the photon waveform through the filter spectral shaping and is better suited to relative delays in the picosecond range @xcite .
we now consider a longitudinal offset group delay shifter ( logds ) built by cascading various systems of the type described in fig .
[ fig::dossalidas ] .
if the transmission bandwidth is small compared to the fsr , multiple delayed outputs of the signal present in a single wavelength channel will be produced at the output , whereas most of the spectral content present in the optical input will be reflected .
this can be used for the realization of complex wdm architectures for generating multiple delayed signals at different channels , as illustrated in figure [ fig::adddrop ] , which could find application , for instance , in advanced optical beamforming networks @xcite .
the frequency selectivity is strictly required only in the first stage of the logds tree that can be implemented with a high @xmath32 single ring stage . after the extraction of the drop channel , the rest of the wdm multiplex is reinjected in the optical transmission line using a circulator . using this procedure
, @xmath33 delayed replicae of the signal at each wdm channel are produced at each respective output .
the systems depicted in figures [ fig::dossalidas ] , [ fig::tdm ] and [ fig::mz ] , for @xmath16 ring crows have been analyzed using the fdtd method @xcite . in each case
, independent simulations have been performed for the subsystems bounded by the dashed - dotted lines in the figures , and the computed transfer functions have been used to obtain the total response .
the computations have been performed on an equivalent 2d reduced model obtained using the effective index method .
all the waveguides in our devices are @xmath34 wide and an effective index of @xmath35 has been assumed .
the bent waveguide sections have an inner radious of @xmath36 in all cases .
the evanescent couplers have been implemented with a waveguide separation of @xmath37 .
the geometry used for the @xmath38 db y - branch couplers and splitters is depicted in fig .
[ fig::acoplador ] .
the coupling and splitting responses obtained are very close to ideal 3 db in all the cases .
figure [ fig::espectros ] ( a ) shows the amplitude responses and fig .
[ fig::espectros ] ( b ) the group delays for the @xmath13 ( solid line ) and @xmath14 ( dashed line ) input / output transfer functions of a @xmath16 ring system with @xmath39 and constant @xmath40 .
as predicted by the expressions , the amplitude responses are nearly identical , and the group delay difference between the two outputs is very close to a constant value of approximately @xmath41 .
the variation of the coupling coefficient with the wavelength is apparent in the results .
figure [ fig::espectromz ] shows the transmission amplitude spectrum for a single input / output ( dashed line ) and the response from the mz configuration in solid line , showing the fsr doubling effect . in this case ,
@xmath42 and @xmath43 .
in this paper we wave analyzed a crow chain realized by means of the longitudinal offset technique .
the inherent accumulated phase imbalance has been used to propose several architectures for photonic processing based on an alternating chain .
we have described the generation of two identically filtered , truly delayed opticals signals , as well as their multiplexing with a controlled time - delay .
a mach - zender structure which permits to double the fsr has also been presented .
all the proposed devices have been proved numerically through 2d fdtd simulation , which accounts for non - ideal effects such as coupler wavelength - dependence and radiation losses , the agreement with the expected results being excellent .
potential applications in optical beamforming antennas and quantum information systems have been discussed . c. k. madsen , g. lenz , a. j. bruce , m. a. cappuzzo , l. t. gomez , and r. e. scotti , integrated all - pass filters for tunable dispersion and dispersion slope compensation , ieee photon .
technol . lett .
16231625 , dec . 1999 .
j. d. domnech , p. muoz , and j. capmany , the longitudinal offset technique for apodization of coupled resonator optical waveguide devices : concept and fabricaiton tolerance analysis , opt .
express , vol .
2105021059 , nov .
domnech , p. muoz , and j. capmany , transmission and group - delay characterization of coupled resonator optical waveguides apodized through the longitudinal offset technique , opt .
2 , pp . 136138 , jan .
2011 .
p. c. humphreys , b. j. metcalf , j. b. spring , m. moore , x - m .
jin , m. barbieri , w. s. kolthammer , and i. a. walmsley , `` linear optical quantum computing in a single spatial mode , '' phys .
111 , 150501 , oct . 2013 .
a. f. oskooi , d. roundy , m. ibanescu , p. bermel , j. d. joannopoulos , and s.g .
johnson , `` meep : a flexible free - software package for electromagnetic simulations by the fdtd method , '' computer physics communications , vol .
181 , 687702 , jan . 2010 .
l. zhuang , c.g.h .
roeloffzen , r.g .
heideman , a. borreman , a. meijerink , w. van etten , `` single - chip ring resonator - based @xmath44 optical beam forming network in cmos - compatible waveguide technology , '' ieee photon .
lett . , vol .
15 , aug . 2007 | we study the differential group delay that can be obtained in coupled micro - ring resonator waveguides with longitudinal offset couplers .
various devices based on this effect are proposed and analyzed using a finite - differences time - domain method and their potential applications are discussed . |
over 30 years ago rafael vera introduced fascinating concepts for mass @xcite .
they have not gained widespread acceptance , possibly because they were thought to be incompatible with general relativity ( gr ) . in this paper
, a careful accounting of observer reference frames shows that vera s ideas are consistent with gr in a schwarzschild field .
this combination is an attempt to begin to resolve the long - standing perplexity over the localization of gravitational energy .
one of vera s ideas is that the rest mass of a particle increases with its gravitational potential @xcite . therefore , gravitational energy would be localized with the masses of bodies .
this must occur in such a way that proper ( locally measured ) rest mass is invariant with position , as perceived from a local observer s mass reference frame .
the mass / velocity relationship for a body falling along a geodesic in a schwarzschild field is shown to be the old - fashioned expression for relativistic mass from special relativity .
this new result is independent of the choice of coordinates , as confirmed for the isotropic and harmonic metrics .
this approach strengthens the correspondence between special relativity and gr , and rehabilitates the undervalued concept of relativistic mass .
reference frames are usually associated with spatial coordinates and time .
the introduction of reference frames for mass is essential for solving this problem .
applying vera s concepts to time - varying gravitational fields is problematic .
it is shown that his concept of an invariant falling mass does not apply to such fields .
unlike static fields , time - varying fields induce a net long - range energy transfer between bodies . the paper proceeds with the following steps :
a dimensionless scalar @xmath0 is introduced ; its important identities are found .
a mass transformation is derived for a one - dimensional , static gravitational field , then generalized to transformations between mass reference frames .
the transformations are applied to the schwarzschild solution and other metrics .
the mass / velocity relationship for falling bodies is evaluated for radial and oblique motion .
it is shown that the proposed interpretation conserves energy and mass within a schwarzschild field while the conventional interpretation does not .
time - varying fields are studied and limitations on the proposed interpretation are imposed . as with maxwell s equations , in which a time - varying electric field causes effects
not seen in static fields , a time - varying gravitational field also causes effects not exhibited by static fields .
this work is a matter of interpretation and is intended to complement gr .
what happens to the energy used to lift a body in a gravitational field ? during the development of gr , einstein derived an expression @xmath1 for the stress - energy of the gravitational field .
that expression is a pseudotensor that varies with the choice of coordinates in an unseemly way .
( for a review see @xcite . )
many attempts have been made to better understand gravitational energy in gr .
tolman @xcite defined gravitational energy as a separate term in the energy - momentum tensor @xmath2 . in 1935 , whittaker @xcite proposed the term potential mass " to describe the contribution of gravitational potential energy to the gravitating mass of a particle .
after much deliberation , the interpretation arose that gravitational energy is not localized in gr @xcite .
thus , it is meaningless to ask where gravitational energy resides . in an influential textbook @xcite , misner , thorne and wheeler
sought to resolve this quandary with the words , anybody who looks for a magic formula for ` local gravitational energy - momentum ' is looking for the right answer to the wrong question . unhappily
, enormous time and effort were devoted in the past to trying to ` answer this question ' before investigators realized the futility of the enterprise . "
yet bondi insists that non - localizable forms of energy are inadmissible in gr @xcite .
although today most relativists seem satisfied with the conventional treatment of gravitational energy as unlocalizable , there remains considerable ambiguity in the interpretation of gravitational energy within gr .
many authors believe that gravitational energy , and even mass , are stored in fields @xcite .
some conclude that gravitational self - energy , which is a type of gravitational energy , contributes to the masses of bodies @xcite .
vera @xcite , savickas @xcite , and ghose and kumar @xcite interpret rest mass as depending on local gravitational potential , while ohanian and ruffini @xcite argue that rest mass is a constant independent of space and time . since energy can not simultaneously be distributed in fields and contribute to particle masses , some of these views appear to be incompatible .
hayward states , it comes as a surprise to many that there is no agreed definition of gravitational energy ( or mass ) in general relativity " @xcite . a few dedicated researchers continue to study gravitational energy and related concepts such as quasi - local energy momentum ( e.g. , @xcite ) , driven partly by the belief , as stated by mirshekari and abbassi @xcite , that one of the old and basic problems in general relativity which is still unsolved is the localization of energy . "
as proposed or implied by others ( e.g. , @xcite ) it is hypothesized that the energy used to lift a body in a static gravitational field increases its rest mass .
it is also assumed that the mass of a body remains constant while falling in a static field @xcite .
this interpretation is illustrated in fig .
1 with the following steps : the energy @xmath3 used to lift a body increases its rest mass by @xmath4 . when the lifted body is allowed to fall from a higher elevation , it maintains a constant mass during its free - fall , equal to its mass at the point it began its fall .
as the body passes its starting point with some velocity , it has a greater mass than an identical body at rest there , in general agreement with special relativity .
when the fall is stopped , the body loses its extra mass and returns to the rest mass it had before being lifted .
the mass that the body loses during its impact is transferred to other bodies ( and to itself ) in the form of vibrational energy , heat , elastic or inelastic energy , etc .
total mass would always be conserved during lifting , falling , and collisions .
no mass or energy would be stored in the gravitational field . with this interpretation
, the rest masses of all bodies would depend on position in a gravitational field .
in accordance with the equivalence principle , any experiments conducted at different elevations should be unable to detect any difference in mass , since all objects would be affected in exactly the same proportion .
this interpretation does for potential energy what special relativity did for kinetic energy gives it a mass equivalent .
the history of various concepts of gravitational energy is a fascinating topic not covered herein .
the concepts shown in fig . 1 appear to work in static fields .
however , it is shown later that the mass of a body can not remain constant while falling in a _ time - varying _ field , in which gravitational energy must be conveyed between bodies via gravitons and/or gravitational waves .
the essential element of fig .
1 that remains applicable in time - varying fields is that energy and mass are primarily localized as part of the masses of bodies rather than distributed in fields . for the purpose of this paper ,
no distinction needs to be made between inertial mass , active gravitational mass , and passive gravitational mass .
they are assumed to be the same @xcite .
also , the term gravitational energy " refers to the gravitational potential energy of a body and is not to be confused with the more general gravitational energy - momentum tensor @xmath2 . as an example of the conservation of mass after an inelastic collision ,
thermal energy increases the velocity of the molecules of a body , with corresponding relativistic mass given by @xmath5 so that the mass of a falling body is in part converted into the mass associated with additional kinetic energy @xmath6 of the molecules @xcite .
after the body cools off , it returns to its original mass .
although the terms rest mass " and relativistic mass " have fallen out of favor @xcite , they are used herein out of necessity . here the term rest mass " describes the mass of a motionless body in any reference frame fixed to a central mass in a static gravitational field . since reference frames span different gravitational potentials , rest masses may vary with elevation .
proper rest mass " is the rest mass of a body when it is co - located with the observer measuring it ; proper rest mass is presumed to be invariant .
relativistic mass " refers to the mass of a body moving relative to an observer fixed with respect to the central mass , as perceived from that observer s mass reference frame . rather than dwell on semantic arguments over the acceptability of the term relativistic mass , " it is more productive to focus on physical arguments and their consequences .
mtw @xcite derived localized gravitational energy for the schwarzschild solution .
however , their approach differs from vera s concepts @xcite .
several variable mass theories of gravity have been proposed @xcite .
however , they do not directly relate changes in mass with potential energy .
the standard model extension ( sme ) has been developed partly to study possible variations in fundamental constants @xcite . in the sme
, the relative masses of different materials might vary with gravitational potential , in disagreement with gr .
that possibility is not considered herein .
modern versions of the e@xmath7tv@xmath7s experiments @xcite support the hypothesis that different materials are affected by gravitational fields in the same way .
because it is necessary to understand static fields before exploring time - varying fields , this work begins with a detailed analysis of the schwarzschild solution . for static gravitational fields it is useful to define the scalar dimensionless ratio @xmath0 as follows : @xmath9 where @xmath10 is
the time interval between two events measured at some point @xmath11 and @xmath12 is the time interval between those two events measured from some other point @xmath13 , using clocks that are motionless relative to the field . if @xmath14 , time passes more slowly at point @xmath11 than at @xmath13 .
( in a static field it is easy to account for the time for a signal to travel between @xmath11 and @xmath13 . ) from the form of ( [ betadef ] ) the following identities for @xmath8 can be found : @xmath15 @xmath16 and @xmath17
the next step is to derive a general transformation for rest mass in a static , one - dimensional gravitational field , from the perspective of observers who are motionless relative to the field .
based on einstein s equivalence principle , an observer at any location should measure the same rest mass for a given body when it is at her location .
this could be achieved experimentally by comparing the mass of the body to the mass of a kilogram standard carried alongside .
consider an observer located at some arbitrary reference elevation @xmath18 .
let the proper rest mass of some body be @xmath19 .
the mass of that body when it is at some other elevation @xmath20 , relative to the reference frame at @xmath18 , is defined to be @xmath21 an important notation for what follows .
the difficulties of measuring mass at another location must be overlooked for now .
when the body is at @xmath18 , the observer there will measure its mass to be @xmath22 . when the body is carried to some other elevation @xmath20
, an observer there will measure its mass to be @xmath23 .
consider a static field causing a proper acceleration @xmath24 in the @xmath25 direction .
let @xmath20 represent proper distance , which will generally differ from the coordinates used in some solution to the field equations .
consider a body with proper rest mass @xmath19 being slowly lifted against the field .
then the proper energy required to lift the body some small distance @xmath26 is given by @xmath27 in the local reference frame at @xmath20 , the small increase in mass as the body is lifted is given by @xmath28 from the reference frame at @xmath18 , the body at @xmath20 has mass @xmath21 , while an observer at @xmath20 measures the mass of the body to be @xmath19 .
the ratio of those masses will apply to any body or mass increment at @xmath20 .
therefore , the incremental mass @xmath29 at @xmath20 observed from @xmath18 and from @xmath20 are related as follows : @xmath30 combining ( [ dmxx ] ) and ( [ mratios ] ) yields @xmath31 this can be integrated to yield the general mass transformation between elevations in a one - dimensional , static gravitational field : @xmath32 this equation , also found in @xcite , indicates that the mass of a body lifted against the field to @xmath20 is greater than the mass of an identical body at @xmath18 .
since there is no preferred elevation in a gravitational field , point @xmath18 can be selected to represent any point .
the equation also shows that proper mass does not vary with elevation . in a weak field , ( [ mxox ] ) reduces to @xmath33 , where @xmath34 is the classical gravitational potential energy of the body . next the general transformations between mass reference frames can be found .
consider the general case of two elevations @xmath11 and @xmath13 .
let @xmath35 and @xmath36 . from the form of the preceding equation , it can be shown that any rest mass @xmath19 at @xmath13 relative to the mass reference frame at @xmath11 , and vice - versa , satisfies identities that are similar to those for @xmath8 , so that @xmath37 @xmath38 and @xmath39 where @xmath40 , for example , is the mass of a body at @xmath11 , with proper rest mass @xmath19 , relative to the mass reference frame at @xmath13 .
observers at different gravitational potentials experience different mass reference frames that are related by these formulas .
next , the mass transformation is applied to the exterior schwarzschild solution of gr , described by the following metric : @xmath41 where @xmath42 the coordinate time @xmath43 is the time interval of a stationary clock at infinity and @xmath44 , where @xmath45 is proper time of a clock corresponding to the spacetime interval @xmath46 . setting @xmath47 in ( [ ds2 ] ) for a motionless clock at @xmath48 yields @xmath49 , where @xmath50 is the time interval of a stationary clock at @xmath48 and @xmath51 is the time of a stationary clock at infinity .
thus , @xmath52 is the ratio of the time rates at @xmath48 and infinity , corresponding to the definition for @xmath8 in ( [ betadef ] ) . that is the origin of the subscripts in ( [ brinf ] )
. equation ( [ brinf ] ) provides @xmath8 at @xmath48 relative to an observer at infinity . at some other radius @xmath53 , @xmath54 from identities ( [ bid2 ] ) and ( [ bid3 ] ) an expression for the @xmath8 s relating elevations @xmath48 and @xmath53
can be found : @xmath55 besides using the identities , this can also be derived from the relative time rates at @xmath48 and @xmath53 .
since the mass transformation ( [ mxox ] ) is based on proper acceleration @xmath24 , the next step is to find proper acceleration relative to a motionless observer in a schwarzschild field . for radial motion , hobson _
_ @xcite find the following geodesic equation of motion : @xmath56 where @xmath57 . in the general case , @xmath58 is the time measured by a clock that may be either moving or stationary . in the case of a stationary clock at @xmath48 ,
the terminology @xmath59 is used to differentiate between those two cases . for a motionless clock at @xmath48 , @xmath60 .
coordinate @xmath48 in the schwarzschild metric has little physical significance . of greater relevance
is the proper distance given by @xmath46 , which in the general case may correspond to a moving meter stick or one that is motionless relative to the central mass . to distinguish between those two possibilities , variable @xmath20 is used to denote proper distance in the reference frame of a motionless observer at @xmath48 . by setting @xmath61 , letting @xmath62 in ( [ ds2 ] ) , and using the appropriate sign for a spacelike spacetime interval , the following relationship is found : @xmath63 which relates proper distance @xmath64 to the radial coordinate distance @xmath65 . using the resulting formula for @xmath64 in place of @xmath65 in ( [ rddot ] ) ,
@xmath66 consider a body at the instant it is released to free - fall , before it has attained appreciable velocity .
then @xmath67 , the proper velocity is @xmath68 , and only the right hand term in ( [ drdot ] ) contributes . using the resulting equation in combination with ( [ rddot ] ) and
( [ drdx ] ) , the proper acceleration of gravity for a slowly falling body at @xmath48 in a schwarzschild field is given by @xmath69 this is the acceleration that must be resisted when slowly lifting a body in the field .
note that @xmath70 is positive towards the central mass .
suppose a small body with proper rest mass @xmath19 is lifted radially from @xmath48 to @xmath53 in a schwarzschild field .
the mass of the body at @xmath53 , relative to the reference frame at @xmath48 , can be found by substituting ( [ drdx ] ) and ( [ gr ] ) into ( [ mxox ] ) , which yields @xmath71 the surprisingly simple solution to this integral is @xmath72 since @xmath73 if @xmath74 , the mass of the lifted body at @xmath53 is perceived to be greater than @xmath19 from the reference frame at @xmath48 . in a weak field
, @xmath75 is its rest mass plus ( or minus ) the mass equivalent of the change in classical potential energy .
if @xmath48 is chosen to be the reference frame at an infinite distance from the central mass and if @xmath76 , then the relative mass of a body at any point @xmath48 perceived from the reference frame at infinity is given by @xmath77 thus , stationary bodies in the vicinity of a central mass have a decreased mass relative to a reference frame at infinity .
it is apparent that @xmath52 plays the role of scale factor for mass as well as time .
the preceding results for the schwarzschild solution suggest the following general transformation for rest mass in any static gravitational field : @xmath78 this generalization , offered without proof , does not depend on the choice of coordinates ; it is not used again in this paper .
the proof of ( [ mxo ] ) for any static field @xmath24 requires topics not covered here .
most studies of localization focus on the gravitating mass rather than test particles . nevertheless , mass reference frames reveal an interesting property of the gravitating mass in a schwarzschild field . in some ways
, the schwarzschild metric is written from the perspective of a reference frame at infinity : coordinate time @xmath79 represents a clock at infinity and radial distance @xmath48 converges to flat space coordinates at infinity .
therefore , the central mass at the origin should be @xmath80 , its mass from the reference frame at infinity . from a reference frame at @xmath48 ,
all masses appear increased from ( [ mrrb ] ) , so that @xmath81 .
substituting this into ( [ gr ] ) yields @xmath82 where @xmath48 is the proper circumference divided by @xmath83 .
thus , newton s law of gravity is preserved for any local observer if the magnitude of the central mass in his reference frame is used .
this result is independent of the choice of coordinates .
now that the mass transformation has been derived for a schwarzschild field , the mass / velocity relationship for a falling body can be found . equation ( [ mrrb ] )
provides the mass of a body that has been lifted from @xmath48 to @xmath53 .
based on the concept in fig .
1 , if that body is allowed to fall from @xmath53 back to @xmath48 , it will maintain the same mass during its fall .
it will therefore arrive at @xmath48 with the relativistic mass @xmath84 given by ( [ mrrb ] ) . in gr it is straightforward to derive the velocity of a body falling radially from @xmath53 to @xmath48 in a schwarzschild field .
the radial geodesic equations of motion @xcite are @xmath85 and @xmath86 expressing @xmath87 in terms of @xmath88 and setting constant @xmath89 so that the coordinate velocity @xmath90 at @xmath53 , the coordinate velocity of the body when it arrives at @xmath48 is @xmath91 using prior formulas , the proper velocity @xmath92 of a body at @xmath48 can be related to its coordinate velocity as follows : @xmath93 where @xmath92 is measured with respect to proper coordinates @xmath20 fixed relative to the central mass .
the preceding two equations can be combined to yield the proper velocity of a body falling from @xmath53 , measured at @xmath48 by a motionless observer there : @xmath94 using ( [ mrrb ] ) to provide the locally measured mass of the falling body @xmath84 , it is found that the relationship between proper velocity and total mass of the falling body at @xmath48 is given by @xmath95 this is the well known relativistic mass / velocity relationship of special relativity , derived for a body falling in a schwarzschild field .
the mass / velocity relationship of ( [ mv ] ) was derived for a body released from rest at @xmath53 .
what if some initial , proper radial velocity @xmath96 is imparted to the body at @xmath53 as it begins its fall ?
then an observer at @xmath53 will perceive the body to have an increased mass as it begins its fall , as given by @xmath97 from the reference frame at @xmath48 , the mass of the body at @xmath53 will be increased above its rest mass @xmath75 in the same proportion .
that is the value of the body s mass that remains constant during the fall from @xmath53 to @xmath48 , and will be measured as @xmath98 when it arrives at @xmath48 . using the equation of motion ( [ rdot2 ] ) and other equations previously derived , it can be shown that , regardless of any initial velocity imparted to the body at @xmath53 , the mass / velocity relationship of ( [ mv ] ) applies when the body arrives at @xmath48 .
for static fields there are two distinct physical explanations for gravitational redshift that can be found in the literature .
the most common view is that the redshift is caused by a loss ( gain ) in the energy of a photon as it climbs ( descends ) in a gravitational field .
another view , argued by okun , selivanov and telegdi @xcite and noted by others ( e.g. , @xcite ) is that the redshift is caused by a change in the energy levels of lifted atoms .
these viewpoints are incompatible ; if both were true the redshift would be doubled .
the concept of mass reference frames supports the view that gravitational redshift is caused by a change in the energy levels ( masses ) of lifted atoms . the energy of a photon would not change while traveling along a null geodesic in a static gravitational field @xcite .
if a photon is emitted at elevation @xmath18 with proper energy @xmath99 , it will arrive at a higher elevation @xmath20 with the same energy . however , all stationary masses and energies at @xmath20 are increased relative to @xmath18 . to an observer at @xmath20 ,
the photon arrives with a reduced energy compared to a photon at @xmath20 emitted by the same process . using a uniform field approximation of @xmath100 and @xmath101
, it can be shown that a photon with proper energy @xmath99 emitted at @xmath18 arrives at @xmath20 with a frequency shift of @xmath102 relative to the frequency the photon would have if it were emitted at @xmath20 with proper energy @xmath99 .
the proper frequency of the photon at @xmath20 is established by its proper energy there , in combination with planck s constant , and is not an inherent property of the photon
. experimental evidence of the local position invariance of planck s constant @xcite is consistent with this interpretation .
a review of gravitational redshift experiments ( e.g. , @xcite ) indicates that they are unable to distinguish whether photons lose energy during travel , or appear to be redshifted due to changes in mass / energy reference frames .
the proposed interpretation also satisfies the thought - experiment of nordtvedt on the conservation of energy in gravitational redshift experiments @xcite .
the proposed existence of mass reference frames appears to be consistent with predictions and observations of gravitational redshift .
this discussion applies only to static fields .
the time - varying field created by the expansion of the universe causes a cosmological redshift that is distinct from that of static fields .
consistent with birkhoff s law @xcite , many metrics describe the field of a central mass , each related to the schwarzschild solution by a coordinate transformation . two such metrics are the isotropic and harmonic metrics @xcite .
the isotropic metric for a central mass can be written in the form @xmath103 where @xmath104 and @xmath105 if a body is lifted from @xmath106 to @xmath53 in the isotropic metric , and then allowed to fall back to @xmath106 , the mass / velocity relationship of ( [ mv ] ) is found at @xmath106 .
if this exercise is repeated for the harmonic metric @xcite , that result is found once again .
what this confirms is that mass reference frames are independent of the choice of coordinates , as should be expected .
it can also be shown that at the same point in space , as determined by well - known coordinate transformations between the three metrics ( e.g. , @xcite ) , both @xmath52 and the local gravitational acceleration @xmath70 are identical for all three metrics .
what this suggests is that , in static fields , gravitational energy is stored as part of the masses of bodies in accordance with the simple scalar function @xmath0 , which transforms in a well - behaved way and is independent of the choice of coordinates .
up to this point , only radial motion relative to a central mass has been considered . next it is shown that motion in any oblique direction
is not only consistent with the concept of mass reference frames , but corresponds to the schwarzschild metric .
consider a stationary observer fixed at some radius @xmath48 from a central mass . to that observer , special relativity and ( [ mv ] )
indicate that a moving or orbiting body at @xmath48 should have a mass @xmath107 of @xmath108 where @xmath107 denotes relativistic mass @xmath98 of the body at @xmath48 in the reference frame at @xmath48 . from a reference frame at infinity
, the mass of the moving body at @xmath48 appears smaller , since all masses at @xmath48 appear decreased relative to those at infinity by a factor @xmath52 , from ( [ mrinf ] ) . from that frame
, the mass of the moving body at @xmath48 is perceived to be @xmath109 where the subscripts on @xmath52 are dropped for now .
if the body is in free - fall around the central mass , moving in any direction with no other external forces , its total mass will remain constant during its fall or orbit .
therefore , the body s total mass from the reference frame at infinity is constant as follows : @xmath110 where @xmath111 is a constant of the motion .
this formula makes it easy to find the velocity at any radius @xmath48 . if @xmath112 the body has sufficient relativistic mass and associated velocity to escape the central mass . if @xmath113 the orbit will be bound .
this result is similar to that of @xcite . using the expressions @xmath114 and @xmath115 , ( [ mvk ] )
can be rearranged to yield @xmath116 where @xmath43 is the time of a clock at infinity .
let @xmath45 represent the time of a clock on the moving body .
those two times are related by @xmath117 using ( [ mvk ] ) and ( [ dtaudt ] ) , it can be shown that the left hand side of ( [ metk ] ) is @xmath118 , so that @xmath119 as previously defined , @xmath64 is proper length in the motionless reference frame at @xmath48 . for
any given metric , @xmath64 can be related to the coordinates used in that solution .
for example , for the standard form of the schwarzschild solution @xmath120 this can be found from the metric at a synchronous instant in time @xmath121 , which gives the spacelike spacetime interval @xmath122 when the appropriate sign is used . substituting this into ( [ ds2x ] ) yields @xmath123
this is the schwarzschild metric of ( [ ds2 ] ) .
other metrics , like the isotropic and harmonic metrics , can similarly be obtained when the corresponding expression for @xmath124 is used .
although this is not an independent derivation of the metrics , it demonstrates the correspondence between mass reference frames and the metrics of gr . since this result
is based on ( [ mvrr ] ) it also indicates that a body moving in any oblique direction will exhibit the mass / velocity relationship of special relativity in a local frame .
the derivation of ( [ mv ] ) confirms that mass and energy are conserved during a lift and fall cycle under the proposed interpretation .
that provides a compelling argument to support the existence of mass reference frames : as shown below , the prevailing interpretation of gravitational energy in gr does not conserve energy and mass during a lift and fall .
consider a body being slowly lifted in a schwarzschild field .
the mass equivalent of the proper lift energy expended over a small distance @xmath64 is given by ( [ dmxx ] ) .
if total mass is indeed invariant between all elevations , then @xmath125 , contrary to ( [ dmxo ] ) . then the mass equivalent of the energy used to lift a body from @xmath18 to @xmath20 would be given by @xmath126 if this is integrated in a schwarzschild field , the total mass equivalent of the energy expended while lifting the body from @xmath48 to @xmath53 would be given by @xmath127 in weak fields , @xmath128 approaches @xmath75 in ( [ mrrb ] ) .
when the body falls from @xmath53 to @xmath48 it must arrive with a relativistic mass given by ( [ mv ] ) in the reference frame of a stationary observer at @xmath48 .
it is easy to show that the lift energy / mass and the additional relativistic energy / mass acquired by the body when it returns to @xmath48 are not conserved , as follows : @xmath129 it takes less energy / mass to lift a body than arrives back at the starting point .
thus , without the use of mass reference frames , energy and its mass equivalent are not conserved within a lift / fall cycle .
this does not indicate any flaw in gr , but suggests that mass reference frames are necessary to complement the theory .
to the extent that distant stars contribute to the local gravitational potential at any point , the proposed interpretation implies that the magnitude of the inertia of any body is directly influenced by masses throughout the universe .
thus , the proposed interpretation brings gr into closer harmony with mach s principle . nevertheless , the concepts developed herein are more likely to fuel the recurring debate on mach s principle than to resolve it .
this paper focuses on static fields , which encompass the most important tests of gr . nevertheless , it is important to show how the proposed interpretation is limited in time - varying fields .
such fields introduce issues that are discussed only briefly here .
the analysis of time - varying fields begins with the simple thought - experiment shown in fig .
2 . consider two masses @xmath130 located deep in intergalactic space and separated by distance @xmath131 .
initially , both masses are approximately co - moving with the cosmological fluid .
the attractive gravitational force between the bodies is @xmath132 .
imagine that a long cable is attached to @xmath133 , with the other end attached to a winch w secured to a distant , massive object .
the winch pulls on the cable with a tension of @xmath134 , so that mass @xmath133 moves away from @xmath135 at the same rate that @xmath135 falls towards @xmath133 . both masses will accelerate towards @xmath136 while maintaining distance @xmath131 between them .
once they have achieved some large velocity @xmath137 , let mass @xmath133 be pulled a great distance away from @xmath135 , leaving @xmath135 moving at velocity @xmath137 in empty space . in this example ,
body @xmath135 has been accelerated by gravity to velocity @xmath137 , imparting relativistic energy / mass to the body . where has that energy come from ?
if vera s interpretation of falling bodies from fig .
1 is applied , then the total mass remains constant and the original rest mass @xmath19 has been converted into relativistic mass @xmath138 . if the body s motion is then stopped , it will lose the relativistic part of its mass and drop to a new rest mass of @xmath139 .
then @xmath140 in a region where the mass reference frame is unchanged .
the body s rest mass has been decreased by the motion induced in fig .
conversely , the rest mass of @xmath133 would necessarily increase
. that interpretation would require the rest masses of all bodies and elementary particles to vary with their prior motion in time - varying gravitational fields .
in the absence of experimental evidence of an appreciable variability in the rest masses of protons or electrons , it is evident that vera s interpretation of a falling body does not work in a time - varying field .
fortunately , all is not lost .
the most essential of vera s concepts is that gravitational energy is stored as part of the masses of bodies rather than in empty fields . in fig .
2 , the energy expended by the winch travels directly to @xmath133 via the cable .
half of that energy then travels through space almost immediately to @xmath135 , limited only by @xmath141 .
no energy would be stored in fields over the long - term . what can be deduced from this exercise
is that gravitational energy must travel through space in a time - varying field .
bondi described the movement of energy across the empty space between two bodies @xcite .
in static fields , energy may be exchanged between gravitating masses , but the net change in the mass of a test particle during free - fall should be zero . thus the proposed interpretation of gravitational energy is compatible with time - varying fields , but is limited accordingly .
this approach is consistent with the transport of energy by gravitational waves , and also suggests that gravitons , if they exist , should convey mass / energy . the transfer of energy in fig .
2 is fundamental and does not arise from the concept of mass reference frames . to show how the proposed interpretation works in time - varying fields , it is next applied to the most important cosmological metric of gr the robertson - walker metric , which can be written in the following form : @xmath142,\ ] ] where @xmath79 is the cosmological time of all observers co - moving with the cosmological fluid , @xmath143 is the scale factor representing the expansion of the universe , and @xmath144 represents radial coordinate distance .
proper distance @xmath64 in the radial direction is related to @xmath144 by @xmath145 at any synchronous time @xmath79 .
suppose that at some time @xmath146 a particle is given an initial proper radial velocity @xmath147 relative to the cosmological fluid , . at some later time @xmath79
, it will arrive at a second point with proper velocity @xmath148 . from the geodesic equations of motion ( e.g. , @xcite )
, the proper radial velocity of the body at @xmath79 can be found : @xmath149 as expected , the equation confirms that if a particle starts out with an initial velocity @xmath150 relative to the cosmological fluid it will continue to move with the cosmological fluid without relative acceleration .
moreover , a photon with an initial proper velocity of @xmath151 will maintain a proper velocity of @xmath141 . as previously discussed for the schwarzschild solution , suppose a particle is slowly lifted " between two points .
what ( [ vr ] ) indicates is that no matter how small the initial velocity @xmath152 , the particle s proper velocity will never subsequently fall to zero
. it will forever continue to drift slowly across the universe .
this means that all points in the expanding universe have the same gravitational potential .
there is no gravitational force or pseudo - force to lift against when slowly moving a particle .
therefore , all observers co - moving with the cosmological fluid experience the same mass reference frame .
however , ( [ vr ] ) shows that as the universe expands , the proper velocity of of the particle decreases monotonically below its initial proper velocity @xmath147 .
therefore , its relativistic mass will decrease with time , relative to cosmological observers .
using ( [ vr ] ) the relativistic mass of a particle as a function of time is @xmath153 where @xmath19 is its rest mass .
thus the relativistic mass decreases with time .
the change in relativistic mass with time can be found from ( [ vr ] ) and ( [ mt ] ) @xmath154 this loss of relativistic mass is directly proportional to the rate of expansion of the universe . in a static universe ,
@xmath155 and there is no loss of mass .
this result is consistent with vera s interpretation that the mass of a falling body remains constant in a static field , while requiring the transfer of mass in a time - varying field .
it is important to recognize that the loss of mass in ( [ dmvdt ] ) does not arise from the concept of mass reference frames , but is a direct consequence of the robertson - walker metric and is implicit in gr .
so what happens to the relativistic mass lost by a body moving relative to the cosmological fluid ?
the concepts developed herein suggest that mass / energy must travel through space to other bodies , most likely at the speed of light .
for the robertson - walker metric , the receiving bodies would be the masses comprising the expanding universe .
similarly , this approach can be applied to any metric to quantify the mass / energy lost or gained by falling bodies .
the concepts developed herein only begin to explore this topic .
equation ( [ mt ] ) provides another way to calculate cosmological redshift .
if @xmath156 is the relativistic mass of a particle at time @xmath146 when it has initial velocity @xmath147 , then the limit of @xmath157 as @xmath158 is @xmath159 , which in terms of energy corresponds to the expected cosmological redshift of a photon @xcite .
the proposed interpretation of gravitational energy provides a means to explore the disposition of gravitational energy in solutions to the field equations .
the localization of gravitational energy in static fields is explained by simple concepts in which the energy is localized with the masses of bodies .
observers at different gravitational potentials would experience different mass reference frames . with this approach
, it is shown that the relativistic mass of a falling body can be calculated directly from gr .
thus , the correspondence between general and special relativity is enhanced .
unlike the pseudotensor @xmath1 , this interpretation is independent of the choice of coordinates .
the use of mass reference frames also conserves mass and energy during a lift / fall cycle , unlike the concept of invariant mass .
inasmuch as there currently exists no widely - accepted treatment of localized gravitational energy in gr , it is hoped that the proposed interpretation will be considered worthy of further study and discussion .
much work remains to be done .
many attempts have been made to incorporate a scalar field into gravitational theory .
examples are the scalar - tensor theory of brans and dicke @xcite and the dilaton field @xcite .
what this paper suggests is that , in static fields , gr already includes an intrinsic scalar field @xmath0 that accounts for the energy of the gravitational field . unlike static fields ,
time - varying fields require a net exchange of energy and mass between gravitating bodies . with this interpretation , gravitational waves and/or gravitons would convey energy and mass . in newtonian gravity ,
potential energy allows total energy to be conserved when lifting a body .
the equivalence of a small amount of mass to a large amount of energy allows total energy and mass to be conserved without it , consistent with everyday experience . | an interpretation of general relativity is developed in which the energy used to lift a body in a static gravitational field increases its rest mass .
observers at different gravitational potentials would experience different mass reference frames .
it is shown that bodies falling in a schwarzschild field exhibit the relativistic mass / velocity relationship from special relativity .
this new result is independent of the choice of coordinates .
the proposed approach provides a physical explanation for gravitational energy , which is localized as a scalar function intrinsic to general relativity . applying this model to the robertson - walker
metric demonstrates that time - varying fields induce a net energy transfer between bodies that is not exhibited in static fields . |
speaking in sicily , home of many distinguished relics of the classical world , it would only be appropriate to begin this talk with a reference to the classical view of accretion disks around black holes and the outflows they drive .
in this view , created in three seminal papers published thirty - plus years ago , matter drifts slowly inward through accretion disks due to the action of a mysterious viscosity " while dissipating an amount of energy per unit mass that is a function solely of the black hole s spin .
more quantitatively , the
viscous " stress is assumed , on the basis of dimensional analysis , to be @xmath0 , where @xmath1 is some unknown number less than or comparable to unity , and @xmath2 is the internal pressure of the accretion flow @xcite .
the reason why the heat liberated is supposed to be a function of black hole spin alone is that all forces are imagined to end at the innermost stable circular orbit ( the isco ) , so that the radiative efficiency can be identified with the binding energy per unit mass of that orbit @xcite .
meanwhile , relativistic jets are driven outward along the rotation axes of spinning black holes , even in the absence of ongoing accretion , as frame - dragging creates an effective electric field parallel to the large - scale magnetic field threading the hole s event horizon @xcite .
the luminosity of these jets is controlled entirely by the strength of the field on the horizon and the spin of the black hole . in recent years , this classical view has been substantially revised .
the stress that transports angular momentum outward so that matter can move inward is now thought to be the result of fluctuating , but correlated magnetic fields : orbital shear forces @xmath3 whenever the orbital frequency gradient @xmath4 . because this stress has nothing to do with the sort of kinetic processes that the name viscosity " connotes , unlike a conventional fluid viscosity , it does not necessarily dissipate any energy where the stress takes place . that the field is strong enough to bring matter inward at a reasonable rate is due to continual stirring by a robust underlying mhd instability , the magneto - rotational instability @xcite .
because this instability has a very rapid linear growth rate ( comparable to the orbital frequency ) and its conditions for growth are very modest ( good enough electrical conductivity to justify the mhd approximation , a seed magnetic field with energy density no more than the total pressure ) , it should be present throughout any disk surrounding a black hole , as well as in the great majority of disks around other , less extreme , gravitating objects .
thus , electromagnetic stress is the primary vehicle for angular momentum transport in accretion disks .
a corollary of attributing disk stresses to magnetic fields is that the classical boundary condition at the isco is likely inappropriate .
its heuristic basis rested on purely hydrodynamic reasoning : as matter accelerates inward through the isco , passing from a domain of stable orbits to one where they are unstable , its density falls and its pressure drops .
it would be hard to envision circumstances in which the small amount of matter resident inside the isco could exert much force upon the far heavier disk outside .
however , as first recognized long ago @xcite and recently recalled to mind @xcite , magnetic fields severely undercut this argument their effectiveness at exerting stresses has nothing to do with the inertia or pressure of the matter .
consequently , there is no particular reason for their stresses to cease at the isco .
in fact , if one estimates the magnetic stress at the isco by supposing that the field is strong enough to drive accretion at the requisite rate in the main disk body and is then advected inward by flux - freezing , the magnetic stress there is likely to be quite sizable @xcite .
a second corollary of this picture is that magnetic fields arrive at the event horizon because they are brought there by the accretion flow . the field strength at the horizon , the basic parameter of the blandford - znajek @xcite mechanism ,
should then be determined by the history and current state of accretion .
analytic tools for studying turbulence are very limited ; large - scale numerical simulation is the only technique we have for obtaining quantitative results . in the past few years both
the capabilities of the hardware and the quality of the computer codes available for this effort have improved dramatically , leading to rapid improvements in accretion disk studies .
relatively recently , it was possible to study disks only in the shearing - box approximation , in which an annular segment is modelled by straightening it out into a rectangular geometry , while retaining a plane - parallel shear , a tidal potential and a coriolis force to approximate the dynamics of orbital motion . given the nature of this approximation , it made sense to perform these simulations in the newtonian limit .
today s codes , on the other hand , can treat global disks , initially with newtonian gravity and now in full general relativity @xcite . under the assumption of ideal mhd
( modulo the inevitable dissipation that occurs on the gridscale ) , they can follow fully three - dimensional dynamics for very long timescales : durations @xmath5 ( for black hole mass @xmath6 , setting @xmath7 ) .
these codes do an excellent job of conserving angular momentum , whether associated with the matter or electromagnetic fields , but their fidelity to nature is much more limited in regard to thermodynamics and radiation .
the de villiers - hawley @xcite and anninos et al .
@xcite codes evolve an internal energy equation and treat the gas equation of state as adiabatic with @xmath8 , adding entropy only in association with shocks through the use of an artificial bulk viscosity .
they therefore capture neither the bulk of the heat production that is a concomitant of accretion nor the photon radiation that cools the gas and produces the light we see . in a sense ,
the first omission compensates for the second .
the gammie - mckinney - tth @xcite and mizuno et al .
@xcite codes evolve a total energy equation . by construction ,
losses in kinetic and magnetic energy are transferred into heat , thereby modeling the turbulent dissipation process , but like the others , these codes also wholly ignore radiation .
most of the global disk simulations done to date by various groups assume either an initial large scale vertical net field threading the disk ( most of these have been newtonian simulations ) , or an initial disk that has no net flux , i.e. , all field lines closing within the disk . simulations with strong net flux generally have little difficulty producing jets and outflows , but we have no way of knowing whether such a strong field passing through the central region of the accretion flow is realistic .
the zero net flux initial condition is the simplest in terms of guessed parameters , as it means the field is simply zero all around the outer boundary .
this is the approach we have adopted for our simulations . to avoid the numerical difficulties arising from subsonic flow across the boundary
, all our simulations place their entire mass supply within the computational volume .
this fact has the consequence that angular momentum removed from the inner accretion flow is given to the mass in the outer disk , which must then move outward .
in other words , only the part of the flow inside a certain radius ( variously 12@xmath9 , depending on the simulation ) actually participates in a genuine accretion flow .
the outer boundary is generally located much farther out ( @xmath10 ) .
a number of results of qualitative importance have been derived from these studies .
most notably , the order - of - magnitude estimates predicting that magnetic stresses at the isco should be important have been strongly vindicated @xcite . as figure [ fig : stress ] demonstrates vividly ,
whereas the classical novikov - thorne picture predicts that the stress ends abruptly at the isco , the actual mhd stresses continue quite smoothly inward toward the black hole . in the case of the non - rotating black hole , there can be no outward stress at the event horizon because there is no source of angular momentum to supply the outward flow of angular momentum that constitutes the stress .
when the black hole rotates , however , the black hole itself is an ample reservoir of angular momentum , and there is no reason why the stress must end somewhere outside the event horizon .
although this process might appear to violate causality , there is no such violation in reality : this process can equally well be thought of as an inward flow of negative angular momentum facilitated by the rotating black hole s ergosphere , where both matter and electromagnetic fields can have negative energy - at - infinity .
[ fig : stress ] .
solid curves are the electromagnetic stress as found in the simulations at a particular time ; dotted curves are the prediction of the novikov - thorne model for an accretion rate matching the time - average of that simulation.,title="fig : " ] .
solid curves are the electromagnetic stress as found in the simulations at a particular time ; dotted curves are the prediction of the novikov - thorne model for an accretion rate matching the time - average of that simulation.,title="fig : " ] regarding the shakura - sunyaev @xmath1 parameter as merely a measure of the stress in pressure units , we find that it rises rapidly with diminishing radius , beginning well outside the isco and continuing through the plunging region ( fig .
[ fig : alpha ] ) . in the disk body , @xmath110.1 .
as one moves closer to the isco , the pressure diminishes ( in relative terms ) even while the magnetic stress increases .
the result is that the nominal @xmath1 is 510 times larger near the isco and another factor of 5 or more larger still deep inside the plunging region .
[ fig : alpha ] . note that the isco for @xmath12 is at @xmath13 , while for @xmath14 , it is at @xmath15.,title="fig : " ] .
note that the isco for @xmath12 is at @xmath13 , while for @xmath14 , it is at @xmath15.,title="fig : " ] throughout the disk , wherever there are stresses transporting angular momentum radially , there is also work done by one ring on its neighbor .
the energy flow into a ring due to the imbalance between the work it does and the work done on it is not necessarily cancelled by the net change in its orbital ( and electromagnetic energy ) content due to accretion .
when novikov and thorne first calculated the radial profile of dissipation on the basis of energy and angular momentum conservation alone , it was exactly this mismatch that lay at the basis of their work .
stress at the isco therefore alters the total dissipation possible in the disk outside that radius @xcite .
recently , we have attempted to estimate the total radiative efficiency , including heat liberated inside the isco ( and therefore beyond the reach of the agol - krolik @xcite formalism ) and also allowing for losses due to photons following orbits that bring them to the black hole rather than infinity @xcite .
unfortunately , as mentioned earlier , we do not actually compute the dissipation rate directly .
instead , we must seek some way of estimating the dissipation rate based on local physical quantities .
although we can not prove that a diagnostic based on guesswork is correct , we can eliminate those that must be wrong : in the disk body , where the inflow time is long compared to the thermal time , energy conservation in the context of a time - steady disk demands a definite dissipation profile . with no stress at the isco , this is the profile predicted by novikov and thorne ; with stress , this form is generalized , as worked out by agol and krolik .
any candidate prescription for estimating the dissipation must therefore match the profile required by energy conservation in the time - steady portion of the disk .
after trying a number of possibilities , all heuristically linked to energy production one way or another , we found that the best agreement with the requirements of energy conservation in a time - steady disk was achieved with the assumption that there is a uniform resistivity @xmath16 , so that the rate of heat generation is @xmath17 , where @xmath18 is the four - current density . to compute the luminosity observed at infinity , we then assumed that photons are emitted isotropically in the fluid frame at a rate proportional to that dissipation function and traced their trajectories until they reach infinity or are swallowed by the hole .
as it turns out , most of the dissipation predicted by this rule takes place outside the isco .
in addition , many of the photons radiated from within the isco are on orbits that ultimately lead to capture by the black hole . consequently , the bulk of the additional radiative efficiency seen by distant observers has its origin near and just outside the isco .
table [ tab : diskeff ] shows the numbers derived from a single snapshot from each of four simulations .
the entries in the second column are the efficiencies predicted by the novikov - thorne model , but after allowance for photons captured by the black hole .
they are therefore smaller than the standard quoted values , by a very small amount for @xmath12 , where the isco is relatively far from the horizon , by a more substantial amount ( about a @xmath19 fractional reduction ) for the case of @xmath20
. ranges rather than single values are shown because some photon orbits intersect the equatorial plane before heading off to infinity .
when they do , they may interact with the matter in the accretion flow .
scattering or absorption followed by reradiation puts the photons onto a new orbit that may or may not go to infinity . which end - result obtains depends very much on the details .
to allow for this very model - dependent uncertainty , we quote ranges in which the lower number is the result of assuming that none of the energy from photons on plane - crossing orbits reaches infinity while the upper number is the result of the opposite assumption .
.estimated radiative efficiency from accretion , in rest - mass units .
[ cols= " < , > , > " , ] not all magnetic geometries are this efficient , however .
the numbers in table [ tab : jeteff ] were computed on the basis of an initial magnetic field shape that is dipolar " in the sense that the field lines wrap concentrically around the pressure maximum of the original matter distribution . when the initial field is , instead ,
quadrupolar " in the sense that the field is divided into two sectors , with one set of field loops centered on a ring offset above the equatorial plane , the other positioned symmetrically below , reconnection in the inner accretion flow is so efficient that the poynting flux in the jet is greatly reduced .
with the progress of the past fifteen years , there have been substantial advances in our understanding of the physics of accretion disks .
the origin of the stress that plays such a fundamental role in driving accretion is no longer unknown it is mhd turbulence , driven by a powerful and robust instability .
although the equations governing such a system are strongly nonlinear , they are based on very well - understood physics : nothing more than classical electromagnetism and relativistic gravity . because we can build upon such a solid foundation
, it makes sense to undertake the arduous task of constructing large - scale simulations .
these are now so feasible that there are several competing groups performing them . * contrary to expectations based on hydrodynamic intuition , mhd stresses continue to be strong all the way through the isco and ( if the black hole rotates ) in to the event horizon .
precise estimates of the additional efficiency due to these stresses are not yet available , but we can begin to make estimates . * when the magnetic field has the right geometry , relativistic electromagnetically - dominated jets can form along the black hole rotation axis . driven by the rotation of the black hole , these jets can carry poynting flux to infinity at a rate that may increase the power output per unit rest mass accreted by factors of order unity relative to that radiated directly from the accretion disk proper .
we are grateful to all our collaborators on this subject : jean - pierre de villiers , shigenobu hirose , kris beckwith , and jeremy schnittman .
this work was partially supported by us national science foundation grant ast-0507455 . | the inward flow of matter through accretion disks is driven by mhd turbulence .
global general relativistic mhd simulations shed quantitative light on this process , revealing a number of aspects of accretion previously unrecognized . among them
are strong stresses in the marginally stable and plunging regions near the black hole and electromagnetically - dominated conical relativistic jets that can form spontaneously from the accretion flow .
the energy release associated with both of these effects can significantly augment the classical energy - release estimates based on purely hydrodynamic models .
address = department of physics and astronomy , johns hopkins university , baltimore md 21218 usa address = department of astronomy , university of virginia , charlottesville va 22904 usa |
a comprehension of the nature of the scalar mesons is very important for some major topics in particle physics such as the qcd vacuum .
however , both parameters of the scalar mesons , obtained from experimental data in various analyses , and even the status of some of them , are still quite ambiguous @xcite . as for the meson parameters ,
let us mention the widely discussed @xmath8 meson ( formerly @xmath9 ) , @xmath10 , and @xmath11 .
a doubtful meson existence can be demonstrated on the case of the @xmath12 state which is apparently required by a bulk of data @xcite , but in some analyses of only the @xmath4 scattering no evidence for its existence was found @xcite .
we have shown that the existence of the @xmath12 does not contradict the data on @xmath13 @xcite . in the hidden gauge unitary approach ,
the @xmath12 appears dynamically generated as a @xmath14 state @xcite and the @xmath15 as generated from the @xmath16 interaction .
note also a situation with scalar states in the 1500-mev region . in our previous model - independent analyses of @xmath17
, we observed a wide state @xmath11 whereas in some other analyses , which included mainly meson production and decay processes and which are cited in the pdg tables @xcite , a rather narrow @xmath11 was obtained .
we have suggested that the wide @xmath11 , observed in the multichannel @xmath4 scattering , is effectively a superposition of two states , the wide and narrow state .
the latter is observed only in decays and productions of mesons .
this suggestion was verified in the model - independent two - channel analysis of data on @xmath18 @xcite . in the presented article
we confirm this assumption in the three - channel analysis of data on @xmath1 and decays @xmath2 from the dm2 , mark iii , and bes ii collaborations @xcite .
this is necessary , especially , as the wide states provoke many questions which should be answered . in view of this situation related to parameters and the status of the scalar mesons ,
there are still many unsolved problems as to determining a qcd nature of the mesons and their assignment to the quark - model configurations in spite of a big amount of work devoted to these problems ( see , e.g. , @xcite and references therein ) . in this article
, we describe the multichannel @xmath4 scattering ( @xmath19 ) using the method based only on the first principles , analyticity and unitarity @xcite , which allows us to avoid any theoretical prejudice in extracting the resonance parameters .
this we call `` the model - independence '' @xcite . the method is applied to the analysis of experimental data on the multichannel @xmath4 scattering and decays @xmath2 .
the @xmath6 decays are described using a formalism from refs .
@xcite , where certain reasonable assumptions about the final - state interactions are made . considering the obtained arrangement of resonance poles on the riemann - surface sheets , the constants of resonance couplings with the channels , and the resonance masses we can draw definite conclusions about the nature of the investigated states .
the article is organized as follows . a basic formalism for the three - channel model - independent method
was already given in our previous paper @xcite ; therefore , in sec .
ii we give only formulas introducing parameters determined in the analysis .
results of the combined coupled - channel analysis of data on isoscalar @xmath0-wave processes @xmath1 and on decays @xmath2 are presented in sec .
discussion of the results and conclusions are given in sec .
the multichannel @xmath0-matrix can be described in our model - independent method , which essentially utilizes a uniformizing variable , without any approximations only in the two - channel case . in the three - channel case , a four - sheeted model of the eight - sheeted riemann surface
has to be constructed to obtain a simple symmetric ( easily interpreted ) picture of the resonance poles and zeros of the @xmath0 matrix on the uniformization plane .
the matrix elements @xmath20 , where @xmath21 denote the channel numbers , have the right - hand cuts along the real axis of the @xmath22 complex plane ( @xmath22 is the invariant total energy squared ) starting with the channel thresholds @xmath23 and the left - hand cuts related to crossed channels .
an influence of the lowest branch point @xmath24 ( @xmath4 ) is neglected but unitarity on the @xmath4 cut is kept .
sheets of the riemann surface are numbered according to the signs of analytic continuations of the square roots @xmath25 @xmath26 , @xmath27 correspond to sheets i , ii,@xmath28 , viii , respectively .
resonances are described on the riemann surface using the formulas for analytic continuations of the @xmath0-matrix elements to all sheets .
the formulas allow us to express the matrix elements on the unphysical sheets by means of the matrix elements on the physical sheet that have only the resonance zeros ( aside the real axis ) , at least , around the physical region @xcite . assuming the resonance zeros on sheet i
, we can obtain an arrangement of poles and zeros of the resonance on the whole riemann surface which we denote as a resonance cluster .
in the three - channel case , we obtain _ seven types _ of the resonance clusters corresponding to possible situations when there are resonance zeros on sheet i only in @xmath29 ( * a * ) ; @xmath30 ( * b * ) ; @xmath31 ( * c * ) ; @xmath29 and @xmath30 ( * d * ) ; @xmath30 and @xmath31 ( * e * ) ; @xmath29 and @xmath31 ( * f * ) ; @xmath29 , @xmath30 , and @xmath31 ( * g * ) .
a three - channel resonance has to be described by one of the seven types of the resonance clusters which is the necessary and sufficient condition for its existence .
the resonances of types ( * a * ) , ( * b * ) and ( * c * ) can be related to the resonances represented by breit - wigner forms but the types ( * d * ) , ( * e * ) , ( * f * ) and ( * g * ) do not have their equivalents in the breit - wigner description .
the cluster type is related to the nature of state .
considering the @xmath4 , @xmath32 and @xmath33 channels , e.g. , a resonance coupled relatively more strongly to the @xmath4 channel than to the @xmath32 and @xmath33 channels is described by the cluster of type ( * a * ) but in the opposite case , the resonance is represented by the cluster of type ( * e * ) ( e.g. , the state with the dominant @xmath34 component ) .
the glueball must be represented by the cluster of type ( * g * ) as a necessary condition for the ideal case .
it is also possible to distinguish , in a model - independent way @xcite , a bound state of colourless particles ( _ e.g. _ , @xmath32 molecule ) from a @xmath35 bound state .
alike in the one - channel case , the existence of the particle bound state means presence of a pole on the real axis below the threshold on the physical sheet . in the two - channel case , therefore , the existence of the bound state in channel 2 ( e.g. , @xmath32 molecule ) that can decay into channel 1 ( @xmath4 decay ) implies the presence of the pair of complex conjugate poles on sheet ii below the second - channel threshold without the corresponding shifted pair of poles on sheet iii . in the three - channel case ,
the bound state in channel 3 ( @xmath33 ) that can decay into the channels 1 ( @xmath4 decay ) and 2 ( @xmath32 decay ) is represented by the pair of complex conjugate poles on sheet ii and by the pair of shifted poles on sheet iii below the @xmath33 threshold without the corresponding poles on sheets vi and vii .
the formulas of the analytic continuations @xcite prescribe that the resonance parameters ( mass , total width , and coupling constants with the channels ) must be calculated using the pole positions on sheets ii , iv , and viii because only on these sheets do the analytic continuations have the forms : @xmath36 , @xmath37 and @xmath38 , respectively , i.e. , the positions of poles on these sheets are at the same points of the complex - energy plane as the resonance zeros on the physical sheet .
the other pole positions are shifted due to the coupling of channels .
the @xmath0-matrix elements of all coupled processes are expressed in terms of the jost matrix determinant @xmath39 using the le couteur - newton relations @xcite .
the jost determinant is a real analytic function with the only square - root branch points at @xmath40 .
the important branch points , which correspond to the thresholds of the coupled and crossed channels , are taken into account in the uniformizing variable . in the uniformizing variable
used here we neglect the lowest @xmath4-threshold branch point but take into account the threshold branch points related to the two remaining channels , and the left - hand branch point at @xmath41 @xcite @xmath42 here @xmath43 , @xmath44 .
this variable maps our model of the eight - sheeted riemann surface onto the uniformization @xmath45 plane divided into two parts by a unit circle centered at the origin .
the semisheets i ( iii ) , ii ( iv ) , v ( vii ) and vi ( viii ) are mapped onto the exterior ( interior ) of the unit disk in the 1st , 2nd , 3rd and 4th quadrants , respectively .
the physical region extends from the @xmath4 threshold @xmath46 on the imaginary axis along this axis down to the point _
i _ on the unit circle ( @xmath32 threshold ) .
then it goes along the unit circle clockwise in the 1st quadrant to point 1 on the real axis ( @xmath33 threshold ) and then along the real axis to the point @xmath47 which is an image of @xmath48 .
the intervals @xmath49 $ ] , @xmath50 , @xmath51 $ ] , @xmath52 are the images of the corresponding edges of the left - hand cut of the @xmath4-scattering amplitude .
each resonance is represented in @xmath29 by the poles and zeros that are symmetric to each other with respect to the imaginary axis .
the representations of all possible types of resonances in @xmath29 on the @xmath45 plane can be found in ref .
@xcite .
the main model - independent effect of multichannel resonances is given by the pole clusters .
assuming that possible small remaining ( model - dependent ) contributions of resonances can be included via the background , the @xmath0-matrix elements are taken as the products @xmath53 [ d_res ] where @xmath54 describes the background and @xmath55 the resonance contributions . on the @xmath45 plane ,
the le couteur - newton relations are somewhat modified taking account of the used model of the initial eight - sheeted riemann surface @xmath56 where the subscripts in the matrix elements @xmath20 denote the channels : @xmath571@xmath4 , 2@xmath32 , 3@xmath33 .
the @xmath58 function for the resonance part in these relations is @xmath59 with @xmath60 a number of resonance zeros . for the background part @xmath54
, the @xmath61 function has the form @xmath62 \label{dbackground}\ ] ] where @xmath63 with @xmath64 the @xmath65 threshold and @xmath66 the effective threshold due to the opening of many channels in the energy region around 1.5 gev ( e.g. , @xmath67 ) .
these thresholds are determined in the analysis .
the expressions ( [ lecouteur - newton ] ) and ( [ dw_res ] ) provide the simplest possible parametrization of the resonance part of the @xmath0 matrix for a given number and type of resonances on the uniformization @xmath45 plane keeping unitarity and analyticity .
the free parameters ( the zeros @xmath68 ) of the @xmath55 with a particular number and type of resonances , are fixed by fitting to the experimental data . the scenario with the smallest @xmath69
is chosen as the most probable hypothesis on the condition of the given data set .
note that this is an opposite approach to that , e.g. , in refs .
@xcite where the @xmath0 matrix is constructed in the physical region and then it is analytically continued to the riemann surface to find poles , for example the @xmath70-meson pole . in our approach
, positions of the poles are obtained directly from the fitting .
an optimal number of poles is the minimal number which guarantees a satisfactory description of the data and which contains only the poles significantly improving the fit .
the poles are introduced according to the formulas for analytic continuations of the @xmath0-matrix elements to all sheets @xcite .
the background part @xmath71 is constructed in the physical region to mimic an influence of the other singularities not included explicitly in the resonance part @xmath55 . the simple form in eq.([dbackground ] )
includes a response to the opening of the channels whose threshold branch points are not taken into account explicitly in the uniformizing variable .
values of the fitted parameters in eq.([dbackground ] ) ( @xmath72 and @xmath73 ) indicate a relative importance of these branch points , e.g. a negative or large value of some background parameter could suggests that the corresponding branch point should be explicitely allowed for in the uniformizing variable .
therefore , in choosing the best variant we also require that the background contribution is negligible , i.e. , the background parameters are small . in our previous analysis of data on @xmath18 we took into account the left - hand branch point at @xmath41 in the uniformizing variable in addition to the @xmath4- and @xmath32-threshold branch points
@xcite . in the analysis of
@xmath1 we allowed rather for the @xmath33-threshold branch point @xcite .
in the presented more elaborate three - channel analysis , unlike in ref .
@xcite , we follow more consistently the spirit of the model - independent description obtaining practically zero background of the @xmath4 scattering .
in the combined analysis of data on the isoscalar @xmath0 waves of processes @xmath19 @xcite we added data on decays @xmath2 from the mark iii @xcite , dm2 @xcite and bes ii @xcite collaborations .
formalism for calculating the di - meson mass distributions of these decays can be found in refs .
@xcite . in this approach
the pairs of the pseudoscalar mesons in the final states are assumed to have @xmath74 and they undergo strong interactions whereas the @xmath75 meson behaves as a spectator .
the amplitudes for @xmath2 decays are related with the scattering amplitudes @xmath76 @xmath77 as follows : @xmath78,\ ] ] @xmath79,\ ] ] where @xmath80 and @xmath81 are functions of the couplings of the @xmath6 to channels 1 and 2 ; @xmath82 , @xmath83 , @xmath84 and @xmath85 are free parameters .
the pole term in @xmath86 approximates possible @xmath87 states which are not forbidden by the ozi rules considering quark diagrams of these processes .
obviously this pole should be situated on the real @xmath22-axis below the @xmath4 threshold .
this is an effective inclusion of the effect of so - called `` crossed channel final - state interactions '' in @xmath88 , which was studied largely , e.g. , in ref .
the di - meson mass distributions are given as @xmath89 where @xmath90 is a normalization constant to data of the experiments determined in the analysis : 0.7512 , 0.2783 , and 5.699 for the mark iii , dm2 , and bes ii data , respectively .
parameters of the @xmath91 functions , obtained in the analysis , are @xmath92 , @xmath93 , @xmath94 , @xmath95 , @xmath96 , and @xmath97 .
the scattering amplitudes @xmath76 are related to the @xmath0 matrix as @xmath98 where @xmath99 . in the analysis we supposed that in the 1500-mev region there are two resonances : the narrow @xmath11 and wide @xmath100 .
the @xmath101 state is described by the cluster of type ( * a * ) , @xmath11 by type ( * c * ) , and @xmath100 by type ( * g * ) .
the @xmath102 is represented only by the pole on sheet ii and shifted pole on sheet iii . however , the representation of the @xmath103 and @xmath15 states is not unique .
these states can be described by clusters either of type ( * b * ) or type ( * c * ) . analyzing only the processes @xmath104 , similarly as it was done in @xcite , it is impossible to prefer any of four indicated possibilities .
moreover , it was found that the data admit two sets of parameters of @xmath5 with a mass relatively near to the @xmath105-meson mass , and with the total widths either @xmath106 or @xmath107 mev , solutions a and b , respectively , like in ref . @xcite . in the extended combined analysis , adding the data on decays @xmath2 , one can prefer the scenarios when the @xmath12 is described by the cluster of type ( * b * ) and @xmath108 by the cluster of type either ( * b * ) or ( * c * ) . to be specific , in the following
we shall discuss the case when the @xmath108 is represented by the cluster of type ( * c * ) .
it is interesting that the di - pion mass distribution of the @xmath109 decay of the bes ii data from the threshold to @xmath110 gev clearly prefers the solution with the wider @xmath5 ( solution b ) .
a satisfactory description of all analyzed processes is obtained with the total @xmath111 where for the @xmath4 scattering , @xmath112 , for @xmath113 , @xmath114 , for @xmath115 , @xmath116 , and for decays @xmath117 , @xmath118 . the combined description ( @xmath69 ) of processes @xmath19 with adding the data on decays @xmath117 is practically the same as in ref .
@xcite performed without considering decays of the @xmath6 mesons .
a comparison of the description with the experimental data is shown in figs . [
fig : pipi]-[fig : j / psi_bes ] .
-scattering @xmath0-wave matrix element .
the data are from ref .
@xcite ( hya73 ) , @xcite ( pro73 ) , @xcite ( other ) , and @xcite ( na48 ) .
[ fig : pipi],title="fig:",scaledwidth=35.0% ] -scattering @xmath0-wave matrix element .
the data are from ref .
@xcite ( hya73 ) , @xcite ( pro73 ) , @xcite ( other ) , and @xcite ( na48 ) .
[ fig : pipi],title="fig:",scaledwidth=35.0% ] @xmath0-wave matrix element .
the data are from ref .
@xcite ( coh80 , cos80 , pol79 , wet76 , etk82 , and mar79 ) and @xcite ( est79).,title="fig:",scaledwidth=35.0% ] @xmath0-wave matrix element .
the data are from ref .
@xcite ( coh80 , cos80 , pol79 , wet76 , etk82 , and mar79 ) and @xcite ( est79).,title="fig:",scaledwidth=35.0% ] @xmath0-wave matrix element .
the data are from @xcite.,scaledwidth=37.0% ] invariant mass distributions in the @xmath109 decay .
panel _ ( a ) _ shows the fit to the data of mark iii and _ ( b ) _ to dm2 .
[ fig : j / psi],title="fig:",scaledwidth=35.0% ] invariant mass distributions in the @xmath109 decay . panel _ ( a ) _ shows the fit to the data of mark iii and _ ( b ) _ to dm2 .
[ fig : j / psi],title="fig:",scaledwidth=35.0% ] invariant mass distributions in the @xmath88 decay . panel _ ( a ) _ shows the fit to the data of mark iii and _ ( b ) _ to dm2 . [
fig : j / psikk],title="fig:",scaledwidth=35.0% ] invariant mass distributions in the @xmath88 decay . panel _ ( a ) _ shows the fit to the data of mark iii and _ ( b ) _ to dm2 .
[ fig : j / psikk],title="fig:",scaledwidth=35.0% ] invariant mass distribution in the @xmath109 decay in comparison with the data of the bes ii collaboration .
[ fig : j / psi_bes],scaledwidth=47.0% ] in table [ tab : clusters ] we show the obtained pole clusters for the resonances on the complex - energy plane @xmath119
. the poles on sheets iii , v , and vii and vi , corresponding to the @xmath100 , are of the second and third order , respectively ( this is an approximation ) .
.the pole clusters for resonances on the @xmath119 plane .
@xmath120 in mev . [ cols="^,^,^,^,^,^,^,^",options="header " , ]
in the combined model - independent analysis of data on @xmath1 in the @xmath121 channel and on @xmath2 from the mark iii , dm2 , and bes ii collaborations , an additional confirmation of the @xmath5 with the pole at @xmath122 mev on sheet ii is obtained , which can be related with the mass @xmath123 mev and width @xmath124 mev via eq .
( [ hmota ] ) .
the real part of the pole is in a good agreement with the results of other analyses cited in the pdg tables of 2012 : the pdg estimation for the @xmath5 pole is @xmath125 mev .
the obtained imaginary part is , however , larger than the pdg estimation . as large values of the imaginary part of the @xmath5 pole appear to be inherent in our method of analysis @xcite , the origin of this interesting result should be understood . in ref .
@xcite we showed that a relatively narrow @xmath70 meson consistent with pdg can be obtained in the analysis of one - channel @xmath4 scattering data but the inclusion of the @xmath126 data into the analysis makes the width significantly larger .
therefore it seems that the large width is tightly connected with the multichannel analysis of data . the value of the mass , which gets a significant contribution from the large width , eq .
( [ hmota ] ) , agrees well with the prediction by weinberg in ref .
@xcite . in this work
it was shown that even where the chiral symmetry is spontaneously broken it can still be used to classify hadron states .
such mended symmetry leads to a quartet of particles with definite mass relations and c parity , giving the prediction @xmath127 .
this prediction is also in agreement with a refined analysis using the large-@xmath128 consistency conditions between the unitarization and resonance saturation which suggests @xmath129 @xcite . in addition , in the soft - wall anti - de sitter / qcd approach @xcite the approach based on gauge / gravity duality the predicted mass of the lowest @xmath3 meson , 721 mev , practically coincides with the value obtained in our work . the above discussion concerns solution b , which is prefered by the analysis presented in this paper .
the imaginary part of the @xmath5 pole in solution a ( 343 mev ) @xcite is still in agreement with the pdg estimation .
however , solution a is inconsistent with data on the @xmath109 decay from the bes ii collaboration : the corresponding curve in fig .
[ fig : j / psi_bes ] lies considerably below the data from the threshold to about 850 mev .
therefore , solution a is not considered in this paper . anyway
the question of too large width of the @xmath5 desires a further investigation , estimating the theoretical uncertainties of our approach .
the obtained results for @xmath10 , @xmath130 mev and @xmath131 mev , indicate that the @xmath10 is a non-@xmath132 state , e.g. , the @xmath33 bound state because it lies slightly above the @xmath32 threshold and is described by the pole on sheet ii and by the shifted pole on sheet iii without the corresponding ( for standard clusters ) poles on sheets vi and vii . in the pdg tables of 2010
its mass is 980@xmath13310 mev .
we found in all combined analyses of the multichannel @xmath4 scattering the @xmath10 slightly above 1 gev , as in the dispersion - relations analysis only of the @xmath4 scattering @xcite . in the pdg tables of 2012 ,
for the mass of @xmath10 an important alteration appeared : now there is given the estimation 990@xmath13320 mev .
we conclude that the @xmath134 and @xmath108 states are dominated by the @xmath34 component in the wave function .
the conclusion about the @xmath134 agrees with results of the work of the crystal barrel collaboration @xcite where the @xmath134 is identified as the @xmath33 resonance in the @xmath135 final state of the @xmath136 annihilation at rest .
this also explains well why one did not find this state considering only the @xmath4 scattering process @xcite .
the conclusion about the @xmath108 is consistent with the experimental fact that this state is observed in @xmath137 @xcite but it is not observed in @xmath138 @xcite . in the 1500-mev region , indeed , there are two states : the @xmath11 ( @xmath139 mev , @xmath140 mev ) and the @xmath100 ( @xmath141 mev , @xmath142 mev ) .
the @xmath100 is interpreted as a glueball taking into account its biggest width among the enclosing states @xcite . as to the large width of the glueball
, it is worth to indicating ref .
there an effective qcd lagrangian with the broken scale and chiral symmetry is used , where a glueball is introduced to theory as a dilaton and its existence is related to the breaking of scale symmetry in qcd .
the @xmath4 decay width of the glueball , estimated using low - energy theorems , is @xmath143 where @xmath144 is the glueball mass .
that is , if the glueball with the mass of about 1 gev exists , then its width would be near 600 mev .
of course , the use of the above formula is doubtful above 1 gev ; however , a trend for the glueball to be wide is apparently seen . on the other hand , in a two - flavour linear sigma model with global chiral symmetry and ( axial-)vector mesons as well as an additional glueball degree of freedom where the glueball is also introduced as a dilaton @xcite
, there arises the rather narrow resonance in the 1500-mev region as predominantly a glueball with a subdominant @xmath145 component . on second thoughts ,
this result can be considered as preliminary due to using a quite rough flavor - symmetry su(@xmath146 ) in the calculations or , e.g. , evaluating the @xmath147 decay , the intermediate state consisting of two @xmath5 mesons is not included . in ref .
@xcite , where the two - pseudoscalar and two - photon decays of the scalars between 1 - 2 gev were analyzed in the framework of a chiral lagrangian and the glueball was included as a flavor - blind composite mesonic field , the glueball was found to be rather narrow .
taking into account the discovery of isodoublet @xmath148 @xcite ( see also @xcite ) , two lower nonets should correspond to two existing isodoublets @xmath149 .
we propose the following sets of the su(3 ) partners for these states excluding the @xmath10 as the non-@xmath35 state @xcite : the lowest nonet consists of the isovector @xmath150 , the isodoublet @xmath148 , and @xmath5 and @xmath12 as mixtures of the eighth component of the octet and the su(3 ) singlet .
the next nonet could consist of the isovector @xmath151 , the isodoublet @xmath152 , and two isoscalars @xmath11 and @xmath15 . since this assignment removes a number of questions that stood earlier when placing the scalar mesons to nonets and
does not put forth any new ones , we think this is the right direction .
an adequate mixing scheme is needed , the search for which is complicated by the fact that , in this case , there is also a remainder of chiral symmetry which , however , makes it possible to predict correctly , _
e.g. _ , the @xmath70-meson mass @xcite .
the authors thank thomas gutsche and mikhail ivanov for useful discussions and interest in this work . this work was supported in part by the grant program of plenipotentiary of the slovak republic at jinr , the heisenberg - landau program , the votruba - blokhintsev program for cooperation of the czech republic with jinr , the grant agency of the czech republic ( grant no .
p203/12/2126 ) , the bogoliubov - infeld program for cooperation of poland with jinr , the dfg under contract no .
ly 114/2 - 1 .
the work was also partially supported under the project 2.3684.2011 of tomsk state university .
this work has been partly supported by the polish ncn grant no 2013/09/b / st2/04382 .
a. zylbersztejn _
et al . _ ,
* 38b * , 457 ( 1972 ) ; p. sonderegger and p. bonamy , in proc .
conference on elementary particles , lund , 1969 , 372 ; j.r .
et al . _ ,
36b * , 134 ( 1971 ) ; j.p .
_ et al . _ ,
33b * , 525 ( 1970 ) ; * 33 * , 528 ( 1970 ) ; p. baillon _
et al . _ ,
38b * , 555 ( 1972 ) ; l. rosselet _
et al . _ ,
d * 15 * , 574 ( 1977 ) ; a.a .
et al . _ , pisma zh .
. fiz . * 25 * , 68 ( 1977 ) ; a.a .
belkov _ et al .
_ , pisma zh .
* 29 * , 652 ( 1979 ) .
w. wetzel _
et al . _ ,
nucl . phys . *
b115 * , 208 ( 1976 ) ; v.a .
et al . _ , phys .
d * 19 * , 1317 ( 1979 ) ; p. estabrooks , phys .
d * 19 * , 2678 ( 1979 ) ; d. cohen _ et al .
d * 22 * , 2595 ( 1980 ) ; g. costa _
et al . _ ,
* b175 * , 402 ( 1980 ) ; a. etkin _
et al . _ ,
phys . rev .
d * 25 * , 1786 ( 1982 ) ; a.d .
martin and e.n .
ozmutlu , nucl . phys . *
b158 * , 520 ( 1979 ) . | a combined analysis of data on the isoscalar @xmath0-wave processes @xmath1 and on decays @xmath2 from the dm2 , mark iii and bes ii collaborations is performed to study @xmath3 mesons .
the method of analysis is based on analyticity and unitarity and uses an uniformization procedure . in the analysis limited only to the multichannel @xmath4-scattering data , two possible sets of parameters of the @xmath5
were found : in both cases the mass was about 700 mev but the total width was either about 680 or 1040 mev .
the extension of the analysis using only the dm2 and mark iii data on the @xmath6 decays does not allow us to choose between these sets . however , the data from bes ii on the di - pion mass distribution in the decay @xmath7 clearly prefer the wider @xmath5 state .
spectroscopic implications from the results of the analysis are also discussed . |
it is an important and controversial issue what causes the low - frequency cutoff in the radio spectrum of gps sources ; synchrotron self - absorption ( ssa ) or free free absorption ( ffa ) .
the discovery of ffa towards the gps galaxy oq 208 ( kameno et al . 2000 ) propounded a question how general is ffa towards gps sources .
the cold dense ffa plasma around the lobes of gps sources could be a cocoon which smothers expansion of jets and lobes ( bicknell et al .
1997 ) .
such an external absorber can be a probe of the viewing angle of jets , so that a test for the unified scheme between quasars and radio galaxies ( barthel 1989 ) , or the unified model between seyfert 1 and 2 galaxies ( antonucci and miller 1985 ) can be done .
both models presume that these classes are intrinsically identical , and suggest that the apparent differences are due to the viewing angle . with respect to barthel s unified scheme
, he showed that the projected distances of double lobes in quasars are significantly smaller than those of radio galaxies , and proposed that a smaller viewing angle causes higher luminosities and apparent presence of the broad line components . on the seyfert unification model , optical polarimetric observations for the seyfert 2 galaxy ngc 1068 detected highly polarized continuum and broad balmer lines ( miller and antonucci 1983 ) .
antonucci and miller ( 1985 ) suggested that hidden broad line components arose via scattering .
presence of the broad line component , which is directly seen in seyfert 1 but unseen in seyfert 2 galaxies , was understood as intrinsic identity of the seyfert 2 with seyfert 1 galaxies . like the unification between quasars and radio galaxies , the viewing angle was thought to play a role in different appearance . when the jet axis is close to the line of sight
, the path length through the external absorber will be longer towards the receding jet than towards the approaching jet .
thus , the ffa opacity towards a double lobe is expected to be rather asymmetric ; ffa opacity should be deeper towards the receding jet .
when the jet axis is nearly perpendicular to the line of sight , on the contrary , we expect rather symmetric ffa opacity towards a double lobe .
based on this idea , we conducted a trichromatic gps survey at 2.3 , 8.4 , and 15.4 ghz using the vlba for 9 objects .
we selected 9 sample objects , based on the gps catalog by de vries et al .
( 1997 ) , under the criteria : 1 .
the peak frequency @xmath2 should stand within our observing range , i.e. , @xmath3 ghz @xmath4 ghz .
this condition is necessary to discriminate between ssa and ffa by spectral fitting , and then to obtain the distribution of opacities .
2 . since all sources must be bright enough to be detected with the vsop and the vlba , we put the criteria @xmath5 jy , @xmath6 jy , and @xmath7 jy . the selected gps sources are listed in table [ tab : imageperform ] .
the optical identifications are based on the nasa / ipac extragalactic database ( ned ) .
hereafter , quasars and seyfert 1 galaxies are categorized as type-1 sources , while radio galaxies and seyfert 2 galaxies belong to type-2 sources .
+ @xmath8one scan corresponds to integration of 11 minutes .
[ tab : imageperform ] + @xmath9optical identifications are based on the nasa / ipac extragalactic database ( ned ) .
qso , rg , and sy stand for quasars , radio galaxies , and seyfert galaxies , respectively .
[ tab : compflux ]
the vlba observations have been carried out on december 15 , 1998 .
table [ tab : imageperform ] lists the sample objects of the vlba observations .
every object was observed at 3 frequencies with 2 to 11 scans , where each scan corresponds to integration of 11 minutes .
observations at three frequencies have been carried out almost simultaneously .
the subreflector switched between the dual - frequency 2.3/8.4 ghz reflector system and the 15.4 ghz feed horn within a 1-minute gap between the cycles .
we used 4 channels of 8-mhz bandwidth at 15.4 ghz , and allocated 2 channels at both 2.3 and 8.4 ghz .
the correlation process was accomplished by the vlba correlator .
we applied fringe - fitting , data flagging , and _ a priori _ amplitude calibration in the nrao aips .
imaging and self - calibration processes were carried out by difmap . synthesized
beam sizes and image qualities are listed in table [ tab : imageperform ] . while relative gain errors among the antennae are corrected though amplitude self - calibration processes
, further flux calibration is necessary to obtain certain spectra across the observing frequencies . for the purpose of absolute flux calibration
, we also imaged four calibrators ; the bl lacertae , da 193 , 3c 279 , and ot 081 .
these calibrators are so compact that the total cleaned flux densities should be the same with the total flux densities measured by a reliable single - dish or a short - baseline - interferometer . basing on the comparison between the total flux measurements by the university of michigan radio astronomy observatory ( umrao ) and the nrao green bank interferometer , and summation of cleaned flux densities , we applied flux scaling for the final results . at 15.4 ghz , for instance ,
total cleaned flux densities before absolute correction were 3.276 , 5.012 , 25.769 , and 4.172 jy for bl lac , da 193 , 3c 279 , and ot 081 , respectively .
the umrao database provides the total flux densities at 14.5 ghz , @xmath10 , @xmath11 , @xmath12 , and @xmath13 jy for each , as averaged over 2 months centered at our observation .
then we calculated a correction factor ( @xmath14 ) , where @xmath15 is the total cleaned flux density of the calibrator images , and @xmath16 is the mean flux densities of calibrators measured by the umrao .
we have @xmath17 at 15 ghz , so that we scaled our image by the factor and obtained the accuracy of the flux scale of 1.8 % .
the accuracy was derived from a standard deviation of the correction factors by the four calibrators . in the same manner
, we scaled images at 2.3 and 8.4 ghz by the correction factors of 0.84 and 0.79 , respectively .
the estimated amplitude accuracies are 4.0 , 7.2 and 1.8 % at 2.3 , 8.4 , and 15.4 ghz , respectively .
all images with uniform weighting are shown in figure [ fig : images ] .
based on the images , we estimated a flux density of each component using three different methods ; tasks ` imfit ' and ` jmfit ' to adopting elliptical gaussian distribution of the brightness , and ` tvstat ' to integrate brightness within a specified area . when the resolution is insufficient to isolate a double structure , in the cases of 2.3 ghz images of 0738 + 313 , 1333 + 459 , and 2149 + 056 , we applied double elliptical gaussian fits for ` imfit ' and jmfit. in these cased , we also tried gaussian model fits in visibilities using ` modelfit ' in difmap instead of ` tvstat ' , because these images are simple enough to be fitted in visibilities while it is difficult to set adequate integral area in these images for ` tvstat ' .
the flux densities of components are shown in table [ tab : compflux ] .
errors in the flux densities are estimated by the root sum squared of the amplitude calibration errors and the uncertainties in measurement of flux densities from the images .
the uncertainties in measurement are evaluated by the standard deviation between results of the three methods , i.e. ` imfit ' , ` jmfit ' , and ` tvstat ' ( or ` modelfit ' ) .
components which are considered to be lobes are labeled as a and b , in the sense that a component with a larger ffa opacity is labeled as a. the identification of lobes or a core is discussed in the next section . since we have observed at only three frequencies , it is impossible to discriminate between ssa and ffa by the spectral fit alone .
instead , we started from the simple ffa model assuming that ssa is negligible , and verify the assumed condition later .
it should be noticed that a solid model fit to discriminate between ssa and ffa requires at least five frequencies .
we are conducting vsop observations at 1.6 and 4.8 ghz for the supplement , which will be reported in continuation .
nevertheless , global properties among type-1 and -2 gps sources can be discussed as shown later .
anyway , we attempted the model : an optically thin synchrotron emission from the lobe is absorbed by external ffa plasma , @xmath18 here , @xmath19 is the observed flux density in jy , @xmath20 is the intrinsic flux density in jy at 1 ghz , @xmath21 is the frequency in ghz , @xmath22 is the intrinsic spectral index of the synchrotron emission , and @xmath23 is the ffa coefficient .
free parameters are @xmath20 and @xmath23 for each gaussian component , and common @xmath22 for all components . therefore , the number of free parameters @xmath24 , where @xmath25 is the number of gaussian components .
the number of data points is given by three frequencies times the number of gaussian components , hence @xmath26 .
consequently , the degree of freedom will be @xmath27 .
when we have two gaussian components , the degree of freedom is 1 .
again , the number of frequencies is too few to verify a statistical confidence for the spectral fit . despite this condition ,
the flux densities , the opacity coefficients , and the residuals of spectral fits are listed in table [ tab : ffafitparam ] .
note that the ffa coefficient @xmath23 directly corresponds to the spectral peak frequency @xmath2 because the spectral peak appears at the frequency where the optical depth approximates to unity .
thus , we have @xmath28 derived best - fit spectra are shown in figures [ fig : spectype1 ] and [ fig : spectype2 ] for type-1 and -2 sources , respectively .
figure [ fig : ffahistogram ] shows the histograms of ffa opacities towards each component .
[ tab : ffafitparam ]
a significant difference between type-1 and -2 gps sources can be seen in the ffa fitting .
spectral peak frequencies of double components in type-1 sources tend to differ significantly , while those in type-2 sources are relatively similar .
in other words , type-1 sources tend to show asymmetric ffa opacities towards double lobes , while type-2 sources have rather symmetric opacities , since the @xmath2 and @xmath23 are related by equation [ eqn : peakfreq ] .
to evaluate the asymmetry in the ffa opacities , we define the ffa opacity ratio @xmath29 ( @xmath30 ) .
@xmath31 is an index of asymmetry in the peak frequencies @xmath2 , too , as @xmath32 even if the spectral peak is caused by ssa , beyond our assumption , @xmath31 represents asymmetry in terms of ssa .
the spectrum of power - law synchrotron radiation with ssa is expressed by @xmath33 , \label{eqn : ssaspectrum}\end{aligned}\ ] ] where @xmath34 is the ssa opacity coefficient .
the spectral peak @xmath2 approximates to @xmath35 .
then the relation between @xmath31 and @xmath36 will be @xmath37 though @xmath31 is derived from the spectral fits using the ffa model .
the histogram of @xmath31 clearly exhibits the difference between type-1 and -2 sources ( see figure [ fig : ffahistogram ] ) .
the weighted means of @xmath31 for type-1 and -2 are @xmath38 and @xmath39 , respectively . here
, we perform a statistical test , whether any significant difference arises between the ffa opacity ratios of type-1 and -2 groups .
let us put a testing hypothesis , which assumes @xmath31 is identical for these two subsets .
then , the @xmath40 value defined as @xmath41 must follow the @xmath42 distribution with the degrees of freedom of 7 . here , @xmath43 and @xmath44 are mean values of a variable @xmath45 in subsets 1 and 2 , respectively , @xmath46 and @xmath47 are the number of data ; @xmath48 and @xmath49 are the sum of residual squared . the calculated value @xmath50 rules out the testing hypothesis .
thus , the opacity ratio of type-1 sources is significantly larger than that of type-2 sources , with the confidence larger than 99 % . even if we take the t - test for logarithms @xmath51 , the values are @xmath52 and @xmath53 for type-1 and -2 sources , respectively .
then , @xmath54 suggests , too , that type-1 sources are more asymmetric than type-2 sources in terms of ffa opacity .
the result is consistent with barthel s unified scheme between rgs and qsos ( barthel 1989 ) .
if the line of sight is close to the jet axis , as thought to be type-1 , a large difference of the path length in external plasma towards the lobes causes an asymmetric ffa . in case of type-2 sources ,
the line of sight is nearly perpendicular to the jet axis , so that a small difference in the path length results in a relatively symmetric ffa . to check if any bias is included in the statistics , we also test the distribution of intrinsic spectral indices @xmath22 and ratios of intrinsic flux densities @xmath55
. the spectral indices are @xmath56 and @xmath57 for type 1 and 2 , respectively .
@xmath58 between @xmath59 and @xmath60 suggests that @xmath22 is likely to be different between the two classes .
the flux density ratios @xmath61 are @xmath62 and @xmath63 for type 1 and 2 , respectively , and @xmath64 indicate no significant difference .
the larger @xmath22 for type-1 objects probably indicates contamination of the core component .
in fact , the flat spectrum of component a in 0646 + 600 and 1333 + 459 implies that these components are the core .
alternatively , let us use @xmath22 of component b ( lobes or jets ) for these two objects , i.e. , @xmath65 and @xmath66 for 0646 + 600 and 1333 + 459 , respectively , as are derived in _
case 2_. the mean @xmath22 for type-1 objects will be @xmath67 , which attributes to no significant difference between that of type-2 objects with @xmath68 . from the above statistical considerations
we conclude that the two groups show no intrinsic difference , but are apparently different in terms of ffa opacities .
what does this mean ? here , we discuss the reason of the apparent difference of the opacity ratios .
this idea is consistent with the unified scheme ( barthel 1989 ) .
it simply explains why the ffa opacities are asymmetric towards type 1s , while no difference in intrinsic properties between type 1 and 2 is found .
the significant difference of the @xmath69 between type-1 and -2 objects ( see figure [ fig : ffahistogram ] ) can be understood in this model : the approaching component would have less path length in the ambient ffa plasma as shown in figure [ fig : unifiedmodel ] .
the external absorption model is self - consistent with the ffa fitting .
this model does not require that ssa is intrinsically asymmetric towards the lobes . in this case
, the inverted spectrum towards the core component is likely due to ssa .
the intrinsic spectral index @xmath22 of a core could be larger than that of a jet or a lobe .
as shown in figure [ fig : ffacorrelation ] , spectral indices of type-1 sources are likely to be larger than those of type-2 sources .
for example , the spectral fits for 0646 + 600 and 1333 + 459 will be much better if a larger @xmath22 is put for components a. in the case of 0646 + 600 , @xmath70 and @xmath71 for components a and b result in the best fit , though we lose degrees of freedom , and we have @xmath72 , @xmath73 , @xmath74 and @xmath75 . in the case of 1333 + 459 ,
@xmath76 and @xmath66 for a and b , respectively , are the best fit to derive @xmath77 , @xmath78 , @xmath79 and @xmath80 . in both cases , components a are likely to be a core , rather than a lobe , because of their flat spectrum . if we assume intrinsic bipolarity of the jet , it is necessary to consider why the counterjet is unseen .
one may consider that the doppler boosting effect possibly causes the apparent one - sided jet .
in this interpretation , component b is approaching towards us to be amplified , while the unseen counterjet is receding to be dimmed .
however , the doppler boosting effect often results in variability of flux densities and a large polarization degree .
these properties are unlikely for gps sources .
an alternative interpretation is that the unseen counterjet is severely obscured via ffa .
this could be an attractive model in which the diffuse emissions opposite to components b in 0646 + 600 and 0738 + 313 can be understood as the counterjets .
deeper imaging capabilities at higher frequencies are necessary to confirm the counterjet .
whatever attenuates the counterjet , a smaller viewing angle in type-1 sources than that of type-2 sources is implied .
thus , this case is also consistent with the barthel s unified scheme .
the path length in the external absorber towards component b of type-1 sources is expected to be shorter than those of type-2 sources ( see figure [ fig : unifiedmodel ] ) .
smaller @xmath69 of type-1 sources than that of type-2 sources can be understood in this context .
when the plasma density decreases as a function of the radius from the nucleus , like the isothermal king model ( king 1972 ) , denser ffa opacities near the center will be produced .
the asymmetry of the opacity can be enhanced , even if the viewing angle remains the same , if the source size is small ( kameno et al . 2001 ) .
this idea is similar to the hypothesis that the ffa plasma is denser towards type-1 sources than towards type-2 sources , as is scaled by the core radius of the ambient ffa plasma .
however , this model requires that the mean opacity of type-1 sources should be larger than that of type-2 sources .
the histogram @xmath81 ( figure [ fig : ffahistogram ] ) shows the opposite behavior , so that this idea is not supported .
consequently , the simplest explanation for the difference in the asymmetry of opacity is the difference of viewing angle .
this should be discussed coupled with the absorption mechanisms , ffa or ssa .
if ssa dominated in the low - frequency cutoff , we expect that the peak frequencies of components a and b would be statistically unbiased in the co - moving frame with each synchrotron emitter .
when the synchrotron sources emanate from the core at a relativistic speed , the peak frequencies of approaching and receding components would become higher and lower , respectively , due to the doppler effect .
together , the flux densities of approaching and receding components would be also amplified and dimmed , respectively , by the doppler effect .
hence , the flux density ratio @xmath82 is expected to correlate with the peak frequency .
however , no significant correlation is found between @xmath83 which is related to the peak frequency , and @xmath84 , with the correlation coefficient of @xmath85 ( see figure [ fig : ffacorrelation ] ) .
these statistics of @xmath31 and @xmath84 does not support the simple ssa - only model , and requires that double lobes are intrinsically asymmetric .
saikia et al .
( 1995 ) reported that compact steep spectrum sources with small linear size are intrinsically asymmetric .
carvalho ( 1998 ) showed that interaction of the jet with a non - homogeneous intragalactic medium can result in asymmetric evolution . since the peak frequency in terms of ssa is anti -
correlated with the linear size ( odea et al .
1998 ) , the intrinsic asymmetry of the ssa opacity can be produced .
the presence of ssa and its intrinsic asymmetry can not be ruled out , nevertheless , the unified scheme coupled with the ffa model can simply account for the difference in asymmetry of opacities .
vlba observations for nine sources at 2.3 , 8.4 , and 15.4 ghz have been carried out , to reveal the morphologies of all objects .
spectral model fitting is applied to obtain spatial distribution of ffa opacities towards the radio emission of individual sources .
a difference between type-1 ( quasars and seyfert 1 galaxies ) and type-2 ( radio and seyfert 2 galaxies ) is found , in terms of the ffa opacity ratio @xmath83 between two components a and b of each object .
type-1 objects tend to show significantly larger opacity ratios than type-2 sources do .
asymmetry in ffa opacities suggests that path lengths through ambient absorbers towards twin lobes are significantly different .
therefore , larger opacity ratios of type 1 objects indicate the axes between lobes are smaller than those of type 2 objects ( see figure [ fig : unifiedmodel ] ) .
this result supports the unified scheme between quasars and radio galaxies proposed by barthel ( 1989 ) .
we the vlba operated by the national radio astronomy observatory , which is a facility of the national science foundation ( nsf ) . this work was supported by grant in aid ( c ) 13640248 from japan society for the promotion of science .
antonucci , r. r. j. , & miller , j. s. 1985 , apj , 197 , 621 | we report very - long - baseline array ( vlba ) observations at 2.3 , 8.4 , and 15.4 ghz towards nine ghz - peaked spectrum ( gps ) sources .
one seyfert 1 galaxy , one seyfert 2 galaxy , three radio galaxies , and four quasars were included in our survey .
we obtained spatial distributions of the free - free absorption ( ffa ) opacity with milliarcsec resolutions for all sources .
it is found that type-1 ( seyfert 1 and quasars ) and type-2 ( seyfert 2 and radio galaxies ) sources showed different distributions of the ffa opacities .
the type-1 sources tend to show more asymmetric opacity distributions towards a double lobe , while those of the type-2 sources are rather symmetric .
our results imply that the different viewing angle of the jet causes the difference of ffa opacity along the external absorber .
this idea supports the unified scheme between quasars and radio galaxies , proposed by barthel ( 1989 ) .
= 2em = 15.5 cm = 22.6 cm = -1.0 cm = 0.5 cm = 0.5 cm @xmath0 national astronomical observatory of japan , 2 - 21 - 1 osawa , mitaka tokyo , japan , 181 - 8588 + [email protected] + [email protected] + [email protected] + @xmath1 the institute of space and astronautical science , 3 - 1 - 1 yoshinodai , sagamihara kanagawa , japan , 229 - 8510 + [email protected] + [email protected] + * keywords : galaxies : active galaxies : nuclei galaxies : jets radio continuum : galaxies techniques : high angular resolution * |
let @xmath3 be an immersion of class @xmath4 of a smooth , compact and oriented two - dimensional manifold @xmath5 into the three dimensional sphere @xmath6 , endowed with the canonical inner product @xmath7 of @xmath8 .
the _ fundamental forms _ of @xmath9 at a point @xmath10 of @xmath11 are the symmetric bilinear forms on @xmath12 defined after spivak @xcite , as follows : @xmath13 here , @xmath14 is the positive unit normal of the immersion @xmath15 tangent to @xmath2 i.e. @xmath16 . in @xmath2 , the coefficients of the second fundamental form @xmath17 in a local positive chart @xmath18 , relative to the normal vector @xmath19 ,
are given by : @xmath20 } { \sqrt { eg -f^2 } } , \ ; f=\frac { \det[\alpha , \alpha_u,\alpha_v,\alpha_{uv}]}{\sqrt{eg - f^2}},\ ; g=\frac { \det[\alpha,\alpha_u,\alpha_v,\alpha_{vv}]}{\sqrt { eg - f^2 } } .$}\ ] ] here @xmath21 , @xmath22 and @xmath23 are the coefficients of the first fundamental form @xmath24 .
the eigenvalues of @xmath25 are called principal curvatures and the corresponding eigenspaces are called principal directions . outside the set of umbilic point @xmath26 where the two principal curvatures coincide , the principal directions define two line fields , called the _ principal line fields _ of @xmath9 , which are of class @xmath27 .
the integral curves of the principal line fields on @xmath28 are called _ principal curvature lines_. the principal curvature lines can be assembled in two one - dimensional orthogonal foliations which will be denoted by @xmath29 and @xmath30 .
the triple @xmath31 is called the _ principal configuration _ of the immersion @xmath15 , @xcite . in a local chart @xmath18
the principal directions of an immersion @xmath15 are defined by the implicit differential equation @xmath32 a principal line @xmath33 is called _ recurrent _ if @xmath34 , where @xmath35 is the _ limit set _ of @xmath33 , and it is called _ dense _ if @xmath36 .
recall that @xmath15 - limit set , @xmath37 , of an oriented leaf @xmath33 is the set of limit points of all convergent sequences of the form @xmath38 , where @xmath39 tends to the positive extreme of definition of the oriented leaf .
the same for the @xmath40 - limit set , @xmath41 , exchanging positive by negative in the previous definition .
clearly , @xmath42 does not depend on the orientation of the leaf .
there is a great similarity between the properties of principal lines in the cases of surfaces in euclidean and in spherical spaces .
they correspond to each other by means of a stereographic projection which , being conformal , preserves the principal lines .
the study of principal curvature lines on surfaces @xmath43 immersed in @xmath1 and @xmath2 is a classical subject of differential geometry .
see @xcite , @xcite , @xcite , @xcite
. in the works of gutierrez and sotomayor , see @xcite , ideas coming from the qualitative theory of differential equations and dynamical systems such as structural stability , recurrence ( and its elimination ) and genericity were introduced into the subject of principal curvature lines .
other differential equations of classical geometry such as asymptotic lines , lines of arithmetic , geometric and harmonic mean curvatures , have been studied in papers by garcia , gutierrez and sotomayor , 1999 , and garcia and sotomayor 1997 , 2001 , 2002 , 2003 . a unification theory for these differential equations
was achieved by garcia and sotomayor , 2004 . a recent survey on classical and recent developments on principal curvature lines , with updated references , can be found in @xcite .
the interest on the study of foliations with recurrent and dense leaves goes back to poincar , birkhoff , denjoy , peixoto , among others .
examples of tori embedded in @xmath1 having only one of the two principal foliation with dense leaves have been given in @xcite .
in this paper are provided examples of embedded tori in @xmath1 with both principal foliations having dense leaves .
these examples , established in theorem [ th:2 ] , are obtained as the stereographic projections of special deformations of the clifford torus in @xmath2 .
these special deformations are defined in the proof of theorem [ th:1p ] .
see section [ sc:3 ] .
in this section will be obtained the variational equations of a quadratic implicit differential equation .
the results will be applied in section [ sc:3 ] . [
prop : eqee ] consider a family of quadratic differential equations , depending on a parameter @xmath44 , of the form @xmath45 let @xmath46 be a solution of with @xmath47 and @xmath48 be a solution of with @xmath49 .
then the following variational equations hold : @xmath50 differentiation of written as @xmath51 leads to : @xmath52 here @xmath53 analogous notation for @xmath54 , @xmath55 and the solution @xmath46 .
evaluation of equation at @xmath56 results in : @xmath57 differentiating equation and evaluating at @xmath58 leads to : @xmath59 similar calculation gives the variational equations for @xmath60 and @xmath61 .
this ends the proof .
[ th:1p ] there are analytic embeddings @xmath62 such that all leaves of both principal foliations , @xmath63 are dense in @xmath0 .
let @xmath64 be the normal vector to the clifford torus @xmath65 , where @xmath66 and @xmath67 , parametrized by @xmath68 we have that , @xmath69 let @xmath70 , @xmath71 being a @xmath72 double periodic function having the symmetry property @xmath73
. for @xmath74 small consider the one parameter family of embedded tori defined by : @xmath75 let @xmath76 be the normal map of @xmath77 .
the coefficients of the first and second fundamental forms of @xmath77 are given by : @xmath78}{4(1+\epsilon^2 h^2)^2}\\ f(u , v,\epsilon)= & \frac{\epsilon [ h_{uv } + \epsilon^2h(h_uh_v - hh_{uv } ) ] } { 2 ( 1+\epsilon^2 h^2)^2 } \\
g(u , v,\epsilon)= & \frac{(1-\epsilon h)[-1 + \epsilon(2h_{vv}-h ) + \epsilon^2(h^2 - 4h_v^2 + 2hh_{vv})+\epsilon^3 h^3]}{4(1+\epsilon^2 h^2)^2}\\ \endaligned $ } \ ] ] from the expressions above , it is important to record the following symmetry relations .
@xmath79 then the coefficients of the differential equation of principal lines of @xmath77 , after multiplication by @xmath80 , are given by : @xmath81 by the equations above and from the symmetry property @xmath73 , it follows that @xmath82 at @xmath58 it follows that @xmath83 consider the reflection @xmath84 defined by @xmath85 and the circle @xmath86 transversal to both principal foliations @xmath87 by equation it follows that @xmath88 and @xmath89 , that is the diffeomorphism @xmath90 preserves the principal configuration of @xmath77 , reversing the order of the foliations .
therefore it also conjugates the poincar first return maps @xmath91 , @xmath92 of the foliations @xmath87 for the basic properties of the rotation number see @xcite and @xcite .
in what follows it will be shown that the rotation number of @xmath93 changes monotonically with @xmath44 , provided it is small . by proposition [ prop : eqee ] with @xmath94
the variational equations of the implicit differential equation @xmath95 with @xmath96 , @xmath97 and @xmath47 are given by : @xmath98 therefore , @xmath99(u , v_0)\endaligned\ ] ] taking @xmath100 it results from the equations above and by equation that : @xmath101 therefore it follows that : @xmath102 consider the poincar map @xmath103 relative to the principal foliation @xmath104 , defined by @xmath105 . here , the circle @xmath106 is a leaf of @xmath30 . therefore @xmath107 and @xmath108 has the following development : @xmath109 therefore the rotation number of @xmath108 changes continuously and monotonically with @xmath44 .
see chapter 12 of @xcite or chapter 5 of @xcite .
the rotation number of @xmath108 is irrational if and only if that of @xmath110 is irrational .
therefore we can take a small @xmath111 such that the rotation numbers of @xmath112 , @xmath113 are both irrational . for @xmath114
all leaves of both principal foliations are dense in @xmath0 .
this ends the proof .
[ th:2 ] there are analytic embeddings @xmath115 such that all leaves of both principal foliations , @xmath116 , are dense in @xmath0
. see figure [ fig : tb ] .
.5 cm stereographic projection of the clifford torus @xmath117 ( left ) and of its @xmath118 deformation for @xmath119 ( right),title="fig : " ] 2.0 cm stereographic projection of the clifford torus @xmath117 ( left ) and of its @xmath118 deformation for @xmath119 ( right),title="fig : " ] take the stereographic projection @xmath120 , @xmath121 , of the example given in theorem [ th:1p ] and define @xmath122 . as @xmath123 is a conformal map
the principal lines of @xmath124 are the images by @xmath123 of the principal lines of @xmath77 .
in this paper it was shown that there exist embeddings of the torus in @xmath2 and @xmath1 with both principal foliations having all their leaves dense .
it is worth mentioning that the consideration of only the first order variational equation , see proposition [ prop : eqee ] , was technically insufficient to achieve the results of this paper .
the same can be said for the technique of local bumpy perturbations used to deform immersions in @xcite , leading to examples with one of the two principal foliations with dense leaves . c. gutierrez and j. sotomayor , lines of curvature and umbilic points on surfaces , @xmath126 brazilian math .
colloquium , rio de janeiro , impa , 1991 .
reprinted as structurally configurations of lines of curvature and umbilic points on surfaces , lima , monografias del imca , ( 1998 ) . | in this paper are given examples of tori @xmath0 embedded in @xmath1 with all their principal lines dense .
these examples are obtained by stereographic projection of deformations of the clifford torus in @xmath2 . |
in nature , there are no true equilibrium phenomena , since all of these require infinite times and infinite thermal reservoirs or perfect insulations . nevertheless , for a large class of systems , it is possible to set up conditions under which predictions from _ equilibrium _ statistical mechanics provide excellent approximations , as many of the inventions of the industrial revolution can attest to .
by contrast , non - equilibrium phenomena are not only ubiquitous , but often elude the powers of the boltzmann - gibbs framework .
unfortunately , to date , the theoretical development of _ non - equilibrium _ statistical mechanics is at a stage comparable to that of its equilibrium counterpart in the days before maxwell and boltzmann . using the intuition developed by studying equilibrium statistical mechanics , we are often `` surprised '' by the behavior displayed by systems far from equilibrium , even if they appear to be in time - independent states . for example
, the well honed arguments , based on the competition between energy and entropy , frequently fail dramatically . in these lectures
, we will present some explorations into the intriguing realm of _ non - equilibrium steady states _
, focusing only on a small class , namely , driven diffusive systems . motivated by both the theoretical interest in non - equilibrium steady states and the physics of fast ionic conductors @xcite , katz , lebowitz and spohn @xcite introduced a deceptively minor modification to a well - known system in equilibrium statistical mechanics : the ising @xcite model with nearest - neighbor interactions . in lattice gas language @xcite ,
the time evolution of this model can be specified by particles hopping to nearest vacant sites , with rates which simulate coupling to a thermal bath as well as an external field , such as gravity .
if `` brick wall '' boundary conditions are imposed in the direction associated with gravity ( particles reflected at the boundary , comparable to a floor or a ceiling ) , and if the rates obey detailed balance , then this system will eventually settle into an _ equilibrium _ state , similar to that of gas molecules in a typical room on earth .
although there is a local bias in the hopping rates ( due to gravity ) , thermal equilibrium is established by an inhomogeneous particle density , at all @xmath0 . however ,
if _ periodic _ boundary conditions are imposed , then translational invariance is completely restored so that , in the final steady state , the particle density is _ homogeneous _
, for all @xmath1 above some finite critical @xmath2 , while a current will be present . clearly ,
for gravity , such boundary conditions can be imposed only in art , as by m.c .
escher @xcite . in physics
, it is possible to establish such a situation with an _ electric _ field , e.g. , by placing the @xmath3 lattice on the surface of a cylinder and applying a linearly increasing magnetic field down the cylinder xis .
if the particles are charged , they will experience a local bias everywhere on the cylinder and a current will persist in the steady state . echoing the physics of fast ionic conductors , we will therefore use the term `` electric '' field to describe the external drive and imagine our particles to be `` charged '' , in their response to this drive .
this is the prototype of a `` driven diffusive system . ''
now , such a system constantly gains ( loses ) energy from ( to ) the external field ( thermal bath ) , so that the time independent state is by no means an equilibrium state , la boltzmann - gibbs .
instead , it is a _ non - equilibrium steady state _
, with an unknown distribution in general . as discovered in the last decade , modifying the ising model to include a simple local bias leads to a large variety of far - from - simple behavior .
the scope of these lectures necessarily limits us to only a bird s eye view . since
a more extensive review has been published recently @xcite , we will restrict these notes to a brief `` introduction '' to this subject . instead , we choose to include some developments since that review . for completeness ,
we summarize , in section ii , the lattice model introduced by kls and the langevin equation believed to capture its essence in the long - time large - distance regime . the next section ( iii ) is devoted to some of the surprising behavior displayed by this system , at temperatures far above , near , and well below @xmath2 .
a collection of recent developments is presented in section iv . in the final section ( v ) , we conclude with a brief summary .
on a square lattice , with @xmath4 sites and toroidal boundary conditions , a particle or a hole may occupy each site .
a configuration is specified by the occupation numbers @xmath5 , where @xmath6 is a site label and @xmath7 is either 1 or 0 .
occasionally , we also use spin language , defining @xmath8 . to access the critical point ,
half - filled lattices are generally used in monte carlo simulations : @xmath9 @xmath10 .
the particles are endowed with nearest - neighbor attraction ( ferromagnetic , in spin language ) , modeled in the usual way through the hamiltonian : @xmath11 , with @xmath12 .
the external drive , with strength @xmath13 and pointing in the @xmath14 direction , will bias in favor of particles hopping `` downwards '' . to simulate coupling to a thermal bath at temperature @xmath1 ,
the metropolis algorithm is typically used , i.e. , the contents of a randomly chosen , nearest - neighbor , particle - hole pair are exchanged , with a probability @xmath15 $ ] . here , @xmath16 is the change in @xmath17 after the exchange and @xmath18 , for a particle attempting to hop ( against , orthogonal to , along ) the drive . note that these dynamic rules conserve particle number . for @xmath19
, this system will eventually settle into an equilibrium state which is precisely the static ising model @xcite . in the thermodynamic limit ,
it is known to undergo a second order phase transition at the onsager critical temperature @xmath20 . when driven ( @xmath21 )
, this system displays the same qualitative properties , i.e. , there is a disordered phase for large @xmath1 , followed by a second order transition into a phase - segregated state for low @xmath1 .
however , with more scrutiny , this superficial similarity gives way to puzzling surprises .
in particular , @xmath22 appears to be monotonically _ increasing _ with @xmath23 , saturating at about @xmath24 @xcite for @xmath25 .
why should @xmath26 be surprising ? consider the following `` argument '' .
for very large @xmath23 , hopping along @xmath27 becomes like a random walk , in that @xmath16 is irrelevant for the rates .
therefore , hops along @xmath27 might as well be coupled to a thermal bath at _ infinite _ @xmath1 ( apart from the bias ) .
indeed , recall that our system gains energy from @xmath23 and loses it to the bath , so that any drive may be thought of as a coupling to a second reservoir with higher temperature .
then , it seems reasonable that @xmath1 must be _ lowered _ to achieve ordering , since this extra reservoir pumps in a higher level of noise , helping to _ disorder _ the system ! to date , there is neither a convincing argument nor a computation which predicts the correct _ sign _ of @xmath28 , let alone the magnitude . in the next section , we will briefly review other puzzling discoveries , only a few of which are understood . to understand collective behavior in the long - time and large - scale limit
, we often rely on continuum descriptions , which are hopefully universal to some extent .
successful examples include hydrodynamics and landau - ginzburg theories .
certainly , the @xmath29 theory , enhanced by renormalization group techniques , offers excellent predictions for both the statics and the dynamics near equilibrium of the ising model .
following these lines , we formulate a continuum theory for the kls model , in arbitrary dimension @xmath30 . in principle , such a description can be obtained by coarse - graining the microscopic dynamics @xcite but we will pursue a more phenomenological approach here . seeking a theory in the long - time , large wavelength limit , we first identify the slow variables of the theory . these are typically * ordering fields which experience critical slowing down near @xmath2 , and * any conserved densities .
the kls model is particularly simple since it involves a single ordering field , namely the local `` magnetization '' or excess particle density , @xmath31 , which is also the only conserved quantity . here
, @xmath32 stands for @xmath33 .
the last entry denotes the one - dimensional `` parallel '' subspace selected by @xmath23 .
thus , we begin with a continuity equation , @xmath34 , and postulate an appropriate form for the current @xmath35 . in the absence of @xmath23 , it simply takes its model b @xcite form , @xmath36 , with the landau - ginzburg hamiltonian @xmath37 . as usual , @xmath38 and @xmath39 . while the first contribution to @xmath35 is a deterministic term , reflecting local chemical potential gradients , the second term , @xmath40 , models the thermal noise .
the noise is gaussian distributed , with zero mean and positive second moment proportional to the unit matrix , i.e. , @xmath41 , @xmath42 . in the presence of @xmath23
, we should expect at least two modifications to @xmath43 , namely first , an additional contribution @xmath44 modeling the nonvanishing mass transport through the system and second , the generation of anisotropies , since @xmath23 singles out a specific direction . by virtue of the excluded volume constraint , the `` ohmic '' current @xmath44 vanishes at densities @xmath45 and @xmath46 , corresponding to @xmath47 .
writing @xmath48 for the coarse - grained counterpart of @xmath23 , pointing along unit vector @xmath49 , the simplest form is therefore @xmath50{\bf \hat{y}}$ ] , where the @xmath51 corrections will turn out to be irrelevant for universal properties .
next , we consider possible anisotropies . clearly , all @xmath52 operators should be split into components parallel ( @xmath53 ) and transverse ( @xmath54 ) to @xmath23 , accompanied by different coefficients . for example , the anisotropic version of the model b term @xmath55 will read @xmath56 @xmath57 , with two different `` diffusion '' coefficients for the parallel and transverse subspaces .
should we expect both of these to vanish as @xmath1 approaches @xmath58 ? or just one of them - but which one ?
recalling that the lowering of @xmath59 , in the equilibrium system , is a consequence of the presence of interparticle interactions , we argue that @xmath23 , stirring parallel jumps much more effectively than transverse ones , should _ counteract _ this effect in the parallel direction , having less of an impact in the transverse subspace .
thus , we anticipate that generically @xmath60 , so that criticality , in particular , is marked by @xmath61 vanishing at positive @xmath62 .
this is borne out by the structure of typical ordered configurations , namely , single strips aligned with @xmath23 , indicating that `` antidiffusion '' , i.e. , @xmath63 , dominates in the transverse directions below @xmath2 .
finally , the drive also induces anisotropies in the noise terms so that the second moment of their distribution should be taken as @xmath64 .
since the transverse subspace is still fully isotropic , we define @xmath65 .
generically , however , we should expect @xmath66 @xmath67 .
summarizing , we write down the full langevin equation : @xmath68 with noise correlations @xmath69 this equation forms the basis for the analytic study of the kls model .
fortunately , we need not be concerned about the detailed dependence of its coefficients on the microscopic @xmath1 , @xmath70 , and @xmath23 .
while they are in principle calculable within an explicit coarse - graining scheme , the properties that we already outlined above suffice for our purposes . moreover , it turns out that we can simplify the most general form , ( [ lang ] ) , depending on the temperature regime considered .
we will return to this discussion in sections 3.1 and 3.3 below . to conclude this section ,
let us take a glance at the structure of ( [ lang ] ) .
its basic form is @xmath71 , with noise correlations @xmath72 and @xmath73 just a positive constant . for simplicity ,
we restrict ourselves to a scalar order parameter @xmath74 and noise `` matrices '' @xmath75 which do not depend on @xmath74 ( more general cases are discussed in , e.g. , @xcite and @xcite ) .
@xmath76 is a functional of @xmath74 and its derivatives . given this basic form , a steady - state solution @xmath77 $ ] for the configurational probability is easily found @xcite _ provided _ @xmath78 is `` hamiltonian '' , i.e. , if it can be written as @xmath75 acting on a total functional derivative : @xmath79 . in this case
, @xmath80 is simply proportional to @xmath81 , irrespective of the choice of @xmath75 .
we will refer to such a dynamics , in an operational sense @xcite , as `` satisfying the fluctuation - dissipation theorem '' ( fdt ) @xcite .
model b falls into this category , with @xmath82 .
in contrast , for our driven system , @xmath83 is just the expression in the @xmath84 brackets of ( [ lang ] ) and @xmath85 , so that @xmath86 .
thus , this @xmath83 is clearly _ not _ hamiltonian ! in this fashion ,
our continuum theory reflects the fact that we are dealing with a generic non - equilibrium steady state .
for completeness , we should point out that certain microscopically non - hamiltonian dynamics may _ become _ hamiltonian if viewed on sufficiently large length scales .
the discussion of these subtleties @xcite , while intriguing , is beyond the scope of these lectures .
for the @xmath3 ising model in equilibrium , thermodynamic quantities are typically analytic , except at a single point , i.e. , @xmath2 . in particular , being a model with only short - range interactions , correlation functions are short - ranged far above @xmath2 , decaying exponentially with distance and controlled by a finite correlation length . far below criticality ,
a half filled system in a square geometry will display two strips of equal width , as a result of the co - existence of a particle rich ( dense ) phase with hole - rich one .
correlations of the fluctuations within each phase are also short - ranged .
more interestingly , the interfaces between the phases do represent soft degrees of freedom , being the goldstone mode of a broken translational invariance @xcite .
one consequence is a divergent structure factor ( fourier transform of the two - point correlation , of deviations from being straight ) .
although this behavior is singular , it is well understood and can be related to that of a simple random walk @xcite . only at criticality does the system possess non - trivial singular properties , the _ nature _ of which became transparent only in the 70s@xcite despite onsager s tour - de - force in 1944@xcite . when driven into _ non - equilibrium steady states _ , this picture changes dramatically .
in particular , non - trivial singular behavior appears at all @xmath1 . while the transition itself remains second order , its properties fall into a non - ising class . here
, we will present some of these surprising features briefly , referring the reader to @xcite for more details . focusing on the behavior far above criticality ,
let us study the two - point correlation @xmath88 ( @xmath89 ) and its transform , the structure factor , @xmath90 .
we first point out that the drive clearly induces an asymmetry between @xmath91 and @xmath27 , i.e. , anisotropy beyond that due to the lattice .
thus , we should not be surprised if the familiar ornstein zernike form of @xmath92 ( @xmath93 ) were to become elliptical rather than circular , e.g. , similar to fig.1a .
however , simulations @xcite showed that there is a _
singularity in @xmath92 at the origin , as in fig.1b . to be more precise
, we may define @xmath94 and measure the discontinuity by @xmath95 . though @xmath96 does approach unity for @xmath97 ,
it is about 4 , for large @xmath23 , even at twice the critical temperature .
further , it diverges as @xmath98 ! we should remind the reader that , for the equilibrium case , @xmath92 may diverge at criticality , but @xmath96 is unity always .
strange as it may seem , this behavior can be understood within the context of the continuum approach , eqn .
( [ lang ] ) . for @xmath99 , _ both _
@xmath100 _ and _ @xmath61 are positive .
thus , the local magnetization @xmath31 fluctuates around a stable minimum at zero , and neither the fourth - order derivatives nor the nonlinearities are necessary for stability . as a consequence , the behavior of the system in the disordered phase can be described by a much simpler , _ linear _ langevin equation , namely , @xmath101 with the help of a fourier transform , we can easily find @xmath102 for any realization of @xmath40 , and then compute averages over the gaussian distributed @xmath40 , with its second moment given by ( [ lang ] ) .
this forms the starting point for the discussion of the disordered phase , quantified by , e.g. , equal - time structure factors or three - point functions @xcite . here
, we focus on the former , @xmath103 , which can be easily computed : @xmath104 so that @xmath105 . in order for this @xmath92 to reduce to the equilibrium ornstein - zernike form
, the fdt has to be invoked , which constrains @xmath96 to unity . for the driven case ,
@xmath96 is no longer constrained , so that a discontinuity develops .
one consequence of such a singularity in @xmath92 is that @xmath106 becomes _ long - ranged _ , decaying as @xmath107 at large @xmath108 .
the amplitude , in addition to being @xmath109 , has a dipolar angular dependence , so that an appropriate angular average of @xmath106 is again short - ranged @xcite . to end this discussion ,
let us follow g. grinstein s argument @xcite that such long - ranged power law decays _ should be expected_. starting with the dynamic two - point correlation @xmath110 , we know that the conservation law leads to the autocorrelation : @xmath111 .
in addition , being a diffusive system , we should expect @xmath112 , at least far from criticality .
using _ naive _ scaling , we would write @xmath113 ! in other words , the generic behavior of @xmath106 in a diffusive system is @xmath114 decay , _ not exponential_. in this case , the equilibrium system is the non - generic one , in which the amplitude of this generic term is forced to vanish , by fdt .
our familiarity with equilibrium systems is so strong that , on first sight , power law decays far above @xmath2 appear quite surprising .
finally , note that a scaling argument of this kind can not produce the angular dependence , a crucial feature of the generic singularities of driven diffusive systems .
next , we turn to systems far below @xmath2 . due to the conservation law
, they typically display phase co - existence , so that the dominant fluctuations and slow modes are those associated with the interface between the phases .
again , for the @xmath3 ising model in equilibrium , there is a wealth of information on the properties of the interface @xcite . in particular , being a _
one_-dimensional object , the interface exhibits behavior identical to a simple random walk @xcite at the large scales , such as widths diverging as @xmath115 . in our case , if we study an interface aligned along the @xmath27 axis , then , to a good approximation , we may specify the configuration by its position along @xmath91 by the `` height '' function @xmath116 .
known as capillary waves @xcite , these fluctuations have a venerable history .
since the interface is a manifestation of broken translational invariance , @xmath116 are the soft goldstone modes @xcite , so that the associated structure factor @xmath117 diverges as @xmath118 .
of course , this property is just the fourier version of the @xmath115 divergence : @xmath119 .
interfaces with divergent widths are also known as `` rough '' ; only for @xmath120 may interfaces in crystalline ising - like models display transitions from rough to smooth phases .
when _ driven _ , however , the interface width appears to approach a _
finite width _ at large distances .
in particular , for @xmath121 , using lattices up to @xmath122 and plotting the widths vs. @xmath123 with various values of @xmath124 , we found that the curves did not straighten out with @xmath125 as low as 0.05 @xcite . as a phenomenon ,
roughness suppression is not novel , gravity being the most common example .
however , the drive here is _
parallel _ to the interface , reminiscent of wind driving across a water surface , which has a _ destabilizing _ effect by contrast .
subsequently , a more detailed simulation study @xcite of the interface with @xmath126 showed that @xmath127 for large @xmath128 , crossing over to @xmath129 for small @xmath128 ( fig .
2 ) . since @xmath130 ,
the small @xmath128 behavior is consistent with @xmath131 .
though @xmath132 appears temptingly close to a simple rational : @xmath133 , there is no viable theory so far ( despite two valiant attempts @xcite ) . finally , let us note that the crossover from rough to smooth occurs at the scale of @xmath134 .
although no precise monte carlo analysis of this crossover has been performed , it is consistent with @xmath23 having the units of 1/length . in this respect ,
such a length is comparable to the capillary length which controls the crossover in the gravitationally stabilized interface .
of course , in that case , the small @xmath128 behavior is simply @xmath135 ! finally , we turn to the critical region , described by @xmath136 in ( [ lang ] ) .
in contrast to the situation for @xmath99 , we now need fourth - order terms in @xmath137 to stabilize the system against large - wavelength fluctuations .
similarly , the nonlinear terms are required to ensure a stable ordered phase below @xmath2 .
however , we still have @xmath138 , so that fourth - order _ parallel _ gradients may safely be neglected .
thus , near criticality , the leading terms on the right hand side of ( [ lang ] ) are @xmath139 and @xmath140 , implying that parallel and transverse wave vectors , @xmath141 and @xmath142 , scale naively as @xmath143 , i.e. , parallel gradients are less relevant than transverse ones .
more systematically , a naive dimensional analysis @xcite reveals that the nonlinearity associated with @xmath48 is the most relevant one , having an upper critical dimension @xmath144 , distinct from the usual value of @xmath145 for the ising model .
dropping irrelevant terms , ( [ lang ] ) simplifies to @xmath146 with noise term ( @xmath147 ) @xmath148 here , we have rescaled @xmath62 to @xmath45 and have kept the ( naively irrelevant ) nonlinearity associated with @xmath149 , to ensure a stable theory below @xmath2 .
this is the starting point for the analysis of universal critical behavior . in this regime
, large fluctuations on all length scales dominate the behavior of the system , so that renormalization group techniques are indispensable .
to summarize very briefly , the langevin equation ( [ laec ] ) is recast as a dynamic functional @xcite , followed by a renormalized perturbation expansion in @xmath150 @xcite .
the quartic coupling @xmath149 must be treated as a dangerously irrelevant operator .
gratifyingly , the series for the critical exponents can be summed , so that we obtain _ quantitatively _ reliable values even in two dimensions .
the details are quite technical @xcite , so that we just review the key results here .
the discussion leading to ( [ laec ] ) already suggests that the critical behavior of the driven system is distinct from its equilibrium counterpart : the upper critical dimension is shifted to @xmath144 , and parallel and transverse wave vectors scale with different exponents .
anticipating renormalization , we reformulate their scaling as @xmath151 , introducing the _ strong anisotropy _
exponent @xmath152 . to illustrate its importance ,
let us consider the wave vector scaling for the equilibrium ising model .
for isotropic exchange interactions , coarse - graining results in the usual landau - ginzburg hamiltonian with gradient term @xmath153 . clearly , this can not lead to anything but @xmath154 .
the _ only _ effect of anisotropies in the microscopic couplings is to give rise to different _ amplitudes _ , so that @xmath152 remains zero .
we refer to this situation as _ weak _ anisotropy , in contrast to _
strong _ anisotropy where @xmath155 .
examples of the latter in equilibrium models include , e.g. , lifshitz points or structural phase transitions @xcite . for any system with strong anisotropy ,
irrespective of its universality class , the renormalization group predicts the general scaling form of , e.g. , the dynamic structure factor near criticality : @xmath156 here , @xmath157 is just a scaling factor .
( [ sf ] ) can be viewed as a _ definition _ of the critical exponents @xmath158 , @xmath159 , @xmath160 and @xmath152 . the appearance of the latter is of course consistent with our earlier discussion .
different universality classes are distinguished by the characteristic values of these exponents , expressed , e.g. , through their @xmath161-expansions . for our model ,
all exponents , _ except _ @xmath152 , take their mean - field values : @xmath162 , @xmath163 , @xmath164 while @xmath165 . a separate analysis yields the order parameter exponent @xmath166 .
note that all of these equalities are _ exact _
, i.e. , all higher order terms in the @xmath167-expansion vanish !
care must be taken on two fronts , both associated with the presence of strong anisotropy , when comparing these exponents to monte carlo data .
first , in the equilibrium ising model , the _ same _ exponent @xmath160 characterizes the divergence of the critical structure factor near the origin in @xmath168-space , @xmath169 as @xmath170 .
in contrast , _ four _ @xmath160-like exponents are needed in the driven case ! fortunately ,
scaling laws relate them to @xmath160 and @xmath152 . for example
, we can _ define _ @xmath171 via @xmath172 , for @xmath173 . using ( [ sf ] )
, we read off the simple result @xmath174 .
similarly , we introduce @xmath175 through the relation @xmath176 , for @xmath177 . keeping track of @xmath152 in the fourier transform
, we obtain the less trivial scaling law @xmath178 . since most simulations have focused on two - dimensional systems , we set @xmath179 , predicting @xmath180 and @xmath181 .
monte carlo data for the pair correlation along the drive beautifully display the expected @xmath182 decay @xcite .
numerical evidence for the strong anisotropy exponent is somewhat more indirect , extracted from an anisotropic finite size scaling analysis @xcite . convincing data collapse is obtained , using anisotropic systems of size @xmath183 with fixed `` aspect '' ratio @xmath184 , for the theoretically predicted values of the exponents .
it is intriguing that the signals of continuous phase transitions in or near equilibrium , namely , a diverging length scale , scale invariance and universal behavior , also mark such transitions in far - from - equilibrium scenarios .
beyond the topics discussed in the previous section , we may arbitrarily name three other `` levels '' of non - equilibrium steady - state systems . the first
are associated with the kls model itself , including fascinating results on higher correlations @xcite , failure of the cahn - hilliard approach to coarsening dynamics @xcite , models with _ shifted periodic _ and/or _ open _ boundary conditions @xcite , and systems with ac or random drives @xcite .
topics in the next `` level '' involve various generalizations of kls , such as other interactions @xcite , multi - layer systems , multi - species models , and systems with quenched impurities @xcite .
further `` afield '' is a wide range of driven systems , e.g. , surface growth , electrophoresis and sedimentation , granular and traffic flow , biological and geological systems , etc . in these brief lecture notes
, we present two topics at the `` intermediate level '' : a driven bi - layer system and a model with two species .
the physical motivation for considering multi - layered systems may be traced to intercalated compounds @xcite .
the process of intercalation , where foreign atoms diffuse into a layered host material , is well suited for modeling by a driven lattice gas of several layers . in both physical systems and monte carlo simulations of a model with realistic parameters @xcite ,
finger formation has been observed .
both are transient phenomena , so that we might ask : are there any novel phenomena in the driven _ steady states _ ? on the theoretical front , there are two motivations . to study critical behavior of the single layer kls model ,
it is necessary to set the overall density at 1/2 , with the consequence that two interfaces develop below criticality .
these interfaces seriously complicate the analysis , since their critical behavior is quite distinct from the bulk .
one way to avoid these difficulties is to consider a bi - layer system @xcite , with _ no inter - layer interactions_. through particle exchange between the layers , the system may order into a dense and a hole - rich layer , neither of which has interfaces .
certainly , this is expected ( and observed ) in the equilibrium case , where the steady state is just two ising models , decoupled except for the overall conservation law .
the other is using multilayer models as `` interpolations '' from two- to three - dimensional systems @xcite .
motivations aside , what the simulations reveal is entirely unexpected . in the simplest generalization ,
a bi - layer system is considered as a fully periodic @xmath185 model .
the first simulations were carried out on the case with zero inter - layer interactions @xcite .
therefore , within each layer , all aspects are identical to the kls model .
particle exchanges _ across _ the layers , unaffected by the drive , are updated according to the local energetics alone , so that these jump rates are identical to a system in equilibrium .
the overall particle density is fixed at 1/2 . in the absence of @xmath23
, there would be a single second order transition at the onsager temperature .
since the inter - layer coupling is zero , phase segregation at low temperatures will occur across the layers , i.e. , the ordered phase is characterized by homogeneous layers of different densities ( opposite magnetizations , in spin language ) . having no interfaces , the free energy of the system is clearly lowest in such a state . at @xmath186 ,
one layer will be full and the other empty , so that this state will be labeled by fe .
intriguingly , when @xmath23 is turned on , _ two _ transitions appear @xcite ! as @xmath1 is lowered , the homogeneous , disordered ( d ) phase first gives way to a state with strips in both layers , reminiscent of two entirely unrelated , yet aligned , single layer driven systems .
we will refer to this state as the strip phase ( s ) .
as @xmath1 is lowered further , a first order transition takes the s phase into the fe state .
why the s state should interpose between the d and fe states was not understood .
this puzzle , together with the presence of interlayer couplings in intercalated compounds , motivated our study of the _ interacting _ bi - layer driven system @xcite . within this wider context , the presence of two transitions is no longer a total mystery . on the other hand , this study reveals several unexpected features , leading to interesting new questions .
our system consists of two @xmath187 ising models with attractive interparticle interactions of _ strength unity_. arranged in a bi - layer structure , the _ inter - layer _ interaction is specified by @xmath70 .
thus , the `` internal '' hamiltonian is just @xmath188 where the first sum is over nearest neighbors _ within _ a given layer while @xmath7 and @xmath189 differ only by the layer index .
the external drive , @xmath23 , is imposed through the jump rates , as in section ii . for simulations
, we kept the overall density at 1/2 and chose @xmath190 @xmath191 $ ] .
note that negative @xmath70 s are especially appropriate for intercalated materials @xcite .
the eventual goal of such a study is to map out the phase diagram in the @xmath1-@xmath70-@xmath23 space .
so far , we have data for only three ( positive ) values of @xmath23 .
not expecting further surprises , we believe that we have uncovered the main features of the phase diagram .
to begin , let us point out various features in the equilibrium case . here
, the @xmath12 and the @xmath192 systems can be mapped into each other by a gauge transformation .
thus , in the thermodynamic limit , the phase diagram in the @xmath193-@xmath70 plane is exactly symmetric .
however , for the lattice gas constrained by a conservation law , the low temperature states of these two systems are not the same , being s or fe , respectively . while the d - s or d - fe transitions are second order ( shown as a line in fig .
3 ) , the s - fe line will be first order in nature ( @xmath1 axis from 0 to 1 in fig .
due to the presence of interfaces , the fe domain includes the @xmath194 line .
for the same reason , it `` intrudes '' into the @xmath12 half plane by @xmath195 , for finite systems . of course , on the @xmath1 axis lies a pair of decoupled ising systems , so that @xmath196=@xmath197 assumes the onsager value , as @xmath198 . in general , @xmath199 is not known exactly , though @xmath196=@xmath200 is precisely @xmath201=@xmath197 , since every configuration of the two layers can be mapped into one in the usual ising model .
the lesson from equilibrium is now clear : the s state is as generic as the fe state in this wider context . as we will see below
, the presence of two transitions does not represent a qualitative change from the equilibrium phase structure . turning to driven systems
, we can no longer expect a symmetric phase diagram , since the drive breaks the ising symmetry .
the data for @xmath202 are shown in fig .
3 , in which the first / second order transitions are marked as solid / open circles .
we see that the only effect of the drive is to shift the phase boundaries .
no new phases appear while the nature of the transitions remains unchanged .
however , there are remarkable features .
one is the _ lowering _ of the critical temperature for large @xmath203 . given that @xmath204 in the kls model , it is quite unexpected that @xmath205 is _ less _ than its equilibrium counterpart .
perhaps more notable is the presence of a small region in the @xmath192 half plane , in which an s - phase is stable .
since particle - rich strips lie on top of each other , such a phase could not exist if either energy or entropy were to dominate the steady state .
concerned that its presence might be a finite size effect , we performed simulations using @xmath206 s up to @xmath207 , with @xmath208 and @xmath209 $ ] . in all cases
, the s - phase prevailed , leading us to conjecture that this region exists even in the thermodynamic limit . on the other hand , for lower @xmath1
, the fe - phase penetrates into the @xmath12 half plane , as in equilibrium cases .
energetics seem to take the upper hand here .
though we believe that , as @xmath198 , this part of the phase boundary will collapse onto the @xmath210 line , we have not checked the finite size effects explicitly .
of course , these `` intrusions '' into the positive / negative @xmath70 region by the fe / s - phase are responsible for the appearance of _ two _ transitions in the studies of the @xmath194 model @xcite
. it would clearly be useful to develop some intuitive arguments which can `` predict '' these qualitative modification of the equilibrium phase structure .
since the usual energy - entropy considerations appear to fail , our attempt @xcite is based on the competition between suppression of short - range correlations @xcite and enhancement of long - range ones @xcite , as a result of driving .
let us focus on the two - point correlations in the disordered phase with @xmath1 being lowered from @xmath211 and see how they are affected by the drive .
on the one hand , the _ nearest - neighbor _ correlations are found to be suppressed by @xmath23 , consistent with the picture that the drive acts as an extra noise in breaking bonds . taken alone , this suppression would lead to the lowering of the critical temperature , as pointed out in section ii . on the other hand ,
as we showed above , the drive changes significantly the large distance behavior of @xmath106 , from an exponential to a power law .
further , the amplitude is positive ( negative ) for correlations parallel ( transverse ) to @xmath23 .
both the positive longitudinal correlations and the negative transverse ones should help the process of ordering into strips parallel to the field .
in other words , the enhanced long - range parts favor the s - phase . taken alone , we expect this effect to increase @xmath58 .
evidently , for the single layer case , the latter effect `` wins '' . for a bi - layer system ,
we need to take into account cross - layer correlations , which are _ necessarily _ short - ranged .
focusing first on the @xmath210 case , where short - range effects due to cross - layer interactions are absent , we are led to @xmath212 . indeed , this ratio is comparable to that in the single - layer case .
next , we consider systems with positive @xmath70 . without the drive ,
@xmath213 is of course enhanced over @xmath214 . for non - vanishing drive
, it is not possible to track the competition between the short- and long - range properties of the transverse correlations .
evidently , for small @xmath70 , the long - range part still dominates , so that @xmath215 however , for @xmath216 , the presence of @xmath23 effectively lowers @xmath70 , since the latter is associated with only short - range correlations .
since a lower effective @xmath70 naturally gives rise to a lower @xmath87 , we would `` predict '' that @xmath217 could decrease considerably as @xmath70 increases . from fig.3
, we see that @xmath218 for @xmath219 the interplay of the competing effects is so subtle that either can dominate , in different regions of the phase diagram . finally , we turn to the @xmath192 case . here
, the two effects tend to co - operate rather than compete , since the long - range parts favor the s - phase over the fe - phase . as a result , the domain of fe is smaller everywhere . in particular , note that the @xmath192 branch of @xmath220 is significantly lower than the @xmath12 branch .
the small region of the s - phase can be similarly understood , at least at the qualitative level . in this picture
, we may argue that the bicritical point and its trailing first order line should be `` driven '' to the @xmath192 half - plane . to end this subsection
, we point out a few other interesting features .
the order parameters of the s and fe phases are conserved and non - conserved , respectively .
based on symmetry arguments , we may expect that the critical properties of the two second order branches will fall into different universality classes .
in contrast , the gauge symmetry forces _ static equilibrium _
properties to be identical along these two branches @xcite .
since the drive breaks this symmetry , we can only speculate that the d - s transitions belong to the same class as the single layer , kls model @xcite while the d - fe ones should remain in the equilibrium ising class @xcite .
since these two classes are distinct , we can also expect a rich crossover structure near the bi - critical point .
assuming these conjectures are confirmed , we can truly marvel at the novelties a driving field can bring .
if the particles in the kls model are considered `` charged '' , it is natural to explore the effects of having both positive and negative charges . defined on a fully periodic @xmath4 square lattice ,
a configuration of such a `` two species '' model can be characterized by two local occupation variables , @xmath221 and @xmath222 , which equal @xmath45 ( @xmath46 ) , if a positive or negative particle is present ( absent ) at site @xmath223 . in spin language
, this corresponds to a spin-@xmath45 model , and novel behavior is to be expected , since we have altered the internal symmetries of the local order parameter . for simplicity
, we will restrict ourselves to zero total charge , @xmath224 . pursuing the analogy to electric charges ,
positive ( negative ) charges move preferentially along ( against ) the drive @xmath23 which is directed along the @xmath14 axis . as a first step ,
we assume that there are no interparticle interactions apart from an excluded volume constraint .
thus , our model can be viewed as the high - temperature , large @xmath23 limit of a more complicated interacting system . to model biased diffusion , a particle with charge @xmath225 jumps onto a nearest - neighbor empty site according to @xmath226 $ ] , where @xmath227 is the change in the @xmath27-coordinate of the particle . to allow charge exchange between neighboring particles , two nearest - neighbor sites carrying opposite charge may exchange their content with probability @xmath228 $ ] .
now , @xmath229 is the change in the @xmath27-coordinate of the _ positive _ particle .
the parameter @xmath230 sets the ratio of the characteristic time scales controlling , respectively , charge exchange and diffusion .
physical motivations for this model come from various directions : fast ionic conductors with several species of mobile charges @xcite , electric breakdown of water - in - oil microemulsions @xcite , and gel electrophoresis of charged polymers @xcite .
it can also be interpreted as a simple model for some traffic or granular flows @xcite .
summarizing our simulation data , we first map out the phase diagram of the model , in the space spanned by @xmath23 , the total mass density @xmath231 , and @xmath230 @xcite . focusing on small @xmath230 ,
the diffusive dynamics is limited by the excluded volume constraint .
for sufficiently small @xmath23 and @xmath232 , typical configurations are disordered , characterized by homogeneous charge and mass densities and fairly large currents . nevertheless , this phase is highly nontrivial , supporting anomalous two - point correlations reminiscent of the kls model : in generic directions , the familiar @xmath233 decay prevails , with the _ remarkable _ exception of the field direction , where a _ novel _
@xmath234 dominates @xcite . with increasing @xmath23 or @xmath232 , the tendency of the particles to impede one
another becomes more pronounced , until a phase transition into a spatially inhomogeneous `` ordered '' phase occurs . beyond the transition line @xmath235 @xmath236
, typical configurations are charge - segregated , consisting of a strip of mostly positive charges `` floating '' on a similar strip of negative charges , surrounded by a background of holes . to explore these transitions more quantitatively , we need to define a suitable order parameter . taking the fourier transforms of the local _ hole _ and _ charge _ densities , @xmath237\exp \{ik_{\bot } x+ik_{\vert } y\}$ ] and @xmath238\exp \{ik_{\bot } x+ik_{\vert } y\}$ ]
, we select the components @xmath239 and @xmath240 since these are most sensitive to ordering into a single transverse strip .
naturally , we choose their averages @xmath241 and @xmath242 as our order parameters .
to distinguish first from second order transitions , we also measure their fluctuations and histograms .
the resulting phase diagram is shown in fig .
typically , hysteresis loops in @xmath243 , @xmath244 and the average current are observed only for @xmath245 , indicating that the transitions are first order in this regime . on the other hand ,
fluctuations develop sharp peaks for larger mass densities , signalling continuous transitions .
larger values of @xmath230 favor the disordered phase , so that the whole transition line shifts to higher @xmath23 , the region between the spinodals narrows , and @xmath246 moves to higher mass densities .
thus , we observe a surface of first order transitions , separated from a surface of continuous ones by a line of multicritical points : @xmath247 .
while it is not surprising that the disordered phase is stable for large @xmath23 provided the mass density is sufficiently small , it may appear rather counterintuitive that it should also dominate the @xmath248 region .
however , setting @xmath249 eliminates every hole in the system so that the dynamics is carried entirely by the charge exchange mechanism .
relabelling positive charges as `` particles '' and negative ones as `` holes '' , the model becomes equivalent to a noninteracting ( i.e. , @xmath210 ) kls model . also termed the asymmetric simple exclusion process ( asep ) in the literature , its steady - state solution is exactly known @xcite to be homogeneous .
thus , for all @xmath250 , our two - species model remains disordered along the entire @xmath249 line .
for sufficiently large @xmath251 , the data indicate quite clearly that a _ finite _ region of disorder persists , for any @xmath23 , just _
below _ complete filling . for smaller
@xmath230 , however , it is less obvious whether such a region exists in the thermodynamic limit , since even a single hole can suffice to induce spatial inhomogeneities in a _
finite _ system : acting as a catalyst for the charge segregation process , the hole creates a strip of predominantly positive charge , separated by a rather sharp interface from a similar , negatively charged domain , located `` downstream '' .
the charge exchange mechanism tends to remix the charges but has little effect for small @xmath252 .
eventually , the hole ends up trapped in the interfacial region !
finally , we turn to larger values of @xmath230 , specifically , @xmath253 : here , the charge exchange mechanism dominates over the excluded volume constraint and suppresses the ordered phase completely .
so far , our discussion has mostly drawn upon monte carlo results .
it is gratifying , however , that the same qualitative picture emerges from a mean - field theory , based upon a set of equations of motion for the local densities .
we briefly summarize the analytic route .
since the numbers of both positive and negative charges are conserved , we begin with a continuity equation for the coarse - grained densities @xmath254 : @xmath255 . for simplicity , we focus on the case @xmath256 , i.e. , particle - hole exchanges , first . as for the kls model , @xmath257 can be written as the sum of a diffusive piece and an ohmic term , @xmath258 .
the density - dependent mobility @xmath259 must vanish with both @xmath260 and the local hole density , @xmath261 , i.e. , @xmath262 . the `` hamiltonian '' @xmath263 is just the mixing entropy associated with distributing @xmath264 positive and @xmath265 negative charges over a lattice of @xmath266 sites .
note that the functional derivative @xmath267 is taken at fixed @xmath268 since we are focusing on particle - hole exchanges here . to model the charge exchange mechanism , we simply add a similar term to @xmath257 , namely , @xmath269 where @xmath270 vanishes with both @xmath271 and @xmath272 , and the functional derivative is taken at fixed _ hole _ density . expressing the resulting equations in the more convenient variables @xmath74 and charge density @xmath273 , we obtain @xmath274 -{\cal e}\varphi ( 1-\varphi ) { \bf \hat{y } } \label{2sp } \\ & & -\frac{\gamma } { 2}{\cal e}[(1-\varphi ) ^{2}-\psi ^{2}]{\bf \hat{y}}\ } \nonumber\end{aligned}\ ] ] these equations have to be supplemented by periodic boundary conditions and constraints on total mass and charge , @xmath275 and @xmath276 .
also , to be precise , all laplacians @xmath52 should be given the appropriate anisotropic interpretation .
using ( [ 2sp ] ) as our starting point , we seek to recapture the major features of the phase diagram .
first , the presence of two phases , uniform vs. spatially structured , is reflected in the existence of both homogeneous and inhomogeneous steady - state solutions of ( [ 2sp ] ) .
the existence of the former is evident , due to the conservation law .
anticipating homogeneity in the transverse directions , we seek the latter in the form @xmath277 , @xmath278 . integrating ( [ 2sp ] ) once , the stationary mass and charge currents appear as natural integration constants
. since the numbers of positive and negative charges are equal , the former must be zero , leaving us with the latter : @xmath70 .
the first equation allows us to eliminate @xmath279 in favor of @xmath74 , and , with @xmath280 as the new variable , we can reduce the second equation to potential form , @xmath281 .
the potential @xmath282 , given explicitly in @xcite , exhibits a minimum for a range of @xmath70 s , so that inhomogeneous , periodic solutions of ( [ 2sp ] ) exist .
we note in passing that these may even be found explicitly , provided @xmath256 @xcite .
otherwise , numerical integration is of course always possible , yielding rather impressive agreement with simulation profiles .
interestingly , our equations of motion predict that @xmath48 enters only through the _ scaling variable _ @xmath283 .
if plotted accordingly , our simulation data confirm this scaling by collapsing rather convincingly , at least within their error bars . however , the mere _ existence _ of two types of steady - state solutions is not sufficient to provide evidence for a phase transition .
therefore , we seek instabilities of these solutions , as the control parameters @xmath48 , @xmath232 and @xmath230 are varied . performing a linear stability analysis , we find that a _ homogeneous _ solution with mass @xmath232 becomes unstable if the drive @xmath48 exceeds a threshold value @xmath284}}$ ] .
the most relevant perturbation is associated with the smallest wave vector in the parallel direction , @xmath285 .
a simple analysis of @xmath286 shows that this instability can only occur within the interval @xmath287 , so that , in particular , the homogeneous phase is always stable for @xmath288 .
clearly , we can not identify this mean - field stability boundary with the true transition line : we have not considered the locus of instabilities associated with _
inhomogeneous _ solutions , and fluctuations have been neglected throughout .
however , it is quite remarkable how well it mirrors the qualitative shape of the phase diagram .
finally , let us consider the _ order _ of the transitions . in principle , two routes can be pursued here .
one of these , namely , the computation of the stability boundary of the _ inhomogeneous _ phase , @xmath289 , is rather subtle , involving three goldstone modes @xcite .
once @xmath290 is known , values @xmath291 for which @xmath292 and @xmath293 coincide can be identified as loci of continuous transitions ; otherwise , @xmath293 and @xmath292 mark the `` spinodals '' , near a first order transition , where the homogeneous / inhomogeneous phases become linearly unstable . alternatively , the adiabatic elimination of the fast modes results in an effective equation of motion for the slow ones which can then be analyzed .
since the technical details of this approach can be found in @xcite , we need only focus on the result which combines the expected with the surprising . letting @xmath294 denote the _ complex _ amplitude of the unique slow mode , associated with the band of wave vectors @xmath295 , its equation of motion takes the ginzburg - landau form : @xmath296 . here , @xmath59 is the soft eigenvalue , which vanishes on the stability boundary and @xmath297 is a rather complicated function . as in standard landau theory
, the sign of @xmath298 determines the order of the transition : if positive , the transition is continuous while negative @xmath298 signals a first order one . since @xmath299
, we can set @xmath300 and discuss @xmath298 on the stability boundary itself : being positive for @xmath248 , it has a unique zero at a critical @xmath301 , below which it becomes negative .
since @xmath246 increases with @xmath230 , we recover the qualitative behavior of the multicritical line observed in the simulations .
the _ surprising _ aspect of this equation of motion , however , resides in the fact that @xmath302 is o(2)-symmetric and @xmath303 spans just a single spatial dimension . given these symmetries
, the mermin - wagner theorem @xcite should forbid the existence of long - range order !
one might hope that a careful analysis of finite - size effects in this system would contribute to the disentanglement of this puzzle .
we conclude with two comments .
first , we noted earlier that the case @xmath304 is exactly soluble .
there are , in fact , two other surfaces in the phase diagram , corresponding to @xmath305 and @xmath306 , for which certain distributions are exactly known .
setting @xmath305 implies that the rates for particle - hole and particle - particle exchanges become equal , so that any given particle , e.g. , a positive charge , can not distinguish between negative charges and holes .
thus , it experiences biased diffusion , equivalent to the non - interacting kls model .
accordingly , the _ marginal _ steady - state distribution of the occupation numbers of _ one _ species is uniform , i.e. , @xmath307 = \sum_{\{n_{{\bf r}% } ^{\mp } \}}p\left [ \{n_{{\bf r}}^{+},n_{{\bf r}}^{-}\}\right ] \propto 1 $ ] , and observables pertaining to a single species are trivial . for example , the two - point correlation functions for equal charges , @xmath308 and @xmath309 , vanish for @xmath310 .
full _ * * distribution , @xmath311 $ ] , however , is nontrivial , so that , e.g. , the two - point function for _ opposite _ charges , @xmath312 , remains long - ranged @xcite .
a completely random system , marked by a uniform @xmath311 $ ] results only if @xmath306 .
finally , we note that the one - dimensional version of our model is exactly soluble by matrix methods @xcite .
second , it is natural to wonder about the consequences of having non - zero charge .
this problem has only been investigated for @xmath256 @xcite , but the findings are quite remarkable .
the system still orders into a charge - segregated strip , while supporting a _ nonvanishing _ mass current , reflected in an overall drift .
if , e.g. , positive charges outnumber negative ones , one might expect that the whole strip would drift in the field direction - in analogy with american football , where the team with fewer players tends to lose ground .
in contrast , the strip wanders _ against _ the field , following the preferred direction of the minority charge ! we should add that this model possesses several other intriguing properties , e.g. , stable configurations with nontrivial winding number ( `` barber poles '' ) @xcite or multiple - valued currents in the mean - field description @xcite . to summarize , the remarkable richness of this deceptively simple system is clearly amazing .
we have presented , within the limited scope of these lecture notes , a brief introduction to the statistical mechanics of driven diffusive systems and some recent developments in this ever expanding field .
focusing only on the prototype model and two of the simplest generalizations , we pointed out a multitude of surprises , when we base our expectations on the experience with equilibrium systems . while some of these , e.g. , the generic presence of singular correlations , are rather well understood , others , such as the nature of ordering in two - species models , remain unresolved .
of course , the holy grail of this whole field , namely , the fundamental understanding and theoretical classification of non - equilibrium steady states , still beckons at the distant horizon .
it is a pleasure to especially acknowledge our collaborators on the recent work reported here : c.c .
hill , g. korniss and k .-
others are too numerous to mention but no less deserving .
we thank the ixth international summer school on fundamental problems in statistical mechanics , and particularly h.k .
janssen and l. schfer , for their hospitality .
this research was supported in part by grants from nato , the sfb 237 of the deutsche forschungsgemeinschaft and the us national science foundation through the division of materials research .
amit , _ field theory , the renormalization group and critical phenomena _ ,
2nd revised edition , ( world scientific , singapore , 1984 ) ; j. zinn - justin , _ quantum field theory and critical phenomena _ , ( oxford university press , oxford , 1989 ) .
b. schmittmann , and r.k.p .
zia , phys .
lett . * 66 * , 357 ( 1991 ) ; b. schmittmann , europhys . lett . * 24 * , 109 ( 1993 ) ; e. praestgaard , h. larsen , and r.k.p .
zia , europhys .
lett . * 25 * , 447 ( 1994 ) and references therein .
leung , b. schmittmann and r.k.p .
zia , phys .
* 62 * , 1772 ( 1989 ) ; g. szab and a. szolnoki , phys . rev .
* b47 * , 8260 ( 1993 ) and references therein ; g. szab , a. szolnoki , t. antal , phys . rev .
* e49 * , 299 ( 1994 ) .
v. becker and h.k .
janssen , europhys .
* 19 * , 13 ( 1992 ) and private communication ; k.b .
lauritsen and h.c .
fogedby , phys . rev . *
e47 * , 1563 ( 1992 ) ; b. schmittmann and k.e .
bassler , phys .
* 77 * , 3581 ( 1996 ) ; b. schmittmann and c.a.laberge , europhys . lett .
* 37 * , 559 ( 1997 ) .
m. rubinstein , phys .
* 59 * ( 1987 ) 1946 ; t.a.j .
duke , phys .
* 62 * ( 1989 ) 2877 ; y. schnidman , in : _ mathematics in industrial problems iv _ , ed .
a. friedman ( springer , berlin 1991 ) ; b. widom , j.l .
viovy and a.d .
desfontaines , j. phys i ( france ) * 1 * ( 1991 ) 1759 .
g. korniss , b. schmittmann and r.k.p .
zia , europhys . lett .
* 32 * , 49 ( 1995 ) and j. stat .
phys . * 86 * , 721 ( 1997 ) .
the second reference discusses the adiabatic elimination procedure . for earlier work ( @xmath315 )
see b. schmittmann , k. hwang , and r.k.p .
zia , europhys . lett . * 19 * , 19 ( 1992 ) as well as refs@xcite . | nonequilibrium steady states in driven diffusive systems exhibit many features which are surprising or counterintuitive , given our experience with equilibrium systems .
we introduce the prototype model and review its unusual behavior in different temperature regimes , from both a simulational and analytic view point .
we then present some recent work , focusing on the phase diagrams of driven bi - layer systems and two - species lattice gases .
several unresolved puzzles are posed . |
theories beyond the standard model based on gauge groups g which contain a @xmath2 gauge symmetry ( b and l are respectively baryon and lepton numbers ) are very interesting for both particle physics and cosmology .
first of all , they predict the existence of right - handed neutrinos and the left - handed neutrinos can acquire very small masses according to the see - saw mechanism @xcite ; the neutrino oscillations discovered by the superkamiokande @xcite are predicted .
next , if the theory is supersymmetric , the lightest superparticle ( lsp ) can remain stable down to low energies and becomes a good cold dark matter ( cdm ) candidate .
this is due to the fact that @xmath3 contains a @xmath4 discrete symmetry which can be left unbroken down to low energies if higgs in _ safe _ representations of g are used to break g down to low energies .
this plays the role of r - parity @xcite and in such theories the lsp can be automatically stable .
the third main interesting point for cosmology is that inflation emerges _ naturally _ , i.e. no field nor any symmetry other than the ones used to build the theory itself are needed for inflation to arise ; and natural values of the parameters are obtain when constraints from the cosmic microwave background explorer ( cobe ) data are applied @xcite .
if the existence of a unified gauge group g is assumed , it must somehow be broken down to the standard model gauge group .
it can either break directly or via one or more intermediate gauge symmetry @xmath5 therefore , if in the very early universe symmetries between particles were described by a gauge group larger than @xmath6 , the universe must have undergone a series of phase transitions associated with spontaneous symmetry breakings .
these symmetry breakings may have lead to non - trivial topologies of the vacuum manifold and topological defects may have formed according to the kibble mechanism @xcite .
it is well known that all unified theories lead to the formation of topological monopoles which are topologically stable down to low energies and are in conflict with observations - this is true as soon as the @xmath7 gauge symmetry of the standard model is embedded in a non - abelian group which does not contain u(1 ) factor .
hence some mechanism has to be invoked to disperse these monopoles .
the standard scenario for solving the monopole problem is inflation .
inflation also solves many of the cosmological problems such as the horizon problem and predicts the formation of the large scale structure .
inflation is always confronted with fine - tuning problems unless supersymmetry is invoked .
the standard scenario for inflation in supersymmetric theories is the so - called f - term hybrid scenario @xcite and it can be implemented in theories which contain @xmath2 @xcite . in the simplest model of f - term inflation
topological defects form at the end of inflation , and hence an intermediate symmetry is needed to solve the monopole problem . in this case , the general symmetry breaking pattern will thus have to be of the form @xmath8 , with h chosen in such a way that monopoles form when g breaks down to h , then inflation takes place driven by some scalar field associated with the breaking of h , and at the end of inflation h spontaneously breaks down to @xmath9 and no unwanted defects must be formed .
it is not easy to find an adequate h. one possibility , is to chose h such that @xmath0 strings form when h breaks down to @xmath10 .
cosmic strings form according to the kibble mechanism during the phase transition associated with the spontaneous symmetry breaking of a gauge group h down to a subgroup k of h if the first homotopy group of the vacuum manifold @xmath11 is non trivial .
@xmath12 cosmic strings form when a gauge group h which contains @xmath13 breaks down to a subgroup k of h which does not contain @xmath2 if @xmath14 . the higgs field which forms the string is a higgs field in a complex representation of h which breaks h and local @xmath12 symmetry when acquiring a non - vanishing vacuum expectation value
; we call it @xmath15 . in supersymmetric theories ,
the superpotential has to be holomorphic and two higgs superfields @xmath16 and @xmath17 in complex conjugate representations are needed to break @xmath12 ; @xmath0 cosmic strings are then made of two higgs fields ( the scalar components of @xmath16 and @xmath17 ) which wind around the string in opposite directions .
the simplest model in which @xmath0 strings are formed is during the symmetry breaking @xmath18 .
as mentioned in the introduction , @xmath0 cosmic strings can form at the end of f - term hybrid inflation .
this is the case when the superpotential in the sector which breaks h and @xmath0 is given by : @xmath19 where @xmath20 is a superfield singlet under h , @xmath21 and @xmath22 are two constants which are taken to be positive and @xmath23 sets the @xmath0 breaking scale . in such models ,
the strings together with inflation generate the cosmological perturbations , and the strings contribution to the cmb is non - negligible but model dependent @xcite . both the strings and the inflation perturbations are proportional to @xmath24 and cobe data
gives the constraint @xmath25 gev .
the mass - per - unit length of the strings is @xmath26 .
the parameter @xmath22 is determined such as to solve the horizon problem and values of order @xmath27 can be found .
a realistic model based on so(10 ) has been built @xcite . numerical work based on a toy u(1 ) model has shown that a mixed scenario with inflation and strings soften the oscillations in the cmb power spectrum predicted by inflation alone and may well better fit the data @xcite .
if this is confirmed by new experimental data and by improved simulations - using a realistic model and taking into account the fact that the strings form at the end of inflation i.e. may have a specific distribution - the motivation for studying such models would be even stronger .
if a superpotential containing only non - renormalisable terms is used to break @xmath0 , such as @xmath28 where @xmath29 is constant taken to be positive and @xmath30 is a superheavy scale beyond which quantum gravity takes place , a period of thermal inflation takes place @xcite .
this can solve the monopole problem if the @xmath0 breaking scale is @xmath31 gev .
but it can not be at the origin of the large scale structure nor solve the horizon problem . at the end of this low energy inflationary period fat @xmath0 cosmic strings formed .
we would like to point out that it is not easy to build such a realistic model .
thermal inflation dilutes the monopoles previously formed but it also dilutes any baryon asymmetry which might have been previously generated ; and it is in general very difficult to produce the observed baryon asymmetry after thermal inflation .
however , here , baryogenesis via leptogenesis can take place at the end of inflation through the decay of the right - handed neutrinos released by decaying @xmath32 cosmic string loops @xcite .
this scenario is independent of the reheat temperature of the universe after the inflationary era .
one of the major problem faced by the standard cosmology is to explain the dark matter component of the universe . in supersymmetric theories with conserved r - parity ,
the lsp is stable and is a good cdm candidate . as mentioned in the introduction , in theories which predict @xmath0 cosmic strings
, the lsp can be automatically stable @xcite . when a network of strings is formed , infinite strings and loops are initially formed .
more loops are formed by the intercommuting of the long strings .
string loops rapidly decay by emitting elementary particles and gravitational radiation .
the main decay channel of @xmath0 cosmic strings loops is into right - handed neutrinos and sneutrinos .
if the lsp is higgsino , then these will directly decay into the lsp and if the lsp is bino as is usually the case in left - right models , subsequent decay will produce the lsp .
a scaling network of @xmath0 cosmic strings can thus produce lsps and these in non - negligible quantities @xcite .
this mechanism of lsp production , as is the leptogenesis mechanism , is non - thermal .
we find that the lsp number density today released by decaying cosmic string loops is given by : @xmath33 where @xmath34 denotes the branching ratio of the right - handed neutrinos into lsp , @xmath29 is the the higgs self coupling and @xmath35 is a numerical string parameter . @xmath36 and @xmath37 count the number of massless degrees of freedom at the critical temperature @xmath38 at which the strings form and at the lsp freeze - out temperature @xmath39 respectively .
@xmath40 is the planck mass . in fig.[fig : t ] we plot @xmath41 a function of the lsp mass @xmath42 .
the region above each curves corresponds to @xmath43 ( @xmath44 respectively ) , and the region below to @xmath45 ( @xmath46 respectively ) ; this region is excluded by observations .
@xmath0 cosmic strings can form in all unified theories based on gauge groups with rank greater than five .
they are very interesting cosmologically because they can explain the matter - antimatter asymmetry of the universe , they can form at the end of inflation and modify the cmb power spectrum , and they can produce cdm in non - negligible quantities .
the author would like to thank her collaborators r. brandenberger and x. zhang .
99 m. gell - mann , p. ramond and r. slansky in _ supergravity _ ,
p. van nieuwenhuizen and d. freeman ( north holland , amsterdam , 1979 ) p. 315 ; + t. yanagida , _ prog .
* b 135 * , 66 ( 1979 ) ; + r. mohapatra and g. senjanovi , . | @xmath0 cosmic strings form in a wide class of theories beyond the standard model which contain a @xmath1 gauge symmetry .
they can form at the end of hybrid inflation and explain , together with inflation , the cosmic microwave background anisotropies and the formation of large scale structure .
they can produce cold dark matter in the form of the lightest superparticle and they can be at the origin of the baryon asymmetry of our universe .
one major advantage of these mechanisms is that they are non - thermal .
* cosmology of b - l cosmic strings * _ ictp , strada costiera 11 , p.o .
box 586 , 34100 trieste , italy _ |
a detailed determination of the wavelength dependence of interstellar extinction , i.e. , the absorption and scattering of light by dust grains , is important for two very different reasons .
first , since it is a product of the optical properties of dust grains , extinction provides critical diagnostic information about interstellar grain populations ( including size distribution , grain structure , and composition ) , providing guidance for interstellar grain models .
second , the accuracy to which the intrinsic spectral energy distributions ( seds ) of most astronomical objects can be determined depends on how well the effects of extinction can be removed from observations . in both cases , a fundamental issue is how accurately the wavelength dependence of extinction can be measured .
consequently , it is essential to have a firm grasp of how measurement errors can affect the determination of this wavelength dependence .
this paper is the culmination of a series of `` techniques '' papers published over the past five years ( fitzpatrick & massa 1999 ; massa & fitzpatrick 2000 ; and fitzpatrick & massa 2005 ; hereafter fm99 , mf00 , and fm05 , respectively ) whose aim has been to develop a technique to _ simultaneously determine _ the wavelength dependence of extinction ( to higher accuracy than previously possible ) and the physical properties of a reddened star .
it represents a continuation of our earlier series on the properties of uv extinction ( fitzpatrick & massa 1986 ; 1988 ; and 1990 , hereafter fm90 ) , and provides a detailed description of a new technique for deriving interstellar extinction curves which does not rely on observations of standard stars , virtually eliminates the effects of `` mismatch '' error , and yields an accurate assessment of the uncertainties .
this `` extinction - without - standards '' technique opens the door to a new class of extinction studies , including regions heretofore inaccessible .
for example , errors in the traditional `` pair method '' approach to extinction strongly limit our ability to study extinction in two important regimes .
the first is low-@xmath0 sightlines , where one might hope to relate extinction properties to the environmental properties of specific physical regions .
the second is extinction derived from mid- to late b stars .
these stars are plentiful , and often constitute the bulk of the stars available for extinction measurements in intrinsically interesting regions , such as the pleiades .
we will demonstrate how the extinction - without - standards approach overcomes both of these problems and present examples of each .
some first results from this program were illustrated by fitzpatrick ( 2004 , hereafter f04 ) .
in addition to the extinction results , we will demonstrate that our approach simultaneously provides accurate stellar parameters which are also astrophysically interesting . in [ measure ] , we provide a broad overview of the basic problem of measuring an extinction curve , and compare the merits of curves derived by the pair method and curves derived by using stellar atmosphere models . in
[ secextinction ] , we describe our new model - based technique in detail , and list the basic ingredients needed to determine an extinction curve using this approach . in [ secdata ] , we describe the data used in the current study . in [ secresults ]
we provide a number of sample extinction curves derived using model atmospheres .
we also demonstrate the high precision of the new curves and the reliability of the error analysis employed . finally in [ secdiscuss ] we describe some to the scientific advantages of this new technique and our plans to exploit them .
to understand how interstellar extinction is measured and to appreciate how different measurement techniques can affect the outcome , we begin with the intrinsic elements of an uncalibrated observation of a spectral energy distribution ( sed ) of a reddened star obtained at the earth , @xmath1 . this can be expressed as @xmath2 where @xmath3 is the intrinsic surface flux of the star at wavelength @xmath4 , @xmath5 is the response function of the instrument , @xmath6 is the angular radius of the star ( where @xmath7 is the stellar distance and @xmath8 is the stellar radius ) , and @xmath9 is the absolute attenuation of the stellar flux by intervening dust ( i.e. , the total extinction ) at @xmath4 .
alternatively , the observed sed can be expressed in terms of magnitudes @xmath10 by @xmath11 where @xmath12 is a term which transforms between the observed magnitude system and absolute flux units , and @xmath13 is the traditional definition of the absolute magnitude of the star at @xmath4 . the difficulty in measuring the total extinction @xmath9 can be seen by rearranging these equations to solve for the extinction term
. equation [ flux1 ] yields @xmath14 while equation [ mag1b ] yields @xmath15 in either case , a true measurement of @xmath16 would require calibrated sed observations , knowledge of the intrinsic sed of the star , and measurements of both the stellar distance and radius , or their ratio @xmath17 .
unfortunately , there are no early - type stars for which both of these latter quantities are known to sufficient accuracy to allow a meaningful measurement of @xmath9 . as a result
, indirect methods must be employed , and @xmath16 is _ always a derived quantity _ , subject to assumptions . in
virtually all extinction studies , the actual measured quantity is a `` color excess '' which describes the extinction at a wavelength @xmath4 relative to that at a fiducial wavelength .
the traditional approach is to adopt the @xmath18 band as the fiducial , since @xmath18 magnitudes are widely available , accurately calibrated , and typically of high quality .
this color excess can be expressed as either @xmath19 as based on equation [ alam1 ] , or @xmath20 as based on equation [ alam2 ] .
thus , the determination of the color excess requires only the measurement of the observed sed and a knowledge of the shape of the intrinsic sed .
there are two basic approaches to determining color excesses , based on the use of either unreddened stars or stellar atmosphere models to represent the intrinsic seds of reddened stars .
these two techniques , and the issue of the normalization of extinction curves , are discussed in the three subsections to follow .
the first approach is the `` pair method . ''
a pair method curve is constructed by comparing the fluxes of a reddened star and an ( ideally ) identical unreddened `` standard star . '' essentially ,
the absolute magnitudes in eq . [ elam2 ] are replaced by the observed magnitudes of the standard star and the curve is usually expressed in the form @xmath21 where the color excesses @xmath22 are normalized by @xmath23 ( ) and the subscripted quantities refer to unreddened indices for the standard star .
( the issue of normalization will be discussed below . )
when the two stars are observed using the same instrument , this method has the advantage that calibration terms cancel , eliminating any dependence on the absolute flux calibration , @xmath5 or @xmath24 .
there are , however , two disadvantages to this technique .
the first is that the grid of unreddened standard stars is necessarily limited , so that some mismatch in the seds of the reddened and unreddened stars , termed `` mismatch error '' , is inevitable .
the second is that there are very few truly unreddened early - type stars , so usually the `` unreddened '' standard must be corrected for some small amount of extinction whose exact magnitude and wavelength dependence are uncertain .
this creates an error that can propagate into the resulting extinction curves .
massa , savage , & fitzpatrick ( 1983 ) presented a detailed study of the uncertainties affecting pair method extinction curves and showed that mismatch effects dominate the error budget . if one uses a single unreddened spectral standard for each spectral class , and assumes that all spectral classifications are perfect , then as a result of spectral binning , mismatch errors will be equal to or less than half a spectral class .
figure [ figmismatch ] shows how such mismatches can affect extinction curves derived from main sequence b stars . in each of the four groups of curves in figure [ figmismatch ] , the true shape of an extinction curve affecting a b1 , b2 , b5 , or b9 star
is indicated by the solid curve .
the extinction curves which would be derived via the pair method in the presence of a @xmath25 spectral class mismatch error are shown by the dash - dot curves for the case of @xmath26 and by the dashed curves for @xmath27 .
( the derived curves fall below the true curve in the uv and above the true curve in the ir when the standard star is cooler than the reddened star , and vice versa . ) in practice , spectral classifications are not perfect and the available unreddened standard stars do not necessarily lie in the middle of the range of properties within a single spectral class . both these effects exacerbate the spectral mismatch problem and thus the uncertainties shown in figure [ figmismatch ] are likely closer to typical errors , rather than extremes .
although it may appear from figure [ figmismatch ] that a discontinuity at the balmer jump at @xmath28 would provide an obvious indicator of the presence of spectral mismatch in an extinction curve , this is almost never practical .
the spectrophotometric data available for constructing extinction curves are generally limited to uv wavelengths ( @xmath29 ) , which are highlighted in figure [ figmismatch ] .
typically , the only data available in the optical and near - uv are photometric indices which straddle the balmer jump , such as the johnson @xmath30 color , and these can not be used to distinguish the effects of mismatch from intrinsic curve shape .
mismatch error clearly can have a profound effect on the shapes of curves derived from stars with low color excesses , particularly in the mid - to - late b spectral range and most particularly in the uv spectral region .
in fact , it is mismatch error which provides the low temperature cutoff to the spectral range of stars from which useful uv extinction measurements can be made .
mismatch also severely limits extinction studies based on stars hotter and/or more luminous than the main sequence b stars . for the o stars , very few unreddened standard stars exist and extreme mismatching of spectral types is often necessary to derive a curve
although , since the intrinsic uv / optical seds of the o stars are not well known , it is not clear how large an effect this introduces in the derived curves .
the use of luminous , evolved stars for extinction studies is particularly problematic since unreddened standards are rare and the sensitivity of the intrinsic seds to both temperature and surface gravity can lead to very severe mismatch effects in the resultant curves ( although see the discussion in cardelli , sembach , & mathis 1992 for results in the early - b spectral range ) .
the second approach to deriving extinction curves is to model the intrinsic sed of the reddened star in order to isolate the effects of extinction .
this technique was first used by whiteoak ( 1966 ) to analyze optical spectrophotometry and has been applied in one form or another many times since .
we refer to it as `` extinction - without - standards '' , since it does not rely upon a set of unreddened standard stars to determine the extinction curve . in this method ,
a model atmosphere of the reddened star is determined from its photometric or spectral properties .
the advantage of this approach is that , in principle , a perfect , unreddened match can be determined for the intrinsic sed of the reddened star , eliminating the mismatch error which plagues pair method curves .
the apparent disadvantages of the approach are not actual disadvantages , but rather requirements , which can limit its accuracy and range of applicability .
the first requirement is that a set of models must exist whose accuracy can be quantified and validated .
the second is that the observations must contain adequate `` reddening free '' information to accurately determine the intrinsic sed of a reddened star .
the third is that the fluxes must be precisely calibrated , i.e. , @xmath5 must well determined .
we have been developing the necessary constituents of this method over the past five years .
we began by demonstrating that the kurucz ( 1991 ) _
atlas9 _ models provide faithful representations of the observed uv and optical seds of near main sequence b stars ( fm99 ) .
we then showed that a combination of an observed uv sed and optical photometry provides adequate reddening independent information to determine _ both _ the appropriate _ atlas9 _
model for a reddened star _ and _ a set of parameters ( defined by fm90 ) which determine the shape of its interstellar extinction curve .
subsequently , we refined the calibration of the _ iue _ data , in order to improve the quality of the fits and the robustness of the physical information derived from them ( mf00 ) .
next , we verified the physical parameters derived from the models through applications of the models to eclipsing binary data , where the results must agree with other constraints ( see , fitzpatrick et al .
2003 and references therein ) .
finally , we have used _ hipparcos _ data of unreddened b stars to derive a consistent recalibration of optical and nir photometry ( fm05 ) and , in the process , once again demonstrated the internal consistency of the models for near main sequence b stars . as a result of these efforts , we now have internally consistent @xmath5 for _ iue _ and optical and nir photometry , and are in a position to apply the results and to quantify the associated errors .
these are the objectives of the current paper , and are discussed fully in the next section .
once color excesses have been determined , we are faced with the problem of how to compare excesses derived for lines of sight with different amounts of extinction .
after all , we are interested in the `` shapes '' of the curves , since these may reveal important clues about the size distribution and composition of the dust .
this desire to compare shapes , brings us to the normalization problem . from a purely mathematical point of view
, a straightforward approach would be to normalize the curves by their norm , @xmath31 with appropriate weighting , this normalization could , on average , minimize the observational error in the normalization factor and reduce systematic effects .
however , such a normalization would be sensitive to the strength of narrow features , such as the 2175 bump , and this could mask the overall agreement between curves over the majority of the wavelength range . a second approach is to search for some immutable feature of the curve and normalize by that .
the idea behind this procedure is that if some aspect of all curves is always the same , then all curves can be normalized by the strength of this property and all of the resulting curves will be directly comparable . this is the motivation for the @xmath32 normalization adopted by cardelli , clayton , & mathis ( 1989 ) .
they assume that all extinction curves have a very similar ( although not identical ) form for @xmath33 . however , there are problems with this normalization as well . in particular , as pointed out above , @xmath32 is not a directly measured quantity , but must be derived from ir photometry and requires assumptions about the shape of extinction curves at very long wavelengths .
the shape of this extinction is often considered to have a universal form , but this has been demonstrated for only a relatively small number of sightlines ( reike & lebofsky 1985 ; martin & whittet 1990 ) .
in addition , measurements of extinction in the ir can be compromised because the stellar seds become increasingly faint at long wavelengths and other sources of light , such as circumstellar emission or scattering from dust in the near stellar environment , can contaminate the sed measurements .
furthermore , ir color excesses are usually small for stars that are detectable in the uv , so uv curves normalized by quantities derived from ir photometry may be affected by large normalization errors .
consequently , the absolute level of such curves can be poorly defined .
finally , even with the advent of the 2mass data base , there are still many stars which do not have ir photometry available . as a result of the complications mentioned above , and since our intent is to demonstrate how precisely curves can be measured while making a minimum number of assumptions , we have opted to use the conventional @xmath0 normalization , as shown in eq .
[ klambda ] .
while the interpretation of @xmath23 ( ) as a measure of the `` amount '' of interstellar dust is not straightforward , its widespread availability , observational precision , and lack of requisite assumptions make it the best choice for this study .
nevertheless , we note that it is a simple matter to transform from one normalization to another , and emphasize that @xmath22 is actually the basic measurable quantity .
our earlier studies ( fm99 , fm05 ) have shown that the observed seds , @xmath34 , of lightly- or unreddened main sequence b stars can be modeled very successfully by using a modified form of eq .
[ flux1 ] , namely , @xmath35 } \;\;.
\label{flux2}\ ] ] the use of absolutely calibrated datasets eliminates the calibration term @xmath5 and the total extinction @xmath16 has been broken down into a normalized shape term ( @xmath36 ) , a normalized zeropoint ( @xmath37 ) , and a scale factor ( @xmath23 ( ) ) . providing that the righthand side of the equation can be represented in a parameterized form
, the equation can be treated as a non - linear least squares problem and the optimal values of the parameters which provide the best fit to the observed sed @xmath34 can be derived along with error estimates . because the stars under study were lightly reddened
, we could replace the extinction curve @xmath36 and the offset term @xmath38 with average galactic values without loss of accuracy . using the kurucz _
atlas9 _ stellar atmosphere models to represent @xmath39 , the results of the fitting procedure were estimates of 6 parameters : @xmath23 ( ) , @xmath17 , and the four parameters that define the best - fitting model , i.e. , @xmath40 , @xmath41 , the metallicity [ m / h ] , and the microturbulence velocity @xmath42 .
we performed the fits using the mpfit procedure developed by craig markwardt the fitting process described above begins to break down when the color excess @xmath23 ( ) of the target stars exceed @xmath430.05 mag . by this
we mean that large systematic residuals begin to appear , which greatly exceed the measurement errors .
the reason is simple : the wavelength dependence of interstellar extinction curves varies greatly from sightline - to - sightline , and once @xmath44 mag the differences between the true shapes of the curves and the assumed mean form begin to exceed the measurement error .
however , fm99 noted that the seds of significantly reddened stars could still be modeled using equation ( [ flux2 ] ) ( and the non - linear least squares approach ) if the wavelength dependence of the extinction curve could be represented in a flexible form whose shape could be adjusted parametrically to achieve a best fit to the observations , and if these parameters were determined _ simultaneously _ with the stellar parameters .
this is the essence of our `` extinction - without - standards '' approach . successfully modeling the shape of reddened stellar seds requires four principal ingredients : 1 ) an observed sed that spans as large a wavelength range as possible , 2 ) an accurate absolute flux calibration ( @xmath5 or @xmath24 ) , 3 ) an extinction curve whose shape can be described by a manageable set of parameters , and 4 ) a grid of stellar surface fluxes , @xmath3 , whose defining parameters can be determined from the observational data . in [ secdata ] we will describe the particular datasets used in this paper to demonstrate our technique .
we have already discussed how mf00 and fm05 have determined the necessary calibrations . in the remainder of this section , we describe our flexible form for the interstellar extinction curve ( [ secflexi ] ) and the grid of stellar surface fluxes with which we have developed our approach ( [ secmodels ] ) .
we adopt a flexible and adjustable form for the uv - through - ir extinction curve , whose shape can be optimized to fit the sed of a reddened star through the adjustment of a specific set of parameters in the least - squares minimization procedure .
this curve is illustrated in figure [ figflexi ] .
it consists of two main regions : 1 ) the uv ( @xmath45 ; solid curve ) where the parameterized form of fm90 is adopted and 2 ) the near - uv / optical / ir ( @xmath46 ; dashed line ) where we use a cubic spline interpolation through a set of uv ( u@xmath47 , u@xmath48 ) , optical ( o@xmath47 , o@xmath48 , o@xmath49 , o@xmath50 ) , and ir ( i@xmath47 , i@xmath48 , i@xmath49 , i@xmath50 ) anchor points to represent the curve . the interpolation is performed using the _ interactive data language ( idl ) _ procedure spline .
we adopt a spline representation for the near - uv / optical / ir curve simply because we do not have reliable , detailed information on the wavelength dependence of the extinction in the near - ir through near - uv region ( 1@xmath51 m 3000 ) .
it is ironic that the portion of the curve that is accessible from the ground is more poorly characterized than the portion accessible only from space . as a result , we do not know whether the optical to near - ir region of the curve can be represented by a compact analytical formula .
our hope is that , by applying our procedure to a large sample of sightlines , we will ultimately be able to characterize the shape of the extinction law in this region by simple relations and determine whether sightline - to - sightline variations are correlated with other aspects of the curve or with interstellar environment .
the placement of the spline points resulted from considerable experimentation , but certainly can not be represented as an objectively determined optimal result .
the current arrangement does , however , allow us to model the major available datasets to a level consistent with the observational errors .
the fm90 parameterization scheme contains 6 free parameters to represent 3 functionally separate features that are summed to produce the uv curve .
an underlying linear component , indicated by the dotted line in figure [ figflexi ] , is specified by an intercept @xmath52 and a slope @xmath53 .
the lorentzian - like 2175 bump is fit by a drude profile @xmath54 , where @xmath55 and @xmath56 specify the position and fwhm of the bump , respectively , whose strength is determined by a scale factor @xmath57 .
finally , the degree of departure of the curve in the far - uv from the underlying linear component is specified by a single parameter @xmath58 .
defining @xmath59 , the complete uv function is given by : @xmath60 where @xmath61 and @xmath62 it is worth emphasizing that the fm90 parameterization is a mathematical scheme only , which allows us to reproduce uv extinction curves in a shorthand form ( and makes the current program possible ) .
however , it is not to be assumed that the functional components of the scheme represent actual separate extinction components arising in distinct dust grain populations .
this parameterization has proven very useful and to the best of our knowledge is able to reproduce all known uv extinction curves to the level of observational error
. its flexibility will be illustrated in
[ secresults ] below .
note that we terminate the fm90 formula at 2700 , although we originally utilized uv data extending to 3000 .
additional experience with the uv data has suggested that real extinction curves begin to exhibit departures from the fm90 formula near 2700 , due to the unrealistic extrapolation of the linear component into the blue - violet region .
the 10 spline anchor points which characterize the near - uv / optical / ir portion of the curve are determined by a least squares fit to the _ iue _ data longward of 2700 and the available optical and ir photometry . although there are 10 anchor points , there are also 6 constraints , so the fit actually introduces only 4 additional degrees of freedom .
we now discuss these constraints in detail .
the two uv anchor points , u@xmath47 and u@xmath48 at 2700 and 2600 , respectively , are fixed at the values of the fm90 uv fitting function at their wavelengths and are not adjustable .
together with o@xmath47 at 3300 , these points guarantee that the curve which passes through the _ iue _ data between 2700 and 3000 will join both the uv and optical portions of the curve smoothly . the four optical anchor points , o@xmath47 , o@xmath48 , o@xmath49 , and o@xmath50 at 3300 , 4000 , 5530 , and 7000 , respectively , are fit under two constraints : that the interpolated curve produces a value of @xmath63 in the v band , and that the curve be normalized to unity in @xmath23 ( ) .
thus , only two free parameters emerge from this region .
the four ir points , i@xmath47 , i@xmath48 , i@xmath49 , and i@xmath50 are located at 0.25 , 0.50 , 0.75 , and 1.0 @xmath51m@xmath64 , respectively .
these four points are constrained to satisfy the formula @xmath65 where the scale factor , @xmath66 , and the intercept , @xmath38 , are the only free parameters .
this is the power - law form usually attributed to ir extinction , with a value for its exponent from martin & whittet ( 1990 ) .
the exponent of the power - law could , potentially , be included as a free parameter in the fitting procedure , and we will investigate this in the future . however , our impression is that the ir data available to us ( primarily _ 2mass _ @xmath67 photometry ) are insufficient to determine this quantity accurately .
all results presented in this paper assume an exponent of -1.84 in equation [ eqnir ] . to represent the intrinsic surface fluxes , @xmath39 , of reddened stars we utilize r.l .
kurucz s line - blanketed , hydrostatic , lte , plane - parallel _ atlas9 _ models , computed in units of erg @xmath68 sec@xmath64 @xmath64 and the synthetic photometry derived from the models by fitzpatrick & massa ( 2005 ) .
these models are functions of four parameters : @xmath40 , @xmath41 , [ m / h ] , and @xmath42 .
all of these parameters can be determined in the fitting process although , because of data quality , it is sometimes necessary to constrain one or more to a reasonable value and solve for the others .
the general technique of deriving extinction curves via stellar atmospheres is , of course , not dependent on the specific set of models used . in the present case , the most important consideration for our adoption of the _
atlas9 _ models is that they work at least within a specific spectral domain . fm99 and
fm05 have shown that these models provide excellent fits to the observed seds for lightly- or un - reddened main sequence b stars throughout the uv through near - ir spectral region .
in addition , experience with eclipsing binaries ( see fitzpatrick et al .
2003 and references within ) has shown that the good sed fits also yield accurate estimates of the physical properties of the stars .
because of the physical ingredients of the models ( specifically lte and plane - parallelism ) we currently restrict our attention to the main sequence b stars .
we plan to investigate how well these models reproduce the seds and properties of somewhat more luminous b stars and also the later o types . also , we will take advantage of more complex models ( e.g. , the non - lte tlusty models ) as the available grids expand their parameter ranges .
to summarize , we model the observed seds of reddened near main sequence b stars by treating equation ( [ flux2 ] ) as a non - linear least squares minimization problem . as a result
we can simultaneously obtain estimates of the physical properties of a reddened star _ and _ the shape of the interstellar extinction curve distorting the star s sed .
a total of 16 parameters specify the righthand side of the equation , including @xmath17 , @xmath0 , four to define @xmath3 , and ten to define the shape of the extinction curve . depending on limitations of the available data , and known properties of the stars or interstellar medium , any subset of the parameters
can be constrained to predetermined values .
in the following section , we will illustrate the potential of our extinction - without - standards technique , utilizing a set of 27 lightly - to - heavily reddened stars .
for this demonstration , and indeed for extinction determinations in general , the ideal sed dataset would consist of absolutely - calibrated spectrophotometry spanning the uv - through - ir spectral regions . such data would allow a straightforward comparison between observations and stellar atmosphere models ( since both are presented in simple flux units ) and would provide the most detailed view of the wavelength dependence of interstellar extinction . while a small amount of such data is available ( see , e.g. , fitzpatrick at al .
2003 ) , the largest existing database of absolutely - calibrated spectrophotometry is the low - resolution archive of the _ international ultraviolet explorer _ satellite ( _ iue _ ) which covers the uv region only ( 11503000 ) . in the optical and near - ir ,
the largest sed databases are photometric in nature , consisting of johnson , strmgren , and geneva photometry in the optical and _ 2mass jhk _ photometry in the near ir . utilizing these resources
, we can examine the uv region for small scale features but can only study the broad scale wavelength - dependence of extinction in the optical and near - ir regions .
we use newsips _ iue _ data ( nichols & linsky 1996 ) obtained from the mast archive at stsci .
these data were corrected for residual systematic errors and placed onto the _ hst_/fos flux scale of bohlin ( 1996 ) using the corrections and algorithms described by mf00 . this step is absolutely essential for our program since our `` comparison stars '' are stellar atmosphere models and systematic errors in the absolute calibration of the data do not cancel out as in the case of the pair method .
( the newsips database is also contaminated by thermally- and temporally - dependent errors , which would not generally cancel out in the pair method see mf00 . )
multiple spectra from each of _ iue _ _ s _ wavelength ranges ( swp or lwr and lwp ) were combined using the newsips error arrays as weights .
small aperture data were scaled to the large aperture data and both trailed and point source data were included .
short and long wavelength data were joined at 1978 to form a complete spectrum covering the wavelength range @xmath69 .
data longward of 3000 were ignored because they are typically of low quality and subject to residual systematic effects .
the _ iue _ data were resampled to match the wavelength binning of the _ atlas9 _ model atmosphere calculations in the wavelength regions of interest .
mean values of the johnson _ ubv _ , strmgren _
uvby_@xmath70 , and geneva _
ubb@xmath47b@xmath48vv@xmath47 g _ photometric magnitudes , colors , and indices for the program stars were acquired from the mermilliod et al .
( 1997 ) archive .
_ 2mass _ _ jhk _ magnitudes for all stars , along with their associated errors , were obtained from the _ 2mass _ all - sky point source catalog at the nasa / ipac infrared science archive .
johnson @xmath71 , @xmath72 , and _ jhk _ data are also available for some of the stars , and were obtained from the mermilliod et al . archive .
in this section , we demonstrate the potential of our extinction - without - standards technique , utilizing a set of the 27 reddened stars , listed in table 1 , which fall into three groups : 1 ) stars in the open cluster ic 4665 ; 2 ) stars with moderate - to - heavy reddening ; and 3 ) lightly reddened stars in a specific region of the sky .
these representative examples illustrate the advantages of our approach and highlight several scientific applications which will be pursed in future studies , using expanded samples of stars . in addition
, they provide confirmation of the error analysis incorporated in our approach . for this demonstration sample ,
the sed fitting procedure was applied as described above , with the following additional details : * the sed data modeled in the fitting procedure include _ iue _ uv spectrophotometry , the johnson @xmath18 , @xmath73 , and @xmath30 indices , the strmgren @xmath74 , @xmath75 , @xmath52 , and @xmath70 indices , the geneva @xmath30 , @xmath76 , @xmath77 , @xmath78 , @xmath79 , and @xmath80 indices , and the _ 2mass _ _ jhk _ magnitudes .
johnson @xmath71 , @xmath72 , and _ jhk _ data are also available for a few of the stars . *
the optical extinction spline point o@xmath50 at 7000 is only well - determined when optical @xmath8 and @xmath81 band photometry are available .
therefore in the examples below , we only solve for o@xmath50 in such cases .
for the other stars , the optical portion of the extinction curve is determined only by the spline points o@xmath47 , o@xmath48 , and o@xmath49 .
* during the @xmath82 minimization , a reddened and distance - attenuated model was created from the current set of input parameters , and then synthetic photometry was performed on this model to produce the photometric indices , which were then compared with observations .
the synthetic photometry was calibrated as described by fm05 , with the calibration extended to redder colors by us .
the uv model fluxes and recalibrated _ iue _ fluxes , both in units of erg @xmath68 sec@xmath64 @xmath64 , were compared directly . *
the initial weighting factors for the various datasets in the @xmath82 minimization were determined from their observational uncertainties , i.e. , @xmath83 .
we then scaled the weights of the optical / near - ir photometry so that their total weight was equal to that of the _ iue _ uv spectrophotometry , thus balancing the fit between the two datasets .
this is the procedure adopted by fm05 , except that we include the strmgren @xmath70 index along with the rest of the optical / near - ir photometry , rather than assigning it its own ( high ) weight .
fm05 weighted @xmath70 heavily in recognition of its value as a surface gravity indicator .
however , we have found that the temperature - sensitivity of @xmath70 ( particularly in the later b stars ) combined with observational errors , can lead to very unsatisfactory fits when @xmath70 is over - emphasized .
treating @xmath70 in the same manner as the rest of the photometric indices seems to be the simplest and most reasonable approach . *
because interstellar h i lyman @xmath84 absorption in reddened stars can have a significant impact on the star s apparent continuum level far from line center at 1215 , we convolve the profile of this heavily damped line with the model atmosphere seds before comparing with observations . along heavily reddened sightlines , where the h i column density @xmath85h i@xmath86 is high and the signature of the atomic absorption strong , the value of @xmath85h i@xmath86 can actually be incorporated into the fitting procedure as a free parameter to be optimized .
we will utilize this capability in future studies . for the present ,
mostly lightly - reddened , sample , however , the @xmath85h i@xmath86 values used to construct the line profiles were taken from the survey of diplas & savage ( 1994 ) or else computed from the general relation @xmath85h i@xmath87 from bohlin , savage , & drake ( 1978 ) .
the inclusion of the lyman @xmath84 line insures that we distinguish the effects of dust extinction from atomic absorption in the far uv region . *
the uncertainties in the best - fit parameters were determined by running 50 monte carlo simulations for each star , during which the input data were randomly varied assuming a gaussian distribution of observational uncertainties and a new fit performed .
the zero points and random photometric uncertainties of the short - wavelength and long - wavelength _ iue _ fluxes were varied as described in fm04 ; the assumed observational errors in the johnson , strmgren , and geneva indices were as given in table 7 of fm04 ; and the uncertainties in the _ 2mass _ data were as obtained from the _
2mass _ archive .
in addition , the @xmath18 magnitude was assumed to have a 1-@xmath88 uncertainty of 0.015 mag .
the adopted 1-@xmath88 uncertainties for each parameter were taken as the standard deviation of the values produced by the 50 simulation .
we begin our examples by examining the extinction towards an open cluster . while multiple scientific rewards can result from the study of extinction towards cluster stars ( see the discussion in [ secdiscuss ] ) , our primary interest in ic 4665 is to demonstrate the `` technical '' advantages of our approach .
namely , the use of cluster extinction curves to help evaluate the magnitude of the uncertainties in the measurement of extinction curves , as discussed in detail by massa & fitzpatrick ( 1986 ) . in particular , and because of its low @xmath0 and late - b stellar population , ic
4665 extinction curves provide an especially sensitive test of the precision and range of our extinction - without - standards approach .
the wavelength dependence of extinction towards ic 4655 was first examined by hackwell , hecht , & tapia ( 1991 ; hereafter hht ) for the purpose of studying the relationship between extinction and ir emission , as measured by the _ infrared astronomical satellite _ ( _ iras _ ) .
this remains one of the most challenging extinction studies yet performed for two reasons : 1 ) the mean reddening in the cluster is very low , @xmath89 mag , and 2 ) the spectral types of the available targets run from mid- to late b. both facts exacerbate errors in the standard pair method approach , as has been shown in figure [ figmismatch ] .
hht recognized these uncertainties and ultimately concluded that the wavelength dependence of extinction among the cluster stars is uniform to within their ability to measure it .
figure [ figsed1 ] shows the results of the sed fits for the nine ic 4665 stars considered here .
the sed s of the best - fitting , reddened models are shown by the histogram - style curves . in the uv ,
the binned _ iue _ fluxes are shown by the small circles . in the optical region ,
johnson _ ubvri _ magnitudes ( converted to flux ) are indicated by circles , strmgren _
_ magnitudes by triangles , and geneva _
_ magnitudes by diamonds . in the near - ir , _
2mass _ and johnson _ jhk _ magnitudes are shown by the large filled and open circles , respectively .
note that the photometric data have been converted to flux units for display purposes only .
the comparison between models and observations was performed in the native photometric format ( i.e. , in magnitudes or colors as noted above ) .
figure [ figext1 ] shows our extinction - without - standards curves for the ic 4665 stars .
the symbols show the actual normalized ratios between the models and the stellar seds , while the thick solid curves show the flexible uv - through - ir extinction curves whose shapes were determined by the fitting procedure .
the curves have been arbitrarily shifted vertically for clarity , but all are shown compared with a similarly - shifted estimate of the average galactic extinction curve for reference ( thick dash - dot curves corresponding to @xmath90 , from fitzpatrick 1999 ) .
as will be discussed further below , we assumed a value of @xmath91 towards the cluster and did not include the ir _ jhk _ data in the fitting procedure .
thus , only the average galactic curve is shown for wavelengths longward of 6000 .
the various parameters determined by the fits are listed in tables 1 and 2 .
the flexible extinction curves themselves , in the form @xmath92 can be reconstructed from the parameters given in table 2 .
the 1-@xmath88 uncertainties of the extinction curves are indicated in figure [ figext1 ] by the grey shaded regions .
the regions are centered on the means of the 50 monte carlo simulations with which we performed our error analysis and their thickness shows the standard deviation of the individual simulations .
figure [ figext1comp ] compares our new curves in the uv with those derived by hht using the standard pair method .
hht s curves were reconstructed from the data in their table 5 .
note that the hht study originally included 17 stars .
we have eliminated 4 a - type stars , 3 chemically peculiar b - type stars , and 1 b - type shell system from consideration here since their extinction curves are particularly uncertain .
thus , figures [ figext1 ] and [ figext1comp ] show only the best - determined curves in the cluster , from the point of view of both the pair method and our technique .
the remarkable aspect of figure [ figext1comp ] is not the scatter among hht s curves it is exactly what should be expected given the limitations of the pair method but rather the impressive improvement in precision gained by the extinction - without - standards technique , as evidenced by the decrease in the curve - to - curve scatter .
clearly , this higher level of precision allows much firmer conclusions to be drawn about the degree of intrinsic variation of extinction across the face of the cluster , and also affords an improved potential to detect correlated behavior between the extinction properties and other aspects of the interstellar environment .
while we will present a full scientific analysis of the ic 4665 results in a future paper , the issue of the intrinsic variation of extinction among the cluster stars is important to consider here , since it can shed light on the accuracy of our error analysis .
the top curve in figure [ figext1 ] shows the mean of the nine individual ic 4665 curves .
the error bars show the actual sample standard deviation and the average galactic curve is again presented for comparison .
the critical result is that , over most of its wavelength range , the standard deviation of the sample about the mean curve is comparable to the predicted uncertainties in the individual curves , as shown by the shaded regions .
this indicates that the ( small ) curve - to - curve variations seen are at a level consistent with our expected uncertainties and that the intrinsic level of variation among the cluster stars must be very low .
another way to approach the issue of variability is to look at the scatter among the various parameters which define the extinction curves .
these are shown in table 3 , where columns 2 and 3 list the weighted mean values of the parameters and their observed standard deviations , respectively .
the predicted uncertainties , i.e. , the rms of the monto carlo - based errors for the individual stars , are listed in column 4 .
if our error analysis is reasonable , then the observed scatter should be the quadratic sum of the expected errors plus any intrinsic variability .
the value of examining cluster extinction curves is that one might reasonably suppose ( at least as a starting point ) that the individual curves , derived for nearly coincident lines of sight , are actually independent measurements of a single `` cluster curve , '' i.e. , no intrinsic scatter . in such a case ,
the measured scatter actually reflects the real measurement errors and provides an important test of the error analysis .
examination of table 3 shows that the predicted and observed scatters are indeed generally very similar , supporting the position that any intrinsic variations among the ic 4665 curves are close to the level of our ability to measure them _ and _ that we have accurately assessed the uncertainties in our results .
the final column of table 3 shows the implied values of the possible small intrinsic variations .
several of the individual parameters merit some additional comment . both the 2175 bump fwhm ( @xmath56 ) and its strength ( @xmath57 or @xmath93 ) show evidence for some small level of variability within the sample , above our expected measurement errors . however , this is not clearcut because these measurements involve the region of the _ iue _ spectra which typically has the lowest quality data and it is possible that weak systematic effects in the data themselves which are not accounted for in the error analysis could produce the small level of variability seen .
one such systematic is a `` reciprocity effect '' in long wavelength data which is not corrected by the mf00 algorithms ( see figures 12 and 13 of mf00 ) .
we conclude conservatively that there is marginal evidence for bump variations within the ic 4665 sample , but will ultimately rely on studies of several open clusters to determine whether our error analysis faithfully predict the real measurement errors in the bump region . the case of the far - uv curvature ( parameter @xmath58 ) ,
for which the implied intrinsic scatter is three times greater than our measurement errors , is more interesting .
it is clear from figures [ figext1 ] and [ figext1comp ] and table 3 that most of this apparent variability arises from two sightlines , namely , those towards hd 161165 and hd 161184 .
in fact , the observed scatter in @xmath58 towards the other seven sightlines is essentially identical to the expected measurement error .
because hd 161165 and hd 161184 are the two coolest stars in the sample , and the only stars with effective temperatures less than 12,000 k , we suspected that the high @xmath58 values in their extinction curves might be artifacts of the analysis , resulting from a failure of the models to accurately portray the far - uv seds of these stars .
the investigation presented below suggests that this could well be the case , with the extinction curve anomalies possibly arising from a difference between the chemical composition profiles of the stars and the atmosphere models .
the data in table 1 show a very uniform `` metallicity '' [ m / h ] for the cluster , with a weighted mean of -0.50 and a standard deviation of 0.04 .
this is close to , and actually slightly smaller than , our estimate of the measurement errors , simultaneously confirming the accuracy of our error analysis and imposing a small upper limit on the intrinsic compositional variability within the cluster .
although our [ m / h ] values are simple scale factors which apply to a template set of _
atlas9 _ solar abundances , fm99 showed that the [ m / h ] derived in our analysis is most sensitive to the abundance of fe due to a very strong opacity signature in the mid - uv and is most analogous to [ fe / h ] . since the fe abundance in the _ atlas9 _ models is @xmath430.2 dex larger than the currently accepted solar value of 7.45 ( where h = 12.00 ; asplund , grevesse , & sauval 2005 ) , our results suggest that the ic 4665 stars are deficient by about a factor of two in fe as compared to the sun .
the small scatter observed in [ m / h ] is both satisfying and expected , since the surface composition of fe is not subject to evolutionary modification in these young stars , which presumably all formed from the same parent material .
we experimented with forcing [ m / h ] to a more solar - like level for the ic 4665 stars and immediately found two effects : 1 ) the @xmath82 values all increased significantly because the fe features in the mid - uv could not be fit as well , and 2 ) most of the extinction curves remained unchanged , but the anomaly in the far - uv region for hd 161165 and hd 161184 decreased dramatically .
the first result was expected .
the second was a surprise , but is understandable .
if the cluster stars ( or at minimum , the two coolest stars hd 161165 and hd 161184 ) have a non - solar ratio of fe to the light metals , e.g. , [ c / fe ] @xmath94 0 , then our best fit models biased towards the fe abundance would underestimate the opacity due to the light metals . for the cooler stars , such elements provide significant opacity in the far - uv region and
the _ atlas9 _ models would not be able to account for such a specific opacity difference .
the fitting procedure would respond by finding a higher far - uv extinction curve .
the curves for the hotter stars are less affected by changing [ m / h ] since the light metal opacity is less significant .
we tentatively conclude that the chemical composition of the b stars in ic 4665 may deviate strongly from that of the sun , with subsolar fe but a more `` normal '' level of the light metals such as c. this suggestion is easily tested with a fine analysis of high resolution stellar spectra and we will pursue this in the future . on the positive side , this result suggests that the uv continua of late b and early a stars might be used to determine _ both _ a scaled [ m / h ] _ and _ a mean [ cno / fe ] index , assuming a grid of models parameterized by both these composition indices is available . on the more sober side ,
it is a reminder that our technique is susceptible to stellar abnormalities . as with the pair method
, all available data should be consulted to determine whether a particular star is suitable for deriving an extinction curve . our final comment on the ic 4665 results concerns the behavior of the ir data . in figure [ figext1 ]
we show two sets of _ jhk _ photometry for each star and for the average cluster curve .
the solid symbols indicate _
2mass _ measurements and the open symbols show hht s data .
the two sets of data systematically differ , with the _ 2mass _ results suggesting a value of @xmath38 somewhat smaller then the average galactic value of 3.1 and the hht data suggesting a slightly larger value ( hht derive a mean of @xmath95 ) .
this latter result is more consistent with expectations , given that the mean far - uv curve is lower than the galactic average ( see , e.g. , cardelli et al .
1989 ) , but this is insufficient evidence for rejecting the _ 2mass _ data . for now
, our solution for this quandary has been to ignore both sets of data in fitting the ic 4665 seds and adopt a default value of @xmath91 . however , the discrepancy for measurements in this very complex region bears further investigation , as does that fact that , in both datasets , the mean curves actually show more extinction in the 2.2 @xmath51 m @xmath96 band than in the 1.65 @xmath51 m @xmath97 band .
the discussion above leads us to three primary conclusions : 1 ) the extinction towards ic 4665 shows at most only a small degree of spatial variability comparable with our measurement errors and that the cluster extinction curve would best be represented by averaging the results for the seven hottest cluster stars studied here ; 2 ) we are able to determine accurate [ m / h ] values from the observations ; and 3 ) our analysis yields reliable estimates of the ( small ) uncertainties in the extinction - without - standards curves and parameters , barring the presence of unusual systematic anomalies in the data sample .
the first and second conclusions are scientific issues which we will pursue further in the future .
the third provides a formidable demonstration of the superiority of the extinction - without - standards technique over the classical pair method for deriving extinction curves in both the precision of the results and the quantification of the uncertainties . with our error
estimates verified , we now examine how curves derived by the current approach compare to pair method curves .
figure [ figsed2 ] shows the uv - through - ir fits to the seds of a set of 9 moderately - reddened early - b stars , most of which are well known for their extinction properties , and figure [ figext2 ] shows the corresponding extinction curves .
these stars were selected because they illustrate the wide range that exists in the wavelength dependence of both uv and ir extinction .
for these stars the ir data allow us to determine the values of @xmath38 and so the flexible extinction curve fits are shown throughout the ir - to - uv domain . as for the ic
4665 stars , all the parameters describing the fits are given in tables 1 and 2 .
also shown in figure [ figext2 ] are uv curves based on the pair method technique .
the curve for hd 294264 is from valencic , clayton , & gordon ( 2004 ) ; the curves for hd 210121 and hd 27778 are from unpublished measurements by us ; and the others are from the catalog of fm90 .
the agreement between the model - based and pair method curves is reasonable , and much better than seen for ic 4665 .
this is consistent with spectral mismatch as the prime cause of the existing discrepancies , since the nine stars have both higher @xmath0 values and earlier spectral types than the ic 4665 stars , both of which tend to reduce the influence of mismatch errors .
note also that the best agreement between the pair method and the model - based curves occurs for the star hd 204827 , for which identical results are found .
again , this is as it should be , since its @xmath0 is the largest of any star in the sample , minimizing the impact mismatch effects .
the 1-@xmath88 uncertainties for our flexible extinction curve fits are indicated by shaded regions around the curves , as in figure [ figext1 ] .
the curves shown in figure [ figext2 ] demonstrate the ability of the parameterized , flexible extinction curve ( see [ secflexi ] ) to conform itself to the wide range of extinction curve shapes encountered in interstellar space .
figure [ figsed3 ] shows a set of nine sed fits for low color excess stars ( 0.10 @xmath98 0.21 ) located along sightlines bounded by the galactic coordinates @xmath99 and @xmath100 .
figure [ figext3 ] shows the corresponding extinction curves .
these stars are all mid - to - late b members of the upper scorpius complex ( garrison 1967 ) and we encountered them while testing a procedure for scanning the _ iue _ archives and automatically generating model - based extinction curves for stars in the appropriate spectral range .
when examining the results for this pilot program which sampled high - latitude stars , it became obvious that curves derived from stars in this specific region demonstrated similar curve morphology which is distinctly different from the galactic average .
given the low reddening and preponderance of late b stars , this strong systematic behavior would be missed by a pair method survey , with its large inherent mismatch errors .
strong regional signatures are important in extinction studies because they may highlight the effects of specific physical processes on dust grain populations .
although we will examine this specific region in the future , two points are worth mentioning .
first , it is important that similar curves result from stars with spectral types ranging from b2.5 to b9 , verifying that the curves are not the result of some temperature - dependent systematic effect in fitting process .
it is also worth noting that the region is near the @xmath101 oph dark cloud .
the star @xmath101 oph a ( hd 147933 , whose extinction curve is shown in figure [ figext2 ] ) is located at a comparable distance and just south of the region , at galactic coordinates of ( @xmath102 , @xmath103 ) = ( @xmath104 ) . the mean curve for the nine lightly - reddened stars
is shown at the top of figure [ figext3 ] , along with the average galactic curve ( dash - dot curve ) and the hd 1479433 curve ( dotted curve ) for comparison . in the uv
, the regional curve is seen to be almost identical to that for the more heavily - reddened hd 147933 ( @xmath105 ) , suggesting that the sightlines sample similar dust populations .
interestingly , however , the curves are not identical in the ir .
2mass _ _ jhk _ data for the nine star sample implies a value of @xmath106 , slightly larger than the galactic mean of 3.1 , while hd 147933 has a value of @xmath107 .
this contrast between uv and ir behavior may indicate different physical processes at work along the higher density sightline towards hd 147933 , or perhaps different timescales in the response of uv and ir extinction to modifications of dust grain properties . in producing the extinction curves for these lightly reddened stars
, we found that , in some cases , the value of @xmath66 was very poorly determined and produced ( presumably ) spurious `` bumps and wiggles '' in the ir portion of the curves .
to eliminate this effect , we imposed a constraint on @xmath66 for the whole sample , namely @xmath108 .
this is taken from f04 , who found a very strong relationship between @xmath38 and @xmath66 from a larger sample of more heavily reddened stars ( see figure 6 of f04 ) .
the ability to impose scientifically reasonable constraints on the fitting procedure is a major advantage of the extinction - without - standards approach , and can potentially allow high quality extinction results to be derived from stars with even lower reddenings than those shown in figure [ figext3 ] .
in the previous sections we first provided an overview of the process used to determine an extinction curve , clarifying the measurement process . we then presented a new method for deriving uv - through - ir extinction curves , based on the use of stellar atmosphere models to provide estimates of the intrinsic ( unreddened ) stellar seds rather than unreddened ( or lightly reddened ) `` standard '' stars .
we have shown that this `` extinction - without - standards '' technique greatly increases the accuracy of the derived extinction curves , particularly in the cases of low reddening and cool spectral types ( i.e. , late - b ) , and allows a realistic estimation of the uncertainties .
a side benefit of the technique is the simultaneous determination of fundamental properties of the reddened stars themselves ( @xmath40 , @xmath41 , [ m / h ] , and @xmath109 ) , making the procedure valuable for both stellar and interstellar studies .
given the physical limitations of the _ atlas9 _ models we currently employ , the technique is limited to near main sequence b stars .
however , in principal , the procedure can be adapted to any class of star for which accurate model seds are available and for which the signature of interstellar reddening can be distinguished from those of the stellar parameters .
although we developed the procedure based on _ iue _ spectrophotometry in the uv , and photometry in the optical and near - ir ( requiring a calibration of synthetic optical and near - ir photometry ) , the ideal application of the technique would be with spectrophotometric data throughout the uv - through - ir domain , allowing the most detailed examination of the wavelength dependence of the extinction curves .
the increased precision in the derived extinction curves allows us to improve our understanding of extinction in a number of ways .
first and most simply , we will determine the basic uv - to - ir wavelength dependence of extinction along a wide variety of sightlines more precisely than has been possible in the past .
second and given that we know that extinction curve morphology varies widely from sightline - to - sightline ( see figure [ figext2 ] ! ) we will be able to study the form of the variability and search for relationships between various features and wavelength domains using a data set with small and well - defined uncertainties
. such relationships , e.g. , the correlation between @xmath38 and the steepness of uv extinction discovered by cardelli et al .
( 1989 ) , provide important constraints on the dust grain population causing the extinction .
non - correlations can be equally important .
for example , the lack of a correlation between the position and width of the 2175 bump demonstrated by fitzpatrick & massa ( 1986 ) remains a strong constraint on models for the bump carrier ( draine 2003 ) . in either case , a precise knowledge of the measurement errors is required for transforming an observation into a scientific constraint .
finally , we will be able to place much stronger and more well - defined limits on the relationship between extinction curve morphology and interstellar environment .
curves derived from lines of sight to specific , localized regions , such as towards open clusters , can be particularly useful for relating curve properties to physical processes occurring in the interstellar medium ( see , e.g. , the study of cepheus ob3 by massa & savage 1984 and trumpler 37 by clayton & fitzpatrick 1987 ) .
the ability to accurately probe low @xmath0 sightlines ( as exemplified in figures [ figext1 ] and [ figext3 ] ) opens the door to studies of dust in regions that have not been thoroughly explored . these include halo dust , dust in very low density regions , and local dust .
halo dust is especially important since we must contend with its effects every time we look out of the galaxy .
there are indications ( kiszkurno - koziej & lequeux 1987 ) that its properties differ systematically from dust at lower galactic altitudes , and this result needs to be verified on a star - by - star basis .
the nature of low density dust provides insights into the processing which occurs in hostile environments .
clayton et al .
( 2000 ) presented observations of dust from low density sightlines , and their results were intriguing .
however , they were forced to select sightlines which accumulate fairly substantial color excesses , introducing the possibility of mixed grain populations .
furthermore , the results for their least - reddened sightlines were , as they acknowledge , poorly determined , forcing them to base their conclusions on a global average of properties of their sample .
finally , measuring the properties of local dust is important because it allows us to search for isolated , relatively homogeneous environments with uniform extinction curve shapes .
these may signal physically and kinematically isolated regions which would be ideal for follow up interstellar line studies .
in addition , _ hipparcos _ parallax data exist for nearby stars and provide an opportunity to study the 3-dimensional structure of local extinction .
these stars are especially important because their space density is higher than that of the early - b stars usually used in extinction studies and thus their inclusion increases the number of stars available to create curves for nearby sightlines .
study of the mid - to - late b stars will enable the examination of the spatial structure of local dust absorption more thoroughly than previously possible and will greatly enhance our understanding of the local interstellar medium . the ability to construct accurate curves for mid - to - late b stars will also allow us to study extinction in open clusters and associations which are dominated by these stars , such as the pleiades , @xmath84 per cluster , and ic 4665 ( see figure [ figext1 ] ) . because our model atmosphere approach does not require human intervention ,
once the basic data have been assembled , it is possible to process large data sets at one time .
naturally , the automated results must be inspected for outliers and data anomalies .
nevertheless , this approach reduces the work load considerably and a first - attempt yielded the regional anomaly shown in figure [ figext3 ] .
furthermore , since the results are produced in a uniform manner , it is a relatively simple matter to inspect them for correlations between various curve properties , for anomalous curve shapes , and for spatial trends on the sky .
in addition to dust parameters , our technique provides a meaningful , quantitative physical properties for the reddened stars .
the temperature and surface gravity information will be useful for population studies of b stars in the field and in clusters .
however , perhaps the most useful property will be the metallicity .
we have demonstrated that several stars in the same cluster , which have a range in temperatures and gravities , all yield the same [ m / h ] .
this verifies the sensitivity of our fitting procedure to this important quantity . as a result , we are confident that application of our procedure to large scale surveys of reddened , near main sequence b stars can provide a census of the distribution of metallicity throughout the galaxy and the local universe .
e.f . acknowledges support from nasa grant nag5 - 12137 , nag5 - 10385 , and nng04gd46 g . d.m .
acknowledges support from nasa grant nng04ec01p . some of the data presented in this paper were obtained from the multimission archive at the space telescope science institute ( mast )
stsci is operated by the association of universities for research in astronomy , inc .
, under nasa contract nas5 - 26555 .
support for mast for non - hst data is provided by the nasa office of space science via grant nag5 - 7584 and by other grants and contracts .
this publication also makes use of data products from the two micron all sky survey , which is a joint project of the university of massachusetts and the infrared processing and analysis center / california institute of technology , funded by the national aeronautics and space administration and the national science foundation .
lllcrcrc + hd 161165 & b8.5 v & @xmath110 & @xmath111 & @xmath112 & @xmath113 & @xmath114 & @xmath115 + hd 161184 & b8 v & @xmath116 & @xmath117 & @xmath118 & @xmath119 & @xmath120 & @xmath121 + hd 161572 & b6 v & @xmath122 & @xmath123 & @xmath124 & @xmath125 & @xmath126 & @xmath127 + hd 161573 & b3 iv & @xmath128 & @xmath129 & @xmath130 & @xmath125 & @xmath131 & @xmath132 + hd 161603 & b5 iv & @xmath133 & @xmath134 & @xmath135 & @xmath125 & @xmath136 & @xmath137 + hd 161660 & b7 v & @xmath138 & @xmath139 & @xmath140 & @xmath141 & @xmath142 & @xmath143 + hd 161677 & b5 iv & @xmath144 & @xmath145 & @xmath146 & @xmath125 & @xmath147 & @xmath148 + hd 161734 & b8 v & @xmath149 & @xmath150 & @xmath112 & @xmath125 & @xmath151 & @xmath152 + hd 162028 & b6 v & @xmath153 & @xmath154 & @xmath155 & @xmath125 & @xmath156 & @xmath127 + + hd 21483 & b3 iii & @xmath157 & @xmath158 & @xmath159 & @xmath125 & @xmath160 & @xmath161 + hd 27778 & b4 v & @xmath162 & @xmath163 & @xmath164 & @xmath125 & @xmath165 & @xmath166 + hd 37061 & b1 v & @xmath167 & @xmath168 & @xmath169 & @xmath170 & @xmath171 & @xmath172 + hd 37367 & b2 iv - v & @xmath173 & @xmath174 & @xmath175 & @xmath176 & @xmath177 & @xmath178 + hd 37903 & b1.5 v & @xmath179 & @xmath168 & @xmath180 & @xmath125 & @xmath181 & @xmath166 + hd 147933 & b2 iv & @xmath182 & @xmath168 & @xmath183 & @xmath125 & @xmath184 & @xmath185 + hd 204827 & b0 v & @xmath186 & @xmath168 & @xmath187 & @xmath170 & @xmath188 & @xmath189 + hd 210121 & b3 v & @xmath190 & @xmath191 & @xmath192 & @xmath125 & @xmath193 & @xmath194 + hd 294264 & b3 vn & @xmath195 & @xmath168 & @xmath196 & @xmath125 & @xmath197 & @xmath198 + + hd 142096 & b2.5 v & @xmath199 & @xmath168 & @xmath200 & @xmath125 & @xmath201 & @xmath202 + hd 142165 & b6 ivn & @xmath203 & @xmath168 & @xmath204 & @xmath125 & @xmath205 & @xmath206 + hd 142315 & b8 vnn & @xmath207 & @xmath208 & @xmath209 & @xmath210 & @xmath211 & @xmath212 + hd 142378 & b3 v & @xmath213 & @xmath168 & @xmath214 & @xmath215 & @xmath216 & @xmath202 + hd 143567 & b9 va & @xmath217 & @xmath168 & @xmath218 & @xmath125 & @xmath219 & @xmath121 + hd 145554 & b9 van & @xmath220 & @xmath221 & @xmath222 & @xmath125 & @xmath223 & @xmath224 + hd 146001 & b7 iv & @xmath225 & @xmath168 & @xmath226 & @xmath227 & @xmath228 & @xmath202 + hd 146029 & b9 va & @xmath229 & @xmath230 & @xmath231 & @xmath232 & @xmath233 & @xmath234 + hd 146416 & b9 vnn & @xmath235 & @xmath236 & @xmath237 & @xmath125 & @xmath238 & @xmath239 + lccccrrcccrccc + hd 161165 & & & & @xmath240 & @xmath241 & @xmath242 & & @xmath243 & @xmath244 & @xmath245 & @xmath246 & @xmath247 & @xmath248 + hd 161184 & & & & @xmath249 & @xmath250 & @xmath242 & & @xmath251 & @xmath252 & @xmath253 & @xmath254 & @xmath255 & @xmath256 + hd 161572 & & & & @xmath257 & @xmath258 & @xmath259 & & @xmath260 & @xmath261 & @xmath262 & @xmath263 & @xmath264 & @xmath265 + hd 161573 & & & & @xmath266 & @xmath267 & @xmath259 & & @xmath268 & @xmath269 & @xmath270 & @xmath271 & @xmath272 & @xmath273 + hd 161603 & & & & @xmath274 & @xmath267 & @xmath259 & & @xmath251 & @xmath275 & @xmath276 & @xmath277 & @xmath278 & @xmath279 + hd 161660 & & & & @xmath280 & @xmath281 & @xmath259 & & @xmath251 & @xmath282 & @xmath283 & @xmath284 & @xmath285 & @xmath286 + hd 161677 & & & & @xmath287 & @xmath288 & @xmath259 & & @xmath289 & @xmath290 & @xmath291 & @xmath292 & @xmath293 & @xmath286 + hd 161734 & & & & @xmath294 & @xmath295 & @xmath242 & & @xmath251 & @xmath296 & @xmath297 & @xmath298 & @xmath299 & @xmath300 + hd 162028 & & & & @xmath301 & @xmath288 & @xmath259 & & @xmath251 & @xmath302 & @xmath303 & @xmath304 & @xmath305 & @xmath306 + + hd 21483 & @xmath307 & @xmath308 & & @xmath309 & @xmath310 & @xmath311 & & @xmath312 & @xmath313 & @xmath314 & @xmath315 & @xmath316 & @xmath317 + hd 27778 & @xmath318 & @xmath319 & & @xmath320 & @xmath321 & @xmath242 & & @xmath322 & @xmath323 & @xmath324 & @xmath325 & @xmath326 & @xmath327 + hd 37061 & @xmath328 & @xmath329 & & @xmath330 & @xmath331 & @xmath242 & & @xmath332 & @xmath333 & @xmath334 & @xmath335 & @xmath336 & @xmath337 + hd 37367 & @xmath338 & @xmath339 & & @xmath340 & @xmath341 & @xmath311 & & @xmath342 & @xmath343 & @xmath344 & @xmath286 & @xmath345 & @xmath346 + hd 37903 & @xmath347 & @xmath348 & & @xmath349 & @xmath331 & @xmath259 & & @xmath350 & @xmath351 & @xmath352 & @xmath353 & @xmath354 & @xmath355 + hd 147933 & @xmath356 & @xmath357 & & @xmath358 & @xmath241 & @xmath311 & & @xmath359 & @xmath360 & @xmath361 & @xmath362 & @xmath363 & @xmath364 + hd 204827 & @xmath365 & @xmath366 & & @xmath367 & @xmath368 & @xmath369 & & @xmath370 & @xmath371 & @xmath372 & @xmath373 & @xmath374 & @xmath375 + hd 210121 & @xmath376 & @xmath377 & & @xmath378 & @xmath379 & @xmath259 & & @xmath380 & @xmath381 & @xmath382 & @xmath383 & @xmath384 & @xmath385 + hd 294264 & @xmath386 & @xmath387 & & @xmath388 & @xmath341 & @xmath369 & & @xmath251 & @xmath389 & @xmath390 & @xmath391 & @xmath392 & @xmath393 + + hd 142096 & @xmath394 & @xmath395 & & @xmath396 & @xmath397 & @xmath242 & & @xmath260 & @xmath398 & @xmath399 & @xmath400 & @xmath401 & @xmath402 + hd 142165 & @xmath403 & @xmath404 & & @xmath405 & @xmath406 & @xmath259 & & @xmath322 & @xmath407 & @xmath408 & @xmath409 & @xmath410 & @xmath411 + hd 142315 & @xmath412 & @xmath413 & & @xmath414 & @xmath250 & @xmath242 & & @xmath415 & @xmath416 & @xmath417 & @xmath418 & @xmath419 & @xmath400 + hd 142378 & @xmath420 & @xmath421 & & @xmath422 & @xmath295 & @xmath242 & & @xmath423 & @xmath424 & @xmath425 & @xmath426 & @xmath427 & @xmath428 + hd 143567 & @xmath429 & @xmath430 & & @xmath431 & @xmath331 & @xmath242 & & @xmath260 & @xmath432 & @xmath433 & @xmath434 & @xmath435 & @xmath436 + hd 145554 & @xmath437 & @xmath438 & & @xmath439 & @xmath341 & @xmath311 & & @xmath268 & @xmath440 & @xmath441 & @xmath442 & @xmath443 & @xmath444 + hd 146001 & @xmath445 & @xmath446 & & @xmath447 & @xmath267 & @xmath242 & & @xmath243 & @xmath448 & @xmath449 & @xmath450 & @xmath451 & @xmath452 + hd 146029 & @xmath453 & @xmath454 & & @xmath455 & @xmath321 & @xmath242 & & @xmath415 & @xmath456 & @xmath457 & @xmath458 & @xmath459 & @xmath460 + hd 146416 & @xmath461 & @xmath462 & & @xmath463 & @xmath464 & @xmath259 & & @xmath465 & @xmath466 & @xmath467 & @xmath468 & @xmath469 & @xmath470 + lrccc @xmath471 & @xmath472 & @xmath473 & @xmath474 & + @xmath475 & @xmath476 & @xmath477 & @xmath478 & @xmath479 + @xmath480 & @xmath481 & @xmath482 & @xmath483 & @xmath482 + @xmath55 & @xmath484 & @xmath485 & @xmath486 & @xmath487 + @xmath56 & @xmath488 & @xmath489 & @xmath490 & @xmath491 + @xmath52 & @xmath492 & @xmath493 & @xmath494 & @xmath495 + @xmath53 & @xmath496 & @xmath497 & @xmath498 & @xmath499 + @xmath57 & @xmath500 & @xmath501 & @xmath502 & @xmath503 + @xmath58 & @xmath504 & @xmath505 & @xmath506 & @xmath507 + @xmath93 & @xmath508 & @xmath509 & @xmath510 & @xmath511 + | in this paper we present a new method for deriving uv - through - ir extinction curves , based on the use of stellar atmosphere models to provide estimates of the intrinsic ( i.e. , unreddened ) stellar spectral energy distributions ( seds ) , rather than unreddened ( or lightly reddened ) standard stars .
we show that this `` extinction - without - standards '' technique greatly increases the accuracy of the derived extinction curves and allows realistic estimations of the uncertainties .
an additional benefit of the technique is that it simultaneously determines the fundamental properties of the reddened stars themselves , making the procedure valuable for both stellar and interstellar studies .
given the physical limitations of the models we currently employ , the technique is limited to main sequence and mildly - evolved b stars .
however , in principle , it can be adapted to any class of star for which accurate model seds are available and for which the signatures of interstellar reddening can be distinguished from those of the stellar parameters .
we demonstrate how the extinction - without - standards curves make it possible to : 1 ) study the uniformity of curves in localized spatial regions with unprecedented precision ; 2 ) determine the relationships between different aspects of curve morphology ; 3 ) produce high quality extinction curves from low color excess sightlines ; and 4 ) derive reliable extinction curves for mid - late b stars , thereby increasing spatial coverage and allowing the study of extinction in open clusters and associations dominated by such stars .
the application of this technique to the available database of uv - through - ir seds , and to future observations , will provide valuable constraints on the nature of interstellar grains and on the processes which modify them , and will enhance our ability to remove the multi - wavelength effects of extinction from astronomical energy distributions . |
gravitational effects on the light propagation provide powerful tools for investigating extrasolar planets , dark matter and dark energy in modern astronomy and cosmology .
the effects are important also in theoretical physics particularly for studying a null structure of a spacetime .
a rigorous form of the bending angle has been studied in understanding a strong gravitational field @xcite .
for example , strong gravitational lensing in a schwarzschild black hole was considered by frittelli , kling and newman @xcite , by virbhadra and ellis @xcite and more comprehensively by virbhadra @xcite ; virbhadra , narasimha and chitre @xcite studied distinctive lensing features of naked singularities .
virbhadra and ellis @xcite and virbhadra and keeton @xcite later described the strong gravitational lensing by naked singularities ; eiroa , romero and torres @xcite treated reissner - nordstrm black hole lensing ; perlick @xcite discussed the lensing by a barriola - vilenkin monopole and also that by an ellis wormhole @xcite , the latter of which had been discussed long before ( for instance , @xcite ) .
moreover , it has been investigated as an observational probe of violations of some energy conditions @xcite . there
exist several forms of the deflection angle by the ellis wormhole @xcite .
a reason for such differences has been clarified by several authors @xcite .
furthermore , tsukamoto and his collaborators @xcite have investigated the lensing in tangherlini spacetime as an extension of schwarzschild one to arbitrary dimensions ; horvath , gergely , and hobill @xcite studied lensing effects with negative convergence by so - called tidal charges in the dadhich et al .
solution @xcite , where , for a brane world black hole , the tidal charge arises from tidal forces acting at the brane - bulk boundary .
note that negative convergence in this case does not require any exotic matter .
it comes from the weyl curvature in higher dimensions .
in addition , gibbons and kodama @xcite have shown that curvature - regular asymptotically flat solitons with negative mass are contained in the myers - perry family , although the soliton solutions in the odd spacetime dimensions might not correspond to astrophysical objects .
kunz and her collaborators have constructed traversable wormholes in the dilatonic einstein - gauss - bonnet theory @xcite , in which they do not require any form of exotic matter for their existence .
therefore , it is interesting to study gravitational lenses for probing an exotic energy or new physics .
furthermore , the arrival time delay of light in schwarzschild spacetime , which is called shapiro time delay , is another gravitational effect on light in a curved spacetime .
the time delay effect has successfully tested the einstein s theory @xcite .
a significant improvement was done by using doppler tracking of the cassini spacecraft on its way to the saturn @xcite .
gravitational time delay in some modified gravity models has been discussed by assuming a small perturbation around the schwarzschild spacetime in the weak field approximation @xcite ; virbhadra and keeton @xcite and deandrea and alexander @xcite have discussed time delay by naked singularities to test the cosmic censorship hypothesis .
time delay by strongly naked singularities are negative , where strongly naked singularities are those that are not covered by any photon sphere . therefore , it will be interesting to study the time delay of light for understanding of other exotic lens models .
furthermore , the frequency shift is caused by a change of the time delay with time , if the emitter of light moves .
note that this frequency shift is different from the gravitational redshift due to a difference between clocks at two places .
see figure [ figure-1 ] , in which the frequency shift is caused by the time delay of each signal . actually , it is the frequency shift by the gravitational time delay that was measured by the cassini spacecraft experiment @xcite .
hence , the present paper will study both the arrival time delay and the induced frequency shift .
in particular , we reexamine the spacetime model that has been discussed by kitamura et al .
@xcite to study exotic gravitational lensing effects .
this spacetime metric depends on the inverse distance to the power of positive @xmath0 in the weak field approximation .
the schwarzschild spacetime and the ellis wormhole correspond to @xmath1 and @xmath2 , respectively , so that these spacetimes can be expressed as a one - parameter family .
this one - parameter model expresses a spherical mass distribution that may be related with violations of some energy conditions for a certain parameter region .
note that birkhoff s theorem could say that cases @xmath3 might be non - vacuum , if the models were interpreted in the framework of the standard einstein equation .
this lens model has suggested possible demagnification of lensed images @xcite , radial shear @xcite , and an anomaly in centroid motions of images @xcite . this paper is organized as follows .
section ii briefly summarizes the basics of the inverse - power lens models @xcite . in section iii ,
we discuss time delay and frequency shift in the modified spacetime .
possible parameter ranges relevant to pulsar timings also are studied .
section iv is devoted to conclusion .
the present paper follows kitamura et al .
@xcite , which assumes that an asymptotically flat , static and spherically symmetric modified spacetime could depend on the inverse distance to the power of positive @xmath0 in the weak field approximation .
we consider the light propagation through a four - dimensional spacetime , though the whole spacetime may be higher dimensional .
the four - dimensional spacetime metric is @xcite @xmath4 where @xmath5 is the circumference radius and @xmath6 and @xmath7 are small book - keeping parameters in iterative calculations .
this model under study might describe a regular distribution of matter for @xmath3 ( see @xcite on this point ) .
the weak - field approximation means @xmath8 and @xmath9 .
namely , we study a far field from the lens object as @xmath10 and @xmath11 .
note that eq .
( [ ds ] ) is not valid in the strong field near @xmath12 ( please see a note in @xcite for more detail ) . here
, @xmath6 and @xmath7 may be either positive or negative , respectively .
negative @xmath6 and @xmath7 for @xmath1 correspond to a negative mass ( in the linearized schwarzschild metric ) . without loss of generality
, we focus on the equatorial plane @xmath13 , since the spacetime is spherically symmetric .
the deflection angle of light becomes at the linear order @xcite @xmath14 where the integral is positive definite , @xmath15 denotes the impact parameter of the light ray , we denote @xmath16 , and we define @xmath17 by @xmath18 for the closest approach @xmath19 . by absorbing the positive integral into the parameter @xmath20 , we rewrite the linear - order deflection angle simply as @xmath21 , where the sign of @xmath22 is the same as that of @xmath20 .
this deflection angle recovers the schwarzschild ( @xmath1 ) and ellis wormhole ( @xmath2 ) cases .
@xmath23 and @xmath24 correspond to positive deflection of light and negative one , respectively .
hence , let us call them a gravitational convex lens and a gravitational concave one , respectively .
it is unlikely that there exists any photon sphere in the concave model .
for instance , see eq .
( 7 ) in ref .
@xcite , which suggests @xmath25 for the existence of a photon sphere .
we mention an effective mass .
a simple application of the standard lens theory @xcite suggests that the deflection angle of light in the form of @xmath21 corresponds to a convergence ( scaled surface mass density ) as @xmath26 which implies an extended spherical distribution of matter ( or energy ) for @xmath3 and a singular source only for @xmath27 . for the weak - field schwarzschild case ( @xmath27 )
, it follows that the convergence vanishes . if @xmath23 and @xmath28 , or if @xmath24 and @xmath29 , the effective surface mass density of the lens object is interpreted as negative in the framework of the standard lens theory @xcite . for these two cases ,
the matter ( and energy ) need to be exotic . if @xmath24 and @xmath28 , or if @xmath23 and @xmath29 , on the other hand , the convergence is positive almost everywhere except for the central singularity and hence exotic matter ( and energy ) are not required in the framework of the standard lens theory , in spite of the gravitational repulsion on light rays .
note that convex lenses ( @xmath23 ) and concave ones ( @xmath24 ) in the above two - parameter models do not have a one - to - one correspondence to positive convergence @xmath30 and negative one @xmath31 .
by using @xmath32 for a light signal , we obtain the orbit equation as @xmath33 + o(\varepsilon_1 ^ 2 , \varepsilon_2 ^ 2 , \varepsilon_1 \varepsilon_2 ) , \label{orbit}\ ] ] where the impact parameter @xmath15 is related with the closest approach @xmath19 as @xmath34 the time - of - flight of a photon from the source ( denoted by @xmath35 ) to the receiver ( denoted by @xmath36 ) becomes @xmath37 where we define @xmath38 .
see figure [ figure-2 ] for the photon path . subtracting the time in the flat spacetime from it provides the time delay at the linear order as @xmath39 where @xmath40 and @xmath41 correspond to the direction from the lens to the receiver and that to the source of light , respectively . for general @xmath0 , the integral in eq .
( [ delay ] ) is not always expressed in terms of elementary functions .
hence , numerical computations are required for eq .
( [ delay ] ) .
if @xmath0 is an integer , however , the integration of powers of trigonometric functions can be done @xcite . for @xmath1 , eq .
( [ delay ] ) becomes @xmath42 + \tilde{\varepsilon}\left [ 2\ln\dfrac{(r_r+x_r)(r_s+x_s)}{r_0 ^ 2}\right ] , \label{delay - n1}\ ] ] which agrees with the shapiro delay formula if @xmath43 equals to schwarzschild radius .
here , we define @xmath44 and @xmath45 . for @xmath2 , we obtain @xmath46 next , we consider the case of an even integer ( @xmath47 ) , where @xmath48 is a positive integer . then , we obtain @xmath49 where the subscript @xmath50 indicates the case of @xmath47 and @xmath51 means @xmath52 .
next , let us consider the case of @xmath53 .
then , we obtain @xmath54 up to this point , integrations have been done without any approximation . for astronomical situations ,
the closest distance of light is much shorter than the distance from the lens to the source and that from the lens to the observer , so that we can take the limit as @xmath55 and @xmath56 .
this leads to @xmath57 and @xmath58 , so that the above expressions can be simplified .
see also figure [ figure-2 ] .
for @xmath47 , we obtain @xmath59 for @xmath53 case , we obtain @xmath60 eqs .
( [ delay - neven - limit ] ) and ( [ delay - nodd - limit ] ) suggest that the time delay @xmath61 is proportional to @xmath62 and also to @xmath63 , if @xmath64 and @xmath65 are large .
note that the sign of the time delay is the same as that of the deflection angle of light .
the time delay @xmath61 for @xmath23 is a downward - convex function of @xmath19 almost everywhere except for @xmath66 , while @xmath61 for @xmath24 is convex upward .
figure [ figure-3 ] shows schematically a motion of the signal emitter with respect to the lens object .
we assume that the emitter and the lens object are in a linear motion during short - time observations , so that we can take @xmath67 . here
, @xmath68 denotes the transverse component of the relative velocity between the lens and the source with respect to the line of sight , @xmath69 denotes the minimum of @xmath19 and @xmath70 is chosen as the passage time of @xmath71 without loss of generality .
figure [ figure-4 ] shows curves for positive time delays for @xmath72 and @xmath73 with @xmath74 .
figure [ figure-5 ] corresponds to negative delays for @xmath75 .
next , we consider the frequency shift caused by the time delay . in most astronomical situations
, observers can not know the exact time of the emission of light .
hence , the arrival time delay can not be measured directly for a single image , although the arrival time difference between multiple images ( e.g. distant quasars ) are astronomical observables @xcite .
note that the round - trip time of a light signal is a direct observable for a spacecraft such as voyager and cassini . on the other hand ,
the frequency shift induced by the time delay becomes a direct observable , if the emitter of a light signal moves with respect to the lens object @xcite .
the frequency shift @xmath76 due to the time delay is defined as @xcite @xmath77 where @xmath78 denotes the intrinsic frequency of light and @xmath79 means the observed one at time @xmath80 . as figure [ figure-1 ] suggests , @xmath81 if the emitter of light approaches the lens and @xmath82 if it recedes . in general , the expression for @xmath76 may become lengthy , though it is simplified for an integer @xmath0 .
first , we differentiate eqs .
( [ delay - neven ] ) and ( [ delay - nodd ] ) with respect to time , where we use @xmath83 for @xmath84 .
next , we take the limit as @xmath57 and @xmath58 . for @xmath47 , the frequency shift is obtained as @xmath85 and for @xmath53 , it becomes @xmath86 figure [ figure-6 ] for @xmath23 shows the frequency shift for @xmath72 , and @xmath73 . here
, the appropriate value of the parameter @xmath19 is chosen such that the peak location can remain the same with each other in order to enable us to see the behavior of @xmath76 .
the timing shift curve falls off more rapidly as the parameter @xmath0 increases .
this is because the gravitational potential decays faster as @xmath0 becomes larger ( see the spacetime metric in eq .
( [ ds ] ) ) .
( [ y2p ] ) and ( [ y2p+1 ] ) suggest that @xmath87 for large @xmath88
. see figure [ figure-7 ] for the @xmath24 case .
there are two observables in time delay measurements .
one is the amplitude of the time delay corresponding to the curve height in figures [ figure-4 ] and [ figure-6 ] , and the other is the duration corresponding to the curve width in figure [ figure-5 ] and [ figure-7 ] .
( [ delay - neven - limit ] ) and ( [ delay - nodd - limit ] ) suggest that the size of the time delay is @xmath89 the duration of the time delay is roughly @xmath90 let us assume that @xmath68 is of the order of the typical rotational velocity in our galaxy , because there are no established theories on the motion of the exotic lens .
the current pulsar timing measurements are done a few times per year for each pulsar .
th root - mean - squared residuals of the timing arrival dispersions are nearly 100 - 1 nanoseconds ( ns ) that are depending on pulsars @xcite .
the relevant value of the closest approach becomes @xmath91 and the exotic lens potential is roughly @xmath92 this is corresponding to the frequency shift @xmath93 .
this accuracy has been already achieved for a milli - second pulsar @xcite .
furthermore , let us consider a possible constraint on the number density of the exotic lens models .
one event would be detectable , if a lens object entered a volume of a cylinder as @xmath94 , where @xmath95 denotes the distance of the source ( pulsar ) from the observer .
the effective survey volume is proportional to both the number of the observed pulsars ( denoted as @xmath96 ) and the total duration of a pulsar timing project ( denoted as @xmath97 ) .
hence , the number density of the lens objects @xmath98 would be constrained by no event detection as @xmath99 although it seems very weak , this constraint might be interesting , because @xmath28 models are massless at the spatial infinity and thus it is unlikely that these exotic objects are constrained by other observations regarding stellar motions , galactic rotation and so on .
we examined the arrival time delay of light and the frequency shift in the lens model with an inverse power law .
the time delay by a gravitational convex lens ( i.e. positive deflection angle of light ) would be positive , even if the lens model had negative convergence like ellis wormholes .
on the other hand , time delay by a gravitational concave lens might become negative , even if the convergence were positive . by using the janis - newman - winicour metric , virbhadra and keeton @xcite and deandrea and alexander @xcite have shown that negative time delay is caused in strongly naked singularities , where strongly naked singularities are those which are not covered within any photon spheres .
the present model of a gravitational concave lens suggests that there is no photon spheres , though the present analysis is limited within the weak - field approximation .
the previous works and the present one suggest that negative time delay might be related with whether there exists a photon sphere in a spacetime .
this is left as a future work . finally , we mention a comparison with the previous works @xcite .
they find distinctive features in lensing effects that are closely related with negative convergence . on the other hand
, the present paper shows that negative time delay might be caused by a gravitational concave lens , even though the convergence is positive .
therefore , the present result might suggest that time delay in such a spacetime will bring us additional information independent of lensing effects .
further study along this course is left as a future work .
we would like to thank f. abe , m. bartelmann , t. harada , s. hayward , t. kitamura , j. kunz , k. nakao , y. sendouda , r. takahashi , n. tsukamoto , and m. visser for the useful conversations on the exotic lens models .
99 s. frittelli , t. p. kling , and e. t. newman , phys .
d * 61 * , 064021 ( 2000 ) .
k. s. virbhadra , and g. f. r. ellis , phys .
d * 62 * , 084003 ( 2000 ) .
k. s. virbhadra , phys .
d * 79 * , 083004 ( 2009 ) .
k. s. virbhadra , d. narasimha , and s. m. chitre , astron . astrophys . * 337 * , 1 ( 1998 ) .
k. s. virbhadra , and g. f. r. ellis , phys . rev .
d * 65 * , 103004 ( 2002 ) .
k. s. virbhadra , and c. r. keeton , phys .
d * 77 * , 124014 ( 2008 ) .
e. f. eiroa , g. e. romero , and d. f. torres , phys .
d * 66 * , 024010 ( 2002 ) .
v. perlick , phys .
d * 69 * , 064017 ( 2004 ) .
h. g. ellis , j. math .
* 14 * , 104 ( 1973 ) . m. s. morris , and k. s. thorne , am
. j. phys . *
56 * , 395 ( 1988 ) .
m. s. morris , k. s. thorne , and u. yurtsever , phys .
61 * , 1446 ( 1988 ) .
l. chetouani , and g. clment , gen .
* 16 * , 111 ( 1984 ) .
g. clment , int .
* 23 * , 335 ( 1984 ) .
m. visser , _ lorentzian wormholes : from einstein to hawking _
( aip , new york , 1995 ) .
m. safonova , d. f. torres , and g. e. romero , phys .
d * 65 * , 023001 ( 2001 ) .
a. a. shatskii , astron .
48 * , 525 ( 2004 ) .
k. k. nandi , y. z. zhang , and a. v. zakharov , phys .
d * 74 * , 024020 ( 2006 ) .
f. abe , astrophys .
j. * 725 * , 787 ( 2010 ) .
y. toki , t. kitamura , h. asada , and f. abe , astrophys .
j. * 740 * , 121 ( 2011 ) .
n. tsukamoto , t. harada , k. yajima , phys .
d * 86 * , 104062 ( 2012 ) .
n. tsukamoto , and t. harada , phys .
d * 87 * , 024024 ( 2013 ) . c. m. yoo , t. harada , and n. tsukamoto , phys .
d * 87 * , 084045 ( 2013 ) .
t. k. dey , and s. sen , mod .
a , * 23 * , 953 ( 2008 ) .
a. bhattacharya , and a. a. potapov , mod .
a , * 25 * , 2399 ( 2010 ) . k. nakajima , and h. asada , phys .
d * 85 * , 107501 ( 2012 ) .
g. w. gibbons , and m. vyska , class .
* 29 * 065016 ( 2012 ) .
n. tsukamoto , t. kitamura , k. nakajima , and h. asada , arxiv:1402.6823 [ gr - qc ] .
z. horvath , l. gergely , and d. hobill , class .
* 27 * , 235006 ( 2010 ) .
n. dadhich , r. maartens , p. papadopoulos , and v. rezania , phys . lett .
b , * 487 * , 1 ( 2000 ) .
g. gibbons , and h. kodama , prog .
* 121 * , 1361 ( 2009 ) .
b. kleihaus , j. kunz , and e. radu , phys .
lett . * 106 * , 151104 ( 2011 ) .
p. kanti , b. kleihaus , and j. kunz , phys .
rev d * 85 * , 044007 ( 2012 ) .
c. m. will , living rev .
relativity , * 9 * , 3 ( 2006 ) , [ http://relativity.livingreviews.org/articles/lrr-2006-3 ] .
b. bertotti , l. iess , and p. tortora , nature , * 425 * , 374 ( 2003 ) .
h. asada , phys .
b , * 661 * , 78 ( 2008 )
. j. p. deandrea , and k. m. alexander , arxiv:1402.5630 [ gr - qc ] . t. kitamura , k. nakajima , and h. asada , phys .
d * 87 * , 027501 ( 2013 ) .
k. izumi , c. hagiwara , k. nakajima , t. kitamura , and h. asada , phys .
d * 88 * , 024049 ( 2013 ) .
t. kitamura , k. izumi , k. nakajima , c. hagiwara , and h. asada , phys .
d * 89 * , 084020 ( 2014 ) .
p. schneider , j. ehlers , and e. e. falco , _ gravitational lenses _
( springer , new york , 1992 ) .
i. s. gradshteyn , and i. m. ryzhik , _ table of integrals , series and products _ , seventh edition , pages 151 - 160 , ( academic press , new york , 2007 ) .
g. hobbs , et al .
, classical and quantum gravity , * 27 * , 8 ( 2010 ) . n. yunes , and x. siemens , living rev . relativity , * 16 * , 9 ( 2006 ) , that is the time measured by a clock at the source of light , while the other is indicating @xmath100 that is the time measured by the receiver . the dotted vertical lines and solid ones are corresponding to the no - delay case and the time delay case , respectively .
signals of light are emitted with the same time interval , so that the intrinsic frequency of the signal can be constant .
if the source is approaching to the lens object , the gravitational time delay of each signal is increasing with time and hence the observed frequency is lower than the intrinsic one , if the source is receding from the lens object , on the other hand , the delay is decreasing and hence the observed frequency is higher than the intrinsic one . , width=377 ] .
the solid , dot - dashed , dashed , and dotted curves correspond to @xmath1 , @xmath102 , @xmath103 and @xmath73 , respectively .
the horizontal axis denotes the time @xmath80 in days and the vertical axis means the time delay @xmath104 in seconds . here , we assume @xmath69 is @xmath105 au and @xmath106 km / s .
the lens is assumed to be a ten solar mass black hole for @xmath1 ( @xmath107 ) , and the parameters for the other @xmath0 are chosen such that the peak height of the time delay curve can remain the same as each other .
, width=377 ] | gravitational lens models , some of which might act as if a concave lens , have been recently investigated by using a static and spherically symmetric modified spacetime metric that depends on the inverse distance to the @xmath0-th power [ kitamura , nakajima and asada , prd 87 , 027501 ( 2013 ) ] .
we reexamine the time delay of light in a gravitational concave lens as well as a gravitational convex one .
the frequency shift due to the time delay is also investigated .
we show that the sign of the time delay in the lens models is the same as that of the deflection angle of light .
the size of the time delay decreases with increase in the parameter @xmath0 .
we discuss also possible parameter ranges that are relevant to pulsar timing measurements in our galaxy . |
since the first brown dwarfs were discovered two decades ago becklin88 , nakajima95 , hundreds more brown dwarfs have been discovered using wide field infrared surveys .
these substellar objects , too low in mass to fuse hydrogen in their cores , range in mass from @xmath4 13 to 75 @xmath5 and are classified by their spectra into l , t , and most recently y dwarfs kirkpatrick05 , cushing11 . without hydrogen fusion as an internal energy source , brown dwarfs never reach a main - sequence state of constant luminosity ; instead , they cool over time and will transition through the brown dwarf spectral sequence as different molecules and condensates form in their atmospheres . to model their atmospheres accurately
requires an understanding of both the chemistry and physics of the materials that will condense into clouds .
grain or condensate formation has been predicted to play an important role in l dwarf atmospheres since before the first brown dwarfs were discovered lunine86 , lunine89 .
modern equilibrium thermochemical models predict that a variety of different condensates will form in l dwarf atmospheres fegley94
, lodders99 ; by comparing models to observations , it is now well - established that a variety of refractory materials condense in l dwarfs [ see , e.g.][]tsuji96 , allard01,marley02,burrows06,cushing08 .
the condensates that appear to dominate , based on the abundances of elements available to condense , are corundum ( al@xmath0o@xmath6 ) , iron ( fe ) , enstatite ( mgsio@xmath6 ) , and forsterite ( mg@xmath0sio@xmath7 ) , and these species form cloud layers , removing atoms found within the clouds from the lower pressure atmosphere above fegley96 , lodders02 , lodders03 , lodders06 , visscher10 . within windows between major molecular absorption bands , there is little gas opacity so , in models without clouds , the emergent flux comes from hotter layers deep within the atmosphere .
cloud opacity tends to suppress the flux in the near - infrared within these windows ; a thick cloud layer limits the depth from which the planet can radiate , removing some of the flux at these wavelengths , and forcing it to other wavelengths am01 .
when the opacity of these clouds is included in radiative - convective equilibrium models of brown dwarf atmospheres , the resulting model spectra match those of observed l dwarfs cushing06 , cushing08 , saumon08 , stephens09 .
observations show that there is a range of colors for a given spectral type , which are believed to be associated with cloud variations or metallicity , but the details of this are not fully understood .
regardless , observed colors and spectra of l dwarfs can not be well - matched without a significant cloud layer burrows06 . as a brown dwarf continues to cool , it undergoes a significant transformation in its observed spectrum when it reaches an effective temperature of approximately 1400 k. objects cooler than this transitional effective temperature begin to show methane absorption features in their near - infrared spectra and , when these features appear , are classified as t dwarfs burgasser02 , kirkpatrick05 . within a small range of effective temperature ,
the iron and silicate clouds become dramatically less important .
marley10 show that this transition could potentially be explained by the breaking up of these cloud layers into patchy clouds , but the details of the transition are still very much unknown .
however , the recent discovery of highly photometrically variable early t dwarfs suggests that cloud patchiness may indeed play a role radigan12 , artigau09 .
regardless , as the clouds dissipate , the atmospheric windows in the near - infrared clear .
flux emerges from deeper , hotter atmospheric layers , and the brown dwarf becomes much bluer in @xmath8 color ( see figure [ intro - colormag ] )
. the first t dwarf discovered , gl229b nakajima95 , oppenheimer95 , was modeled by marley96 , allard96 , fegley96 , and tsuji96 using cloud - free models .
these models assume that the condensate - forming materials have been removed from the gas phase , but do not contribute to the cloud opacity .
early t dwarfs are generally quite well - modeled using cloudless atmospheric models . however , recent observations of cooler t dwarfs suggest that t dwarfs of type t8 or later ( @xmath1@xmath9 800 k ) appear to be systematically redder in @xmath8 and @xmath10 colors than the cloudless models predict ( see figure [ intro - colormag ] )
. one of the challenges of modeling brown dwarf spectra is the uncertainties in the absorption bands of major gas species such as methane and ammonia , as well as absorption due to collisional processes .
recent work by saumon12 has modeled a range of brown dwarfs using improved line lists for ammonia from yurchenko11 and an improved treatment of the pressure - induced opacity of h@xmath0 collisions from richard12 .
this work improves the accuracy of model near - infrared spectra and reddens the @xmath8 colors of the model spectra with effective temperatures between 500 and 1500 k. the color shift is due to decreased opacity in @xmath11 band from collision - induced absorption and , for @xmath1@xmath9 500 k model only , increased ammonia opacity in @xmath12 band .
however , these improvements do not change the colors enough to match the colors of the coolest t dwarfs .
color is plotted against the absolute @xmath13 magnitude for all known brown dwarfs with measured parallax .
m dwarfs are plotted as black circles , l dwarfs as red circles , and t dwarfs as blue circles .
observational data are from dupuy12 .
models are plotted as solid lines .
blue lines are cloudless models and red lines are cloudy ( @xmath3=2 ) models that include iron , silicate , and corundum clouds . each labeled temperature marks the approximate location of the model with that effective temperature .
the surface gravity of all models is log @xmath2 = 5.0 ( 1000m / s@xmath14 ) . _
bottom : _ same as above , but @xmath8 color is plotted against the absolute @xmath11 magnitude . , width=345 ] color is plotted against the absolute @xmath13 magnitude for all known brown dwarfs with measured parallax .
m dwarfs are plotted as black circles , l dwarfs as red circles , and t dwarfs as blue circles .
observational data are from dupuy12 .
models are plotted as solid lines .
blue lines are cloudless models and red lines are cloudy ( @xmath3=2 ) models that include iron , silicate , and corundum clouds . each labeled temperature marks the approximate location of the model with that effective temperature .
the surface gravity of all models is log @xmath2 = 5.0 ( 1000m / s@xmath14 ) .
_ bottom : _ same as above , but @xmath8 color is plotted against the absolute @xmath11 magnitude .
, width=345 ] clouds are a natural way to redden near - infrared spectra .
cloud opacity limits the emergent flux most prominently in @xmath12 band , so it reddens the @xmath8 and @xmath10 colors of the models .
burgasser10 suggest that the remnants of the iron and silicate clouds could redden these cool t dwarfs , but here we suggest instead that the formation of other condensates , which naturally arise from equilibrium chemistry calculations , may play an important role .
the proposed spectral class y encompasses brown dwarfs that have cooled below @xmath1@xmath4500 k ; a handful of these cool objects have recently been discovered cushing11 , kirkpatrick12 . at these temperatures ,
nh@xmath6 begins to play a more significant role in shaping the near - infrared spectra , and sodium and potassium wane in importance in the optical because they condense into clouds .
appreciable amounts of h@xmath0o and nh@xmath6 will condense into clouds at @xmath1@xmath4350 k and @xmath4200 k , respectively , and will further alter y dwarf spectra . as we discover and characterize more of these cold objects
, the study of clouds will be crucial to understand their spectral characteristics .
silicate , iron , and corundum , which are the condensates that dominate the cloud opacity in our l dwarf models , are not the only condensates that thermochemical models predict will form in substellar atmospheres as they cool .
other condensates will form at lower temperatures and add to the cloud opacity via the same physical processes that formed the iron and silicate cloud layers . in cool substellar atmospheres ,
na@xmath0s ( sodium sulfide ) has been predicted to play a potentially significant role lodders99 , lodders06 , visscher06 .
other species expected to condense at these lower temperatures ( roughly 600 to 1400 k ) include cr , mns , zns , and kcl . to our knowledge ,
none of these five condensates have been included in a brown dwarf atmosphere model before now .
marley00 estimated column optical depths for several of these species and recognized that na@xmath0s could be important at low @xmath1 but did not include this species in subsequent modeling because of lack of adequate optical constant data .
burrows01 , burrows02 noted that na@xmath0s and kcl will condense in cool t dwarfs , but also noted that the indices of refraction are difficult to find .
helling and collaborators helling06 also recognized that some of these species will form condensates in some cases but also did not compute model atmospheres that included this opacity source .
fortney05c noted that some of these species might be detectable in extrasolar planet transit spectra which probe a longer path length through the atmosphere .
to model cloudy t dwarf atmospheres , we modify the am01 cloud model .
this model has successfully been used to model the effects of the iron , silicate , and corundum clouds on the spectra of l dwarfs saumon08,stephens09 . here
, we do not include the opacity of iron , silicate , and corundum clouds ; based on observed trends , we assume that the opacity of these clouds becomes negligible for the early t dwarfs .
we instead include cr , mns , na@xmath0s , zns , and kcl .
the am01 approach avoids treating the highly uncertain microphysical processes that create clouds in brown dwarf and planetary atmospheres .
instead , it aims to balance the advection and diffusion of each species vapor and condensate at each layer of the atmosphere .
it balances the upward transport of vapor and condensate by turbulent mixing in the atmosphere with the downward transport of condensate by sedimentation .
this balance is achieved using the equation @xmath15 where @xmath16 is the vertical eddy diffusion coefficient , @xmath17 is the mixing ratio of condensate and vapor , @xmath18 is the mixing ratio of condensate , @xmath19 is the convective velocity scale , and @xmath3 is a parameter that describes the efficiency of sedimentation in the atmosphere .
this calculation provides the total amount of condensate at each layer of the atmosphere .
the distribution of particle sizes at each level of the atmosphere is represented by a log - normal distribution in which the modal particle size is calculated using the sedimentation flux .
a high sedimentation efficiency parameter @xmath3 results in vertically thinner clouds with larger particle sizes , whereas a lower @xmath3 results in more vertically extended clouds with smaller particles sizes . as a result
, a higher @xmath3 corresponds to optically thinner clouds and a lower @xmath3 corresponds to optically thicker clouds .
the am01 cloud model code computes the available quantity of condensible gas above the cloud base by comparing the local gas abundance ( accounting for upwards transport by mixing via @xmath20 ) to the local condensate vapor pressure @xmath21 . in cases where the formation of condensates does not proceed by homogeneous condensation
we nevertheless compute an equivalent vapor pressure curve as described in section [ cloudcondchem ] .
the cloud code is coupled to a 1d atmosphere model that calculates the pressure - temperature profile of an atmosphere in radiative - convective equilibrium .
the atmosphere models are described in mckay89 , marley96 , burrows97 , mm99 , marley02 , saumon08 , fortney08b .
this methodology has been successfully applied to modeling brown dwarfs with both cloudy and clear atmospheres marley96 , marley02 , burrows97 , saumon06 , saumon07 , leggett07a , leggett07b , mainzer07 , blake07 , cushing08 , geballe09 , stephens09 . in the atmosphere model
, the thermal radiative transfer is determined using the `` source function technique '' presented in toon89 . within this method , it is possible to include mie scattering of particles as an opacity source in each layer .
our opacity database for gases , described extensively in freedman08 , includes all the important absorbers in the atmosphere .
this opacity database includes two significant updates since freedman08 , which are described in saumon12 : a new molecular line list for ammonia yurchenko11 and an improved treatment of the pressure - induced opacity of h@xmath0 collisions richard12 .
both the cloud model and the chemical equilibrium calculations ( see section [ cc ] ) are coupled with the radiative transfer calculations and the pressure - temperature profile of the atmosphere ; this means that a converged model will have a temperature structure that is self - consistent with the clouds and chemistry .
s index of refraction .
the real and imaginary parts of the sodium sulfide index of refraction from the two sources used are plotted as a function of wavelength .
montaner79 observational data are shown as a blue dashed line .
khachai09 calculations are shown as a pink dashed line .
the interpolated values used for the mie scattering calculation are shown as pink circles .
, width=336 ] s index of refraction .
the real and imaginary parts of the sodium sulfide index of refraction from the two sources used are plotted as a function of wavelength .
montaner79 observational data are shown as a blue dashed line .
khachai09 calculations are shown as a pink dashed line .
the interpolated values used for the mie scattering calculation are shown as pink circles .
, width=336 ] we calculate the effect of the model cloud distribution on the flux using mie scattering theory to describe the cloud opacity . assuming that particles are spherical and homogeneous , we calculate the scattering and absorption coefficients of each species for each of the particle sizes within the model . in order to make these scattering calculations ,
we need to understand the optical properties ( the real and imaginary parts of the index of refraction ) of each material .
the optical properties were found from a variety of diverse sources , summarized in table [ indexofref ] .
to calculate mie scattering within the model atmosphere , we use a grid of optical properties at wavelengths from 0.268 to 227 . where data were not available , we extrapolated the available data , following trends for similar known molecules .
the molecules with the most complete published optical properties are zns and kcl , both of which are obtained from querry87 , who tabulates the optical constants for 24 different minerals .
optical properties for cr are published in stashchuk84 from 0.26 to 15 .
the optical properties from 15 to 227 were linearly extrapolated from these experimental data following the trend of the optical properties of fe .
various extrapolations were tested ; the choice of optical properties beyond 15 does not change the results of the calculations in any meaningful way .
optical properties for mns are published in huffman67 , from 0.09 to 13 .
optical properties from 15 to 227 are extrapolated , following the trends of the other two studied sulfide condensates zns and na@xmath0s .
the optical properties for na@xmath0s , the clouds with the largest optical depth , are combined from two different sources .
montaner79 provides experimental data in the infrared , from 25 to 198 .
khachai09 provides first principles calculations of the optical properties from 0.03 to 91 . in the region of overlap ,
the montaner79 laboratory values are used .
the real and imaginary parts of the index of refraction are plotted in figure [ na2s_n ] .
lll kcl & querry87 & 0.22 - 167 + zns & querry87 & 0.22 - 167 + mns & huffman67 & 0.09 - 13 + cr & stashchuk84 & 0.26 - 15 + na@xmath0s & montaner79 & 25 - 198 + & khachai09 & 0.03 - 91 + [ indexofref ] the abundances of molecular , atomic , and ionic species are calculated using thermochemical equilibrium following the models of fegley94 , fegley96 , lodders99 , lodders02,lodders02b , lodders03 , lodders06 , lodders09 .
we adopt solar - composition elemental abundances from lodders03 .
the differences between lodders03 and newer abundance measurements [ e.g.][]asplund09 are not large enough to significantly alter the condensation temperatures considered in the paper .
lodders03 abundances were therefore selected for consistency with previous modeling efforts by our groups .
the abundances of condensate - forming elements are listed in table 2 .
we assume uniform heavy element abundance ratios over a range of metallicities from [ fe / h ] = -0.5 to [ fe / h ] = + 0.5 in order to explore the metallicity dependence of the condensation temperature expressions .
a simplified equilibrium condensation approach is used to calculate saturation vapor pressures and condensation curves ( see figure 3 ) for cr , mns , na@xmath22s , zns , and kcl as a function of pressure , temperature , and metallicity , based upon the more comprehensive thermochemical models of lodders & fegley ( see [ gaschem ] ) and visscher06,visscher10 . in each case
, we consider condensation from the most abundant cr- , mn- , na- , zn- , and k - bearing gas phases at the cloud base as predicted by the chemical models .
the relative mass of each cloud ( relative to na@xmath22s ) is listed in table 3 , assuming complete removal of available condensate material from the gas phase .
lll fe & @xmath23 & fe + si & @xmath24 & mg@xmath0sio@xmath7 , mgsio@xmath6 + mg & @xmath25 & mg@xmath0sio@xmath7 , mgsio@xmath6 + o & @xmath26 & mg@xmath0sio@xmath7 , mgsio@xmath6 , al@xmath0o@xmath6 , h@xmath0o + al & @xmath27 & al@xmath0o@xmath6 , caal@xmath28o@xmath29 , + & & caal@xmath0o@xmath7 , ca@xmath0al@xmath0sio@xmath30 + na & @xmath31 & na@xmath0s + zn & @xmath32 & zns + mn & @xmath33 & mns + s & @xmath34 & na@xmath0s , zns , mns + cr & @xmath35 & cr + k & @xmath36 & kcl + cl & @xmath37 & kcl + [ abundances ] chromium metal is the most refractory of the clouds considered here and condenses from monatomic cr gas via the reaction @xmath38 where ` ( s ) ' indicates a solid phase . the condensation condition for cr - metal is defined by @xmath39 where @xmath40 is the saturation vapor pressure of cr gas in equilibrium with cr - metal and @xmath41 is the partial pressure of cr below the cloud for a solar - composition gas ( @xmath42 , where @xmath43 is the mole fraction abundance of cr and @xmath44 is the total atmospheric pressure ) . upon condensation ,
the thermodynamic activity of cr - metal is unity and the equilibrium constant ( @xmath45 ) expression for reaction ( [ rxn : cr ] ) can be written as @xmath46 substituting for the temperature - dependent value of @xmath47 , the saturation vapor pressure of cr above the cloud base can be estimated using the expression @xmath48 for @xmath49 in kelvin and @xmath50 in bars . below the cloud , we assume that cr gas is approximately representative of the elemental cr abundance in solar composition gas ( see table [ cloudstable ] ) : @xmath51.\ ] ] the condensation temperature as a function of the total atmospheric pressure ( @xmath44 ) and metallicity can therefore be approximated by setting @xmath52 and rearranging to give @xmath53.\ ] ] this expression yields a condensation temperature near @xmath54 k at 1 bar and solar metallicity [ cf.][]lodders06 , and shows that greater total pressures and/or metallicities will lead to higher condensation temperatures .
condensation of cr - metal effectively removes gas - phase chromium from the atmosphere , and the abundances of cr - bearing gases rapidly decrease with altitude above the cloud .
our modeling of sulfide condensation chemistry follows that of visscher06 , and the condensation reactions and temperature - dependent expressions presented here are taken from that study .
the deepest sulfide cloud expected in brown dwarf atmospheres is mns , which forms via the reaction @xmath55 the formation of the mns cloud is limited by the total manganese abundance , which is 2% of the sulfur abundance in a solar - composition gas .
the condensation curve for mns is thus derived by exploring the chemistry of monatomic mn , which is the dominant mn - bearing gas near the cloud base . using results from visscher06 , the saturation vapor pressure of mn above the cloud
is given by @xmath56,\ ] ] where the metallicity dependence comes from h@xmath22s ( the dominant s - bearing gas ) remaining in the gas phase above the mns cloud base . by setting @xmath57 ,
the mns condensation curve is approximated by visscher06 : @xmath58,\ ] ] giving a condensation temperature near @xmath59 k at 1 bar in a solar - metallicity gas .
the na@xmath22s cloud is the most massive of the metal sulfide clouds expected to form in brown dwarf atmospheres because na is more abundant than either mn or zn in a solar - composition gas ( see table [ cloudstable ] ) .
sodium sulfide condenses via the net thermochemical reaction @xmath60 the mass of the na@xmath22s cloud is limited by the elemental abundance of sodium , which is 13% of the abundance of sulfur in a solar composition gas . using results from visscher06 , the saturation vapor pressure of na above the cloud base
is given by @xmath61,\ ] ] where the metallicity dependence results from h@xmath22s remaining in the gas phase above the na@xmath22s cloud and from the stoichiometry of na and h@xmath22s in the condensation reaction . the condensation temperature ( where @xmath62 ) is given by visscher06 : @xmath63,\ ] ] indicating condensation near @xmath64 k at 1 bar in a solar - metallicity gas .
the zns cloud layer forms via the reaction @xmath65 the formation of the zns cloud is limited by the total zn abundance , which is 0.3% of the s abundance in a solar - composition gas . using results from visscher06 , the saturation vapor pressure of zn over condensed zns
is given by @xmath66\ ] ] the condensation curve ( where @xmath67 ) is approximated by visscher06 : @xmath68,\ ] ] giving a condensation temperature of @xmath69 k at 1 bar in a solar - metallicity gas .
our treatment of kcl condensation chemistry is similar to that for cr - metal and the metal sulfides . with decreasing temperatures
, kcl replaces neutral k as the dominant k - bearing gas in brown dwarf atmospheres lodders99,lodders06 .
the kcl cloud layer is thus expected to form via the net thermochemical reaction @xmath70 and condenses as a solid over the range of conditions considered here . the vapor pressure of kcl above condensed kcl(s ) is given by @xmath71 derived from the equilibrium constant expression for the condensation reaction .
the mass of the kcl cloud is limited by the total potassium abundance , which is 70% of the chlorine abundance in a solar - composition gas lodders03 .
note that other k - bearing species may remain relatively abundant near cloud condensation temperatures , particularly at higher pressures ( e.g. , see @xcite and @xcite for a more detailed discussion of chemical speciation ) . however , kcl is the dominant k - bearing gas near the cloud base for the relevant @xmath72 conditions expected in cool brown dwarf atmospheres ( see figure 3 ) over the range of metallicities ( -0.5 to + 0.5 dex ) considered here . for simplicity
we therefore assume that kcl is approximately representative of the elemental k abundance below the cloud , given by @xmath73.\ ] ] the condensation temperature as a function of pressure and metallicity is estimated by setting @xmath74 and rearranging to give @xmath75,\ ] ] yielding a condensation temperature near @xmath76 k at 1 bar in a solar - metallicity gas [ cf.][]lodders99,lodders06 . in general , the condensation curve expressions demonstrate that condensation temperatures increase with total pressure , as illustrated in figure 4 .
furthermore , higher metallicities are expected to result in higher condensation temperatures and more massive cloud layers in brown dwarf atmospheres . in each case , the saturation vapor pressures of cloud - forming species rapidly decrease with altitude above the cloud layers .
llc cr & @xmath77 & 0.30 + mns & @xmath78 & 0.36 + na@xmath22s & @xmath79 & @xmath80 + zns & @xmath81 & 0.05 + kcl & @xmath82 & 0.12 + fe & @xmath83 & 20.85 + [ cloudstable ] the am01 model is one method of several that have been applied to cloudy brown dwarf atmospheres .
helling08 review various cloud modeling techniques and compare model predictions for various cases .
the most important conceptual differences between these approaches lies in the assumptions of how condensed phases interact with the gas . in the chemical equilibrium approach [ e.g.][]allard01 , condensed phases remain in contact with the gas phase and
can continue to react with the gas even at temperatures well below the condensation temperature . as an example , when following this approach , fe grains which first condense at temperatures of over 2000 k react with atmospheric @xmath84 to form fes below 1000 k. in the condensation chemistry approach we employ here , the condensed phases are assumed to sediment out of the atmosphere and are not available to interact with gas phases at temperatures below the condensation temperature .
thus fe grains form a discrete cloud layer and do not react to form fes . @xmath84
consequently remains in the gas phase and reacts to form condensates as outlined in section 2.4.2 .
jupiter is an excellent example of the applicability of this framework , as the presence of h@xmath0s in the observable atmosphere is only possible because fe is sequestered in a deep cloud layer , which prevents the formation of fes which otherwise deplete other gas phase s species fegley94 .
the presence of alkali absorption in t dwarfs likewise demonstrates the applicability of condensation chemistry @xcite .
a detailed comparison of true equilibrium condensation and cloud condensate removal from equilibrium can be found in fegley94 , lodders06 and references therein .
a different approach is taken by helling & woitke helling06 who follow the trajectory of tiny seed particles of tio@xmath0 that are assumed to be emplaced high in the atmosphere and sink downwards .
as the seeds fall through the atmosphere they collect condensate material . in helling06 and
numerous follow on papers helling08 , witte09 , witte11 , dekok11 this group models the microphysics of grain growth given these conditions . because the background atmosphere is not depleted of gaseous species until the grains fall through the atmosphere , a compositionally very different set of grains
are formed . in particular
they predict ` dirty ' grains composed of layers of varying condensates .
a direct comparison between the predictions of the various cloud modeling schools is often difficult because of differing assumptions of elemental abundances and the background thermal profile .
modeling tests in which predictions of the various groups are compared to data would be illuminating , but this is far beyond the scope of the work reported here . in order to calculate absolute magnitudes of the modeled brown dwarfs , we use the results of evolution models which determine the radius of a brown dwarf as it cools and contracts over its lifetime .
we use the evolution models of saumon08 with the surface boundary condition from cloudless atmospheres . using a cloudless boundary condition instead of one consistent with these clouds changes the calculated magnitudes of the models very slightly , but does not change the overall trends or results . to analyze the effect of these clouds ,
we generate a grid of 182 model atmospheres at effective temperatures and surface gravities spanning the full range of t dwarfs .
we calculate pressure - temperature profiles and synthetic spectra for atmospheres from 400 to 1300 k ( 50 to 100 k increments ) , with log(@xmath2 ) ( cgs ) of 4.0 , 4.5 , 5.0 , and 5.5 and cloud sedimentation efficiency parameter @xmath3=2 , 3 , 4 , and 5 . for this study
, we use only solar metallicity composition .
we then compare these , both photometrically and spectroscopically , to observed t dwarfs .
= 4.5 ; for the hotter models , log @xmath2=5.0 . we show cloudless models in blue , and cloudy models with @xmath3=2 ( red ) and 4 ( orange ) .
condensation curves for each condensate species are plotted .
the cloudy models include the condensates cr , mns , na@xmath0s , zns , and kcl .
note that for each case , increasing the cloud thickness increases the temperature at a given atmospheric pressure .
the 1 - 6 photosphere of each model is shown as a thicker line .
, width=336 ] in figure [ pt - results ] , we show the pressure - temperature profiles of models with effective temperatures of 400 k , 600 k , 900 k , and 1300 k. the surface gravity of the 400 k models is log @xmath2=4.5 ; for the hotter models , log @xmath2=5.0 .
we plot models with two different cloud sedimentation efficiencies which include only the na@xmath0s , mns , zns , cr , and kcl clouds . because na@xmath0s and mns are by far the most dominant cloud species ( see section [ cloudstructures ] ) , as a shorthand we refer to this collection of clouds as ` sulfide clouds . '
the condensation curves of major and minor species predicted to form by equilibrium chemistry are also plotted .
the location of a given cloud base is expected to be where the pressure - temperature profile of the model atmosphere crosses the condensation curve .
each model crosses the condensation curve of each species at very different pressures and , to a lesser extent , temperatures , so we expect that the significance of the clouds will be strongly controlled by the effective temperature of the model .
the cloudy 400 k and 600 k models have two convection zones .
all 900 and 1300 k models have a single deep convection zone . for all models ,
it is clear that , as in previous cloudy models of l dwarfs , adding cloud opacity has a `` blanketing '' effect on the model , increasing the temperature of the atmosphere for a given atmospheric pressure .
as the cloud becomes optically thicker , the entire pressure - temperature profile becomes hotter ; thus , on a plot of pressure - temperature profiles such as figure [ pt - results ] , increasing the cloud opacity moves the whole profile to the right .
= 1300 k , log @xmath2=5.0 ; @xmath1=900 k , log @xmath2=5.0 ; @xmath1=600 k , log @xmath2=5.0 ; @xmath1=400 k , log @xmath2=4.5 .
we show cloudy models with @xmath3=2 and 4 which include the condensates cr , mns , na@xmath0s , zns , and kcl and cloudless models for comparison .
note that for the @xmath1=400 k , @xmath1=600 k and @xmath1=900 k models , the cloudy models are progressively fainter in @xmath85 and @xmath12 bands and brighter in @xmath11 band as the sedimentation efficiency decreases .
in contrast , for the @xmath1=1300 k case , the clouds do not significantly change the spectrum .
, width=336 ] = 1300 k , log @xmath2=5.0 ; @xmath1=900 k , log @xmath2=5.0 ; @xmath1=600 k , log @xmath2=5.0 ; @xmath1=400 k , log @xmath2=4.5 .
we show cloudy models with @xmath3=2 and 4 which include the condensates cr , mns , na@xmath0s , zns , and kcl and cloudless models for comparison .
note that for the @xmath1=400 k , @xmath1=600 k and @xmath1=900 k models , the cloudy models are progressively fainter in @xmath85 and @xmath12 bands and brighter in @xmath11 band as the sedimentation efficiency decreases .
in contrast , for the @xmath1=1300 k case , the clouds do not significantly change the spectrum . , width=336 ] = 1300 k , log @xmath2=5.0 ; @xmath1=900 k , log @xmath2=5.0
; @xmath1=600 k , log @xmath2=5.0 ; @xmath1=400 k , log @xmath2=4.5 .
we show cloudy models with @xmath3=2 and 4 which include the condensates cr , mns , na@xmath0s , zns , and kcl and cloudless models for comparison .
note that for the @xmath1=400 k , @xmath1=600 k and @xmath1=900 k models , the cloudy models are progressively fainter in @xmath85 and @xmath12 bands and brighter in @xmath11 band as the sedimentation efficiency decreases .
in contrast , for the @xmath1=1300 k case , the clouds do not significantly change the spectrum .
, width=336 ] = 1300 k , log @xmath2=5.0 ; @xmath1=900 k , log @xmath2=5.0 ; @xmath1=600 k , log @xmath2=5.0 ; @xmath1=400 k , log @xmath2=4.5 .
we show cloudy models with @xmath3=2 and 4 which include the condensates cr , mns , na@xmath0s , zns , and kcl and cloudless models for comparison .
note that for the @xmath1=400 k , @xmath1=600 k and @xmath1=900 k models , the cloudy models are progressively fainter in @xmath85 and @xmath12 bands and brighter in @xmath11 band as the sedimentation efficiency decreases . in contrast , for the @xmath1=1300 k case , the clouds do not significantly change the spectrum .
, width=336 ] , from top to bottom , @xmath1=1300 k , log @xmath2=5.0 ; @xmath1=900 k , log @xmath2=5.0 ; @xmath1=600 k , log @xmath2=5.0 ; @xmath1=400 k , log @xmath2=4.5 .
we show cloudy models with iron / silicate / corundum clouds ( no sulfide clouds ) with @xmath3=2 and cloudless models for comparison .
note that these clouds , unlike the sulfide clouds in figure [ specs - results ] , significantly change the shape of the 1300 k model .
, width=336 ] , from top to bottom , @xmath1=1300 k , log @xmath2=5.0 ; @xmath1=900 k , log @xmath2=5.0 ; @xmath1=600 k , log @xmath2=5.0 ; @xmath1=400 k , log @xmath2=4.5 .
we show cloudy models with iron / silicate / corundum clouds ( no sulfide clouds ) with @xmath3=2 and cloudless models for comparison .
note that these clouds , unlike the sulfide clouds in figure [ specs - results ] , significantly change the shape of the 1300 k model .
, width=336 ] , from top to bottom , @xmath1=1300 k , log @xmath2=5.0 ; @xmath1=900 k , log @xmath2=5.0 ; @xmath1=600 k , log @xmath2=5.0 ; @xmath1=400 k , log @xmath2=4.5 .
we show cloudy models with iron / silicate / corundum clouds ( no sulfide clouds ) with @xmath3=2 and cloudless models for comparison .
note that these clouds , unlike the sulfide clouds in figure [ specs - results ] , significantly change the shape of the 1300 k model .
, width=336 ] , from top to bottom , @xmath1=1300 k , log @xmath2=5.0 ; @xmath1=900 k , log @xmath2=5.0 ; @xmath1=600 k , log @xmath2=5.0 ; @xmath1=400 k , log @xmath2=4.5 .
we show cloudy models with iron / silicate / corundum clouds ( no sulfide clouds ) with @xmath3=2 and cloudless models for comparison .
note that these clouds , unlike the sulfide clouds in figure [ specs - results ] , significantly change the shape of the 1300 k model .
, width=336 ] in figures [ bt ] and [ specs - results ] , we show the spectra of the same example models at 1300 , 900 , 600 , and 400 k : figure [ bt ] shows the wavelength - dependent brightness temperatures from these models , while [ specs - results ] shows the model fluxes computed from the top of the atmosphere .
the brightness temperature gives some insight into the depth into the atmosphere probed at each wavelength .
flux from 0.8 to 1.3 comes from the deepest , hottest layers of the atmosphere .
clouds change the flux in this wavelength range by limiting the depth from which flux emerges ; conversely , the clouds do not change the depth probed between @xmath42 and 5 because the clouds form below the layers from which most of the flux is emerging .
however , the hotter atmospheric temperatures at a given pressure ( see figure 3 ) lead to slightly higher fluxes at these wavelengths .
though not plotted here , flux in the mid - infrared also comes from above the cloud layer and is slightly higher because the entire pressure - temperature profile is hotter .
for the hottest of these models ( @xmath1=1300 k ) , the cloudy spectra look almost identical to the cloudless spectrum .
this model is too hot to have much mass of these condensed species form in the photosphere . for a cooler model ( @xmath1=900 k ) ,
the cloudy spectra look different from the cloudless spectrum .
as we decrease the sedimentation efficiency @xmath3 in the model , the flux in @xmath85 and @xmath12 bands decreases and the flux in @xmath11 band increases . for the coldest two models shown ( @xmath1=400 , 600 k ) ,
the cloudy spectra look dramatically different from the cloudless spectrum in the near - infrared ; even the thinnest cloud considered here ( @xmath3=5 ) causes the flux in @xmath85 and @xmath12 to decrease by 50% and the flux in @xmath11 to correspondingly increase .
decreasing the sedimentation efficiency enhances this effect .
figure [ specs - silicate ] shows the effect of the iron and silicate clouds on the spectra of models with the same effective temperatures ( 1300 k , 900 k , 600 , and 400 k ) and surface gravity . note that unlike the sulfide clouds , these iron and silicate clouds substantially change the shape of the 1300 k and 900 k models by suppressing the flux in @xmath85 , @xmath12 , and @xmath13 and increasing the flux in @xmath11 band and the mid - infrared
this strong effect at higher temperatures is due to the large amount of iron and silicate condensed in the visible atmosphere at those temperatures .
figure [ colopd ] shows the distribution of clouds in the model atmospheres for the three example cases .
the locations of iron , silicate , and corundum clouds , using models that only include those clouds
the standard saumon08 cloud configuration are plotted for reference .
the column optical depth is given by equation 16 in am01 , which calculates the cumulative geometric scattering optical depth by cloud particles through the atmosphere .
for the 1300 k model , all of the sulfide clouds have tiny optical depths in the photosphere and do not affect the emergent spectra .
the silicate and iron clouds would have significant optical depth ( @xmath86=2 - 3 ) and would substantially change the emergent spectra .
s , zns , kcl , cr , and mns .
the dashed lines show the column optical depths of models that include only the more refractory clouds corundum ( al@xmath0o@xmath6 ) iron ( fe ) , and forsterite ( mg@xmath0sio@xmath7 ) to show where those clouds would form in comparison to the sulfide clouds .
all models use @xmath3=2 .
the shaded grey area shows the region of the atmosphere which lies within the @xmath87 1 to 6 photosphere .
note that the na@xmath0s cloud is by far the most important of the added clouds for the 600 k model in the near - infrared .
also note that if the al@xmath0o@xmath6 , fe , and mg@xmath0sio@xmath7 persisted to effective temperatures of 900 - 1300 k , they would be quite visible , which would not match observations .
, width=319 ] s , zns , kcl , cr , and mns .
the dashed lines show the column optical depths of models that include only the more refractory clouds corundum ( al@xmath0o@xmath6 ) iron ( fe ) , and forsterite ( mg@xmath0sio@xmath7 ) to show where those clouds would form in comparison to the sulfide clouds .
all models use @xmath3=2 .
the shaded grey area shows the region of the atmosphere which lies within the @xmath87 1 to 6 photosphere .
note that the na@xmath0s cloud is by far the most important of the added clouds for the 600 k model in the near - infrared .
also note that if the al@xmath0o@xmath6 , fe , and mg@xmath0sio@xmath7 persisted to effective temperatures of 900 - 1300 k , they would be quite visible , which would not match observations .
, width=319 ] s , zns , kcl , cr , and mns .
the dashed lines show the column optical depths of models that include only the more refractory clouds corundum ( al@xmath0o@xmath6 ) iron ( fe ) , and forsterite ( mg@xmath0sio@xmath7 ) to show where those clouds would form in comparison to the sulfide clouds .
all models use @xmath3=2 .
the shaded grey area shows the region of the atmosphere which lies within the @xmath87 1 to 6 photosphere .
note that the na@xmath0s cloud is by far the most important of the added clouds for the 600 k model in the near - infrared .
also note that if the al@xmath0o@xmath6 , fe , and mg@xmath0sio@xmath7 persisted to effective temperatures of 900 - 1300 k , they would be quite visible , which would not match observations .
, width=319 ] s , zns , kcl , cr , and mns .
the dashed lines show the column optical depths of models that include only the more refractory clouds corundum ( al@xmath0o@xmath6 ) iron ( fe ) , and forsterite ( mg@xmath0sio@xmath7 ) to show where those clouds would form in comparison to the sulfide clouds .
all models use @xmath3=2 .
the shaded grey area shows the region of the atmosphere which lies within the @xmath87 1 to 6 photosphere .
note that the na@xmath0s cloud is by far the most important of the added clouds for the 600 k model in the near - infrared .
also note that if the al@xmath0o@xmath6 , fe , and mg@xmath0sio@xmath7 persisted to effective temperatures of 900 - 1300 k , they would be quite visible , which would not match observations .
, width=319 ] s cloud above the bottom of the 1 - 6 photosphere in each model is plotted as a function of model effective temperature .
the curves connecting the points are there to guide the eye .
three different surface gravities are shown and all models use @xmath3=2 . the column optical depth peaks at temperatures of about 600 k , and models with higher surface gravity have a greater na@xmath0s column optical depth . , width=336 ] for the 900 k model ,
all of the sulfide clouds have a column optical depth smaller than 1 in the photosphere .
kcl and zns have tiny optical depth ( @xmath88 ) and will not create an observable change in the spectrum .
na@xmath0s and mns have optical depth between 0.1 and 1 and will change the model spectra slightly . for the 600 k model , na@xmath0s is the most important condensate opacity source .
kcl has a small optical depth and zns has a negligible optical depth .
this result is expected , based on the abundances of each species ( see table [ cloudstable ] ) .
the other two clouds , mns and cr , are below the near - infrared photosphere , so also do not change the spectrum .
the silicate and iron clouds would also be below the photosphere .
using our full grid of models , we can examine the importance of the na@xmath0s cloud as a function of @xmath1 and surface gravity .
figure [ maxtau ] shows how the column optical depth of this cloud varies from 400 - 1300 k and log @xmath2 from 4.0 - 5.0 for a constant value of the parameter @xmath3 .
moving to @xmath1 values below 1300 k , the na@xmath0s cloud grows in importance as it forms progressively deeper in the atmosphere , so that there is larger mass of condensate in the cloud .
the maximum optical depth of the na@xmath0s cloud at pressure levels above the bottom of the 1 - 6 photosphere is largest for models with @xmath1 of 600 k. at lower @xmath1 , much of the cloud opacity is below the visible atmosphere .
the optical depth within the photosphere is significant ( @xmath89 1 ) for models between 400 and 700 k. figure [ maxtau ] indicates that the cloud optical depth within the photosphere is largest for higher surface gravity atmospheres for a constant value of @xmath3 .
however , this does not necessarily predict that higher gravity brown dwarfs will have thicker clouds than lower gravity brown dwarfs because the parameter @xmath3 is not necessarily independent of gravity .
in figure [ colormag - results ] , we plot the photometric colors in the near - infrared of all brown dwarfs with measured parallaxes and apparent @xmath12 and @xmath11 magnitude errors smaller than 0.2 magnitudes dupuy12 , faherty12 .
we also plot the calculated photometric colors of our suite of cloudless and cloudy models from 400 - 1300 k with several representative surface gravities . in section [ generaltrends ]
we discuss the general trends of the photometric colors of models at various effective temperatures and cloud sedimentation efficiencies . in section [ comparison ]
we compare our model results to the photometric observations .
as discussed in sections [ modelspecs ] and [ cloudstructures ] , clouds in t dwarf atmospheres tend to suppress the flux in @xmath85 and @xmath12 bands and increase the flux in @xmath11 band .
the flux shift from @xmath12 to @xmath11 gives cloudy models larger ( redder ) @xmath8 and @xmath10 colors than cloudless models . in figure [ colormag - results
] , the hottest cloudy models have nearly the same near - infrared colors as cloudless models .
as we decrease the effective temperature of a cloudy model , more cloud material condenses ; the model has a redder photometric color than a cloudless model with the same @xmath1 .
if we reduce the sedimentation efficiency ( @xmath90 ) of the cloud , the cloud becomes optically thicker , and the model has a redder photometric color .
the upper panels , which show @xmath8 photometric colors , show that our sulfide cloud models can easily reach the colors of red t dwarfs , with @xmath3 values of 4 - 5
. the bulk of the t dwarf population is bluer than the @xmath3=5 model .
however , the cooler t dwarfs are generally well - matched by the models . in @xmath10 , the color directly affected by cloud opacity limiting the depth to which one sees , the model colors are an excellent match to the data .
the cloudy models are a much better match than the corresponding cloud - free models . based on saumon08 evolution models and assuming that observed brown dwarfs will have ages less than 10 gyr
, we expect that the coldest objects modeled , between 400 - 600 k , will have surface gravities less than log @xmath2=5.2 .
hotter objects , between 1000 and 1300 k , will have surface gravities less than log @xmath2=5.5 . for the l dwarfs
, we are generally able to match photometric colors by including silicate , iron , and corundum clouds with a sedimentation efficiency parameter of @xmath3=2@xmath911 stephens09 , saumon08 . however , for these cooler t dwarfs , models with these sulfide clouds with @xmath3=2 are redder than observed brown dwarfs . if we assume @xmath3 is larger around 4 or 5we are able to match observed colors . the cloud model does not explicitly suggest any physical mechanism for why @xmath3 would be different .
however , since these objects are about 1000 k cooler than l dwarfs , it would not be surprising if these objects populate a different physical regime , and would have substantially different rates of atmospheric mixing and cloud condensation . indeed ,
a large increase in @xmath3 with lower @xmath1 values is one way to quickly clear away the silicate and iron clouds knapp04 cushing11 announced the discovery of six proposed y dwarfs found by the wise mission .
objects around the t - to - y transition are cold enough to have nh@xmath6 absorption features in the near infrared .
we have obtained near - infrared photometry for two proposed y0 dwarfs discovered using the wide - field infrared survey explorer ( wise ) by cushing et al .
@xmath92 was obtained for wisep [email protected] and @xmath94 for [email protected] , using the near - infrared imager niri hodapp03 on the gemini - north telescope on mauna kea , hawaii .
the photometry is on the mauna kea observatories system tokunaga05.the data were obtained via the gemini queue program gn-2012a - q-106 on 2012 february 10 ; the program aims to supplement and improve on the photometry presented in the discovery paper .
individual exposure times of 60 s were used at @xmath85 and @xmath12 , and 30 s at @xmath13 ; a 9-position dither pattern with 10 " offsets was repeated as necessary for sufficient signal to noise .
the total exposure time for wisep j140518.40 + 553421.5 was 9 minutes at @xmath85 and @xmath12 and 58.5 minutes at @xmath13 ; for wisep j154151.65 - 225025.2 the total exposure time was 18 minutes at each of @xmath85 and @xmath12 .
data were reduced in a standard fashion and flatfielded with calibration lamps on the telescope .
the ukirt faint standards fs 133 and 136 were used for photometric calibration ; @xmath12 and @xmath13 were taken from leggett06 , and @xmath85 from the ukirt online catalog ( http://www.jach.hawaii.edu/ukirt/astronomy/calib/phot cal / fs zy mko wfcam.dat ) .
the final reduced magnitudes are : @xmath96 , @xmath97 and @xmath98 for wisep j140518.40 + 553421.5 ; @xmath99 and @xmath100 for wisep j154151.65 - 225025.2 .
figure [ wise ] shows how clouds will effect these cold objects .
@xmath101 is a useful temperature indicator for these objects , while @xmath10 is sensitive to both the cloud structure and gravity .
as the @xmath1 of the non cloudy models decreases from 800 to 500 k , the models become progressively bluer in @xmath10 color . however , most of the proposed wise y dwarfs are redder than the cloudless model .
the models that include the sulfide clouds match their colors better .
this result tentatively suggests that for objects colder than t dwarfs , the sulfide clouds remain important .
of course , for objects with effective temperatures of @xmath4350 k , water will condense ; at that point , h@xmath0o clouds should contribute to the spectra [ e.g.][]burrows03 .
we now compare model spectra to two relatively cold , red t dwarfs , ross 458c and ugps 072205 .
the near - infrared spectra of these two objects are not well - matched by cloudless t dwarf spectra ; by including our neglected clouds , which for these cool objects are dominated by the na@xmath0s cloud , we match their spectra more accurately .
we compare models to both near - infrared spectra and near- and mid - infrared photometry . as in previous studies of brown dwarfs
cushing08 , stephens09 , we find that in different bands , the observations are best fit by models of different parameters . in this study , we focus on finding models that fit the shape of the spectra in the near - infrared where clouds play a significant role .
ross 458c is a late - type t dwarf ( t79 ) which is separated by over 1100 au from a pair of m star primaries .
it has anomalously red near - infrared colors ( @xmath102 ) .
burgasser10 obtained spectroscopic observations with the fire spectrograph simcoe08 , simcoe10 on the magellan baade 6.5 meter telescope at las campanas observatory .
they fit the spectrum using cloudless and cloudy models ( which include only the opacity of the iron , silicate , and corundum clouds ) and find that cloudy models fit significantly better than cloudless models .
burgasser10 conclude that cloud opacity is necessary to reproduce the spectral data and invoke a reemergence of the iron and silicate clouds .
we instead assume that the iron and silicate clouds are depleted , as we observe generally in other t dwarfs , and investigate the effect of the sulfide clouds . in figure [ ross458-spec ]
we show the fire spectrum and the best fitting cloudy and cloudless models .
we also show photometry in @xmath12 , @xmath13 , and @xmath11 dupuy12 , wise photometry kirkpatrick12 , and spitzer photometry burningham11 .
the spectra used to generate these results differ somewhat from those in burgasser10 because we use models that include recent improvements to the opacity database saumon12 for both ammonia and the pressure - induced opacity of h@xmath0 collisions .
all models are fit by eye to the observations . like burgasser10
, we find that clouds are essential to match the spectrum of ross 458c .
figure [ ross458-spec ] shows the best fitting cloudless model and the two best fitting cloudy models ( one including the iron and silicate clouds and the other including the sulfide clouds ) . the cloudless model is a poor representation of the spectral data
; the flux in @xmath85 and @xmath12 is too high and the flux in @xmath13 and @xmath11 is too low .
the cloudy models are better representations of the relative flux in each band .
burgasser10 found that the surface gravity of ross 458c must be low ( log @xmath2=4.0 ) for models to match the observed spectrum .
likewise , we find that our best - fitting models have surface gravities of 4.0 ( cloudless ) , 3.7 ( silicate clouds ) , and 4.0 ( sulfide clouds ) .
we conclude that we do not need to invoke a reemergence of iron and silicate clouds into the photosphere of ross 458c to reproduce the observed spectrum .
instead , we are able to reproduce the spectrum using the sulfide clouds which are naturally expected to form in the photospheres of cool t dwarfs .
section [ sulfsil ] contains additional discussion on which cloud species we expect to be important .
the very red slope to the l band spectrum of ross 458c much redder than all the models is reminiscent of the behavior of some cloudy l dwarfs , including 2mass2224 and de 0255 ( l3.5 and l9 respectively ) and may be a signature of very small dust grains stephens09 .
the discrepancies at 4.5 are likely to be a result of non - equilibrium chemistry , which is not included in these models .
this effect is discussed in more detail in section [ results - non - eq ] .
ugps j072227.51 - 054031.2 ( hereafter ugps 072205 ) is a t9 or t10 dwarf with an effective temperature of approximately 500 k , discovered by lucas10 .
previous spectral analysis with cloudless models has been unsuccessful at modeling the flux in the near - infrared in @xmath85 and @xmath12 bands leggett12 . in figure [ 0722-spec ]
we plot the near - infrared spectra published in lucas10 and leggett12 with the cloudy and cloudless models that are fit by eye to be the closest representations of the data .
we also show @xmath12 , @xmath13 , @xmath11 and spitzer photometry lucas10 and wise photometry kirkpatrick12 .
these models have similar temperatures and gravities to previous studies ; leggett12 presented fits with @xmath1 between 492 and 550 k and log g=3.52 to 5.0 , whereas our fits have @xmath1 of 600 k ( cloudless ) and 500 k ( with sulfide clouds ) and both have log @xmath2=4.5 .
note that the flux in @xmath85 and @xmath12 in cloudless models is systematically too high .
the opacity of the sulfide clouds provides a natural mechanism to decrease the flux in @xmath85 and @xmath12 and increase the flux in @xmath13 and @xmath11 bands .
the match to the models is still not perfect .
this may be due in part to incomplete line lists for methane ; these cold objects have a significant amount of ch@xmath7 in the atmosphere , which absorbs strongly in the near infrared .
the discrepancy at 4.5 is most likely due to absorption of co as a result of non - equilibrium chemistry ( see section [ results - non - eq ] ) .
outstanding issues in t dwarf modeling are discussed in section [ outstandingissues ] .
we note that both of our preferred sulfide cloud models predict brighter @xmath103 band ( 4.5 ) photometry than is observed .
this is likely a consequence of our neglect of non - equilibrium mixing of co in this study .
as first predicted by fegley96 , such mixing is an important process in the atmospheres of brown dwarfs noll97 , marley02 , saumon06 , stephens09 as it is in solar system giants barshay78 and the relative impact of the mixing increases with decreasing gravity hubeny07 , barman11 .
absorption by excess atmospheric co decreases the thermal emission in m band and is likely responsible for the mismatches seen in figures [ ross458-spec ] and [ 0722-spec ] , particularly for the lower gravity ross 458c . the formation of the clouds considered in this study does not involve the species most affected by non - equilibrium chemistry such as co and ch@xmath7 .
the cloud models will therefore be only minimally affected by the changes in the pressure - temperature profile of the atmosphere due to the changes in gas opacity .
however , the overall spectra of models will look quite different in regions where co absorbs strongly , such as the 4.5 feature , so future , more comprehensive fits of sulphide cloud models to observations will have to include non - equilibrium models .
clouds must form in brown dwarf atmospheres as they cool ; there is no way to avoid the condensation of different species as the atmosphere reaches lower effective temperatures . in these models ,
we parameterize the opacity of clouds by creating a distribution of cloud material in the atmosphere which has a distribution of cloud particle sizes . within the models
, we can change those distributions .
a cloud that sediments into a small number of large particles will settle into a thin layer and will not significantly change the emergent spectrum ; the same cloud material organized into an extended cloud with small particle sizes will have a dramatic effect on the model spectrum .
for these reasons , we require a model of cloud particle sizes and distribution as well as the underlying chemistry .
when we consider models that include new or different clouds , we do not change any of the underlying chemistry of condensation ; we change the opacity of the condensate particles and in doing so modify the effect that the cloud formation has on emergent flux .
burgasser10 invoked the reemergence of silicate clouds to explain the spectrum of ross 458c .
we suggest instead that the initial emergence of sulfide clouds would have a similar effect on the spectrum and provide a more natural explanation for the results . from observations ,
it is clear that the range in t dwarf colors just following the l / t transition is small ; spectra of t dwarfs show no evidence that clouds still affect the emergent flux for objects slightly cooler than this transition .
this observation suggests that the iron and silicate clouds have dissipated between 1400 and 1200 k ( for typical field dwarfs ) and are no longer important in t dwarf atmospheres .
if iron and silicate clouds were sometimes important in t dwarf atmospheres , we would expect to see a population of relatively quite red objects at effective temperatures between that of ross 458c and the l dwarfs ; no brown dwarf with these properties has been observed . as t dwarfs cool , the range in observed infrared colors increases ; a population of red t dwarfs develops , which are redder in the near - infrared than cloudless models predict . based on these observations ,
we favor a mechanism that can not strongly alter @xmath1@xmath4900 - 1200 k t dwarf atmospheres , but naturally reddens @xmath1@xmath9800 k t dwarfs .
the emergence of sulfide clouds provides that natural explanation for this range in t dwarf colors at low effective temperatures .
just as the iron and silicate clouds condense in the photospheres of l dwarfs and change their observed spectra , the sulfide clouds begin to condense in the photospheres of t dwarfs with temperatures cooler than 900 k , changing their observed spectra .
we have not yet investigated whether the sulfide clouds will have identifiable spectral features that would confirm their presence in t dwarf atmospheres , but given the features in the sulfide indices of refraction ( see figure [ na2s_n ] ) these features would likely be in the mid - infrared .
there are several challenges in modeling t dwarfs that have not yet been addressed in these calculations .
because of the high densities in brown dwarf atmospheres , sodium and potassium bands at optical wavelengths are extremely pressure - broadened in t dwarf spectra tsuji99 , burrows00 , allard05 , allard07 .
the wings of these broadened bands extend into the near - infrared in @xmath85 and @xmath12 bands , creating additional opacity at those wavelengths .
for these calculations , we use the line broadening treatment outlined in burrows00 , which is somewhat _ ad hoc _ and potentially creates some inaccuracies in the model flux in @xmath85 and @xmath12 bands .
a calculation of the molecular potentials for potassium and sodium in these high pressure environments , as is carried out in allard05 , allard07 , would improve the accuracy of these models .
another challenge in modeling t dwarf spectra is the inadequacies of methane opacity calculations ; methane is the only important gas - phase absorber with inadequate opacity measurements or calculations .
uncertainties in methane absorption bands create inaccuracies in t dwarf models , especially in @xmath13 band where it is a very strong absorber saumon12 .
sulfide clouds could form partial cloud layers with patchy clouds .
one way to infer patchy cloud cover is to observe variability in photometric colors ; variability can indicate high - contrast cloud features rotating in and out of view .
radigan12 studied objects at the l / t transition and inferred that the iron and silicate clouds could be in the process of breaking up and forming patchy clouds in those atmospheres .
a similar study of the variability of cool t dwarfs could reveal a similar physical process in sulfide clouds . a larger number of high fidelity spectra of the coldest t dwarfs would help to constrain these cloudy models .
currently there are a few objects with effective temperatures cooler than 700 k with moderate resolution spectra .
a larger sample of objects would give us better statistics on the overall population of t dwarfs , with different surface gravities , metallicities , and cloud structures . at cooler effective temperatures
, water clouds will condense in brown dwarf atmospheres burrows03 .
oxygen is 300 times more abundant than sodium in a solar - composition gas and the silicate clouds only use 20% of the total oxygen in the atmosphere , so water clouds will be much more massive and important in shaping the emergent flux . as missions like wise find colder objects kirkpatrick11 , cushing11 and these objects are observed spectroscopically , future models of brown dwarfs will need to include the condensation of these more volatile clouds .
before the water clouds condense , lodders99 , lodders06 predict that rbcl and cscl will condense ; however , the abundances of cs and rb are very low lodders03 so these clouds will be optically thin .
if equilibrium conditions prevail , nh@xmath7h@xmath0po@xmath7 would also condense fegley94 , visscher06 with a mass similar to that of the na@xmath0s cloud . whether nh@xmath7h@xmath0po@xmath7 condenses or p remains in the gas phase as ph@xmath6 ( as on jupiter and saturn ) deserves further study to examine potential effects on the spectra of the coolest brown dwarfs .
observations and models of t dwarfs provide a testbed to study planetary atmospheres . while brown dwarfs are more massive than planets , the atmospheres of t dwarfs have similar effective temperatures to those of young giant planets burrows97,fortney08b .
the study of t dwarfs provides crucial tests of cloudy atmosphere models that will be applicable to directly - imaged exoplanet atmospheres .
cloud models designed originally for brown dwarfs are already being applied to exoplanets .
cloud models with iron and silicate clouds were originally developed to model l dwarf atmospheres ; these models have been successfully applied to observations of the only directly imaged multiple planet system , hr 8799 .
several studies of the hr 8799 planets have shown that the iron and silicate clouds play a significant role in their atmospheres marois08 , barman11 , galicher11 , bowler10 , currie11 , madhu11c , marley12 .
as instruments like the gemini planet imager and sphere begin to discover new planets in the next few years , we will be able to apply brown dwarf models to colder planetary atmospheres in which clouds will likely play a key role in shaping their spectra .
cloud formation is a natural and unavoidable phenomenon in cool substellar atmospheres . at temperatures cooler than those of l dwarfs
, chemistry dictates that additional condensates , beyond the silicates and iron , must form .
we have examined the effect of including the most abundant of these lower - temperature condensates , cr , mns , na@xmath0s , zns , and kcl , in brown dwarf model atmospheres . within the framework of the am01 cloud model
, we have investigated the opacity of these clouds over a wide range of parameter space , across the relevant range of t dwarfs , to the warmest y dwarfs . from our suite of models from 400 to 1300 k , log @xmath2=4.0 to 5.5 , @xmath3=2 to 5 ,
we have shown the likely role that these low-@xmath1 clouds , dominated by sulfides , play in these cool atmospheres .
model spectra were compared to two t dwarfs , ross 458c and ugps 072205 .
these two objects have red near - infrared colors and are not well - matched by cloudless models .
model spectra that include the sulfide clouds match the observed spectra of both objects more accurately than cloudless models . the photometric colors of the cloudy models were calculated and compared to the full population of brown dwarfs with parallax measurements .
this analysis shows that the sulfide clouds provide a mechanism to match the near - infrared colors of observed brown dwarfs .
the agreement is particularly good in @xmath10 , while in @xmath8 the models are somewhat too red .
wise observations of the coolest t dwarfs and warmest y dwarfs indicate the these new models fit observations better than cloud - free models .
we thank rabah khenata and the rest of his team in algeria for providing calculations of optical properties for sodium sulfide .
we thank adam burgasser for providing the spectrum for ross 458c .
we thank katharina lodders for providing the condensation chemistry tables used in the model calculations as well as other very helpful advice and suggestions .
we also thank the anonymous referee for his or her suggestions .
based on observations obtained at the gemini observatory , which is operated by the association of universities for research in astronomy , inc .
, under a cooperative agreement with the nsf on behalf of the gemini partnership : the national science foundation ( united states ) , the science and technology facilities council ( united kingdom ) , the national research council ( canada ) , conicyt ( chile ) , the australian research council australia),ministrio da cincia , tecnologia e inovao ( brazil ) and ministerio de ciencia , tecnologa e innovacin productiva ( argentina ) . | as brown dwarfs cool , a variety of species condense in their atmospheres , forming clouds .
iron and silicate clouds shape the emergent spectra of l dwarfs , but these clouds dissipate at the l / t transition .
a variety of other condensates are expected to form in cooler t dwarf atmospheres .
these include cr , mns , na@xmath0s , zns , and kcl , but the opacity of these optically thinner clouds has not been included in previous atmosphere models . here
, we examine their effect on model t and y dwarf atmospheres . the cloud structures and opacities
are calculated using the am01 cloud model , which is coupled to an atmosphere model to produce atmospheric pressure - temperature profiles in radiative - convective equilibrium .
we generate a suite of models between @xmath1 = 400 and 1300 k , log @xmath2=4.0 and 5.5 , and condensate sedimentation efficiencies from @xmath3=2 to 5 .
model spectra are compared to two red t dwarfs , ross 458c and ugps 072205 ; models that include clouds are found to match observed spectra significantly better than cloudless models .
the emergence of sulfide clouds in cool atmospheres , particularly na@xmath0s , may be a more natural explanation for the `` cloudy '' spectra of these objects , rather than the re - emergence of silicate clouds that wane at the l - to - t transition .
we find that sulfide clouds provide a mechanism to match the near- and mid - infrared colors of observed t dwarfs .
our results indicate that including the opacity of condensates in t dwarf atmospheres is necessary to accurately determine the physical characteristics of many of the observed objects . |
the subject of this note is illustrated in fig . [
fig : phase ] , where i have sketched the phase diagram of qcd in the temperature , baryochemical potential , isospin chemical potential space .
lattice calculations aim at a quantitative analysis of the qcd phase diagram , using the qcd lagrangian as a sole input .
i will review the progress towards this goal since qm2006@xcite examining the various thermodynamic regions indicated in fig.[fig : phase ]
. plane .
mature calculations are possible around @xmath0 , for small @xmath1 , in the region marked ( * 1 * ) in the plot , and in the domain of the strongly interactive qgp ( * 2 * ) .
progress is being made in the parameter range of potential interest to fair ( * 3 * ) , while the physics of cold , dense matter ( * 4 * ) can be explored only with a nonzero isospin density , or by use of two - color qcd.,scaledwidth=40.0% ]
a critical point in the t,@xmath1 plane has been predicted by model calculations , and several lattice studies have searched for it , without finding a general agreement , see e.g. @xcite for a review . in ref.@xcite
an alternative strategy was proposed , based on the analysis of the slope @xmath2 of the critical surfaces in the @xmath3 space stemming from the critical line which limits the first order transition area of the @xmath4 plane : @xmath5 a positive slope would be indicative of a critical endpoint .
@xmath2 was computed for three degenerate quark masses @xmath6 , giving @xmath7 @xcite .
-9 truecm plane .
, title="fig:",scaledwidth=70.0% ] -3 truecm plane . , title="fig:",scaledwidth=70.0% ] [ fig : njl ] -2 truecm -2 truecm an analogous calculation , .
but for isospin baryochemical potential , give @xmath8 @xcite .
close to the zero density critical temperature the difference between effects produced by baryochemical potential and isospin chemical potential is small ( for instance , large @xmath9 calculation predict differences @xmath10 @xcite ) , so the two results @xcite consistently suggest that the region of first order phase transition becomes smaller at nozero chemical potential , at least till @xmath11 mev .
a mean field calculation carried out in the @xmath12 njl model shows instead the opposite trend@xcite , see fig .
[ fig : njl ] , right . all in all we are observing qualitative differences between qcd and purely fermionic models at small chemical potential .
indeed chiral symmetry is most likely broken by long distance forces in qcd , and by short distance vector forces in njl , and these differences might well have an impact on the existence of the critical endpoint , see also ref . @xcite .
the results @xcite have been obtained on lattices with @xmath13 .
it would be very important to confirm these predictions closer to the continuum limit .
indeed , the relevance of taking the continuum limit can be hardly overemphasised in qcd @xcite
. one substantial step towards the continuum limit of the critical line has been presented at this meeting @xcite .
the slope of the critical line has been computed for @xmath14 , and the claim is that there is a reasonable control over the @xmath15 extrapolation , see fig.3 .
note that while the ( pseudo)critical temperatures at @xmath16 estimated from the peak of the polyakov loop susceptibility differs from the one estimated from the peak of the chiral susceptibility , the slope of the two lines are the same within errorbars .
1.0 truecm are shown in the left plot , while the slope of the critical line , computed from two different susceptibilities , as a function of the lattice spacing are summarised in the right plot .
courtesy c. guse.,title="fig:",scaledwidth=60.0% ] are shown in the left plot , while the slope of the critical line , computed from two different susceptibilities , as a function of the lattice spacing are summarised in the right plot .
courtesy c. guse.,title="fig:",scaledwidth=30.0% ]
thermodynamics studies on the lattice are based on the analysis of the number density @xmath17 and its associated susceptibilities @xmath18 the latter are a significant probe of the fluctuational behaviour of the system@xcite , and are also the taylor coefficients of the excess pressure @xmath19 which contains information about baryon density effects in the eos . by exploiting these observables
it is was established already at the time of qm2006 that the hadron resonance gas model , where @xmath20 describes well the system up to rather high temperatures @xmath21 @xcite . for zero chemical potential ,
new results for the susceptibilites have been presented by the rbc - bielefeld collaboration@xcite .
they confirm the hadron gas behaviour at low temperature , and the approach to the free gas at high temperature .
most important , they note that these behaviours are not smoothly connected , rather there is a peak in between associated with the critical behaviour , see fig .
[ fig : chr ] .
-11 truecm -1.0 truecm is well fitted by a conventional critical behaviour ( left , from ref .
@xcite ) , as well as by a quasiparticle model with @xmath22 dependent coefficients@xcite . once continued to real chemical potential , this gives a modified form of the stefan - boltzmann law ( right , upper curve , from @xcite ) , which by increasing t gradually approaches the free field behaviour , corresponding to the lower curve in the plot , title="fig:",scaledwidth=120.0% ] -2.7 truecm -0.5 truecm is well fitted by a conventional critical behaviour ( left , from ref .
@xcite ) , as well as by a quasiparticle model with @xmath22 dependent coefficients@xcite . once continued to real chemical potential , this gives a modified form of the stefan - boltzmann law ( right , upper curve , from @xcite ) , which by increasing t gradually approaches the free field behaviour , corresponding to the lower curve in the plot , title="fig:",scaledwidth=110.0% ] -1.0 truecm at nonzero chemical potential , it turns out that it is useful to consider the phase diagram of qcd in the @xmath23 plane .
it has been found @xcite that the strongly interactive quark gluon plasma can be described by @xmath24 , implying @xmath25 , where @xmath26 is the critical point at imaginary chemical potential , see fig .
5 . note that the simple polynomial behaviour of the free field limit would be recovered when @xmath27 , and that @xmath28 , resulting from the numerical analysis , implies a slower increase of the particle number in the critical region with respect to the free case .
it would be interesting to repeat this investigation using the generalised method of ref.@xcite .
data at imaginary chemical potential are amenable to an easy comparison with analytic studies , and indeed the results in the strongly interactive region compares favourably with a quasiparticle study@xcite , once an explicit dependence of the self - energy parts on @xmath29 and @xmath30 , as well as an implicit dependence via the effective coupling @xmath31 has been taken into account : @xmath32 at larger temperature the data should compare successfully with a resummed perturbation theory @xcite , and it remains to be seen which is the lower limit of its applicability .
while the results of the two previous sections are robust , and we believe they can be systematically improved much in the same way as standard lattice qcd calculations , in this colder , denser region the sign problem becomes more severe and the results should still be considered exploratory @xcite .
the mass spectrum has been computed via the qcd strong coupling expansion , observing the expected signatures of chiral symmetry restoration , and a @xmath33 mass decreasing with temperature@xcite .
calculations have also been performed in the double limit : @xmath34 fixed , corresponding to an evolved ` quenched approximation ' in the presence of charged matter .
the order parameter has been computed ( note the ridge in the @xmath35 plane in fig .
[ fig : ale ] , left diagram ) , allowing the observation of the phase transition .
there are also emerging indication of a tricritical point , and studies of diquark are in progress @xcite.similar results come from a study using overlap fermions at strong coupling @xcite . -5 truecm 2 truecm 0.3 truecm from a numerical study at large mass(left @xcite ) and the pattern of chiral symmetry as seen from the meson spectrum at low temperature as a function of @xmath22 from strong coupling qcd ( right,@xcite),title="fig:",scaledwidth=80.0% ] -4 .
truecm -19 truecm from a numerical study at large mass(left @xcite ) and the pattern of chiral symmetry as seen from the meson spectrum at low temperature as a function of @xmath22 from strong coupling qcd ( right,@xcite),title="fig:",scaledwidth=210.0% ] -2 truecm new results in the canonical formalism , in which the canonical partition function was evaluated following hasenfratz - toussaint@xcite : @xmath36 allowed the identification of a first order line , a coexistence region , and the associated critical @xmath22 and critical densities , for four - flavor qcd@xcite . a variant of the canonical approach , based on the evaluation of the grand partition function via a taylor expansion produced results for @xmath37 , indicating a qualitative change at @xmath38 , and a first order line @xcite .
the density of states method - a reordering of the functional integral based on the costrained partition function ( which in condensed matter parlance might be called a mesoscopic approach ) - @xmath39 applied to the @xmath40 theory allowed the identification of two phase transition lines , and gave indication of a triple point @xcite ( fig.7 , left ) .
-6.0 truecm -2.9 truecm plane , from a ( still exploratory ) density of states calculation ( left diagram @xcite ) .
the phase diagram of qcd in the @xmath41 plane ( right diagram , @xcite).,title="fig:",scaledwidth=110.0% ] -3 truecm -0.1 truecm plane , from a ( still exploratory ) density of states calculation ( left diagram @xcite ) .
the phase diagram of qcd in the @xmath41 plane ( right diagram , @xcite).,title="fig:",scaledwidth=85.0% ]
finally in the cold and dense phase , the sign problem is very severe and at this moment i can only show result for qcd at non - zero density of isospin , and two - color qcd at nonzero baryon density .
these theories have extra symmetries , which protect the determinant from becoming complex , and allow standard montecarlo simulation , see e.g ref .
@xcite for reviews . at zero temperature
the onset for thermodynamics equals , as usual , the mass of the lowest excitation carrying the relevant charge , i.e. @xmath42 in the case of isospin density , and @xmath43 , for two- color qcd , so the physics of a finite density of isospin , and of finite baryon density have many analogies .
the phase diagram of qcd in the @xmath44 plane has been recently calculated , and the superfluid bec phase has been clearly identified @xcite ( fig.7 , right ) .
early studies , both analytic and numeric , have identified a superfluid phase also in two - colour qcd , by inspecting the behaviour of fermionic observables , and confirming the prediction of chiral perturbation theory , which is applicable here @xcite .
there is an emerging consensus that the bec phase of two - colour qcd is still confining , and preliminary indications of a bec / bcs crossover have been reported @xcite , with the superfluid order parameter assuming the scaling consistent with cooper pairing at a fermi surface .
results from ongoing work@xcite indicate that a bec / bcs transition persists at a finer lattice spacing . at the hadronic / bec interface
there are signal of criticality in the gluonic sector , as well as non trivial modifications of the glueball propagators in the superfluid phase @xcite ( fig.8 ) .
-6 truecm , indicating non trivial modification of the baryon - dense gluonic medium ( right , from ref .
@xcite).,scaledwidth=75.0% ] -4.5 truecm -5.7 truecm , indicating non trivial modification of the baryon - dense gluonic medium ( right , from ref .
@xcite).,title="fig:",scaledwidth=130.0% ] -1 truecm one final comment concerns the large @xmath9 limit , where a scenario has been proposed in which deconfinement and chiral transition are separated@xcite .
it is very tempting to speculate on possible analogies of these observation with the behaviour of @xmath45 , where indeed chiral transition ( even if between two phases where chiral symmetry remains broken ) and deconfinement transition are different phenomena .
i wish to thank the organisers of qm2008 for a most pleasant and interesting conference .
i am very grateful to gordon baym , philippe de forcrand , massimo delia , francesco di renzo , alessandra feo , zoltan fodor , christa guse , simon hands , tetsuo hatsuda , frithjof karsch , edwin laermann , larry mclerran , owe philipsen , christian schmidt , zsolt szep , and aleksi vuorinen for many helpful discussions and correspondence .
10 t. hatsuda , j. phys .
g * 34 * , s287 ( 2007 ) [ arxiv : hep - ph/0702293 ] .
d. toublan , phys .
b * 621 * , 145 ( 2005 ) [ arxiv : hep - th/0501069 ] .
p. kovacs and z. szep , phys .
d * 75 * , 025015 ( 2007 ) [ arxiv : hep - ph/0611208 ] .
k. fukushima , this volume .
see e.g. z. fodor , pos * lat07 * , 011 ( 2006 ) [ arxiv:0711.0336 [ hep - lat ] ]
. f. karsch , pos * lat07 * , 015 ( 2006 ) .
[ arxiv:0711.0661 [ hep - lat ] ] .
z. fodor , c. guse , s. d. katz and k. k. szabo , pos * lat07 * , 189 ( 2006 ) [ arxiv:0712.2702 [ hep - lat ] ] ; z. fodor , this volume .
r. v. gavai and s. gupta , phys . rev .
d * 73 * ( 2006 ) 014004 [ arxiv : hep - lat/0510044 ] . c. bernard
_ et al . _ ,
d * 77 * , 014503 ( 2008 ) [ arxiv:0710.1330 [ hep - lat ] ] . c. r. allton _ et al .
_ , phys .
d * 71 * , 054508 ( 2005 ) [ arxiv : hep - lat/0501030 ] .
f. karsch , k. redlich and a. tawfik , phys .
b * 571 * ( 2003 ) 67 [ arxiv : hep - ph/0306208 ] .
m. delia and m. p. lombardo , phys .
d * 70 * , 074509 ( 2004 ) [ arxiv : hep - lat/0406012 ] . c. schmidt for the rbc - bielefeld collaboration , this volume .
m. delia , f. di renzo and m. p. lombardo , phys .
d * 76 * , 114509 ( 2007 ) [ arxiv:0705.3814 [ hep - lat ] ] .
v. azcoiti , g. di carlo , a. galante and v. laliena , nucl .
b * 723 * , 77 ( 2005 ) [ arxiv : hep - lat/0503010 ] .
m. bluhm and b. kampfer , phys .
d * 77 * , 034004 ( 2008 ) [ arxiv:0711.0590 [ hep - ph ] ] .
a. vuorinen , phys .
d * 68 * ( 2003 ) 054017 [ arxiv : hep - ph/0305183 ] , a. ipp , k. kajantie , a. rebhan and a. vuorinen , phys .
d * 74 * , 045016 ( 2006 ) [ arxiv : hep - ph/0604060 ] ; see j. p. blaizot , nucl .
a * 785 * , 1 ( 2007 ) [ arxiv : nucl - th/0611104 ] . for a review .
see e.g. m. p. lombardo , prog .
suppl . *
153 * , 26 ( 2004 ) [ arxiv : hep - lat/0401021 ] .
s. muroya , a. nakamura , c. nonaka and t. takaishi , prog .
phys . *
110 * , 615 ( 2003 ) [ arxiv : hep - lat/0306031 ] .
n. kawamoto , k. miura and a. ohnishi , arxiv:0710.1720 [ hep - lat ] ; a. ohnishi , n. kawamoto and k. miura , arxiv:0803.0255 [ nucl - th ] .
r. de pietri , a. feo , e. seiler and i. o. stamatescu , phys .
d * 76 * ( 2007 ) 114501 [ arxiv:0705.3420 [ hep - lat ] ] .
x. l. yu and x. q. luo , mod .
a * 22 * , 537 ( 2007 ) .
[ arxiv : hep - lat/0508032 ] .
a. hasenfratz and d. toussaint , nucl .
b * 371 * ( 1992 ) 539 .
p. de forcrand and s. kratochvila , nucl .
* 153 * , 62 ( 2006 ) [ arxiv : hep - lat/0602024 ]
. s. ejiri , arxiv:0710.0653 [ hep - lat ] ; s. ejiri , arxiv:0804.3227 [ hep - lat ] .
z. fodor , s. d. katz and c. schmidt , jhep * 0703 * , 121 ( 2007 ) [ arxiv : hep - lat/0701022 ] .
s. hands , prog .
suppl . * 168 * , 253 ( 2007 ) [ arxiv : hep - lat/0703017 ] ; d. k. sinclair , j. b. kogut and d. toublan , prog .
suppl . *
153 * , 40 ( 2004 ) [ arxiv : hep - lat/0311019];m . g. alford , a. kapustin , f. wilczek , physical review * d59 * ( 1999 ) 054502 .
de forcrand , m. a. stephanov and u. wenger , pos * lat2007 * , 237 ( 2007 ) [ arxiv:0711.0023 [ hep - lat ] ] .
s. hands , s. kim and j. i. skullerud , eur .
j. c * 48 * , 193 ( 2006 ) [ arxiv : hep - lat/0604004 ] .
s. hands , s. kim and j. i. skullerud , in preparation , private communication with s. hands . m. p. lombardo , m. l. paciello , s. petrarca and b. taglienti , pos * lat07 * , 206 ( 2006 ) [ arxiv:0710.3251 [ hep - lat]];arxiv:0804.4863 [ hep - lat ] , epjc , in press . l. mclerran and r. d. pisarski , nucl . phys .
a * 796 * , 83 ( 2007 ) [ arxiv:0706.2191 [ hep - ph ] ] . | a concise review of the progress of lattice calculations at non - zero density since qm2006 , with emphasis on the high baryon density , low temperature domain .
possibilities for exploring densities higher than those studied by standard techniques are analysed .
the phase transitions of cold , dense matter , where the sign problem remains severe , are discussed in the context of qcd - like models and approximations to qcd . |
the basis of leptogenesis @xcite are the ` sphaleron processes ' , effective nonperturbative interactions @xmath2 of all left - handed quarks and leptons in the standard model @xcite ( cf .
fig . 1 ) , which change baryon number ( b ) and lepton number ( l ) by a multiple of three , while preserving @xmath0-@xmath1 , @xmath3 here @xmath4 , the chern - simons number , is an integer characterizing the sphaleron gauge field configuration . at high temperatures , between the critical temperature @xmath5 of the electroweak phase transition and a maximal temperature @xmath6 , @xmath7 these processes
are believed to be in thermal equilibrium @xcite .
although uncontroversial among theorists , it has to be stressed that this important phenomenon has so far not been experimentally tested !
it is therefore very interesting that the corresponding phenomenon of chirality changing processes in strong interactions might be observable in heavy ion collisions at the lhc @xcite .
sphaleron processes relate baryon and lepton number and therefore strongly affect the generation of the cosmological baryon asymmetry . analyzing the chemical potentials of quarks and leptons in thermal equilibrium , one obtains an important relation between the asymmetries in @xmath0- , @xmath1- and @xmath0-@xmath1-number , @xmath8 where @xmath9 . in the standard model one
has @xmath10 .
this relation suggests that lepton number violation can explain the cosmological baryon asymmetry
. however , lepton number violation can only be weak at late times , since otherwise any baryon asymmetry would be washed out .
the interplay of these conflicting conditions leads to important contraints on neutrino properties , and on extensions of the standard model in general . because of the sphaleron processes
, lepton number violation can replace baryon number violation in sakharov s conditions for baryogenesis .
leptogenesis is an immediate consequence of the seesaw mechanism , which explains the smallness of light neutrino masses in terms of the largeness of heavy majorana neutrino masses .
the heavy mass eigenstates @xmath11 and the light mass eigenstates @xmath12 are given by @xmath13 where @xmath14 is the dirac neutrino mass matrix .
for third generation yukawa couplings @xmath15 , as in some so(10 ) gut models , one obtains the heavy and light neutrino masses , @xmath16 remarkably , the light neutrino mass @xmath17 is compatible with @xmath18 ev , as measured in atmospheric @xmath12-oscillations .
this suggests that neutrino physics probes the mass scale of grand unification ( gut ) , although other interpretations of neutrino masses are possible as well .
the heavy majorana neutrinos have no gauge interactions .
hence , in the early universe , they can easily be out of thermal equilibrium .
this makes @xmath19 , the lightest of them , an ideal candidate for baryogenesis , in accord with sakharov s condition of departure from thermal equilibrium . in the simplest form of leptogenesis the heavy majorana neutrinos
are produced by thermal processes , which is therefore called ` thermal leptogenesis ' .
the @xmath20 violating @xmath19 decays into lepton - higgs pairs lead to a lepton asymmetry @xmath21 , which is partially converted to a baryon asymmetry @xmath22 by the sphaleron processes . in early work on leptogenesis , it was anticipated that the light neutrino masses are then required to have masses @xmath23 @xcite . after the discovery of atmospheric neutrino oscillations , more stringent upper bounds on neutrino masses could be derived , and leptogenesis became increasingly popular .
the generated baryon asymmetry is proportional to the @xmath20 asymmetry in @xmath19-decays . for hierarchical heavy neutrinos it satisfies the upper bound @xcite @xmath24 which depends on the mass of @xmath19 and on @xmath25 , the mass splitting in atmospheric neutrino oscillations .
the nonequilibrium process of baryogenesis via leptogenesis in the hot early universe is usually described by a set of boltzmann equations , where also important washout processes have to be taken into acount . solving these equations , one obtains lower bounds on the heavy neutrino mass @xmath26 , and upper bounds on the light neutrino mass @xmath27 @xcite . in the simplest case of hierarchical heavy neutrinos , and summing over lepton flavours , one obtains a mass window of light neutrino masses , which is favoured by leptogenesis @xcite @xmath28 note , however , that both , lower and upper bound , are modified by lepton flavour effects , which have been extensively studied in recent years ( for reviews see , for example , refs .
@xcite ) . in view of eq .
( [ window ] ) , knowledge of the absolute neutrino mass scale is of crucial importance .
hence , a measurement of the neutrino masses @xmath29 in tritium @xmath30-decay @xcite and @xmath31 in neutrinoless double @xmath30-decay @xcite , or the determination of the sum @xmath32 from cosmology @xcite , consistent with eq .
( [ window ] ) , would strongly support the leptogenesis mechanism .
how does leptogenesis depend on the phases of the dirac and majorana neutrino mass matrices ? due to the large value of @xmath33 reported at this conference @xcite ,
measurement of the pmns phase @xmath34 now appears feasible .
this is certainly important , and in some models the generated baryon asymmetry strongly depends on the phase @xmath34 ( cf .
3 ) . in general
, however , this is model dependent@xcite , and in the case of hierarchical heavy neutrinos the effect of the pmns phase @xmath34 is unimportant @xcite .
so far we have considered right - handed neutrinos with gut scale majorana masses . in pinciple , the three right - handed ( sterile ) neutrinos could have much smaller masses , of order gev or even kev .
it is remarkable that in this case the standard model with three sterile neutrinos ( @xmath12msm scenario ) can account for neutrino oscillations , baryogenesis and dark matter @xcite .
the standard model lagrangian is extended by dirac and majorana mass terms , @xmath35 and the active - sterile mixings are described by the matrix @xmath36 ( @xmath37 ) .
the scenario has recently been studied in detail quantitatively @xcite .
the lightest sterile neutrino @xmath19 provides dark matter , with a mass in the range @xmath38 , and tiny mixings , @xmath39 , constrained by x - ray observations .
following ref .
@xcite , baryogenesis is achieved by cp - violating oscillations of @xmath40 and @xmath41 . to obtain the right amount of baryon asymmetry and dark matter , resonant enhancement of cp violation
is needed , with a high degeneracy of the sterile neutrinos , @xmath42 .
the observed dark matter abundance @xmath43 requires @xmath44 masses in the range from 2 - 10 gev ( cf .
fig . 4 ) .
sterile neutrino ( @xmath19 ) dark matter with mass in the kev range can also be realized in models with left - right symmetric electroweak interactions based on the gauge group @xmath45 .
the baryon asymmetry is then generated by @xmath46 decays , which requires @xmath44 masses of order @xmath47 gev @xcite .
a further low scale alternative is leptogenesis at the electroweak scale , with sterile neutrino masses @xmath48 gev and additional scalar fields , which leads to specific predictions testable at the lhc @xcite .
more models of leptogenesis at the tev scale can be found in ref .
@xcite .
as already discussed in the previous section , the seesaw mechanism does not only work for right - handed neutrino masses at the gut scale and yukawa couplings similar to quark and charged lepton yukawa couplings , it is also applicable for heavy neutrino masses at the tev scale and very small yukawa couplings . in this case
heavy neutrino self - energy effects have to dominate the @xmath20 asymmetry @xcite , leading to resonant leptogenesis in the case of quasi - degenerate right - handed neutrinos @xcite . in a particular neutrino mass model successful leptogenesis
is achieved for masses at the electroweak scale , @xmath49 , with a degeneracy @xmath50 @xcite .
it is well known that in supersymmetric models there is a close connection between leptogenesis and lepton flavour changing processes like @xmath51 @xcite .
it is interesting that in the case of resonant leptogenesis , one can have large lepton flavour changing rates also without supersymmetry ( cf .
fig . 5 ) , within the reach of the meg experiment @xcite .
leptogenesis is a nonequilibrium process taking place in an expanding universe with decreasing temperature .
it involves quantum interferences in a crucial manner , which implies that the standard treatment by means of boltzmann equations is theoretically unsatisfactory @xcite . in particular , it is currently not possible to quote a theoretical error on the predicted bayon asymmetry . within quantum field theory ,
leptogenesis can be treated on the basis of the schwinger - keldysh formalism @xcite , and during the past years significant progress has been made towards a theory of leptogenesis @xcite .
very important in this context is also the calculation of quantum corrections to decay widths and scattering cross sections at high temperature @xcite . in the schwinger - keldysh formalism one
considers green s functions @xmath52 for heavy neutrino @xmath19 , lepton and higgs on a complex time contour starting at some initial time @xmath53 ( cf .
these green s functions satisfy schwinger - dyson equations with self - energies @xmath54 , @xmath55 it is then convenient to consider two particular correlation functions , the spectral functions @xmath56 , which contain information about the system , and the statistical propagators @xmath57 , which depends on the initial state at time @xmath53 , @xmath58\rangle\ .\end{aligned}\ ] ] they satisfy the kadanoff - baym equations @xcite @xmath59 where we have assumed spatial homogeneity , and @xmath60 is the dalembert operator for a particular momentum mode @xmath61 . solving these kadanoff - baym equations
, one can describe the change of the system from an initial state of zero baryon number to a final state of non - zero baryon number , i.e. the process of baryogenesis .
recently , this technique has been applied to resonant leptogenesis @xcite .
the obtained effective enhancement of the @xmath20 asymmetry is shown in fig .
the maximal enhancement predicted by boltamann equations reads @xmath62 note that for equal masses and widths of the two heavy neutrinos @xmath19 and @xmath46 , @xmath63 is singular , and therefore unphysical .
this singularity is cured by memory effects contained in the kadanoff - baym equations , which yield the result @xmath64 in summary , the generic effect of a possible resonant enhancement of the @xmath20 asymmetry is confirmed by the full quantum mechanical treatment . however , its size is reduced .
thermal leptogenesis requires a rather large reheating temperature , @xmath65 . in supersymmetric theories
this causes a potential problem because of the thermal production of gravitinos , which yields the abundance @xmath66 where @xmath67 , and @xmath68 is the reheating temperature . for unstable gravitinos , one has to worry about consistency with primordial nucleosynthesis ( bbn ) whereas stable gravitinos may overclose the universe . as a possible way out , nonthermal production of heavy neutrinos has been suggested @xcite , which allows to decrease the reheating temperature and therefore the gravitino production . on the other hand ,
it is remarkable that for typical gravitino and gluino masses in gravity mediated supersymmetry breaking , a reheating temperature @xmath69 yields the right order of magnitude for the dark matter abundance if the gravitino is the lsp . but
why should the reheating temperature be as large as the temperature favoured by leptogenesis , i.e. @xmath70 ?
it this context it is interesting to note that for typical neutrino mass parameters in leptogenesis , @xmath71 , @xmath72 , the heavy neutrino decay width takes the value @xmath73 if the early universe in its evolution would reach a state where the energy density is dominated by nonrelativistic heavy neutrinos , their decays to lepton - higgs pairs would lead to a relativistic plasma with temperature @xmath74 which is indeed the temperature wanted for gravitino dark matter ! is this an intriguing hint or just a misleading coincidence ?
it is remarkable that an intermediate heavy neutrino dominance indeed occurs in the course of the cosmological evolution if the initial inflationary phase is driven by the false vacuum energy of unbroken @xmath0-@xmath1 symmetry @xcite .
consider the supersymmetric standard model with right - handed neutrinos , described by the superpotential ( in @xmath75 notation : @xmath76 , @xmath77 ) , @xmath78 supplemented by a term which enforces @xmath0-@xmath1 breaking , @xmath79 the higgs fields @xmath80 and @xmath81 break electroweak symmetry and @xmath0-@xmath1 symmetry , respectively .
it is very interesting that the last term , @xmath82 , can successfully describe inflation with @xmath83 as inflaton field @xcite .
inflation ends in tachyonic preheating where @xmath0-@xmath1 is spontaneously broken .
the false vacuum energy density is then rapidly converted into a nonrelativistic gas of @xmath0-@xmath1 higgs bosons and a small admixture of a relativistic plasma of standard model particles produced during preheating .
heavy neutrinos @xmath19 are thermally produced from the plasma and nonthermally in decays of the @xmath0-@xmath1 higgs bosons ( @xmath84 ) .
this generates entropy , baryon asymmetry and gravitino abundance .
the result of a quantitative analysis is shown in fig . 8 for typical parameters : @xmath85 , @xmath86 , @xmath87 and @xmath88 . the final baryon asymmetry and dark matter abundance are @xmath89 , @xmath90 .
a systematic parameter scan yields a lower bound on the gravitino mass , @xmath91 , and a range of heavy neutrino masses , @xmath92 .
note that the described mechanism for the generation of dark matter also works for very heavy gravitinos , whose decays before bbn produce nonthermal higgsino or wino dark matter @xcite . as fig .
8 shows , most radiation , @xmath0-@xmath1 asymmetry and gravitino abundance are generated during a reheating period where the cosmic scale factor increases from @xmath93 to @xmath94 , and the equation of state changes from matter dominance to radiation dominance .
on the other hand , the ` reheating temperature ' @xmath68 is roughly constant , since there is a balance between temperature decrease due to expansion and temperature increase due to @xmath0-@xmath1 higgs boson decays .
note that contrary to conventional reheating mechanisms , the reheating temperature @xmath68 is now determined by neutrino parameters ! for neutrino masses consistent with leptogenesis and dark matter , the reheating temperature varies between @xmath95 and @xmath96 ( cf .
fig . 9 ) .
* standard thermal leptogenesis is an elegant explanation of the cosmological matter - antimatter asymmetry .
it is consistent with gut scale heavy neutrino masses , and it ( successfully ) predicts bounds on light neutrino masses and , in a model dependent way , also restricts cp phases .
* there exist various viable low - scale alternatives , with kev sterile neutrinos as dark matter or right - handed neutrinos with @xmath97 masses .
one then assumes highly degenerate masses to obtain a resonance enhancement of @xmath20 violation in heavy neutrino decays , which allows successful leptogenesis .
contrary to gut scale leptogenesis , these models can be directly tested in astrophysical observations and at the lhc . * recently , considerable progress has been made in treating the nonequilibrium process of leptogenesis within quantum field theory , based on the schwinger - keldysh formalism , with a first application to resonant leptogenesis .
furthermore , important quantum corrections to decay rates and scattering processes have been evaluated at finite temperature .
one can expect that these efforts will eventually lead to a quantitative theory of leptogenesis. * the possible connection between leptogenesis , dark matter and inflation is a fascinating possibility . it is intriguing that in a supersymmetric extension of the standard model with right - handed neutrinos , a false vacuum with unbroken @xmath98 symmetry indeed provides an initial state whose decay , via an intermediate state of heavy neutrino dominance , can explain the observed entropy , matter - antimatter asymmetry and dark matter abundance .
the author thanks pasquale di bari , valerie domcke , mathias garny and kai schmitz for helpful discussions during the preparation of this talk .
this work has been supported by the german science foundation ( dfg ) within the collaborative research center 676 `` particles , strings and the early universe '' .
00 m. fukugita and t. yanagida , phys .
b * 174 * ( 1986 ) 45 .
g. t hooft , phys .
* 37 * ( 1976 ) 8 .
v. a. kuzmin , v. a. rubakov and m. e. shaposhnikov , phys .
b * 155 * ( 1985 ) 36 .
d. e. kharzeev , l. d. mclerran and h. j. warringa , nucl .
a * 803 * ( 2008 ) 227 [ 0711.0950 [ hep - ph ] ] .
t. kalaydzhyan and i. kirsch , phys .
* 106 * ( 2011 ) 211601 [ 1102.4334 [ hep - th ] ] .
w. buchmuller and m. plumacher , phys .
b * 389 * ( 1996 ) 73 [ hep - ph/9608308 ] .
s. davidson and a. ibarra , phys .
b * 535 * ( 2002 ) 25 [ hep - ph/0202239 ] .
k. hamaguchi , h. murayama and t. yanagida , phys .
d * 65 * ( 2002 ) 043512 [ hep - ph/0109030 ] .
w. buchmuller , p. di bari and m. plumacher , nucl .
b * 665 * ( 2003 ) 445 [ hep - ph/0302092 ] ; annals phys . *
315 * ( 2005 ) 305 [ hep - ph/0401240 ] .
g. f. giudice , a. notari , m. raidal , a. riotto and a. strumia , nucl .
b * 685 * ( 2004 ) 89 [ hep - ph/0310123 ] .
e. nardi , nucl .
* 217 * ( 2011 ) 27 .
s. davidson , e. nardi and y. nir , phys .
* 466 * ( 2008 ) 105 [ 0802.2962 [ hep - ph ] ] .
p. di bari , contemp .
* 53 * ( 2012 ) issue4 [ 1206.3168 [ hep - ph ] ] .
j. f. wilkerson , these proceedings e. molinaro and s. t. petcov , eur .
j. c * 61 * ( 2009 ) 93 [ 0803.4120 [ hep - ph ] ] .
for a recent discussion and references see , for example , g. c. branco , r. g. felipe and f. r. joaquim , 1111.5332 [ hep - ph ] .
g. c. branco , t. morozumi , b. m. nobre and m. n. rebelo , nucl .
b * 617 * ( 2001 ) 475 [ hep - ph/0107164 ] .
s. davidson , j. garayoa , f. palorini and n. rius , phys .
. lett .
* 99 * ( 2007 ) 161801 [ 0705.1503 [ hep - ph ] ] .
t. asaka , s. blanchet and m. shaposhnikov , phys .
b * 631 * ( 2005 ) 151 [ hep - ph/0503065 ] ; phys .
b * 620 * ( 2005 ) 17 [ hep - ph/0505013 ] .
l. canetti , m. drewes and m. shaposhnikov , 1204.3902 [ hep - ph ] ; l. canetti , m. drewes , t. frossard and m. shaposhnikov , e. k. akhmedov , v. a. rubakov and a. y. .smirnov , phys . rev .
* 81 * ( 1998 ) 1359 [ hep - ph/9803255 ] .
f. bezrukov , a. kartavtsev and m. lindner , 1204.5477 [ hep - ph ] . b. kayser and g. segre , phys .
b * 704 * ( 2011 ) 570 [ 1011.6362 [ hep - ph ] ] .
m. flanz , e. a. paschos and u. sarkar , phys .
b * 345 * ( 1995 ) 248 [ erratum - ibid .
b * 382 * ( 1996 ) 447 ] [ hep - ph/9411366 ] .
l. covi , e. roulet and f. vissani , phys .
b * 384 * ( 1996 ) 169 [ hep - ph/9605319 ] .
a. pilaftsis , phys .
d * 56 * ( 1997 ) 5431 [ hep - ph/9707235 ] .
a. pilaftsis and t. e. j. underwood , nucl .
b * 692 * ( 2004 ) 303 [ hep - ph/0309342 ] .
for a recent discussion and references see , for example , a. ibarra and c. simonetto , jhep * 0908 * ( 2009 ) 113 [ 0903.1776 [ hep - ph ] ] . f. f. deppisch and a. pilaftsis , aip conf .
* 1467 * ( 2012 ) 159 [ 1205.6994 [ hep - ph ] ] .
r. sawada , these proceedings j. s. schwinger , j. math .
* 2 * ( 1961 ) 407 .
l. v. keldysh , zh . eksp .
fiz . * 47 * ( 1964 ) 1515 [ sov
. phys .
jetp * 20 * ( 1965 ) 1018 ] .
w. buchmuller and s. fredenhagen , phys .
b * 483 * ( 2000 ) 217 [ hep - ph/0004145 ] .
a. anisimov , w. buchmuller , m. drewes and s. mendizabal , annals phys .
* 324 * ( 2009 ) 1234 [ 0812.1934 [ hep - th ] ] ; phys .
* 104 * ( 2010 ) 121102 [ 1001.3856 [ hep - ph ] ] ; annals phys .
* 326 * ( 2011 ) 1998 [ 1012.5821 [ hep - ph ] ] .
m. garny , a. hohenegger , a. kartavtsev and m. lindner , phys .
d * 80 * ( 2009 ) 125027 [ 0909.1559 [ hep - ph ] ] ; phys .
d * 81 * ( 2010 ) 085027 [ 0911.4122 [ hep - ph ] ] .
m. garny , a. hohenegger and a. kartavtsev , phys .
d * 81 * ( 2010 ) 085028 [ 1002.0331 [ hep - ph ] ] . v. cirigliano , c. lee , m. j. ramsey - musolf and s. tulin , phys .
d * 81 * ( 2010 ) 103503 [ 0912.3523 [ hep - ph ] ] .
m. beneke , b. garbrecht , m. herranen and p. schwaller , nucl .
b * 838 * ( 2010 ) 1 [ 1002.1326 [ hep - ph ] ] ; m. beneke , b. garbrecht , c. fidler , m. herranen and p. schwaller , nucl . phys .
b * 843 * ( 2011 ) 177 [ 1007.4783 [ hep - ph ] ] . c. fidler , m. herranen , k. kainulainen and p. m. rahkila , jhep * 1202 * ( 2012 ) 065 [ 1108.2309 [ hep - ph ] ] ; m. herranen , k. kainulainen and p. m. rahkila , jhep * 1202 * ( 2012 ) 080 [ 1108.2371 [ hep - ph ] ] ; b. garbrecht and m. herranen , nucl .
b * 861 * ( 2012 ) 17 [ 1112.5954 [ hep - ph ] ] .
d. besak and d. bodeker , jhep * 1005 * ( 2010 ) 007 [ 1002.0022 [ hep - ph ] ] ; jcap * 1203 * ( 2012 ) 029 [ 1202.1288 [ hep - ph ] ] .
a. anisimov , d. besak and d. bodeker , jcap * 1103 * ( 2011 ) 042 [ 1012.3784 [ hep - ph ] ] .
m. laine and y. schroder , jhep * 1202 * ( 2012 ) 068 [ 1112.1205 [ hep - ph ] ] ; a. salvio , p. lodone and a. strumia , jhep * 1108 * ( 2011 ) 116 [ 1106.2814 [ hep - ph ] ] . c. kiessig and m. plumacher , jcap
* 1207 * ( 2012 ) 014 [ 1111.1231 [ hep - ph ] ] ; jcap * 1209 * ( 2012 ) 012 [ 1111.1235 [ hep - ph ] ] . l. p.
kadanoff and g. baym , quantum statistical mechanics , benjamin , new york , 1962 .
m. garny , a. kartavtsev and a. hohenegger , 1112.6428 [ hep - ph ] .
g. lazarides and q. shafi , phys .
b * 258 * ( 1991 ) 305 .
h. murayama , h. suzuki , t. yanagida and j. i .
yokoyama , phys .
* 70 * ( 1993 ) 1912 .
t. asaka , k. hamaguchi , m. kawasaki and t. yanagida , phys .
b * 464 * ( 1999 ) 12 [ hep - ph/9906366 ] ; phys . rev .
d * 61 * ( 2000 ) 083512 [ hep - ph/9907559 ] .
s. antusch , j. p. baumann , v. f. domcke and p. m. kostka , jcap * 1010 * ( 2010 ) 006 [ 1007.0708 [ hep - ph ] ] .
w. buchmuller , k. schmitz and g. vertongen , phys .
b * 693 * ( 2010 ) 421 [ 1008.2355 [ hep - ph ] ] ; nucl .
b * 851 * ( 2011 ) 481 [ 1104.2750 [ hep - ph ] ] .
w. buchmuller , v. domcke and k. schmitz , nucl .
b * 862 * ( 2012 ) 587 [ 1202.6679 [ hep - ph ] ] .
e. j. copeland , a. r. liddle , d. h. lyth , e. d. stewart and d. wands , phys .
d * 49 * ( 1994 ) 6410 [ astro - ph/9401011 ] .
g. r. dvali , q. shafi and r. k. schaefer , phys .
* 73 * ( 1994 ) 1886 [ hep - ph/9406319 ] .
w. buchmuller , v. domcke and k. schmitz , phys . lett .
b * 713 * ( 2012 ) 63 [ 1203.0285 [ hep - ph ] ] . | after a brief discussion of baryon and lepton number nonconservation , we review the status of thermal leptogenesis with gut scale neutrino masses , as well as low scale alternatives with kev neutrinos as dark matter and heavy neutrino masses within the reach of the lhc .
recent progress towards a full quantum mechnical description of leptogenesis is described with resonant leptogenesis as an application .
finally , cosmological @xmath0-@xmath1 breaking after inflation is considered as origin of the hot early universe , generating entropy , baryon asymmetry and dark matter .
leptogenesis , nonequilibrium processes , dark matter |
the most influential role cepheid variables play in galactic and extragalactic astronomy is through their period - luminosity ( pl ) relation .
the rather easy measurement of the period of their regular pulsation allows to estimate an absolute magnitude or luminosity and therefore to derive eventually the distance to the variable star .
the high luminosities of cepheids favors them to anchor the extragalactic distance scale . due to the still debated value of the hubble constant ,
a secure calibration and a solid understanding of the pl relation is of fundamental importance .
references to the enormous body of literature on the pl relation and directions to historic accounts can be found e.g. in cox ( 1980 ) , madore and freedman ( 1991 ) , gieren and fouqu ( 1993 ) .
theoretical pl relations of cepheids were commonly determined by referring to mass - luminosity ( ml ) relations which were derived from stellar evolution computations .
thereupon separate stellar envelope models were calculated whose stability properties were eventually studied ( e.g. @xcite , @xcite , @xcite , @xcite ) .
the authors are not aware of any pl - relation studies in which stability properties of stellar evolution models were computed directly . in terms of recent theoretical work on the pl relation of cepheids
we should mention stothers ( 1988 ) .
he used various published sources of stellar evolution computations with differing composition and different physical assumptions to derive ml and pl relations .
the periods of the radial fundamental modes were already parameterized through a period - mass - radius relation which was fitted through numerical data .
the aim of stother s study was to quantify the relative dependences of the pl and period - luminosity - color ( plc ) relation on chemical composition on purely theoretical grounds .
later , chiosi et al . (
1993 ) performed pulsation computations on envelope models which were calculated on the basis of ml relations derived from stellar evolution tracks from the literature and from a few unpublished computations .
a range of chemical compositions which bracketed galactic as well as magellanic cepheids were considered .
the effect of overshooting was included in the modeling ; for mild overshooting @xmath1 of the of the ml relation increased by 0.25 and it was shifted by @xmath1 by 0.5 for full overshooting .
eventually , analytical fits to the numerical data were transformed into astronomical observables ( colors and magnitudes in different band - passes ) .
the authors favored models with considerable overshoot over ones neglecting it as the former ones apparently solved the long - standing mass discrepancy problem for the cepheids .
this particular problem was , however , considerably defused by the application of the new generation of opacity data ( @xcite ) .
therefore , the question concerning a realistic overshooting length of convective elements at the edge of the convective core has to be re - addressed from a different viewpoint .
due to recent high - precision observations with hst and hipparchos accuracy considerations of the distance - scale calibration relying on cepheids gained new momentum ( @xcite , @xcite , @xcite ) .
however , suggestions of the pl relation being sensitive to metallicity effects ( for an extreme viewpoint see @xcite ) overshadowed the distance determination of galaxies with different mean metallicities .
a solid understanding of the theoretical basis of the pl relation is indispensable ( sandage 1996 ) to prevent possible systematic misinterpretation of the data .
this paper presents self - consistently derived blue - edges of radial fundamental modes of model stars which were obtained from evolution computations .
heavy - element and helium abundances were varied and evolutionary tracks of stars with various masses were recomputed to find possible shifts of the location of the blue edge of the instability strip . from the data which was found directly from the nonadiabatic pulsation computations the blue edge could be derived .
the red edge of the instability strip is an ill - defined quantity in astrophysics . to have a handle what happens pulsationally at cooler temperatures we assumed ad hoc a red line which was displaced by a fixed @xmath2 relative to the blue edge . from the self - consistently determined blue edge as well as from the intersection of the evolutionary tracks with the red line we derived the corresponding theoretical pl relations . as a side aspect
, we looked into the variation of the pulsation constant @xmath0 .
the numerical results and analytical fits to them are presented in section 3 .
we compare our findings with other theoretical studies in section 4 and attempt to give some explanations for encountered deviations .
evolutionary tracks of 4 , 5 , 6 , 7 , and 10 @xmath3 stars were computed with a henyey - type stellar evolution program . for the stellar opacity we used the opal tables which were extended to the low - temperature domain below @xmath4 using the alexander and ferguson ( 1994 ) data .
convective overshooting was neglected ; this aspect will be commented on in section 4 .
the computations started always with chemically homogeneous main - sequence models .
the evolution of the model stars was followed through the helium burning so that the major blue loops which cross the instability strip were included . to investigate the influence of different metallicities on stellar evolution and in particular on stellar pulsations we adopted the following range of ( x , z ) values : ( 0.7,0.02 ) , ( 0.7 , 0.004 ) , and ( 0.7 , 0.001 ) .
the influence of helium abundance on stellar evolution and on pulsational stability was studied by computing one set of stellar masses , as listed above , with ( x , z ) = ( 0.75 , 0.004 ) .
figure 1 shows examples of the evolutionary tracks for stellar models with initial homogeneous composition ( x , z ) = ( 0.7,0.004 ) and for 4 and 10 @xmath3 with ( 0.75 , 0.004 ) , respectively .
nonadiabatic radial stellar pulsation computations were _ performed directly on those evolutionary models _ which are located in the vicinity of the instability strip .
this contrasts the previous approaches where envelope models of cepheids were computed based on mass - luminosity relations derived from stellar evolution calculated for other purposes .
these values are kept constant throughout the instability strip .
this implies constant @xmath5 across the strip .
therefore , our approach is to be considered as more fundamental since it is based directly on consistently computed stellar evolution models .
computationally , the pulsation code is of a finite - difference relaxation type and its basics were described in saio et al ( 1983 ) .
the location of the fundamental blue edge ( fbe ) was interpolated on the hr diagram between adjacent evolutionary models for which the imaginary part of the fundamental mode changed sign from being stable to being overstable . the corresponding red edge ( fre )
could not be located in our computations since the indispensable pulsation - convection interaction is not included in the code .
the resulting lack of convective - flux leak causes the damping to be too weak to overcome the he@xmath6-partial ionization driving so that pulsational instability extended to much too low temperatures .
therefore , this other natural limit for pulsating stars could not be used to study the mapping of the strip onto the p
l plane and to derive lines of constant period . for these purposes , however , it is sufficient to _ define _ an arbitrary line , which we chose to lie close to the supposed fre
. unfortunately , the width of the instability region is , observationally as well as theoretically , a poorly known quantity ( cf .
@xcite , @xcite ) .
we adopted here our red line to be shifted by @xmath7 to the red from the fbe at any luminosity .
this particular choice was motivated by the theoretical analysis of buchler et al .
( 1990 ) . for comparisons with observational data ,
we emphasize that the blue - edge data should be considered since they depend neither on any assumptions concerning pulsation - convection interaction nor on observational or on any other kind of secondary calibration .
in this section we present the results of radial fundamental modes at the blue edges , at the location of our red line , and the data that can be derived therefrom .
in the figures , the blue - edge data will be referred to by open symbols whereas red - line data will be displayed by filled ones .
the correspondence of the different chemical compositions with appropriate graphical symbols can be read off from the panel inserted in fig . 2 .
figure 1 shows
superposed on some evolutionary tracks the location of the blue - edge of the instability strip for the radial fundamental mode .
despite the tracks with equal masses being displaced considerably from each other for different chemical compositions , the position of the fbes is not very sensitive to abundances and evolutionary phase .
a linear least - square fit to the fbe data ( including all different chemical compositions ) leads to the following parameterization @xmath8 for quite some purposes such a mean fit is sufficient
. a closer look at the dependencies of the fbe position on chemical composition revealed @xmath9 the coefficients were determined at @xmath10 .
the sensitivity on the helium abundance has to be taken with caution as it is based on two composition choices only : ( x , z)=(0.7,0.004 ) and ( 0.75,0.004 ) , respectively .
our helium dependence is indeed considerably weaker than the ones given by iben and tuggle ( 1972 ) or chiosi et al .
nevertheless , the compensating action of @xmath11 and @xmath12 remains unchanged .
when the periods of the fundamental mode of the stellar models at the fbe are plotted versus their luminosity we arrive directly at the pl relation .
the same procedure was applied along the red line of the instability strip .
the results for the blue edge ( open symbols ) and the red line ( filled symbols ) are shown in fig .
only crossings of the instability strip during the cepheid blue - loops ( cf .
@xcite ) were considered .
we notice no systematic difference neither in slope nor in zero - point of the pl relation on metallicity or helium abundance . with good accuracy ,
the red - line data are shifted by a constant amount relative to the blue - edge data .
again , no systematic dependence of the corresponding pl relation on metallicity or helium was found . to quantify the pl relation , we performed a least - square straight - line fit through the blue - edge and the red - line data separately .
again , only second and later crossings of the instability strip are included in the data .
for the blue edge we found @xmath13 with a mean @xmath14 deviation from the fit of @xmath15 .
the red - line data can be described by @xmath16 for which the mean @xmath14 deviation from the fit amounts to @xmath17 .
the slopes of both edges deviate from each other by about 10 % .
this means that the constant width in @xmath18 of the instability strip translates into a relation of only slightly variable width on the @xmath19
@xmath1 plane .
the zero - points of the pl relation shift by @xmath20 at @xmath21 .
it is worth noticing that the pl relation hardly changes even if the first crossing of the instability strip ( before the onset of core helium burning ) of the stellar models is included in the fits to the pl relation .
when computing the nonadiabatic fundamental modes for our stellar models we also obtained the corresponding pulsation constants @xmath22 .
this quantity was frequently used in the past in qualitative semi - analytical arguments on the pl relation of cepheids in the literature , therein it was usually treated as a constant . since stellar evolution during the cepheid phase can not be mimicked by simple homology transformations between different evolutionary phases within the instability strip we quantify its variation along the instability strip and for the masses and luminosities of cepheids .
figure 3a displays the nonadiabatically derived pulsation constants as a function of effective temperature of the blue - edge and the red - line , respectively .
evidently , @xmath0 is not constant .
connecting open and filled symbols belonging to the same branch of an evolutionary track , we find @xmath0 to increase slightly when a star of a given mass crosses the instability strip from blue to red .
the largest variation of @xmath0 , by roughly 50 % , occurs upon increasing luminosity along the instability strip .
as already observed in the pl relation , also the @xmath0-values do not depend on particular choices of metallicity or helium abundance . when plotting the @xmath0-values versus period the blue - edge
red - line offset which is seen in fig .
3a reduces considerably ( cf . fig .
. only a slight vertical shift of at most 0.005 in @xmath0 remains at a given period between blue - edge and red - line data .
the following quadratic fits the data satisfactory : @xmath23 ^ 2\;. \eqno ( 5)\ ] ] our theoretical data delineating the positions of the fbe , the red line of the strip , and the associated pl relation allow the computation of lines of constant period along the instability strip on the hr diagram .
since we restricted ourselves to the computation of periods at the fbe and along the red line , our loci of constant period are straight .
this must , however , not necessarily be so .
we found that the slope is a function of the position
parameterized by the period in the instability strip .
it can be approximated by the following linear relation : @xmath24 the slope changes therefore by 20 % when going from @xmath25 day to @xmath26 days .
the existence of the slope change is equivalent to the fact that the strength of the temperature dependence of q - value depends on period as seen in fig .
if we integrate equation ( 6 ) , we see that the coefficient of the color term in a period - luminosity - color ( plc ) relation should be a function of period .
when plc relations are derived from observational data , however , it is always assumed that the coefficient of the color correction term is independent of period .
this assumption is , strictly speaking , inconsistent with our theoretical results .
we are not aware of any attempt to take this effect into consideration and we hypothesize that this might be the reason for the failure of establishing a plc relation for galactic cepheids ( cf .
fernie 1990 ) .
according to our evolution / stability computations , incorporating new opal opacity data and low - temperature extensions provided by alexander and ferguson ( 1994 ) we find _ no _ significant metallicity and no helium dependence of the @xmath27 relation of cepheids .
our results apply to bolometric quantities , however .
any observational realization of the pl relation such as e.g. a @xmath28 relation will depend to some degree on the the particular chemical composition of the cepheids through the metallicity sensitivity of the bolometric correction .
in fact , the metallicity dependence of stothers(1988 ) @xmath29 relation ( which is weak though ) mostly comes from the bolometric correction .
the magnitude of the metallicity dependence of the bolometric correction is dictated by the choice of the filter passband .
the quantification of this effect is not the purpose of this paper and has to be addressed by the theory of stellar atmospheres .
for the blue - edge and the red - line data , the slopes of the pl relation are 1.270 and 1.244 , respectively .
our fbe slope is similar to what chiosi et al .
( 1993 ) ( 1.280 ) found but larger than the 1.19 which was given as a mean value by stothers ( 1988 ) . from observational data ,
cox ( 1980 ) derived a slope of the bolometric pl relation of 1.15 .
this pl relation was based on bolometrically corrected @xmath30 data .
stothers ( 1988 ) refers to the infrared pl relation which should be close to the bolometric one and cites slopes between 1.16 and 1.39 .
we conclude from this that the slopes of our pl relations lie within the range of previous theoretical ones .
therefore , the latest improvements in opacities do not influence the _ slope _ of the cepheid pl relation .
compared with chiosi et al.s ( 1993 ) mild overshooting case our zero - point ( measured at @xmath31 days ) is brighter by 0.02 in @xmath32 .
the particular treatment of overshooting influences mainly the ml relation . compared with the ml relation the luminosity shift
due to overshooting is about a factor of five less in the pl relation .
the slope of the pl relation is even more stable with regard to changes in the treatment of overshooting . in view of the experience of chiosi et al .
( 1993 ) we did not further elaborate here on computing series of stellar models with different overshooting parameterizations .
the present physical calibration is slightly brighter than observational ones . using galactic cepheids as calibrators gives for a ridge - line cepheid with @xmath33 d @xmath34 ( sandage and tammann 1968 ) .
based on an adapted lmc modulus of ( @xmath35 ) = 18.57 , madore and freedman ( 1991 , 1997 ) find a corresponding value of @xmath36 .
using stellar radii as calibrators di benedetto ( 1997 ) finds @xmath37 . a 10-days cepheid with @xmath38
has a bolometric correction of bc@xmath39 ( lang 1991 ) . with this , one obtains @xmath40 .
our physical calibration gives for a corresponding cepheid on the ridge line ( half - way between the fbe and our red line ) @xmath41 .
the steeper slopes of the theoretical pl relations and the shifts in the zero points , say at 10 days period , could be understood by noticing that the cepheid population of the instability strip is not uniform .
fernie ( 1990 ) showed evidence that cepheids favor to populate regions close to the blue - edge at lower luminosities and close to the red line a higher luminosities . a ridge line determined from an arbitrary sample of cepheids neglecting their inhomogeneous distribution across the instability strip
will lead to a pl relation with a shallower slope than what results from an instability strip with a homogeneous distribution of variable stars . if this inhomogeneity is not accounted for properly in observational data a shallower ` ridge - line ' pl relation results when compared with the theoretically predicted one .
finally we quantified the variation of the pulsation constant @xmath0 across and along the instability strip as far as it is relevant for cepheid studies .
the variation is considerable as a function of luminosity : going from 4 to 10 @xmath3 , @xmath0 changed by roughly 50 % .
an appropriate fit was presented in sect .
an important result is that the @xmath0 values change very little as helium or heavy - element abundances are changed .
furthermore , we showed how the lines of constant period change their slopes ( when they are assumed to be straight lines across the instability strip ) as a function of period .
the fit to the data allowed a quantification of the effect so that it could be incorporated into semi - analytical discussions of pulsation properties of cepheids .
acknowledgments : we are grateful to g.a .
tammann and a.r .
sandage for encouraging this study .
acknowledges financial support from the japanese society for promotion of science and the swiss national science foundation .
was financially supported by the swiss national science foundation through a profil2 fellowship .
alexander , d.r . , and ferguson , j.w .
1994 , , 437 , 879 becker , s.a . , iben , i. jr . , and tuggle , r.s .
1977 , , 218 , 633 benvenuti , p. , macchetto , f.d . , and schreier , e.j .
1996 , science with the hubble telescope - ii , proceedings of paris workshop buchler , j.r .
, moskalik , p. , and kovcs g. 1990 , , 351 , 617 cox , j.p .
1980 , theory of stellar pulsation , princeton : princeton univ .
press choisi , c. , wood , p.r . , and capitanio , n. 1993 , , 86 , 541 deupree , r.g .
1980 , , 236 , 225 di benedetto , g.p .
1997 , , in press feast , m.w . , and catchpole , r.m .
1997 , , 286 , l1 fernie j.d .
1990 , , 354 , 295 gieren w.p . , and fouqu p. , 1993
, in new perspectives on stellar pulsation and pulsating variable stars , iau coll .
139 , eds .
j.m . nemec and j.m .
matthews , cambridge : university press , p. 72
iben , i. jr . , and renzini , a. 1984 , phys .
, 105 , 329 iben , i. jr .
, and tuggle , r.s .
1972 , , 178 , 445 kippenhahn , r. , and weigert a. 1990 , stellar structure and evolution , berlin : springer lang , k.r .
1991 , astrophysical data : planets and stars , new york : springer , p.142 madore , b.f . , and freedman , w.l .
1991 , , 103 , 933 madore , b.f . , and freedman , w.l .
1997 , , in press moskalik , p. , and buchler , j.r . , marom , a. 1992 , , 385 , 685 saio h. , winget d.e . , and robinson , e.l .
1983 , , 265 , 982 sandage a.r .
1996 , , 28 , 52 sandage a.r . , and tammann g.a .
1968 , , 151 , 531 sandage , a.r . , and tammann , g.a .
1997 , , submitted sekiguchi , m. , and fukugita , m. 1997 , submitted to observatory stothers , r. 1988 , , 329 , 712 | the theoretical period - luminosity relation for cepheids was deduced from stellar evolution and pulsation computations .
questions concerning the metallicity dependence of the relation as well as the constancy of the pulsation ` constant ' @xmath0 in the instablility strip were addressed .
we found no significant metallicity dependence of either the slope or of the zero - point of the period - luminosity relation .
the @xmath0-value , however , varies across the instability strip , in effective temperature as well as in luminosity ; its metallicity dependence , on the other hand , is marginal . |
in general relativity , as well as in any metrical theory of gravitation of some generality and scope , a common approach to cosmological modelling commences with a four - dimensional space - time manifold ( representing the physical events ) endowed with a lorentzian metric ( necessary to ensure the local validity of the well established special relativity theory ) . representing the physical phenomena in this manifold we have fields satisfying appropriate local differential equations ( the physical laws ) .
finally , fields and geometry are coupled according to the corresponding gravitational theory one is dealing with . the space - time geometries arising as solutions of the gravitational field equations constrain to some extent the dynamical behavior of the physical fields
this eminently metrical approach to model the physical world has led a number of physicists to implicitly ( or explicitly ) restrict their studies to purely geometric features of space - time , either by ignoring the role of topology or by considering just a limited set of topological alternatives for the space - time manifold .
if , on the one hand , the topological properties of a manifold antecede and are more fundamental than the metrical features and the differentiable structure on which tensor analysis is based upon , on the other hand , it is well known that geometry constrains but it does not dictate the topology of a space - time .
it is therefore important to determine whether and ( or ) to what extent physical results concerning a space - time geometry depend on or are somehow constrained by the topology of the underlying manifold . in a recent work @xcite we studied the role played by topology in the time evolution of the energy @xmath0 of a radiating system in static flat frw space - time manifolds
@xmath1 whose @xmath2 spacelike sections @xmath4 are endowed with different topologies , namely the simply - connected euclidean manifold @xmath5 , and six topologically non - equivalent flat multiply - connected orientable _ compact _
3-manifolds @xcite
clearly these flat space - time manifolds @xmath1 are orientable and decomposable into @xmath8 .
the radiating system we have examined , which has been used to study a wide class of oscillating phenomena @xcite
@xcite , is formed by a pointlike harmonic oscillator ( energy source ) coupled with a massless scalar field ( scalar radiation waves propagating at speed of light ) . through a _ numerical _ integration of the time evolution equation for the harmonic oscillator ( with the topological constraints suitably considered ) , we have shown that there is an exponential damping of the energy @xmath0 of the harmonic oscillator when the spacelike section @xmath4 is the euclidean space @xmath5 , whereas when @xmath4 is endowed with one of the six compact and orientable flat topologies the energy @xmath0 exhibits a loosely speaking reverberation pattern , with discontinuities in the derivative of @xmath0 and relative minima and maxima ( both the discontinuities and the extrema are due to the travelling waves which `` reflect '' from the `` topological walls '' ) , followed by a growth of the energy with the time .
we have also shown that , for these six compact cases , the energy @xmath0 diverges exponentially when @xmath9 , in striking contrast with the damping of @xmath0 in the @xmath5 case .
this unexpected divergent behavior of the energy for flat space - times with compact spacelike sections illustrates that ( _ totally _ ) compact topologies may give rise to rather important dynamic changes in the behavior of a physical system .
this type of sensitivity has been referred to as topological fragility and can occur without violation of any local physical law @xcite . in the present work
we extend these investigations by performing a rigorous _ non - numerical _ study of the behavior of the same physical system under a less restrictive topological setting , namely we shall consider static flat space - time manifolds @xmath1 whose @xmath2 spacelike sections @xmath3 are compact in at least one spatial direction .
moreover , no assumption will be made as to whether or not these 3-manifolds are orientable .
this amounts to saying that the results of the present work hold for all flat space - time manifolds whose @xmath2 spacelike sections are endowed with one of the 17 flat topologies discussed by wolf @xcite , ellis @xcite , and others @xcite ( see also gomero @xcite ) .
it should be stressed from the outset that in spite of being the same physical system discussed in @xcite , the present work generalizes the results of that article in several respects .
firstly , the underlying topological setting is now much more general in that here we do not assume that the spacelike 3-manifolds @xmath4 are compact and orientable .
secondly , contrarily to the _ numerical _ integration performed in @xcite , in the present paper we have obtained a closed _ exact _ solution of the evolution equation for the harmonic oscillator in the above - mentioned general topological setting .
thirdly , the case study we have examined in section [ case ] , where we obtain the explicit solution for the radiation reaction equation and discuss in details the graphs which reveal the time evolutions of @xmath0 and @xmath6 , is completely new .
finally here we have also performed a lengthy discussion on the conservation and on the balance of the total energy of our physical system , making apparent how ( here as well as in ref .
@xcite ) the total energy of the whole system is conserved for any time @xmath10 .
the plan of this article is as follows .
we derive ( section [ radreaceq ] ) the time evolution equation for the harmonic oscillator ( also referred to as radiation reaction equation ) when @xmath11 is endowed with a non - trivial flat topology , i.e. , when the @xmath12 spacelike sections are any multiply - connected ( compact in at least one direction ) flat manifold @xmath4 .
it turns out that the evolution equation for this case is formally the same as that obtained by bernui _
et al . _
@xcite for the case where @xmath4 is any ( totally ) compact and orientable 3-manifold .
however , they differ in the non - homogeneous term that accounts for the difference in the topology distinct degrees of connectedness of the spacelike sections give rise to different non - homogeneous terms .
we also emphasize in section [ radreaceq ] that since our system is composed only of test fields in static flat space - time backgrounds , there is no need to consider the back - reaction of the fields on the geometry .
we present in section [ sol ] an _ exact _ solution of the evolution equation for the harmonic oscillator in flat space - time whose @xmath2 spacelike sections are any multiply - connected flat 3-manifold . through a heuristic analytical reasoning
we also show in section [ sol ] that this exact solution @xmath13 and the corresponding energy of the harmonic oscillator @xmath0 both exhibit a divergent _ asymptotical _ exponential behavior .
clearly the energy can be limitlessly `` extracted '' from the interaction term in the multiply - connected cases only because the system permits this extraction , since it is not bounded from below .
but as we shall discuss in this work it is indeed the topology that `` excites '' this available `` physical mode '' of our system .
the balance and conservation of the total energy of our system is also discussed in section [ sol ] , where we show that it is finite and conserved for all finite time @xmath10 .
explicit exact solution of the evolution equation for the case where @xmath14 is discussed in details in section [ case ] .
for this special case we also present and analyze graphs which reveal how the amplitude @xmath13 and the energy @xmath0 of the _ overdamped _ harmonic oscillator vary with the time .
the graph of @xmath0 presents a reverberation pattern with relative minima and maxima , and discontinuities of the derivative @xmath15 .
the topological origin of the reverberations for the overdamped oscillator is also made clear and stressed in that section .
the graphs in this section also confirm ( within the limit of accuracy of the plots , of course ) the divergent asymptotical exponential behavior for both @xmath13 and @xmath0 , which is shown to take place in section [ sol ] .
these results make apparent that , contrarily to what one might infer from @xcite , it is not necessary to have flat space - time manifolds with ( totally ) compact spacelike sections @xmath4 for the topological induction of the reverberations and ( or ) the divergent behavior of the energy @xmath0 the compactness of @xmath4 in just one direction suffices .
we begin the section [ finals ] by emphasizing the importance of topological considerations in physical problems , and give an example that makes explicit how the overall electric charge of universes whose @xmath12 spacelike sections are orientable and compact flat 3-manifolds is related to the topology .
we also show that topological fragility can arise not only in the usual cosmological modelling , but also in the ordinary multiply - connected flat space - time manifolds we consider in the present paper .
there we stress the origin of the topological reverberations we found , and show that it constitutes an example of topological fragility , which takes place without violation of any local physical law .
the extent to which this paper generalizes ref .
@xcite is discussed in details in section [ finals ] .
finally the relation of our results with cosmology is also addressed .
the flat space - time manifolds we shall be concerned with are decomposable into @xmath8 , where the real line @xmath16 represents a well defined global time of minkowski space - time .
corresponding to the possible topologically distinct 3-manifolds @xmath4 there exists a simply - connected covering manifold @xmath17 , such that each manifold @xmath11 is obtained from @xmath18 by identifying points which are equivalent under the action of a discrete subgroup of isometries of the euclidean space @xmath5 without fixed points . in other words ,
each manifold @xmath19 is obtained by forming the quotient space @xmath20 , where @xmath21 is a discrete group of isometries of @xmath18 acting freely on @xmath17 and referred to as the covering group of @xmath4 ( for a complete classification of euclidean 3-manifolds see wolf @xcite ) .
the physical system we shall be concerned throughout this paper is represented by a pointlike harmonic oscillator coupled with a scalar field .
this system has been treated by many people , since the work by schwalb and thirring @xcite in 1964 , as a simplified model to study a few features of oscillating ( radiating ) phenomena @xcite @xcite .
when the underlying manifold is @xmath5 this system may serve as a model for an electric dipole coupled to electromagnetic radiation or an impurity atom interacting with acoustical waves @xcite .
it has also been used as simple model to parallel classical electrodynamics in the dipole approximation as discussed by kampen @xcite and kramers @xcite .
it is worth mentioning that similar systems have also been used to mimic the basic properties , and thus to study the most relevant features of pulsating stellar systems @xcite @xcite .
the essential idea that permeates these latter papers is to work with simple models of oscillating fields interacting with oscillating sources in order to develop some analytical physical background to understand the relationship between gravitational waves and their sources ( radiation emission and radiation reaction ) .
consider that the harmonic oscillator of our physical system is located at an arbitrary point @xmath22 . without loss of generality
, one can assume that the covering map @xmath23 maps the origin to @xmath24 , namely @xmath25 . in what follows
we shall denote by @xmath26 the orbit @xmath27 of @xmath28 under the action of @xmath21 on @xmath5 ( see fig . 1 ) .
now , the time evolution of the harmonic oscillator at @xmath22 can be obtained by studying the equivalent system on the covering manifold , which is formed by an infinite set of indistinguishable harmonic oscillators each one located at a point of the orbit @xmath29 , subject to identical sets of initial conditions and interacting with the scalar field @xmath30 in the same way as that of the original oscillator at @xmath22 ( see fig . 1 ) .
thus , the dynamics of our physical system in @xmath4 can be derived from the following functional action for the equivalent system in the universal covering space @xmath17 : @xmath31 where @xmath32 is a scalar field term , @xmath33 corresponds to the oscillators , and @xmath34 is a coupling term between the scalar field and each harmonic oscillator .
these three terms are , respectively , given by @xmath35 \;,\label{so } \\
s_i & = & \lambda \sum_{\vec{p } \,\in\ , { \cal o}_q}\ , \int d^4x \,\ , \rho_{\vec{p}}\,(\vec{x})\,\ , \theta\,(t)\,\ , \varphi(t,\vec{x})\,\ , q_{\vec{p}}\,(t ) \;,\label{si}\end{aligned}\ ] ] where @xmath36 , @xmath37 is the massless scalar field , @xmath38 is the amplitude of the oscillator located at @xmath39 , overdot represents derivative with respect to the time @xmath40 , @xmath41 is the angular frequency of each oscillator and @xmath42 is the coupling constant .
hereafter , @xmath43 is a non - negative normalized density function ( @xmath44 ) with compact support ( corresponding to the region of interaction ) contained in the interior of a fundamental domain of @xmath4 and centered at the origin , and finally @xmath45 , where @xmath46 and @xmath47 [ note that @xmath48 .
the step function @xmath49 is used to indicate that the interaction starts at @xmath50 .
it should be noticed that the interaction term ( [ si ] ) clearly is not bounded from below , potentially permitting a limitless extraction of energy by the oscillator . however , as we shall discuss later in this article , it is the topology of the 3-space that `` excites '' this physically available mode of the system .
a word of clarification is in order here . in dealing with the behavior of physical fields in the curved space - times of general relativity one has to consider that these fields are not only influenced by , but they also have gravitational effects , i.e. , they change the geometry . in most ordinary situations , however , these back - reaction effects are small enough that can be neglected . in such cases the fields
are treated as _ test _ fields in the corresponding curved space - time background .
had we considered our fields @xmath13 and @xmath51 as _ non - test _ fields in a _ non - static curved _ space - time we would have to take into account the back - reaction of the fields on the geometry , and so besides @xmath52 a new interaction term would have to be added to the action ( [ act ] ) .
however , our backgrounds are _ static flat _ space - times ( endowed with any one of the 17 possible non - trivial flat topologies ) , and our system contains only _ test _ fields with a mutual interaction , thus there is no place for such a term in the action of our system . varying the action ( [ act ] ) with respect to @xmath53 and @xmath54 one obtains the coupled equations of motion of the equivalent system , namely @xmath55 where @xmath56 denotes the dalembertian operator on @xmath5 .
note that the sum in ( [ eqmot1 ] ) is not an infinite sum . actually owing to the disjoint
compact supports of the functions @xmath57
the right - hand side of that equation contains at most one summand which does not vanish . to obtain the radiation reaction equation from the above equations ( [ eqmot1 ] ) and ( [ eqmot2 ] ) we recall that the general solution of eq .
( [ eqmot1 ] ) , considered as an initial value problem , can be written in the form @xmath58 , where @xmath59 satisfies the corresponding homogeneous equation , and where @xmath60 indicates the solution of the inhomogeneous equation . from now on we shall assume , for simplicity , that the whole energy of the system is initially stored in the oscillator .
accordingly , we shall use as initial conditions for the equations ( [ eqmot1 ] ) and ( [ eqmot2 ] ) that @xmath61 and that @xmath62 , @xmath63 , where @xmath64 and @xmath65 are real arbitrary constants .
these conditions imply that @xmath66 , and therefore @xmath67 .
now since the interaction begins at time @xmath50 , the green function for the dalembertian operator in ( [ eqmot1 ] ) is clearly given by , or on the covering space @xmath5 . on @xmath11 , as it has been considered in @xcite , it has to be invariant under the covering transformations .
on @xmath5 , however , which we are considering in this section , the invariance under @xmath21 must not be imposed .
note additionally that the covering transformations have been taken into account in eqs .
( [ so ] ) ( [ eqmot2 ] ) and in the subsequent equations ( [ setrreq ] ) ( [ rreq2 ] ) . ]
@xmath68 ) can be formally written as @xmath69 where we have used @xmath70 as dummy index instead of @xmath71 . finally , inserting ( [ solphi ] ) into equation ( [ eqmot2 ] ) one easily obtains the radiation reaction equations for the oscillators , namely @xmath72
as the infinite set of identical harmonic oscillators ( each one located at a point of @xmath26 ) are subject to the same set of physical constraints , the amplitudes @xmath73 must evolve identically .
thus , the set of radiation reaction equations ( [ setrreq ] ) reduces to just one differential equation , namely @xmath74 which holds for each oscillator in @xmath75 , and whose solution @xmath13 gives the time evolution for the amplitude of the harmonic oscillator in @xmath4 .
one might think at first sight that the sums on the right hand side of eqs .
( [ solphi ] ) ( [ rreq ] ) are divergent , and a regularisation scheme is needed to cope with the divergences .
however , this is not true because for any finite time @xmath76 the sums in those equations are in fact finite owing to the disjoint compact supports of the functions @xmath57 together with the cut - off effects of the step function @xmath77 .
in other words , we do not have formal series to be carefully handled , but finite expressions . to deal with the pointlike coupling between the harmonic oscillator and the scalar field [ @xmath78
, we shall consider an infinite family of oscillator field systems ( @xmath79-system , for short ) , with the same coupling constant @xmath80 , and such that each element of the @xmath79-system is characterized by a _ bare _ angular frequency @xmath81 and density function @xmath82 , satisfying the generalized aichelburg - beig condition @xcite , namely latexmath:[\[\label{dampcond } \omega_n^2 - \frac{\lambda^2}{4 \pi }
\int{d^3x \ , d^3y \ , \frac{\rho_n(\vec{x } ) \rho_n(\vec{y}\,)}{| \vec{x } - \vec{y}\ , constant for each member of the @xmath79-system , and @xmath84 is a @xmath85-sequence in the sense of distributions @xcite .
it should be noticed that the inequality ( [ dampcond ] ) is a sufficient condition for having radiation damping of the energy @xmath0 in the pointlike coupling limit case when the oscillator - field system lies in the minkowski space - time manifold @xmath86 ( see , for example , ref .
@xcite ) . now ,
since eq .
( [ rreq ] ) holds for each member of the @xmath79-system , using relation ( [ dampcond ] ) one finds @xmath87 where we have set @xmath88 , @xmath89 and @xmath90 . in the limit @xmath91 ( see appendix ) we have that the following equation holds for all @xmath92 : @xmath93 where @xmath94 , @xmath95 is the number of points @xmath96 such that @xmath97 , the sequence @xmath98 is ordered by increasing values , and clearly @xmath99 when @xmath100
. it should be noticed that the right - hand side of the radiation reaction equation ( [ rreq2 ] ) formally contains an infinite countable number of retarded terms of topological origin .
however , due to the @xmath101 function , at any given finite time @xmath40 only a finite number of terms will contribute to the sum of ( [ rreq2 ] ) .
these terms are different for distinct multiply - connectedness of the 3-manifolds @xmath4 , and vanish for the simply - connected manifold @xmath17 . loosely speaking they give rise to
`` reflected '' waves from the `` topological walls '' , and thus account for both the extrema of @xmath0 and the discontinuities of @xmath15 . in other words
they are responsible for the topological reverberations .
clearly for the simply - connected manifold @xmath17 , there is an exponential decay of @xmath0 and no reverberation takes place in this case .
in this section we shall present an exact recursive solution of the radiation reaction equation ( [ rreq2 ] ) , which holds for any multiply - connected flat manifold @xmath4 . to this end
, we first consider the differential operator @xmath102 which acts on real - valued piecewise smooth functions @xmath103 according to @xmath104 it is straightforward to show that if @xmath105 is some integrable function , then the solution of the equation @xmath106 with initial conditions @xmath107 is @xmath108 with @xmath109 and @xmath110 in what follows we shall discuss in detail how to find the solution of ( [ rreq2 ] ) for the underdamped case , in which @xmath111 .
the solution for the remaining two cases , however , can be similarly worked out , and for the sake of brevity we will present later only the final results .
our purpose now is to find a real - valued function @xmath112 of class @xmath113 such that @xmath114 and @xmath63 and satisfying ( [ rreq2 ] ) , which in terms of the operator @xmath115 can be rewritten as @xmath116 where the @xmath117 are positive real numbers and @xmath98 is a strictly increasing sequence of positive real numbers ( @xmath118 ) , such that @xmath119 when @xmath100 .
thus , for @xmath120 and taking into account ( [ deq ] ) and ( [ nhomsol ] ) the solution of ( [ rreq3 ] ) can be written as @xmath121 where the function @xmath122 must satisfy the homogeneous equation @xmath123 with the conditions @xmath124 , @xmath125 .
clearly one can rewrite the solution ( [ impsol1 ] ) in the form @xmath126 where , from ( [ deq ] ) and ( [ nhomsol ] ) , the function @xmath127 fulfils the differential equation @xmath128 with the initial conditions @xmath129 .
the solution ( [ impsol2 ] ) is given in an implicit form .
an explicit form can be obtained , though . indeed , as we are concerned only with real - valued functions defined for @xmath130 , if we recursively define functions @xmath131 by @xmath132 with the conditions @xmath133 for @xmath134 , and functions @xmath135 by @xmath136 with @xmath137 for @xmath138 , then we can show by induction that @xmath139 notice that the first of these equations ( for @xmath140 ) is nothing but equation ( [ impsol2 ] ) . an explicit form for the solution of ( [ rreq3 ] ) can now be obtained as follows .
equations ( [ imps ] ) and ( [ impsn1 ] ) can be integrated to give @xmath141 where @xmath142 , @xmath143 , and where obviously we have used ( [ opeq2 ] ) ( [ hfun ] ) .
on the other hand , if we define @xmath144 then the explicit form of the solution of equation ( [ rreq3 ] ) is simply given by @xmath145 it should be noticed that since the above solution is ultimately given in terms of @xmath146 , equations ( [ exps ] ) ( [ expsol1 ] ) make apparent that @xmath13 is completely determined by the pair of initial conditions [ @xmath147 as one would have expected from the outset . another important point to be stressed regarding the above solution is that the right - hand side of ( [ expsol1 ] ) is not an infinite sum as it appears at first sight .
actually , for a given finite time @xmath148 the above sum will clearly be truncated by the presence of the step function [ @xmath149 for @xmath150 , @xmath151 for @xmath92 ] , i.e. only those @xmath153 with @xmath154 will contribute to the sum ( [ expsol1 ] ) . moreover
, each @xmath153 as given by ( [ qieq ] ) is itself a finite sum , again due to the cut - off effects of the step function @xmath155 . before proceeding to the discussion of the asymptotical behavior of @xmath13 we mention that the solution of ( [ rreq2 ] ) for the critically damped ( @xmath156 ) and overdamped ( @xmath157 ) cases
is again given by ( [ expsol1 ] ) with ( [ qieq ] ) , but now with the functions @xmath122 and @xmath158 given by @xmath159 where @xmath160 and @xmath161 are real constants , which are determined by the initial conditions , and the functions @xmath162 are given by equations ( [ hfun ] ) .
now , for any non - trivial flat topology the asymptotic behavior of the solution @xmath13 of equation ( [ rreq3 ] ) can be figured out by a procedure similar to that used in @xcite .
indeed , if one _ ad hoc _ assumes an exponential asymptotical behavior for @xmath54 of the form @xmath163 with @xmath164 and @xmath165 real constants , then in the limit @xmath166 equation ( [ rreq3 ] ) reduces to @xmath167 this equation has only one real solution for @xmath164 .
indeed , the right - hand side of ( [ asingen ] ) is a positive monotone decreasing function @xmath168 with @xmath169 and @xmath170 . thus @xmath168 lies entirely in the first quadrant of the plane and crosses it from the top - left to the bottom - right ( see fig . 2 ) .
now , since @xmath171 then for a given pair ( @xmath172 ) the left - hand side of ( [ asingen ] ) is a parabola curved upwards with vertex at @xmath173 .
therefore , it always intersects the curve for @xmath168 in just one point , which is in the first quadrant ( see fig . 2 ) .
this amounts to saying that there is only one real number @xmath174 solution to equation ( [ asingen ] ) .
hence the amplitude @xmath13 , and consequently the energy of the oscillator @xmath175 $ ] , both exhibit an asymptotical exponentially divergent behavior ( confirmed , within the limits of accuracy of the plots , by figures 3 and 4 for the special case of an overdamped oscillator discussed below ) .
one can sum up by stating that the compactification in just one direction is sufficient to topologically induce a asymptotic divergent behavior of @xmath0 .
a question which naturally arises here is whether the total energy of our system is conserved .
as our system contains only mutually interacting test fields @xmath176 and @xmath177 in static flat background , the energy of the overall system must be conserved .
indeed , for @xmath10 we have @xmath178 and the action ( [ act ] ) ( [ si ] ) does not depend explicitly on the time .
thus , for any well - behaved density function @xmath179 the total energy of our system @xmath180 clearly is conserved , which squares with the fact that our system is a test system in a static background .
for the sake of completeness , in what follows we shall show in more details how the total energy of our system @xmath180 is finite and conserved for all time @xmath36 even in the pointlike limit case @xmath181 .
we shall also qualitatively discuss how the balance of @xmath180 can be handled . from the functional action ( [ act ] )
( [ si ] ) the energy @xmath180 of our system can clearly be written in the form @xmath182 where @xmath183 ^ 2\,\}\ ; , \label{ef } \\ { \cal e}_{o}(t ) & = & \frac{1}{2}\;[\,\dot{q}^2\,(t ) + \omega^2\,q^2\,(t)\ , ] \;,\label{eo } \\ { \cal e}_i(t )
& = & - \lambda\;\ , q(t ) \int_{{\cal d } } d^3x \ ; \rho(\vec{x } ) \ , \varphi(t,\vec{x } ) \ ; , \label{ei}\end{aligned}\ ] ] for @xmath184 . here
@xmath185 denotes a fundamental domain of @xmath4 . as we have mentioned above for any time @xmath186 the energy @xmath180 is conserved since the corresponding hamiltonian does not depend explicitly on the time . from equation ( [ solphi ] ) the interaction component @xmath187 , for which the field @xmath51 is evaluated at @xmath188 , i.e. in the location of the pointlike oscillator
, clearly presents a divergent behavior of the form @xmath189 in the neighborhood of @xmath188 ( see below for details ) .
now , to deal with the balance of the total energy @xmath180 , we introduce the divergent quantity @xmath190 in terms of which a finite energy @xmath191 can be defined for all @xmath192 , namely @xmath193 thus , for each time @xmath40 in that interval , clearly one can extract from @xmath187 an infinite amount of negative energy to obtain a regularized finite negative energy @xmath191 .
the energy @xmath194 as given by ( [ ef ] ) also diverges .
in fact , from ( [ solphi ] ) one finds that in the neighborhood of @xmath188 the scalar field @xmath51 takes the form @xmath195 where @xmath196 denotes the non - diverging terms for any @xmath184 .
so , in the neighborhood of @xmath188 we have @xmath197 , and the integral corresponding to this term in eq .
( [ ef ] ) , namely @xmath198 , has a divergent contribution of the form @xmath199 when @xmath200 .
now since the contribution of the term @xmath201 to the integral on the right - hand side of ( [ ef ] ) is finite for all @xmath184 , one can similarly introduce a regularized finite energy @xmath202 by @xmath203 which is clearly finite for all finite time @xmath204 .
the energy component corresponding to oscillator @xmath205 as given by ( [ eo ] ) can be dealt with by using the frequency @xmath206 given by eq .
( [ dampcond ] ) in the limit case @xmath181 , namely using @xmath207 . indeed , with this renormalized frequency one can define @xmath208 where the regularized energy @xmath209 $ ] is finite for all finite time @xmath10 . the energy of the oscillator @xmath210 has also been denoted simply by @xmath0 in many places of this paper . finally , we note that according to eqs .
( [ eif ] ) , ( [ eff ] ) and ( [ eof ] ) , the divergent terms of the form @xmath211 cancel out in the sum @xmath212 .
thus , from ( [ tenergy ] ) one obtains that the total energy of the system @xmath180 given by ( [ tenergy ] ) reduces to @xmath213 which is constant and finite for all time @xmath10 .
the fact that the total energy of the system @xmath214 is finite and conserved for all time @xmath10 is undoubtedly the most important physical requisite our system obeys .
however , it is the fact that the interaction energy is not bounded from below which makes possible and understandable the asymptotical divergent behavior of the energy of the oscillator and of the scalar field , since they both can `` extract '' energy from the interaction term .
remarkably this latent degree of freedom of our system is excited only when it lies in the multiply - connected manifolds , since when it is in @xmath5 there is a damping of energy of the oscillator . in other words ,
the topology ( multiply - connectedness ) excites this available `` physical mode '' of our system .
this remarkable feature of this system has not been perceived since the sixties , when it was first examined by schwalb and thirring ( see , for example , refs .
@xcite
@xcite ) .
it is certainly desirable that quantum physical systems have the interaction energy bounded from below .
however , at a classical level this is not an imperative .
when one considers , for example , a system of two isolated pointlike charges with opposite signs in a flat space ( vacuum medium ) , the smaller is the distance @xmath215 between the charges the greater is the absolute value of potential energy of the system .
even if one assumes that the charges have already been renormalized , in the limit @xmath216 the potential energy formally diverges .
this sort of divergences are formal , since the limit @xmath216 can not be attained in practice .
furthermore , one ought to bear in mind that the classical theory used to infer this divergence does not hold for every small @xmath215 ( compared with , e.g. , the bohr radius ) .
actually , the distance @xmath215 is assumed to be larger than the bohr radius since otherwise the self - fields of the two interacting charges would be rather distorted , and one should expect much more involved physical effects .
similarly in our system the energy of the oscillator @xmath217 is finite for all finite @xmath76 , but formally diverges in the limit @xmath218 .
this limit case may perhaps be excluded on physical grounds
. however , the most remarkable point here is that a well - behaved physical system when it lies in the ordinary simply - connected euclidean manifold @xmath5 , will exhibit an unexpected behaviour ( reverberation pattern followed by a growth of @xmath0 ) when it is in any possible flat manifold with non - trivial topology , without violating any local physical law .
in this section we shall focus our attention on the time evolution of the harmonic oscillator in a flat space - time whose @xmath12 spacelike section is the simplest multiply - connected euclidean 3-manifold , namely @xmath219 . in this case , denoting by @xmath220 the distance between two equivalent points in the covering manifold @xmath5 , one can easily find that @xmath221 , where @xmath222 ( @xmath223 stands for a positive integer ) .
hence , from ( [ qieq ] ) and for @xmath224 one obtains the solution @xmath225 which together with ( [ expsol1 ] ) and the first eq .
( [ qieq ] ) give the behavior of the amplitude @xmath13 and of the energy @xmath175 $ ] for the following different types of oscillator : ( i ) the underdamped ( @xmath226 ) for which @xmath122 and @xmath158 are given by ( [ exps ] ) and ( [ expsn1 ] ) ; ( ii ) the critically damped ( @xmath156 ) where ( [ scd ] ) and ( [ sncdod ] ) furnish @xmath122 and @xmath131 ; ( iii ) the overdamped ( @xmath227 ) for which @xmath122 and @xmath158 are given by ( [ sod ] ) and ( [ sncdod ] ) .
as far as the asymptotical behavior of @xmath13 for this special case ( @xmath228 ) is concerned , the right - hand side series of ( [ asingen ] ) can easily be evaluated and equation ( [ asingen ] ) becomes @xmath229 clearly , for a specific oscillator , i.e. for a given pair ( @xmath172 ) , this equation can be numerically solved for @xmath164 by using , for example , a computer algebra system such as maple @xcite .
for an overdamped oscillator with @xmath230 and @xmath231 , which we shall use in our numeric calculations , taking @xmath232 and using maple one easily obtains the approximate solution @xmath233 . in the remainder of this section
we shall discuss the figures 3 and 4 and make some further remarks .
we begin by emphasizing that we have chosen to discuss the dynamical behavior of both the amplitude and the energy of the oscillator in a space - time whose spacelike section is @xmath234 because this is the simplest multiply - connected flat manifold ( since it is compact in just one direction ) having therefore the lowest degree of connectedness for a flat 3-manifold @xcite .
this amounts to saying that the dynamical effects exhibited by the physical quantities in this manifold will also be present in any other multiply - connected flat 3-manifold . as a matter of fact , the greater is the degree of connectedness of @xmath4 the more reinforced will be these dynamical effets of topological nature .
the limiting simply - connected case @xmath5 will manifest no sign of such effects . a second point of general order which is worth noting is that we have decided to plot the graphs for an overdamped oscillator because we knew from the outset that the amplitude @xmath13 and the energy @xmath0 for this oscillator , when in @xmath17 , exhibit an exponential decay with no relative extrema .
this choice amounts to freezing the degrees of oscillations , and therefore rules out any sort of resonance , since clearly there is no frequency to be coupled with to resonate .
besides , our choice also makes easier and neater the comparison of the dynamical behavior of the oscillator in the simply and multiply connected cases . taking into account these considerations
we have plotted the figures 3 and 4 for an overdamped harmonic oscillator with @xmath235 , @xmath236 and with initial conditions @xmath237 and @xmath238 , in a flat space - time whose @xmath12 spacelike section @xmath4 is endowed with the topology @xmath239 , and where we have taken the distance @xmath232 in eqs .
( [ q_cilinder ] ) and ( [ asincil ] ) .
the analysis of these figures is given in what follows .
figure 3 shows the graph for the amplitude @xmath13 which exhibits relative maxima and minima followed by a growth of @xmath13 .
this time evolution of the amplitude sharply contrasts with the behavior of the amplitude @xmath13 for the same oscillator in @xmath5 , where neither relative extrema nor growth of @xmath13 take place .
this figure also contains the graph of the function @xmath240 , which within the accuracy of numeric calculations performed for the plots confirms an exponential form for the divergent behavior suggested by the graph of @xmath13 and shown to indeed take place in the previous section . in this graph
we have used @xmath241 , which is the root of eq .
( [ asincil ] ) for @xmath235 , @xmath236 and @xmath232 .
figure 4 shows the behavior of the energy @xmath0 which exhibits a few relative minima and maxima followed by an asymptotic divergent behavior . this divergent behavior is clearly in agreement with the behavior of @xmath13 when @xmath242 shown in figure 3 .
the time evolution of the energy for the oscillator in this multiply - connected manifold is in striking contrast with the behavior of the energy @xmath0 in the simply - connected case @xmath5 , where there is an exponential damping of @xmath0 with no relative extrema .
figure 4 also contains the graph of the derivative @xmath6 , exhibiting discontinuities at each @xmath243 ( @xmath223 stands for a positive integer ) , which again do not occur in the simply - connected case @xmath5 .
the graph of @xmath6 also reveals that @xmath0 has a finite number of extrema .
it should be also noticed that the derivative @xmath244 of the energy function exhibits discontinuities not only in the specific case handled in this section , but also in the general underlying topological setting of this work . indeed , from equation ( [ rreq2 ] ) one easily obtains @xmath245 \,.\ ] ] from this equation one has that the discontinuities occur at @xmath246 , i.e. they come about each time a new term @xmath247 is taken into acount in the right - hand side of eq .
( [ rreq2 ] ) . the relative extrema and asymptotic divergent behavior of @xmath0 as well as the discontinuities of @xmath15 , which occur only in the multiply - connected flat manifolds , are due to the retarded terms of topological origin in the evolution equation ( [ rreq2 ] ) .
these topological terms are different for distinct connectedness of the @xmath12 section @xmath4 of the flat space - time manifold , and vanish for the simply - connected case @xmath5 .
they are ultimately responsible for those features ( extrema , discontinuities and asymptotic behavior ) of @xmath0 . in brief , as it has been made apparent for the overdamped oscillator , both the suppression of the radiation damping and the reverberation pattern of @xmath0 in those multiply - connected flat space - times are of purely topological origin , and on the other hand , evince that our system is topologically fragile @xcite .
since the physical laws are usually expressed in terms of local differential equations the topological considerations may not be prominent at first sight .
nevertheless they are often necessary in many problems of physics . in general relativity and cosmology , which handles the dynamic and the global structure of space - time manifolds ,
the study of topological features is even more significant and acquires a dynamical meaning in a sense .
topological considerations are necessary in many other situations in physics . when one considers , for example , the electric fields produced by bounded sources , one quite often chooses as boundary condition that the field vanishes at spatial infinity .
this choice is possible and even convenient if the @xmath12 spacelike section of the minkowski space - time @xmath1 is the ordinary simply - connected euclidean manifold @xmath5 .
however , as far as multiply - connected spacelike three - spaces @xmath4 are concerned the choice of satisfactory boundary conditions for this problem is not as simple as that .
so , for example , in the exam of the electric field produced by an isolated point charge in any one of the six possible orientable compact ( multiply - connected ) topologically distinct euclidean 3-manifolds @xmath4 , clearly this condition at infinity can not be imposed . actually , it is easy to show by assuming the gauss s law that one can not have a net electric charge in these compact flat manifolds , making clear that the overall electric charge in these manifolds is related to the topology @xcite .
this example evinces that changes in the assumption of the space - time topology may induce important physical consequences even in the case of static flat space - time manifolds .
this type of sensitivity , which is also present in cosmological modelling @xcite , has been referred to as topological fragility and can clearly occur without violation of any local physical law @xcite . as a matter of fact , in the above example the topological fragility arises when we impose the local validity of an ordinary law of classical electromagnetism to a system which lies in a flat multiply - connected manifold . in the same context , in this work we have studied the role played by multiply - connectedness in the time evolution of the energy @xmath0 of a radiating system that lies in static flat space - time manifolds @xmath1 whose @xmath2 spacelike sections @xmath3 are compact in at least one spatial direction .
so , it may even have the lowest degree of multiply - connectedness .
this topological setting is general enough to encompass the entire set of possible classes of multiply - connected flat 3-manifolds discussed , for example , by wolf @xcite .
we have shown that the behavior of the radiating energy @xmath0 of the oscillator changes remarkably from exponential damping , when the underlying 3-manifold is @xmath5 , to a reverberation behavior for @xmath0 when the spacelike @xmath12 sections of @xmath1 are any possible multiply - connected flat 3-manifold .
thus , our study shows that the topological fragilities can arise not only in the usual cosmological modelling @xcite , but also in ordinary static flat space - time manifolds as long as they are multiply connected .
as we have emphasized in section [ radreaceq ] and [ sol ] this striking behavior of @xmath0 occurs with no violation of any physical law , and has a purely topological origin .
although the physical system we have investigated here is the same system discussed in @xcite , the present article generalizes the results of that paper in several respects .
firstly , the underlying topological setting is much more general since we did not assume that our spacelike 3-manifolds @xmath4 are compact and orientable .
actually we only require that they are multiply - connected , which means that @xmath4 can be endowed with any one of the 17 multiply - connected flat topologies @xcite , including , of course , the orientable compact 3-manifolds dealt with in @xcite . secondly , contrarily to the _ numerical _ integration performed in @xcite , here we have obtained a closed _ exact _ solution of the evolution equation for the harmonic oscillator in the above - mentioned general topological setting .
thirdly , in the case study we discussed in details in section [ case ] we have found the ( exact ) time evolution of our system in a specific _ non - compact _ ( although multiply - connected ) manifold , while in @xcite the time behavior of our system was ( numerically ) treated only for six _ compact orientable _ background manifolds .
finally in the present article we have also discussed the conservation and the balance of the total energy of our physical system , making apparent that in both papers the total energy is finite and conserved for any time @xmath10 . before closing it is worth mentioning that to the extent that we have studied the dynamical behavior of a test system in
flat frw space - time backgrounds our results have a moderate , though suggestive and well - founded , relation to cosmology .
besides , it should be noticed that the net role played by topology is better singled out in this static case , where the dynamical degrees of freedom have been frozen . on the other hand
, it appears that , apart from the inflationary expansion and perhaps a few other cases , the restriction we have made will not be decisive for the pattern of the behavior radiating energy @xmath0 in a number of flat expanding locally homogeneous and isotropic frw cosmological models .
finally , it is worth noting that since geometry constrains , but does not dictate the topology of space - time manifolds , in cosmological modelling one often is confronted with the basic question of what topologies are physically acceptable for a given space - time manifold .
an approach to this problem is to study the possible physical ( observational ) consequences of adopting particular topologies for the space - time ( for a fair list of references on this approach see the review article by lachize - rey and luminet @xcite ) .
this paper constitutes an example of this relevant approach .
we would like to express our thanks to professor john a. wheeler for stimulating correspondence on this subject matter .
we are also grateful to the brazilian scientific agencies capes , faperj and cnpq for the grants under which this work was carried out .
our aim in this appendix is to show how one can derive from equation ( [ rreq1 ] ) , which holds for an arbitrary density function @xmath248 of a @xmath85-sequence , the radiation reaction equation ( [ rreq2 ] ) for the limiting case of pointlike coupling between the scalar field and the oscillator [ @xmath249 .
as we mentioned earlier the pointlike limit case can be achieved by taking the limit @xmath256 of the above @xmath79-system equation . to this end
we firstly remind that as we are dealing with density functions @xmath248 whose supports are connected sets , the following reworded version of the mean value theorem can be stated : let @xmath257 be an arbitrary continuous function defined in @xmath258 , then there exists a point @xmath259 such that @xmath260 holds .
let now @xmath261 be the radius of @xmath262 and consider a point @xmath263 , then by the mean value theorem there exists a sequence of pair of points @xmath264 such that for each @xmath265 we have the sequence of numbers @xmath266 where @xmath267 . since @xmath268 when
@xmath269 , the mid subinterval shrinks to the point @xmath270 , then it is irrelevant for our purpose the knowledge of @xmath271 in that region .
furthermore , since @xmath272 we have @xmath273 the limit of @xmath274 can be obtained as follows . for a given fixed value @xmath265 , there exists a natural number @xmath275 such that for all @xmath276 we have @xmath277 .
thus by the mean value theorem there also exists a pair of points @xmath278 such that @xmath279 with @xmath280 when @xmath269 .
since @xmath281 has continuous first derivative we can expand it about the fixed value @xmath40 up to first order to obtain @xmath282 therefore the following limit : @xmath283 holds for all @xmath10 . | we study the role played by multiply - connectedness in the time evolution of the energy @xmath0 of a radiating system that lies in static flat space - time manifolds @xmath1 whose @xmath2 spacelike sections @xmath3 are compact in at least one spatial direction .
the radiation reaction equation of the radiating source is derived for the case where @xmath4 has any non - trivial flat topology , and an exact solution is obtained .
we show that the behavior of the radiating energy @xmath0 changes remarkably from exponential damping , when the system lies in @xmath5 , to a reverberation pattern ( with discontinuities in the derivative @xmath6 and a set of relative minima and maxima ) followed by a growth of @xmath0 , when @xmath3 is endowed with any one of the 17 multiply - connected flat topologies .
it emerges from this result that the compactness in at least one spatial direction of minkowski space - time is sufficient to induce this type of topological reverberation , making clear that topological fragilities can arise not only in the usual cosmological modelling , but also in ordinary flat space - time manifolds .
an explicit solution of the radiation reaction equation for the case where @xmath7 is discussed in details , and graphs which reveal how the energy varies with the time are presented and analyzed . |
a ( marked ) _ circular polygon _ is a closed disk @xmath2 with marked boundary points @xmath3 , which are called _ corners _ , enumerated cyclically according to the positive orientation of @xmath4 , equipped with a riemannian metric of constant curvature @xmath5 with conic singularities at the corners , and such that each _ side _ @xmath6 has constant geodesic curvature .
two such polygons @xmath2 and @xmath7 are _ congruent _ if there is an orientation - preserving isometry between them which sends each corner @xmath8 of @xmath2 to the corner @xmath9 of @xmath7 .
polygons with @xmath10 and @xmath11 are called _ digons , triangles _ and _ quadrilaterals _ , respectively .
if each side has zero geodesic curvature then @xmath2 is called a _
spherical polygon_. at every corner @xmath8 , an interior angle @xmath12 is defined , and in what follows we measure all angles in half - turns .
so angle @xmath13 means an angle of @xmath14 radians , in particular , `` integer angle '' is an integer multiple of @xmath15 radians .
a circular quadrilateral @xmath2 whose angles are odd multiples of @xmath1 is called a _
circular rectangle_. if such a quadrilateral is spherical , it is called a _ spherical rectangle_. in this paper
we describe the set of spherical rectangles with prescribed angles . as every surface of positive curvature @xmath5
is locally isometric to a piece of the unit sphere , every circular polygon can be described in terms of the _ developing map _ @xmath16 which is an analytic function on @xmath17 mapping every side into a circle on the riemann sphere . for spherical polygons these circles
are geodesic ( great circles ) .
this function @xmath18 is a local homeomorphism at each point except the corners , and at a corner @xmath19 satisfies @xmath20 where @xmath21 is the angle at this corner .
( if @xmath22 , the right - hand side has to be replaced by @xmath23 . )
each such function defines a circular polygon by the pull - back of the spherical metric from @xmath24 to @xmath2 .
if none of the @xmath25 equals @xmath5 , then the pair @xmath26 defines the polygon uniquely .
two pairs @xmath27 and @xmath28 define congruent polygons if @xmath29 , where @xmath30 is a rotation of the riemann sphere , and @xmath31 a conformal automorphism of the disk with the property @xmath32 for all @xmath33 .
this paper is a part of the project whose goal is to understand metrics of constant positive curvature with conic singularities on compact surfaces .
@xcite . an important class of such metrics can be obtained by gluing a spherical polygon to its mirror image isometrically along the boundary .
metrics of positive curvature on the sphere obtained in this way are characterized by the symmetry property : all conic singularities belong to a circle on the sphere , and the metric is symmetric with respect to this circle . in this paper , we classify spherical rectangles .
the cases when at least one of the angles of a spherical quadrilateral is integer were considered in @xcite and @xcite .
if we use the upper half - plane as @xmath2 , then the developing map of every circular quadrilateral is a ratio of two linearly independent solutions of the heun equation @xmath34 where @xmath35 with the standard normalization @xmath36 , @xmath37 . here
@xmath25 are the angles at the corners , and @xmath38 is a real _ accessory parameter_. each pair of linearly independent solutions of such an equation defines a circular quadrilateral .
different pairs of linearly independent solutions of the same equation define _ equivalent _ quadrilaterals : their developing maps are related by post - composition with linear - fractional transformations .
we will later see that an equivalence class may contain at most one spherical quadrilateral , up to congruence .
the condition that an equivalence class contains a spherical quadrilateral translates to the following condition on the heun equation : the projective monodromy group must be conjugate to a subgroup of @xmath39 .
so the problem of classification of spherical quadrilaterals with prescribed corners and angles is equivalent to the problem of classification of heun s equation with prescribed @xmath8 and @xmath25 whose monodromy is _ unitarizable _ , that is conjugate to a subgroup of @xmath39 .
the correspondence between the metrics on the sphere and heun s equations with unitarizable monodromy is bijective .
each symmetric metric on the sphere corresponds to two spherical quadrilaterals which are related by an anti - conformal involution , and to a unique normalized heun equation with real @xmath40 and unitarizable monodromy . in the case that all angles are odd multiples of @xmath1 , heun s equation can be explicitly solved in terms of elliptic integrals .
this fact was discovered by darboux @xcite who generalized hermite s work @xcite for the lam equation .
for the study of general circular rectangles in connection with heun s equation with @xmath25 multiples of @xmath1 we refer to the paper of van vleck @xcite .
we recall how this explicit solution is obtained .
* theorem a. * _ suppose that all @xmath25 in ( [ heun])([heun3 ] ) are odd multiples of @xmath1 .
then there are two linearly independent solutions of ( [ heun ] ) whose ratio is of the form @xmath41 where @xmath42 and @xmath43 is a real polynomial in all three variables .
this polynomial satisfies the third order linear differential equation @xmath44 where @xmath45 and @xmath46 are the coefficients in front of @xmath47 and @xmath48 in ( [ heun ] ) .
_ equation ( [ hermite ] ) has one - dimensional space of polynomial solutions of degree @xmath49 this permits to find @xmath50 by rational operations . that @xmath50 satisfies ( [ hermite ] ) guarantees that all residues of the integrand in ( [ integral ] ) are of the form @xmath51 with some real @xmath52 .
the condition @xmath53 defines @xmath50 up to a sign .
periods of the integral ( [ integral ] ) , other than those coming from the residues , form a lattice generated by two canonical periods ( integrals over adjacent real segments ) .
of these two canonical periods one is real and another is pure imaginary .
the condition that the monodromy of ( [ heun ] ) is unitarizable means that both periods must be imaginary , therefore the real period must vanish . for a fixed real @xmath19 , and
given angles , this imposes a transcendental equation on @xmath38 .
it is not clear how to determine or estimate the number of real solutions of this equation , but for small angles @xmath54 it can be solved numerically .
the results of computation are described in section [ limits ] .
instead we use a geometric method which allows us to classify spherical rectangles , and describe their geometry .
the following elementary statement was proved in @xcite .
[ onecircle ] let @xmath18 be the developing map of a spherical rectangle @xmath2 .
then there are two opposite sides of @xmath2 whose @xmath18-images are contained in the same circle , and the other pair of opposite sides is mapped to distinct circles .
thus the boundary of a spherical rectangle @xmath2 is mapped by @xmath18 to the union of three great circles , one of them , say @xmath55 , being orthogonal to the other two , @xmath56 and @xmath57 .
let @xmath58 be the angle between the circles @xmath56 and @xmath57 .
there are two choices for this angle ( the other one being @xmath59 ) .
see definition [ df : theta ] below for the unique choice of the angle @xmath60 associated with a spherical rectangle @xmath2 .
the @xmath18-preimage in @xmath2 of the three circles is called the _ net _ of @xmath2 .
the net is a combinatorial invariant of @xmath2 , defined up to an orientation - preserving homeomorphism of @xmath2 respecting its initial corner @xmath61 .
[ isometry ] a marked spherical rectangle @xmath2 is defined uniquely up to isometry by its net and the angle @xmath58 .
this will be proved in section [ nets ] . in section [ nets ] , we will explicitly describe all possible nets of spherical rectangles ( theorems [ classification ] and [ existence ] ) . as a consequence
we will obtain the following necessary and sufficient conditions on the angles of a spherical rectangle : [ theorem1 ] let @xmath62 be non - negative integers , and @xmath63 then for the existence of a spherical rectangle with angles @xmath64 it is necessary and sufficient that one of the following conditions be satisfied : either @xmath65 or @xmath66 to state our next result we need some definitions .
we recall that we consider _ marked _ spherical rectangles .
two of them are congruent if there is an orientation - preserving isometry of @xmath2 respecting the initial corner @xmath61 .
every quadrilateral can be mapped conformally onto a flat rectangle with vertices @xmath67 and all angles @xmath1 , such that @xmath61 maps to @xmath68 .
the number @xmath69 is called the _ modulus _ of the quadrilateral .
each pair @xmath70 , where @xmath71 is a net and @xmath58 , defines a marked spherical rectangle @xmath72 ( see theorem [ existence ] ) .
thus the set of all spherical rectangles with given angles consists of curves @xmath73 parameterized by @xmath60 and labeled by the nets .
the modulus @xmath69 of @xmath74 is a continuous function of @xmath60 .
there are two kinds of these curves : on the curves of the first kind , @xmath75 as @xmath76 , while @xmath69 tends to a non - zero limit @xmath77 as @xmath78 . on the curves of the second kind ,
@xmath79 as @xmath76 , while @xmath69 tends to a non - zero limit @xmath77 as @xmath78 . this is proved in theorem [ limit ] , and we give few examples of computation of the limits @xmath80 in section [ limits ] . in all our examples @xmath69 is a monotone function of @xmath60 .
this is proved in @xcite for the simplest family of spherical quadrilaterals with angles @xmath81 but it is unlikely that this property holds in general .
however , it is true for sufficiently small and large values of @xmath69 .
[ small - large ] each curve @xmath74 has finitely many intervals on which @xmath69 is monotone .
in particular , for sufficiently small ( resp . , large ) @xmath82 , there is a unique spherical rectangle with the modulus @xmath69 in a curve @xmath74 of the first ( resp . , second ) kind .
this follows from the general theory of o - minimal structures ( see , e.g. , @xcite ) , since the integral in ( [ integral ] ) is a pfaffian function in the sense of khovanskii @xcite .
it was shown in @xcite that the structure generated by pfaffian functions is o - minimal .
our final results count the nets for spherical rectangles with given angles .
the quadruple @xmath83 is _ special _ if @xmath84 in ( [ delta ] ) is an odd integer and one of the following holds : either @xmath85 and @xmath86 or @xmath87 and @xmath88 .
define @xmath89,\ ] ] @xmath90,\ ] ] @xmath91 note that conditions ( [ 1 ] ) and ( [ 2 ] ) are satisfied when @xmath92 and @xmath93 , respectively .
[ theorem2 ] for a special quadruple @xmath94 satisfying either ( [ 1 ] ) or ( [ 2 ] ) there exist @xmath95 one - parametric families of congruence classes of marked spherical rectangles with angles @xmath96 . if @xmath94 is not special but satisfies ( [ 1 ] ) ( resp . , ( [ 2 ] ) ) then there exist @xmath97 ( resp . , @xmath98 ) one - parametric families of congruence classes of marked spherical rectangles with angles @xmath96 .
each family is parameterized by @xmath58 ( see definition [ df : theta ] ) .
each family contains either rectangles of arbitrarily small moduli or arbitrarily large moduli but not both .
the numbers of families of both types are equal , so for each type this number is either @xmath99 or @xmath100 or @xmath101 , depending on @xmath83 .
[ exact ] theorem [ theorem2 ] and proposition [ small - large ] imply that the number of spherical rectangles with given angles @xmath96 is exactly @xmath99 or @xmath100 or @xmath101 , depending on @xmath83 , for sufficiently small and large values of @xmath69 . for fixed angles
@xmath102 , consider the two - parametric family of heun s equations ( [ heun ] ) with parameters @xmath103 equivalence classes of circular rectangles are in correspondence to such heun s equations .
one - parametric families of spherical rectangles of theorem [ theorem2 ] correspond to smooth disjoint curves in the half - plane @xmath104 , each having one end in this half - plane . on the other end , either @xmath105 or @xmath106 .
when @xmath107 is rational , the monodromy group of the developing map is finite , so @xmath18 is an algebraic function . in this case ,
@xmath108 , where @xmath109 and @xmath110 is a rational belyi function , which means that the only critical values of @xmath110 are @xmath111 , @xcite .
function @xmath112 is also a belyi function , it is called the fundamental rational function of the dihedral group @xcite .
the set @xmath113 consists of the unit circle and @xmath46 lines @xmath114 in the simplest case @xmath115 , the image @xmath116 is contained in the union of the unit circle @xmath55 , real line @xmath56 , and imaginary line @xmath57 .
the monodromy is the klein viergroup @xmath117 , represented as @xmath118 and @xmath112 has the property that @xmath119 .
then it is easy to see that our net , together with its reflection in the real line , coincides with @xmath120 .
the set @xmath120 for a belyi function @xmath110 is a triangulation of the sphere with all vertices of even degree .
therefore our classification of the nets can be restated as classification of triangulations @xmath121 of the sphere with the following properties : \a ) @xmath121 is symmetric with respect to @xmath122 , and @xmath122 is contained in the @xmath5-skeleton of @xmath121 , \b ) there are four vertices @xmath8 of @xmath121 on the real line of prescribed orders @xmath123 .
\c ) all other vertices of @xmath121 have order @xmath11 . for @xmath124 and @xmath125 ,
algebraic developing maps are explicitly written in @xcite .
in this section we prove a preliminary result for classification of nets . roughly speaking
it says that every spherical rectangle is a union of two spherical triangles . as we prove this result by induction , it is convenient to consider a more general class of spherical polygons , characterized by the property that the developing map sends their sides to three transversally intersecting great circles and corners to the intersection points of these circles .
the net @xmath71 defines a triangulation of such a polygon @xmath2 , each face of it being mapped by @xmath18 one - to - one onto one of the triangles into which the three circles partition the sphere .
this triangulation satisfies the following properties:(p1 ) each vertex inside @xmath2 has degree @xmath11;(p2 ) all boundary vertices , other than corners of @xmath2 , have degree @xmath126.combining this triangulation with its mirror copy , we obtain a triangulation @xmath121 of the sphere satisfying the following properties:(s1 ) @xmath121 is symmetric with respect to a circle @xmath127 contained in the 1-skeleton of @xmath121;(s2 ) each vertex of @xmath121 has even degree , and all its vertices not contained in @xmath127 have degree @xmath11.it is easy to show that the nets of spherical polygons with all sides mapped to three transversal circles and all corners to intersection points of those circles are in one - to - one correspondence with triangulations of the sphere satisfying ( s1 ) and ( s2 ) .
two nets are combinatorially equivalent if they can be obtained from each other by an orientation - preserving homeomorphism ( mapping corners to corners and sides to sides ) preserving the initial corner .
if @xmath55 is any of the three circles , its preimage in @xmath2 is called @xmath55-_net _ , denoted @xmath128 .
an _ arc _ of the net @xmath128 ( or an arc of @xmath71 if @xmath55 is not specified ) is a connected component of @xmath129 .
since @xmath18 is a local homeomorphism on the interior of @xmath2 , an arc may be homeomorphic to either an open interval with both ends on the boundary of @xmath2 ( possibly , at the same corner of @xmath2 ) or a circle in the interior of @xmath2 .
we ll show below ( see corollary [ nocircle ] ) that an arc of a spherical rectangle @xmath2 must have at least one end at a corner of @xmath2 .
in particular , an arc of a spherical rectangle can not be a circle .
an arc is called _ short _ if it does not intersect other arcs of @xmath71 .
any two arcs of the same net @xmath128 are disjoint .
[ triangulation ] let @xmath2 be a spherical @xmath130-gon such that all its sides are mapped to three transversal great circles by the developing map , and all its corners are mapped to intersection points of those circles .
then either @xmath131 or there is a triangulation of @xmath2 by @xmath132 disjoint arcs of its net , each of them connecting two non - adjacent corners of @xmath2 . _ proof .
_ it is enough to show that , unless @xmath2 is a digon or a triangle , there exists an arc of its net @xmath71 connecting two of its non - adjacent corners .
we prove this by induction on the number @xmath99 of faces of @xmath71 .
if @xmath71 has one face then , since any face of @xmath71 is a triangle , @xmath2 is a triangle . if @xmath133 then there exists an arc @xmath134 of @xmath71 adjacent to a point @xmath45 on the boundary of @xmath2 .
otherwise the face of @xmath71 adjacent to its boundary would not be simply connected .
if @xmath134 connects two distinct corners @xmath45 and @xmath46 of @xmath2 then either @xmath45 and @xmath46 are non - adjacent and we are done , or @xmath45 and @xmath46 are adjacent corners of @xmath2 , and @xmath134 partitions @xmath2 into a digon and a polygon @xmath7 with the same number of corners as @xmath2 and a smaller than @xmath99 number of faces of its net . by inductive hypothesis , unless @xmath7 ( and thus @xmath2 ) is a digon or a triangle , there is an arc @xmath135 of @xmath7 connecting two of its non - adjacent corners . in the latter case
, @xmath135 is also an arc of @xmath71 connecting two non - adjacent corners of @xmath2 , and we are done .
suppose now that @xmath71 does not have any arcs connecting two corners of @xmath2 .
if @xmath134 has both ends at the same point @xmath45 then @xmath2 can be replaced by a @xmath136-gon @xmath7 having all sides of @xmath2 plus @xmath134 as its sides , with the number of faces of the net @xmath137 of @xmath7 smaller than @xmath99 .
there is a mapping @xmath138 such that any two distinct points of @xmath7 map to distinct points of @xmath2 , except the two ends of the side @xmath134 of @xmath7 that both map to @xmath45 . by the inductive hypothesis
, there is an arc @xmath135 of @xmath137 connecting two non - adjacent corners @xmath139 and @xmath140 of @xmath7 .
then @xmath141 is an arc of @xmath71 connecting two ( possibly , adjacent ) corners of @xmath2 , a contradiction .
thus we may suppose that @xmath134 has two distinct ends @xmath45 and @xmath46 on the boundary of @xmath2 , at least one of them not a corner of @xmath2
. then @xmath134 partitions @xmath2 into two polygons @xmath7 and @xmath142 , with the number of corners @xmath143 and @xmath144 respectively , where @xmath145 . if @xmath146 then at least one of @xmath143 and @xmath144 is greater than @xmath126 . since both @xmath7 and @xmath142 have the number of faces of their nets smaller than @xmath99 , by the inductive hypothesis at least one of them has an arc @xmath135 of its net connecting two non - adjacent corners .
then @xmath135 is also an arc of @xmath71 connecting two non - adjacent corners of @xmath2 .
this completes the proof . and
@xmath147.,width=384 ] with three digons @xmath148 attached to its sides .
( b ) nets of spherical rectangles.,width=480 ] [ nocircle ] if @xmath2 is a spherical polygon satisfying the condition of theorem [ triangulation ] then an arc @xmath134 of the net of @xmath2 is an open interval with at least one end at a corner of @xmath2 .
_ it follows from classification of spherical triangles ( see sections 10 and 12 of @xcite , section 6 of @xcite and figs .
[ rect - triang ] , [ rect - digon ] and [ triang - rectang]a ) that an arc of the net of a spherical triangle with all sides mapped to three circles @xmath149 and all corners to intersection points of those circles must have at least one end at its corner .
indeed , such a triangle @xmath150 contains a primitive triangle @xmath151 ( either one of triangles @xmath152 or one of triangles @xmath147 , see fig .
[ rect - triang ] ) with its apex at the intersection of two circles , say @xmath55 and @xmath56 , and its base on the third circle @xmath57 .
any arc of the net of @xmath151 connects its apex with its base .
the triangle @xmath150 is obtained by attaching digons to the sides of @xmath151 ( see fig .
[ triang - rectang]a ) .
an arc of the net of a digon @xmath153 either has an end at one of its corners or the ends on both its sides , but can not have both ends on one side of @xmath153 .
thus an arc of @xmath150 must have at least one end at its corner .
since the intersection of @xmath134 with any triangle @xmath150 of a triangulation of @xmath2 by disjoint arcs of its net connecting its non - adjacent corners is either a side of @xmath150 or an arc of the net of @xmath150 , it must have at least one end at a corner of @xmath150 .
but all corners of @xmath150 are also corners of @xmath2 .
as was shown in @xcite ( see proposition [ onecircle ] ) any spherical rectangle @xmath2 has two opposite sides mapped to the same circle by its developing map @xmath18 , and two other opposite sides mapped to distinct circles .
thus there are two types of marked spherical rectangles : in the first type the images of @xmath154 and @xmath155 belong to the same circle , and in the second type the images of @xmath156 and @xmath157 belong to the same circle .
it is enough to classify spherical rectangles of the first type , as all rectangles of the second type can be obtained from those of the first type by orientation - reversing isometry preserving the marked corner .
* assumption 1 . *
_ unless stated otherwise , all spherical rectangles below are assumed to be of the first type .
_ let @xmath55 be the circle to which two sides @xmath154 and @xmath158 of a spherical rectangle @xmath2 are mapped , and let @xmath56 and @xmath57 be the circles to which the sides @xmath156 and @xmath157 of @xmath2 are mapped ( see fig . [ tt ] ) .
theorem [ triangulation ] implies that there is an arc of the net @xmath71 of @xmath2 connecting two opposite corners of @xmath2 .
such an arc must be mapped to the circle @xmath55 , since two opposite corners of @xmath2 are mapped to intersection points of @xmath55 with two distinct circles other than @xmath55 .
this implies that @xmath2 can not have two arcs of @xmath71 connecting two pairs of its opposite corners : such arcs would have an intersection point inside @xmath2 , while any two arcs mapped to the same circle @xmath55 are disjoint .
* assumption 2 . *
_ unless stated otherwise , we choose the initial corner @xmath61 of a marked spherical rectangle @xmath2 so that there is an arc @xmath134 of @xmath71 connecting the corners @xmath159 and @xmath160 of @xmath2 .
_ such an arc @xmath134 partitions @xmath2 into two spherical triangles @xmath151 and @xmath161 , where @xmath151 has an integer corner at @xmath160 and the base @xmath156 mapped to @xmath56 , while @xmath161 has an integer corner at @xmath159 and the base @xmath157 mapped to @xmath57 ( see fig .
we ll show below ( see remark [ angles ] ) that the angles of such rectangle @xmath2 satisfy the inequality @xmath162 .
any rectangle of the first ( resp .
, second ) type with an arc of its net connecting its corners @xmath61 and @xmath163 can be obtained from a rectangle of the second ( resp , first ) type with an arc of its net connecting its corners @xmath159 and @xmath160 by choosing @xmath159 instead of @xmath61 as an initial corner , and relabeling the corners accordingly .
the angles of such rectangle satisfy the inequality @xmath164 .
thus it is enough to classify spherical rectangles satisfying assumptions 1 and 2 .
[ classification ] let @xmath2 be a marked spherical rectangle satisfying assumptions 1 and 2 . then @xmath2 is a union of two primitive triangles @xmath152 and @xmath165 having integer angles @xmath166 and @xmath167 , respectively , digon @xmath168 with the sides mapped to @xmath55 having common sides with both @xmath152 and @xmath165 , digon @xmath169 with the sides mapped to @xmath56 attached to the base of @xmath152 , digon @xmath170 with the sides mapped to @xmath57 attached to the base of @xmath165 , digon @xmath171 attached to the remaining side of @xmath152 , and digon @xmath172 attached to the remaining side of @xmath165 .
the sides of @xmath172 and @xmath171 are mapped to @xmath55 .
here @xmath173 are non - negative integers satisfying @xmath174 , that is , @xmath175 only if @xmath176 , @xmath177 only if @xmath178 . the value @xmath68 for @xmath179 means there is no digon attached
. _ proof . _ assumptions 1 and 2 imply that an arc @xmath134 of the net @xmath71 of @xmath2 connecting its corners @xmath159 and @xmath160 partitions @xmath2 into two spherical triangles @xmath151 and @xmath161 , where @xmath151 has an integer corner with an angle @xmath166 at @xmath160 , and @xmath161 has an integer corner with an angle @xmath167 at @xmath159 , for some non - negative integers @xmath180 and @xmath181 .
classification of spherical triangles with one integer corner ( see sections 10 and 12 of @xcite ) implies that the triangle @xmath151 ( resp . , @xmath161 )
is combinatorially equivalent to a primitive triangle @xmath152 ( resp . ,
@xmath165 ) having an angle @xmath166 ( resp . , @xmath167 ) at its integer corner , with digons attached to its sides ( see figs .
[ rect - triang ] , [ rect - digon ] and [ triang - rectang]a ) .
no digons may be attached to the base @xmath156 of @xmath151 ( resp .
, the base @xmath157 of @xmath161 ) if @xmath182 ( resp . , @xmath183 ) .
each digon @xmath184 has equal integer angles @xmath130 at its two corners .
the sides of @xmath152 and @xmath165 are mapped to @xmath55 and can not contain preimages of intersection points of @xmath55 with either @xmath56 or @xmath57 , other than the corners of @xmath2 .
this implies that the union of digons attached to the sides of @xmath152 and @xmath165 and having @xmath134 as their common side is a digon @xmath168 with even integer angles @xmath185 at its two corners .
thus a net of a spherical rectangle satisfying assumptions 1 and 2 must have the structure shown schematically in fig .
[ triang - rectang]b .
this proves theorem [ classification ] .
[ angles ] the angles at the corners @xmath186 of a marked spherical rectangle @xmath2 in theorem [ classification ] have the integer parts @xmath187 respectively . in particular , @xmath162 . for a marked spherical rectangle ( of either first or second type ) with an arc of its net connecting its corners @xmath61 and @xmath163 ,
the integer parts of its angles satisfy @xmath164 .
[ shortarc ] theorem [ classification ] implies that a spherical rectangle @xmath2 satisfying assumptions 1 and 2 has at least one short arc @xmath134 connecting its corners @xmath159 and @xmath160 , and that all such short arcs are mapped to the same arc @xmath188 of the circle @xmath55 with the ends at the intersection points of @xmath55 with @xmath56 and @xmath57 .
[ df : theta ] the angle @xmath58 between the circles @xmath56 and @xmath57 is defined as @xmath189 where @xmath13 is the length of any short arc @xmath134 of the net of a spherical rectangle @xmath2 connecting its opposite corners , divided by @xmath15 . alternatively , @xmath13 is the length of the arc @xmath188 of @xmath55 to which @xmath134 is mapped , divided by @xmath15 .
_ proof of proposition [ isometry ] .
_ we want to show that two marked spherical rectangles @xmath2 and @xmath7 with equivalent nets @xmath71 and @xmath137 , and the same angle @xmath60 , are congruent , i.e. , there is an orientation - preserving isometry @xmath190 mapping the initial corner @xmath61 of @xmath2 to the initial corner @xmath191 of @xmath7 . according to definition [ df : theta ] , the developing map @xmath16 maps the sides of @xmath2 to three great circles @xmath55 , @xmath56 , @xmath57 , such that both @xmath56 and @xmath57 are orthogonal to @xmath55 , so that the sides @xmath154 and @xmath155 of @xmath2 are mapped to @xmath55 , the sides @xmath156 and @xmath157 of @xmath2 are mapped to @xmath56 and @xmath57 , respectively , and any short arc of @xmath71 connecting the corners @xmath159 and @xmath160 of @xmath2 is mapped to an arc @xmath188 of @xmath55 of length @xmath14 where @xmath192 , the ends @xmath50 and @xmath193 of @xmath188 being the images of @xmath159 and @xmath160 , respectively . in particular , @xmath50 and @xmath193 are intersection points of @xmath55 with @xmath56 and @xmath57 , respectively .
similarly , the developing map @xmath194 maps the sides of @xmath7 to three great circles @xmath127 , @xmath195 , @xmath196 , such that both @xmath195 and @xmath196 are orthogonal to @xmath127 , and any short arc @xmath135 of @xmath137 connecting the corners @xmath197 and @xmath198 of @xmath7 is mapped to an arc @xmath199 of @xmath127 of length @xmath14 , the ends @xmath200 and @xmath201 of @xmath199 being the images of @xmath197 and @xmath198 , respectively . applying if necessary rotation of @xmath24 , we may assume that @xmath202 , @xmath203 , @xmath204 , @xmath205 , @xmath206 , and @xmath207
. then the equivalence of the nets @xmath71 and @xmath137 implies that the developing maps @xmath18 and @xmath112 send the corresponding faces , edges and vertices of partitions of @xmath2 and @xmath7 to the same triangles , segments and intersection points of the partition of @xmath24 by the three circles @xmath55 , @xmath56 and @xmath57 .
in particular , @xmath61 and @xmath191 are mapped by @xmath18 and @xmath112 to the same point of @xmath24 .
[ existence ] for any non - negative integers @xmath173 satisfying @xmath174 there is a unique , up to combinatorial equivalence , net @xmath71 of the type described in theorem [ classification ] . for any such net @xmath71 and any @xmath58 there
exists a unique spherical rectangle @xmath208 having @xmath71 as its net , sides mapped to three circles @xmath55 , @xmath56 , @xmath57 , and a short arc of length @xmath209 connecting its corners @xmath159 and @xmath160 .
_ to define the net @xmath71 , we start with a digon @xmath168 obtained by combining @xmath210 copies of digon @xmath211 shown in the middle of the first row of fig .
[ rect - digon ] . if @xmath212 then there is no such digon , and we proceed with gluing together triangles @xmath152 and @xmath165 ( see fig .
[ rect - triang ] ) so that the side of each of these triangles that follows its base in the counterclockwise cyclic order becomes their common side . if @xmath213 then triangles @xmath152 and @xmath165 are attached to opposite sides of @xmath168 so that the side of each of these triangles that follows its base in the counterclockwise cyclic order becomes its common side with @xmath168 .
the integer corner of @xmath152 ( resp . , @xmath165 ) coincides with a non - integer corner of @xmath165 ( resp . , @xmath152 ) .
next , we attach digons @xmath172 and @xmath171 , obtained by combining @xmath214 and @xmath215 copies , respectively , of the digon @xmath148 shown in the left side of the first row of fig .
[ rect - digon ] , so that any two adjacent digons have either a common short side or a common long side , and so that each of the two resulting digons has at least one short side , as is shown in examples of @xmath211 and @xmath216 in the first row of fig .
[ rect - digon ] .
then digons @xmath172 and @xmath171 are attached to the free sides of @xmath165 and @xmath152 , respectively .
the free sides of these triangles are preceding their respective bases in the counterclockwise cyclic order .
if @xmath217 ( resp . , @xmath218 ) then no digon @xmath172 ( resp . , @xmath171 ) is attached . finally , if @xmath176 and @xmath175 ( resp . ,
@xmath178 and @xmath177 ) then a digon @xmath169 ( resp . , @xmath170 ) , obtained by combining @xmath219 and @xmath220 copies , respectively , of the digon @xmath148 shown in the second row of fig .
[ rect - digon ] , is attached to the base of @xmath152 ( resp . , @xmath165 ) .
examples of such digons are shown in the second row of fig .
[ rect - digon ] . if we label the sides of @xmath152 and @xmath165 by @xmath55 , the base of @xmath152 by @xmath56 and the base of @xmath165 by @xmath57 then all edges of @xmath71 can be uniquely labeled so that the sides of each of its triangles are labeled by three distinct labels .
consider the standard sphere @xmath24 with three great circles @xmath55 , @xmath56 and @xmath57 such that @xmath55 is orthogonal to both @xmath56 and @xmath57 , and one of the two complementary angles between @xmath56 and @xmath57 is @xmath60 , the other one being @xmath192 .
then there is a mapping of @xmath2 to @xmath24 , locally one - to - one everywhere except at the corners , which is unique up to a homeomorphism of @xmath2 preserving all vertices and edges of @xmath71 and a rotation of the sphere preserving the three circles , such that any arc of the boundary of @xmath168 ( or a common side of @xmath152 and @xmath165 if @xmath212 ) maps to an arc of @xmath55 of length @xmath14 , and each edge of @xmath71 maps to an arc of the circle corresponding to its label .
this defines on @xmath2 a metric of the spherical rectangle @xmath72 . ,
width=480 ] , width=384 ] , width=384 ] , and their deformations , width=480 ]
each admissible set of integers @xmath173 in theorem [ classification ] defines a net @xmath71 and the corresponding one - parametric family of marked spherical rectangles @xmath72 satisfying assumptions 1 and 2 , parameterized by the angle @xmath60 between the circles @xmath56 and @xmath57 ( see definition [ df : theta ] ) .
these two circles intersect the circle @xmath55 at the right angle . in figs .
[ rect-3circles]a and [ rect-3circles]b , two projections of the three circles are shown , and two of the triangles of the partition of the sphere defined by these circles are shaded .
let @xmath50 and @xmath193 be the images of the corners @xmath159 and @xmath160 , respectively , of a marked spherical rectangle @xmath2 , so that the arc @xmath223 of @xmath55 is the image of a short arc @xmath134 of the net @xmath71 of @xmath2 connecting @xmath159 and @xmath160
. then the shaded areas in figs .
[ rect-3circles]a and [ rect-3circles]b contract to arcs when @xmath76 and expand to half - disks when @xmath78 .
if all circles remain geodesic , then @xmath56 and @xmath57 converge to the same circle when @xmath76 and when @xmath78 .
however , applying a linear - fractional transformation depending on @xmath60 , so that the arc @xmath224 of @xmath55 in figs .
[ rect-3circles]a and [ rect-3circles]b contracts to a point while the arc @xmath225 does not , we can obtain in the limit @xmath78 a non - geodesic configuration shown in two different projections in figs .
[ rect-3circles]c and [ rect-3circles]d , where the shaded areas are the limits of the shaded areas in figs .
[ rect-3circles]a and [ rect-3circles]b .
[ 1313 ] a net of a spherical rectangle @xmath2 with the angles @xmath226 considered in @xcite is shown in fig .
[ rectangle1313 ] ( left ) .
the shaded area corresponds to preimage of the shaded areas in figs .
[ rect-3circles]a and [ rect-3circles]b . the arc connecting @xmath159 and @xmath160
is mapped to the arc @xmath225 of @xmath55 , the corners @xmath61 and @xmath163 are mapped to @xmath227 and @xmath99 , respectively . when @xmath76 , the sides @xmath154 and @xmath155 of @xmath2 are contracted to points , while the distance between them has a positive limit .
thus the modulus of @xmath2 has limit @xmath68 as @xmath76 ( see @xcite , section 15 , and @xcite ) . when @xmath78 , applying a linear - fractional deformation depending on @xmath60 to the three - circle configuration , and the corresponding transformation to @xmath2 ( this does not change the modulus of @xmath2 which is a conformal invariant ) we can get in the limit a non - geodesic circular rectangle shown in fig . [ rectangle1313 ] ( right ) .
thus the modulus of @xmath2 tends to a finite positive value when @xmath78 .
computation in @xcite shows that this value is @xmath228 .
[ limit ] let @xmath71 be one of the nets described in theorem [ classification ] , and @xmath58 .
then the modulus of @xmath72 tends to @xmath68 when @xmath76 , and to a finite positive value when @xmath78 .
_ when @xmath76 , none of the arcs connecting @xmath159 and @xmath160 contract , while the triangle @xmath152 ( resp . , @xmath165 ) contains either a short arc of @xmath71 or a side @xmath155 ( resp . , @xmath154 ) of @xmath2 that maps to either the arc @xmath229 or the arc @xmath230 of @xmath55 , connecting its apex @xmath160 ( resp . ,
@xmath159 ) with a point @xmath231 ( resp . , @xmath232 ) on its base .
these two arcs contract to points when @xmath76 .
hence the distance between the sides @xmath156 and @xmath157 of @xmath2 tends to @xmath68 when @xmath76 . at the same time
, there are no short arcs of @xmath71 having one end on @xmath154 and another end on @xmath155 , other than those connecting @xmath159 and @xmath160 which do not contract as @xmath76 .
thus the distance between the sides @xmath154 and @xmath155 does not tend to @xmath68 as @xmath76 .
this implies that the modulus of @xmath2 tends to @xmath68 as @xmath76 ( see @xcite , section 15 ) . when @xmath78 , the short arcs connecting @xmath159 and @xmath160 contract , thus both the distance between @xmath156 and @xmath157 and the distance between @xmath154 and @xmath155 tend to @xmath68 .
to understand the limit of the modulus of @xmath2 , we apply a linear - fractional transformation depending on @xmath60 as in example [ 1313 ] to the three - circle configuration , so that the short arc @xmath224 of @xmath55 contracts , while the short arc @xmath225 does not . in the non - geodesic limit ( see fig .
[ rect-3circles]c ) the circles @xmath56 and @xmath57 become tangent ( when their tangency point is mapped to @xmath233 as in fig .
[ rect-3circles]d , they become parallel lines ) .
all short arcs of @xmath71 connecting @xmath159 and @xmath160 map to @xmath225 , and all arcs of @xmath55 connecting the apex of a triangle @xmath152 ( resp . ,
@xmath165 ) with a point on its base , map to either @xmath229 or @xmath230 .
since neither @xmath229 nor @xmath230 contracts when @xmath78 , and @xmath225 does not contract after the linear - fractional transformation , the distances between opposite sides of @xmath2 do not tend to @xmath68 in the limit .
thus @xmath2 converges to a non - geodesic circular rectangle ( see figs .
[ rectangle1313 ] , [ rectangle3535 ] , [ rectangle3737 ] , and [ rectangle5757 ] ) and the modulus of @xmath2 tends to a finite positive value . * remarks on computation of limit moduli @xmath69 .
* the boundary of the limit rectangle described in the proof of theorem [ limit ] is mapped by developing map into three straight lines ( see fig . [
rect-3circles]d ) . this allows to represent the developing map by the schwarz
christoffel formula .
condition that the points @xmath50 and @xmath2 are on the same vertical line imposes one real equation which permits to determine the modulus of the rectangle @xmath69 .
see @xcite where the simplest example is described in all detail .
the number of solutions to this equation is the number of nets with given angles . to determine which solution corresponds to which net , we use the evident inequalities between the moduli of degenerate rectangles ( shown in the right of figs .
[ rectangle3535 ] , [ rectangle3737 ] , and in the bottom of fig .
[ rectangle5757 ] ) . [ 3535 ] two combinatorially distinct nets of spherical rectangles with the angles @xmath234 , and the nets of their non - geodesic limits when @xmath78 , are shown in figs .
[ rectangle3535]a and [ rectangle3535]b . the moduli @xmath235 and @xmath236 of the limiting rectangles
are @xmath237 and @xmath238 , respectively .
[ kk ] shows schematically the areas of existence of these spherical rectangles ( nets a and b ) and their involution - symmetric rectangles ( nets @xmath239 and @xmath240 ) for different values of the modulus @xmath69 .
[ 5757 ] three
combinatorially distinct nets of spherical rectangles with the angles @xmath241 , and the nets of their non - geodesic limits when @xmath78 , are shown in figs .
[ rectangle5757]p , [ rectangle5757]q , [ rectangle5757]r .
the moduli @xmath242 , @xmath243 , @xmath244 of the limiting rectangles are @xmath245 , @xmath246 , @xmath247 , respectively .
[ kkk ] shows schematically the areas of existence of these spherical rectangles ( nets p , q , r ) and their involution - symmetric rectangles ( nets @xmath248 , @xmath249 , @xmath250 ) for different values of the modulus @xmath69 . [ 3737 ] a net of a spherical rectangle with the angles@xmath251 , and the net of its non - geodesic limit when @xmath78 , is shown in fig .
[ rectangle3737 ] .
the modulus @xmath69 of the limiting rectangle is @xmath252 . , width=384 ] , width=480 ] * remarks and conjectures . *
each of the three nets p , q , r in fig .
[ kkk ] produces a continuous family where the modulus @xmath69 can be arbitrarily small .
so for sufficiently small @xmath69 there are at least three different marked spherical quadrilaterals with the angles @xmath241 .
similarly , one can conclude from fig .
[ kkk ] that there are at least two quadrilaterals with modulus @xmath253 at least three for @xmath254 , at least two for @xmath255 , at least three for @xmath256 , at least two for @xmath257 , and at least three for @xmath258 .
similar conclusions apply to fig .
our computations show that in fact these lower estimates are equalities , in all cases which we computed .
actually @xmath69 is a monotone function of @xmath60 in all these cases , but we do not expect this monotonicity to hold for all angles .
so theorem [ theorem2 ] gives only lower estimates for the number of quadrilaterals with given angles and modulus , when this modulus is small or large .
this lower estimate is @xmath99 or @xmath259 ( see ( [ k1 ] ) , ( [ k2 ] ) , ( [ n ] ) .
we conjecture that these lower estimates are exact , and that we have equality for large and small moduli .
all computed examples confirm this .
we want to answer the following question : given four non - negative integers @xmath260 , how many nets of marked spherical rectangles with the integer parts @xmath260 of the angles at their corners @xmath261 do exist ? it is enough to answer this question for the spherical rectangles of the first type ( satisfying assumption 1 ) , since all rectangles of the second type can be obtained then by an involution preserving the marked corner .
also , we may assume that @xmath162 , which is true for the marked spherical rectangles satisfying assumption 2 ( see remark [ angles ] ) .
the answer for the rectangles with @xmath164 , having an arc connecting corners @xmath61 and @xmath163 , can be obtained then by a different choice of the initial corner .
we start with listing operations on the nets which do not change the angles of a spherical rectangle @xmath2 . assuming notation of theorem [ classification ] and remark [ angles ]
, we have expressions ( [ inequality ] ) for the angles of @xmath2 .
\a ) if @xmath213 and @xmath280 then there is a unique operation among operations i - v that is applicable to @xmath2 and results in a rectangle with the same angles as @xmath2 , with @xmath210 reduced by @xmath5 .
_ we start with a proof of ( a ) , assuming @xmath213 . if @xmath283 then operation v is applicable . if @xmath284 then @xmath176 and operation iii is applicable . if @xmath285 then @xmath176 and operation i is applicable . if @xmath286 then @xmath178 and operation iv is applicable . if @xmath287 then @xmath178 and operation ii is applicable
it is easy to check that only one of the operations i - v is applicable in each of these cases . to prove ( b ) note first that inverses to operations i - v are only possible when @xmath280 .
next , for given @xmath180 and @xmath181 , conditions on @xmath180 and @xmath181 for applicability of the operations inverse to operations i - v may hold for at most one of these operations .
[ exist ] for given @xmath260 , the set of values of @xmath210 that may appear in the nets of marked spherical rectangles of the first type with @xmath290 is either empty ( in which case spherical rectangles with such angles do not exist ) or an interval @xmath291 $ ] , for some integer @xmath292 depending on @xmath260 . in the latter case , there are exactly @xmath293 combinatorially distinct nets of marked spherical rectangles of the first type with given @xmath260 and @xmath280 . [ kappamax ]
let @xmath2 be a marked spherical rectangle with @xmath280 satisfying assumptions 1 and 2 .
the net of @xmath2 can not be obtained from a net of some other spherical rectangle by one of the operations inverse to operations i - v if and only if one of the following twelve conditions is satisfied : @xmath295 ; @xmath296 ; @xmath297 ; @xmath298 ; @xmath299 ; @xmath300 ; @xmath301 ; @xmath302 ; @xmath303 ; @xmath304 ; @xmath305 ; @xmath306.for given @xmath260 with @xmath307 , a net of a marked spherical rectangle with @xmath280 satisfying assumptions 1 and 2 may satisfy at most one of these conditions.at most one ( up to combinatorial equivalence ) net of a marked spherical rectangle @xmath2 satisfying assumptions 1 and 2 , with given @xmath260 and @xmath280 , may satisfy any of these conditions . the value of @xmath210 for such rectangle @xmath2 is @xmath308.\ ] ] _ proof .
_ one can easily check case by case that none of the operations inverse to operations i - v can be applied if and only if one of the conditions ( a)-(l ) is satisfied .
note that it is enough to assume @xmath309 , and to check that with this assumption none of the operations inverse to operations i , iii , v can be applied if and only if one of the conditions ( a),(b),(d),(e),(g),(i),(k ) is satisfied .
the case @xmath310 follows by rotation of the net exchanging @xmath61 with @xmath163 , @xmath159 with @xmath160 , @xmath180 with @xmath181 , @xmath219 with @xmath220 , and @xmath214 with @xmath215 .
if @xmath2 satisfies ( a ) then @xmath311 is an odd integer , @xmath312 , @xmath313 . if , in addition , @xmath314 then @xmath315 and @xmath316 , thus @xmath317 , and the net of @xmath2 is completely determined by its angles .
such a net exists when @xmath318 thus @xmath319 .
the case when @xmath320 is treated similarly , and a net with @xmath320 satisfying ( a ) exists when @xmath321 .
the net with @xmath283 satisfying ( a ) exists when @xmath322 .
if @xmath2 satisfies ( b ) then @xmath323 thus @xmath324 , @xmath325 , @xmath326 , @xmath314 , @xmath315 , @xmath327 thus @xmath328 , @xmath329 thus @xmath330 . a net satisfying ( b )
is completely determined by its angles .
it exists when @xmath318 , thus @xmath331 .
similarly , if @xmath2 satisfies ( c ) then @xmath323 , and its net is completely determined by its angles .
it exists when @xmath332 , @xmath333 , @xmath334 .
if @xmath2 satisfies ( e ) then @xmath337 is a positive even integer , @xmath338 , @xmath339 , @xmath314 , @xmath340 , @xmath315 , @xmath341 thus @xmath342 , @xmath343 , and the net is completely determined by its angles .
it exists when @xmath318 thus @xmath344 .
similarly , if @xmath2 satisfies ( f ) then @xmath337 , and its net is completely determined by its angles .
it exists when @xmath345 , @xmath346 , @xmath347 .
if @xmath2 satisfies ( g ) then @xmath348 thus @xmath349 , @xmath350 , @xmath351 , @xmath314 , @xmath315 , @xmath352 thus @xmath353 , @xmath354 thus @xmath355 . a net satisfying ( g )
is completely determined by its angles .
it exists when @xmath318 , thus @xmath356 .
similarly , if @xmath2 satisfies ( h ) then @xmath348 , and its net is completely determined by its angles .
it exists when @xmath350 , @xmath351 , @xmath357 .
if @xmath2 satisfies either ( i ) , ( j ) , ( k ) or ( l ) then @xmath358 is an odd integer , @xmath359 . if @xmath2 satisfies ( i ) then @xmath360 , @xmath283 , @xmath361 , @xmath315 .
if @xmath2 satisfies ( j ) then @xmath362 , @xmath283 , @xmath336 , @xmath363 .
if @xmath2 satisfies ( k ) then @xmath364 , @xmath314 , @xmath340 , @xmath315 .
a net satisfying ( k ) exists when @xmath365 , thus @xmath366 . if @xmath2 satisfies ( l ) then @xmath367 , @xmath320 , @xmath336 , @xmath368 . a net satisfying ( l )
exists when @xmath369 , thus @xmath370 . a net satisfying either ( i ) , ( j ) , ( k ) or ( l ) is completely determined by its angles .
[ kappamaxplus1 ] a marked spherical rectangle @xmath2 satisfying assumptions 1 and 2 , with given @xmath260 and @xmath280 exists if and only if @xmath371 and there are exactly @xmath372\ ] ] combinatorially distinct nets of such spherical rectangles .
_ it follows from corollary [ exist ] and lemma [ kappamax ] that existence of a rectangle @xmath2 satisfying conditions of lemma [ kappamaxplus1 ] implies that the number in ( [ eq : kappamax ] ) is non - negative , which is equivalent to ( [ eq : exist ] ) , and that the number of combinatorially distinct nets of such rectangles in that case is the number in ( [ eq : kappamax ] ) plus @xmath5 , which is the number in ( [ eq : kappamaxplus1 ] ) .
thus we have only to prove that for any @xmath260 satisfying ( [ eq : exist ] ) there exists a rectangle @xmath2 satisfying conditions of lemma [ kappamaxplus1 ] .
it follows from corollary [ exist ] that , if a rectangle with given @xmath260 satisfying conditions of lemma [ kappamaxplus1 ] exists , there exists also a rectangle with @xmath212 with the same angles satisfying the same conditions .
we are going to construct a net of such a rectangle for any @xmath260 satisfying ( [ eq : exist ] ) .
there are three possible cases : ( i ) @xmath373 and @xmath374 ; ( ii ) @xmath373 and @xmath375 ; ( iii ) @xmath376 and @xmath374 .
note that @xmath376 and @xmath375 is not possible because @xmath377.in case ( i ) let @xmath283 , @xmath378 , @xmath379 , @xmath380 , @xmath381.in case ( ii ) let @xmath382 , @xmath383 , @xmath379 , @xmath384 , @xmath385.in case ( iii ) let @xmath386 , @xmath387 , @xmath378 , @xmath388 , @xmath389.this completes the proof of lemma [ kappamaxplus1 ] .
[ special ] a marked spherical rectangle @xmath2 satisfying assumptions 1 and 2 with given @xmath260 is special if an only if there exists a rectangle with the same angles whose net has @xmath295 . for given @xmath260 satisfying conditions of definition [ df : special ] with @xmath392 , there are @xmath393 special rectangles satisfying assumptions 1 and 2 , with @xmath295 and @xmath281 .
_ if @xmath281 for the net of @xmath2 then @xmath295 . from lemma [ kappa ] ( c ) , iteration of operation vi applied to @xmath2 results in a unique rectangle @xmath394 with the same angles as @xmath2 , @xmath295 and @xmath280
thus @xmath394 satisfies lemma [ kappamax ] ( a ) .
if @xmath280 for the net of @xmath2 then , from corollary [ exist ] , there is a unique rectangle with the same angles as @xmath2 satisfying one of the conditions ( a)-(l ) of lemma [ kappamax ] .
one can easily check that @xmath84 is an odd integer , @xmath312 and @xmath313 only in case ( a ) of lemma [ kappamax ] .
conversely , a rectangle satisfying condition ( a ) of lemma [ kappamax ] is clearly special , thus any rectangle @xmath2 with the same angles is also special . finally , the number in ( [ extra - special ] ) is obtained by counting distinct rectangles that can be obtained from a rectangle with @xmath295 by iterating operation vi and its inverse ( compare with lemma 11.2 in @xcite ) .
[ count - all ] let @xmath2 be a marked spherical rectangle satisfying assumptions 1 and 2 , with given @xmath260
. if @xmath2 is not special then there are ( [ eq : kappamaxplus1 ] ) combinatorially distinct nets of marked spherical rectangles satisfying assumptions 1 and 2 with the same angles as @xmath2 .
if @xmath2 is special then the number in ( [ eq : kappamaxplus1 ] ) is @xmath395 , and there are additionally ( [ extra - special ] ) nets of marked spherical rectangles satisfying assumptions 1 and 2 with the same angles as @xmath2 , thus the total number of combinatorially distinct nets is @xmath396 _ proof of theorems [ theorem1 ] and [ theorem2 ] .
_ lemmas [ kappamaxplus1 ] and [ special ] imply that a marked spherical rectangle satisfying assumptions 1 and 2 exists if and only if ( [ eq : exist ] ) holds .
since this condition is symmetric with respect to @xmath397 and @xmath398 , it remains true for spherical rectangles of the second type satisfying @xmath399 , as any such rectangle can be obtained from a rectangle satisfying assumptions 1 and 2 by a reflection preserving @xmath61 and @xmath163 , exchanging @xmath159 and @xmath160 . if @xmath2 is a marked spherical rectangle of either first of second type satisfying @xmath400 , replacing @xmath61 by @xmath159 as an initial corner , and relabeling the corners accordingly , results in a marked spherical rectangle @xmath7 of either second or first type , with the integer parts of the angles @xmath401 and @xmath402 . applying the above arguments to @xmath7 we see that a marked spherical rectangle @xmath2 with @xmath400 exists if and only if @xmath403 combining ( [ eq : exist ] ) and ( [ eq : exist1 ] ) we get the statement of theorem [ theorem1 ] .
the statement of theorem [ theorem2 ] follows from ( [ eq : kappamaxplus1 ] ) and ( [ total - special ] ) applied to either @xmath2 or @xmath7 in a similar way .
_ due to proposition [ isometry ] , it is enough to prove that the net @xmath71 of @xmath2 is symmetric with respect to a transformation exchanging @xmath61 with @xmath163 and @xmath159 with @xmath160 , and that this symmetry of @xmath71 maps any short arc of @xmath71 connecting @xmath159 with @xmath160 is mapped to a ( possibly , different ) short arc connecting @xmath159 with @xmath160 .
to show this , we have only to check ( assuming that @xmath2 satisfies assumptions 1 and 2 ) that @xmath408 , @xmath409 , and @xmath410 in the notations of theorem [ classification ] .
suppose first that @xmath295 .
then @xmath411 implies @xmath412 , and @xmath413 implies @xmath414 ( see ( [ inequality ] ) in remark [ angles ] ) . adding up these two equalities yields @xmath409 , and subtracting them yields @xmath410 . if @xmath182 but @xmath178 then @xmath320 , thus @xmath336 , @xmath417 , @xmath368 , and @xmath418 . since @xmath411
, we have @xmath419 , thus @xmath420 , a contradiction . similarly , @xmath176 and
@xmath183 is not possible .
this completes the proof .
[ thm : symmetric ] a spherical rectangle @xmath2 with the angles at two of its opposite corners equal @xmath13 , and the angles at two other opposite corners equal @xmath188 , exists if and only if @xmath421 .
if @xmath422 is even then there are @xmath423 non - isometric spherical rectangles with these angles , @xmath424 of them satisfying assumption 1 .
if @xmath422 is odd then there are @xmath425 non - isometric spherical rectangles with these angles , @xmath426 of them satisfying assumption 1 .
the net in fig .
[ rectangle1313 ] is special , with @xmath427 , @xmath428 , @xmath429 .
according to ( [ total - special ] ) and theorem [ thm : symmetric ] , there is a unique net of a marked spherical rectangle of the first type with these angles .
the two nets in fig . [ rectangle3535 ] are special with @xmath430 , @xmath431 , @xmath429 . according to ( [ total - special ] ) and theorem [ thm : symmetric ] , there are two nets of marked spherical rectangles of the first type with these angles .
the net in fig . [ rectangle3737 ] is not special , with @xmath430 , @xmath432 , @xmath433 .
according to ( [ eq : kappamaxplus1 ] ) and theorem [ thm : symmetric ] , there is a unique net of a marked spherical rectangle of the first type with these angles .
the three nets in fig .
[ rectangle5757 ] are special with @xmath434 , @xmath432 , @xmath429 . according to ( [ total - special ] ) and theorem [ thm : symmetric ] , there are three nets of marked spherical rectangles of the first type with these angles
. 1 a. khovanskii , on a class of systems of transcendental equations , soviet math . dokl . , 22 ( 1980 ) , 762765 .
ching - li chai , chang - shou lin and chin - lung wang , mean field equations , hyperelliptic curves and modular forms : i , arxiv:1502.03297 .
g. darboux , sur une quation linaire , c. r. acad .
paris , t. 94 ( 1882 ) 16451648 .
a. eremenko , metrics of positive curvature with conic singularities on the sphere , proc .
( 2004 ) , 33493355 a. eremenko and a. gabrielov , on metrics of curvature 1 with four conic singularities on tori and on the sphere , arxiv:1508.06510 .
a. eremenko , a. gabrielov , m. shapiro and a. vainshtein , rational functions and real schubert calculus , proc .
ams , 134 ( 2006 ) , no .
4 , 949957 .
a. eremenko , a. gabrielov and v. tarasov , metrics with conic singularities and spherical polygons , arxiv:1405.1738 a. eremenko , a. gabrielov and v. tarasov , metrics with four conic singularities and spherical quadrilaterals , arxiv:1409.1529 a. eremenko , a. gabrielov and v. tarasov , spherical quadrilaterals with three non - integer angles , arxiv:1504.02928 .
s. finch , mathematical constants , cambridge up , cambridge , 2003 .
hermite , sur lquation de lam , extrait de feuilles authographies du course danalyse de lcole polytechnique , @xmath435 division , 1872 - 73 , @xmath436 leon .
oeuvres , t. iii , p. 118122 .
paris , gauthier - villars , 1912 .
f. klein , vorlesungen ber das insider und die aufsnug der clonking von fften grade , birkhuser verlag , basel , 1993 .
chang - shou lin and chin - lung wang , elliptic functions green functions and the mean field equations on tori , ann . math .
, 172 ( 2010 ) 911954 .
f. luo and g. tian , liouville equation and spherical convex polytopes , proc .
( 1992 ) , no . 4 , 1119.1129 .
g. mondello and d. panov , spherical metrics with conical singularities on a 2-sphere : angle constraints , arxiv:1505.01994 .
l. schneps , dessins denfants on the riemann sphere , in : the grothendieck theory of dessins denfants ( luminy , 1993 ) , 4777 , cambridge univ . press , cambridge , 1994 .
p. speissegger , the pfaffian closure of an o - minimal structure , j. fr die reine und angew . math . ,
508 ( 1999 ) 189211 .
m. troyanov , prescribing curvature on compact surfaces with conical singularities , trans .
, 324 ( 1991 ) 793821 .
l. van den dries , tame topology and o - minimal structures , london math .
lecture notes series 248 , cambridge university press , cambridge 1998 .
e. van vleck , a determination of the number of real and imaginary roots of the hypergeometric series , trans .
soc . , 3 ( 1902 ) 110131 . | we study spherical quadrilaterals whose angles are odd multiples of @xmath0 , and the equivalent accessory parameter problem for the heun equation .
we obtain a classification of these quadrilaterals up to isometry . for given angles ,
there are finitely many one - dimensional continuous families which we enumerate . in each family
the conformal modulus is either bounded from above or bounded from below , but not both , and the numbers of families of these two types are equal .
the results can be translated to classification of heun s equations with real parameters , whose exponent differences are odd multiples of @xmath1 , with unitary monodromy .
msc 2010 : 34m03 , 30c20 , 35j91 , 33e05 .
keywords : surfaces of positive curvature , accessory parameters , heun s equation , elliptic integrals , belyi functions . |
significant progress has been made in recent years in understanding the theory of the launching of jets from accretion disk / black hole systems @xcite .
simulations are able to produce jets starting with an accretion disk without ordered magnetic fields .
the fields are stretched by the differential rotation of the disk , and loops rise above , but remain attached to the disk , and ultimately the black hole .
a structure forms along the rotation axis that can have a relativistic poynting jet in the center surrounded by a magnetic field twisted into a helical shape by the disk rotation and filled by material from a disk wind or corona .
the jet is not fully accelerated and collimated until several hundred schwarzschild radii ( r@xmath0 ) . within this scale
, it might be possible to see evidence for the acceleration and collimation process and to constrain the theory .
there are very few objects in which structures in a jet of under a few hundred r@xmath0 can be observed , even with high frequency vlbi . a favorable combination of high mass black hole , proximity to the earth , and bright jet
is required .
galactic objects are close , but too low mass except for the galactic center .
the black hole at the galactic center has a high angular size black hole , but no observed jets .
the best source for such a study is m87 .
m87 has a super - massive black hole with a mass of about @xmath1 @xcite and , as the dominant galaxy of the virgo cluster , has a distance of only 16 mpc @xcite . for these parameters , the schwarzschild radius of about 60 au subtends 3.7 microarcseconds .
m87 has a bright jet that can be studied at the highest frequencies , and hence , highest resolutions available to very long baseline interferometry ( vlbi ) .
other galaxies that are as close or closer either have lower mass black holes or do nt have adequately bright jets to allow detailed observations .
the basic structure on sub - parsec scales of m87 was first published by @xcite based on 43 ghz global vlbi observations .
that structure includes a very wide opening angle in the first few tenths of a milli - arcsecond ( mas ) and a pronounced limb - brightened jet structure .
the wide opening angle suggests that the collimation region is beginning to be resolved .
the jet has been observed with vlbi many times since .
typically motions significantly less than the speed of light are seen , including recent results at 15 ghz where features are almost stationary @xcite . at lower resolution ,
farther from the core , superluminal motions of @xmath2 to @xmath3 have been seen , for example at the hst1 knot @xcite .
but at higher resolutions , the situation is unclear because any fast motions would be seriously under - sampled by the previous observations .
we report here on the status of our efforts to delineate properly the motions in m87 on scales from around 60 to several hundred r@xmath0 in projection using observations with the very long baseline array ( vlba : * ? ? ?
* ) at 43 ghz every 3 weeks for a year .
higher resolution could have been obtained using 86 ghz , or a global array , but the imaging difficulties and practical scheduling constraints made any such project far more difficult . for the declination of m87 ( @xmath4 ) ,
the typical resolution of the vlba at 43 ghz of @xmath5 mas ( elongated north - south ) corresponds to 110 by 54 r@xmath0 or 0.031 by 0.015 pc . note that distance along the jet is foreshortened if the jet is near the line - of - sight as suggested by the observation of superluminal motions .
for example , for the @xmath6 needed to see @xmath3 motions , the angle to the line - of - sight is likely to be near @xmath7 , giving a foreshortening by a factor of about 6 .
our study of the inner jet of m87 began with a study of archival data from 5 vlba observations at 43 ghz spaced at intervals of roughly a year .
the data were originally taken for various reasons , usually to use m87 as a phase reference source for observations of a nearby weak source .
these data were reprocessed and examined for indications of structural evolution @xcite . while the general character of the structure was consistent from epoch to epoch , it only seemed possible to identify corresponding components in more than one image by using the two most closely spaced observations .
if those components really were related , apparent speeds of @xmath8 to @xmath9 were seen . but
any higher speeds would have made it impossible to identify components at more than one epoch and , given the superluminal speeds seen in other sources and elsewhere in m87 , such a case could certainly not be excluded .
the images from the archival data also revealed emission on the opposite side of the `` core '' , or brightest feature , at 3 epochs .
that feature was also seen in the 15 ghz observations of @xcite who see it extending to about 3 mas from the core .
it is an open question as to whether that feature is the counterjet , or the inner jet with the radio core offset from the actual black hole location .
the facts that it is seen out to 3 mas , or about 800 r@xmath0 in projection and seems to get wider with distance from the core , make the offset black hole possibility seem unlikely , as does the observation of circular polarization at the core position @xcite . if it is the far side jet of a symmetric system , and a speed and brightness ratio can be determined , it can give strong evidence for the true speed and orientation to the line - of - sight of the jet , a possibility explored in @xcite .
the archival data raised a serious question about possible under - sampling of any motions so a project to make a fully sampled vlba movie at 43 ghz was begun .
the first step of that project was to determine the proper sampling rate . that was done using 6 observations at spacings of between 3 and 97 days during 2006 .
based on the results of those observations , an interval of 3 weeks was chosen for the movie and a year - long series of 18 observations was proposed .
the movie observations began on 2007 january 27 and were still in progress at the time of this meeting .
images from the first 11 observations were available for the meeting .
they are shown figure [ montage ] .
the appearance of the source is generically similar at all epochs .
the source is always a distinctly edge brightened jet with emission filling in the central region .
the feature on the counterjet side of the brightest emission is clearly seen at all epochs . in some of the epochs
, there appears to be a central spine .
the jet has structure , but that structure has more of a lumpy but continuous nature and is not easily described as a series of `` components '' . some features do seem to appear in multiple images , but no very clearly defined features persist for many epochs . to improve the sensitivity to weak stuctures and to bring out the persistent aspects of the structure ,
a weighted average image was made from the first 9 epochs and is shown in figure [ stack ] .
note that any detailed , changing structure is washed out in the average .
when the images are presented as a montage as in figure [ montage ] , any motions are not immediately apparent .
but when assembled as a movie and played reasonably fast , it is clear that there is overall fast , outward motion .
it is something like watching a smoke plume that is moving in bulk , but that also has rapid evolution of its internal structure .
it is impossible to display such a movie here , but one can be found at ` http://www.aoc.nrao.edu/~cwalker/m87 ` . that website will be updated as more
epochs become available .
the movie of m87 shows fast motions . to quantify the speed in the absence of persistent components , the difference in positions of apparently related features in adjacent epochs were measured .
such differences consistently came out to be about 0.5 mas indicating a speed of about 9 mas yr@xmath10 or about @xmath11 .
that is much faster than motions seen in other vlbi observations of the inner regions of this source , but those observations could not have measured such speeds because of undersampling .
in fact , even the current observations are undersampled because the motion is about 2.5 times the beam width per epoch .
once this situation was clear , the pilot project was revisited , but any such fast motions were still not clear in those images .
it seems a significant sequence is required for them to be obvious . as a result of this speed determination , a proposal was made to extend the sequence with 10 additional observations at 5 day intervals .
those observations are in progress as this is being written .
the m87 43 ghz movie project is a work in progress .
but it has already shown that the inner jet , on scales near 100 r@xmath0 , has mildly relativistic motions , contrary to the conclusions of previous observations .
it has an edge brightened structure with a possible central spine and with a counterjet side feature with motions whose character is not yet clear .
as the observations and processing are completed , attempts will be made to delineate the velocity field and to constrain models of jet formation .
m87 presents a prime target for vsop2 .
ground observations with the vlba at 43 ghz , and with a global array at 86 ghz @xcite are starting to resolve the wide - opening - angle base of the jet .
another factor of four in resolution over our vlba results will allow significant details to be seen in this important region and will help guide the theoretical understanding of the structure and dynamics of the jet launch region .
but the observations must be made frequently enough that they are not under - sampled .
the data presented here show motions of about 0.025 mas per day , which , with a vsop2 beam of 0.05 mas , should be sampled daily much more frequently than now envisioned . to obtain sufficient information on the dynamics , this observing pace must be maintained for at least a month .
this will be a major strain on observing resources , but this is probably the best opportunity for vsop2 to have an impact on the understanding of the physics in the launch region of an astrophysical jet .
this research has made use of the nasa astrophysics data system .
w. junor was partially supported through nsf ast 98 - 03057 . c. ly acknowledges support through the nrao graduate research program .
p. hardee is supported by nasa / msfc cooperative agreement ncc8 - 256 and nsf award ast-0506666 to the university of alabama .
the national radio astronomy observatory is a facility of the national science foundation , operated under cooperative agreement by associated universities , inc .
krichbaum , t. p. , lee , s. s. , lobanov , a. p. , marscher , a. p. , & gruwell , m. a. 2008 , in asp conf .
386 , extragalactic jets : theory and observation from radio to gamma ray , eds . t. a. rector and d. s. de young , in press ( arxiv:0708.3915v1 ) | m87 provides the best opportunity to study the base of a jet where it is collimated and accelerated .
the size of that region scales with the mass of the black hole , and m87 has the best combination of high mass , proximity to the earth , and presence of a bright jet .
vlbi observations of m87 can probe regions under 100 gravitational radii where theoretical studies suggest that the jet formation and acceleration occurs . a one - year sequence of 43 ghz observations every 3 weeks on the vlba is being used to study the structure and dynamics in this region .
initial results from that effort are reported here , including the observation of rapid motions - sufficiently rapid that more frequent observations are planned for early 2008 .
the contribution ends with a discussion of prospects for future vlbi observations of m87 with vsop2 . for vsop2
, a strong recommendation is made that a series of daily observations at 43 ghz be planned . |
the structure of amorphous semiconductors is believed to be well represented by the continuous random network ( crn ) model introduced by zachariasen more than sixty years ago @xcite . as a result
, the generation of high quality crns has been the subject of investigation for many years .
the first crns were built by hand , see for instance the work of polk @xcite .
nowadays , the generation of crns is mostly carried out on computers .
the first computer - generated networks , which date back to the sixties and seventies , typically contain a few hundred particles . more advanced algorithms and faster computers have increased the size of the networks that can be handled to a few thousand atoms , with simulation cells of up to . as the simulation cells increase in size , actual devices
have decreased in size .
for example , the thickness of solar cells based on amorphous silicon has already decreased to ; and because in - plane periodicity after approximately is expected to be a good approximation of the macroscopic lateral size , a reasonable solar cell model would require a simulation cell of , containing approximately 45,000 atoms .
this is only one order of magnitude larger than currently feasible . for other electronic devices ,
lithography on technology is expected to be reached in the coming decade . in this work ,
we present a computational approach to generating large crns , and discuss the properties of high quality networks containing up to 20,000 particles .
this achievement shows that the generation of device - size atomic configuration networks is within reach .
we begin by describing the algorithm of wooten , winer , and weaire ( www ) , which has been the basis of the best crns generated to date .
we then move on to describe a number of improvements made to the original www algorithm by barkema and mousseau in 1999 @xcite .
these improvements accelerate the relaxation by two orders of magnitude or more .
both the original and the improved www algorithm , however , scale poorly with system size , since the computational effort per attempted local atomic rearrangement increases linearly with system size . in this work ,
we introduce local force and energy evaluations and improve the scaling of computation time with system size @xmath0 significantly , namely to a constant per attempted move plus @xmath1@xmath2 per accepted move .
we also demonstrate how parallel processing can be used to realize an additional speedup , with parallel efficiencies of over 50% .
the significance of these improvements is demonstrated by generating 10,000-atom and 20,000-atom crns .
we then discuss the structural and electronic properties of these models and conclude with an outlook on future research , aiming towards the generation of device - size atomic networks .
in 1985 , wooten , winer , and weaire presented an algorithm for the generation of four - fold coordinated crns @xcite . in their approach ,
a configuration consists of the coordinates of @xmath0 atoms and a list of the @xmath3 bonds between them .
the structural evolution consists of a sequence of bond transpositions as illustrated in fig .
[ fig : wwwmove ] .
diagram depicting the www bond transposition .
four atoms a , b , c , and d are selected following the geometry shown left ; two bonds , ab and cd , are then broken and atoms a and d are reassigned to c and b , respectively , creating two new bonds , ac and bd shown right.,width=226 ] within the original www approach , the generation of a crn starts with a cubic diamond structure which is randomized by a large number of such bond transpositions .
after randomization , the network is relaxed through a sequence of bond transpositions , accepted with the metropolis acceptance probability @xcite : @xmath4,\ ] ] where @xmath5 is the boltzmann constant , @xmath6 is the temperature , and @xmath7 and @xmath8 are the total quenched energies of the system before and after the proposed bond transposition . with an explicit list of neighbors , it is possible to use a simple interaction such as the keating potential @xcite to calculate energy and forces : @xmath9 where @xmath10 and @xmath11 are the bond - stretching and bond - bending force constants , and @xmath12 is the si - si strain - free equilibrium bond length in the diamond structure .
usual values are @xmath13 ev @xmath14 and @xmath15 . with the approach described above wooten and weaire generated a 216-atom model with an angular distribution as low as 10.9 degrees @xcite .
a decade later , using the same approach but more computing power , djordjevi , thorpe , and wooten produced two large 4096-atom networks of even better quality , with a bond - angle distribution of 11.02 degrees for configurations without four - membered rings and 10.51 degrees when these rings are allowed @xcite .
the www algorithm in its original form is capable of producing high quality amorphous networks containing of the order of a thousand atoms ; it is not well suited to generate much larger networks .
this is mostly due to the fact that for each proposed bond transposition , about one hundred energy and force calculations are required , each scaling as @xmath1@xmath2 with system size @xmath0 .
these @xmath1@xmath2 operations are the bottleneck of the algorithm . in 1999 ,
barkema and mousseau ( bm ) presented a number of modifications to the original www algorithm , partially aimed at resolving these poor scaling properties @xcite .
their modifications are summarized below : 1 .
starting point for the relaxation in this case is a truly random configuration whereby the atoms are placed at random locations in a periodic box at the crystalline density .
this guarantees that the resulting network is not contaminated by some memory of the crystalline state .
2 . after a bond transposition in the original www approach ,
the structure is always completely quenched , i.e. , all the atomic coordinates are fully relaxed .
after the quench , the bond transposition is either accepted or rejected based on the metropolis probability .
in contrast , bm determines a threshold energy before quenching . during
the quench the final quenched energy is continuously estimated .
relaxation is stopped when it becomes clear that the threshold energy can not be reached so that the bond transposition will eventually have to be rejected .
this leads to a large reduction in the number of force evaluations associated with rejected bond transpositions .
a local relaxation procedure is used whenever possible . immediately after a bond transposition
, only a small cluster of atoms in the model experiences a significant force .
this cluster consists of the atoms directly involved in the bond transposition ( marked a , b , c , and d in fig .
[ fig : wwwmove ] ) and of nearby atoms , typically up to the fourth neighbor shell of the four transposition atoms .
the number of atoms in such a cluster is about 150 .
it therefore suffices to calculate the force _ locally _
( i.e. , only for the 150 or so atoms inside the cluster ) rather than _ globally _ ( i.e. , for all the atoms in the model ) .
+ calculating the force on a cluster of atoms is an @xmath1(1 ) operation , which means that it is independent of the total system size .
local force calculations are therefore much cheaper than global @xmath1@xmath2 force calculations . by using a local relaxation scheme we can increase the efficiency of the algorithm significantly . still , to make the final accept / reject decision on the proposed move , the total keating energy of the system has to be calculated , which is again an @xmath1@xmath2 operation . in practice
, a switch must be made from local to global relaxation , usually after about ten local relaxation steps .
the zero temperature case is treated specifically . using the improved www algorithm ,
barkema and mousseau generated two 1000-atom models with bond angle deviations as low as 9.20 degrees @xcite .
furthermore , using the same algorithm they generated a 4096-atom model with an angular deviation of 9.89 degrees .
all models show structural and electronic properties in excellent agreement with experiments .
while the improved www algorithm can successfully generate networks containing several thousand particles , it does not deal well with systems of 10,000 particles or more .
each attempted bond transposition still requires one or more @xmath1@xmath2 energy evaluations . in this section
, we present an algorithm for attempting bond transpositions that is completely local , i.e. , free of @xmath1@xmath2 operations for unsuccessful bond transpositions .
to exploit the local nature of the bond transpositions , we need to introduce the concept of local energy : we assign to each atom @xmath16 an energy @xmath17 such that with @xmath18 the keating energy of the system given by eq .
( [ eq : keating ] ) .
one way to achieve this is to divide the energy due to two - body interactions equally between the two participating atoms and to assign the energy of three - body interactions to the central atom of the corresponding triple .
thus , we obtain : @xmath19.\end{aligned}\ ] ] here , the constants @xmath10 , @xmath11 , and @xmath20 are defined as in eq .
( [ eq : keating ] ) ; @xmath21 represents the vector pointing in the direction of the @xmath22-th bond away from atom @xmath16 .
the energy @xmath23 of a cluster @xmath24 of atoms can now be calculated using : @xmath25 the force on the atoms inside the cluster is obtained from the derivative of the keating energy with respect to the atomic coordinates .
care has to be taken for atoms on the edge of the cluster since these atoms also interact with atoms outside the cluster : due to the two- and three - body terms in the keating potential , all atoms interact with their first and second nearest neighbors ; for atoms located on the edge of the cluster , some of these neighbors are outside the cluster .
starting point is a random configuration generated using the method described in ref . .
this guarantees that the resulting configurations are not contaminated by some memory of the crystalline state .
assuming that the total keating energy of the initial configuration is known and equals @xmath18 , www moves can be attempted locally as follows : 1 .
a threshold energy @xmath26 is determined by using the equation : @xmath27 where @xmath28 is a random number uniformly drawn from the interval @xmath29 .
the move is accepted if the attempted bond transposition leads to a configurational energy below the threshold energy ; otherwise it is rejected .
the four atoms involved in the attempted bond transposition and all atoms up to the fourth neighbor shells of these four atoms are grouped into a cluster .
such a cluster contains about 150 atoms .
a list is constructed of all the bonds that contribute to the force on the atoms inside the cluster . as was explained above ,
some of these bonds involve atoms outside the cluster .
for each bond we store the labels of the two atoms constituting the bond , the @xmath30 , @xmath31 , and @xmath32 components of the bond vector ( taking care of the periodic boundary conditions ) , and the square of the bond length .
we then calculate the cluster energy @xmath23 using eq .
( [ eq : clusterenergy ] ) . in the calculation of the cluster energy most bonds
are encountered more than once . to increase efficiency ,
a bond ( i.e. , its set of three components ) is calculated only once during an energy or force evaluation ; once a bond has been calculated it is time - stamped with an integer flag and the bond information is stored .
later references to the same bond are then retrieved from memory .
+ we also store the energy of the atoms that remain outside the cluster : .
we then perform the bond transposition to obtain the geometry shown in the right frame of fig .
[ fig : wwwmove ] .
the system is relaxed locally , i.e. , only atoms inside the cluster are allowed to move . at each relaxation step
we use eq .
( [ eq : clusterenergy ] ) to calculate the energy of the cluster @xmath23 and the atomic forces , again making sure each bond is calculated only once , and perform structural relaxation as in the original and improved www algorithms . at each relaxation
step the total energy of the system is equal to .
local relaxation is continued until the energy has converged or until it becomes clear that the threshold energy can not be reached . in the local relaxation procedure above , the computational effort per attempted bond transposition does not grow with the system size .
local relaxation alone , however , is not sufficient and we also have to relax globally to relieve any strain that may have built up between atoms on the edge of a cluster and non - cluster atoms . for clusters extending up to the fourth neighbor shell around the atoms directly involved in the bond transposition we find that global relaxation can lower the configurational energy typically by less than .
we therefore switch from local to global relaxation when , during local relaxation , the energy comes to within of the threshold energy . in most cases
, this leads to the move being accepted .
we have developed a parallel version of our algorithm with the aim of harnessing the tremendous power of parallel computers .
the parallel algorithm is in bulk synchronous parallel ( bsp ) style @xcite with alternating phases of computation and communication , separated by a synchronization of all the processors .
the parallel algorithm has two main parts , local relaxation and global relaxation .
the local relaxation is done in parallel by letting every processor try a sequence of randomly chosen bond transpositions , until one of the processors finds an acceptable transposition .
the processors work independently but synchronize at regular intervals to communicate their success or failure to the others .
if more than one processor succeeds , an arbitrary bond transposition is chosen as the winner .
this approach requires the replication of all the atomic data .
fortunately , the memory storage needed is limited to an array of @xmath33 atomic coordinates and a few other arrays of size @xmath0 , which usually can be stored on every processor .
( the www algorithm and its variants are demanding in cpu time , but not in memory requirements . ) furthermore , this approach also requires refreshing the atomic data when the positions change , causing communication between the processors .
this only happens after a bond transposition is accepted , which is a relatively rare event ( of the order of once every thousand attempts ) . for these reasons ,
we choose to replicate the data instead of distributing them , and develop a parallel local relaxation algorithm based on replicated data .
it is crucial to choose a suitable time interval between successive synchronizations .
if this interval is too short , the time of the synchronization itself will become dominant , or fluctuations in the amount of work of the different processors will become visible ; in longer intervals such fluctuations are averaged out and have less impact .
if the interval is too long , it becomes likely that one ( or more ) accepted moves are found in every time interval .
part of the work in a successful time interval is wasted , because the processor that finds an accepted move waits until the others have finished their ( useless ) computations .
thus a high success rate means that much time is wasted . from the point of view of parallel efficiency
, the ideal situation occurs when most time intervals fail to produce a successful bond transposition . on average ,
little cpu time is then wasted .
a processor decides to synchronize based on the total number of relaxation iterations performed during all its bond transposition attempts .
each iteration requires of the order of @xmath34 floating point operations ( flops ) . simply counting the number of attempts
would not give a good indication of the total amount of work performed by a processor , since the number of iterations per attempt may vary , depending for instance on the observed energy decrease .
thus , a processor synchronizes after every @xmath35 iterations .
we determined the parameter @xmath35 empirically , and found on our machine , a cray t3e , that values in the range @xmath35=10200 give the best performance on @xmath36 processors ; for higher numbers of processors this range becomes smaller and the choice of @xmath35 becomes critical . for @xmath37 , we used @xmath38 .
the bsp cost model @xcite can be helpful in choosing @xmath35 .
for instance , the bsp parameter @xmath39 , representing the synchronization time of the parallel computer , can be used to find a lower bound for @xmath35 .
the global relaxation is done in parallel by partitioning the simulation cell over the @xmath40 processors of the parallel computer and letting every processor compute the energies @xmath17 , forces , and displacements for the atoms in its own part of the cell .
in contrast to the local relaxation , it is now justified to have all the processors participate in one relaxation : the amount of work , @xmath41@xmath2 , in an iteration is much more than in the case of the local relaxation .
in fact , processors are even obliged to participate , because there is no other useful work to do : most likely the global relaxation succeeds and provides the starting point for the remainder of the computation .
communication arises in the global relaxation because processors need data from other processors concerning atoms that lie near inter - processor boundaries .
thus at the end of an iteration , a processor has to communicate the changes in the positions of its boundary atoms ( i.e. , atoms within two bonds from an atom on another processor ) .
also , some atoms may move to another processor . to reduce the size of the boundary region , we use three types of partitioning @xcite : standard cubic ( sc ) , which splits the simulation cell into @xmath42 subcubes ; body - centred cubic ( bcc ) , which splits the cell into @xmath43 truncated octahedra centred at the lattice sites of the bcc lattice ; and face - centred cubic ( fcc ) , which splits the cell into @xmath44 rhombic dodecahedra centred at the lattice sites of the fcc lattice . the new bcc and fcc partitionings generate about 10% less communication than the commonly used sc partitioning . with these three partitionings
, we can choose from a wide range of processor numbers @xmath40 , and in particular we can employ every parallel computer where @xmath40 is a power of two .
we have implemented the parallel algorithm using the bsplib communications library @xcite on a cray t3e . in the local relaxation
, we have achieved a speedup of 18.6 on 32 processors , corresponding to an efficiency of 58% .
the efficiency loss is due both to the fluctuations in work load per local iteration ( caused by small differences in the numbers of cluster atoms ) and to waiting time at the end of successful time intervals . at the optimum value of @xmath35 ,
both effects are significant .
losses due to the synchronizations themselves are negligible , as one synchronization takes less than 1% of the time of a local iteration .
the replication time was found to be the same as the time of 14 local iterations , but overall this time is negligible since data only need to be replicated after a successful attempt .
in the global relaxation , we have achieved a speedup of 19.3 on 32 processors for the 20,000-atom model , corresponding to an efficiency of 60% , see @xcite for more details . here
, the efficiency loss is mainly due to redundant computations for boundary regions , and communication of data for boundary atoms .
the total speedup of our parallel version depends on the mixture of local and global relaxations needed .
this mixture is influenced by a variety of parameters such as the temperature @xmath6 in the metropolis acceptance criterion ( [ eq : metropolis ] ) and the expected reduction due to the global relaxation . in our simulations ,
the amounts of cpu time spent on local relaxation and on global relaxation were nearly equal .
using the scalable www algorithm we have generated one 10,000-atom amorphous silicon network and one 20,000-atom network . in this section , we discuss the structural and electronic properties of these networks . in table
[ tab : properties ] , we compare our configurations relaxed with the keating potential with those of djordjevi , thorpe , and wooten @xcite and with models generated by barkema and mousseau using the improved www algorithm @xcite .
we also provide irreducible ring statistics . [ cols="^,<,<,<,<,<,<,<",options="header " , ] the properties of the networks after relaxation with the ( modified ) sw potential are reported in table [ tab : mswproperties ] . for all configurations ,
the bond angle distribution widens and the density decreases .
[ fig : rdf ] shows the rdf for the 10,000-atom and 20,000-atom models compared to the experimental rdf obtained by laaziri _
et al . _ on annealed _
a_-si samples prepared by ion bombardment @xcite .
agreement is excellent .
however , configurations differing widely in topology can easily produce similar rdfs .
agreement with the experimental rdf must therefore be regarded as a minimum demand on a high quality crn .
radial distribution function for the 10,000-atom model ( top ) and the 20,000-atom model ( bottom ) after relaxation with the modified stillinger - weber potential ( solid line ) .
the dashed line shows the experimental result from ref . ; distances are in .,title="fig:",width=302 ] radial distribution function for the 10,000-atom model ( top ) and the 20,000-atom model ( bottom ) after relaxation with the modified stillinger - weber potential ( solid line ) .
the dashed line shows the experimental result from ref . ; distances are in .,title="fig:",width=302 ] a more stringent criterion that can be used to evaluate the quality of a model is the coordination number of the atoms . using the minimum of the rdf beween the first and second neighbor peak as the nearest neighbor cut - off distance and after relaxation with the modified stillinger - weber potential , we observe that the 10,000-atom and 20,000-atom model develop 0.9% and 0.2% of coordination defects , respectively . while structural averages provide good insight into the overall quality of a model , they do not say much regarding local environments . it is therefore also important to look at the electronic properties of our models : even small densities of highly strained geometries or defects will be picked up as states in the gap of the electronic density of states ( edos ) . in fig .
[ fig : edos ] we show the edos of the 10,000-atom and 20,000-atom models .
the _ fireball _ local - basis ab - initio code @xcite was used to obtain the edos .
a remarkable feature of the state densities shown here is the absence of states in the gap , leading to a perfect gap of for both models .
we have presented here a scalable version of the www algorithm which allows for local atomic rearrangements to be tried using only @xmath1(1 ) operations .
we have developed an efficient parallel version which achieves good load balance and limits communication .
the scalable performance of the algorithm has been demonstrated by generating one 10,000-atom and one 20,000-atom model .
structural and electronic properties of these models are excellent and they compare well to experiments . these high - quality models have the long term goal of accurately modeling devices such as solar cells . at this point , using periodic - boundary conditions in the two extended directions , we are able to simulate _ a_-si films with thinkness of about .
once such atomic configurations become available , the role of various structural and electronic defects can be studied .
we thank dave drabold for communicating to us the edos of the 10,000-atom and 20,000-atom models . | the atomic structure of amorphous materials is believed to be well described by the continuous random network model .
we present an algorithm for the generation of large , high - quality continuous random networks .
the algorithm is a variation of the _ sillium _ approach introduced by wooten , winer , and weaire . by employing local relaxation techniques , local atomic rearrangements
can be tried that scale almost independently of system size .
this scaling property of the algorithm paves the way for the generation of realistic device - size atomic networks . |
in the last few years , several experiments and observations have shown strong evidences that the expansion of our universe is accelerating , under the influence of a mysterious dark energy component@xcite .
the dark energy presents an equation of state parameter @xmath3 and represents about 74% of the universe . when @xmath4 we have the special case of the cosmological constant - or vacuum energy - as responsible for the universe acceleration ( it is the standard @xmath5 cosmological model ) .
we also know that around @xmath6 26% of the universe are composed by matter ( baryonic + dark matter ) ; analyzing the growth of these matter density fluctuation , we can describe the formation of the large structures , such as clusters and super - clusters of galaxies .
the mass function of galaxies is a powerful tool to study the large scale structure formation in the universe .
the standard analytical approach was developed in 1974 by press and schechter@xcite ( hereafter ps ) , and it is still used today , for its success and simplicity .
the ps approach uses a gaussian distribution to describe the primordial matter density fluctuation field . inspired by tsallis @xmath2-nonextensive statistics@xcite and kinetic theory@xcite ,
we have proposed a new mass function@xcite , replacing the gaussian by a non - gaussian distribution - a power law ( pl ) distribution .
the pl mass function has a physically motivated free @xmath2 parameter ( related with the long range gravitational correlations between particles ) , which provides malleability to fit observational data .
also , if @xmath7 we recover the original gaussian distribution .
the pl mass function is an extension of the ps one , and presents the same analytical simplicity .
we have compared the ps mass function with the pl one , in the special case of the @xmath5 cosmological model@xcite . using the x - ray flux - limited sample of galaxy clusters from reiprich and boehringer@xcite ( hiflugcs ) in a @xmath8 statistical analysis , and applying independent cosmological tests ( bao and shift parameter ) to better constrain the results
, we concluded that the ps approach presents incompatibilities with the independent cmb data .
on the other hand , using the pl mass function , we have better cosmological parameters , and we note an overlap with the cmb data for a large range of values of its @xmath2 free parameter . these results , although encouraging , are limited to the @xmath5 model ( @xmath4 ) . the mass function could be used to probe the dark energy behavior in the universe , and works using homogeneous and non - homogeneous dark energy in the large scale structure formation are beginning to flourish in the literature .
in this work we use a model of homogeneous dark energy from percival@xcite . by allowing other values for the @xmath1 dark energy parameter
, we can compare the applicability of the ps and the pl mass functions to probe dark energy models .
the distribution of bound objects with masses between @xmath9 and @xmath10 , using the gaussian distribution to describe the primordial density fluctuation field @xmath11 ( the ps approach ) reads@xcite : @xmath12 where @xmath13 is the mean squared fluctuation , and @xmath14 is the critical @xmath15 for collapse . now , if instead of gaussian we use the pl distribution for the initial fluctuations , we derive the follow expression@xcite @xmath16 ^{\frac{1}{\left ( 1-q\right ) } } \label{fm - pl}\end{aligned}\ ] ] where the factor @xmath17 is a one - dimensional normalization constant . in the limit
@xmath18 the above pl expression reduces to the standard gaussian approach .
we performed a @xmath8 statistical procedure using the x - ray hiflugcs data sample@xcite , and so we determined the statistical confidence contours ( from @xmath19 to @xmath20 errors ) for the pair of parameters , @xmath21 and @xmath22 .
we applied the homogeneous dark energy model of percival@xcite , and we marginalized over all possible values of the dark energy equation of state parameter @xmath1 . in figure [ fig1 ]
we show the contours in the @xmath22-@xmath21 plane obtained by using the ps and the pl approaches .
the left panel shows the ps results : the best - fit to the hiflugcs x - ray galaxy cluster data are @xmath23 , @xmath24 and @xmath0 .
the best fit values of @xmath22 and @xmath21 are , respectively , very low and very high when compared to the nowadays independent cmb results@xcite .
the solid vertical and horizontal lines show the minimum and maximum cmb limits ( from wmap 3 years data ) for the @xmath22 and @xmath21 parameters .
the contours using the gaussian distribution ( the left panel ) does not intercept the cmb values . in the other hand , by using the pl mass function with a certain @xmath2 free parameter ( the right panel of figure [ fig1 ] ) , we could overlap the wmap independent data , for almost the same dark energy @xmath1 parameter ( in this example , we use @xmath25 ) . in figure
[ fig1 ] we see that , for almost the same particular model of dark energy ( @xmath26 ) , we obtain better cosmological parameters using the pl mass function , while the standard ps mass function can not explain the independent wmap data even at @xmath27 confidence level . in fig .
[ fig2 ] , we plot two statistical contours in the @xmath22-@xmath21 plane , using the hiflugcs data with the pl mass function , under the dark energy model of percival@xcite .
again we marginalize over all possible values of @xmath1 . by fixing @xmath28 in the pl mass function ,
we obtain as best - fit a dark energy @xmath29 , @xmath30 and @xmath31 ( and the plot overlap the independent wmap data with this configuration ) .
when we fix @xmath32 the plot continues to overlap the wmap data , with almost the same best - fit for @xmath22 and @xmath21 , but it presents a huge modification in the best - fit dark energy @xmath1 parameter ( @xmath33 here , close to the cosmological constant case ) .
so , in the range @xmath34 , the pl mass function provides better fits to the hiflugcs x - ray galaxy data and the independent data from cmb .
also , the sensibility over the @xmath1 parameter , as we change the free @xmath2 parameter , indicates that the pl mass function could be a powerful tool to constrain dark energy models .
it must to be noted , however , that we are using the standard press & schechter approach , in this present work .
this approach smooths the initial density fluctuations , and uses the spherical model to calculate the epoch of critical overdensity for collapse of the density perturbations ( @xmath35 - critical linear density contrast at collapse time ) . such simple model eventually fail in detail , given structure formation complexities like , for example , the asymmetrical gravitational collapse ; meanwhile , recent mass function improvements from numerical simulations have now quantified these problems - e.g. the sheth & tormen ellipsoidal collapse ( 1999 - hereafter st)@xcite , and the universal function of jenkins et al .
( 2001 - hereafter jenkins)@xcite .
in this work , we use the percival critical overdensity for collapse ( @xmath35 ) , with growth factor derived from the special case of flat dark energy models with constant @xmath1 parameter ( equations 19 to 21 from reference @xcite ) .
this critical overdensity is only weakly dependent on cosmological parameters , and the cumulative mass function , using the numerical fit of st , presents very little difference for different cosmologies ( different @xmath1 dark energy parameters ) at low redshifts ( @xmath36 close to zero ) ; but as we go further back in time ( towards higher redshifts ) , the difference becomes appreciable ( figure 6 from @xcite ) .
we see that the evolution of the mass function is strongly dependent on @xmath1 , due to the evolution of the linear growth factor ( which affects @xmath35 directly ) .
it is worth to discuss , at this point , the results from the work of matthew et al .
( 2009 - hereafter matthew)@xcite .
matthew studied the role of the linear density contrast at collapse time , @xmath35 , in early dark energy ( ede ) models - where dark energy has a non - negligible importance since the beginning of the structure formation process . using a pure numerical approach to compute @xmath35 ,
matthew found that , at redshift @xmath37 , the ede mass functions are not greatly altered compared to the lcdm cosmological model , but the difference increases with redshift - the same conclusion as francis et al .
( 2008 , hereafter francis)@xcite .
matthew also demonstrates that the ede model presents a basic agreement between the st and the jenkins mass functions , even if we use a cosmology dependent @xmath35 , derived from the spherical collapse model .
however , when we compare the jenkins mass function ( which is blind to the growth history of the universe ) to the st mass function with a fixed @xmath35 ( @xmath38 ) , or deriving st @xmath35 from the spherical model ( where the history of the universe - the growth history - counts ) , we can see small differences between these mass functions in the high mass end ( @xmath39 ) ; despite these differences , the most up - to - date numerical simulations have not sufficient accuracy at this high mass range to discriminate between these two approaches .
so , we observe a general agreement between the conclusions of percival , matthew and francis , concerning dark energy models in the structure formation .
all dark energy mass functions , st or jenkins , have almost the same behavior as the lcdm model at @xmath37 , and at high redshifts we can perceive differences between the cosmological models , due to the evolution of the linear growth function . in the work of percival , the focus is the difference based on the dark energy @xmath1 parameter , and in the work of matthew and francis , the focus is the difference between the st ( with fixed or evolving @xmath35 ) and the jenkins mass function . in this present work
we use , as said before , the standard ps approach , with a gaussian and a non - gaussian ( pl ) initial distribution function .
we use in this work a low - redshift x - ray galaxy cluster survey , but even at small redshifts we show a strong mass function dependence on the dark energy @xmath1 parameter , using the pl approach , and the @xmath8 statistical procedure show significant contour differences between the dark energy mass function and the lcdm cosmological model .
this show clearly the power of the pl mass function @xmath2 free parameter to fit observational data ; and if this malleability is observed to galaxy clusters at @xmath40 , we should expect even more exciting results using high redshifts catalogs , to probe the characteristics of the dark energy , in a close future . also , we intend to study these dark energy models using st and jenkins mass functions as well , studying the @xmath35 influence in the process , to compare our results with the previous conclusions of percival , matthew and francis .
in this work we perform a statistical analysis , using the x - ray galaxy cluster data named hiflugcs@xcite , and the independent cosmic microwave background radiation ( cmb ) data from wmap@xcite , to study the influence of galaxy mass functions under a homogeneous dark energy background , from the work of percival@xcite . performing a @xmath8 analysis and marginalizing over the @xmath3 dark energy equation of state parameter
, we compare the standard gaussian press - schechter mass function@xcite to the pl mass function ( based on the non - extensive thermodynamics and kinetic theory ) , which presents a physically motivated free parameter , @xmath2@xcite .
we conclude that the ps mass function can not explain at the same time the x - ray and the cmb data , even at @xmath20 confidence level , which imposes a problem to the standard gaussian scenario .
also , the ps approach presents a best fit dark energy equation of state parameter of @xmath0 , which is clearly distant from independent estimates of @xmath41 ( a cosmological constant ) . observing the contours of figure [ fig2 ] ,
in the range @xmath34 , the pl mass function provides better fits to the hiflugcs and the cmb independent data , with almost the same best - fit for @xmath22 and @xmath21 , but with a huge modification in the best - fit dark energy @xmath1 parameter , as we change @xmath2 .
this strongly suggests that the pl mass function could be a powerful tool to constrain dark energy models .
future analysis , based on the next generation of cluster surveys in medium and high redshifts , will certainly provide better constraints to the dark energy model that accelerates de universe .
the author are grateful to ect - ufrn and fapesp , for the financial support provided , and to lima j.a.s . for helpful discussions .
a. g. riess _
et al . _ ,
j. * 659 * , 98 ( 2007 ) ; r. g. carlberg _
et al . _ ,
apj . * 462 * , 32 ( 1996 ) ; a. friaca , j. s. alcaniz and j. a. s. lima , mon . not .
r. astron .
soc . * 362 * , 1295 ( 2005 ) , arxiv:0504031 [ astro - ph ] ; s. w. allen , r. w. schmidt and a. c. fabian , mnras * 334 * , l11 ( 2002 ) ; j. a. s. lima , j. v. cunha and j. s. alcaniz , phys .
d * 68 * , 023510 ( 2003 ) , arxiv:0303388 [ astro - ph ] ; j. v. cunha , l. marassi and r. c. santos , ijmpd 16 , 403 ( 2007 ) ; j. v. cunha , l. marassi and j. a. s. lima , mnras 379 , l1-l5 ( 2007 ) , arxiv:0611934 [ astro - ph ] .
d. n. spergel et al . , astrop .
148 , 175 ( 2003 ) ; d. n. spergel et al . , astrophys . j. supl . 170 , 377 ( 2007 ) ; j. dunkley et al . ,
arxiv:08030586 [ astro - ph ] ; e. komatsu _ et al .
_ , arxiv:08030547 [ astro - ph ] . | several independent cosmological tests have shown evidences that the energy density of the universe is dominated by a dark energy component , which cause the present accelerated expansion .
the large scale structure formation can be used to probe dark energy models , and the mass function of dark matter haloes is one of the best statistical tools to perform this study .
we present here a statistical analysis of mass functions of galaxies under a homogeneous dark energy model , proposed in the work of percival ( 2005 ) , using an observational flux - limited x - ray cluster survey , and cmb data from wmap .
we compare , in our analysis , the standard press - schechter ( ps ) approach ( where a gaussian distribution is used to describe the primordial density fluctuation field of the mass function ) , and the pl ( power law ) mass function ( where we apply a nonextensive q - statistical distribution to the primordial density field ) .
we conclude that the ps mass function can not explain at the same time the x - ray and the cmb data ( even at 99% confidence level ) , and the ps best fit dark energy equation of state parameter is @xmath0 , which is distant from the cosmological constant case .
the pl mass function provides better fits to the hiflugcs x - ray galaxy data and the cmb data ; we also note that the @xmath1 parameter is very sensible to modifications in the pl free parameter , @xmath2 , suggesting that the pl mass function could be a powerful tool to constrain dark energy models . |
there have been sincere hopes that the inverse tully - fisher ( tf ) relation could overcome the distance dependent selection bias , the so - called malmquist bias of the 2nd kind ( teerikorpi @xcite ) , influencing the determination of the value of the hubble constant .
first , schechter ( @xcite ) pointed out that the inverse relation of the form @xmath3 where @xmath4 is absolute magnitude ( or linear diameter ) and @xmath5 is an observable parameter not affected by selection effects ( e.g. the maximum rotation velocity @xmath0 in the tf relation ) , can be derived even from magnitude limited samples in an unbiased manner .
secondly , teerikorpi ( 1984 , or @xcite ) showed explicitly that the inverse relation gives unbiased average distances for a sample of galaxies ( if @xmath5 is not affected by selection ) , a result later confirmed by tully ( @xcite ) via simulations ( see also hendry & simmons @xcite ) .
thirdly , ekholm & teerikorpi ( 1997 , or @xcite ) showed that with the inverse relation , one does not necessarily need a volume - limited calibrator sample , which on the contrary is quite critical for the use of the direct relation : @xmath6 the direct tf relation has been succesfully applied for derivation of @xmath2 by the method of normalized distances that was first used by bottinelli et al .
( @xcite ) .
theureau et al .
( @xcite ) have used an improved version of this method in an analysis of a sample of 5171 spiral galaxies that leads to the value of @xmath7 . in this method , which is rather similar to the spaenhauer diagram method of sandage ( @xcite )
, one identifies the so - called unbiased plateau in the @xmath8 vs. normalized distance diagram for a magnitude or angular diameter limited sample , from which @xmath2 is determined without the need to make further model dependent correction due to selection .
normally , the unbiased plateau contains only 10 - 20 percent of the total sample . that one can use the whole sample without bias correction
is a remarkable advantage of the inverse relation approach .
however , several problems balance the advantages of the inverse relation ( fouqu et al .
@xcite , teerikorpi @xcite , sandage et al .
@xcite , @xcite ) , and one has not yet been able to use it reliably for an independent determination of @xmath2 ( see theureau @xcite , ekholm et al .
@xcite ) . in the present paper ,
we study two important points which must be clarified before a successful use of the inverse tf relation : 1 . how would possible cutoffs in @xmath0 be seen in the data ? 2 .
how large an influence do upper and lower cutoffs in @xmath0 have on derived average distances ( and hence on @xmath2 ) ? 3 .
how does the nature of the calibrator sample influence the derived average distances if the inverse tf slopes for calibrators and field galaxies differ ?
as regards the first point , we emphasize that our approach follows the lines which we have adopted and found useful with the direct tf relation : to detect and overcome the bias without detailed modeling , by studying how the cutoffs would appear in kinematical applications ( c.f .
the increase of @xmath9 as a function of true ( or normalized ) distance , when the direct tf relation is used ) .
another , more technical approach to the direct tf relation is that of willick ( @xcite ) , based on corrections to individual galaxies within an iterative scheme . in principle , a good knowledge of the @xmath0 selection function could also permit willick s method to work for the inverse tf relation , in order to derive the slope as it would be without the selection .
our approach necessarily cuts away a part of the sample .
however , as shown by ekholm et al .
( @xcite ) for the klun sample , the remaining subsample suitable for the determination of @xmath2 with the inverse tf method is still much larger than the unbiased subsample for the direct tf relation .
the observational motivation for studying question 2 ) comes from theureau et al .
( @xcite ) , who discuss the detection rate of 21 cm line profiles at the nanay radio telescope , in connection with the angular size limited klun galaxy sample .
the detection rate is relatively high , 86 percent for galaxies with a previous redshift measurement and 61 percent for those with unknown @xmath10 .
however , more important than the average detection rate , is _ how detection depends on @xmath5 , i.e. the selection function @xmath11_. unfortunately this is difficult to derive from the raw data .
theureau ( @xcite ) notes that one expects loss of galaxies with either a narrow or broad 21 cm line width .
narrow profiles are difficult to detect among the noise , while broad profiles , being low , also tend to be missed .
hence , as a first approximation we suppose that the distribution of @xmath0 is affected by sharp lower and upper cutoffs , @xmath12 and @xmath13 , although the selection @xmath11 more probably depends on various galactic parameters in a complicated manner .
how such cutoffs influence , e.g. the value of @xmath2 as determined by the inverse relation method , has not been previously _
quantitatively _ discussed .
we study the above questions using normal distributions , which allow analytic expressions .
this forms a natural sequel to our previous discussions of the tf relations , both direct and inverse , where the assumption of gaussianity has been adopted ( teerikorpi 1984 , @xcite , @xcite , @xcite ) .
it is worth noting that though we occasionally refer to the klun sample ( see e.g. theureau et al .
@xcite ) , the theoretical ideas concerning the above mentioned points 13 have general validity . as an extensive sequel to the present work , ekholm et al .
( @xcite ) have investigated whether the large value of @xmath2 , derived with a straightforward application of the inverse tf relation for the klun sample , could be due to the errors arising from cutoff and/or calibrator sample biases .
they show that the calibrator sample bias is more important .
in this paper we also aim at a better general understanding of the inverse relation as a distance indicator , and present some arguments in a heuristic manner , since the method has been justly criticised as difficult to visualize , in comparison with the `` more natural '' direct relation approach .
this paper is structured as follows : in sect .
2 we show qualitatively how cutoffs in @xmath0 influence the distance determination . in sect .
3 the method of normalized distance @xmath14 is introduced for the inverse tf relation . in sect . 4
, the behaviour in the @xmath15 vs. @xmath14 diagram is calculated analytically . in sect . 5
, we derive an analytical expression for the average bias in @xmath15 when there is an upper or lower cutoff in @xmath0 , and give some examples in sect . 6 .
the fundamental problems of slope and calibration are discussed in sects . 7 and 8 .
section 9 contains concluding remarks , emphasizing the essential points and consequences of the present study . a note on the nomenclature : we use capital @xmath16 s for the diameter of a galaxy , @xmath17 is the usual apparent 25th - mag isophotal diameter ( @xmath18=0.1 $ ] arcmin )
, @xmath16 is the linear diameter ( @xmath19=$ ] kpc ) . in sect . 4 , to make the formulae more readable , we use @xmath20 for the @xmath21 s .
small @xmath22 s are used for inferred distances , @xmath14 is the normalized and @xmath23 the kinematical distance .
it is assumed that kinematical distances have been calculated from a realistic velocity field model , giving reliable relative distances ( in practice from a virgo - centric infall model as e.g. in theureau et al .
the true distance is denoted by @xmath24 .
let us assume that the selection may be described as cutoffs on the wings of the symmetric @xmath5 distribution . to illustrate ,
let us consider the simple case when there is only an upper cutoff @xmath13 , so that all @xmath25 are missed , while all @xmath26 are measured .
the situation is sketched in fig . 1 .
the calibrator sample ( at a fixed distance ) is on the right , while a distant cluster of galaxies is on the left .
the @xmath13-limit is indicated by a dotted line .
the correct regression lines , with the same slopes , are shown .
note that unaware of the @xmath5 selection , one would derive from the cluster galaxies ( or more generally from field galaxies ) a too shallow slope . for the moment we simply assume that the correct slope is known .
we return to the problem of slope in sect . 7 , and this issue will be discussed elsewhere ( ekholm et al.@xcite ) in connection with real data . in inverse tf method ( itf ) for deriving distances ,
one basically shifts one of the regression lines over the other , and the required shift of @xmath27 gives the ( logarithmic ) ratio of cluster to calibrator distance . in practice ,
one calculates for each cluster galaxy the itf `` distance '' @xmath28 using the calibrating relation @xmath29 , so that @xmath30 .
averaging over all cluster galaxies gives the unbiased distance .
this is equivalent to `` shifting the regression lines '' .
the itf `` distance '' has a systematic error depending on @xmath5 , though one does not know how large the error is ( c.f .
sect . 5 in @xcite ) . in the depicted situation ,
even if one knows the correct slope , the attempt to force - fit it on the cluster data will shift the line towards larger @xmath27 , because the upper cutoff has eliminated more galaxies from above the `` correct '' ( unbiased ) regression line than from below it .
hence , the inevitable result of an upper cutoff is that the itf distance to the cluster will be underestimated .
now , because of the limiting angular size , not all galaxies of the cluster are included .
so one should add to the diagram a vertical line , only galaxies to the right of which are large enough to enter the sample .
note that this cutoff does not bias the distance indicator if there is no @xmath5-cutoff .
if one moves the cluster closer ( to the right ) , the angular size limit cuts away less of the cluster galaxies .
then the influence of the @xmath13-cutoff on the derived distances becomes weaker . in terms of the calculated hubble constant ( actually up to an unknown factor , because @xmath31 is still unknown ) , one should see in a @xmath15 vs. @xmath23 diagram first at small kinematical distances ( proportional to the true ones ) a constant `` plateau '' when the angular size limit keeps out of the cluster .
then there is an increase due to the larger influence of the @xmath13-cutoff .
the plateau in the @xmath15 vs. @xmath23 diagram is not an unbiased plateau ; it still gives too large a value of @xmath32 , because of a residual effect of the @xmath13-limit .
the bias can be reduced by applying an upper cutoff to the cluster sample in @xmath27 , as illustrated by the dashed vertical line in fig . 1 .
the galaxies between the upper and lower @xmath27 cutoffs now provide the best itf distance estimate . the position of the critical righthand side vertical line in fig . 1 , which cuts away the `` bad '' @xmath27 interval , depends on the distance of the cluster ( compare the cluster and the calibrators )
. however , its position is practically constant on the _ linear size scale _ , independently of the cluster distance .
this is clearly seen if one shifts in fig . 1 the cluster galaxies to the right , i.e. decreases their distance until it coincides with the distance of the calibrator cluster .
because the tf relations are essentially the same , the critical vertical lines will almost coincide .
hence , in the general case of a field galaxy sample , one may shift all galaxies independent of their distances , to one diagram of @xmath0 vs. @xmath33 , where the abscissa is up to an unknown constant ( as far as @xmath2 is unknown ) the same as linear diameter , @xmath21 .
if one now accepts galaxies left of the ( common ) critical line only , an unbiased average distance is derived .
one might term @xmath33 ( or @xmath34 ) the normalized distance , analogously to the case of the direct tf normalized distance introduced in @xcite . then in the @xmath15 vs. @xmath35 diagram , using the correct slope @xmath36 , there is at small normalized distances an unbiased plateau , after which @xmath8 starts to grow .
the resulting unbiased plateau should be horizontal , if the slope @xmath36 is correct .
@xcite pointed out a simple device of deciding which is the correct inverse slope : in the @xmath15 vs. @xmath23 diagram the run of points should be horizontal .
this was , however , made under the premise of no @xmath5-selection .
now we understand that instead of @xmath23 one should use @xmath37 on the abscissa , in order to identify the cutoff @xmath14 .
only galaxies below this cutoff are expected to show a horizontal run on the @xmath15 vs. @xmath23 diagram , if @xmath36 is correct .
figure 2 shows how in the more general case , the lower cutoff causes too small @xmath8 below the unbiased plateau .
ekholm et al .
@xcite will discuss the evidence for upper and lower cutoffs in @xmath0 in the klun sample and how the described method can in practice be implemented .
though in this manner one may detect sharp cutoffs in @xmath0 and derive an unbiased value of @xmath2 without detailed modelling , it is important to know how large an influence the cutoffs would have on @xmath2 , if their presence were ignored .
we discuss this in sects .
assuming gaussian distributions and sharp cutoffs at @xmath12 and @xmath13 , it is easy to calculate the expected run of @xmath8 vs. normalized distance @xmath14 . for simplicity , we write the inverse relation as @xmath38we use @xmath39 as shorthand notation for the logarithm of the linear diameter . at any fixed @xmath20
the dispersion of @xmath5 is @xmath40 .
now the task is to calculate the bias in @xmath15 at each @xmath20 .
this is done starting from the same initial formula ( 28 ) of @xcite which was used to show the absence of bias . however , in this case one must add the selection function @xmath11 : @xmath41where @xmath42}$ ] and @xmath43 .
this formula gives for the sample galaxies with true ( log ) diameter @xmath20 , the ( log ) error in the diameter deduced by the inverse relation ( using @xmath44 ) .
it is directly reflected to the derived value of @xmath8 at constant @xmath20 ( i.e. , at fixed normalized distance @xmath45 ) . with the present choice of the selection function ( sharp upper and lower cutoffs ) , and after a change of variable , @xmath46eq .
( 4 ) becomes @xmath47here the term @xmath48 can be evaluated in terms of the error function @xmath49 : @xmath50where @xmath51 and @xmath52 correspond to the cutoffs @xmath12 and @xmath13 via eq .
inspection of the formula confirms the qualitative conclusions of sect.2 .
putting @xmath53 , so that only the upper cutoff remains ( and recalling that @xmath54 ) , the bias is seen to be positive : the diameters are underestimated , leading to underestimated distances , hence to overestimated @xmath15 .
note that this calculation does not tell us the average bias in the sample .
in order to have an idea of how large an influence the @xmath5-cutoffs have on the value of @xmath8 calculated from the whole sample using a ( correct ) inverse relation , we derive the average bias for a special case of radial space density behaviour : @xmath55 where @xmath24 is the radial distance from our galaxy and @xmath56 corresponds to homogeneity . it is easy to understand that the average bias depends on how the observed galaxies are distributed along the @xmath14-axis , which must depend on how galaxies are distributed in space . rather than trying to start with averaging the derived @xmath57 over @xmath14 ( eq . 6 )
, we use another route , based on the fact that the bivariate distribution @xmath58 is known for an angular diameter limited sample , when space density behaves as eq .
we assume for simplicity that the cosmic distribution of @xmath5 ( as well as of @xmath20 ) is normal , which is a rough approximation for fixed hubble types , see e.g. fig . 1 in theureau et al .
( @xcite ) . as illustrated by fig . 1 in @xcite for the case of magnitude tf relation , in magnitude and volume limited samples the direct regression lines are parallel and separated by @xmath59 ( when @xmath56 )
. the separation happens so that the volume - limited line ( carrying with it the bivariate distribution ) glides along the unchanged inverse line which goes through the average @xmath4 of both the volume ( @xmath60 ) and magnitude limited ( @xmath61 ) bivariate distributions .
figure 3 gives a similar diagram for the diameter tf relation , showing the three kinds of regression lines for a synthetic sample of field galaxies . in the present calculation
we need to know how the diameter direct relation is constructed from the parameters defining the itf ( c.f .
@xcite , teerikorpi @xcite , where this was given for the magnitude tf relation ) : @xmath62where @xmath63 the malmquist shifted ( angular size limited ) direct relation is @xmath64and the inverse relation is @xmath65 one can express the average bias at a fixed @xmath5 using the inverse and ( malmquist - shifted ) direct relations : @xmath66where @xmath67 so that @xmath68this simple expression is then easy to integrate over all @xmath5 in the sample , when one notes that the distribution of @xmath5 is @xmath69where @xmath70 gives the malmquist - shifted average value of @xmath5 ( shifted by @xmath36 times the shift of @xmath71 ) : @xmath72averaging @xmath73 over all @xmath5 from @xmath12 to @xmath13 gives : @xmath74 where @xmath51 and @xmath52 are @xmath75note that the information on the space density ( via @xmath76 and eq .
16 ) is contained in @xmath51 and @xmath52 , which show how far away from the malmquist - shifted average @xmath70 the cutoffs are , in terms of the dispersion @xmath77 of the cosmic ( gaussian ) distribution function of @xmath5 .
the factor before @xmath48 can also be written in terms of @xmath40 and @xmath78 .
then @xmath79such an analytic calculation is possible only in the given special cases of space density .
generally , the distribution of points in the @xmath5 vs. @xmath35 diagram will differ from the gaussian bivariate case : at each @xmath5 , the shift of @xmath80 will be due to different malmquist biases of the 1st kind .
however , even in this case the inverse relation is not affected and the calculation and prediction of sect .
3 is valid .
we are interested in seeing how large an effect on @xmath15 cutoffs of different severities have .
table 1 shows the values of @xmath81 ( @xmath82 ) and the bias in @xmath83 for different values of upper ( lower ) cutoff @xmath84 , and the corresponding completeness percentage of the sample .
.bias in @xmath8 due to @xmath5 cutoffs .
the upper sign in the @xmath85 symbols refers to @xmath13 while the lower sign refers to @xmath12 . [
cols=">,>,^,^ " , ] in principle , the problem of the @xmath5-distribution of the calibrators does not affect the determination of the zero - point @xmath86 when one forces the correct slope @xmath87 on the calibrators .
this makes one ask what happens if @xmath87 and @xmath86 are then used for the field sample .
first of all one expects a distance - dependent bias , according to eq.15 from teerikorpi ( @xcite ) , here adopted to diameters : @xmath88here @xmath89 is the average ( log ) diameter for the galaxies at true distance @xmath24 . now one may ask what is the average bias for the sample ?
the average bias clearly depends on the average of @xmath90 .
this is actually the malmquist bias of the 1st kind for gaussian distribution of diameters with dispersion @xmath91 , and in the case of a space density distribution @xmath92 one gets : @xmath93when @xmath56 the distribution is homogeneous . in order to use this as a correction
, one must know both @xmath87 and @xmath36 . clearly , if the field and calibrator slopes are identical , then @xmath8 is unbiased , and this is so irrespective of the nature of the calibrator sample .
note also that this formula contains the width @xmath91 of the total diameter function , not the smaller dispersion @xmath94 .
hence the effect can be formidable even when the slopes @xmath87 and @xmath36 do not differ very much . as a concrete example ,
consider the klun field galaxies following the inverse slope @xmath95 which is in agreement with the simple disk model of theureau et al .
( @xcite ) .
this slope , when forced through the calibrators , gives @xmath96 .
above it was noted that the calibrator sample can not represent the cosmic distribution of @xmath0 , and this is the situation when a too high @xmath2 is derived if the actual slope is steeper than the slope from the field galaxies .
it is natural to ask , how large a slope @xmath97 would in this manner explain the difference between @xmath98 and @xmath99 .
taking @xmath100 , we get predicted @xmath101 for different slopes @xmath97 as given in table 2 .
the last column corresponds to a decrease in the average space density , corresponding to density @xmath102 as derived by teerikorpi et al .
( @xcite ) .
it may be concluded that in order to explain the difference between @xmath103 and 55 solely in this manner , it is required that actually @xmath104 0.65 0.75 , instead of @xmath105 .
ekholm et al .
( @xcite ) show that such a steep calibrator slope really is consistent with the data .
in the present study we have searched for answers to a few basic questions concerning the inverse tf relation as a distance indicator . _ first _ , how to see the signature of observational incompleteness in the tf parameter @xmath0 , analogously to the well - known increase in @xmath15 at increasing ( normalized ) distance , in the case of the direct tf relation ?
_ secondly _ , in which sense and how _ significant _ is the influence of the @xmath0 incompleteness alone on the derived value of @xmath15 ( supposing one knows the correct inverse tf slope ) ?
_ thirdly _ , how to work with the situation when the inverse slope obeyed by the calibration sample differs from the relevant slope for the field sample ( or for a cluster ) , and how important is in this regard the nature of the calibration sample ?
we have generalized the `` fine - tuning '' scheme discussed by @xcite , and have pointed out that there is a useful @xmath15 vs. @xmath14 representation for the inverse relation , revealing both the relevant inverse slope and cutoffs in @xmath0 . here @xmath14 is the normalized distance applicable to the inverse tf relation , actually the ( log ) linear diameter within an unknown constant .
we have derived simple analytical formulae which give the influence of the @xmath0 cutoffs on derived average distance ( hence , on @xmath15 ) .
the treatment is based on gaussian distributions , and , in the case of the average bias , on the assumption that the space density is proportional to @xmath106 .
the probable case that there is an upper observational cutoff in @xmath0 , would cause an overestimated value of @xmath2 .
however , this effect is not at all dramatic even if the upper cutoff excludes half of the galaxies , @xmath2 would be increased typically by only 14 percent .
it may be concluded that a cutoff in @xmath0 can not alone explain derived large values of @xmath2 .
the problems of the relevant inverse tf slope and the nature of the calibrator sample lead to serious consequences on the use of the inverse tf relation as a distance indicator .
it has been previously realized that the inverse slope @xmath36 applicable to the field sample is not necessarily the same as the slope @xmath97 obeyed by the calibrators , which may result in a biased value of @xmath2 ( teerikorpi @xcite ) . in the present paper , such situations were systematically studied .
if the two slopes are the same ( in practice , when accuracies of calibrator distances are high and the photometric measurements for the field sample are not less accurate than those for the calibrators ) , then there is no bias , and it is also true that the nature of the calibrator sample ( volume - limited , magnitude - limited , etc . ) is not important ( @xcite ) .
if the sample inverse slope @xmath36 is shallower than the calibrator slope ( as normally expected ) , then matters come to depend critically on the nature of the calibrator sample .
first , there is the possibility that the calibrator sample is a true , volume - limited representation of the gaussian cosmic distribution function of @xmath0 from which the field sample has been taken ( the latter subject only to malmquist bias ) . in this ideal case
, it is permitted to use the field sample slope , with the zero - point obtained by a force - fit through the calibrators .
this was the solution suggested by teerikorpi ( @xcite ) , and in this case it is not necessary to know the value of the calibrator slope .
however , generally the calibrator sample s @xmath107 is shifted from the cosmic @xmath108 .
if one knows the shift @xmath109 ( or @xmath110 ) _ and _ the slopes @xmath97 and @xmath36 , then one may calculate the resulting error in @xmath15 , when the field sample s @xmath36 is used .
the systematic error is given by eq .
( 25 ) .
another , more convenient route , which bypasses the explicit need for @xmath109 , is to use the calibrators slope @xmath87 ( and corresponding zero - point ) and to calculate analytically the resulting systematic error in terms of the malmquist bias of the 1st kind .
this assumes that the average space number density around us is proportional to @xmath106 .
the systematic error is given by eq .
( 27 ) .
we repeat that in the general case both @xmath87 and @xmath36 ( with corresponding zero - points from force - fits through calibrators ) result in erroneous average distance estimates .
fortunately , we have now some quantitative control on this error .
bottinelli , l. , gouguenheim , l. , paturel , g. , teerikorpi , p. , 1986 , a&a 156 , 157 ekholm , t. , teerikorpi , p. , 1997 , a&a 325 , 33 ( et97 ) ekholm , t. , teerikorpi , p. , theureau , g. , hanski , m. , paturel , g. , 1998 a&a ( submitted ) fouqu , p. , bottinelli , l. , gouguenheim , l. , paturel , g. , 1990 , apj 349 , 1 hendry , m.a . , simmons , j.f.l . , 1994 , vistas astron . 39 , 297 sandage , a. , 1994 , apj 430 , 1 sandage , a. , tammann , g.a . , federspiel , m. , 1995 , apj 452 , 1 schecter , p. , 1980
, aj 85 , 801 teerikorpi , p. , 1984
, a&a 141 , 407 ( t84 ) teerikorpi , p. , 1987
, a&a 173 , 39 teerikorpi , p. , 1990
, a&a 234 , 1 teerikorpi , p. , 1993
, a&a 280 , 443 teerikorpi , p. , 1997
a&a 35 , 101 teerikorpi , p. , hanski , m. , theureau , g. et al . , 1998 , a&a 334 , 395 theureau , g. , 1997 , thse de doctorat , ( paris observatory ) theureau , g. , hanski , m. , teerikorpi , p.
et al . , 1997a , a&a 319 , 435 theureau , g. , hanski , m. , ekholm , t. et al . , 1997b , a&a 322 , 730 theureau , g. , bottinelli , l. , coudreau , n. et al . , 1998 , a&as ( in press ) tully , r.b . , 1988 ,
nature 334 , 209 willick , j. , 1994 , apjs 92 , 1 | we study the influence of the assumption behind the use of the inverse tully - fisher relation : that there should be no observational cutoffs in the tf parameter @xmath0 .
it is noted how lower and upper cutoffs would be seen in a @xmath1 vs.``normalized distance '' diagram .
analytical expressions , under the simplifying assumption of a normal distribution and the use of the correct tf slope , are derived for the resulting biases , especially the average bias which @xmath0 cutoffs produce in the derived value of @xmath2 .
this bias is shown to be relatively weak , and as such can not explain the large differences in the reported values of @xmath2 derived from direct and inverse tf relations .
some problems of slope and calibration are shown to be more serious .
in particular , one consequence of fitting through the calibrators either the slope relevant for field galaxies or the steeper slope followed by calibrators is that the derived value of the hubble constant comes to depend on the nature of the calibrator sample .
if the calibrator sample is not representative of the cosmic distribution of @xmath0 , large errors in the derived value of @xmath2 are possible .
analytical expressions are given for this error that we term the calibrator sample bias . |
the nonlinear schrdinger ( nls ) equation plays an important role in many fields of nonlinear physics @xcite .
the cubic nls equation is shown to be completely integrable for both self - focusing and defocusing kerr nonlinearities @xcite , which can be used to describe the propagation of optical pulses in kerr - type optical media @xcite , or to describe the dynamics of matter waves in bose - einstein condensates , known as the gross - pitaevskii equation @xcite .
the nls equations with real external potentials and gain - and - loss distributions have been studied in many works @xcite because the refractive index of optical materials is usually complex , i.e. , @xmath2 with @xmath3 and @xmath4 being the real and imaginary parts , respectively . in optics ,
the propagation of a signal is stable unless the propagation constant of the light is in real spectrum range .
this requirement can be efficiently achieved if the gain - and - loss distributions in the medium are exactly balanced to ensure the relation @xmath5 ( or @xmath6 and @xmath7 ) , which is known as the parity - time ( @xmath0 ) symmetric systems @xcite .
@xmath0-symmetry may exhibit entirely real spectrum of the respective optical potential in some parameter regions , referred to as the unbroken @xmath0-symmetry @xcite . beyond these regions , in the broken @xmath0-symmetry , the spectrum becomes complex and propagating waves may be either grow or decay .
recently , various @xmath0-symmetric potentials have been introduced to the nls equations , which have been shown to possess stable and unstable solitons of different types @xcite .
examples include the nls equations with @xmath0-symmetric scarff ii potential @xcite , periodic potential @xcite , harmonic potential @xcite , rosen - morse potential @xcite , gaussian potential @xcite , sextic anharmonic double - well potential @xcite , time - dependent harmonic - gaussian potential @xcite , self - induced potential @xcite , and etc .
( see , e.g. , @xcite ) .
one the other hand , due to the significant progress achieved in recent years on developing optical materials with adjustable refractive index , @xmath8-symmetric optical systems made of solid - state waveguides and fiber networks @xcite , multi - level atomic systems @xcite , and microcavities @xcite have been suggested or realized experimentally .
these practical systems provide a solid ground for the study of solitons in the nls model with @xmath0-symmetric potentials .
the @xmath0-symmetric potentials play a key role in the wave propagation in the nls equations in fibre and waveguide optics . as an example , the wave propagation with a @xmath0-symmetric scarff ii potential in both linear and nonlinear models has been studied in recent years @xcite . the @xmath0-symmetric scarff ii potential can support stable bright solitons in the nls models within particular parameter regions @xcite .
more importantly , the real and imaginary parts of @xmath0-symmetric @xmath1-wavenumber scarff ii potential @xcite both approach to zero as @xmath9 [ cf .
( [ poten ] ) , for @xmath10 , we have @xmath11 and @xmath12 as @xmath13 , that is to say , they have limited effect on the nonlinear waves for the nls equation only in a boundary region ( e.g. , @xmath14 for @xmath15 ) [ cf .
( [ ode ] ) ] and can more easily support the existence of bright solitons in the nls equation , compared with the @xmath0-symmetric harmonic and optical lattice potentials , which always have the effect on the nonlinear waves in the whole region ( see , e.g , @xcite ) .
however , a complete analysis of the soliton stability in a @xmath0-symmetric @xmath1-wavenumber scarff ii potential is still lacking in a full parameter space ( e.g. , the space consists of amplitudes of real and imaginary parts of @xmath0-symmetric @xmath1-wavenumber scarff ii potential and the wavenumber @xmath1 , see eq .
( [ poten ] ) ) .
in addition , the study of optical solitons could be very useful in the information science .
for example , one can use optical solitons in coding for secured optical communication , etc . in this work
we propose a unified study of the optical bright solitons governed by self - focusing and defocusing nonlinear schrdinger ( nls ) equations with generalized @xmath0-symmetric scarff ii potentials .
particularly , we consider a @xmath0-symmetric @xmath1-wavenumber scarff ii potential and a multi - well scarff ii potential , respectively . for the @xmath1-wavenumber scarff ii potential , we show that by using two @xmath1-wavenumber rays and a @xmath1-wavenumber parabola one can divide the parameter space @xmath16 ( @xmath17 and @xmath18 are , respectively , the real and imaginary parts of the potential ) in different regions , corresponding to unbroken and broken @xmath0-symmetry and the bright solitons with self - focusing and defocusing kerr nonlinearities . for the multi - well scarff ii potential we obtain bright solitons by using a periodically space - modulated kerr nonlinearity .
then , we analyze the linear stability of bright solitons for both potentials in details by using numerical simulations .
surprisingly , stable and unstable bright solitons are found in both regions of unbroken and broken @xmath0-symmetry due to the existence of the nonlinearity .
furthermore , we explore the bright solitons in three - dimensional ( 3d ) self - focusing and defocusing nls equations with a generalized @xmath0-symmetric scarff ii potential . turning to the possible experimental implementation of the models we will study
( [ nls ] ) with @xmath0-symmetric @xmath1-wavenumber and multi - well scarff ii potentials ( [ poten ] ) and ( [ poten2 ] ) ] , we can exploit a @xmath0-symmetric refractive index profile ( e.g. , the considered @xmath0-symmetric @xmath1-wavenumber scarff ii potential ) imprinted in a cold gas of two atomic isotopes of @xmath19-type configuration ( say of @xmath20rb and @xmath21rb isotopes ) loaded in an atomic cell , as suggested in refs . @xcite . by the interference of two raman resonances ,
the required spatial distribution of the refractive index ( i.e. , the @xmath1-wavenumber and multi - well scarff ii potentials ) can be achieved by a proper combination of a control laser field and a far - off - resonance stark laser field . since the refractive index of the atomic vapor
is determined by two external laser fields , whose intensities can be lowered to several microwatts achievable for the lasers nowadays , the system has the advantages of real - time all - optical tunable capability and controllable @xmath0-symmetry accuracy .
in addition , the studied model supports a large kerr nonlinearity ( which is at least @xmath22-order larger than those measured for usual nonlinear optical materials ) due to the raman resonant character , which favors to the formation of optical bright solitons .
the rest of this paper is arranged as follows .
ii gives the bright spatial solitons in one - dimensional @xmath0-symmetric potentials .
both @xmath0-symmetric @xmath1-wavenumber and multi - well scarff ii potentials are considered .
the linear stability of optical bright solitons is analyzed .
stable solitons are found for both unbroken and broken @xmath0-symmetric potentials .
iii gives the bright solitons in 3d self - focusing and defocusing nls equations with a generalized @xmath0-symmetric scarff ii potential .
finally , the last section contains a summary of main results obtained in this work .
now we uniformly investigate the exact localized solution modes and their stabilities of self - focusing and defocusing nls equations with a @xmath0-symmetric potential . in the dimensionless form
, the physical model can be written as @xcite @xmath23\psi+g|\psi|^2\psi=0 , \end{aligned}\ ] ] where @xmath24 , @xmath25 is a complex field of @xmath26 ( @xmath27 and @xmath28 are , respectively , dimensionless space and time ) , @xmath29 is a real external potential , @xmath30 is a real gain - and - loss distribution , @xmath31 characterizes the self - focusing ( @xmath32 or defocusing @xmath33 kerr nonlinearity , respectively .
( [ nls ] ) is associated with a variational principle @xmath34 with the lagrangian density ( ) = i(^*_t-^*_t ) + 2|_x|^2 + + 2[v(x)+iw(x)]||^2+g||^4 .
the @xmath0-symmetric potential @xmath35 leads to the sufficient ( but not necessary ) condition @xmath36 and @xmath37 .
the quasipower and power of eq .
( [ nls ] ) are given by @xmath38 and @xmath39 , respectively .
one can readily get that @xmath40dx$ ] and @xmath41 . notice that if we set @xmath42 in eq .
( [ nls ] ) , where @xmath43 denotes the propagation distance @xcite , then the following results about eq .
( [ nls ] ) still hold as @xmath42 .
we focus on the stationary solutions of eq .
( [ nls ] ) in the form @xmath44 , where @xmath45 is the real propagation constant and the complex function @xmath46 satisfies the stationary nls equation with varying parameter modulation ( x)=+(x)+g|(x)|^2(x),[ode ] which can be solved by using numerical methods . in the following ,
we consider two different types of @xmath0-symmetric potential , i.e. , the @xmath0-symmetric @xmath1-wavenumber scarff ii potential and the periodically space - modulated scarff ii potential , and study the exact localized solutions and their linear stability in both regions of unbroken and broken @xmath0-symmetry . to study the soliton solutions of eq .
( [ ode ] ) , we first consider the @xmath0-symmetric @xmath1-wavenumber scarff ii potential @xcite , which is given as [ poten ] v(x)=v_0sech^2(kx ) , w(x)=w_0sech(kx)(kx ) , where @xmath47 denotes the wavenumber , and @xmath48 and @xmath18 are real parameters and modulate amplitudes of the external potential and gain - and - loss distribution , respectively . the wavenumber @xmath1 and
amplitude @xmath17 can modulated the well width and depth of the potential @xmath29 , respectively .
@xmath18 can modulate the impact of the gain - and - loss distribution . for the case
@xmath49 , @xmath0-symmetric @xmath1-wavenumber scarff ii potential ( [ poten ] ) becomes the usual @xmath0-symmetric scarff ii potential @xcite .
figure [ fig-1 g ] displays the @xmath0-symmetric @xmath1-wavenumber scarff ii potential ( [ poten ] ) for different amplitudes @xmath50 and wavenumber @xmath1 .
it follows from eq .
( [ poten ] ) that @xmath51 as @xmath9 and/or @xmath52 , and @xmath53 as @xmath54 and/or @xmath55 .
if @xmath56 , then @xmath29 and @xmath30 satisfy the family of elliptic curves : ( -12)^2+=14 with the centers being @xmath57 in the @xmath58-space .
it is easy to see that the family of elliptic curves is dependent on the wavenumber @xmath1 . in general , for the @xmath0-symmetric @xmath1-wavenumber scarff ii potential @xmath59 with @xmath29 and @xmath60 given by eq .
( [ poten ] ) for any real parameters @xmath61 there are the following three cases : * when @xmath62 , the potential @xmath63 is a real , hermitian , and @xmath0-symmetric potential ; * when @xmath64 , the potential @xmath65 is a pure imaginary , non - hermitian , and @xmath0-symmetric potential ; * when @xmath56 , the potential @xmath59 is a complex , non - hermitian , and @xmath0-symmetric potential . in the following , we mainly consider the potential @xmath59 in case iii ) with @xmath48 .
ccccc + [ -2.0ex ] region & linear problem ( @xmath66 ) : & linear problem ( @xmath66 ) : & nonlinear problem ( @xmath67 ) : & nonlinear problem ( @xmath68 ) : + [ 0.5ex ] & unbroken @xmath0-symmetry & broken @xmath0-symmetry & soliton ( [ solu ] ) & soliton ( [ solu ] ) + [ 1.0ex ] + [ -2.0ex ] @xmath69 & yes & no & yes & no + [ 2.0ex ] + [ -1.6ex ] @xmath70 & yes & no & no & no + [ 1.0ex ] + [ -2.0ex ] ii & no & yes & yes & no + [ 1.0ex ] + [ -2.0ex ] iii & yes & no & no & yes + [ 1.0ex ] [ table ] the linear eigenvalue problem l_pt(x)=(x ) , l_pt=-_x^2-[v(x)+iw(x ) ] with @xmath29 and @xmath30 given by eq .
( [ poten ] ) exhibits an entirely real spectrum provided that two amplitudes @xmath71 and the wavenumber @xmath1 satisfy [ ptc ] |w_0|v_0 + , where @xmath72 and @xmath73 are eigenvalue and engenfunction , respectively .
the inequality ( [ ptc ] ) is called the @xmath0 symmetry - unbreaking condition . if @xmath74 , the spectrum becomes complex . for @xmath49 ,
the condition ( [ ptc ] ) reduces to the well - known one , @xmath75 @xcite .
therefore , two families of @xmath1-wavenumber open rays @xmath76 [ l1 ] w_0=v_0+k^2/4,v_0>0 and @xmath77 [ l2 ] w_0=-(v_0+k^2/4 ) , v_0>0 divide the parameter space @xmath16 into two regions , i.e. , the unbroken @xmath0-symmetric region @xmath78 and the broken @xmath0-symmetric region @xmath79 ( see fig . [ fig-1 ] ) . for the given @xmath0-symmetric @xmath1-wavenumber scarff ii potential
( [ poten ] ) with @xmath48 and @xmath80 , eq . ( [ ode ] ) admits the unified bright soliton for both self - focusing ( @xmath32 and defocusing @xmath33 nonlinearities [ solu ] ( x)=sech(kx)e^i(x ) , where @xmath81 , @xmath82>0 $ ] , and the phase is ( x)=. in particular , when @xmath83 or @xmath84 , the solution ( [ solu ] ) reduces to the known ones as given in refs . @xcite .
now we analyze the conditions for the existence of bright soliton ( [ solu ] ) for parameters @xmath85 , and @xmath31 . for the positive ( self - focusing ) nonlinearity @xmath67 , the condition for the existence of bright soliton ( [ solu ] ) is given by [ con1 ] w_0 ^ 2>9k^2(v_0 - 2k^2 ) .
the real and imaginary parts of the soliton are displayed in fig .
[ fig-2](a ) . for the negative ( defocusing ) nonlinearity @xmath68
, the condition for the existence of bright soliton ( [ solu ] ) is given by [ con2 ] |w_0|<3k , v_0>2k^2 .
the real and imaginary parts of the soliton are illustrated in fig .
[ fig-2](b ) . taking into account the conditions ( [ con1 ] ) and ( [ con2 ] ) for the existence of bright soliton ( [ solu ] ) for @xmath67 and @xmath68
, the parameter space @xmath16 can be divided into another two fundamental regions by the family of parabolas @xmath86 with vertex being @xmath87 .
these two conditions ( [ con1 ] ) and ( [ con2 ] ) for bright solitons with @xmath67 and @xmath68 , respectively , are shown in fig . [ fig-1 ] .
notice that the bright soliton ( [ solu ] ) with @xmath67 ( attractive case for eq .
( [ nls ] ) ) can also exist for the region @xmath88 even if we do not study the @xmath0-unbroken / broken of the linear eigenvalue problem in this region .
two branches of @xmath1-wavenumber open rays @xmath89 given by eqs .
( [ l1 ] ) snd ( [ l2 ] ) are tangent to the parabola @xmath90 w_0 ^ 2=9k^2(v_0 - 2k^2 ) at the points @xmath91 , respectively . in another word
, @xmath92 always holds except for two tangent points @xmath93 , at which they are equal ( see fig .
[ fig-1 ] and note @xcite for the details ) .
therefore , the parameter space @xmath94 can be divided into eight regions , i.e. , l \{(v_0 , w_0)| v_0>0 , w_0 } + = i_1+i_2+i_3+ii+iii+l_1+l_2+l_3 , where ` @xmath95 ' denotes the union of two sets , these regions @xmath96 denote open sets without the corresponding boundaries ( see fig .
[ fig-1 ] ) . to make it become clearer , we summarize these different regions in table [ table ] .
one can see that the region for bright solitons with self - focusing nonlinearity contains both regions for unbroken and broken @xmath0-symmetry , whereas the region for bright solitons with defocusing nonlinearity contains only the region for unbroken @xmath0-symmetry . for any complex solution @xmath97 with @xmath98 bering a real phase of @xmath46
, we find that its corresponding transverse power - flow or poynting vector and the gradient of phase obey the relation @xmath99 .
for the bright soliton ( [ solu ] ) , its corresponding transverse power - flow or ` poynting vector ' is given by [ s ] s(x)=(-v_0 + 2k^2)sech^3(kx ) , which implies that @xmath100 for any @xmath27 and @xmath101 ( see fig . [ fig-2](c ) ) since @xmath47 and @xmath102 are required for the conditions of the existence of bright solitons .
therefore , the power always flows in one direction , i.e. , from the gain toward the loss .
in addition , we have the conserved power related to solution ( [ solu ] ) is @xmath103 .
next we study the linear stability of the bright solitons ( [ solu ] ) in the above - mentioned different regions . to this end
, we considered a perturbed bright soliton solution @xcite @xmath104\right\}e^{i\mu t},\ ] ] where @xmath105 is a stationary solution of eq .
( [ nls ] ) , @xmath106 , and @xmath107 and @xmath108 are the eigenfunctions of the linearized eigenvalue problem .
substituting eq .
( [ pert ] ) in eq .
( [ nls ] ) and linearizing with respect to @xmath109 , we obtain the following linear eigenvalue problem @xmath110 where @xmath111 .
the stability of the perturbed soliton @xmath112 is related to the imaginary parts @xmath113 of all eigenvalues @xmath114 .
if @xmath115 , then the solution @xmath112 will grow exponentially with @xmath28 ( i.e. , it is unstable ) , otherwise the solution @xmath112 is stable . in the following ,
we study the stability of exact soliton ( [ solu ] ) under an initial random noise perturbation up to @xmath116 of its amplitude for different parameters since the regions i@xmath117 and i@xmath118 are symmetric about the @xmath17-axes , we only need to consider one of them ( say i@xmath117 ) .
thus we leave four regions for studying the linear stability of the bright solitons ( [ solu ] ) .
the regions i@xmath117 , i@xmath119 and ii correspond to bright solitons for the self - focusing nonlinearity @xmath120 , whereas the region iii corresponds to bright solitons for the defocusing nonlinearity @xmath121 . to be specific
, we study the stability of bright solitons in the regions for unbroken and broken @xmath0-symmetry with @xmath49 ( the usual @xmath0-symmetric scarff ii potential ) and @xmath122 .
notice that we only consider @xmath123 as @xmath30 is an odd function .
if @xmath49 and @xmath120 : ( i ) for @xmath124 in the region ii ( broken @xmath0-symmetry ) , we numerically obtain the spectrum of eigenvalues @xmath114 ( see fig . [ fig3](a ) ) and a stable propagation of the soliton intensity @xmath125 shown in fig .
[ fig3](b ) ; ( ii ) for @xmath126 in the region @xmath127 ( unbroken @xmath0-symmetry ) , we numerically obtain the spectrum of eigenvalues @xmath114 ( see fig . [ fig3](c ) ) and a stable propagation of the soliton intensity @xmath125 shown in fig . [ fig3](d ) ; ( iii ) for @xmath128 in the region @xmath129 ( unbroken @xmath0-symmetry ) , we numerically obtain the spectrum of eigenvalues @xmath114 ( see fig . [ fig3](e ) ) and an unstable propagation of the soliton intensity @xmath125 shown in fig . [ fig3](f ) ; ( iv ) for @xmath130 in the region @xmath88 , we numerically obtain the spectrum of eigenvalues @xmath114 ( see fig . [ fig3](g ) ) and an unstable propagation of the soliton intensity @xmath125 shown in fig . [ fig3](h ) .
if @xmath49 and @xmath121 , for @xmath131 in the region iii ( unbroken @xmath0-symmetry ) , we numerically obtain the spectrum of eigenvalues @xmath114 ( see fig .
[ fig3c](a ) ) and a stable propagation of the soliton intensity @xmath125 shown in fig .
[ fig3c](b ) .
we stress that the stability ( instability ) of the bright solitons in broken ( unbroken ) @xmath0-symmetry region is due to the existence of kerr nonlinearity .
if @xmath122 and @xmath120 : ( i ) for @xmath132 in the region ii ( broken @xmath0-symmetry ) , we numerically obtain the spectrum of eigenvalues @xmath114 ( see fig . [ fig3k](a ) ) and a stable propagation of the soliton intensity @xmath125 shown in fig .
[ fig3k](b ) ; ( ii ) for @xmath133 in the region @xmath127 ( unbroken @xmath0-symmetry ) , we numerically obtain the spectrum of eigenvalues @xmath114 ( see fig . [ fig3k](c ) ) and a stable propagation of the soliton intensity @xmath125 shown in fig .
[ fig3k](d ) ; ( iii ) for @xmath134 in the region @xmath129 ( unbroken @xmath0-symmetry ) , we numerically obtain the spectrum of eigenvalues @xmath114 ( see fig .
[ fig3k](e ) ) and an unstable propagation of the soliton intensity @xmath125 shown in fig .
[ fig3k](f ) ; ( iv ) for @xmath135 , in the region @xmath88 , we numerically obtain the spectrum of eigenvalues @xmath114 ( see fig . [ fig3k](g ) ) and an unstable propagation of the soliton intensity @xmath125 shown in fig . [ fig3k](h ) . if @xmath122 and @xmath121 , for @xmath136 in the region iii ( unbroken @xmath0-symmetry ) , we see that @xmath137 ( see fig . [ fig3kc](a ) ) and a stable propagation of the soliton intensity @xmath125 is obtained as shown in fig .
[ fig3kc](b ) .
now we consider another type of @xmath0-symmetric scarff ii potential , i.e. the @xmath0-symmetric multi - well scarff ii potential , reading [ poten2 ] l v(x)=^2(x ) , + w(x)=w_0sech(x)(x ) , where @xmath138 and @xmath139 denotes the wavenumber . if @xmath140 and @xmath141 , then @xmath29 becomes the usual scarff ii potential @xcite , which exhibits the shape of single well , but for non - zero wavenumber @xmath142 , @xmath29 exhibits the shape of multi - well , which can provide more abundant structures . for @xmath143 , the external potential @xmath29 exhibits double - well ( @xmath144 ) , singe - well ( @xmath145 ) , and multi - well ( @xmath146 ) structures , respectively ( see fig .
[ fig - cos - v ] ) .
in fact , for the fixed parameters @xmath18 and @xmath142 , one can change the parameter @xmath147 to control the number of the wells .
notice that in the figure we plot the profile of @xmath148 because the hamiltonian of eq .
( [ nls ] ) is @xmath149 . as @xmath150 ( or @xmath151 ) , the linear problem associated with the @xmath0-symmetric potential ( [ poten2 ] ) admits an entirely real spectra provided that @xmath152 .
if @xmath153 , the condition for the unbroken-@xmath0 symmetry is complicated . for the fixed parameter @xmath143 , we find the regions of the unbroken-@xmath0 symmetry : ( i ) @xmath154 , @xmath155 ; ( ii ) @xmath156 , @xmath157 ; ( iii ) @xmath158 , @xmath159 ; and ( iv ) @xmath160 , @xmath161 . for the @xmath0-symmetric multi - well scarff ii potential ( [ poten2 ] )
, we consider the periodically space - modulated nonlinearity @xmath162 in eq .
( [ ode ] ) with the form [ g ] g(x)=(x ) .
for the zero wavenumber @xmath140 , nonlinearity @xmath163 is a constant , @xmath164 , however , for the given non - zero wavenumber @xmath142 , the nonlinearity @xmath163 is either positive or negative as @xmath27 increases .
based on the potential ( [ poten2 ] ) and the nonlinearity ( [ g ] ) we have exact bright solitons of eq .
( [ ode ] ) [ soluc ] ( x)=sech(x)e^i(x ) , where @xmath165 and the phase is ( x)=. the corresponding transverse power - flow or poynting vector is given by s(x)=(_x^*-^*_x)=^3(x ) , which implies that @xmath100 for any @xmath27 .
therefore , the power always flows in one direction , i.e. , from the gain toward the loss .
moreover , the conserved power is @xmath166 .
now we consider the linear stability of the soliton solution ( [ soluc ] ) using eq .
( [ stable ] ) with @xmath31 being replaced by the periodic function ( [ g ] ) with @xmath143 and different values of @xmath167 .
we find that for the fixed @xmath142 the soliton ( [ soluc ] ) is more ( less ) stable if @xmath168 ( @xmath169 ) . for examples ,
we show the linear stability eigenvalues and the stable propagation of the soliton ( [ soluc ] ) for @xmath170 ( see fig .
[ fig3-cos - g](a)-(b ) ) , @xmath171 ( see fig .
[ fig3-cos - g](c)-(d ) ) , @xmath172 ( see fig .
[ fig3-cos - g](e)-(f ) ) , and @xmath173 ( see fig . [ fig3-cos - g](g)-(h ) ) . for the fixed parameters @xmath174 , when @xmath175 , for which the @xmath0-symmetry is unbroken , the soliton ( [ soluc ] ) is still stable ( see fig . [ fig3-cos - b](b ) ) , however , when @xmath176 , in which @xmath0-symmetry is unbroken , the soliton ( [ soluc ] ) becomes unstable ( see fig .
[ fig3-cos - b](d ) ) .
thus , for @xmath174 , the parameter @xmath147 has significant influence on the linear stability of the soliton solution ( [ soluc ] ) .
a negative @xmath147 favors to obtain the stable solitons .
we now consider the 3d nls equation with the @xmath0-symmetric potential i_t + ^2++g||^2=0 , where @xmath177 is a complex field with respect to @xmath178 , @xmath179 , @xmath180 and @xmath181 are both real - valued functions related to external potential and gain - and - loss distribution , respectively , and @xmath31 is a real constant with @xmath101 .
the @xmath0-symmetric potential requires the sufficient ( but not necessary ) condition @xmath182 and @xmath183 .
the stationary solution can be solved in the form @xmath184 , where @xmath45 is the real propagation constant and @xmath185 satisfies [ ode2 ] = ^2+[v(x , y , z)+iw(x , y , z)]+g||^2 .
[ 0.42 ] with ( a ) @xmath120 , ( b ) @xmath186 , ( c ) @xmath187 , and ( d ) @xmath121 ; ( e ) the gain - and - loss distribution @xmath188 .
the parameters are @xmath189 and @xmath190.,title="fig : " ] [ 0.42 ] ( [ vf ] ) ; ( b ) the flux density @xmath191 ( [ div ] ) .
the parameter are @xmath189 and @xmath192 .
, title="fig : " ] we consider the generalized 3d @xmath0-symmetric scarff ii potential [ potend ] l v(x , y , z)=_=x , y , z(+2k_^2)sech^2(k _ ) + -g_0 ^ 2_=x , y , zsech^2(k _ ) , + w(x , y , z)=w_0_=x , y , zsech(k_)(k _ ) , where @xmath193 are the wavenumbers in @xmath194-directions , respectively , and @xmath195 are real constants . in particular , the @xmath0-symmetric potential ( [ potend ] ) reduces to the scarff ii potential ( [ poten ] ) when @xmath196 and @xmath197 .
since @xmath194 are symmetric in @xmath180 and @xmath181 thus we plot their profiles in @xmath198-space with @xmath199 ( see fig .
[ fig4-vw ] ) . for the self - focusing nonlinearity @xmath67 , the external potential @xmath200 exhibits the different profiles as @xmath31 becomes large ( see fig . [ fig4-vw](a)-(c ) ) . for the above - mentioned 3d potential ( [ potend ] ) we obtain the exact bright solitons of eq .
( [ ode2 ] ) [ solu2 ] ( x , y , z)=_0sech(k_xx)sech(k_yy)sech(k_zz)e^i(x , y , z ) , where @xmath201 and the phase is ( x , y , z)=_=x , y , zk_^-2 .
the real and imaginary parts and intensity of the solution ( [ solu2 ] ) are shown in fig .
[ fig5](a)-(c ) for @xmath202 and @xmath189 .
the velocity field @xmath203 of the solitons ( [ solu2 ] ) have the form [ vf ] rl * v*=&(x , y , z ) = & [ k_x^-1sech(k_xx ) , k_y^-1sech(k_yy ) , k_z^-1sech(k_zz ) ] , which is shown in fig .
[ fig6](a ) with @xmath189 .
it follow from eq .
( [ vf ] ) that the divergence of velocity field @xmath203 ( alias the flux density ) is given by [ div ] rl div*v*(x , y , z)=&*v*(x , y , z)=^2(x , y , z ) + = & -_=x , y , zsech(k_)(k _ ) , which measures the flux per unit area and is dependent on both @xmath18 and space position .
[ fig6](b ) shows the flux density in @xmath204-space .
in addition , we also have the relation * v*(x , y , z)=^2(x , y , z)=-13 w(x , y , z ) . from eq .
( [ div ] ) we have the following proposition . for the given parameter @xmath123 , we have : * the fluid flows outward if @xmath205 ; * the fluid flows inward if @xmath206 ; * the fluid does not flow if @xmath207 .
where we have introduced @xmath208 . for the case @xmath209
, we also have the corresponding results . for any complex solution @xmath210
, we find that its corresponding transverse power - flow or poynting vector and the gradient of phase obey the relation rl ( x , y , z ) & = ( ^*-^ * ) + & = |(x , y , z)|^2(x , y , z ) .
the transverse power - flow or poynting vector related to the solution ( [ solu2 ] ) is given by rl ( x , y , z)&=_0 ^ 2_=x , y , zsech^2(k_)*v*(x , y , z ) + & = _ = x , y , zsech^2(k _ ) + & , for either @xmath120 or @xmath211 ( i.e. it does not depend on the sign of the nonlinearity ) , which is exhibited in fig .
[ fig5](d ) .
in addition , the conserved power related to the solution ( [ solu2 ] ) is given by @xmath212 for either @xmath120 or @xmath121 .
in conclusion , we have presented a unified theoretical study of the optical bright solitons governed by self - focusing and defocusing nls equations with @xmath0-symmetric scarff - like ii potentials .
particularly , a @xmath0-symmetric @xmath1-wavenumber scarff ii potential and a multi - well scarff ii potential are considered , respectively . for the @xmath1-wavenumber scarff ii
potential , the parameter space can be divided into different regions , corresponding to unbroken and broken @xmath0-symmetry and bright solitons with self - focusing and defocusing kerr nonlinearities . for the multi - well scarff ii potential
the bright solitons can be obtained by using a periodically space - modulated kerr nonlinearity .
the linear stability of bright solitons with @xmath0-symmetric @xmath1-wavenumber and multi - well scarff ii potentials is analyzed in details by using numerical simulations .
stable and unstable bright solitons are found in both regions of unbroken and broken @xmath0-symmetry due to the existence of the nonlinearity .
furthermore , the bright solitons in 3d self - focusing and defocusing nls equations with a generalized @xmath0-symmetric scarff ii potential are explored .
the used method can also be used to study optcial solitons in other @xmath0-symmetric k - wavenumber potentials .
the results we obtained in this work provide a new way of control over soliton stability by using generalized @xmath0-symmetric scarff ii potentials in different parameter domains .
this may have potential applications in the field of optical information transmission and processing based on optical solitons in nonlinear dissipative but @xmath0-symmetric systems .
the authors would like to thank the referees for their invaluable suggestions .
this works of zyy and zcw were partially supported by nsf - china under grants no .
61178091 , and nkbrpc under grants no .
the work of c.h . was supported by nsf - china under grants no .
11475063 .
z. y. yan and d. m. jiang , phys .
e * 85 * , 056608 ( 2012 ) ; z. y. yan and v. v. konotop , phys .
e * 80 * , 036607 ( 2009 ) ; z. y. yan , nonlinear dyn .
* 79 * , 2515 ( 2015 ) ; y. q. yang , z. y. yan , and d. mihalache , j. math
* 56 * , 053508 ( 2015 ) .
the scarff potential is given by @xmath213 with @xmath214 .
the extension of the scarff potential in the complex space is ususally called the scarff ii potential , which is written as @xmath215 with @xmath214 .
the family of @xmath1-wavenumber open rays @xmath216 : @xmath217 and the family of @xmath1-wavenumber parabolas @xmath70 : @xmath218 have only one crossover point @xmath219 , i.e. , the open ray @xmath216 is tangent to the parabola @xmath70 at the point @xmath220 .
similarly , another family of @xmath1-wavenumber open rays @xmath221 : @xmath222 is also tangent to the parabola @xmath70 at the point @xmath223 . | we present a unified theoretical study of the bright solitons governed by self - focusing and defocusing nonlinear schrdinger ( nls ) equations with generalized parity - time ( @xmath0)-symmetric scarff ii potentials .
particularly , a @xmath0-symmetric @xmath1-wavenumber scarff ii potential and a multi - well scarff ii potential are considered , respectively . for the @xmath1-wavenumber scarff ii
potential , the parameter space can be divided into different regions , corresponding to unbroken and broken @xmath0-symmetry and the bright solitons for self - focusing and defocusing kerr nonlinearities . for the multi - well scarff ii potential
the bright solitons can be obtained by using a periodically space - modulated kerr nonlinearity .
the linear stability of bright solitons with @xmath0-symmetric @xmath1-wavenumber and multi - well scarff ii potentials is analyzed in details using numerical simulations .
stable and unstable bright solitons are found in both regions of unbroken and broken @xmath0-symmetry due to the existence of the nonlinearity .
furthermore , the bright solitons in three - dimensional self - focusing and defocusing nls equations with a generalized @xmath0-symmetric scarff ii potential are explored .
this may have potential applications in the field of optical information transmission and processing based on optical solitons in nonlinear dissipative but @xmath0-symmetric systems . |
since its introduction in 1975 @xcite the sherrington - kirkpatrick ( sk ) model for spin glasses has been one of the major challenges for the physicists interested in complex systems .
although it is a mean - field model the exact solution is still not completely certain .
nonetheless it is known @xcite that in the low - temperature phase the replica symmetry is spontaneously broken and this makes the solution highly non trivial .
in other words it can be rigorously proved that it is not possible that the connected correlation function of the spins at different points goes to zero when the total number of spins @xmath0 goes to infinity ; consequently the probability distribution of the overlap @xmath1 ( defined below ) can not be a single delta function as happens in usual model ( e.g. ferromagnetic model ) .
the existence of fluctuating intensive quantities ( like @xmath1 ) implies that it is not possible that only one equilibrium state is present in the thermodynamic limit . as usual extensive quantities
do fluctuate when more than one equilibrium state is present and consequently we could say , with some abuse of language ( see @xcite for a more precise discussion ) that in the sk model , for large values of @xmath0 more that one equilibrium state is present .
the presence of many equilibrium states implies that any analytic solution of the model should tell us something on the nature of these states , on their relative relations and on the probability at equilibrium of finding the system in one of these states .
of course this information should be of probabilistic nature given the presence of the quenched disorder in the system .
using the replica formalism an ansatz was proposed almost twenty years ago @xcite which makes some hypothesis on the nature of the states , the most notable being ultrametricity .
roughly speaking ultrametricity implies that the distance among the different states is such that they can be put in a taxonomic or genealogical tree such that the distance among two states is consistent with their position on the tree .
this hierarchical ansatz seems every day more reliable and , although the physical origin of ultrametricity was not fully evident , it is widely believed that it provides the correct solution of the sk model .
the ultrametric solution has passed many numerical tests and it is in agreement with all the known analytical results @xcite .
it is quite possible , and in agreement with the numerical simulations , that the ultrametric organization of the equilibrium configurations is also present in finite dimensional spin glasses @xcite .
it is certainly very interesting to find which are the physical assumptions at the basis of the hierarchical ansatz for two reasons : it would be easier to understand if the assumptions make sense also in the finite dimensional case and it could be easier to prove them of to extract their consequences .
a considerable progress has been done in recent years when it was realized that one of the main hypothesis at the basis of the hierarchical ansatz was stochastic stability : many compulsory arguments can be given for the validity of stochastic stability and the correctness of this hypothesis can be directly tested in experiments measuring the fluctuations and the response to a perturbation of the appropriate quantities . in the replica language _ stochastic stability _
is equivalent to the usual assumption of _ replica equivalence _
( i.e. each replica is equivalent to the others ) .
the other pillar of the hierarchical ansatz turned out to be ultrametricity .
indeed it can be shown that if we assume stochastic stability and ultrametricity , the whole hierarchical ansatz can be reconstructed @xcite .
the aim of this note is to show that there is a simpler property which is equivalent to ultrametricity .
this property can be called _ separability , _ in the replica language , or _ overlap equivalence _ : it states that all the mutual information about a pair of equilibrium configurations is encoded in their mutual distance or overlap .
in other words according to the principle of overlap equivalence any possible definition of overlaps should not give information additional to that of the usual overlap .
it is not clear if there are strong compulsory reasons for assuming overlap equivalence , however , the results presented here show that the hierarchical ansatz is certainly the simplest one that we may think for a stochastically stable system with many equilibrium states : any other possible proposal should include the presence of at least two inequivalent definitions of distance .
these two assumptions , stochastic stability and overlap equivalence , are quite general and can be applied to many other systems beyond the sk model . a direct test or an analytic proof of the validity of both properties would have direct implications on the validity of the ultrametric solution .
moreover as we have already remarked the results that we are going to present in this paper is interesting because they show the root of ultrametricity : ultrametricity is the unique possibility we have if we stay within the simple framework where stochastic stability and separability hold .
it should be clear that the whole discussion applies to systems in which the overlap fluctuates also when the volume is very large and consequently replica symmetry is broken .
for systems in which the overlap does not fluctuates and replica symmetry is exact we have nothing to say ( ultrametricity is satisfied , but in a trivial way ) .
it should also be clear that the arguments presented here can not be used to argue if replica symmetry breaking happens or not in a particular system .
we do not discuss here the criticism that have been done to the replica approach by @xcite ( who cast some doubts on the viability of replica symmetry breaking in finite dimensional systems ) : the reader may find a quite long discussion in @xcite . in the second section we will recall the replica formalism , in the third section we will present our assumptions . next in the forth section we will present our main results on the relation among overlap equivalence and ultrametricity .
finally we will present our conclusions .
some of the arguments needed to show that overlap equivalence implies separability are presented in appendix i and a part of the tedious algebra we have to do to reach the results is confined in appendix ii .
in this paper we make use of the replica formalism ( we address the reader to refs .
@xcite for an introduction on the issue ) .
let for simplicity restrict the discussion to systems with quenched random disorder in the hamiltonian .
a similar discussion can be done for system without quenched disorder ( like structural glasses ) if we substitute the average over the quenched disorder with the average over the size of the system .
we consider a system with @xmath0 spins characterized by an hamiltonian @xmath2 ( where @xmath3 represents the quenched disorder ) .
we define @xmath4 to be the probability distribution of finding two equilibrium configurations @xmath5 and @xmath6 at the same inverse temperature @xmath7 with overlap @xmath1 , the overlap @xmath1 being defined as @xmath8 let us assume that , in the large @xmath0 limit , the function @xmath4 does not becomes a delta function ( otherwise we have nothing interesting to say ) and therefore the function @xmath4 has a non trivial shape also for large @xmath0 .
when this happens , we are interested to find out the probability distribution of the function @xmath4 in the limit where @xmath0 goes to infinity . for a given value of @xmath0 different choices of @xmath3 may produce different functions @xmath4 and we can introduce the functional @xmath9 as the probability distribution of @xmath4 at fixed @xmath0 ( we assume that the @xmath3 have a given probability distribution ) .
eventually we would like to know @xmath10 we are also interested to control the behavior of the probability distribution of the mutual overlaps among three or more equilibrium configurations ( e.g. the probability @xmath11 which will be properly defined later ) .
the origin of our interest in the probability distribution of the overlap is due to the fact that they control many others physical properties of the systems : e.g. in some models one finds @xcite that the magnetic susceptibility is given by @xmath12 where @xmath13 is the average over @xmath3 of the equilibrium expectation value of @xmath1 . in the replica formalism the behavior of the these probability distributions
is encoded by an @xmath14 symmetric matrix @xmath15 in the limit of @xmath16 ( taken after the analytical continuation of @xmath17 from integer to real values ) .
the limiting matrix depends on all the matrices with any value of @xmath17 and so the general solution has an infinite number of parameters and the analytical continuation of the matrix @xmath15 is , in general , dependent on an extremely high number of parameters .
this is quite natural as far as the matrix @xmath15 encodes the properties of the functional which controls the probability distribution of finding for a random @xmath3 a set of probability distributions for the overlaps ( i.e. @xmath4 , @xmath11 , ) . in the hierarchical ansatz
the @xmath17 replicas are divided into many groups of equal sizes , such that , if the replica indices @xmath18 and @xmath19 belong to the same group , then has a higher value than whether @xmath18 and @xmath19 are in different groups .
the groups are then divided in subgroups and so on for an infinite number of times .
this kind of solution can be summarized in an infinite set of parameters ( the size of the group and the value of the overlap at each level ) . in the limit @xmath16 these parameters
can be conveniently represented by a function defined for @xmath20 $ ] , where is the probability of finding in the matrix @xmath15 an element of value @xmath1 . to
every ultrametric matrix @xmath15 corresponds one and only one probability distribution function . in the paramagnetic phase all the elements of @xmath15 are equal and
the function is a delta function . in the spin glass phase
the elements take different values and the acquires a finite width .
the relation of this function with the probability distribution function of the overlap is @xmath21 where the bar denotes the average over @xmath3 at fixed @xmath0 .
the equality of the two functions and @xmath22 is one of the many relations among probability distribution functions of the overlaps and the matrix @xmath15 .
more complicated probability distribution functions ( pdf ) can be defined , considering the joint probability of more than one overlap .
for example a crucial role is played by the joint pdf of 3 real replicas @xmath23 , defined as the @xmath3-average of @xmath24 where @xmath25 , @xmath26 and @xmath27 are three equilibrium configurations and @xmath28 with the indices @xmath29 and @xmath7 running from 1 to 3 . if ultrametricity holds , this probability distribution has the following property @xmath30 as soon as the ultrametricity relations @xmath31 are not satisfied .
the um property can be easier understood by a geometrical picture . given three configurations , that is three points in the configurational space , um implies that they can be the vertices of two kind of triangles only : equilateral or isosceles , with the two equal edges larger than the third one .
the other kind of isosceles triangles , together with the scalene triangles , can not be obtained with any tern of equilibrium configurations , if these configurations are organized in an um fashion .
this property , which is satisfied in the hierarchical ansatz , has rather strong consequences .
firstly , as far as probabilities can not be negative , the previous relations implies that also for any @xmath3 ( with probability one ) we have that @xmath32 as soon as the relations in eq.([eq : dis ] ) are not satisfied .
this result has the consequence that any equilibrium configuration can be assigned ( for fixed @xmath3 ) to a leave of a tree constructed in such a way that the overlap among two equilibrium configuration is related to the distance of the two configurations on the tree .
ultrametricity implies for example that , if two equilibrium configurations 1 and 2 are at overlap @xmath33 , any equilibrium configuration 3 such that @xmath34 satisfies also the relation @xmath35 .
ultrametricity is very interesting because it implies that many pdf of more than three overlaps are zero in a wide region and reduces the whole problem to the construction of the statistical properties of the aforementioned tree . moreover , if stochastic stability is valid , the ultrametricity completely determines the @xmath36 given the .
one can show that @xcite @xmath37 where @xmath38 and @xmath39 in other words stochastic stability and ultrametricity allow us to obtain all the pdf of the overlap starting from the knowledge of the function .
it is clear that it is extremely difficult to arrive to some general conclusions on these probability distributions without doing extra assumptions .
we now show that two rather simple assumptions : replica equivalence ( or equivalently stochastic stability ) and separability are giving very strong constraints .
let us firstly consider replica equivalence as formulated in replica formalism .
as we have already stated the properties of the probability distribution of the overlaps can be obtained in terms of the matrix . even in the low - temperature phase ,
when the matrix elements are not constant , we may expect no physical difference between the replicas ( which have been introduced as a mathematical trick ) .
replica equivalence states that the observable which involve only one replica are replica symmetric , i.e. the assume the same value .
for example replica equivalence implies that we must have that @xmath40 does not depend on @xmath18 .
replica equivalence is equivalent to the stochastic stability property introduced by guerra @xcite and aizenman and contucci @xcite which is valid under general conditions , i.e. if we introduce an arbitrary small random long range hamiltonian ( see @xcite for a more careful discussion ) .
eq.([re ] ) implies that each line ( column ) of the matrix @xmath15 is a permutation of the other lines ( columns ) .
moreover it has interesting consequences : with some algebra the following equalities can be proven @xmath41 the proof of these equations can be done by recalling some relations among the matrix and the probability functions @xmath42 where the primed sum is done over all different replica indices .
let us denote @xmath43 the sum does not depend on @xmath18 as consequence of replica equivalence .
it is also evident that @xmath44 if we now look to the consequences of the previous equation and we use the relations in eq.([prima ] ) both for @xmath45 and @xmath46 , we obtain the two relations in eq.([gue_rel ] ) .
identical relations have been proven by guerra @xcite , using stochastic stability .
we can very safely assume that they must be valid in any scheme of replica symmetry breaking .
eq.([gue_rel ] ) determine all the joint pdf of 2 real replicas in terms of the .
the consequences of stochastic stability have been lengthy discussed in @xcite . in the nutshell stochastic stability implies that the system is a generic random system and it does not have any special properties : its properties are smooth functions of any external random perturbation .
the second assumption we made is the separability ( also know as non - degeneracy ) of the matrix @xmath15 @xcite , which correspond to the following statement .
let us consider all the matrices which can be generated from the matrix @xmath15 in a permutational covariant fashion .
some example are @xmath47 separability states that , if we take two pair of indices ( @xmath48 and @xmath49 ) , we have that @xmath50 where @xmath51 is a generic matrix of the set generated by the rules shown in eq.([set ] )
. in other words pairs of indices which have different properties have different values of the overlap .
it means that we can classify a pair of replicas in terms of their mutual overlap @xcite and that no finer classification of their mutual properties is possible .
the physical meaning of separability can be understood if we introduce another concept , the overlap equivalence .
let us consider an arbitrary local observable @xmath52 .
simple examples of such an observable are @xmath53 where @xmath54 and @xmath55 are appropriate functions ( e.g. they decrease sufficiently fast at infinity )
. many more complex choices of the local observable @xmath56 can be constructed , for example those involving more than two spins . for any choice of the operator @xmath56 we could define a generalized overlap @xcite : @xmath57 in the hierarchical ansatz it turns out @xcite that for any reasonable choice of the observable @xmath56
, @xmath58 is a function of @xmath1 . in other words when we change the two equilibrium configurations and the couplings @xmath3 the values of @xmath1 and @xmath58 fluctuates also for very large @xmath0 , while the value of @xmath58 restricted on those pairs of configurations with a fixed value of @xmath1 does not fluctuate when @xmath0 goes to infinity ( that is a scattered plot of @xmath1 and @xmath58 should collapse on a curve in the limit of large @xmath0 ) . in other words
overlap equivalence implies that in the case where replica symmetry is broken and all overlaps fluctuate in the usual thermodynamical ensemble , these fluctuations disappear in the fixed @xmath1 ensemble .
in other words overlap equivalence states that for a system composed by two replicas the overlap is a good , complete order parameter in the same way as the magnetization is for ferromagnetic systems .
if we stay in the usual thermodynamic ensemble , there are many quantities that fluctuates also at large distances , however if we consider the restricted ensemble where the order parameter takes a given value , all fluctuations at large distance disappear and the connected correlation functions go to zero at infinity .
this happens only if the order parameter has been chosen in such a way to carry enough information : in a ferromagnetic heisenberg model the order parameter must be the three component vector of the magnetization , one or two components of the magnetization would be not enough to fully characterize the state of the system in case of a spontaneous magnetization
. a direct check of overlap equivalence can be done in usual numerical simulations and it would be very interesting to see the results .
this property is called overlap equivalence because it states that all possible definitions of the overlap are equivalent and there is an unique correspondence among the values of the different overlaps .
it is clear that the overlap equivalence is a very strong simplification .
in general we could have that the mutual relations among two equilibrium configurations are characterized by a large , possibly infinite set of independent overlaps and therefore their mutual relations are characterized by a large ( or infinite ) set of parameters .
the property of overlap equivalence implies a much simpler situation , where only one parameter ( the overlap @xmath1 ) characterizes the mutual relations among two equilibrium configurations .
we can argue that separability is the way to code overlap equivalence in the replica formalism . both properties state that once the overlap among two _ objects _ is fixed , all the mutual relations among the two objects are also fixed .
the difference among these two statements is that in the case of replica equivalence the two objects are equilibrium configurations while in the case of separability the two objects are replicas .
the identification of separability with overlap equivalence is quite natural because the structure of the matrix @xmath15 in replica space mirrors the structure of the mutual overlaps of equilibrium configurations . in appendix
i we present some more detailed considerations which points toward the correctness of the identification of these two properties , however a more general and formal proof of this statement would be welcome . it is interesting to note that in the simplest model leading to ultrametricity , i.e. a branching random process in the infinite dimensional space the condition of overlap equivalence is satisfied @xcite .
indeed if we consider a random vector @xmath59 in a finite dimensional space ( of dimension @xmath0 ) the quantities @xmath60 convey different information when @xmath29 changes from 1 to @xmath0 and can be used as different measures of the distance . on the other end ,
when @xmath0 goes to infinity at fixed @xmath61 , thanks to the rotational invariance , we have that for each @xmath29 , @xmath62 .
then if we introduce generalized distances parameterized by @xmath63 ( where @xmath64 ) and defined as @xmath65 it is easy to check that with probability 1 ( if the probability distribution is rotational invariant ) in the limit @xmath66 @xmath67 therefore in this simply model overlap equivalence is automatically satisfied .
let us consider what happens in the usual hierarchical ansatz . in this case ,
when replica symmetry is broken , there is a subgroup of the group of permutations that commutes with the matrix @xmath15 .
let us consider the orbits in the space of pairs of indices .
it can be checked that the values of the elements of the matrix @xmath15 and of any matrix derived using the rules in eq.([set ] ) are constant of the orbits and that different values of @xmath1 do correspond in general to different orbits .
moreover it can be checked that both separability and overlap equivalence hold in this case .
maybe the simplest non - trivial example of a non - ultrametric system is given by the union of separately ultrametric system with a non - trivial distribution of the overlaps @xcite .
it is easy to check that @xmath68 does not satisfy the ultrametric condition also if @xmath69 and @xmath70 do satisfy it .
however it is clear that in this example @xmath69 and @xmath70 are generalized overlaps which are not function of @xmath1 .
both ultrametricity and overlaps equivalence disappear at the same time .
the separability condition is extremely powerful in determining the expectation values of higher order moments of the probability distribution .
let us study a simple example and let us consider two matrices @xmath51 and @xmath71 constructed with the rules in eq .
( [ set ] ) .
it is evident that we can write @xmath72 where we have defined @xmath73 indeed separability implies that the matrix elements @xmath74 and @xmath75 are constant in the region where @xmath76 and their value is denoted @xmath77 and @xmath78 respectively . in the same way we have that : @xmath79 therefore separability implies that quantities like those in eq.([prod ] ) can be computed from the knowledge of those in eq.([due ] ) .
if we introduce the functions @xmath80 , @xmath81 and @xmath82 such that @xmath83 the previous equations implies that @xmath84 the last equation can also be written as @xmath85 if we apply the previous formula to the case where @xmath51 and @xmath71 have the form @xmath86 and we consider all the possible values of the @xmath87 s , we find ( after separating the contributions where some of the indices are equal ) the rather surprising formula @xmath88 similar results can be obtained for other probability distributions with more overlap .
eq.([woh ] ) is particular interesting because integrating over @xmath1 it implies that @xmath89 + \nonumber \\ + 2 \int { { \rm d}}q { p^{12,23,31}(q , q_1,q_2 ) p^{12,23,31}(q , q_3,q_4 ) \over p(q ) } \quad .
\label{main}\end{aligned}\ ] ] the previous equation is remarkable not only because it give the full expression of the probability with four overlap in terms of the probability with three overlaps but also because it puts hard constraints on the possible values of the function @xmath90 .
indeed the l.h.s . of eq.([main ] ) is by definition invariant under cyclic permutation of the @xmath1 s , while the r.h.s . of the same equation
is not invariant for a generic choice of the function @xmath36 .
what is the form of the generic function @xmath36 that satisfy eq.([main ] ) ?
we will argue in the next section that it must be ultrametric .
our problem is now that of finding the most general matrix @xmath15 ( or equivalently the most general probability distribution ) compatible with the replica equivalence [ and then with guerra s relations , eq.([gue_rel ] ) ] and with separability [ and then in particular with eq.([main ] ) ] . we will show that the most general matrix is the ultrametric one .
we consider the case when few values ( @xmath91 , @xmath92 or @xmath93 ) are allowed for the matrix elements .
the generalization to more than 5 values is straightforward and we hope that our conclusions will still be valid for a generic which have a continuous distribution of possible values . when the overlap ( or the matrix elements ) can take only @xmath87 different values the function is the sum of @xmath87 delta functions @xmath94 where the weights @xmath95 are , by definition , positive and such that @xmath96 , and the values @xmath97 are different .
also the joint pdf of 3 overlaps @xmath36 ( that hereafter we will call for brevity ) is the the sum of @xmath98 delta functions on the points @xmath99 with @xmath100 and so we should give the @xmath98 weights @xmath101 in order to determine .
we can lower the number of free parameters @xmath101 using some symmetries and the guerra relations
. the weight of the tern @xmath99 must be the same of any permutation of it , i.e.@xmath102 .
then the number of really independent parameters in is @xmath103 .
more relations between the @xmath101 can be obtained exploiting the following equation , which is based essentially on the first guerra s relation @xmath104 these are @xmath105 relations that lower the degrees of freedom of to @xmath106
. then we have to determine the values of these @xmath106 parameters which are compatible with eq.([main ] ) . to fix the ideas , let us write down some formul for the easier case ( @xmath91 ) where we have 27 parameters @xmath101 : @xmath107 .
the symmetries imply that @xmath108 while the guerra s relations imply some equalities like @xmath109 we end with only 4 free parameters ( @xmath110 ) : @xmath111 the way we have ordered the probabilities is meaningful : we call @xmath112 the weight of the scalene triangle ( which is forbidden in the ultrametric solution ) and we call @xmath113 the weights of the isosceles triangles ( which are also forbidden in the um ansatz if we assume the overlap ordering @xmath114 ) . if we do not fix any order in the values of the @xmath97 , we have to keep in mind that , whether we exchange two of the overlap values , the forbidden isosceles triangle changes .
for example , if @xmath114 then ultrametricity implies @xmath115 , while when we reverse the second inequality , i.e. @xmath116 , we have that @xmath117 . then we note that @xmath118 while the reversed implication is not true , because um also holds for different parameters values , e.g. @xmath119 and @xmath120 ( that corresponds to the ordering @xmath116 ) . for a generic @xmath87 we have @xmath121 scalene parameters @xmath122 which must be all identically zero in order that um hold ( @xmath123 ) , while the @xmath124 isosceles parameters @xmath125 must be zero or @xmath126 , depending on the order of @xmath97 and @xmath127
. we will now use eq.([main ] ) to determine the values of all these parameters .
the l.h.s . of eq.([main ] ) is invariant under cyclic permutations of the four overlaps .
this allows us to obtain useful relations simply taking two of these equations ( the second one with the overlaps cycled with respect to the first one ) and equating the right hand sides .
the number of non - trivial equations we can obtain in this way is large enough to fix all the parameters . in the particular case of
@xmath91 we have that all the non - trivial equations are equal ( this is highly fortuitous ) and read @xmath128 using the relations that comes from the sixth equation in ( [ grande ] ) @xmath129 we can write eq.([brutta ] ) as @xmath130 where in @xmath131 we put all the terms that survive once we set @xmath132 .
the coefficients of @xmath112 and @xmath133 are positive defined ( thanks to the positiveness of all the probabilities ) and @xmath131 in non - negative ( as we will show in a while )
. then eq.([bella ] ) is equivalent to @xmath134 as a first result we obtain that scalene triangles are completely forbidden .
let us now introduce the following symbol @xmath135 for @xmath136 $ ] we have that @xmath137 and the equality @xmath138 only holds on 6 of the 12 edges of the cube ( those in bold face in fig .
[ fig ] ) .
= 0.4 if we introduce the new parameters @xmath139 that belong to the range @xmath140 $ ] thanks to the positiveness of the probabilities , then the second equality in eq.([s0 ] ) can be rewritten in a very compact form as @xmath141 this form makes clear that @xmath131 is non - negative , as we claimed above .
eq.([corta ] ) is not so stringent as ultrametricity would , but the deviations from um are small . in fact , in the cube @xmath142 $ ] ( see fig .
[ fig ] ) strict um only holds on the vertices marked by a circle . while eq.([corta ] ) is satisfied along the bold lines too .
for example , on the segment @xmath143 and @xmath144 they seem to co - exist non - zero probabilities @xmath145 and @xmath146 and it would be a small violation of um .
however we know that @xmath147 correspond to the ordering @xmath114 , while @xmath143 and @xmath148 correspond to @xmath149 .
then we believe that the points on the segment between this two um points corresponds to the case @xmath150 , when there is no difference between @xmath145 and @xmath146 ( but we can not still prove it ) . in conclusion , in the case with @xmath91 overlaps
, we have that the scalene triangle and two of the three `` wrong '' ( in the um sense ) isosceles triangles are forbidden . as we will see below , the um violations become smaller and smaller as @xmath87 is increased . in this section
we would like to sketch how the information we need about can be systematically derived from eq.([main ] ) . to make this section more readable , the formul relatives to the cases
@xmath152 will be presented in appendix ii .
the method we use to obtain the results does not depend on @xmath87 and so we will be able to generalize our findings to whichever that is the sum of a finite number of delta functions .
the many equations derivable from eq.([main ] ) can be divided into three classes : those with 2 , 3 and 4 different values of the overlaps .
these equation are not independent : those with 2 ( respectively 4 ) different overlaps can be expressed as the sum ( resp .
difference ) of those with 3 overlaps .
there are many ways of solving the equations . here
we present the simplest one we were able to find : we have to consider only the equations with 2 and 3 overlaps , i.e. those which respectively correspond to the equalities @xmath153 and @xmath154 each one of these equations can be identified giving a pair or a tern of numbers : @xmath155 or @xmath156 .
the l.h.s . of these equations will be called respectively @xmath157 and @xmath158 , for brevity .
our demonstration follows two steps : first we show that the equations of the kind of eq.([dueq ] ) can be solved only if all the scalene parameters @xmath101 ( with @xmath159 different ) are zero , then we find all the solutions for the simplified set of equations corresponding to eq.([treq ] ) .
our demonstration is essentially based on the non - negativity of the @xmath157 expressions and on the properties of the double - parenthesis symbol , previously introduced .
first of all we note ( see appendix ii ) that when we set to zero all the scalene parameters @xmath101 ( with @xmath159 different ) , every @xmath157 becomes the sum of some double - parenthesis symbols , and so it is non - negative . moreover in some of the @xmath157 expressions all the scalene parameters have positive defined coefficients , and then we should set them to zero in order to solve the equation , @xmath160 . in appendix
ii we present a possible way of choosing the @xmath157 expressions in order to systematically set to zero all the scalene parameters .
once the scalene parameters have been set to zero , we prefer working with eq.([treq ] ) , because each equation identified by @xmath156 takes a very simple form : @xmath161 where we choose the indexes such that @xmath162 . in general , given 3 different overlap values , we can easily write down the corresponding equation [ of the kind of eq.([treq ] ) ] , which gives , once we set all the scalene parameters to zero , the corresponding double - parenthesis symbol [ of the kind of eq.([final ] ) ] . what about the equation with 2 and 4 different overlaps ?
when we set all the scalene parameters to zero , we have that the expressions @xmath157 are the sum of @xmath163 of these double - parenthesis symbols , those derived from the overlap terns @xmath164 with @xmath165 and @xmath166 ( that is those where the parameter @xmath167 appears ) . on the other hand the equations with 4 different overlaps are identically satisfied and they are useless .
then we conclude that , in the more general solution , eq.([final ] ) must hold for every overlaps tern @xmath168 .
what does it imply in terms of the ultrametric properties of ? for any pair of overlaps @xmath169 we have two different isosceles triangles : a `` right '' one ( i.e. allowed by um ) with probability @xmath170 and a `` wrong '' one ( i.e. forbidden by um ) with probability @xmath171 . in the solution we have found , almost all the wrong isosceles triangles are forbidden .
more precisely , for any pair of overlaps @xmath169 , such that there is an overlap @xmath172 in between ( @xmath173 ) , we have that @xmath174 and the wrong isosceles triangle is not allowed . that can be easily proved noting that , for any @xmath173 , the equation @xmath175 forces @xmath174 .
small um violations can appear only when one considers nearest neighbors overlaps pairs .
in this case both the right and the wrong isosceles triangles are allowed .
however , for any fixed @xmath87 , the maximum number of wrong isosceles triangles allowed is @xmath176 $ ] , while the total number of isosceles triangles is proportional to @xmath177 .
so in the limit @xmath178 the probability of having wrong isosceles triangles tends to zero .
moreover , if in the continuum limit the @xmath22 is dense on a single compact domain , the distance between any pair of nearest neighbors overlaps tends to zero for @xmath178 and then strict um holds for any finite overlap difference @xmath179 .
lastly it should be noted that we have not exploited all the available information and maybe even these small um violations could be ruled out with some more work .
in fact we believe that the solution with both @xmath180 and @xmath170 different from zero , actually corresponds to the case @xmath181 and it is not really an um violation
. maybe it should be also possible to find a direct proof of our result directly in the continuum limit , without considering the intermediate case in which the number of steps are finite , using maybe the techniques introduced by ruelle @xcite , however we have not succeeded in this task .
we have seen that in systems where the function @xmath22 is non trivial and the overlap is a fluctuating quantity , stochastic stability and separability imply ultrametricity .
the reader should notice that the proofs presented here are likely to be too involved and it is quite possible that there is a direct proof of the fact that replica equivalence implies ultrametricity .
at this end we recall that in a dynamical approach it was show that one can identify a dynamical equivalent of separability : i.e. we can assume that in the aging regime all the possible overlaps among two configurations at two quite different times ( @xmath182 and @xmath183 ) are functions of the usual overlap among the two configurations at the same times ( @xmath182 and @xmath183 ) .
it was possible to prove that this dynamical overlap equivalence implies a dynamical form of ultrametricity .
this result implies that , if we do not give up stochastic stability ( which is a general property of generic equilibrium systems ) violations of ultrametricity may be found only in systems for which the separability conditions does not hold and the mutual relations among equilibrium configurations is described by two or more overlaps .
the probability distribution of such a system ( if it exists in the framework of equilibrium statistical mechanics ) would be much more complex of that of the usual ultrametric ansatz .
we can thus conclude that the ultrametric solution is the simplest one .
our arguments imply that it would be particularly interesting to check the numerical validity of overlap equivalence .
this task can be done at high precision using present numerical technology .
this could be done just doing numerical simulations in the ensemble with fixed overlap and looking to the fluctuations of the other overlaps .
we thank s. franz and m. a. virasoro for many interesting discussion on the subject of the paper .
in this appendix we present some arguments in order to show that _ separability _ implies _ overlap equivalence_. to this end let us consider a specific spin glass model @xmath184 where the variables @xmath3 are gaussian uncorrelated random variables with zero average and variance @xmath185 in short range models @xmath186 is a fast decreasing function of the distance among the two points @xmath187 and @xmath87 while in the sk model @xmath188 , @xmath0 being the total number of spins . in this model
we can define not only the usual overlap but also a modified overlap among two configurations @xmath5 and @xmath6 which we denote by @xmath189 : @xmath190 in the same way we can define an overlap among two replicas , which we denote by @xmath191 @xmath192 we notice that by simple integration by part on the gaussian variables @xmath3 one can prove that @xmath193 with @xmath194 where by the overbar we denote the average over the random quenched variables @xmath3 .
a similar computation tells us that the fluctuations of the quantity @xmath189 at fixed @xmath1 are ( neglecting terms which go to zero with the volume ) the same as the fluctuations of @xmath195 at fixed @xmath196 . on the other end
it is evident that @xmath197 separability states that the last sum takes a fixed value which does not fluctuates if we stay in the ensemble of fixed @xmath198 . at this level
the relation among separability and overlap equivalence is clear : the first is equivalent to the statement that the quantity @xmath199 does not fluctuate ( in the fixed @xmath196 ensemble ) when @xmath200 is large , while for overlap equivalence we need the same quantities does not fluctuate when the distance among @xmath187 and @xmath87 is fixed .
the two properties seems to be slightly different if the interaction is short range . on the contrary
if the interaction is long range the two formulations are the same . in order to be more precise
we can consider a model in which a small long range interaction has been added and the same argument as before can be used to prove that overlap equivalence implies separability .
moreover the presence of a long range term should not affect too much the properties according to the principle of stochastic stability which tells us that the system should be stable with respect to a small random perturbation .
the argument we have presented here tell us that replica equivalence implies separability .
in this appendix we show some details of the computation we have done in the cases @xmath92 and @xmath93 .
in particular we show exactly how to derive the solution for the @xmath92 case , while we simply sketch it in the @xmath93 case .
let s consider the equation of the kind of eq.([dueq ] ) with the two greatest overlaps ( @xmath204 and @xmath205 in the @xmath92 case ) , @xmath206 using some of eqs.([def ] ) and multiplying the previous equation by @xmath207 we end with the following equation @xmath208 ( p_{432}+p_{431 } ) + \nonumber \\ + [ 4 p_1 p_3 p_4 ] p_{432 } p_{431 } + [ 2 p_1 p_4 ( p_2+p_3 ) ] p_{432}^2 + [ 2 p_2 p_4 ( p_1+p_3 ) ] p_{431}^2 = 0 \quad .
\quad \label{eq43}\end{aligned}\ ] ] in @xmath209 there are all the terms that survive from the expression @xmath210 when we set all the scalene parameters to zero . in general all the expressions
@xmath211 are non - negative defined ( see below ) . the coefficients of @xmath212 and @xmath213 in eq.([eq43 ] )
are positive defined , thanks to the inequalities @xmath214 , that are direct consequence of eqs.([def ] ) and of probabilities positiveness .
then we have that eq.([eq43 ] ) is equivalent to @xmath215 and @xmath216 .
to force the two remaining scalene parameters to zero , it is enough to consider the equation analogous to eq.([ciao ] ) with @xmath217 and @xmath218 instead of @xmath205 and @xmath204 .
once we set @xmath215 we obtain @xmath219 p_{421 } + \nonumber \\ + [ 2 p_1 p_3 p_4 ( a_{42}+a_{32 } ) + p_1 p_2 p_4 ( p_3 p_2 -2 a_{32 } + p_3 p_1 -2 a_{31 } ) ] p_{321 } + \nonumber \\ + [ 2 p_3 p_4(p_1+p_2 ) ] ( p_{421 } + p_{321})^2 = 0 \quad . \label{eq21}\end{aligned}\ ] ] again we note that all the coefficient are positive defined thanks to eqs.([def ] ) and to the positiveness of the probabilities .
eq.([eq21 ] ) implies @xmath220 and @xmath221 . in order to obtain this result
we have made only one assumption , about the non - negativity of @xmath209 and of @xmath222 , which we now show to be correct . using the rescaled variables , @xmath223 , those expressions read @xmath224
note that @xmath209 ( resp .
@xmath222 ) is the sum of the two double - parenthesis symbols containing @xmath225 ( resp .
@xmath226 ) . once we set to zero all the scalene parameters @xmath227 , we found easier to work with equations of the kind of eq.([treq ] ) .
for example , considering the 3 overlaps @xmath114 , we have that @xmath228 in general for every 3 given overlaps we end with an equation like eq.([final ] ) .
the way to force the scalene parameters to zero should be now clear : it exploit the coefficient positiveness in the equations with 2 different overlaps .
maybe it is still not so clear if there is a systematic way to set all those parameters to zero , without getting lost in the many @xmath157 expressions .
we found such a systematic way and we will illustrate it in the case with @xmath93 different overlaps .
let s always consider first the equation with the two greatest overlaps ( @xmath229 and @xmath204 in this particular case ) .
it implies @xmath230 note that all the scalene probabilities forced to be zero contain both @xmath229 and @xmath204
. then let s substitute the just found solution ( @xmath231 ) into all the other equations and let s go forward in the same way @xmath232 at this point we end with the same scalene parameters we work with in the @xmath92 case and then let s follow the same steps as in the previous section @xmath233 g. parisi , phys .
a * 73 * , 203 ( 1979 ) ; phys .
* 43 * , 1754 ( 1979 ) ; j. phys .
a * 13 * , l115 ( 1980 ) ; j. phys . a * 13 * , 1101 ( 1980 ) ; j. phys . * a 13 * , 1887 ( 1980 ) ; phys . rev
. lett . * 50 * , 1946 ( 1983 ) . | in this paper we show that in systems where the probability distribution of the the overlap is non trivial in the infinity volume limit , the property of ultrametricity can be proved in general starting from two very simple and natural assumptions : each replica is equivalent to the others ( _ replica equivalence _ or _ stochastic stability _ ) and all the mutual information about a pair of equilibrium configurations is encoded in their mutual distance or overlap ( _ separability _ or _ overlap equivalence _ )
. = -1.5 cm = 24.0 cm = 15.8 cm = 0 cm |
a number of observations , mainly at radio frequencies , have established that relativistic particles and magnetic fields are part of the intracluster medium ( icm ) of galaxy clusters ( e.g. , * ? ? ?
* ) . the large ( @xmath8mpc ) scale , diffuse structures known as radio halos and relics are produced by relativistic electrons spiraling around @xmath8@xmath7 g magnetic fields .
the synchrotron emission is a product of both the particle and magnetic field energy densities , the latter of which is not well constrained globally from these or other observations .
however , the electron population can be independently detected through inverse compton ( ic ) scattering off of ubiquitous cosmic microwave background ( cmb ) photons , which are up - scattered to x - ray energies and may be observable if the electron population is sufficiently intense @xcite . for single electrons or populations with power law energy distributions ,
the ratio of ic to synchrotron flux gives a direct , unbiased measurement of the average magnetic field strength @xmath9 in the icm of a cluster .
the magnetic field plays a potentially important role in the dynamics and structure of the icm , such as in sloshing cool cores where @xmath9 may be locally amplified so that the magnetic pressure is comparable to the thermal pressure @xcite .
detections of ic emission , therefore , probe whether the non - thermal phase is energetically important or , particularly if the average magnetic field is large , it is sizable enough to affect the dynamics and structure of the thermal gas .
the quest for the detection of ic emission associated with galaxy clusters began with the launch of the first x - ray sensitive sounding rockets and satellites , although the origin of extended , @xmath8 kev x - rays from clusters was soon recognized to be thermal ( e.g. , * ? ? ? * ; * ? ? ? * ) .
even so , in clusters with radio halos or relics , ic emission _ must _ exist at some level , since the cmb is cosmological .
thermal x - ray photons are simply too numerous at @xmath10 kev for a reliable detection of the ic component ; at higher energies , however , the bremsstrahlung continuum falls off exponentially , allowing the non - thermal ic emission to eventually dominate and produce `` excess '' flux in the spectrum .
while the first ic searches with yielded only upper limits , and thus lower limits on the average strength of icm magnetic fields , @xmath11 g @xcite , the next generation of hard x - ray capable satellites and
produced detections in several clusters , although mostly of marginal significance ( for a review , see , e.g. , * ? ? ?
the most recent observatories and however , have largely failed to confirm ic at similar levels @xcite .
one exception is the bullet cluster ( a.k.a .
1e 0657 - 56 , rx j0658 - 5557 ) , although the detection significance of the non - thermal component is marginal in both the and data alone .
the observation of the bullet cluster s had x - ray emission was not very constraining , but the overall spectrum from the pca and hexte instruments , fit jointly with mos data , favored a non - thermal tail at not quite @xmath12 significance @xcite .
a two - temperature model fit the data equally well , but the higher temperature component had a nearly unphysically high temperature ( @xmath13 kev ) for a large ( 10% ) fraction of the total emissivity . in a similar analysis ,
the data were simultaneously fit with a spectrum from the bat all sky survey , and the non - thermal component was confirmed at the @xmath14 confidence level @xcite .
a two - temperature model technically did a better job of describing the spectra , although the secondary temperature component was very low ( 1.1 kev ) , causing the authors to reject this interpretation .
while this low temperature component is certainly not physical , the fact that a model can fit the data so well when an extra component is added solely at low energies indicates that the non - thermal component is not being strongly driven by the bat data .
further confirmation of an ic component in the bullet cluster is clearly necessary to rule out a purely thermal description of the hard band emission and uphold the implied magnetic field strength of @xmath15 g .
the intriguing evidence for a non - thermal excess at hard energies coupled with its smaller angular size makes the bullet cluster an ideal galaxy cluster target for the x - ray observatory @xcite .
is the first focusing hard x - ray telescope with a bandpass between 3 and 80 kev and is the first telescope with the ability to focus x - rays in the hard x - ray band above 10 kev .
it has an effective area at 30 kev of @xmath16 @xmath17 and imaging half power diameter ( hpd ) of @xmath18 .
while the effective area is somewhat lower than that of previous instruments , the focusing capability vastly reduces the background level and its associated uncertainties . whereas collimators onboard , , and have quite large ,
@xmath19fields of view ( fovs ) that include substantial emission from cosmic x - ray background ( cxb ) sources , the equivalent region of the bullet cluster within spans @xmath20 less solid angle on the sky . also , for clusters that fit well within s @xmath21fov , simultaneous offset regions can be used to precisely characterize the background to an extent not possible with collimated instruments .
we describe the two observations and their generic processing in section [ sec : obs ] . in section [ sec : cal ] , the modeling of the background and its systematics and the overall flux calibration are briefly described ( see appendices [ sec : appendixbgd ] and [ sec : appendixsim ] for details ) .
we examine hard band images and the character of the global spectrum in section [ sec : analy ] .
lastly , the implications of these results are discussed in section [ sec : disc ] .
we assume a flat cosmology with @xmath22 and @xmath23 km s@xmath4 mpc@xmath4 .
unless otherwise stated , all uncertainties are given at the 90% confidence level .
the bullet cluster was observed by in two epochs .
the optical axis fell near the centroid of the large - scale x - ray emission in the first observation and near the western shock driven by the bullet subcluster in the second .
the first pointing was carried out over a little under 3 days , 1820 october 2012 , for a total unfiltered exposure of 231 ks .
for the second pointing , the bullet cluster was observed for a slightly longer raw exposure of 287 ks from 14 november 2012 . to filter the events , standard pipeline processing ( heasoft v6.13 and nustardas v1.1.1 )
was applied along with stricter criteria regarding passages through the south atlantic anomaly ( saa ) and a `` tentacle''-like region of higher activity near part of the saa ; in the call to the general processing routine that creates level 2 data products , nupipeline , the following flags are included : saamode = strict and tentacle = yes .
these additional flags reduce the cleaned exposure time by @xmath24% from what it would otherwise be , but also reduce background uncertainties .
no strong fluctuations are present in light curves culled from the cleaned events , suggesting a stable background , so no further time periods were excluded . from the cleaned event files , we directly extract images like those shown in figure [ fig : rawimgs ] and light curves using xselect , create exposure maps using nuexpomap , and extract spectra and associated response matrix ( rmf ) and auxiliary response ( arf ) files using nuproducts .
the call to nuproducts includes extended = yes , most appropriate for extended sources , which weights the rmf and arf based on the distribution of events within the extraction region , assuming that to be equivalent to the true extent of the source .
although the effective smoothing of the source due to the point spread function ( psf ) is not folded in with the weighting , the relatively narrow full width at half maximum ( fwhm ) of @xmath8 18lessens the impact of this omission .
the response across a given detector is largely uniform , so the rmfs of the four detectors are simply averaged by the weighted fraction each detector contributes to a region .
in addition to the mirror response , the arf includes low energy absorption in the detectors ( due to a czt dead layer and platinum electrodes ) and is also `` corrected '' to a canonical power law crab spectrum of photon index 2.1 and normalization 9.7 photons s@xmath4
@xmath5 kev@xmath4 at 1 kev .
the remaining products necessary to analyze the spectra background spectra and a psf - corrected flux calibration are tailored for this analysis and described in section [ sec : cal ] .
ccccccc 700055002 & 104.63207 & -55.924552 & 231 & 126 + 700056002 & 104.53211 & -55.919636 & 287 & 140 events are displayed grey or black , respectively .
top row : obsid 700055002 images ; bottom row : obsid 700056002 images .
the left and right columns show the data from the a and b telescopes , respectively .
[ fig : rawimgs],title="fig:",width=188 ] events are displayed grey or black , respectively .
top row : obsid 700055002 images ; bottom row : obsid 700056002 images .
the left and right columns show the data from the a and b telescopes , respectively .
[ fig : rawimgs],title="fig:",width=188 ] events are displayed grey or black , respectively .
top row : obsid 700055002 images ; bottom row : obsid 700056002 images .
the left and right columns show the data from the a and b telescopes , respectively .
[ fig : rawimgs],title="fig:",width=188 ] events are displayed grey or black , respectively .
top row : obsid 700055002 images ; bottom row : obsid 700056002 images .
the left and right columns show the data from the a and b telescopes , respectively .
[ fig : rawimgs],title="fig:",width=188 ]
one of s pioneering technologies , at least for an astrophysics x - ray mission , is the separation of its optics and focal plane modules by an open mast structure that was extended after launch .
the telescope is thus open and subject to stray light , which dominates the background at low energies and creates a spatial gradient across the fov .
the stray light must be distinguished from the instrumental background , which varies from detector to detector but is otherwise spatially uniform , in order to use local background regions for any source region .
also , because the psf scatters some emission outside our extraction region , we must estimate the fraction of the emission collected within the region by convolving the cluster s true spatial distribution with the psf , which varies with off - axis angle .
our solution to these challenges is outlined below . as is typical
, the background has both intrinsic and extrinsic components , which for vary in relative importance both spectrally , spatially , and somewhat temporally . for faint sources where the background is a significant fraction of the source counts , it is to some degree inappropriate to naively extract and rescale a spectrum from elsewhere in the fov to use as a background .
however , because the background components are reasonably well understood and stable , we can model its instantaneous composition from source - free regions and , using what we know about the spatial variations of each component , extrapolate that model to the source region . the physical origin of the background components are briefly described below ; for details on the specific models and how the background is actually fit with them , see appendix [ sec : appendixbgd ] .
these components are all identified in the spectra shown in figure [ fig : app : stack ] , and it may benefit the reader to refer to it and the following section simultaneously . *
internal * : the radiation environment of s orbit leads to a roughly flat background across all channels .
an underlying featureless continuum is produced primarily , but probably not entirely , by high energy gamma rays that either pass through the anti - coincidence shield and compton scatter in the detector or scatter untriggered in the shield itself .
the remainder of the internal background consists of various activation and fluorescence lines , which are mostly resolved and only dominate the background between 22 - 32 kev . above these energies
weaker lines are still present , but the continuum dominates . more details can be found in appendix [ sec : appendixbgd : overview ] * aperture stray light * : because the space between the optics and focal plane benches is not fully baffled , a series of aperture stops protrude from the focal plane bench to block unfocused x - rays from striking the detectors ( for a diagram of this geometry , see figure [ fig : app : geometry ] ) . due to technical implementation limitations
, the aperture stop does not exclude 100% of the stray light , leaving a few degree window centered on each mirror module .
although the optics bench itself blocks much of the fov , there remain lines of sight connecting every detector pixel , through the aperture stop , to regions of open sky .
the amount of sky visible to any given pixel is location - dependent . since the cxb is roughly uniform on large scales , the stray light from the cxb through the aperture stop ( hereafter called the aperture " background ) produces a smooth gradient across the detector plane that depends on the orientation of the detectors and the apparent position of the optics module .
the cxb spectral shape is consistent with that found by previous missions , and we adopt the canonical _
heao-1 a2 _ spectral model , valid from 360 kev @xcite .
due to cosmic variance , the precise normalization for any given observation should be measured intrinsically ( see appendix [ sec : appendixsim ] for details ) .
* reflected and scattered stray light * : besides direct exposure to sources of stray light , the open geometry of the spacecraft is susceptible to reflected and scattered x - rays from the entire sky .
one possible reflecting surface along with many other parts of the observatory , including the mast is the backside of the aperture stops , which are clearly visible to the detectors .
there are three potential sources of reflected emission : the cxb , the earth , and the sun .
because such a large fraction of the sky is visible to the backside of the aperture stops , they are capable of reflecting a contribution of 10 - 20% of the total unfocused ( i.e. , aperture " ) cxb emission despite their smaller solid angle and low reflectivity . assuming the spectrum is unchanged and uniformly illuminates the detectors , this extra emission simply adds to that coming through the aperture stops .
emission from the sun ( solar " ) , and potentially the earth s albedo , is much softer and also much more variable . during episodes of high solar activity ,
the background below @xmath25 56 kev will be dominated by a @xmath26 kev thermal spectrum of solar abundance , but even during less active periods this component accounts for @xmath27% of the @xmath28 kev total .
the `` solar '' emission is only present when the satellite is illuminated by the sun , so there is no doubt as to its origin .
there are also some weak fluorescence lines from material elsewhere on the spacecraft , such as the mast , that contribute to the background , although their origin and contribution is still under active investigation .
* focused cosmic background * : unlike the above components , there always exists an inherent `` background '' from other unresolved foreground / background sources within the fov that are not of primary scientific interest . while subdominant at all energies , the focused cxb ( fcxb " ) contributes noticeably below 15 kev having roughly 10% the flux of the aperture " cxb with a slightly softer spectrum than the aperture " cxb since it has been modulated by the mirror effective area , which begins to decline above 10 kev .
although we directly measure the contribution of each component , we do not do so with infinite precision or accuracy .
inaccurately estimated systematic offsets can easily lead to `` detections , ''
especially when the associated precision of a component is overestimated .
faint spectral components , such as ic emission in galaxy clusters , fall into in this category since they tend to reside in background - dominated regimes .
therefore , we must have some sense of the systematic uncertainty intrinsic to the background , and as much as possible to each component of the background . for some components , like the internal background , the systematic uncertainty could in theory be arbitrarily close to 0% . in practice , of course , uncertainties of less than a few percent are difficult to achieve
. components with a cosmic origin , however , have systematic uncertainty floors due to their very natures . while these uncertainties are sometimes large , they may also be well known , as in the case of the cxb . at higher ( @xmath29 kev ) energies , where the internal background strongly dominates , performing the background fitting procedure outlined in appendix [ sec : appendixsim ] on the first - pass ecdfs survey fields results in an accurate reconstruction of the background level with a standard deviation of @xmath30% after accounting for the effect of statistical fluctuations .
although the real uncertainty may be smaller , the large statistical uncertainties due to the shorter exposure time ( @xmath27 ks / field ) make it difficult to surmise with greater precision .
we adopt a conservative uncertainty of 3% for the entire energy range . because much of this regime is dominated by lines , whose normalizations have independent systematic uncertainties that are dwarfed by their statistical uncertainties , a global shift up or down maximizes this background s impact on fits to the cluster spectrum .
the shape of the cxb spectrum has been well - measured by other missions ( e.g. , * ? ? ?
* ) , and although it may vary on small scales , the larger scales relevant to are unlikely to exhibit noticeable deviations from the average spectrum .
the overall normalization , however , depends critically on the number of more rare , brighter sources , which varies from one location to another on the sky . because we have no way to exclude the brightest sources , even the variance on large scales ( 0.310 deg@xmath31/pixel over a total solid angle of 37.2 deg@xmath31 )
can be high .
we can eliminate much of this uncertainty by directly measuring it in the non - source regions of an observation .
this technique is especially powerful thanks to the strong correlation of the cxb normalization between source and non - source regions .
each cxb point source produces an aperture - shaped ( circular aperture stop opening modulated by any fraction blocked by the optics bench ) `` plateau '' of emission across the detectors , so many pixels see " the same sources seen by other pixels , especially those nearby
. however , background and source regions will not contain all the same cxb sources , so a residual uncertainty remains .
based on simulations of the @xmath32@xmath33 from @xcite , we find that for the approximate location of the cluster on the detectors the residual systematic for the aperture cxb is 8% ( @xmath34 )
. in principle , scattered and reflected x - rays ( contributing at the lowest energies ) should be nearly perfectly correlated between all pixels , even if their spatial distribution is not necessarily uniform . because we do not know exactly where the scattering is taking place , we can not predict the appearance of this emission like we can for the aperture " background .
it does not appear to be flat ; independent fits of spectra from the various detectors give different normalizations .
unfortunately , it is not yet feasible to empirically determine the shape any more finely than that at this time . based on the same exercise used to constrain the internal value ,
we find a systematic uncertainty for the solar " component of 10% ( @xmath34 ) . for the fcxb " emission , we can apply a straightforward shorthand estimate of cosmic variance , consistent with the method used for the aperture " component but based on an empirical estimate of the variance .
we assume a conservative point source detection threshold of @xmath35 ergs s@xmath4 @xmath5 ( 2030 kev ) , below which individual sources would not be obvious embedded within the cluster emission ( see section [ sec : analy : imgs ] for details ) .
the variance scales as @xmath36 , where @xmath37 is the solid angle on the sky and @xmath38 is the flux limit for excised point sources . for our elliptical source region ,
shown in figure [ fig : imgs ] , @xmath39 arcmin@xmath31 .
we can estimate the variance in our observation relative to another measurement assuming a @xmath32@xmath33 relation of @xmath40 . using the _
heao-1 a2 _ estimate @xcite with @xmath41 , @xmath38(2030 kev ) @xmath42 erg s@xmath4 @xmath5 , and @xmath43 ( 1@xmath44 )
, we find a variance and thus systematic uncertainty of @xmath45% ( @xmath34 ) for our extraction region .
since we want the total cluster flux , to first order we could simply use as large a region as possible and assume that includes all the emission .
however , the psf wings cause a fraction of the flux to get redistributed far from its true origin on the sky , which results in some emission being scattered beyond the fov as defined by the detectors .
detector gaps also miss flux , and one just happens to fall across the brightest part of the bullet during the second observation .
these effects require careful correction so that the exposure across the field is accurate . as mentioned in section [ sec : obs ] , the arf for an extended source is created by averaging the vignetting function across the region , weighted by the distribution of events .
extended source arfs are not additionally corrected for any source emission scattered out of the region through the wings of the psf . to get a proper total flux for the bullet cluster spectrum
, we must take the psf and estimate the fraction of the total emission captured inside the region .
this task is not entirely trivial since not only does the psf shape vary with off - axis angle , but the off - axis angle varies for any given position on the sky over the course of an observation
. normally one could neglect the variation in shape , as it only becomes a measurable effect for large ( @xmath46 ) off - axis angles .
because the placement of the cluster in the second observation results in large off - axis angles for its eastern parts , we include these minor adjustments to the psf shape .
following @xcite , we can construct composite or effective psfs for our particular observations across the entire cluster , so that each position has an appropriate psf associated with it .
now armed with a set of position - dependent psfs ( but not energy - dependent ) , the flux in the wings can be directly computed .
note that the psf varies weakly as a function of energy ; below @xmath47 kev , the fwhm is up to 10% broader than it is at higher energies , although the encircled energy fraction within a radius of @xmath48 agrees to within a few percent at all energies .
the latter behavior justifies our use of an energy - independent psf .
ideally , we would like to take the true flux distribution from the cluster and convolve it with the psfs to estimate the redistributed fraction , but above @xmath8 7 kev is the only telescope capable of making a reliable image . to estimate the fraction of the total flux in the 320 kev energy range within our spectral extraction region , we generate psfs in a 25@xmath4928 grid
each position separated by 1 fwhm of 18 and roughly fit them to the a and b telescope images .
the extraction region encompasses 95% of the intrinsic flux from the cluster , and a net @xmath50% of that is scattered out of the region by the psf .
thus , in terms of total cluster emission , our spectrum captures @xmath51% of the total 320 kev flux . when comparing to past observations , our quoted model normalizations and fluxes
would then be 10% lower ; however , the overall effective area given in the calibration used here is @xmath52% lower than that needed to match with xrt and epic fluxes , which means our fluxes should also be decreased 15% ( this adjustment is present in later caldb releases ) .
since these corrections roughly cancel out , and given the uncertain nature of absolute calibration between telescopes , we do not further adjust the normalizations and fluxes derived from model fits to the spectra .
although the goal of this paper is to determine the character of the hardest emission in the cluster , we must first confirm that no reasonably bright point sources contaminate that emission . unlike all previous observatories , s
unprecedented spatial resolution at hard energies makes a task heretofore impossible as simple as examining the images .
the pipeline - filtered event files are sufficiently processed to produce images , which can be done in arbitrary energy bands by further filtering on the pha column in , e.g. , xselect .
however , calibrated images also require exposure - correction and background - subtraction ; the necessary images are generated from nuexpomap and nuskybgd , respectively .
the latter is not part of the software distribution , but was developed independently as part of this work .
( see appendices [ sec : appendixbgd ] and [ sec : appendixsim ] ) .
we create exposure maps at single energies for each band , which roughly correspond to the mean emission - weighted energy of the band .
to mosaic the two epochs along with the data from both telescopes , we also need to correct for offsets due to the @xmath50uncertainty in the reconstructed astrometry .
no obvious point sources appear within the fov , so we estimate the necessary shifts using the global distribution of the cluster emission and find slight offsets of 0 to 3 pixels relative to the first epoch s a telescope astrometry . because the 2.46pixels significantly oversample the psf , the final images are smoothed by 5 pixels , more consistent with the psf s fwhm of @xmath53 .
images in four energy bands ( top : 38 kev , 815 kev ; bottom : 1530 kev , 30 - 40 kev ) are presented in figure [ fig : imgs ] .
the white ellipse shows the extraction region for spectra discussed in section [ sec : analy : spec ] . from 38 kev ,
the cluster resembles the or images blurred by the larger psf , except that the `` bullet '' to the west is relatively de - emphasized since it is composed of cooler ( @xmath54 kev ) gas than is the main subcluster ( @xmath55 kev ) and s response is more sensitive to harder emission in this band . above 8 kev ,
the `` bullet '' essentially disappears , although the halo of shocked gas surrounding it is clearly visible .
the cluster begins to approach the level of the background above 15 kev , and above 30 kev whatever detectable emission remains is highly background - dominated . while the overall morphology changes slightly with energy a subject of a future paper it does not deviate appreciably from what one would expect extrapolating a temperature map measured at energies @xmath56 kev , suggesting the origin of the @xmath57 kev emission is also mostly , if not entirely , thermal as well . most critically , there is no indication of a background agn whose emission could masquerade as the non - thermal emission we are searching for .
the bullet cluster is generally free of bright point sources ; the contribution of obvious point source emission in the 0.84 kev band from a 0.5 ms mosaic ( courtesy m. markevitch ) is @xmath58% of the total cluster emission , or a flux of roughly @xmath59 ergs s@xmath4 @xmath5 . of course
, considering these sources alone does not protect us from contamination by absorbed or very hard sources .
while there are no bright point sources in our images , s large psf makes it more difficult to distinguish point sources embedded within the diffuse cluster emission .
we can estimate the approximate brightness of point sources that we would be able to identify visually by adding a fake source to the data and noting the flux above which the source becomes readily apparent .
the resulting flux limit is likely higher than what might be achieved with wavdetect or some other point source identification method , but relying on simple visual inspection of images is straightforward and sufficient for these purposes . in images covering the entire relevant energy band ( 340 kev ) ,
point sources would be clearly identified within a radius of @xmath26 if they are @xmath60% of the total cluster flux and @xmath19% outside this radius , corresponding to flux limits of 29 @xmath61 ergs s@xmath4 @xmath5 . at higher energies ,
the signal - to - noise rapidly declines ; in the 2030 kev band point sources only become obvious when they have @xmath62% of the cluster emission at those energies , or @xmath63 ergs s@xmath4 @xmath5 .
these limits are roughly comparable , with the 5% limit above translating to a flux in the 2030 kev band of @xmath64 ergs s@xmath4 @xmath5 , assuming a power law spectrum with a typical photon index of 1.8 .
note that the entire fov of is at least a factor of two smaller than the effective psf of the hxd - pin and bat instruments , further reducing the comparative chance of a point source contaminating the hard x - ray spectrum .
figure [ fig : specraw ] displays the four raw spectra extracted with nuproducts from the elliptical region illustrated in figure [ fig : imgs ]
. the consistency between the four spectra demonstrate the very similar effective areas between the two telescopes , the shallow vignetting function below 20 kev ( the primary difference between the two epochs is the off - axis angle of the cluster centroid ) , and the stability of the background ( whose dominance coincides with the appearance of strong lines just above 20 kev ) .
the detection of an excess above the thermal tail clearly depends critically on the reproducibility of the background . via the procedure
discussed in detail in appendices [ sec : appendixbgd ] and [ sec : appendixsim ] , we have an empirical model for what the background emission should be in this region , based on blank field observations , which has been fit to non - source regions from these observations .
this is our best guess for the background spectrum of each observation and telescope , but it is only the most likely state of the background ; the actual background may be somewhat different given systematic and statistical uncertainties .
a proper background should mimic the statistical impact of the actual background , having both the same area and exposure time of the source region .
typical backgrounds are often taken from larger regions or longer exposures in order to minimize the impact of statistical fluctuations in background regions that could bias the background level in the source region .
however , this procedure underestimates the background - subtracted error per channel since the true background suffers from larger statistical uncertainties , which is important wherever the background dominates . one solution is to jointly fit the background and source data together , but this requires simultaneously fitting 20 spectra each with 3675 unbinnned channels , making it computationally challenging just to find a good fit let alone calculate errors on parameter values . to circumvent this difficulty
, we separate the background and source modeling phases but attempt to retain a statistically appropriate treatment of the background .
as described in appendix [ sec : appendixsim ] , a nominal background model is found for the source region for each epoch and telescope .
we then simulate , for the same exposure time , a background for the region from this model using fakeit in xspec , including poisson fluctuations .
of course , the resulting background spectra fail to incorporate any systematic offsets from the nominal model , and statistical fluctuations introduced to the spectrum have the potential to bias fit parameters as well .
although not the only path forward , we choose to simulate many realizations of the four backgrounds , fitting the spectra with each set .
this procedure naturally allows systematic uncertainties to be incorporated as well , since the several background component model normalizations can be randomly varied to reflect those uncertainties .
each background thus represents a possible version of the true background , ideally in proportion to the likelihood that it matches the true background .
a similar approach was taken to incorporate background systematic uncertainties in @xcite .
we assume gaussian fluctuations about the normalizations of each component with magnitudes given in section [ sec : cal : bgd : sys ] ; appendix [ sec : appendixsim ] outlines the specific methodology in detail . for continuum - driven fits on data
binned to just above the gaussian limit ( 25 - 30 counts / bin ) , the @xmath65 statistic is known to be biased , especially for fits using a large number of bins @xcite .
briefly , the weights @xmath66 on bins with negative fluctuations are overestimated while bins with positive fluctuations are underestimated , since @xmath67 , so the @xmath65 statistic drives the global best - fit model below the data .
the model derived from fits to the background spectra , for example , are biased by 23% when @xmath65 is used as the fit statistic . to avoid this and similar issues with fitting the bullet cluster spectra
, we use the xspec command statistic cstat , which applies the w statistic , a cash - like statistic appropriate for fits with unmodeled background spectra .
bins with no counts have a tendency to confuse the implementation of this statistic in xspec , so we also group the spectra such that there are at least 3 counts in each bin in _ both _ data and background spectra .
armed with a reliable way to deal with the background , we can confidently evaluate the nature of the hardest detectable emission from the bullet cluster .
a strong motivation for these observations was to confirm and better characterize the non - thermal component claimed in @xcite .
in clusters with radio halos and relics , such as the bullet cluster , ic emission the only diffuse interpretation for
a non - thermal tail must be present at some level .
if the ic emission is bright and begins to dominate the spectrum over the thermal emission at a low enough energy , then the spectrum will be trivial to model .
the characterization of a weaker ic component , however , depends much more on the model employed to discriminate it from the thermal emission .
of course , our spectrum falls within the latter regime .
the range of models considered is somewhat restricted , but appropriate for the data , consisting of single temperature ( 1 t ) , two temperature ( 2 t ) , and single temperature plus power law ( t+ic ) components . the thermal components are calculated using the version of the @xcite plasma code implemented in xspec .
the 1 t model provides the simplest possible description of the spectrum .
emission dominated by isothermal or nearly isothermal gas will be satisfactorily characterized with a single temperature component , since s 0.4 kev fwhm resolution does not allow us to easily separate the k@xmath68 line complexes near he - like and h - like fe at 6.7 and 6.9 kev , respectively .
therefore , we are entirely reliant on the shape of the largely featureless continuum to discern multi - temperature gas . given our broad bandpass ( 330 kev ) , the 1 t model is unlikely to account for all the truly thermal emission . from and ,
spatially resolved spectroscopy clearly demonstrates that the icm contains gas spanning a large range of temperatures @xcite , which one would expect for an ongoing merger @xcite .
we do not know the true temperature structure , however , only the emission - weighted line - of - sight projected temperature distribution , which is also folded through the effective area and is thus dependent on the calibration and energy band .
for the global spectrum , we are not particularly concerned with describing the true temperature structure , since that is not possible .
instead , we wish to accurately represent the part of the temperature distribution seen by , which is more heavily weighted toward the hotter regions and thus may not entirely agree with the projected temperature structure measured within a lower energy bandpass .
because thermal continua are fairly featureless , the 2 t model will likely encompass the full range of significant gas temperatures .
if the ic emission is sufficiently bright , however , then the higher temperature component of the 2 t model will be skewed to an unphysically high value . in this case
, the t+ic model should provide a better description of the overall spectrum .
although the thermal component would be imperfectly suited to the true thermal distribution , the harder non - thermal component would better capture the spectral shape at higher energies .
note that the statistical power resides at low energies where the majority of counts are , so the non - thermal excess at high energies must be sufficiently strong to overcome the worsening of the fit quality at the low end . for the non - thermal component of the t+ic model , we fix the power law photon index to 1.86 , the best - fit value found by @xcite .
we also allowed the index to be a free parameter , but in nearly all cases the index became steeper ( @xmath69 ) , where it was most likely mimicking the lower temperature component of the 2 t model .
this appropriating of the ic component directly results from the greater statistical power of the counts at the low end of the energy range driving the fit .
although the radio synchrotron spectrum basically agrees with this best - fit index , implying @xmath70 for the ic index @xcite , the electrons producing the radio emission are more energetic ( for @xmath71 @xmath7 g , @xmath72 23,000 where @xmath73 is the `` relativistic gamma '' of the electron ) than the ones producing the ic ( @xmath74 at 30 kev ) , so there is no guarantee the photon index would directly follow , and there is good reason to assume the index flattens at lower energies as is seen in , e.g. , the coma cluster @xcite . for comparison purposes and for our primary result , we choose to fix @xmath75 , which is similar to the typically assumed value of @xmath76 in any case .
typical fits to the four spectra , using a background spectrum generated from the nominal background model , illustrate the subtle differences between the 1 t , 2 t , and t+ic descriptions of the bullet cluster s spectrum in figures [ fig : specallmodels ] and [ fig : specratio ] .
( note : all uncertainties quoted in this subsection are purely statistical and are derived using the the nominal backgrounds displayed in the above figures . ) in each of these figures , the data , backgrounds , and models for the four spectra have been grouped together for clarity , although the models are folded through each response separately during the fit .
the background is also shown to highlight where the spectrum becomes dominated by the background . to zeroth order ,
the 1 t fit is quite good , with a typical @xmath77 .
the global temperature of @xmath78 kev agrees quite well with the acceptable best - fit temperature range of 13.6 kev @xmath79 kev @xcite , which varies depending on the value of @xmath80 used .
although higher than the best - fit global temperature of @xmath81 kev @xcite , we would expect the average temperature in the 330 kev band to be slightly higher than measured in the 110 kev band .
the and temperature disagreement almost certainly comes down to their respective calibrations , e.g. , @xcite . despite the somewhat coarse spectral resolution around the fe lines ,
the large effective area and exposure time allows the abundance to be well constrained at @xmath82 of solar , consistent with those determined from previous observatories , such as ( @xmath83 , * ? ? ?
although a 1 t model can largely explain the detected emission , a very slight curvature in the residuals of the fit indicates that the spectrum is not of a truly isothermal plasma .
because our sensitivity extends up to higher energies , we can test whether that extra curvature is more likely to come from the true multi - temperature structure of the cluster or an ic component .
the 2 t model approximates what is actually a fairly smooth , somewhat bimodal , temperature distribution ( e.g. , * ? ? ?
* ) , so the best fit thermal components in this model only roughly correspond to the actual temperatures .
even so , the temperatures we find for the two components are reasonable , with @xmath84 kev and @xmath85 kev
. figure [ fig : specallmodels ] shows the relative importance of the fainter component , with the lower temperature accounting for only @xmath50% of the 330 kev flux .
the hard spectral tail up to 30 kev is fully consistent with a thermal spectrum of @xmath86 kev , only a little higher than the ambient , non-``bullet '' icm temperature of @xmath87 kev seen with @xcite and the 14.2 kev temperature found here with the 1 t model . given that at a minimum there is recently shocked gas at much higher temperatures @xcite , a rise in @xmath88 of this magnitude is not surprising . also , the range in temperatures for @xmath89 agrees very well with temperatures common in both the `` bullet '' region and nearby @xcite .
lastly , we evaluate the likelihood of an ic excess at high energies with the t+ic model .
the near success of the 1 t model suggests that if a detectable ic component lies at harder energies , the thermal emission should be well accounted for by a single temperature component .
we again find a very reasonable temperature of @xmath90 kev , and the resulting 50100 kev ic flux is @xmath91 ergs s@xmath4 cm@xmath31 .
again , the thermal component is entirely consistent with that found with previous observatories .
the ic flux , on the other hand , falls nearly a factor of 3 below the expected value of @xmath93 ergs s@xmath4 cm@xmath31 @xcite .
unfortunately , the best - fit ic component only surpasses the thermal component at @xmath94 kev , where the cluster emission becomes so faint it is lost in statistical fluctuations of the background . to first order
, the spectrum appears equally well - fit by the addition of a non - thermal model as by the addition of another thermal model , and the statistical significance of the ic component is possibly high enough to warrant a detection .
the inclusion of systematic uncertainties and a detailed comparison of the 2 t and t+ic fitting results outlined in the following subsection , however , preclude us from making such a claim .
the mean parameter values and statistical errors were reported in section [ sec : analy : spec : fits ] ; however , true uncertainty ranges must include the impact of both statistical and systematic fluctuations in the background on the fits .
the distribution of best - fit temperatures for the three models found using 1000 realizations of the background for our 4 spectra are shown in figure [ fig : ktdist ] , immediately illustrating the impact of both background uncertainties and the model we choose to use on our ability to evaluate the spectrum . when the shape of the model is determined by one parameter , as in the 1 t case , background uncertainties have only a slight effect on the temperature , creating a spread of only 0.18 kev ( compared to the statistical uncertainty of @xmath95 kev ) . adding another parameter that more finely controls the broadband shape of the model ( 2 t or t+ic cases ) allows background fluctuations to play a more significant role . for the 2 t model , the best - fit temperatures for each component shown in green in the main panel and inset panel of figure [ fig : ktdist ] are much more sensitive to background fluctuations than in either the 1 t or t+ic cases , mostly owing to the greater flexibility of the model to adjust to small changes in the shape of the spectrum .
background variations primarily affect the @xmath88 component , since a slightly higher / lower background will cause the spectrum to turn over at a lower / higher energy , thus pushing @xmath88 to lower / higher values .
the @xmath89 component then adjusts to `` correct '' the low energy part of the spectrum ; the two temperatures are strongly correlated for a given fit , such that a higher than typical @xmath88 will have a higher than typical @xmath89 . in the t+ic model , the temperature component dominates at all relevant energies and thus maintains the precision of the 1 t model s temperature ( despite having a larger statistical error of 0.35 kev )
. the ic flux , in principle , should be much more sensitive to background systematics than to statistical uncertainties , for the simple reason that its shape more closely matches the background and any systematic shift up or down of the background will correspondingly shift the ic normalization .
tellingly , the uncertainty due to the background on the ic flux ( @xmath96 ergs s@xmath4 cm@xmath31 , 50100 kev ) is slightly less than its statistical uncertainty ( @xmath97 ergs s@xmath4 cm@xmath31 ) .
a true non - thermal excess at high energies should be more affected by background fluctuations .
since the ic flux is not , it is likely driven more by `` correcting '' ( as in the 2 t case ) the model shape at lower energies , where the signal - to - noise is higher .
the ic component is not accounting for truly non - thermal flux in these fits ; instead , it substitutes for additional thermal components missing from the single temperature model , likely at both the high _ and _ low energy ends of the spectrum . allowing the ic photon index to be
a free parameter further confirms this explanation . for the nominal background , the index steepens to @xmath98 and mirrors the contribution of the @xmath89 component of the 2 t model at low energies , where its continuum shape is nearly identical to that of a @xmath50 kev plasma .
as shown by the cyan histograms in figure [ fig : ktdist ] , the temperature of the thermal component in this t+ic model is actually hotter than for the 1 t case .
the hard emission is modeled entirely by the thermal component , while the ic appears to be mimicking the @xmath89 component .
this argument alone does not invalidate the ic hypothesis , since in principle the gas could be sufficiently isothermal to allow ic emission to be contributing excess flux at lower energies where the component is mostly being constrained , and it is only coincidental that the spectrum runs out of counts just when the ic component begins to dominate the hard emission . assuming this viewpoint
, a combination of the statistical and systematic uncertainties gives a most likely ic flux of @xmath99 ergs s@xmath4 cm@xmath31 , or a positive fluctuation of less than 2@xmath44 .
considering we have strong reason to believe this component is thermal in nature , it is clear we do not detect ic emission in the global bullet cluster spectrum with .
the results for each model , including statistical and systematic uncertainties , are summarized in table [ tab : fit ] .
even so , the t+ic model may fit the spectrum better than the 2 t model , in which case we might still argue that the spectrum shows evidence of an ic component .
the relative quality of the t+ic versus 2 t fits depends on the background realization being used , and figure [ fig : delcstat ] demonstrates that certain backgrounds do in fact favor the t+ic over the 2 t model . in this figure , the fits with the 1000 background realizations have been binned according to the difference in c - statistic values between these two models , with values to the right of the vertical lines favoring the 2 t model and values to their left favoring the t+ic model . the solid histogram / vertical line correspond to fits with @xmath100 fixed to a value of 1.86 and the dashed versions to fits with @xmath100 as a free parameter . in the majority of background realizations ,
the 2 t model is preferred , and in only 1.2% of them can the same be said for the t+ic model where @xmath100 is fixed . the t+ic model is favored 7.6% of the time when @xmath100 is free , although in this case the ic component may simply be mimicking a second thermal component .
so while it is most likely the case that the spectrum can best be characterized with a pure thermal model , we can not rule out an ic flux within the range of fluxes in the right panel of figure [ fig : ktdist ] .
based on this analysis , a fair 90% upper limit on the ic flux should correctly incorporate both the systematic and statistical uncertainties already discussed . to capture the fact that the t+ic fits prefer a non - zero ic flux , we sum the mean flux with the quadrature - summed uncertainties , yielding an upper limit of @xmath3 ergs s@xmath4 cm@xmath31 in the 50100 kev band .
lcccccc 1 t & @xmath101 & @xmath102 & @xmath103 & - & - & @xmath104 + 2 t & @xmath105 & @xmath106 & @xmath107 & @xmath108 & @xmath109 & @xmath110 + t+ic & @xmath111 & @xmath112 & @xmath113 & 1.86(fixed ) & @xmath114 & @xmath115 + t+ic & @xmath116 & @xmath117 & @xmath118 & @xmath119 & @xmath120 & @xmath121
the bullet cluster was observed by in two epochs for a cumulative 266 ks of conservatively - cleaned exposure time .
the cluster is clearly detected below @xmath122 kev with an energy - dependent morphology consistent with the extrapolation of projected temperature maps obtained with and . above @xmath122 kev , potential emission associated with the icm consists of @xmath123% of the counts per channel .
the average temperature of the global spectrum is @xmath124 kev , in good agreement with estimates from @xmath125(14.5 kev , * ? ? ? * ) and ( 14.8 kev , * ? ? ?
* ) , but somewhat higher than independent estimates from and ( @xmath126 kev , * ? ? ?
given the differences between instrument sensitivity and the accuracy of their respective calibrations , we do not suggest any significant discrepancy . in order to search for a non - thermal excess above the thermal emission at hard energies
, we invested a good deal of effort to understand the largest uncertain factor : the background .
we constructed an empirical , spatial - spectral model of the background from blank sky data and applied it to our observations to derive a `` most likely '' model background spectrum for the region containing cluster emission . after evaluating the important systematic uncertainties in the model , 1000 realizations of the background
are generated and each subtracted from the spectrum , which is fit with three spectral models representing a simple ( 1 t ) or more realistic ( 2 t ) thermal - only origin , or a significant ic component at the highest detectable energies ( t+ic ) , for the emission . in over 98% of the fits , the 2 t model was statistically favored over the t+ic model , and reasonable values are obtained for both temperatures in the former .
we therefore conclude that no significant non - thermal emission has been detected in the observations of the bullet cluster and place an upper limit on the ic flux of @xmath3 ergs s@xmath4 @xmath5 ( 50100 kev ) .
this flux falls below that reported by and . as mentioned in section [ sec : intro ] , @xcite first suggested the existence of significant ic emission at hard energies in the bullet cluster based on a joint analysis of and spectra .
the uncertainty in the measurement of @xmath127 ergs s@xmath4 @xmath5 ( 50100 kev ) is too large to justify a claim of detection .
however , a more recent analysis @xcite , using a bat spectrum found a flux of @xmath128 ergs s@xmath4 @xmath5 ( 50100 kev ) , roughly consistent with that from @xcite .
both fluxes are only barely in conflict with our conservative upper limit , but our most likely ic flux of @xmath129 ergs s@xmath4 @xmath5 ( 50100 kev ) is clearly inconsistent with these previous measurements .
the origin of the discrepancy has two potential explanations : either the spectra from the various instruments disagree ; or the approach to modeling the spectra disagree .
while even minor calibration differences between the characterization of the telescope responses and of the backgrounds can significantly affect results , a comparison of the , , and spectra fit to 1 t or 2 t models implies these are not responsible .
none of the instruments on these satellites reliably detect emission above 30 kev from the bullet , and below this energy there is no compelling excess above a reasonable thermal - only model in figure 2 of @xcite , the lower left panel of figure 5 of @xcite , or figure [ fig : specallmodels ] of this paper . at higher energies
, the background dominates the count rate and its treatment becomes crucial , where even small fluctuations can result in a false ic signal .
it is beyond the scope of this paper to evaluate the backgrounds from the other two missions , but no causes for worry are evident in the analyses of the and data .
if the spectra are all consistent with each other , we must attribute the conflicting conclusions to differences in how the spectra are modeled . in principle
there should be no difference , since 1 t , 2 t , and t+ic models are each tried in all three analyses . the crucial distinction between them is the minimum energy used in the fits : 1 kev @xcite , 0.5 kev @xcite , or 3 kev ( this work ) .
the lower end of the energy range matters because the thermal gas of the bullet cluster is decidedly _ not _ isothermal @xcite , and the fraction of the emission any temperature component contributes strongly varies with energy , with low temperature components dominating at soft energies but essentially disappearing from the hard band .
merging clusters , especially those like the bullet where one subcluster hosts a cool core , may have components of roughly equal emission measure that span a factor of two in temperature .
in particular , the emission coming from the cool core ranges from @xmath130 kev up to 7 kev , has a higher abundance , and mostly contributes at the lowest energies .
the gas associated with the main subcluster is hotter , with a central @xmath131 kev and shocked regions to the w and also to the slight se with @xmath132 kev ( m. markevitch , priv .
comm . ) . given the extreme range in temperatures , even a 2 t model may provide an insufficient description of the data over a broad energy range .
ironically , the t@xmath125ic model might better fit the _
purely thermal _ emission more successfully in this case , since a power law with free photon index is able to simultaneously account for emission from components at either extreme of the temperature distribution ( e.g. , a3112 , * ? ? ?
* ; * ? ? ?
* ) . by including data below 3 kev in order to better constrain the thermal component , in all likelihood
the larger consequence is to bias the characterization of the thermal component , since only simple spectral models are considered . because the response of s epic instruments peaks between 12 kev and shot noise , which has a fractional error decreasing with energy , sets the signal - to - noise ratio , fit minimization routines are overly biased to find good fits at these lower energies . the second model in the multi - component fits of @xcite , from this perspective ,
are focused on artificially `` fixing '' the residuals below 1 or 2 kev with either the second temperature or ic component , and the slope of the ic s photon index is determined mostly by the data alone , given that the t@xmath125ic model over - predicts almost every bat data point .
this explanation is less compelling for the @xmath125analysis of @xcite . in this case , the fact that fits to both the ( over 110 kev ) and ( over 330 kev ) yield the same temperature despite the different energy bands is worrisome ; given the multi - temperature structure , one would expect the 310 kev temperature from to be hotter than this average , and the 310 kev temperature from to be cooler or unchanged .
in contrast , the temperatures in our 2 t model roughly agree with the approximately bimodal temperature distribution seen with , lending credence to the still imperfect thermal model approximated with only two components .
the much improved spectral resolution of over that of and undoubtedly helps the fit find physical temperatures . for the t@xmath125ic model ,
when the photon index is left free , it tends toward a somewhat larger or steeper value where it only influences the lowest energy channels .
the ic component , when exhibiting this behavior , mimics a lower temperature thermal component more than it tries to account for any excess emission at high energies , further refuting the existence of a significant non - thermal excess . by combining the synchrotron spectrum at radio frequencies with an ic estimate or upper limit
, we can directly constrain the volume averaged magnetic field strength . following the arguments and expression for @xmath9 in equation 14 of @xcite
, we use the total radio halo flux of 78 mjy at 1300 mhz and a radio spectral index of 1.21.4 @xcite .
the radio spectrum exhibits no flattening at lower frequencies as in @xcite for the coma cluster , so we assume the spectrum continues as a power law to lower frequencies where the electron population producing the synchrotron is the same as those producing the ic .
the upper limit on ic emission translates to a lower limit on the magnetic field strength of @xmath6 @xmath7 g , which is comparable to values found in other clusters using and data ( e.g. , * ? ? ?
* ; * ? ? ?
unlike estimates of @xmath1330.2 @xmath7 g , such lower limits are more consistent with equipartition estimates ( @xmath8 1 @xmath7 g for the bullet cluster , * ? ? ?
* ) and faraday rotation measure estimates in other clusters , which typically place the field strength at a few @xmath7 g ( e.g. , * ? ?
* ; * ? ? ?
* ; * ? ? ?
while it is possible to reconcile these estimates with a lower volume averaged value of @xmath9 , our lower limit does not requires it . in order to detect diffuse , faint ic emission in galaxy clusters
, the ic signal must be teased from both thermal and instrumental `` backgrounds , '' both of which are likely to be brighter than the ic emission itself .
while going to harder energies reduces contaminating emission from the thermal gas , it requires a large effective area at high energies and/or low and well - characterized instrumental and/or cosmic backgrounds . regarding the background , focusing optics like those onboard have clear advantages over non - focusing ones , such as collimators and coded - mask telescopes . the effective area or equivalent sensitivity , however , remains a greater challenge for reflective optics due to the large number and thus weight of mirror shells needed .
ic photon intensity also declines rapidly with energy , making it exceedingly difficult to detect such emission at high energies given the statistical fluctuations of a realistic background level without a very large effective area .
in the foreseeable future , ic emission in hot clusters will only be detectable as a subtle inflection of the thermal tail .
such non - thermal inflections , however , are complicated by having plausible alternative origins , such as background agn , clumps of super hot gas , and slightly underestimated overall backgrounds .
these difficulties , combined with magnetic field equipartition estimates nearly an order of magnitude larger than the field strengths inferred by ic measurements , emphasize the need for a conservative approach .
the recent history of ic searches seems to justify this view .
@xcite nicely summarizes some , , , and detections and upper limits in their figure 10 , which shows that clusters may exhibit an ic signal in the dataset of one observatory but not another sometimes , but not often , contradictorily .
the reasons behind these differences are not always clear , but likely include some combination of relative instrumental calibration , background treatment , and telescope capabilities .
detections are only mildly statistically significant and are in danger of being compromised by the complications mentioned above .
the clusters expected to host ic - producing electrons are those undergoing mergers , which produce possibly extreme
multi - temperature distributions .
such distributions should in principle be straightforward to separate from a non - thermal component , _ if _ the ic component begins to dominate the spectrum at an energy where the signal - to - noise is sufficiently high , including systematic uncertainties .
for the bullet cluster , we reach this point around 2030 kev .
the next mission capable of detecting ic emission associated with radio halos is _ astro - h _ , which will include a hard x - ray telescope ( hxt ) and imager ( hxi ) , with a sensitivity similar to , as well as substantial soft x - ray capabilities with the soft x - ray imager ( sxi ) and x - ray calorimeter spectrometer ( xcs ) .
although the hxi alone provides for no improvement over , the sxi and especially the xcs should allow for a more detailed and complete accounting of the thermal components of target clusters through emission line diagnostics .
a better understanding of the thermal continuum will make marginal non - thermal - like excesses at hard energies more significant and upper limits more constraining .
if the average magnetic field strength in galaxy clusters hosting radio halos is typically closer to @xmath26 @xmath7 g than the @xmath134 @xmath7 g implied by past detections , even _
astro - h _ is unlikely to be enough of a technical advance . because the ratio of synchrotron to ic flux scales with the energy density of the of the magnetic field ( @xmath135 ) , a @xmath136 stronger @xmath9 requires a @xmath137 more sensitive telescope than currently exists .
ic emission at this level would only compete with the thermal emission of a bullet - like cluster between 3050 kev , and given how faint the cluster is at these energies relative to the background ( e.g. , figures [ fig : specallmodels ] and [ fig : specsig ] ) , it is likely that most of the sensitivity gain will come from increasing the effective area .
an increase in effective area over of not quite an order of magnitude would be achieved by the proposed probe class _ hex - p _ mission , so a substantial decrease in background and its systematic uncertainty would still be necessary . in terms of past ic detections
, it may be the case that what has been measured is not ic emission associated with large scale radio halos . instead of being associated with the electrons producing radio halos and relics ,
the ic emission might originate from electrons accelerated by accretion shocks at the virial radius ( e.g. , * ? ? ?
* ; * ? ? ?
* ) . non - imaging telescopes unlike would pick up this emission , which peaks in surface brightness @xmath138 mpc from cluster centers . given our restricted extraction region around the bullet cluster , we are not sensitive to these electrons .
however , the fov does partially include the virial region , where we characterized the background , so in principle this ic emission could exist at very faint levels ; a cursory check for a non - thermal component was made when the background was fit , but no such signal beyond the generic background model was apparent .
note that these observations are not ideally suited for searches of this emission , which would be better served by several offset pointings around the periphery of the cluster .
even so , the emission would be strongest at the low energy end , where we attribute extra flux detected in the background regions to scattered thermal photons .
it should be feasible to constrain these models , but only after a more detailed accounting of the bullet cluster s thermal structure has been undertaken , in order to separate local emission from scattered photons from various regions in the cluster .
we will address this issue in a future paper focussed on the hard x - ray weighted temperature structure , including extreme temperature shock regions .
this research was supported by an appointment to the nasa postdoctoral program at the goddard space flight center , administered by oak ridge associated universities through a contract with the national aeronautics and space administration ( nasa ) and made use of data from the mission , a project led by the california institute of technology , managed by the jet propulsion laboratory , and funded by nasa .
we thank the operations , software and calibration teams for support with the execution and analysis of these observations .
this research has made use of the data analysis software ( nustardas ) jointly developed by the asi science data center ( asdc , italy ) and the california institute of technology ( usa ) .
the authors wish to thank maxim markevitch for providing a 0.5 ms image of the bullet cluster to confirm the cluster region lacks bright point sources .
46 natexlab#1#1 ajello , m. , rebusco , p. , cappelluti , n. , et al .
2009 , , 690 , 367 ajello , m. , rebusco , p. , cappelluti , n. , reimer , o. , bhringer , h. , la parola , v. , & cusumano , g. 2010 , , 725 , 1688 andersson , k. , peterson , j. r. , & madejski , g. 2007 , , 670 , 1010 barcons , x. , mateos , s. , & ceballos , m. t. 2000 , , 316 , l13 boldt , e. 1987 , observational cosmology , 124 , 611 bonafede , a. , feretti , l. , murgia , m. , et al .
2010 , , 513 , a30 bonamente , m. , nevalainen , j. , & lieu , r. 2007 , , 668 , 796 brunetti , g. , & blasi , p. 2005
, , 363 , 1173 buote , d. a. 2001 , , 553 , l15 clarke , t. e. , kronberg , p. p. , bhringer , h. 2001 , , 547 , l111 govoni , f. , & feretti , l. 2004 , international journal of modern physics d , 13 , 1549 govoni , f. , markevitch , m. , vikhlinin , a. , et al .
2004 , , 605 , 695 harrison , f. a. , craig , w. w. , christensen , f. e. , et al .
2013 , , 770 , 103 humphrey , p. j. , liu , w. , & buote , d. a. 2009 , , 693 , 822 kaastra , j. s. 1992 , an x - ray spectral code for optically thin plasmas ( internal sron - leiden rep .
, updated version 2.0 ) keshet , u. , kushnir , d. , loeb , a. , & waxman , e. 2012 , arxiv:1210.1574 kim , k .-
t . , kronberg , p. p. , dewdney , p. e. , & landecker , t. l. 1990 , , 355 , 29 kim , m. , wilkes , b. j. , kim , d .- w . , et al .
2007 , , 659 , 29 kushnir , d. , & waxman , e. 2010 , journal of cosmology and astroparticle physics , 2 , 25 leccardi , a. , & molendi , s. 2007 , , 472 , 21 lehto , t. , nevalainen , j. , bonamente , m. , ota , n. , & kaastra , j. 2010 , , 524 , a70 liang , h. , hunstead , r. w. , birkinshaw , m. , & andreani , p. 2000 , , 544 , 686 mantz , a. , allen , s. w. , ebeling , h. , & rapetti , d. 2008 , , 387 , 1179 markevitch , m. , gonzalez , a. h. , david , l. , et al .
2002 , , 567 , l27 markevitch , m. 2006 , the x - ray universe 2005 , 604 , 723 markwardt , c. b. 2009 , astronomical data analysis software and systems xviii , 411 , 251 mitchell , r. j. , culhane , j. l. , davison , p. j. n. , & ives , j. c. 1976 , , 175 , 29p moretti , a. , gastaldello , f. , ettori , s. , & molendi , s. 2011 , , 528 , a102 nevalainen , j. , david , l. , & guainazzi , m. 2010 , , 523 , a22 nynka , m. , hailey , c. j. , mori , k. , et al .
2013 , , 778 , l31 ota , n. 2012 , research in astronomy and astrophysics , 12 , 973 petrosian , v. , madejski , g. , & luli , k. 2006 , , 652 , 948 rephaeli , y. 1979 , , 227 , 364 rephaeli , y. 1987 , , 225 , 851 rephaeli , y. , & gruber , d. e. 1988 , , 333 , 133 rephaeli , y. , nevalainen , j. , ohashi , t. , & bykov , a. m. 2008 , , 134 , 71 revnivtsev , m. , gilfanov , m. , sunyaev , r. , jahoda , k. , & markwardt , c. 2003 , , 411 , 329 sarazin , c. l. 1999 , , 520 , 529 schuecker , p. , bhringer , h. , reiprich , t. h. , & feretti , l. 2001 , , 378 , 408 shafer , r. a. 1983 , phd .
thesis , university of maryland snowden , s. l. , mushotzky , r. f. , kuntz , k. d. , & davis , d. s. 2008 , , 478 , 615 solinger , a. b. , & tucker , w. h. 1972 , , 175 , l107 thierbach , m. , klein , u. , & wielebinski , r. 2003 , , 397 , 53 tucker , w. , blanco , p. , rappoport , s. , et al .
1998 , , 496 , l5 trler , m. , chernyakova , m. , courvoisier , t. j .- l . , et al . 2010 , , 512 , a49 wik , d. r. , sarazin , c. l. , finoguenov , a. , matsushita , k. , nakazawa , k. , & clarke , t. e. 2009 , , 696 , 1700 wik , d. r. , sarazin , c. l. , zhang , y .- y . , et al .
2012 , , 748 , 67 vanderlinde , k. , et al .
2010 , arxiv:1003.0003 vikhlinin , a. , et al .
2009 , , 692 , 1060 zuhone , j. a. , markevitch , m. , & lee , d. 2011 , , 743 , 16
the observatory design gives rise to various , independent background components that vary spatially across the fov , complicating standard background estimation techniques .
the purpose of this appendix is to describe the behavior of the spatial variation of the various background components in the detector plane and how to translate to the sky frame used in source analysis . in this framework
, we make assumptions about the spatial and spectral features of the several background components , that are physically motivated but empirically determined , to allow a model of the background for the entire fov to be based on the characterization of only non - source regions .
the model itself derives from fits to stacked `` blank '' field observations , taken from the deep ( ecdfs ) and medium ( cosmos ) survey data , listed in table [ tab : app : blankfields ] .
lcc 60022001_ecdfs_mos001 & 60022001002 & 41.2 + 60022002_ecdfs_mos002 & 60022002001 & 43.1 + 60022003_ecdfs_mos003 & 60022003001 & 43.2 + 60022004_ecdfs_mos004 & 60022004001 & 43.7 + 60022005_ecdfs_mos005 & 60022005001 & 42.3 + 60022006_ecdfs_mos006 & 60022006001 & 41.6 + 60022007_ecdfs_mos007 & 60022007002 & 44.3 + 60022008_ecdfs_mos008 & 60022008001 & 43.2 + 60022009_ecdfs_mos009 & 60022009001 & 41.7 + 60022010_ecdfs_mos010 & 60022010001 & 42.7 + 60022011_ecdfs_mos011 & 60022011001 & 43.4 + 60022012_ecdfs_mos012 & 60022012001 & 43.9 + 60022013_ecdfs_mos013 & 60022013001 & 44.6 + 60022014_ecdfs_mos014 & 60022014001 & 44.8 + 60022015_ecdfs_mos015 & 60022015001 & 45.1 + 60022016_ecdfs_mos016 & 60022016001 & 42.1 + 60022001_ecdfs_mos001 & 60022001003 & 40.9 + 60022002_ecdfs_mos002 & 60022002002 & 41.2 + 60022003_ecdfs_mos003 & 60022003002 & 41.0 + 60022004_ecdfs_mos004 & 60022004002 & 41.2 + 60022005_ecdfs_mos005 & 60022005002 & 41.2 + 60022006_ecdfs_mos006 & 60022006002 & 41.3 + 60022007_ecdfs_mos007 & 60022007003 & 41.7 + 60022008_ecdfs_mos008 & 60022008002 & 41.8 + 60022009_ecdfs_mos009 & 60022009003 & 41.6 + 60022010_ecdfs_mos010 & 60022010002 & 27.9 + 60022010_ecdfs_mos010 & 60022010004 & 13.2 + 60022011_ecdfs_mos011 & 60022011002 & 41.6 + 60022012_ecdfs_mos012 & 60022012002 & 41.9 + 60022013_ecdfs_mos013 & 60022013002 & 41.6 + 60022014_ecdfs_mos014 & 60022014002 & 44.1 + 60022015_ecdfs_mos015 & 60022015003 & 43.5 + 60022016_ecdfs_mos016 & 60022016003 & 43.9 lcc 60021001_cosmos_mos001 & 60021001002 & 18.8 + 60021002_cosmos_mos002 & 60021002001 & 22.6 + 60021003_cosmos_mos003 & 60021003001 & 20.5 + 60021004_cosmos_mos004 & 60021004001 & 21.8 + 60021005_cosmos_mos005 & 60021005001 & 21.5 + 60021006_cosmos_mos006 & 60021006001 & 21.7 + 60021007_cosmos_mos007 & 60021007001 & 22.9 + 60021008_cosmos_mos008 & 60021008001 & 23.3 + 60021009_cosmos_mos009 & 60021009002 & 22.8 + 60021010_cosmos_mos010 & 60021010001 & 24.4 + 60021011_cosmos_mos011 & 60021011001 & 25.9 + 60021012_cosmos_mos012 & 60021012001 & 23.0 + 60021013_cosmos_mos013 & 60021013001 & 25.3 + 60021014_cosmos_mos014 & 60021014001 & 22.7 + 60021015_cosmos_mos015 & 60021015001 & 23.4 + 60021016_cosmos_mos016 & 60021016001 & 25.5 + 60021017_cosmos_mos017 & 60021017001 & 22.9 + 60021018_cosmos_mos018 & 60021018001 & 24.0 + 60021019_cosmos_mos019 & 60021019001 & 28.9 + 60021020_cosmos_mos020 & 60021020002 & 28.0 + 60021021_cosmos_mos021 & 60021021001 & 27.4 + 60021022_cosmos_mos022 & 60021022001 & 22.0 + 60021023_cosmos_mos023 & 60021023001 & 24.6 + 60021024_cosmos_mos024 & 60021024001 & 25.4 + 60021025_cosmos_mos025 & 60021025001 & 22.1 + 60021026_cosmos_mos026 & 60021026001 & 28.9 + 60021027_cosmos_mos027 & 60021027002 & 23.9 + 60021028_cosmos_mos028 & 60021028001 & 22.9 + 60021029_cosmos_mos029 & 60021029001 & 22.9 + 60021030_cosmos_mos030 & 60021030001 & 24.0 + 60021031_cosmos_mos031 & 60021031001 & 22.2 + 60021032_cosmos_mos032 & 60021032001 & 25.8 + 60021033_cosmos_mos033 & 60021033001 & 21.7 + 60021034_cosmos_mos034 & 60021034001 & 19.4 + 60021034_cosmos_mos034 & 60021034003 & 9.4 + 60021035_cosmos_mos035 & 60021035002 & 21.4 + 60021036_cosmos_mos036 & 60021036002 & 22.7 + 60021037_cosmos_mos037 & 60021037002 & 24.1 + 60021038_cosmos_mos038 & 60021038001 & 23.1 + 60021039_cosmos_mos039 & 60021039001 & 22.3 + 60021040_cosmos_mos040 & 60021040001 & 23.8 + 60021041_cosmos_mos041 & 60021041001 & 22.2 + 60021042_cosmos_mos042 & 60021042002 & 22.4 + 60021043_cosmos_mos043 & 60021043001 & 23.8 + 60021044_cosmos_mos044 & 60021044002 & 21.7 + 60021046_cosmos_mos046 & 60021046002 & 17.6 + 60021046_cosmos_mos046 & 60021046004 & 13.0 + 60021047_cosmos_mos047 & 60021047002 & 24.9 + 60021048_cosmos_mos048 & 60021048002 & 24.9 + 60021049_cosmos_mos049 & 60021049002 & 24.7 + 60021053_cosmos_mos053 & 60021053002 & 11.4 + 60021053_cosmos_mos053 & 60021053004 & 5.7 + 60021053_cosmos_mos053 & 60021053006 & 6.2 consists of two separate telescopes ( two sets of optics , housed in the optics module , focusing onto two focal planes , housed in the focal plane module ) the telescopes , or associated data / response functions , are referred to as a and b. each focal plane consists of a 2@xmath492 array of cdznte detectors with a 32@xmath4932 array of pixels . in principle , each pixel has a unique background response , but in practice all the pixels on a single detector excepting edge pixels behave similarly . due to differences in thickness and other properties of the detectors , the instrumental background for each detector is somewhat unique .
the benches containing the optics and detectors are separated on two ends of an unenclosed mast . pointing variations throughout a given observation cause a given detector pixel to sample several times more sky than without this wobble . because the light path is open to space , stray light from the cosmic x - ray background ( cxb ) is able to skirt between the optics bench and the aperture stops in front of the two focal planes ( figure [ fig : app : geometry ] ) ; the geometry of this window produces highly non - uniform background gradients across the detectors at low energy ( @xmath139 kev ) . at the lowest energies , scattered solar x - rays reflected from other parts of the observatory structure are visible to the detectors , due to its open design .
the low altitude and inclination orbit of minimizes saa activation and proton flares , so the instrumental background dominating at higher energies is low and stable . ,
title="fig:",height=453 ] , title="fig:",height=453 ] the background spectrum can generally be decomposed into four broadband components of fixed spectral shape . for certain observations near the galactic plane , an additional component to account for diffuse galactic ridge emission ( grxe )
may also be needed .
the spectral components , fit to a stacked spectrum of the observations from table [ tab : app : blankfields ] for the entire fov , are shown in figure [ fig : app : stack ] . , width=415 ] below @xmath140 kev , the background is dominated by stray light from unblocked sky leaking through the aperture stop ; when the origin of this emission is the cxb , this component is referred to as the `` aperture '' background . by their nature , the background produced by the cxb , grxe , or bright sources with a line of sight through the stop is spatially non - uniform
. the fov samples a solid area of 37.2 deg@xmath31 on the sky , with any individual pixel exposed to something in the range 0.310 deg@xmath31 . below @xmath50 kev
, there is a strong , soft additional component ( `` solar '' ) that is most likely due to reflected solar x - rays as evidenced by its persistence in spectra from earth observations when the satellite is illuminated by the sun and its absence when not .
this component can undergo significant fluctuations due to solar activity .
although thought to come primarily from reflections off the backside of the aperture stop , this conjecture has yet to be confirmed and thus we have no way to predict the spatial pattern it produces on the detectors , but it is likely non - uniform as well .
the other low energy contributor to the background is from the cxb `` focused '' by the optics ( fcxb " ) .
the fcxb " includes both truly focused events ( photons reflected off of both mirrors ) and scattered events or ghost rays ( photons reflecting off of only one mirror ) from the many unresolved sources both within and outside the fov .
its shape is roughly flat across the detector plane despite vignetting due to an increase in scattered light from sources outside the fov at larger off - axis angles . above @xmath52@xmath141
kev , the internal or instrumental background dominates .
it is made up of gamma rays compton scattered by the detector and shield , lines activated by interactions between the spacecraft / detectors and the radiation environment in orbit , and a few fluorescence lines .
most of the lines are driven by frequent if glancing
passages through the saa , when protons activate material in the focal plane module near or in the detectors .
unstable elements are created by proton spallation and secondary neutron capture by cadmium , which then radioactively decay with half - lives typically longer than s orbital period .
the strongest of these activation lines appear in the complex from 2225 kev .
while the strengths of these lines depend on the spacecraft s recent orbital history , there is as yet no evidence for spatial variations across individual detectors , and the relative strength of a given line between detectors which depends on properties unique to each detector such as its thickness does not vary .
the strongest instrumental lines are due to k - shell fluorescence of cesium and iodine at 28 kev and 31 kev , respectively , residing in the anti - coincidence shield .
the continuum , meant to represent the compton scattered component and any other featureless instrumental components , is modeled as a broken power law with a break at 124 kev .
the lines and line complexes are modeled with 29 lorentzian - profile lines , _ empirically _ added to the spectra in figure [ fig : app : stack ] until the fit can no longer be reasonably improved .
initially , the line energies and widths , which are tied between the a and b spectra , are allowed some freedom during the fitting process as is the temperature describing the `` solar '' component and the indices of the internal continuum but at some arbitrary point the model is deigned to be `` good enough '' and those parameters fixed thereafter .
the internal ( and perhaps `` solar '' ) components exhibit no detectable spatial variation within individual detectors , but they do between detectors .
to first order , the cxb has a constant surface brightness across the sky .
the intensity detected by a given pixel thus depends on the solid angle of visible sky , which is solely a function of the observatory s geometry .
each pixel `` sees '' a solid angle of @xmath126 deg@xmath31 defined by the circular aperture stop .
the view is blocked , however , by the apparent position of the optics bench , which depends on the location of the pixel in the focal plane , so the level of cxb flux smoothly varies across the detectors . despite understanding this geometry ,
the absolute position of the focal plane detectors in the bench is uncertain at the 1 mm level . also , just as for the `` solar '' component , cxb emission from the entire rear hemisphere of the sky
except that blocked by the earth can be scattered by the backside of the aperture stop and other parts of the observatory into the focal plane , thus modulating its spatial distribution .
using cxb focal plane maps generated by ray traces through the observatory s geometry , we can adjust the precise position of the detectors within the focal plane and the proportion of unmodulated , scattered cxb flux until we obtain a good match to stacked images from the blank sky observations . to isolate the `` aperture '' component ,
we stack 715 kev images from all the observations in detector ( det1 ) coordinates , which has a finer spatial resolution than the native pixels ( possible due to probability distribution functions relating to event grades obtained from pencil - beam ground calibrations ) ; the stacked images are shown in the left panels of figure [ fig : app : det1im ] .
the 40@xmath4940 mm detector plane is binned into cells with sides 2 to 4 mm long and fit to the ray trace model using the @xmath65 statistic and minimization package mpfit @xcite .
in addition to fitting for the `` aperture '' model @xmath142 and @xmath143 positions and normalizations , we also include simple spatial models for the internal , `` solar , '' and `` fcxb '' components , the normalizations of which are allowed to vary .
the relative flux assigned to each component is initially inconsistent with the expectation from figure [ fig : app : stack ] , when the aperture " spectral component only includes direct emission through the aperture stop . to reconcile the spatial and spectral aperture " models ,
the spatial model requires extra flat emission of uncertain origin .
one possibility is that earth albedo or cxb photons are scattered off of the backside of the aperture stop and elsewhere and into the detector housing .
spectra extracted during earth - occulted periods , for example , exhibit a component below 15 kev with the same spectral shape as the `` aperture '' component , even though the cxb is not directly visible ( see figure [ fig : app : solar ] ) .
alternatively , a contributor to the internal continuum component may rise with decreasing energy instead of following the simple power law spectrum we assume .
the spectral shape of this component is hard to predict , and we make no attempt to do so , so such a rise is very plausible and would be consistent with it being spatially flat . for simplicity , the spatial model of the `` aperture '' component
is modified to include this extra emission , rather than adjusting the spectral shape of internal continuum .
the amount of extra emission added is increased until we achieve self - consistency between spectral and spatial fits to the data without large shifts in the position of the detectors .
we also performed simultaneous spectro - spatial fits of similarly binned regions , and while they are the most comprehensive , they are too computationally intensive and fickle to arrive at best - fit detector offsets and extra `` aperture '' emission . however , these fits were useful to explore the parameter space , as were fits to the full fov spectra , informing the level to which each component should contribute to the 715 kev images .
this iterative procedure results in extra
aperture " fractions of 13% for each focal plane and position offsets of ( @xmath144 , @xmath145 ) and ( @xmath146 , @xmath147 ) in ( @xmath142 , @xmath143 ) for a and b , respectively .
these exact values are irrelevant as long as the `` aperture '' shape is correct ; they only matter if one wants to extract an absolute flux for the cxb using this data . the model created with these values
is shown in the right panels of figure [ fig : app : det1im ] . because the origin of the `` aperture '' component of the background is well understood , we were able to characterize its variation across the fov with high confidence .
the spatial distribution of the two other cosmic sources `` fcxb '' and `` solar '' of the background have yet to be as well - constrained .
although the `` fcxb '' distribution can in principle be simulated , at the time of this writing the model of the mirror modules is still being refined to account for observed ghost ray patterns .
( ghost rays are photons typically scattered by the optics , usually once - reflected , from sources within @xmath148 of the optical axis . )
the pre - launch model predicts a distribution somewhat following the vignetting function ; however , the blank sky fields show no evidence of such a spatial modulation . observed ghost ray patterns produced by bright sources near to but outside the fov show an additional halo farthest from the source , likely due to reflections off the back sides of the mirrors .
the extra contribution due to this halo from cxb sources outside the fov may act to compensate for the drop in flux from higher off - axis sources within the fov . empirically , the spatial shape of the `` fcxb '' is consistent with a flat distribution , although due to its relative faintness it is difficult to discern otherwise .
we assume a flat distribution hereafter for simplicity .
the `` solar '' component has only recently been recognized as originating from the sun through reflections off the observatory structure .
no study of its likely spatial distribution has been undertaken , and in any case the distribution may vary with sun angle . to allow for spatial variations ,
we treat the `` solar '' continuum and associated 3.5 kev and 4.5 kev lines as if they had an instrumental origin and thus should only vary between detectors and not within them .
although this treatment amounts to a very coarsely defined spatial model , this component is typically only important below 5 kev where sources are brightest .
we note , however , that x - ray emission from the sun is highly variable and that during flares this component can dominate up to 10 kev ; such periods are not currently handled by the background model described here , since the spectrum itself is likely to evolve from the quiescent one we include .
when data are split between the periods that the spacecraft is and is not illuminated by the sun , the correlation between the soft emission and a solar origin is clear , as shown in the examples in figure [ fig : app : solar ] , the four panels of which also demonstrate its variability . in contrast
, the components of the background with an instrumental origin should not depend on position within the fov , as long as the detectors are all uniform and identical .
consisting of single cdznte crystals , each individual detector should be very close to uniform , which agrees with the lack of spatial fluctuations across any given detector in the stacked high energy images of the blank sky fields .
the detectors are not identical , however , and the variation between them in thickness and charge transport properties lead to slight differences in overall background level and line strengths . for any given observation ,
the overall level and strengths depend on the orbital history through the south atlantic anomaly and other higher radiation zones .
since all the detectors share this history , the _ relative _ strengths of the internal components should always be the same . to complete our empirical spectro - spatial model of the background
, we simply need to determine the ratios between these components for the detectors on each of the two focal planes .
we separate each full fov spectrum from figure [ fig : app : stack ] into four spectra corresponding to each detector , which each share the same spectral model shapes . having previously determined the `` aperture '' and `` fcxb '' model spatial shapes , the relative proportion of their flux falling on each detector
is fixed appropriately , but all other model normalizations are left free .
the four spectra are then fit simultaneously , and independently for each telescope , and the resulting fits are shown in figure [ fig : app4det ] .
many of the lines have similar strengths on each detector , but that is not universally true .
table [ tab : app : ratios ] gives the fraction of the model normalization of each component associated with each detector . ,
title="fig:",width=340 ] , title="fig:",width=340 ] rr|cccc|cccc 3.54 & 0.43 & 0.217 & 0.228 & 0.326 & 0.228 & 0.235 & 0.231 & 0.209 & 0.324 + 4.51 & 0.54 & 0.165 & 0.141 & 0.393 & 0.301 & 0.305 & 0.219 & 0.157 & 0.319 + 10.20 & 0.64 & 0.035 & 0.000 & 0.425 & 0.540 & 0.308 & 0.194 & 0.148 & 0.350 + 19.65 & 0.23 & 0.295 & 0.057 & 0.220 & 0.427 & 0.167 & 0.156 & 0.407 & 0.270 + 21.89 & 0.57 & 0.182 & 0.140 & 0.248 & 0.430 & 0.179 & 0.259 & 0.330 & 0.232 + 22.97 & 0.15 & 0.257 & 0.204 & 0.235 & 0.305 & 0.230 & 0.228 & 0.311 & 0.231 + 24.75 & 1.96 & 0.204 & 0.166 & 0.273 & 0.357 & 0.225 & 0.248 & 0.290 & 0.237 + 25.30 & 0.15 & 0.316 & 0.256 & 0.231 & 0.198 & 0.250 & 0.245 & 0.279 & 0.226 + 27.75 & 1.71 & 0.543 & 0.457 & 0.000 & 0.000 & 0.232 & 0.252 & 0.265 & 0.251 + 28.08 & 2.06 & 0.000 & 0.000 & 0.511 & 0.489 & 0.250 & 0.250 & 0.250 & 0.250 + 28.55 & 0.27 & 0.285 & 0.310 & 0.195 & 0.210 & 0.266 & 0.251 & 0.244 & 0.239 + 30.17 & 0.71 & 0.222 & 0.182 & 0.281 & 0.315 & 0.245 & 0.220 & 0.277 & 0.257 + 30.86 & 0.45 & 0.273 & 0.272 & 0.235 & 0.220 & 0.248 & 0.259 & 0.255 & 0.238 + 32.19 & 0.86 & 0.236 & 0.201 & 0.259 & 0.304 & 0.267 & 0.250 & 0.244 & 0.239 + 35.03 & 0.82 & 0.249 & 0.277 & 0.277 & 0.198 & 0.265 & 0.255 & 0.227 & 0.254 + 39.25 & 9.13 & 0.238 & 0.150 & 0.299 & 0.313 & 0.252 & 0.235 & 0.253 & 0.260 + 39.40 & 0.52 & 0.231 & 0.215 & 0.234 & 0.320 & 0.225 & 0.239 & 0.269 & 0.267 + 47.56 & 6.82 & 0.261 & 0.193 & 0.286 & 0.260 & 0.246 & 0.252 & 0.245 & 0.256 + 52.50 & 1.60 & 0.241 & 0.193 & 0.260 & 0.306 & 0.233 & 0.245 & 0.266 & 0.256 + 57.99 & 4.72 & 0.237 & 0.162 & 0.257 & 0.345 & 0.247 & 0.255 & 0.233 & 0.265 + 65.01 & 5.24 & 0.230 & 0.162 & 0.270 & 0.338 & 0.228 & 0.218 & 0.271 & 0.283 + 67.06 & 0.53 & 0.233 & 0.279 & 0.193 & 0.296 & 0.278 & 0.267 & 0.210 & 0.245 + 75.18 & 5.59 & 0.204 & 0.300 & 0.191 & 0.305 & 0.242 & 0.205 & 0.259 & 0.294 + 85.82 & 7.58 & 0.219 & 0.236 & 0.274 & 0.271 & 0.274 & 0.241 & 0.260 & 0.225 + 87.90 & 0.58 & 0.299 & 0.279 & 0.199 & 0.223 & 0.241 & 0.260 & 0.238 & 0.260 + 92.67 & 0.64 & 0.279 & 0.282 & 0.210 & 0.228 & 0.242 & 0.257 & 0.247 & 0.254 + 105.36 & 0.46 & 0.272 & 0.335 & 0.172 & 0.221 & 0.237 & 0.246 & 0.238 & 0.279 + 122.74 & 2.30 & 0.275 & 0.458 & 0.076 & 0.192 & 0.204 & 0.261 & 0.263 & 0.272 + 144.56 & 0.74 & 0.300 & 0.294 & 0.193 & 0.213 & 0.253 & 0.260 & 0.229 & 0.258 + & 0.222 & 0.189 & 0.216 & 0.373 & 0.279 & 0.262 & 0.195 & 0.264 + & 0.239 & 0.252 & 0.243 & 0.266 & 0.254 & 0.244 & 0.252 & 0.250 based on the above description , each identified component making up s background has been assigned a fixed spectral shape and spatial distribution across the fov .
given these assumptions , one can directly measure a `` local '' background for any subset of the fov and use that to accurately predict the background for anywhere in the entire fov .
the quality of the background is of course limited by the statistics available in the observation used to constrain the background model , but one advantage of separating out the different components is that separate systematic uncertainties associated with each component can be applied individually .
now that we have a background model , we can use the events far from the cluster to determine the precise level of each background component , which are unique to the conditions of these observations . to apply the model defined in appendix [ sec : appendixbgd ]
, we have developed a small suite of idl routines called nuskybgd , whose purpose is to take regions defined in sky coordinates , compute the relative strengths of each background component based on their location in the detector plane , and create an xspec - readable script that sets up and fits for all observation - specific component normalizations , much in the spirit of the background treatment in the extended source analysis software package ( as introduced in * ? ? ?
those normalizations correspond to a complete spectro - spatial background model from which images in any energy band or spectra for any region can be produced . in principle
, we could extract a single spectrum of the non - cluster part of the fov for each telescope and epoch and fit the model to that .
the downside of this approach is that all spatial information is lost , which can cause the various components especially the `` aperture '' component to obtain unphysical best - fit normalizations . to incorporate this information
while also keeping the computational load to a minimum , we divide the non - cluster area into four rectangular regions for each focal plane and epoch , shown in figure [ fig : bgdreg ] .
we also try to minimize the `` contamination '' of these regions with cluster emission , mostly originating from the brightest parts of the cluster and carried far away from its true location by the wings of the psf .
the ellipse in figure [ fig : bgdreg ] indicates the parts of the background regions excluded for this reason .
even so , residual cluster emission remains , which we must also model to avoid biasing the background model . :
the images have been smoothed by gaussian kernel of width 5 pixels and scaled between 0 ( white ) and 1 ( black ) counts to bring out structure in the background .
[ fig : bgdreg],title="fig:",width=340 ] : the images have been smoothed by gaussian kernel of width 5 pixels and scaled between 0 ( white ) and 1 ( black ) counts to bring out structure in the background .
[ fig : bgdreg],title="fig:",width=340 ] : the images have been smoothed by gaussian kernel of width 5 pixels and scaled between 0 ( white ) and 1 ( black ) counts to bring out structure in the background .
[ fig : bgdreg],title="fig:",width=340 ] : the images have been smoothed by gaussian kernel of width 5 pixels and scaled between 0 ( white ) and 1 ( black ) counts to bring out structure in the background .
[ fig : bgdreg],title="fig:",width=340 ] for the most part , the regions from each telescope and epoch are fit independently , but the cxb components between telescopes and epochs are correlated and can be tied together to improve their precision .
the `` fcxb '' component , being the unresolved contribution of sources in that region of the sky , will be identical for telescopes a and b as long as the regions are roughly coincident . similarly , because the roll angle is very similar between the two epochs , the part of the sky producing the `` aperture '' component for each telescope is almost entirely identical . fitting all 16 spectra simultaneously therefore permits the normalizations of these parameters to be appropriately tied together , reducing the number of free parameters and the chance that any component gets pushed to an unphysical value by preventing the fit from heading down a local minimum .
we pursue this strategy because below @xmath52 kev all of the background components contribute at non - negligible levels , making it easier for the fit minimization procedure to be misled by mere statistical fluctuations . despite having a conservative exclusion region around the cluster
, a small but noticeable number of cluster photons are scattered into the background regions , roughly at the level of the `` fcxb '' component . because its contribution is fairly modest ,
the spectral model used to account for its emission does not have to be extremely accurate ; we take a single temperature model at the global average temperature of 14.1 kev and abundance relative to solar of 0.15 convolved with the same arf used by the `` fcxb '' component . the spectral shapes and normalizations of the scattered cluster emission and
`` fcxb '' components turn out to be very similar , which means that if both were left free they may very well take on unphysical values . even under these circumstances , however , the overall background model should not suffer , since the scattered component is not included in it and the `` fcxb '' normalization , while on average constant across the fov , can significantly vary location - to - location due to cosmic variance . as that flux in the background regions can not be directly applied at the cluster location , we simply fix the fcxb " component to its average value and allow the scattered cluster emission component to be free , which may compensate for variations in the cxb flux in each region as well .
the fit to all background region spectra is shown in figure [ fig : bgdspec ] . the background model is defined both spatially and spectrally , and its parameters
have now been determined for our specific observations , allowing a background spectrum to be generated from the model for any location in the fov with nuskybgd . to realistically assess the impact of both statistical and systematic uncertainties associated with the background on fits to the cluster spectrum , we perform monte carlo simulations of the background including those uncertainties as fluctuations from the expected model . because the background is broken up into separate components , each one can be varied based on its own systematic error , as specified in section [ sec : cal : bgd : sys ] .
each of the simulated background spectra is generated in two steps from the predicted model for the source region .
first , the normalizations of each component are randomly shifted , assuming a normal distribution about their systematic uncertainty .
then , a counts spectrum with poisson fluctuations is created from the adjusted model for an exposure time equal to that of the observation using the fakeit command in xspec . while counting statistics should not bias the modeling of the bullet cluster spectrum in principle ,
the true background can be thought of as one such realization ; a conspiracy of high or low shot noise at just the right energies would act just like a systematic offset .
our procedure captures the likelihood of such occurrences and thus more realistic error ranges for the cluster model parameters .
we simulate 1000 background spectra , enough to characterize the standard deviation at each energy and confirm the naive expectation that the fluctuations are roughly gaussian out to @xmath149 . considering the full gamut of likely background spectra , as opposed to the nominal model derived from local background regions ,
puts several intriguing or worrying features in the proper context . in figure
[ fig : specsig ] , the hard band of the bullet cluster spectrum is shown relative to the range bound by the background simulations ( red / green or gray shaded regions ) and relative to the average 1 t thermal model ( blue / dashed line ) . above @xmath13 kev
, the spectrum generally agrees well with the mean expectation of the background , and deviations from the mean fall appropriately distributed within the range .
a few energy ranges , however , show more systematic deviations from the mean . from @xmath14022 kev ,
the spectrum quickly rises above the 1 t model , and from @xmath15050 kev the spectrum stays slightly , but consistently , below the mean background level .
when considered relative to the allowed range of the background , it is clear that the deviations are not worryingly extreme .
a common systematic fluctuation in the 3550 kev background could cause the @xmath151 offset , and the blip at 22 kev is most likely an imperfectly calibrated background line or lines ( see table [ tab : app : ratios ] ) . | the search for diffuse non - thermal inverse compton ( ic ) emission from galaxy clusters at hard x - ray energies has been undertaken with many instruments , with most detections being either of low significance or controversial . because all prior telescopes sensitive at @xmath0 kev do not focus light and have degree - scale fields of view , their backgrounds are both high and difficult to characterize .
the associated uncertainties result in lower sensitivity to ic emission and a greater chance of false detection . in this work ,
we present 266 ks observations of the bullet cluster , which is detected in the energy range 330 kev .
s unprecedented hard x - ray focusing capability largely eliminates confusion between diffuse ic and point sources ; however , at the highest energies the background still dominates and must be well understood . to this end , we have developed a complete background model constructed of physically inspired components constrained by extragalactic survey field observations , the specific parameters of which are derived locally from data in non - source regions of target observations . applying the background model to the bullet cluster data , we find that the spectrum is well but not perfectly described as an isothermal plasma with @xmath1 kev . to slightly improve the fit
, a second temperature component is added , which appears to account for lower temperature emission from the cool core , pushing the primary component to @xmath2 kev .
we see no convincing need to invoke an ic component to describe the spectrum of the bullet cluster , and instead argue that it is dominated at all energies by emission from purely thermal gas .
the conservatively derived 90% upper limit on the ic flux of @xmath3 erg s@xmath4 @xmath5 ( 50100 kev ) , implying a lower limit on @xmath6 @xmath7 g , is barely consistent with detected fluxes previously reported . in addition to discussing the possible origin of this discrepancy , we remark on the potential implications of this analysis for the prospects for detecting ic in galaxy clusters in the future . |
on january 6 , 2010 , p/2010 a2 ( linear ) , a comet - like object , was discovered by the linear sky survey .
the orbital elements of the object ( @xmath2 = 2.29 au , @xmath3 = 0.12 and @xmath4 = 5.26 @xmath5 ) are typical of an inner main - belt asteroid belonging to the flora collisional family , and suggest that it is unlikely to have originated in the classical comet source regions ( i.e. , the kuiper belt or the oort cloud ) .
the object reached perihelion , at a heliocentric distance of @xmath6=2.0 au , on 4 december 2009 .
observations taken a few days after its discovery showed an inactive nucleus lying outside a dust cloud that looks like a cometary tail without central condensation @xcite .
this suggested that the observed dust cloud and the nucleus were the result of an impact between two previously unknown asteroids : the dust cloud is a plume of dust and the nucleus is what remains from the largest of the asteroids that collided .
owing to its cometary - like aspect and orbital parameters , p/2010 a2 can be classified as a main - belt comet ( mbc , * ? ? ?
in contrast with the other known mbcs , however , p/2010 a2 has a significantly smaller semi - major axis ( 2.29 au versus 2.7 au and @xmath73.2 au ) . until now
, the activity observed in mbcs was found to be compatible with a water - ice driven activation mechanism suggesting that those asteroids retained ice layers below their surface and , under certain conditions , become `` activated asteroids '' .
the presence of water - ice on the surface of 24 themis , the parent asteroid of the themis family mbcs @xcite strongly support this .
one of the activation mechanisms could be a collision between two asteroids .
collisions are known to take place regularly , but they are so rare that none of the dust plumes that they should generate has ever been seen
. if p/2010 a2 is the debris of a collisional event , this would be the first time that the ejecta from such a collision is observed soon after it happened .
it would provide a unique opportunity to learn something about asteroid collision processes and about the internal composition of asteroids . alternatively ,
if the observed activity is sustained in time as in comets , this leaves open two interesting problems : ( 1 ) if water - ice sublimation is the activation mechanism , how does water - ice survive in an asteroid with such a small semi - major axis ; or ( 2 ) is there any other mechanism capable of ejecting dust in a similar manner as water - ice sublimation does ? in this paper we present and analyze images of p/2010 a2 obtained with three telescopes at the roque de los muchachos observatory ( orm ) , la palma , spain . in sect .
2 the observations and data reduction are presented . in sect .
3 we analyze the combined images using the inverse monte carlo dust tail fitting method ( e.g. * ? ? ?
* ; * ? ? ?
* ) to study possible ejection scenarios .
the conclusions are presented in sect .
images of p/2010 a2 were obtained in january 2010 with the optical system for imaging and low resolution integrated spectroscopy ( osiris ; * ? ? ? * ; * ? ? ?
* ) camera - spectrograph at the gran telescopio canarias ( gtc ) , with the auxiliary camera - spectrograph ( acam ) at the william herschel telescope ( wht ) , and with the andalucia faint object spectrograph and camera ( alfosc ) at the nordic optical telescope ( not ) all located at the orm .
the observational circumstances are given in table 1 .
for the observations with the not , the telescope was set at the comet s rate motion , while sidereal tracking was used for both wht and gtc . in all cases
the images were obtained in service mode by telescope staff and were reduced by subtracting the overscan level and flat fielded using standard procedures .
the observations with osiris at the gtc were made with the sloan g@xmath8,r@xmath8 , and i@xmath8 filters .
osiris provides a field of view [email protected]@xmath10 with a gap of 9.2@xmath11 in the middle and a pixel scale of 0.125@xmath11pixel@xmath12 . to increase the signal - to - noise the data were binned in 2@xmath02 pixels .
the images were calibrated using photometric zeropoints determined from standard star observations .
the observations with acam at the wht were made with the sloan g@xmath8 and r@xmath8 filters .
acam is mounted permanently at a folded - cassegrain focus of the telescope and has a circular field of view of 8@xmath10 diameter with a pixel scale of 0.25 @xmath11pixel@xmath12 .
the resulting calibrated wht images show brightness levels in agreement with those obtained at gtc within 10% . the observations with alfosc at the not were made with standard johnson - cousins @xmath13 and @xmath14 filters .
alfosc provides a field of view of [email protected]@xmath10 with a pixel scale of 0.19 @xmath11pixel@xmath12 .
these images were calibrated with stars in the field of view using magnitudes from the usno - b1.0 catalog @xcite , which provides a photometric accuracy of @xmath70.3 mag . using the transformation equations of @xcite and the magnitude of the sun in the standard johnson - cousins @xmath13 filter ( @xmath15=26.75 , * ? ? ?
* ) , we derive @xmath16=26.96 . since @xmath17=27.29 , the @xmath18 images of the object obtained at gtc and wht were transformed to @xmath14 standard johnson - cousins magnitudes by subtracting 0.33 mag , where we assume for the object the same spectral dependence as for the sun within the bandpasses of these two red filters .
the resulting gtc , wht , and not @xmath14 magnitudes are consistent with their errors and we conclude that the object did not change brightness significantly during our observations . the available images in each night were combined in order to improve the signal - to - noise ratio and converted to solar disk intensity units appropriate for the analysis in terms of dust tail models .
figure 1 depicts some representative images obtained on the different dates with the instruments mentioned above . in this figure
one can clearly see the inactive nucleus ( marked with an arrow ) lying outside the dust cloud that looks like a cometary tail .
the absence of a dust cloud surrounding the nucleus indicates that the dust emission has stopped before the observations .
using aperture photometry , we determined the nucleus @xmath14 magnitudes for the gtc , wht , and not images whenever possible ( see table 1 ) .
the aperture size was of 1.5@xmath11 , and the nucleus profiles were stellar - like .
for the gtc images , which have the best s / n ratio , and using the formalism by @xcite with a slope parameter of @xmath19=0.15 , we obtained a nucleus absolute magnitude of @[email protected] .
hence , a nucleus diameter of @xmath22=220@xmath2140 m can be determined , assuming a bulk albedo of @xmath23=0.11 , typical of a s - type asteroid .
this assumption is based on the fact that the s - type asteroids are the most common objects in the inner main asteroid belt ( e.g. * ? ? ?
the combined images from gtc and wht obtained on the 17th and 21st january show the best s / n ratio ( see fig .
1 ) , and were selected for a more detailed analysis . to carry out the analysis in terms of dust tail models an additional rebinning was made of the images .
the wht image was rebinned to 4 pixels , giving a spatial resolution of 772.4 km pixel@xmath12 , while the gtc image was rebinned to 3 pixels to give a resolution of 576.3 km pixel@xmath12 .
finally , the rebinned images were rotated to the @xmath24 coordinate system , where @xmath25 is the extended radius vector from the sun , and @xmath26 is perpendicular to @xmath25 and directed opposite to the object s motion along its orbit @xcite .
we have performed an analysis of both the wht and gtc images by the inverse monte carlo dust tail fitting code ( e.g. * ? ? ?
* ; * ? ? ?
* ) , which is based on the same procedures as that developed by fulle ( e.g. * ? ? ?
* ; * ? ? ?
briefly , a large amount of particles ( typically 10@xmath1 ) of a selected size range are released from the surface of the object in a given time interval , with an assumed velocity law , which may be a function of time and particle size .
the particles are then submitted to the radiation pressure and gravity field of the sun , and their keplerian orbits are computed . the ratio of the force exerted by the solar radiation pressure and the solar gravity is given by @xmath27 , which can be expressed as @xmath28 , where @xmath29=1.19@xmath0 10@xmath30 kg m@xmath31 , and @xmath32 is the particle density . for particle radii
@xmath33 0.25 @xmath34 m , the radiation pressure coefficient is @xmath35 1 @xcite .
the inverse method involves the inversion of the overdetermined system of equations @xmath36 , where @xmath37 is the kernel matrix containing the dust tail model , i.e. , the surface density of the sample of particles integrated over time and @xmath27 , @xmath38 is the output vector which contains the time - dependent distribution of @xmath27 , and @xmath39 is the brightness of the input image in the selected region of the photographic @xmath24-plane .
the physical appearance of the tail , detached from the nucleus and sun - ward oriented ( see figures 1 and 2 ) , suggests that most of the dust was already ejected sometime well before the date of the observation , and should be composed of relatively large particles , otherwise they would have been blown away by radiation pressure .
this is clearly demonstrated in figure 2 , where the combined wht image is shown along with the synchrones @xcite corresponding to ejection times from 272 days before perihelion to 3 days after perihelion . assuming a particle density of 1000 kg m@xmath30
, these synchrones would correspond to particle sizes between 0.001 cm and 1 cm . to start the simulations , and based on the synchrone map , we assumed initially that dust has been ejected at any time starting approximately a year before the observations , with sizes in the range 0.001 cm to 1 cm .
we fixed the particle density to @xmath32= 1000 kg m@xmath30 , and the albedo times the phase function to be @xmath40=0.04 , i.e. , a halley - like value . in principle , we assumed isotropic ejection
. the better spatial resolution of the gtc image compared to the wht image might suggest anisotropic ejection , which is particularly indicated by the shape of the isophotes near the nucleus .
we will come back to this point later .
for lack of better information , we assumed an ejection velocity as given by @xmath41 , which from hydrodynamical considerations is typical from gas drag by sublimation processes .
we did not assume any temporal variation in the ejection velocities , which is reasonable taking into account the low eccentricity of the orbit .
we found that the best fits to the images are found when the velocity is given by @xmath42 cm s@xmath12 .
this imply a velocity of 8.5 cm s@xmath12 for a grain of @xmath6=1 cm .
using the formulae by @xcite for a nucleus of 220@xmath2140 m in diameter as given above and heating efficiency factor of 0.1 we get , at a heliocentric distance of 2 au , a range of velocities of 1214 cm s@xmath12 ( and smaller values for lower values of the heating efficiency ) , which are not far from our model results .
this would then be consistent with a typical cometary activity scenario where the outgassing is driven by water - ice sublimation .
the result of the model fit to the wht image is shown in figure 3(a ) , and the derived dust mass loss rates as a function of time are shown in figure 4 .
the model isophotes reproduce quite closely those of the wht image .
the variation of the mass loss rate indicates that the onset of the activity occurred in late march 2009 , with a maximum of activity around early june 2009 , with a peak of dust loss rate of about 5 kg s@xmath12 , and decreasing to 0.1 kg s@xmath12 near perihelion in early december 2009 .
the integrated dust loss is 5.4@xmath010@xmath1 kg . assuming a spherically - shaped nucleus of diameter of 220 m , and a bulk density of 3000 kg m@xmath30 , the dust mass in the tail corresponds to 0.3% of the total nucleus mass .
we tried other model combinations to search for other possible solutions .
in particular , we tried to fit the data by assuming narrower time intervals for the dust ejection , which could support an interpretation of a single collision event without any further outgassing activity , but we could not find any acceptable solutions even modifying also the particle sizes and ejection velocities with values well above and below the expected values that occur in the laboratory experiments such as presented by @xcite . if the isotropic ejection model with the same input parameters is applied to the gtc image , we obtain the result shown in figure 3(b ) . in this case ,
due to the fact that a certain portion of the tail is clipped , the dust mass loss differs from that found for the wht image , particularly in that the mass loss becomes zero at times earlier than 180 days pre - perihelion ( see figure 4 ) . for the same reason ,
the integrated ejected dust mass is also lower at 1.7@xmath010@xmath1 kg .
looking closely at the fitted isophotes , some mismatch in the tail head near the nucleus can be seen , and also in the outermost isophote towards the sun - ward region .
therefore , we decided to incorporate ejection anisotropy in the model by considering emission from a selected active area at a particular region of a spherically symmetric rotating nucleus .
the rotation period was set to 3 hours . after trying tens of models with varying rotational parameters and location of the active area ,
a very good fit was obtained by taken the argument of the sub - solar meridian at perihelion as @xmath43=180@xmath44 , and the obliquity as @xmath39=95@xmath44 , with an active area occupying a part of the southern hemisphere of the object in the latitude range [ 90@xmath44,40@xmath44 ] .
the resulting fit to the gtc observation is shown in figure 3(c ) .
if a collision triggered the outgassing event , it could have occurred at high southern latitudes on the object .
although this model seems highly speculative , and other alternative scenarios exist , it does produce an excellent fit to the data . for the resulting time - averaged size distribution
, a fit indicates a power - law with an index of [email protected] , a value typically found in the analysis of dust ejected from comets ( e.g. * ? ? ?
the above fits indicate that the observations can not be explained by dust released in a sudden event such as a collision and subsequent ejection of asteroid surface material , but it does need a sustained cometary - like activity during several months after the onset of the activity as described above . whether the onset of the activity was triggered by a collision event or by another mechanism that might be similar to those occurring in other main - belt comets @xcite , can not be decided from the available data .
the application of the inverse dust tail monte carlo fitting method to images of the tail of p/2010 a2 tail has revealed that the observed brightness distribution is only consistent with a sustained ( possibly water - ice driven ) cometary - like activity that spanned approximately 8 months , which is estimated to have started in late march 2009 .
the total amount of the ejected mass was about 5@xmath010@xmath1 kg , when assuming an albedo times phase function of 4% , and a particle density of 1000 kg m@xmath30 .
it is not possible to asses whether the outgassing event was triggered by a collision or by another unspecified mechanism .
however , the detection of a sustained cometary - like activity over a considerable period of time has obvious consequences for the internal structure of the object .
it might indicate that some subsurface ice layer exists in this object .
another option is that there exists another mechanism which is capable of causing ejection of dust in a manner similar to water - ice sublimation .
a more sophisticated ( but also more speculative ) dust tail model indicates that the dust ejection was localized and originated on the southern hemisphere of the nucleus , with rotation parameters @xmath43=180@xmath44 , obliquity @xmath39=95@xmath44 , and a rotation period of @xmath73 hours .
these parameters can be confirmed with further follow - up observations .
this work is based on observations made with the gran telescopio canarias ( gtc ) , the william herschel telescope ( wht ) , and the nordic optical telescope ( not ) , installed in the spanish observatorio del roque de los muchachos of the instituto de astrofsica de canarias , in the island of la palma .
the nordic optical telescope is operated on the island of la palma jointly by denmark , finland , iceland , norway , and sweden , in the spanish observatorio del roque de los muchachos of the instituto de astrofisica de canarias .
the instrument alfosc on the not is owned by the instituto de astrofsica de andaluca ( iaa ) and operated at the nordic optical telescope under agreement between iaa and the nbifafg of the astronomical observatory of copenhagen . this work was supported by contracts aya2007 - 63670 , aya2008 - 01720e , aya2009 - 08190 , and fqm-4555 ( proyecto de excelencia , junta de andaluca ) . j. licandro gratefully acknowledges support from the spanish `` ministerio de ciencia e innovacin '' project aya2008 - 06202-c03 - 02 .
j. l. ortiz gratefully acknowledges support from the spanish `` ministerio de ciencia e innovacin '' project aya2008 - 06202-c03 - 01 . 14 jan 2010 & not 2.56 m & 2.013 & 1.039 & 5.53 & [email protected] + 17 jan 2010 & gtc 10.4 m & 2.014 & 1.047 & 7.20 & [email protected] + 21 jan 2010 & wht 4.2 m & 2.016 & 1.061 & 9.32 & [email protected] + 23 jan 2010 & not 2.56 m & 2.016 & 1.065 & 9.87 & | the dust ejecta of main - belt comet p/2010 a2 ( linear ) have been observed with several telescopes at the at the observatorio del roque de los muchachos on la palma , spain .
application of an inverse dust tail monte carlo method to the images of the dust ejecta from the object indicates that a sustained , likely water - ice driven , activity over some eight months is the mechanism responsible for the formation of the observed tail . the total amount of dust released is estimated to be 5@xmath010@xmath1 kg , which represents about 0.3% of the nucleus mass .
while the event could have been triggered by a collision , this can not be decided from the currently available data . |
galaxies with unusual morphologies appear at high redshift in the hubble deep fields north ( williams et al .
1996 ) and south ( volonteri , saracco , & chincarini 2000 ) , the hawaiian deep field ( cowie , hu , & songaila 1995 ) , the tadpole galaxy field ( tran et al . 2003 ) , the goods field ( giavalisco et al . 2004 ) and the hubble ultra deep field ( hdf ; beckwith et al .
. some are recognizable as elliptical and spiral galaxies , but many at @xmath3 have more clumpy and irregular structures ( e.g. , abraham et al .
1996a , b ; conselice 2004 ) .
chain galaxies , for example , are nearly straight alignments of a half - dozen clumps ( cowie , hu , & songaila 1995 ; van den bergh et al . 1996 ; elmegreen , elmegreen & sheets 2004 , hereafter paper i ) .
tadpole galaxies are curved thin structures with a big clump near one end ( van den bergh et al .
1996 ; paper i ; straughn et al .
luminous diffuse objects are oval distributions of clumpy emission resembling disks ( reshetnikov , dettmar , & combes 2003 ; conselice et al .
2004 ) ; a subcategory of these , having no bulges or exponential disk light profiles , has been called clump - clusters ( elmegreen , elmegreen , & hirst 2004a , hereafter paper ii ) .
double - clump galaxies have two big clumps .
most of these types are not present in the modern universe although some probably evolve into normal hubble types .
characteristic of all the irregular types are enormous clumps of star formation , @xmath4 times more massive the the largest star complexes in today s spiral galaxies , as well as thick disks and extreme asymmetries that are most likely the result of interactions and merging . here
we present morphological data on 884 galaxies in the hubble ultra deep field ( udf ) .
magnitudes , colors , and axial ratios of the galaxies are studied .
udf images were obtained with the hubble advanced camera for surveys ( acs ) by beckwith et al .
( 2005 ) and are available on the space telescope science institute ( stsci ) archive .
the images are 10500 x 10500 pixels with a scale of 0.03 arcsec per px ( 315 arcsec x 315 arcsec ) in 4 filters : f435w ( b band , hereafter b@xmath5 ; 134880 s exposure ) , f606w ( v band , v@xmath6 ; 135320 s ) , f775w ( i band , i@xmath7 ; 347110 s ) , and f850lp ( z band , z@xmath8 ; 346620 s ) .
we used the i@xmath7 image , the deepest of the four images , to search by eye to identify and study galaxies that have major axes larger than 10 pixels ( corresponding to 0.3 arcsec ) .
the whole udf field was subdivided into 50 fields of 800x800 pixels for identification of objects .
each of these subfields was further subdivided and displayed in iraf in 100x100 pixel increments , which were scanned by eye to search for objects larger than 10 pixels ( confirmed by measuring with contour plots out to the 2@xmath9 limit ) .
our full sample includes 884 galaxies .
morphological classifications were made based on the i@xmath7 images and aided by contour plots and radial profiles .
the galaxies were divided into 6 categories : chain ( 114 total ) , clump - cluster ( 178 ) , double ( 126 ) , tadpole ( 97 ) , spiral ( 269 ) , and elliptical ( 100 ) .
figure [ fig : examples ] shows 8 examples of each type ; the lines correspond to 0.5 arcsec .
galaxy morphology can vary with wavelength , so we viewed many of the catalogued objects at other acs passbands and with nicmos ( thompson et al . 2005 ) .
generally the morphological classification does not change significantly with wavelength ( e.g. , dickinson 2000 ) because it is based on only the most fundamental galaxy characteristics , such as elongation and number of giant clumps .
also , the nicmos images have a factor of 3 lower resolution , so they do not reveal the same fine structure as the other images .
the distinguishing characteristics of the main types we classified are : * chain : linear objects dominated by several giant clumps and having no exponential light profiles or central red bulges . * clump cluster : oval or circular objects resembling chain galaxies in their dominance by several giant clumps , and having no exponential profiles or bulges . *
double - clump : systems dominated by two similar clumps with no exponential profile or bulge .
* tadpole : systems dominated by a single clump that is off - center from , or at the end of , a more diffuse linear emission . *
spiral : galaxies with exponential - like disks , evident spiral structure if they have low - inclination , and usually a bulge or a nucleus .
edge - on spirals have relatively flat emission from a midplane , and often extended emission perpendicular to the midplane , as well as a bulge . *
elliptical : centrally concentrated oval galaxies with no obvious spiral structure .
chain galaxies were first recognized by cowie et al .
( 1995 ) using the same definition as that here .
tadpole galaxies were defined by van den bergh et al .
( 1996 ) and examples from the udf were discussed by straughn et al .
tadpole galaxies with short tails were classified as `` comma '' type in the morphology review by van den bergh ( 2002 ) .
van den bergh et al .
( 1996 ) also noted objects like clump - clusters and called them proto - spirals .
conselice et al .
( 2004 ) called these clump - dominated young disk galaxies `` luminous diffuse objects , '' although some of their sample included galaxies with bulges and exponential - like profiles , unlike the clump - clusters here .
binary galaxies , like our doubles , were noted by van den bergh ( 2002 ) .
other classifications of galaxies in deep hubble images , based on the hubble classification or ddo systems , were made by brinchmann et al .
( 1998 ) , van den bergh et al . ( 2000 ) , and others , who generally noted that an increasing fraction of distant objects fall outside these conventional morphologies .
distinctions between morphological types do not necessarily imply there are significant physical differences .
for example , double - clump galaxies may be smaller versions of chain galaxies , and the single - clump tadpoles may be smaller yet .
the origins of these types are not known .
also , chain galaxies and clump clusters may be the same type of object viewed at different orientations ( dalcanton & schectman 1996 ; oneil et al .
2000 ; paper ii ) .
spiral galaxies are extremely varied .
many look highly disturbed , as if by interactions with smaller companions .
figure [ fig:4spirals ] shows four more examples of spirals .
the galaxy on the left is very asymmetric and has a giant star formation region along the bottom arm .
the galaxy on the right is dominated by two round clumps and an extended bright arm ( `` the cigar smoker '' ) .
images of 10 different clump clusters from the udf are in elmegreen & elmegreen ( 2005 , hereafter paper iii ) , where their clump properties were measured . images of 8 different udf ellipticals are in elmegreen , elmegreen , & ferguson ( 2005 , hereafter paper iv ) , which concentrated on those having giant blue clumps in their centers ( one of the ellipticals shown in fig .
[ fig : examples ] , udf 4389 , has blue clumps in its core , but they can not be seen in this image ) .
ellipse fits were done on all of the spirals using iraf to search for radial variations in position angle and ellipticity .
bars have a characteristic signature on such a plot as a result of the twisting inner isophotes .
the images were also viewed directly to look for bars . in general ,
the most highly inclined galaxies do not readily show bars even if they are present , and the short type of bar that is commonly found in local late - type galaxies could not be resolved .
nevertheless , we found 26 barred spirals out of 269 spiral galaxies , a bar fraction of @xmath0% .
this is a factor of @xmath2 smaller than in the tadpole field , considering even the inclination effects ( elmegreen , elmegreen , & hirst 2004b , hereafter paper v ) , and it is a factor of @xmath10 smaller than in the goods field ( jogee et al .
the tadpole field contained 43 barred galaxies in a sample of 186 spirals and showed clear correlations between apparent bar fraction and both inclination and galaxy angular size .
the goods field contained 80 bars out of 258 nearly face - on spirals . both previous studies used radial profiles of ellipticity and position angle to identify bars , as we do here .
deeper fields like the udf might be expected to show fewer bars as physical resolution and surface brightness dimming become more of a problem at high redshift .
bar destruction by severe interactions ( athanassoula & bosma 2003 ) could play a role at high @xmath11 too .
a more thorough study of the udf bars will be addressed in another paper .
early searches for barred galaxies found relatively few examples ( e.g. , van den bergh et al .
2002 ) , probably because of the lower angular resolution of the wfpc2 camera .
sheth et al .
( 2003 ) was the first to suggest that the relative numbers of bars might be the same as that found locally .
two examples of barred spirals in the udf are shown in figure [ fig : bar ] , along with the radial profiles of ellipticity and position angle .
they are the galaxies udf 1971 ( left ) and udf 9341 .
the bars and spirals are more irregular than in local galaxies , suggesting chaotic stellar orbits and poorly defined resonances .
the bars could therefore be young .
star formation is enhanced in many small clusters near the top end of the bar in ugc 1971 .
the distribution functions of the colors and magnitudes of the clumps that have been measured in chains , clump - clusters , doubles , and tadpoles , are all about the same ( paper i ) .
similar also are the colors and magnitudes of clumps in the cores of 30 ellipticals in the udf ( paper iv ) .
these colors and magnitudes give clump masses in the range from @xmath12 m@xmath13 for the smallest ellipticals ( paper iv ) to @xmath14 m@xmath13 for the best studied clump - clusters ( paper iii ) , based stellar population models from bruzual & charlot ( 2003 ) .
the clumps are clearly smaller and bluer than the bulges of spiral galaxies ( paper i ) .
the irregular morphology of the first four types suggests a history dominated by galaxy interactions , clump accretions , and large - scale instabilities leading to star formation . the highly clumpy structure of the elliptical galaxies suggests a similar history of clump accretions .
most of the galaxies in this catalog are also listed in the udf catalog without morphological classifications ( beckwith et al .
2005 , listed on the stsci website ) .
some of the galaxies here are not listed in the udf catalog because they appear near bright stars or near the edge of the field of view .
the parameters used for sextractor were chosen to optimize different aspects of the udf than our morphological system , so there is not a one - to - one correspondence .
our catalog has 9 chain galaxies , 14 clump - cluster galaxies , 10 double galaxies , and 1 tadpole galaxy with multiple udf numbers , and there are 12 chains , 16 clump - clusters , 11 doubles , 8 tadpoles , 28 spirals and 6 ellipticals which are not in the udf catalog at all .
one object , udf 5898 , is two separate spiral galaxies which sextractor called one , using the midpoint as a position ; we call these udf 5898a and udf 5898b . the unlisted and multiply - listed galaxies amount to 154 objects , or 17% of the total sample .
the integrated ab magnitudes were measured in all four filters from counts within a rectangular box whose edges were defined by the i@xmath7 isophotal contours 2@xmath9 above the sky noise ( corresponding to a surface brightness of 26.0 mag arcsec@xmath15 ) .
the integrated magnitudes depend only weakly on the choice of outer isophotal contour : 10-pixel extensions changed the magnitudes of isolated objects by less than 0.1 mag .
the integrated i@xmath7-z@xmath8 colors of the galaxies range from -0.7 to 1.3 .
the integrated i@xmath7 magnitudes range from 18.5 to 29 mag , with surface brightnesses from 22.5 to 27 mag arcsec@xmath15 .
the details of some clump properties are given in separate studies of different subsamples ( papers iii , iv ) .
tables 1 - 6 list the galaxies by morphological type . in the first part of each table , the galaxies with a single udf number are listed . in the second part
, galaxies with multiple udf numbers are given along with the combined average surface brightnesses in i@xmath7 , @xmath16 , the apparent magnitudes , i@xmath7 , and the three colors . in the third part
, galaxies with no udf numbers are given along with the coordinates of their centers and other properties .
we do not repeat the colors and magnitudes of the single sources because these measurements are in the udf catalog ; they are the same as what we measured independently to within 0.1 mag ( which is the same as the measuring error for udf magnitudes ) .
barred galaxies are indicated with an asterisk in the spiral list .
the average color and magnitude values for each galaxy type are shown in figure [ fig : colormagall ] .
this plot has more observational relevance than physical because the galaxies are a mixture of redshifts and k - corrections .
the average i@xmath7 surface brightness is fairly constant at @xmath17 mag arcsec@xmath15 for all types except ellipticals , which are about 0.5 mag brighter .
this constancy is likely the result of the surface brightness limit of the survey ( @xmath18 is 26.0 mag arcsec@xmath15 ) , with the averages being larger than this limit because of central light concentration .
the average i@xmath7 integrated magnitude is about 26.3 for chains , doubles , and tadpoles , 24.5 for spirals , 24.2 for ellipticals , and 25.2 for clump clusters .
the ellipticals have the reddest colors in all indices .
the surface brightness distribution of all the galaxies peaks at @xmath19 mag arcsec@xmath15 , rapidly declines beyond the 2@xmath9 limit of 26.0 mag arcsec@xmath15 , and is virtually null beyond the 3@xmath9 limit of 26.6 mag arcsec@xmath15 ( see also fig . [
fig : wlvsmag ] below ) .
figure [ fig : overlay ] plots b@xmath5-v@xmath6 versus @xmath20 for all galaxies in our sample , divided according to morphology . note the lyman break galaxies of various types , which have extremely red b@xmath5-v@xmath6 colors . except for these , the color distributions are similar with a large number of galaxies at b@xmath5-v@xmath21 and a second diagonal branch going up to b@xmath5-v@xmath6@xmath2 .
the curves superposed on the data are bruzual & charlot ( 2003 ) evolution models that assume star formation began at @xmath22 for all cases and then decayed exponentially over time with e - folding times of 0.01 gy , 0.03 gy , 0.1 gy , 0.3 gy , 1 gy , 3 gy , and infinity .
each curve traces the sequence of decay times with the longest decay times corresponding to the bluest colors .
the different curves are for different galaxy redshifts , from 0 to 4 in steps of 0.5 ( see color coding in figure caption ) .
the models are described more in paper iii .
the distribution of colors reveals several things about these galaxies .
the chains , clump - clusters , doubles , and tadpoles have similar color distributions , and the spirals have about the same distribution too , but the spirals do not have the very red b@xmath5-v@xmath6 and i@xmath7-z@xmath8 components that are present in the more irregular systems .
these red points correspond to the highest redshifts , @xmath23 , so the difference in distribution suggests the spirals do not generally go to as high a redshift as the more irregular and clumpy galaxies .
there is a spiral , udf 7036 , that has the colors of a b - band drop - out , however , so it could be at @xmath23 ( see sect .
[ sect : lbg ] and table [ tab : lbg ] ) .
selection effects could remove the most easily recognized spirals , which are the face - on spirals , from our sample ( see sect .
[ sect : axial ] ) .
the irregular and spiral galaxies both populate the bluer regions in figure [ fig : overlay ] , which suggests that both types linger around in cosmological time after the spirals become obvious ( see also franceschini et al .
1998 ) .
the ellipticals also extend in figure [ fig : overlay ] to the high v@xmath6-i@xmath7 and b@xmath5-v@xmath6 values of the clumpy types .
the ellipticals are slighlty redder than the other types , lacking the big clump of points around i@xmath7-z@xmath24 and b@xmath5-v@xmath21 that all the others have .
evidently the ellipticals have much less star formation at the epoch of observation , as is well known ( e.g. , ferguson , dickinson & williams 2000 ) .
figure [ fig : fractions ] plots the number and fraction of galaxies of each type as a function of i@xmath7 magnitude . the counts of galaxies with spiral or elliptical morphologies peak at about the same apparent magnitude , which is slightly brighter than the peak of the clump clusters .
the counts of chain , double , and tadpole galaxies all peak at about one magnitude fainter .
the fractions follow a similar trend , with spirals dominating the bright galaxies , while chains , doubles , and tadpoles dominate the faint galaxies .
as is well known , the galaxies with the most unusual morphologies increase in dominance at the faintest magnitudes , beyond i@xmath25 .
the trend for chains , doubles , and tadpoles to dominate at the faintest magnitudes illustrates an important selection effect discussed in the next section .
we believe on the basis of this trend , and on the basis of the distribution of axial ratios for spirals , that the spirals begin to disappear at i@xmath25 because the face - on spirals become too faint to see at the surface brightness limit of the survey ( @xmath26 mag arcsec@xmath15 for @xmath18 detections ) .
edge - on spirals have higher surface brightnesses than face - on spirals , and are observed to fainter magnitudes . in the distribution of axial ratios ( see below ) , there is a clear and unusual drop in the count for face - on spirals . in the same way
, the chain galaxies extend to fainter magnitudes than the clump - clusters in figure [ fig : fractions ] because chain galaxies are probably edge - on clump clusters ( paper ii ) .
edge - on galaxies tend to have higher surface brightnesses and fainter apparent magnitudes than face - on galaxies ( see models in paper i ) .
thus , the faintest galaxies at the surface - brightness limit of detection are observed to be linear structures , many of which are probably flattened objects viewed edge - on .
the tadpole galaxies , for example , could be strongly interacting galaxies with a @xmath27 m@xmath13 merger remnant seen as a single clump at one end and a long tidal tail seen as the diffuse lane ( e.g. , see straughn et al
figure [ fig : ellfractwl ] compares the distribution of axial ratios , @xmath28 , for elliptical galaxies in the udf field with the distribution for local ellipticals in the third reference catalogue of bright galaxies ( de vaucouleurs et al .
1991 ; rc3 ) .
the udf ratios came from the ellipse fits in the published udf catalog .
the error bars equal @xmath29 for number @xmath30 . for the udf distribution ,
the dashed histogram is for the half of the sample with the brightest average surface brightness ( @xmath31 mag arcsec@xmath32 ) .
this bright half is slightly rounder on average than the faint half , as observed locally ( tremblay & merritt 1996 ) .
the overall distribution for udf ellipticals resembles the distribution for local galaxies , particularly for the high surface brightness ellipticals , suggesting that elliptical galaxies relaxed to their current forms at very early times .
the distribution of axial ratios for spiral galaxies is shown in figure [ fig : spiralwl ] .
the spiral galaxies in the deep field background of the tadpole galaxy ( tran et al . 2003 ; paper i ) are shown on the upper left .
the udf spirals are on the bottom left .
we divided the udf spirals into the faint surface - brightness half and the bright surface - brightness half and plotted the separate distributions on the right of figure [ fig : spiralwl ] .
for the udf spirals , the solid - line histograms use the ellipticities given in the udf catalog and the dashed - line histograms use ellipticities derived individually for each galaxy from ellipse fits in iraf .
because the spirals are irregular , we did not want to use the sextractor ellipticities exclusively . to check the first two ellipticity measures
, we also determined the axial ratios ( @xmath33ellipticity ) by eye , based on the length ( @xmath34 ) and width ( @xmath35 ) of the average 2-@xmath9 contour limits for each galaxy .
all three axial ratio determinations correlated well with each other , to within 0.1 in @xmath28 . in what follows
we consider only the axial ratios determined by the udf automatic ellipse fit and by our iraf ellipse fits , which are more objective than the visual fits to contour shapes .
the distribution of axial ratios for spirals does not resemble the local distributions for spirals , which is reproduced in figure [ fig : spiralwlrc3 ] using data from the rc3 .
the local distribution is relatively flat from @xmath36 down to @xmath37 , depending on hubble type , indicating that the objects are round disks with intrinsic axial ratios comparable to this lower limit . the only exception is for spirals of de vaucouleurs type 9 , which are sm irregulars and known to be relatively thick .
the udf and tadpole - field spirals have a distribution of axial ratios that is peaked in the middle , which means that the catalog lacks face - on examples ( @xmath36 ) .
the fall - off at low @xmath28 is probably from the intrinsic disk thickness , as for local galaxies , although these spiral disks are evidently @xmath2 times thicker than the local versions ( e.g. , reshetnikov et al .
2002 ; papers ii , iii ) . the fall - off at high
@xmath28 could arise if the spirals are intrinsically elongated disks , i.e. , not circular , as suggested for local irregular galaxies ( binggeli & popescu 1995 ; sung et al . 1998 ) . alternatively , they could be circular disks that tend to be overlooked when viewed face - on .
we consider these two options in detail now .
figure [ fig : wl ] shows several models for the projected axial ratios of intrinsically oval , i.e. , triaxial , disks .
the two on the left show simple cases and the two on the right are sample fits to the udf or tadpole data .
the model on the bottom left is for circular thick disks with a range of relative thicknesses , @xmath38 to 0.375 , viewed in projection at random angles .
the flat part extending to @xmath39 is a result of the circular form of the disk itself , and the rise and fall below @xmath40 is for the edge - on case , when the minimum axial ratio is viewed .
the top left is for an ellipsoid with an axial ratio of 1:0.6:0.2 .
the peak at @xmath41 is from the main disk axial ratio , when the disk is viewed face - on .
the two peaks at @xmath42 and 0.33 are for edge - on versions ; the lower ratio corresponds to the case where the oval disk is viewed edge - on parallel to the minor axis ( showing the major axis length in projection ) , and the higher ratio corresponds to the case where the oval disk is viewed edge - on along the major axis ( showing the minor axis length in projection ) .
neither of these ideal cases matches the data , but mixtures of them do .
the models on the right span a range of intrinsic @xmath28 and @xmath43 ratios , viewed in projection for orientations of the ellipsoid at random angles in 3d space .
evidently , the distribution of axial ratios observed for the udf is consistent with intrinsically elongated spiral disk galaxies .
if the disks are actually distorted like this , then tidal forces would be a likely explanation .
such an interpretation is plausible in view of the importance of galaxy interactions at high redshift .
however , visual examination of each udf spiral suggests that very few of the elongated cases have companion galaxies or other direct evidence of a strong interaction .
thus we consider another explanation , as follows .
recall that figure [ fig : spiralwl ] showed that the distribution of axial ratios for the low surface brightness spirals is slightly more skewed toward edge - on than the distribution for the high surface brightness spirals .
this distortion is shown again in figure [ fig : wlvsmag ] , which plots the i@xmath7 magnitudes and average surface brightnesses ( out to the @xmath18 contour ) versus the axial ratios .
the solid red line is the average of the plotted points , which are from the udf catalog .
the dashed red line is the average for the w / l values we measured based on iraf ellipse fits ( these values are not plotted ) . the density of points on this plot increases for intermediate @xmath28 , reflecting the peak at intermediate @xmath28 in figure [ fig : spiralwl ] . the density of points shifts to even lower @xmath28 at the faintest surface brightnesses ( top part of the top panel ) , and this shift also parallels the trend in figure [ fig : spiralwl ] .
the bottom part of figure [ fig : wlvsmag ] shows that the average flux gets @xmath2 mag brighter as the spiral galaxies become more face - on .
this brightening also occurs for local spirals because the total flux increases as the projected area increases and stars become less obscured by midplane dust .
the top part of figure [ fig : wlvsmag ] indicates that the average surface brightness is either constant or increases slightly ( @xmath44 mag ) for more face - on spirals , which is contrary to expectations .
locally the surface brightness decreases for face - on disks because the path length through the stars decreases ( paper i ) .
the situation changes , however , if most of the udf surface brightnesses for spirals are below the detection limit .
this limit is 26 mag arcsec@xmath15 for the faintest contour that defines our sample , so the average surface brightness of a typical galaxy inside this contour is greater , perhaps by a magnitude .
the two dashed green lines in this figure represent the approximate region where we begin to lose galaxies at the surface brightness limit . above @xmath45 mag arcsec@xmath15
, the integrated surface brightness would be fainter than the lowest contour considered here ( @xmath18 ) and the galaxy would not likely be in our survey .
thus we should ask what a distribution of surface brightnesses should look like if the average surface brightness decreases with increasing @xmath28 as for local galaxies , and as it decreases , the detection limit is passed so the fainter galaxies become lost in the noise .
the solid green line shows what the average distribution of surface brightnesses should be if a face - on galaxy is at 25.9 mag arcsec@xmath15 ( where @xmath39 ) and edge - on galaxies get brighter in proportion to the path length through the disk ( i.e. , @xmath46 ) .
this solid line is well populated with observed galaxies at low @xmath28 so the number of points in the plot is large there , but as @xmath28 increases , the surface brightness limit comes in and the number of points centered on the line decreases .
another way to see this surface brightness limit is with figure [ fig : fractions ] .
there we noted that the spiral count ( in the left panel ) drops precipitously for galaxies fainter than i@xmath47 mag , which is about where the distributions of chains , doubles , and tadpoles all linear systems begin to peak .
this drop may be combined with a model result in paper i to explain the factor - of-5 drop in the axial ratio distribution function between @xmath48 and 1 ( fig .
[ fig : spiralwl ] ) .
paper i showed in its figure 8 a radiative transfer model through a thick exponential disk with various levels of extinction ; the average disk surface brightness of the projected disk was plotted versus the axial ratio . as @xmath28 increased from 0.5 to 1 , the average surface brightness decreased by factors of 1.92 , 1.78 , and 1.74 for perpendicular extinctions to the midplane at the galaxy center equal to 0.33 , 1.33 , and 1.77 optical depths .
the limiting case with no extinction would be a factor of 2 , which is the inverse ratio of the two @xmath28 values , as this intensity ratio is mostly the ratio of path lengths along the line of sight .
for these three extinctions , the surface brightness dimming in going from @xmath48 to 1 corresponds to 0.71 , 0.63 , and 0.61 mag arcsec@xmath15 .
such magnitude changes correspond , in figure [ fig : fractions ] , to more than half of the full drop in spiral galaxy counts ( left - hand panel ) .
this means that at the limit of the udf survey , projection effects alone can change a galaxy that would be detected at @xmath49 to one that would be below the limit of detection at @xmath36 .
a simple model for the probability of detection of disk galaxies near the surface brightness limit of a survey is shown in figure [ fig : faceon ] .
this model assumes that the intrinsic surface brightness distribution of face - on spiral galaxies , in magnitudes arcsec@xmath15 , is gaussian , i.e. , that the distribution of intensity is log - normal .
this assumption is consistent with the narrow range of surface brightnesses for modern spiral galaxies ( freeman 1970 ) and with the tully & fisher ( 1977 ) relation , which suggests the same narrow range . in terms of the model on the top of figure
[ fig : wlvsmag ] , the breadth of the gaussian would give the intrinsic vertical range of points around the solid green line .
for figure [ fig : faceon ] , we integrate over this log - normal distribution to find the relative number of galaxies detected above some minimum intensity level .
the absolute value of the intensity at the peak of the gaussian is arbitrary so we take it to be @xmath50 , and we take the gaussian dispersion to be @xmath9 .
the observed intensity of a galaxy with face - on intensity @xmath51 is assumed to be @xmath52 , which is approximately the case without extinction ( for @xmath28 not too small ) .
this means that if an observed intensity is @xmath53 , then the intrinsic face - on intensity for use in the distribution function is @xmath54 .
thus we plot the detection probability : @xmath55 ^ 2\right ) d\ln i } \over{\int_{-\infty}^\infty \exp\left(-0.5\left[\ln\left\{iw / l\right\}/\sigma\right]^2\right ) d\ln i}}. \label{eq : p}\ ] ] the intensity threshold @xmath56 and intrinsic dispersion were chosen to match the observed fall - off in the distribution of axial ratios between @xmath48 and 1 .
this gives @xmath57 between 0.25 and 0.5 , or @xmath56 between @xmath58 and @xmath59 times the intensity at the peak of the log - normal distribution for the two cases plotted .
the fit also gives @xmath60 .
the fits in figure [ fig : faceon ] indicate that the fall - off in spiral galaxy counts between @xmath48 and 1 corresponds to a most - probable , face - on , galaxy surface brightness that is fainter than the threshold of detection by a factor between 1.3 and 1.6 .
thus a high fraction of face - on spiral galaxies are likely to be missed and high fraction of inclined spiral galaxies are likely to be seen ( if they satisfy our @xmath61 pixel size selection ) .
the fall - off in galaxy count for face - on spirals does not appear for local galaxies ( fig .
[ fig : spiralwlrc3 ] ) .
this is because there is a limited range in surface brightness for most spiral galaxies and the imaging surveys that made the rc3 went considerably deeper than the average galaxy surface brightness .
the fact that there is also not a large increase in local counts at small @xmath28 indicates that there is not a relatively large number of thin , ultra - low surface brightness galaxies in the local universe .
the situation is apparently different in the udf because even though galaxies tend to be intrinsically more intense than they are locally , i.e. , because of their higher star formation rates , cosmological surface brightness dimming , which is proportional to @xmath62 , knocks down their apparent surface brightnesses by large factors , corresponding to 3 , 4.8 , 6 , and 7 magnitudes arcsec@xmath15 for @xmath63 , 2 , 3 and 4 .
thus only the brightest of the local galaxies , viewed at these redshifts , would be detected in the udf survey , and most would be lost unless they were viewed nearly edge - on .
the distribution of axial ratios for chain and clump cluster galaxies in the udf was shown in paper iii .
it is flatter than the spiral galaxy distribution , perhaps because the clumps in clump clusters are very bright and they can still be seen in the face - on versions even if the interclump medium is close to or below the survey limit . figure [ fig : fractions ] showed a large difference in the apparent magnitude distribution for spirals compared to the linear systems ( chains , doubles and tadpoles ) and a modest difference compared to clump clusters , which may be face - on versions of chains .
the linear systems dominate at the faintest magnitudes and some of the reason for this may be the surface brightness selection effects just described , especially since the spiral sample in figure 6 contains the full range of axial ratios
. however , not all of the difference in these distributions is likely to result from surface brightness selection .
paper i showed that even the edge - on spirals in the tadpole field , as identified by their bulges and exponential disks , tend to have brighter apparent magnitudes and slightly brighter surface brightnesses than the chains , doubles , and tadpoles .
this is true also for the edge - on spirals in the udf sample ( not shown here ) .
this difference among the purely edge - on systems could be the result of a smaller average redshift for the spirals than for the chains , doubles and tadpoles , which would be consistent with evolution over time from irregular structures to smooth spiral disks .
for example , we noted in paper i that the spirals in the tadpole field tend to have slightly larger angular sizes than the doubles and tadpoles ( by 50% on average ) .
the redshift distributions of galaxies with various morphologies will have to be studied in order to determine the relative importance of surface brightness selection and evolutionary effects .
steidel et al . ( 1999 and references therein ) identified high z galaxies photometrically by their redshifted lyman break .
adelberger et al .
( 2004 ) and others applied the technique to galaxies between redshifts 1 and 3 . in the udf field ,
lyman break galaxies ( lbgs ) appear as b - band drop - outs at @xmath23 and v - band drop - outs at @xmath64 .
giavalisco et al .
( 2004 ) gave color criteria for lyman break galaxies , based on a sample of lbgs in the goods / acs survey .
based on their criteria , we find in our udf sample 30 b - band drop - outs ( 10 chains , 4 clump clusters , 8 doubles , 4 tadpoles , 1 spiral , and 3 ellipticals ) and 13 v - band drop - outs ( 1 chain , 3 clump - clusters , 8 doubles , and 1 tadpole ) .
these galaxies are identified by their type and udf catalog number in table [ tab : lbg ] . in figure
[ fig : overlay ] , many of the objects redder than @xmath65 in b@xmath5-v@xmath6 are b - band drop - outs .
there are many other lbg galaxies in the udf that are smaller than our 10 pixel major axis limit ( bouwens et al .
bunker et al .
( 2004 ) and yan & windhorst ( 2004 ) reported over 100 i - band drop - out ( @xmath66 ) galaxies in the udf ; one of our v - band drop - out chain galaxies was also reported at redshift z=5.4 by rhoads et al .
( 2004 ) .
most of the drop - outs in table [ tab : lbg ] are linear galaxies ( doubles , chains , tadpoles ) and relatively few are disks ( spiral ) .
this is consistent with the common perception that disks form late , but it may also contain a selection bias like that discussed in the previous section if most face - on disks are fainter than the survey limit .
madau et al .
( 1998 ) determined star formation rates ( sfrs ) from models of star clusters with salpeter initial mass functions based on observed uv fluxes . using their conversions , and assuming that the rest wavelength of 1500 is shifted approximately to i@xmath7-band for the @xmath23 galaxies and to z@xmath8 band for @xmath67 , we calculated the sfrs ( in m@xmath13/year ) of our lbgs .
the results are listed in table [ tab : lbg ] .
the star formation rates for the v - band drop - outs in this sample tend to be larger than for the b - band drop - outs . in comparison , immeli et al .
( 2004 ) predicted a sfr of about 20 m@xmath13 yr@xmath32 from their chain galaxy models , which assumed galaxy sizes a factor of five larger than those of our lbgs .
somerville , primack , & faber ( 2001 ) got rates of @xmath68 m@xmath13 yr@xmath32 for the largest galaxies considered in their lyman break models .
we classified and measured 884 galaxies larger than 10 pixels in the hubble udf .
six prominent types were used : chain , clump - cluster , double , tadpole , spiral , and elliptical .
their colors , magnitudes , and surface brightnesses are similar , reflecting their likely formation over a large range of times .
the distribution of axial ratios for elliptical galaxies is similar to that in the local field , suggesting that the elliptical shapes were quickly established and remained unchanged over time .
the distribution of axial ratios for spiral galaxies was distinctly different than for local galaxies with a deficit of face - on spirals in the deep field .
models illustrate how this could be the result of an intrinsically oval shape , as might be expected for highly perturbed galaxies . however , other models suggest that most of the face - on spirals could be lost below the surface brightness limit of the survey , while the more highly inclined spirals could still be seen .
a significant lack of spiral galaxies with low axial ratios implies an intrinsic disk thickness of 0.2 to 0.3 times the disk length .
this implies the spiral disks are thicker than local spirals by a factor of @xmath2 to 3 .
twenty - six barred spiral galaxies were found , a fraction of the total spirals equal to about 10% . this is a factor of 2 less than other deep field studies but not inconsistent with them considering projection effects and surface brightness dimming .
in addition , bars become harder to distinguish at high redshifts because the spiral galaxies generally get more irregular .
we thank the vassar college ursi program and the keck northeast astronomy consortium exchange program for undergraduate research support for m.s . and
is supported by the national science foundation through grant ast-0205097 . | galaxies in the hubble ultra deep field ( udf ) larger than 10 pixels ( 0.3 arcsec ) have been classified according to morphology and their photometric properties are presented .
there are 269 spirals , 100 ellipticals , 114 chains , 126 double - clump , 97 tadpole , and 178 clump - cluster galaxies .
we also catalogued 30 b - band and 13 v - band drop - outs and calculated their star formation rates .
chains , doubles , and tadpoles dominate the other types at faint magnitudes .
the fraction of obvious bars among spirals is @xmath0% , a factor of 2 - 3 lower than in other deep surveys .
the distribution function of axial ratios for elliptical galaxies is similar to that seen locally , suggesting that ellipticals relaxed quickly to a standardized shape .
the distribution of axial ratios for spiral galaxies is significantly different than locally , having a clear peak at @xmath1 instead of a nearly flat distribution .
the fall - off at small axial ratio occurs at a higher value than locally , indicating thicker disks by a factor of @xmath2 .
the fall - off at high axial ratio could be from intrinsic triaxial shapes or selection effects .
inclined disks should be more highly sampled than face - on disks near the surface brightness limit of a survey .
simple models and data distributions demonstrate these effects . the decreased numbers of obvious spiral galaxies at high redshifts could be partly the result of surface brightness selection . |
how do galaxies get their gas is a key question to better understand their formation and evolution . in the conventional picture ,
the cold gas is accreted onto the dark matter halo , and is shock heated to the virial temperature of the halo .
the hot gas then loses pressure support due to radiative cooling , then settles into a centrifugally supported disc and begins to form stars @xcite .
this picture has been supported by early simulations which have shown that a substantial fraction of cold gas was indeed shock heated after accretion @xcite .
a number of theoretical and simulation - based works in the past decade , however , have found two distinct modes of gas accretion , named as hot and cold accretion @xcite .
cold gas flowing along dark matter filaments can be directly accreted onto the central galaxy without being shock heated , cold streams are the main source of gas supply in galaxies especially in low - mass halos of @xmath2 at high redshift @xcite . simulations also predict that these cold streams , being optically thick lyman - limit systems with low metallicities , could be observable via absorption lines against background - objects @xcite .
@xcite proposed that cold - flow streams will produce an orbiting circum - galactic component of cool gas with high angular momentum , which might be observable as well .
however , the existence of cold streams has not been confirmed by solid observational results .
observations from @xcite , showed outflow rather than inflow of cold gas in the circum - galactic medium .
@xcite argued that the covering factor might be smaller than @xmath3 at @xmath1 , so the observation of cold streams is very difficult .
very recently , several observations @xcite reported distinct signatures of strong metal poor absorbing systems in the vicinity of galaxies at @xmath4 by observing background quasars
. features of these systems are broadly consistent with the prediction of accreted cold gas in simulations . however , to confirm the existence of cold streams , information on the morphology of the cold gas is needed .
if cold streams do appear at some epoch , the velocity shear between them and hot gas will drive kelvin - helmholtz ( k - h ) instability . when the contact surface between cold and hot gas is not aligned with the gravity gradient
, rayleigh - taylor ( r - t ) instability will also grow .
in addition , thermal evaporation will occur .
another source of instability for the cold flows may come from the very high merger rate of dark matter haloes predicted in the hierarchical scenario , especially at high redshift @xcite .
all these possible sources of perturbation naturally raise the question : how long could the cold streams survive against such instabilities ? in this letter
we make a first attempt in trying to determine the stability of cold flows .
we employ high resolution hydrodynamical eulerian simulations combined with a simplified set - up that allows us to better understand the impact of all these different sources of instability .
the letter is organized as follows : we present our numerical methods and cold flow model in 2 .
hydrodynamical simulation results are shown in 3 .
discussion and conclusion are given in 4 .
we assume that the cold streams are penetrating through a hot gaseous halo which is in hydrostatic equilibrium with the gravitational potential of the hosting dark matter halo .
the dark matter halo has a mass of @xmath5 and we assume a nfw profiles @xcite with a concentration parameter of 10 following @xcite .
all the cosmological parameters are set in agreement with the wmap5 results @xcite .
the hot halo gas is initially set to be isothermal and in hydrostatic equilibrium under the gravitational potential of the dark matter halo .
the self gravity of gas is not included .
the central ( stellar ) galaxy component is not considered as we are mainly interested in the hydrodynamical evolution of cold streams within the virial radius .
the temperature of the hot halo is set to @xmath6 , and we assume a poly - tropic index @xmath7 polytropic . finally we extend the gas distribution to 1.5 times the virial radius of the dark matter halo , ( to what we call the boundary radius @xmath8 ) , in order to reduce any possible boundary effect .
the hot gas accounts for @xmath9 of the total mass within @xmath10 .
the morphology , temperature , density and velocity of cold streams are set according to recent theoretical and simulation works @xcite .
@xcite performed a very broad statistical study of the cold streams velocity and temperature as function of distance from the halo center for dark matter haloes with @xmath11 at @xmath1 . according to their results we assume the velocity of cold streams at @xmath12 to be @xmath13 .
the cold stream is firstly assumed to be a simple cone geometry , following simulations results @xcite .
the radius of the stream at the boundary @xmath12 is given by @xmath14 where @xmath15 is the stream opening angle and could be derived from the covering factor @xmath16 as @xmath17 , where @xmath18 is the number of cold streams in the halo , we adopt the value of @xmath19@xcite . on the other hand the covering factor
@xmath16 is more difficult to estimate .
@xcite predicted that gas at temperature @xmath20 with column densities @xmath21 should cover @xmath22 the area at radii from 20 to 100 kpc in haloes with @xmath23 at @xmath24 .
@xcite measured the covering factor for a milky way type progenitor lyman - break galaxy in a cosmological simulation and found a smaller value , which is @xmath25 for the lyman limit systems and @xmath26 for the damped ly@xmath27 absorbers within @xmath10 .
we adopt a value of @xmath28 within @xmath12 , in agreement with recent studies by @xcite . therefore , @xmath29
. the density of cold gas @xmath30 at @xmath12 can be determined by the accretion rate @xmath31 , @xmath32 , the velocity and the mass ratio of cold accretion mode as , @xmath33 where , @xmath34 is the cold accreted gas mass fraction . using the so called extended press & schechter approach @xcite for structure formation , @xcite derived that the growth rate of the baryonic mass at the virial radius should be @xmath35 where @xmath36 , and @xmath37 is the baryonic fraction in the haloes in units of the cosmic mean . for a halo with @xmath38 ,
it is about @xmath39 at @xmath1 .
however , @xcite found a mean value of @xmath40 and @xmath41 for the total and cold baryonic mass accretion respectively . for our selected cold accretion rate , @xmath42 , and the number of cold streams , @xmath43 ,
the net accretion rate for one single cold stream is @xmath44 . in our simulation
we only model one such cold stream .
the density of the cold stream at the boundary , @xmath45 is @xmath46 obtained from eq.2 .
the temparature of cold gas is set to @xmath47 , which is almost in thermal pressure quilibrium . in fig.1
we give a schematic of the initial condition , and the simulation parameters are listed in table [ table ] .
the satellite is described by a single , mass varying particle with an initial mass of @xmath48 @xmath49 , leading to a merger mass ratio of 1/33 .
the orbit of the satellite is predetermined and extracted from a similar high resolution n - body simulation in @xcite .
the initial position of the satellite is at @xmath12 , locating at the opposite side of the cold accretion with respect to the halo center .
the initial radial and tangential velocity of the satellite are respectively set to @xmath50 and @xmath51 times of the virial velocity of the main halo , @xmath52.the mass evolution of the satellite is also extracted , giving @xmath53 at 2 gyr .
finally we assume the orbit plane of the satellite is co - planar with the cold stream plane .
such a choice is motivated by results from n - body simulations @xcite , and by recent observation of a possible `` cosmic alignment '' of galactic satellites in m31 @xcite .
a sketch of our initial configuration and the satellite orbit are presented in figure [ figure1 ] .
.parameters used in the hydrodynamic simulation [ cols= " < , < " , ] , the dashed curve is the orbit of the satellite and the red circle mark the position of the satellite at 2 gyrs.,scaledwidth=32.0% ] to track the evolution of cold streams , we run hydrodynamical simulations with a fixed grid code using the positive preserving weno scheme @xcite , which can effectively keep positive density and pressure in hypersonic regions , and can well capture the motions of the baryonic gas in cosmological simulations @xcite .
simulations are performed in a cubic box of 300 kpc , with a number of grid cells @xmath54 , i.e , a resolution of @xmath55 @xmath56 .
analytical gravitational potentials based on nfw profiles are given at each grid cell initially .
we impose hydrostatic equilibrium for the boundary condition except for the stream .
density and velocity in the ghost grid cells outside the boundary of the simulation box are assumed to be zero - gradient along the radial direction and the pressure is given by a first - order differentiation of the hydrostatic equilibrium equation .
radiative cooling is followed by using the cooling function given in @xcite .
recent simulation works @xcite showed that the metallicity abundance ranges from @xmath57 to @xmath58 @xmath59 for the cold gas .
we adopt the value @xmath60 , and further assume the hot gas has the same value for the sake of simplicity .
thermal conduction is also included with the unsaturated regime assumption , and the thermal conductivity is set to @xmath61 @xcite , with @xmath62 , where @xmath63 is the coulomb logarithm , which is only weakly dependent on @xmath64 and @xmath65 ( @xmath66 in our simulation ) . from our model
set up , the cold gas is injected into the simulation box within an aperture radius of @xmath67 kpc at @xmath12 with a constant rate of @xmath44 , with the density , velocity and temperature given in table.1 .
the constant inflow rate lasts for the whole simulation run , i.e. , 4 gys .
it is found that a stable cold stream towards the halo central region is soon established under the gravity of the host halo , with time less than 1.5 gyr ( top left panel of fig.2 ) .
note that for the cold stream we assume an impact parameter @xmath68 respect to the halo center , as suggested by cosmological simulations @xcite .
we perform two sets of simulations , with and without the inclusion of the merging satellite , which are referred to as _ col - mer-512 _ and _ col - ref-512 _ respectively .
the latter simulation is intended to track the effect of thermal conduction and k - h instability , while the former is used to study the impact of an orbiting satellite on the stability of cold flows .
fig . [ figure2 ] presents the slices of density and temperature distribution of gas at 1.5 , 2.5 , 3.0 and 4.0 gyrs in the simulations without / with the satellite in the upper / lower panels , respectively .
the evolution before 1.5 gyrs , during the formation of the stream , is practically the same in the two simulations and it is not shown . in the simulation without satellite ,
only subtle changes in morphology , density and temperature happen to the stream after its formation .
thermal conduction results a thin transition layer between the two gas phases and evidence for k - h instability is hard to find .
the cold gas could flow into the central region ( @xmath69 ) with a velocity as high as @xmath70 .
after passing through the inner region , a small fraction of gas flows out of @xmath71 . in a full cosmological simulation
this gas will be retained by the central galaxy and used for star formation . in the _ col - mer-512 _ simulation
, obvious disturbances have been triggered by the satellite at 2.5 and 3.0 gyrs .
there is practically no flow of cold gas inside @xmath71 at 3.0 gyr .
interestingly enough the cold stream is re - established after 4.0 gyr , when the satellite is far away from the stream . in order to quantify the effect of satellite on the cold stream , we present in fig .
[ figure3 ] the evolution of the net accretion rate of cold gas through a spherical shell at @xmath10 and at @xmath71 .
gas with temperature below @xmath72 k is defined as cold .
the _ col - ref-512 _ run shows an almost constant accretion rate at both radius .
this implies that thermal conduction and gravitational heating are able to convert only a small fraction of the incoming cold gas to the hot phase .
the situation is very different for the _ col - mer-512 _ run .
the accretion rate at both radius shows sharp drops delayed by a couple of hundred myr with respect to the crossing of the stream by the satellite
. the accretion rate at at 0.5 @xmath10 is practically suppressed for a period of @xmath73 gyrs .
this very reduced cold accretion is clearly connected with the close passage of the satellite at @xmath74 gyrs , marked by the red line in the figure . despite the very destructive effect of the satellite on the stream ,
the cold flow is able to re - build itself quite rapidly in less than 0.3 gyrs .
this shows that in the presence of a continuous accretion of cold gas from cosmological filaments , even if disturbed by orbiting satellites , cold flows are able to regenerate themselves quite rapidly . ) . in the middle and bottom panels ,
the red line indicate the distance between the satellite and the cold stream.,scaledwidth=38.0% ]
in this letter , inspired by recent simulation works and observations , we have studied the stability of cold streams penetrating through hot gaseous halos in dark matter halos with mass @xmath75 at @xmath1 by hydrodynamic simulations in a simplified scenario . without external perturbations ( e.g. satellites ) ,
the cold gas is able to flow into the inner region of the central galaxy .
the k - h instability and thermal conduction are not able to inhibit the cold gas accretion .
k - h instability might have been damped by the supersonic motion of the gas due to the low cooling rate @xcite .
when we consider the disturbing action of a satellite ( with a mass ratio of 1:33 ) merging with the central halo , the accretion rate of cold gas at half of the virial radius experiences a dramatic drop for @xmath76 0.5 gyrs . in this period the supply of cold gas into the central region decreases by more than @xmath77 .
cosmological simulations have shown that the merger rate at @xmath1 for our satellite / host mass combination is about two mergers per gyr@xcite .
however , the cross section between isotropically distributed stream and satellites can be crudely estimated by @xmath78 where @xmath79 and @xmath80 are the volume of cold stream and halo respectively .
therefore , the realistic distribution of cold stream and satellite would play an important role . even in the case of a direct interaction between the satellite and the stream ,
the flow of cold gas towards the halo center is able to re - establish itself in less than 0.3 gyrs , under the assumption of a continuous flow of cold gas from cosmic filaments .
what we present in this letter is a pilot study for the stability of cold flows .
we plan to improve and extend it in forthcoming works by increasing the resolution of the simulation , considering different satellite orbits and impact parameters , adding the self gravity of the gas and comparing different hydrodynamical approaches ( see @xcite for the effect of the hydro solver on the cold flow stability in cosmological simulations ) .
meanwhile , we expect that further observations may reveal more information about both cold and hot gas in the halo , and more details on galaxy formation and growth .
we acknowledge useful discussions with greg stinson and salvo cielo , we thank kester smith for revising the text .
the simulations are performed at the hyper performance computing center of nanjing university .
lw and wsz thank the mpia for its hospitality during the preparation of this work .
wsz thanks the support from the china postdoctoral science foundation and the fundamental research funds for the central universities .
wsz , llf and xk are supported by the national natural science foundation of china under grant 11203012 , 11273060 , 91230115 , 11333008 , 11073055 .
lw , avm , jc and xk acknowledge support from the mpg - cas through the partnership program between the mpia group lead by a. macci and the pmo group lead by x. kang .
benson a.j . ,
frenk c.s . ,
baugh c.m .
, cole s. , lacey c.g
. , 2001 , mnras , 327 , 1041 birnbiom y. , dekel a. , 2003 , mnras , 345 , 349 bouch n. , murphy m.t .
et al . , 2013 , science , 341 , 50 chang j. , macci a.v . , kang x. , 2013 , mnras , 431 , 3533 crighton n. , hennawi j. , prochaska j. , 2013 , apjl , 776 , l18 danovich m. , dekel a. , hahn o. , teyssier r. , 2012 , mnras , 422 , 1732 dekel a. , birnbiom y. , 2006 , mnras , 368 , 2 dekel a. , birnbiom y.,engel g. et al . , 2009 , nature , 457 , 451 fall s.m . ,
efstathiou g. , 1980 , mnras , 193 , 189 fakhouri o. , ma , c - p . , boylan - kolchin m. , 2010 , mnras , 406 , 12 faucher - gifure c.a . , kere d. , 2011 , mnras , 412 , l118 fumagalli m. , prochaska j. , kasen d. , dekel a. , ceverino d. , primack j. r. , 2011 , mnras , 418 , 1796 fumagalli m. , meara j. m. , prochaska j. x. , worseck g. , 2013 , apj , 775 , 78 goerdt t. , dekel a. , sternberg a. , gnat o. , ceverino d. , 2012 , mnras , 424 , 2292 helly j.c . , cole s. , frenk c.s . ,
baugh c.m . ,
benson a. , lacey c. , pearce f.r . , 2003 , mnras , 338 , 913 ibata r.a . , lewis g.f .
et al . , 2013 ,
nature , 2013 , 493 , 62 kere d. , katz n. , weinberg d.h .
, dave r. , 2005 , mnras , 363 , 2 kere d. , hernquist l. , 2009 , apj , 700 , l1 nelson , d. , et al .
2013 , mnras , 429,3353 ocvirk p. , pichon c. , teyssier r. , 2008 , mnras , 390 , 1326 spitzer l. , 1962 , physics of fully ionized gases ( new york : interscience ) steidel c. , erb d. shapley a. et al . , 2009 ,
apj , 700 , l1 stewart k.r . ,
kaufmann t. , bullock j.s . ,
barton e.j . , et al . , 2011 ,
apj , 738 , 39 sutherland r.s . ,
dopita m.a . , 1993 ,
apjs , 88 , 253 vietri m. , ferrara a. , miniati f. , 1997 , apj , 483 , 262 white simon d. m. , frenk carlos s. , 1991 , apj , 52 , 379 van de voort f. , schaye j. , 2012 , mnras , 423 , 2991 zentner a. r. , kravtsov a. v. , gnedin o. y. , klypin a. a. , 2005 , apj , 629 , 219 zhang x. , shu c.w . , 2010 , journal of computational physics , 229 , 8918 zhu , w.s . , feng , l.l .
, xia , y.h . ,
shu , c .- w .
, gu , q.s . , & fang , l. z. , 2013 , apj , 777 , 48 | we use high - resolution eulerian simulations to study the stability of cold gas flows in a galaxy size dark matter halo ( @xmath0 ) at redshift @xmath1 .
our simulations show that a cold stream penetrating a hot gaseous halo is stable against thermal convection and kelvin - helmholtz instability .
we then investigate the effect of a satellite orbiting the main halo in the plane of the stream .
the satellite is able to perturb the stream and to inhibit cold gas accretion towards the center of the halo for 0.5 gyr .
however , if the supply of cold gas at large distances is kept constant , the cold stream is able to re - establish itself after 0.3 gyr .
we conclude that cold streams are very stable against a large variety of internal and external perturbations .
[ firstpage ] galaxies : formation ; cooling flows ; numerical ; galaxies : interactions |
radio galaxies and quasars produce powerful relativistic plasma jets , accelerating particles up to tev energies .
jets are not unique to agn , but are also observed in other objects like in x - ray binaries , micro - quasars , young stellar objects and probably also in gamma - ray bursters . in modern theory ,
the launching and acceleration of the relativistic jets is explained via mhd processes , with magnetic fields either attached to the rotating accretion disk or directly linked to the ergosphere of the central and possibly rotating super - massive black hole ( e.g. * ? ? ?
in order to better understand the process of jet formation , a study of the inner most regions of compact radio sources is necessary .
vlbi at short millimeter wavelengths ( mm - vlbi ) provides , via its unrivaled angular resolution of only a few ten micro - arcseconds ( at @xmath1 mm ) , an almost unique tool to test and discriminate between the different theoretical models .
in addition to the high spatial resolution ( corresponding to @xmath2 schwarzschild radii for nearby sources like sgra * or m87 ) , the high observing frequency allows to image emission regions , which are self - absorbed and therefore not observable at longer wavelengths . in the following
we present some results from global 3mm - vlbi performed with the global millimeter vlbi array ( gmva ) at 86ghz .
vlbi observations at shorter wavelengths ( @xmath3 mm , @xmath4ghz ) are possible but are still limited by the number of available mm - telescopes ( typically @xmath5 ) , their sensitivity and the limited observing bandwidth ( see e.g. * ? ? ?
* and references therein ) . to date , it is therefore not yet possible to produce a reliable map of a radio source at these shorter wavelengths .
the gmva combines the sensitive and mm - capable telescopes in europe ( 100 m effelsberg , 6x15 m plateau de bure , 30 m pico veleta , 20 m onsala , & 14 m metshovi ) with the vlba .
the gmva is open to the scientific community and calls for observing proposals twice per year ( proposal deadlines : feb .
@xmath6 , oct .
@xmath6 ) .
it enables imaging of compact radio sources at an angular resolution of @xmath7as .
the transatlantic detection sensitivity ( @xmath8 ) ranges between @xmath9mjy on the most sensitive baselines ( for 512mbps ) .
this should be compared with the @xmath10mjy baseline sensitivity of the stand - alone vlba . in a typical 12hr gmva observing run , an image rms of @xmath11mjy is reached . .
vlbi surveys at 86ghz : col.1 gives the reference , col.2 the number of antennas , col.3 the typical detection sensitivity ( @xmath8 ) in [ jy ] , col.4 the number of observed sources and col.5 the percentage of detected objects ( detection rate ) . [ cols="<,^,^,<,^",options="header " , ] ll & + + @xmath12:&inhomogeneous array ( 6x10m+9x6 m ) + @xmath13:&10 different antennas with elevations ranging from sea level to 4100 m + @xmath14:&present value , aim is @xmath15 m + @xmath16:&after upgrade + krichbaum , t.p .
, agudo , i. , bach , u. , witzel , a. , zensus , j.a . , in : new developments in vlbi science and technology , 8th evn symposium , ed .
a. marecki et al . , torun observatory , poland , 2007 , in press ( astro - ph/0611288 ) | we study the most central regions of agn jets with an angular resolution of tens of micro - arcseconds using vlbi at millimeter wavelengths ( mm - vlbi ) .
we present and discuss a new 86ghz vlbi survey of compact radio sources .
we show new high dynamic range images of two nearby radio galaxies ( 3c120 and m87 ) . in m87 the size of the compact vlbi core ( the jet base ) is @xmath0 schwarzschild radii .
future mm - vlbi observations at 1 mm and shorter wavelengths should lead to images of galactic and extragalactic radio sources with a spatial resolution down to a few schwarzschild radii of the central super massive black holes . to achieve this ,
the participation of large and sensitive millimeter and sub - millimeter telescopes in vlbi is essential . |
in this paper , we are investigating variational methods and partial differential equations for filtering displacement errors in imaging data .
such types of errors appear when measurement data are _ sampled _ erroneously . in this work we consider measurement data @xmath0 , which are considered perturbations of an ideal function @xmath1 at random sampling locations @xmath2 : that is , @xmath3 a particular case of a displacement error @xmath4 appears in computerized tomography ( here the @xmath5-component denotes the @xmath6-ray beam direction ( @xmath7 below ) ) when the angular sampling is considered erroneous . in this case
the recorded data are @xmath8(\theta+d_1(\theta),l).\ ] ] here @xmath9 $ ] denotes the _ radon transform _ or _ x - ray transform _ of the function @xmath10 , and @xmath7 and @xmath11 denote the beam direction and beam distance , respectively .
displacement errors of the form @xmath12 are jittering errors , and the filtering and compensation of such has been considered in @xcite .
our work is motivated by @xcite , where partial differential equations for denoising image data destructed by _
general _ sampling errors of the form @xmath13 have been stated .
the nonlinear evolution equations have been derived by mimicking a convex semi - group for a non - convex energy .
the pdes from @xcite revealed properties similar to the mean curvature motion equation @xcite . in comparison to @xcite , here , we are considering displacement errors in the @xmath5-component only .
the paper is organized as follows : in section [ sec : regularization ] , we review the state of the art of non - convex regularization models for sampling error corrections : in particular , we comment on algorithms for recovering different types of displacements in a discrete setting . in section [ sec :
flows ] , we consider nonlinear filtering flows motivated from non - convex regularization energies . for these flows we present numerical experiments , which suggest particular properties of the pdes .
finally , we present an application of the novel filtering techniques for correcting tomographical image data with errors in the beam directions .
we study the following problem : let @xmath14 or @xmath15 be fixed . given noisy image data @xmath0 , the goal is to simultaneously recover the ideal function @xmath1 the displacement @xmath16 satisfying .
figure [ fig : displacement ] shows the effect of displacement errors @xmath17 and @xmath18 on some image data , respectively .
the two displacement errors result in image data with orthogonal visual perturbations . and @xmath19 ) .
_ bottom : original and jittered image .
the two different kinds of errors results appear as visually complementary data perturbations .
this observation is the basis of regularization methods , where jittering is corrected by regularizing in @xmath20 direction and tomographic problems are corrected by filtering in @xmath5 direction.___,title="fig:",scaledwidth=45.0% ] and @xmath19 ) . _
bottom : original and jittered image .
the two different kinds of errors results appear as visually complementary data perturbations .
this observation is the basis of regularization methods , where jittering is corrected by regularizing in @xmath20 direction and tomographic problems are corrected by filtering in @xmath5 direction.___,title="fig:",scaledwidth=45.0% ] + and @xmath19 ) . _
bottom : original and jittered image .
the two different kinds of errors results appear as visually complementary data perturbations .
this observation is the basis of regularization methods , where jittering is corrected by regularizing in @xmath20 direction and tomographic problems are corrected by filtering in @xmath5 direction.___,title="fig:",scaledwidth=45.0% ] and @xmath19 ) . _
bottom : original and jittered image .
the two different kinds of errors results appear as visually complementary data perturbations .
this observation is the basis of regularization methods , where jittering is corrected by regularizing in @xmath20 direction and tomographic problems are corrected by filtering in @xmath5 direction.___,title="fig:",scaledwidth=45.0% ] to this end we consider the following optimization problem : @xmath21 here @xmath22 denotes some proper quasi - norm or norm of functions @xmath23 and the choice of @xmath24 depends on the a - priori information on regularity of the function @xmath1 .
below , we are considering _ discrete _ and _ continuous _ optimization formulations of problem .
* in the discrete optimization problem , @xmath25 denotes the pixel displacements of columns ( @xmath14 ) and rows ( @xmath26 ) , respectively .
the image data @xmath1 can be represented as a matrix with non - negative integer values , that is @xmath27 , where each matrix entry denotes the discrete image intensity of the image data at a position @xmath28 .
moreover , the derivatives are considered in the sense of finite differences . * in a continuous formulation , @xmath29 and the image is considered a function @xmath30 . for @xmath26 , the discrete optimization problem has been investigated in @xcite and the continuous optimization problem has been studied in @xcite .
the computational complexity of the discrete optimization algorithms varies significantly for solving the discrete problem with @xmath14 and @xmath26 .
* nikolova @xcite introduced a highly efficient optimization algorithm with exhaustive search for the case @xmath26 .
the complexity is @xmath31 for @xmath27 . * for @xmath14
the discrete optimization problem is an _
assignment _ problem .
even the simplified problem of alignment of the columns in each of the @xmath32 non - intersecting sub - blocks has already an exponential complexity @xmath33 by exhaustive search .
+ we note that the complexity of assignment problems depends on the properties of the given cost functionals . for linear assignment problem ( see e.g. @xcite ) , the hungarian algorithm @xcite has a complexity @xmath34 .
however , nonlinear assignment problems @xcite , such as problem , are usually _ np - hard_. the discrete optimization algorithm for solving problem can be used to correct for large displacement errors .
small ( including subpixel ) displacement errors can be corrected for by using partial differential equations : if the displacement @xmath16 is small , following @xcite , we consider a first order approximation of the continuous data @xmath30
. then @xmath35 such that @xmath36 we aim for simultaneous minimizing the displacement error @xmath16 and maximizing the smoothness of @xmath1 by the minimization of the functional @xmath37 with some fixed parameter @xmath38
. our particular choice of the regularization functional is motivated from the structure of the data ( see figure [ fig : displacement ] ) , where we observed that correcting for line jittering requires regularization in @xmath20-direction and angular displacements require regularization in @xmath5-direction .
the functional @xmath39 is non - convex with respect to @xmath1 and has a singularity when @xmath40 vanishes . for the practical minimization we consider an approximation consisting of a sequence of convex minimization problems : @xmath41 where @xmath42 is a small real number , and @xmath43 is the set of convex functionals defined by @xmath44 in the following we give a convergence result inspired from @xcite .
let @xmath45 , @xmath46 , @xmath47 , and let @xmath48 be the sequence of minimizers from . * the sequences @xmath49 and @xmath50 both are monotonically decreasing .
* if there exist a constant @xmath51 and if @xmath52 , then @xmath53 from the definition of @xmath54 in it follows that @xmath55 is proper , strictly convex , and lower semi - continuous for every @xmath56 . thus there exists an unique minimizer @xmath57 minimizing @xmath55 .
we are able to infer the following inequalities @xmath58 and @xmath59 which shows that both @xmath49 and @xmath50 are non - negative and monotonically decreasing .
for the second statement , @xmath60 from the uniform boundedness assumption of @xmath61 it follows that @xmath62 since the sequence @xmath63 is bounded from below and monotonically decreasing it follows that @xmath64 .
@xmath65 identifying @xmath66 , the formal optimality condition for characterizes the solution of by @xmath67 on the other hand , if @xmath68 is relatively large , the estimate of the displacement by the error measure in is unrealistic , and a least squares error measure @xmath69 might be more efficient . to be able to compensate for relatively large and small displacement errors simultaneously , we , therefore , propose to use the geometric mean of the two error measures , which is @xmath70 in this case we end up with the following variational model @xmath71 again identifying @xmath66 , the associated optimality condition for the relaxed functional is @xmath72
let @xmath73 , @xmath46 , @xmath45 , @xmath74 be fixed and assume that @xmath0 is given .
@xmath75 solving ( corresponding to @xmath76 ) , ( corresponding to @xmath77 ) can be considered a numerical approximation of the flow @xmath78 at time @xmath79 . here
@xmath80 and @xmath81 denotes the derivative of @xmath1 with respect to @xmath82 .
we also can consider as the flow according to the non - convex functional @xmath39 ( which is depended on @xmath83 ) . in practical simulations
the unbounded domain @xmath84 is replaced by @xmath85 and the flow is associated with boundary conditions : @xmath86 the case of @xmath26 has been considered in @xcite . when @xmath14 , the right hand side of only involves partial derivatives of @xmath1 with respect to @xmath5 , such that it reduces to a system of independent equations defined for the functions @xmath87 for every @xmath88 . in the following , we present some numerical simulations with .
* for @xmath14 reads as follows @xmath89 + figure [ fig : diffusion ] shows numerical simulations of for different choices of @xmath90 and @xmath91 with @xmath76 . in all test cases the initial data
@xmath0 is a function representing a narrow white strip on a black background .
we visualize the solutions at @xmath92 with identical time unit .
we observe diffusion of the white strip for all choices of @xmath90 and @xmath91 except in the case @xmath93 , with @xmath76 .
+ we emphasis that when @xmath94 and @xmath76 reads as follows @xmath95 of which the differential operator on the right - hand side is a one dimensional _
p - laplacian _ with @xmath96 ( see @xcite for some regularity properties , where they have a general result for _ p - laplacian _ in @xmath97 ) . hence the equation is nothing but a system of independent one dimensional _
p - laplacian _ flow , of which the equilibrium ( stationary point ) is a function of the form @xmath98 .
+ . from top to bottom , and from left to right , the images are corresponding to the initial value @xmath0 and the results of the evolution with @xmath76 and various parameters : @xmath99 ; @xmath100 ; @xmath93 ; @xmath101 in . , title="fig:",scaledwidth=30.0% ] . from top to bottom , and from left to right , the images are corresponding to the initial value @xmath0 and the results of the evolution with @xmath76 and various parameters : @xmath99 ; @xmath100 ; @xmath93 ; @xmath101 in . , title="fig:",scaledwidth=30.0% ] + . from top
to bottom , and from left to right , the images are corresponding to the initial value @xmath0 and the results of the evolution with @xmath76 and various parameters : @xmath99 ; @xmath100 ; @xmath93 ; @xmath101 in .
, title="fig:",scaledwidth=30.0% ] . from top to bottom , and from left to right , the images are corresponding to the initial value @xmath0 and the results of the evolution with @xmath76 and various parameters : @xmath99 ; @xmath100 ; @xmath93 ; @xmath101 in . , title="fig:",scaledwidth=30.0% ] . from top to bottom , and from left to right , the images are corresponding to the initial value @xmath0 and the results of the evolution with @xmath76 and various parameters : @xmath99 ; @xmath100 ; @xmath93 ; @xmath101 in . , title="fig:",scaledwidth=30.0% ] + . from top to bottom , and from left to right , the images are corresponding to the initial value @xmath0 and the results of the evolution with @xmath76 and various parameters : @xmath99 ; @xmath100 ; @xmath93 ; @xmath101 in . , title="fig:",scaledwidth=32.0% ] . from top to bottom , and from left to right , the images are corresponding to the initial value @xmath0 and the results of the evolution with @xmath76 and various parameters : @xmath99 ; @xmath100 ; @xmath93 ; @xmath101 in . , title="fig:",scaledwidth=32.0% ]
+ . from top to bottom , and from left to right , the images are corresponding to the initial value @xmath0 and the results of the evolution with @xmath76 and various parameters : @xmath99 ; @xmath100 ; @xmath93 ; @xmath101 in . , title="fig:",scaledwidth=32.0% ] . from top to bottom , and from left to right , the images are corresponding to the initial value @xmath0 and the results of the evolution with @xmath76 and various parameters : @xmath99 ; @xmath100 ; @xmath93 ; @xmath101 in . , title="fig:",scaledwidth=32.0% ]
from top to bottom , and from left to right , the images are corresponding to the initial value @xmath0 and the results of the evolution with @xmath76 and various parameters : @xmath99 ; @xmath100 ; @xmath93 ; @xmath101 in . , title="fig:",scaledwidth=32.0% ] * for @xmath26 reads as follows @xmath102 we investigate the long time behavior of the solution of equation initialized with some curved interface data @xmath0 .
the numerical results are presented in figure [ fig : lineflow ] .
we find that the curved interface evolves into a vertical line ( for all cases of @xmath90 and @xmath91 , with @xmath76 , we have tested ) , which we assume to be a general analytical property of the pdes .
now , we are applying the displacement correction methods for problems arising in tomography .
as data we consider angular disturbed sinograms of the radon transform of 2d images and the final goal is to reconstruct the corresponding attenuation coefficients .
if the sinogram is recorded with angular perturbations , without further processing , application of the inverse radon transform may produce outliers in the reconstruction ( see figure [ fig : phanton ] ) . )
.,title="fig:",scaledwidth=50.0% ] ) .,title="fig:",scaledwidth=50.0% ] as a test image we use the shepp - logan phantom of size @xmath103 , and discretize the angular axis of the sinogram domain @xmath104 with uniform step size @xmath105 .
the synthetic data are generated by evaluating the radon transform along the line in direction @xmath106 .
the ( inverse ) radon transforms are implemented with the matlab toolbox . in the first series of experiments ( see figure [ fig : s_angular_correction ] ) , we allow the random perturbations on the beam directions to be @xmath107 $ ] , which is relatively small . in the example shown in figure [ fig : s_angular_correction ] these sampling errors do not cause significant mismatch in the reconstruction
. + + although the flows in are suitable for correcting small displacement errors , they may not be very qualified for the data with larger displacement errors and with additive noise .
in the latter case , the recorded sinogram is considered to be @xmath108(\theta+d_1(\theta),l)+\eta,\ ] ] where @xmath109 denotes some additive noise . here
the filtering by the flow defined in with @xmath77 outperforms the filtering by the flow with @xmath76 . in the numerical experiments , we tested with an example of the angular perturbation @xmath110 $ ] ( see figure [ fig : displacement ] ) .
for @xmath77 we show the filtering by with @xmath111 and @xmath112 .
that is , we use the equation @xmath113 for filtering .
our numerical results are reported in figure [ fig : l_angular_correction ] and figure [ fig : l_noisy_angular_correction ] , which is based on the displacement error shown in figure [ fig : displacement ] . as a comparison
, we also show the results when problem is considered a discrete optimization problem and is solved with a heuristic discrete optimization algorithm , which is generalized from @xcite .
the right one on the top of figure [ fig : l_angular_correction ] shows the final result , having no preference against standard fbp . the bottom images in figure [ fig : l_angular_correction ] show the results which are obtained by applying the nonlinear flow for correcting the unknown angular perturbations . in this example
, the nonlinear flows have a better performance in comparing with the heuristic discrete optimization algorithm . from the results shown in figure [ fig : l_noisy_angular_correction ] , we find that the proposed equation is able to correcting the angle displacement error and denoise simultaneously .
in this paper , we have considered two families of nonlinear flows for correcting two different kinds of displacement errors .
our numerical analysis revealed interesting properties of the flows , which should be confirmed by further theoretical studies in the future .
we have also presented some applications to tomography , where the novel pdes are able to correct for angular displacement errors .
some other application area of our methods are electron tomography in single particle analysis @xcite .
the work of os has been supported by the austrian science fund ( fwf ) : geometry and simulation , project s11704 ( variational methods for imaging on manifolds ) , and interdisciplinary coupled physics imaging , project p26687-n25 .
g. dong , a.r .
patrone , o. scherzer , and o. ktem .
infinite dimensional optimization models and pdes for dejittering . in _ scale space and variational methods in computer vision 5th international conference , ssvm 2015 , lge - cap ferret , france , may 31 - june 4 , 2015 , proceedings _ , pages 678689 .
springer .
f. lenzen and o. scherzer . a geometric pde for interpolation of @xmath114-channel data . in _
ssvm 09 : proceedings of the second international conference on scale space and variational methods in computer vision _ ,
pages 413425 , berlin , heidelberg , 2009 .
springer - verlag .
s. voss .
heuristics for nonlinear assignment problems . in p.
m. pardalos , and l. pitsoulis , editors , _ nonlinear assignment problems , algorithms and applications _ , volume 7 of the series combinatorial optimization , pages 175215 , 2000 .
kluwer academic publishers . | in this paper we derive nonlinear evolution equations associated with a class of non - convex energy functionals which can be used for correcting displacement errors in imaging data .
we study properties of these filtering flows and provide experiments for correcting angular perturbations in tomographical data . |
let @xmath0 be an oriented surface of finite type , i.e. @xmath0 is a closed surface of genus @xmath1 from which @xmath2 points , so - called _ punctures _ , have been deleted .
we assume that @xmath4 , i.e. that @xmath0 is not a sphere with at most @xmath5 punctures or a torus without puncture
. in particular , the euler characteristic of @xmath0 is negative . then the _ teichmller space _
@xmath6 of @xmath0 is the quotient of the space of all complete hyperbolic metrics of finite volume on @xmath0 under the action of the group of diffeomorphisms of @xmath0 which are isotopic to the identity .
the _ mapping class group _ @xmath7 of all isotopy classes of orientation preserving diffeomorphisms of @xmath0 acts properly discontinuously on @xmath6 with quotient the _ moduli space _ @xmath8 .
a _ geodesic lamination _ for a fixed choice of a complete hyperbolic metric of finite volume on @xmath0 is a _ compact _ subset of @xmath0 foliated into _
simple _ geodesics .
a _ measured geodesic lamination _ is a geodesic lamination together with a transverse translation invariant measure .
the space @xmath9 of all measured geodesic laminations on @xmath0 , equipped with the weak@xmath10-topology , is homeomorphic to @xmath11 where @xmath12 is the @xmath13-dimensional sphere .
the mapping class group @xmath7 naturally acts on @xmath9 as a group of homeomorphisms preserving a radon measure in the _ lebesgue measure class_. up to scale , this measure is induced by a natural symplectic structure on @xmath9 ( see @xcite for this observation of thurston ) , and it is ergodic under the action of @xmath7 . moreover , the measure is _ non - wandering_. by this we mean that @xmath9 does not admit an @xmath7-invariant countable borel partition into sets of positive measure .
if @xmath0 is a once punctured torus or a sphere with 4 punctures , then the teichmller space of @xmath0 has a natural identification with the upper half - plane @xmath14 .
up to passing to the quotient by the hyperelliptic involution if @xmath0 is the once punctured torus , the mapping class group @xmath7 is just the group @xmath15 acting on @xmath16 by linear fractional transformations .
the action of @xmath7 on measured lamination space can in this case be identified with the quotient of the standard _ linear _ action of @xmath17 on @xmath18 under the reflection at the origin ( see the book @xcite of bekka and mayer for more and for references and compare the survey @xcite ) . extending earlier work of furstenberg , in 1978
dani completely classified all @xmath17-invariant radon measures on @xmath18 .
he showed that if such a measure @xmath19 is ergodic under the action of @xmath17 then either it is non - wandering and coincides with the usual lebesgue measure @xmath20 up to scale , or it is _ rational _ , which means that it is supported on a single @xmath17-orbit of points whose coordinates are dependent over @xmath21 .
if the surface @xmath0 is _ non - exceptional _ ,
i.e. if @xmath3 , then the @xmath7-invariant radon measures on @xmath9 which naturally correspond to the rational measures for exceptional surfaces are defined as follows .
a _ weighted geodesic multi - curve _ on @xmath0 is a measured geodesic lamination whose _ support _ is a union of simple closed geodesics .
the orbit of a weighted geodesic multi - curve under the action of @xmath7 is a discrete subset of @xmath9 ( see section 5 for this easy and well known fact ) and hence it supports a ray of invariant purely atomic radon measures which we call _
rational_. this definition coincides with the one given above for a once punctured torus or a forth punctured sphere . for a non - exceptional surface @xmath0 , there are additional @xmath7-invariant radon measures on @xmath9 .
namely , a _ proper connected bordered subsurface _
@xmath22 of @xmath0 is a connected component of the space which we obtain from @xmath0 by cutting @xmath0 open along a collection of disjoint simple closed geodesics .
then @xmath22 is a connected surface with non - empty geodesic boundary and of negative euler characteristic .
if two boundary components of @xmath22 correspond to the same closed geodesic @xmath23 in @xmath0 then we require that @xmath24 contains a connected component which is an annulus with core curve @xmath23 .
let @xmath25 be the space of all measured geodesic laminations on @xmath0 which are contained in the interior of @xmath22 .
the space @xmath26 can naturally be identified with the space of measured geodesic laminations on the surface @xmath27 of finite type which we obtain from @xmath22 by collapsing each boundary circle to a puncture .
the stabilizer in @xmath7 of the subsurface @xmath22 is the direct product of the group of all elements which can be represented by diffeomorphisms leaving @xmath22 pointwise fixed and a group which is naturally isomorphic to a subgroup of finite index of the mapping class group @xmath28 of @xmath27 .
let @xmath29 be a weighted geodesic multi - curve on @xmath0 which is disjoint from the interior of @xmath22 .
then for every @xmath30 the union @xmath31 is a measured geodesic lamination on @xmath0 which we denote by @xmath32 .
let @xmath33 be an @xmath28-invariant radon measure on @xmath26 which is contained in the lebesgue measure class .
the measure @xmath33 can be viewed as a radon measure on @xmath9 which gives full measure to the laminations of the form @xmath32 @xmath34 and which is invariant and ergodic under the stabilizer of @xmath35 in @xmath7 .
the translates of this measure under the action of @xmath7 define an @xmath7-invariant ergodic wandering measure on @xmath9 which we call a _ standard subsurface measure_. we observe in section 5 that if the weighted geodesic multi - curve @xmath29 contains the boundary of @xmath22 then the standard subsurface measure defined by @xmath33 and @xmath29 is a radon measure on @xmath36 . the goal of this note is to show that every @xmath7-invariant ergodic radon measure on @xmath9 is of the form described above . 1 .
an invariant ergodic non - wandering radon measure for the action of @xmath7 on @xmath9 coincides with the lebesgue measure up to scale .
an invariant ergodic wandering radon measure for the action of @xmath7 on @xmath9 is either rational or a standard subsurface measure .
the organization of the paper is as follows . in section 2
we discuss some properties of geodesic laminations , quadratic differentials and the curve graph needed in the sequel . in section 3 we introduce conformal densities for the mapping class group .
these conformal densities are families of finite borel measures on the projectivization @xmath37 of @xmath9 , parametrized by the points in teichmller space .
they are defined in analogy to the conformal densities for discrete subgroups of the isometry group of a hyperbolic space . up to scale , there is a unique conformal density in the lebesgue measure class .
we show that this is the only conformal density which gives full measure to the @xmath7-invariant subset of @xmath37 of all projective measured geodesic laminations whose support is minimal and fills up @xmath0 .
every conformal density gives rise to an @xmath7-invariant radon measure on @xmath9 .
the investigation of invariant radon measure which are not of this form relies on the structural results of sarig @xcite . to apply his results we use train tracks to construct partitions of measured lamination space which have properties similar to markov partitions .
section 4 summarizes those facts about train tracks which are needed for this purpose .
the proof of the theorem is completed in section 5 . in the appendix we present a result of minsky and weiss @xcite in the form needed for the proof of our theorem .
after this paper was posted on the arxiv i received the preprint @xcite of lindenstrauss and mirkzakhani which contains another proof of the above theorem .
in this introductory section we summarize some properties of ( measured ) geodesic laminations , quadratic differentials and the curve graph which are needed later on .
we also introduce some notations which will be used throughout the paper . in the sequel
we always denote by @xmath0 an oriented surface of genus @xmath1 with @xmath2 punctures and where @xmath4 .
a _ geodesic lamination _ for a complete hyperbolic structure of finite volume on the surface @xmath0 is a _ compact _ subset of @xmath0 which is foliated into simple geodesics .
a geodesic lamination @xmath20 is called _ minimal _ if each of its half - leaves is dense in @xmath20 .
thus a simple closed geodesic is a minimal geodesic lamination .
a minimal geodesic lamination with more than one leaf has uncountably many leaves and is called _ minimal arational_. every geodesic lamination @xmath20 consists of a disjoint union of finitely many minimal components and a finite number of isolated leaves .
each of the isolated leaves of @xmath20 either is an isolated closed geodesic and hence a minimal component , or it _ spirals _ about one or two minimal components @xcite .
a geodesic lamination is _ maximal _ if its complementary regions are all ideal triangles or once punctured monogons .
a geodesic lamination _ fills up @xmath0 _ if its complementary regions are all topological discs or once punctured topological discs .
a _ measured geodesic lamination _ is a geodesic lamination @xmath20 together with a translation invariant transverse measure .
such a measure assigns a positive weight to each compact arc in @xmath0 which intersects @xmath20 nontrivially and transversely and whose endpoints are contained in the complementary regions of @xmath20 .
the geodesic lamination @xmath20 is called the _ support _ of the measured geodesic lamination ; it consists of a disjoint union of minimal components @xcite .
the space @xmath9 of all measured geodesic laminations on @xmath0 equipped with the weak@xmath10-topology is homeomorphic to @xmath38 .
its projectivization is the space @xmath37 of all _ projective measured geodesic laminations_. the measured geodesic lamination @xmath39 _ fills up @xmath0 _ if its support fills up @xmath0 .
the projectivization of a measured geodesic lamination which fills up @xmath0 is also said to fill up @xmath0 .
a measured geodesic lamination is called _ uniquely ergodic _ if its support admits a single transverse measure up to scale .
there is a continuous symmetric pairing @xmath40 , the so - called _ intersection form _ , which extends the geometric intersection number between two simple closed curves .
the smooth fibre bundle @xmath41 of all _ holomorphic quadratic differentials of area one _ over the teichmller space @xmath6 of the surface @xmath0 can naturally be viewed as the unit cotangent bundle of @xmath6 for the _ teichmller metric_. the _ teichmller geodesic flow _
@xmath42 on @xmath41 commutes with the action of the mapping class group @xmath7 of all isotopy classes of orientation preserving diffeomorphisms of @xmath0 .
thus this flow descends to a flow on the quotient @xmath43 , again denoted by @xmath42 .
a measured geodesic lamination can be viewed as an equivalence class of _ measured foliations _ on @xmath0 @xcite .
therefore every holomorphic quadratic differential @xmath44 on @xmath0 defines a pair @xmath45 where the _ horizontal measured geodesic lamination _ @xmath46 corresponds to the _ horizontal measured foliation _ of @xmath44 which is expanded under the teichmller flow , and where the _ vertical measured geodesic lamination _
@xmath47 corresponds to the vertical measured foliation of @xmath44 which is contracted under the teichmller flow .
the area of the quadratic differential is just the intersection @xmath48 . for a quadratic differential @xmath49 define the
_ strong unstable manifold _
@xmath50 to be the set of all quadratic differentials @xmath51 whose vertical measured geodesic lamination equals the vertical measured geodesic lamination for @xmath44 .
similarly , define the _ strong stable manifold _ @xmath52 to be the set of all quadratic differentials @xmath51 whose horizontal measured geodesic lamination coincides with the horizontal measured geodesic lamination of @xmath44 .
the _ stable manifold _ @xmath53 and the _ unstable manifold
_ @xmath54 are submanifolds of @xmath41 which project homeomorphically onto @xmath6 @xcite . the sets @xmath55 ( or @xmath56 )
@xmath57 define a foliation of @xmath41 which is invariant under the mapping class group and hence projects to a singular foliation on @xmath58 which we call the _ stable foliation _ ( or the _ strong stable , strong unstable , unstable foliation _ ) .
there is a distinguished family of lebesgue measures @xmath59 on the leaves of the stable foliation which are conditional measures of a @xmath42-invariant borel probability measure @xmath20 on @xmath58 in the lebesgue measure class .
the measure @xmath20 is ergodic and mixing under the teichmller geodesic flow ( see @xcite and also @xcite ) .
every area one quadratic differential @xmath49 defines a singular euclidean metric on @xmath0 of area one together with two orthogonal foliations by straight lines , with singularities of cone angle @xmath60 for some @xmath61 .
this metric is given by a distinguished family of isometric charts @xmath62 on the complement of the zeros ( or poles at the punctures ) of @xmath44 which map the distinguished foliations to the foliation of @xmath63 into the _ horizontal _ lines parallel to the real axis and into the _ vertical lines _
parallel to the imaginary axis .
let @xmath64 be the compactification of @xmath0 which we obtain by adding a point at each puncture .
a _ saddle connection _ for @xmath44 is a path @xmath65 which does not contain any singular point , whose image under a distinguished chart is an euclidean straight line and which extends continuously to a path @xmath66\to \hat s$ ] mapping the endpoints to singular points or punctures .
a saddle connection is _ horizontal _ if it is mapped by a distinguished chart to a horizontal line segment .
the group @xmath67 acts on the bundle @xmath41 by replacing for @xmath49 and @xmath68 each isometric chart @xmath69 for @xmath44 by @xmath70 where @xmath71 acts linearly on @xmath72 .
this preserves the compatibility condition for charts .
the teichmller geodesic flow @xmath42 then is the action of the diagonal group @xmath73 the so - called _ horocycle flow _
@xmath74 is given by the action of the unipotent subgroup @xmath75 the _ curve graph _
@xmath76 of the surface @xmath0 is a metric graph whose vertices are the free homotopy classes of _ essential _ simple closed curves on @xmath0 , i.e. curves which are neither contractible nor freely homotopic into a puncture of @xmath0 . in the sequel we often do not distinguish between an essential simple closed curve and its free homotopy class whenever no confusion is possible .
two such curves are connected in @xmath76 by an edge of length one if and only if they can be realized disjointly .
if the surface @xmath0 is non - exceptional , i.e. if @xmath3 , then the curve graph @xmath76 is connected .
any two elements @xmath77 of distance at least 3 _ jointly fill up @xmath0 _ , i.e. they decompose @xmath0 into topological discs and once punctured topological discs .
the mapping class group naturally acts on @xmath76 as a group of simplicial isometries . by bers theorem
, there is a number @xmath78 such that for every complete hyperbolic metric @xmath79 on @xmath0 of finite volume there is a _ pants decomposition _ of @xmath0 consisting of @xmath80 pairwise disjoint simple closed geodesics of length at most @xmath81 . on the other hand ,
the number of essential simple closed curves @xmath82 on @xmath0 whose hyperbolic length @xmath83 ( i.e. the length of a geodesic representative of its free homotopy class ) does not exceed @xmath84 is bounded from above by a universal constant not depending on @xmath79 , and the diameter of the subset of @xmath76 containing these curves is uniformly bounded as well .
define a map @xmath85 by associating to a complete hyperbolic metric @xmath79 on @xmath0 of finite volume a curve @xmath86 whose @xmath79-length is at most @xmath81 . if @xmath87 is any other choice of such a map then @xmath88 . by the discussion in @xcite there is a constant @xmath89 such that @xmath90 where by abuse
, we use the same symbol @xmath91 to denote the distance on @xmath6 defined by the teichmller metric and the distance on the curve graph @xmath76 . for
a quadratic differential @xmath49 define the _
@xmath44-length _ of an essential closed curve @xmath82 on @xmath0 to be the minimal length of a representative of the free homotopy class of @xmath82 with respect to the singular euclidean metric defined by @xmath44 .
we have . [ simple1 ] for every @xmath92
there is a constant @xmath93 with the following property . for any quadratic differential @xmath49 the diameter in @xmath76 of the set of all simple closed curves on @xmath0 of @xmath44-length at most @xmath94 does not exceed @xmath95 . by lemma 5.1 of @xcite
( see also lemma 5.1 of @xcite for an alternative proof ) there is a number @xmath96 and for every singular euclidean metric on @xmath0 defined by a quadratic differential @xmath44 of area one there is an embedded annulus @xmath97 of width at least @xmath98 .
if we denote by @xmath23 the core curve of @xmath99 , then the @xmath44-length of every simple closed curve @xmath29 on @xmath0 is at least @xmath100 . as a consequence , for every @xmath101 the intersection number with @xmath23 of every simple closed curve @xmath29 on @xmath0 whose @xmath44-length is at most @xmath94 is bounded from above by @xmath102 .
this implies that the set of such curves is of uniformly bounded diameter in @xmath76 ( see @xcite ) .
by possibly enlarging the constant @xmath78 as above we may assume that for every @xmath49 there is an essential simple closed curve on @xmath0 of @xmath44-length at most @xmath81 ( see the proof of lemma [ simple2 ] for a justification of this well known fact ) .
thus we can define a map @xmath103 by associating to a quadratic differential @xmath44 a simple closed curve @xmath104 whose @xmath44-length is at most @xmath81 . by lemma [ simple1 ] ,
if @xmath105 is any other choice of such a map then we have @xmath106 for all @xmath49 where @xmath91 is the distance function on @xmath76 .
recall the definition of the map @xmath107 which associates to a complete hyperbolic metric @xmath79 on @xmath0 of finite volume a simple closed curve @xmath29 of @xmath79-length at most @xmath81 .
let @xmath108 be the canonical projection .
the following simple observation is related to recent work of rafi @xcite .
we include its short proof for completeness .
[ simple2 ] there is a constant @xmath109 such that @xmath110 for all @xmath49 . by lemma [ simple1 ] ,
it is enough to show that for every @xmath49 and every simple closed curve @xmath82 on @xmath0 whose length with respect to the hyperbolic metric @xmath111 is bounded from above by @xmath81 , the @xmath44-length of @xmath82 is uniformly bounded . for this
observe that by the collar lemma of hyperbolic geometry , a simple closed geodesic @xmath82 on a hyperbolic surface whose length is bounded from above by @xmath81 is the core curve of an embedded annulus @xmath99 whose _ modulus _ is bounded from below by a universal constant @xmath96 .
then the _ extremal length _ of the core curve of @xmath99 is bounded from above by a universal constant @xmath112 .
now the area of @xmath44 equals one and therefore the @xmath44-length of the core curve @xmath82 of @xmath99 does not exceed @xmath113 by the definition of extremal length ( see e.g. @xcite ) .
this shows the lemma .
choose a smooth function @xmath114 $ ] with @xmath115\equiv 1 $ ] and @xmath116 where as before , @xmath81 is a bers constant for @xmath0 . for every @xmath117
we obtain a finite borel measure @xmath118 on the curve graph @xmath76 by defining @xmath119 where @xmath120 denotes the dirac mass at @xmath29 .
the total mass of @xmath118 is bounded from above and below by a universal positive constant , and the diameter of the support of @xmath118 in @xmath76 is uniformly bounded as well . moreover , the measures @xmath118 depend continuously on @xmath117 in the weak@xmath10-topology .
this means that for every bounded function @xmath121 the function @xmath122 is continuous .
the curve graph @xmath76 is a hyperbolic geodesic metric space @xcite and hence it admits a _ gromov boundary _ @xmath123 . for every @xmath124
there is a complete distance function @xmath120 on the gromov boundary @xmath125 of @xmath76 of uniformly bounded diameter and there is a number @xmath126 such that @xmath127 for all @xmath128 .
the distances @xmath120 are equivariant with respect to the action of @xmath7 on @xmath76 and on @xmath123 . for @xmath117
define a distance @xmath129 on @xmath123 by @xmath130 clearly the metrics @xmath129 are equivariant with respect to the action of @xmath7 on @xmath6 and @xmath125 .
moreover , there is a constant @xmath131 such that @xmath132 for all @xmath133 ( here as before , @xmath91 denotes the teichmller metric ) .
namely , the function @xmath134 is smooth , with uniformly bounded differential .
moreover , for every simple closed curve @xmath124 , the function @xmath135 on @xmath6 is smooth , with unformly bounded differential with respect to the norm induced by the teichmller metric ( see @xcite ) .
since @xmath134 is supported in @xmath136 $ ] , this implies that for each @xmath124 the function @xmath137 on @xmath6 is smooth , with uniformly bounded differential . as a consequence , for all @xmath138
the function @xmath139 is smooth , and the differential of its logarithm is uniformly bounded with respect to the teichmller norm , independent of @xmath140 . from this and the definitions ,
the estimate ( [ delta ] ) above is immediate . via enlarging the constant @xmath141
we may also assume that @xmath142 for every @xmath117 .
in this section we study _ conformal densities _ on the space @xmath37 of projective measured geodesic laminations on @xmath0 .
recall that @xmath37 equipped with the weak topology is homeomorphic to the sphere @xmath12 , and the mapping class group @xmath7 naturally acts on @xmath37 as a group of homeomorphisms . by the hubbard masur theorem @xcite , for every @xmath143 and every @xmath144
there is a unique holomorphic quadratic differential @xmath145 of area one on @xmath146 whose horizontal measured geodesic lamination @xmath147 is contained in the class @xmath20 .
for all @xmath148 there is a number @xmath149 such that @xmath150 .
the function @xmath151 is continuous , moreover it satisfies the _ cocycle identity _
@xmath152 for all @xmath153 and all @xmath154 .
[ condense ] a _ conformal density of dimension @xmath155 _ on @xmath37 is an @xmath7-equivariant family @xmath156 @xmath157 of finite borel measures on @xmath37 which are absolutely continuous and satisfy @xmath158 almost everywhere . the conformal density @xmath156 is _ ergodic _ if the @xmath7-invariant measure class it defines on @xmath37 is ergodic .
there is an ergodic conformal density @xmath159 of dimension @xmath160 in the lebesgue measure class induced by the symplectic form on the space @xmath9 of all measured geodesic laminations on @xmath0 , see @xcite .
note that since the action of @xmath7 on @xmath37 is minimal , the measure class of a conformal density is always of full support . in section 5 we will see that every conformal density gives rise to an @xmath7-invariant radon measure on @xmath9 ( see @xcite ) , so the classification of conformal densities is essential for the classification of invariant radon measures on @xmath9 .
following @xcite , we construct from a conformal density @xmath161 of dimension @xmath82 an @xmath7-invariant family @xmath162 of locally finite borel measures on strong unstable manifolds @xmath50 @xmath57 which transform under the teichmller geodesic flow @xmath42 via @xmath163 . for this
let @xmath164 be the natural projection which maps a quadratic differential @xmath49 to its horizontal projective measured geodesic lamination .
let @xmath108 be the canonical projection .
the restriction of the projection @xmath165 to the strong unstable manifold @xmath50 is a homeomorphism onto its image @xcite and hence the measure @xmath166 on @xmath37 induces a borel measure @xmath167 on @xmath50 .
the measure @xmath168 on @xmath50 defined by @xmath169 is locally finite and does not depend on the choice of @xmath44 .
the measures @xmath168 on strong unstable manifolds transform under the teichmller flow as required .
the _ flip _
@xmath170 maps strong stable manifolds homeomorphically onto strong unstable ones and therefore we obtain a family @xmath171 of locally finite borel measures on strong stable manifolds by defining @xmath172 .
let @xmath173 be the usual lebesgue measure on the flow lines of the teichmller flow .
the locally finite borel measure @xmath174 on @xmath41 defined by @xmath175 is invariant under the teichmller geodesic flow @xmath42 and the action of the mapping class group . if we denote by @xmath176 the diagonal in @xmath177 then the desintegration @xmath178 of @xmath174 along the flow lines of the teichmller flow is an @xmath7-invariant locally finite borel measure on @xmath179 .
let @xmath47 be the @xmath42-invariant locally finite borel measure on @xmath58 which is the projection of the restriction of @xmath174 to a borel fundamental domain for the action of @xmath7 .
for the conformal density @xmath159 in the lebesgue measure class the resulting @xmath42-invariant measure @xmath20 on @xmath58 is finite . call a quadratic differential
@xmath180 _ forward returning _
if there is a compact subset @xmath181 of @xmath58 and for every @xmath182 there is some @xmath183 with @xmath184 . call a projective measured geodesic lamination @xmath185 _ returning _
if there is a quadratic differential @xmath49 whose horizontal measured geodesic lamination is contained in the class @xmath186 and whose projection to @xmath58 is forward returning .
the set of returning projective measured geodesic laminations is a borel subset of @xmath37 which is invariant under the action of the mapping class group and is contained in the set of all uniquely ergodic projective measured geodesic laminations which fill up @xmath0 @xcite .
note however that cheung and masur @xcite constructed an example of a uniquely ergodic projective measured geodesic lamination which fills up @xmath0 and is not returning .
call a quadratic differential @xmath180 _ forward recurrent _ if @xmath44 is contained in the @xmath187-limit set of its own orbit under @xmath42 .
call a projective measured geodesic lamination @xmath185 _ recurrent _ if there is a quadratic differential @xmath49 whose horizontal measured geodesic lamination is contained in the class @xmath186 and whose projection to @xmath58 is forward returning and contains a forward recurrent point @xmath188 in its @xmath187-limit set .
the set @xmath189 of all recurrent projective measured geodesic laminations is an @xmath7-invariant borel subset of @xmath37 which has full lebesgue measure .
we have . [ ergodicity ] for an ergodic conformal density @xmath161 of dimension @xmath82 the following are equivalent . 1 .
@xmath161 gives full mass to the returning projective measured geodesic laminations .
@xmath161 gives full mass to the set @xmath189 of recurrent projective measured geodesic laminations .
the measure @xmath190 on @xmath179 is ergodic under the diagonal action of @xmath7 .
the teichmller geodesic flow @xmath42 is conservative for the measure @xmath47 on @xmath58 .
the teichmller geodesic flow @xmath42 is ergodic for @xmath47 .
we follow sullivan @xcite closely .
namely , for @xmath191 let @xmath192 be the open @xmath7-invariant subset of all complete hyperbolic structures on @xmath0 of finite volume whose systole ( i.e. the length of a shortest closed geodesic ) is bigger than @xmath193 and define @xmath194 .
then the closure @xmath195 of @xmath196 projects to a compact subset @xmath197 of @xmath58 . for @xmath198
we have @xmath199 and @xmath200 . for @xmath201 define @xmath202 to be the set of all projective measured geodesic laminations @xmath186 such that there is a quadratic differential @xmath49 with @xmath203 and a sequence @xmath204 with @xmath205 for all @xmath206 .
the set @xmath207 is invariant under the action of @xmath7 , and @xmath208 is the set of all returning projective measured geodesic laminations .
let @xmath161 be an ergodic conformal density of dimension @xmath82 which gives full mass to the set @xmath209 of returning projective measured geodesic laminations . by invariance and ergodicity ,
there is a number @xmath191 such that @xmath210 gives full mass to @xmath207 .
let @xmath47 be the @xmath42-invariant radon measure on @xmath58 defined by @xmath161 . by the results of @xcite , for every forward returning quadratic differential @xmath180 and every @xmath211 we have @xmath212 @xmath213 .
thus for @xmath47-almost every @xmath180 the @xmath42-orbit of @xmath44 enters the compact set @xmath214 for arbitrarily large times .
then there is a number @xmath215 and for @xmath47-almost every @xmath216 there are infinitely many integers @xmath217 with @xmath218 . as a consequence ,
the first return map to @xmath219 of the homeomorphism @xmath220 of @xmath58 defines a measurable map @xmath221 which preserves the restriction @xmath222 of @xmath47 . since @xmath47 is a radon measure , @xmath223 is finite and hence the system @xmath224 is conservative .
but then the measure @xmath47 is conservative for the time - one map @xmath220 of the teichmller flow .
moreover , by the poincar recurrence theorem , applied to the measure preserving map @xmath225 , we obtain that @xmath47-almost every @xmath180 is forward recurrent .
thus 1 ) above implies 2 ) and 4 ) . using the usual hopf argument
@xcite we conclude that the measure @xmath47 is ergodic .
namely , choose a continuous positive function @xmath226 such that @xmath227 ; such a function @xmath228 exists since the measure @xmath47 is locally finite by assumption . by the above ,
if @xmath229 are typical for @xmath47 and contained in the same strong stable manifold then the orbits of @xmath42 through @xmath230 are forward asymptotic @xcite . by the birkhoff ergodic theorem , for every continuous function @xmath231 with compact support
the limit @xmath232 exists almost everywhere , and the hopf argument shows that the function @xmath233 is constant along @xmath47-almost all strong stable and strong unstable manifold . from this ergodicity follows as in @xcite . in particular ,
1 ) above implies 5 ) .
the remaining implications in the statement of the lemma are either trivial or standard .
since they will not be used in the sequel , we omit the proof . every ergodic conformal density either gives full measure or zero measure to the @xmath9-invariant borel subset @xmath189 of recurrent projective measured geodesic laminations .
the main goal of this section is to show that a conformal density @xmath161 which gives full measure to @xmath189 is contained lebesgue measure class . for this
we adapt ideas of sullivan @xcite to our situation .
namely , the set @xmath189 can be viewed as the radial limit set for the action of @xmath7 on @xmath6 , where @xmath37 is identified with the thurston boundary of @xmath6 .
this means that every point in @xmath189 is contained in a nested sequence of neighborhoods which are images of a fixed set under elements of the mapping class group .
the mass of these sets with respect to the measures @xmath234 ( where as before , @xmath159 is a conformal density in the lebesgue measure class ) are controlled , and this allows a comparison of measures . to carry out this approach
, we have to construct such nested sequences of neighborhoods of points in @xmath189 explicitly .
we use the gromov distances on the boundary @xmath123 of the curve graph @xmath76 for this purpose .
this boundary consists of all unmeasured minimal geodesic laminations which fill up @xmath0 , equipped with a coarse hausdorff topology .
if we denote by @xmath235 the borel set of all projective measured geodesic laminations whose support is a geodesic lamination which is minimal and fills up @xmath0 then the natural forgetful map @xmath236 which assigns to a projective measured geodesic lamination in @xmath237 its support is a continuous @xmath7-equivariant surjection @xcite .
the restriction of the projection map @xmath238 to the borel set @xmath239 is injective @xcite . for @xmath49 and @xmath240
define @xmath241 then @xmath242 is a compact neighborhood of @xmath44 in @xmath50 with dense interior which depends continuously on @xmath44 in the following sense . if @xmath243 in @xmath41 then @xmath244 in the hausdorff topology for compact subsets of @xmath41
we have .
[ closed ] 1 .
the map @xmath245 is continuous and closed .
if the horizontal measured geodesic lamination of the quadratic differential @xmath49 is uniquely ergodic then the sets @xmath246 @xmath247 form a neighborhood basis for @xmath248 in @xmath125 .
the first part of the lemma is immediate from the description of the gromov boundary of @xmath76 in @xcite . to show the second part , note first that for every @xmath49 the restriction of the projection @xmath249 is a homeomorphism of @xmath50 onto an _
open _ subset of @xmath37 . to see this , identify @xmath37 with a section @xmath250 of the fibration @xmath251 .
then @xmath252 is contained in @xmath253 if and only if the function @xmath254 on @xmath250 is _
positive_. since the intersection form @xmath255 on @xmath9 is continuous and since @xmath250 is compact , this is an open condition for points @xmath252 .
now if @xmath49 is uniquely ergodic then by @xcite we have @xmath256 and hence if @xmath240 is arbitrary then @xmath257 is a closed subset of @xmath125 which does not contain @xmath248 .
in particular , @xmath258 is a neighborhood of @xmath248 in @xmath125 . from this and continuity of @xmath238 the second part of the lemma follows .
define @xmath259 recall from section 2 that there is a number @xmath131 and there is an @xmath7-equivariant family of distance functions @xmath129 ( @xmath260 ) on @xmath123 such that @xmath261 for all @xmath133 . for @xmath262 and @xmath101
define @xmath263 to be the closed @xmath264-ball of radius @xmath94 about @xmath265 .
the following lemma is a translation of hyperbolicity of the curve graph into properties of the distance functions @xmath129 .
[ ballestimate ] 1 . for every @xmath266
there is a number @xmath267 such that @xmath268 for every @xmath262 and every @xmath269 .
2 . there is a number @xmath270 with the following property . for every @xmath262 and every @xmath191
there is a number @xmath271 such that @xmath272 for every @xmath273 . by the results of @xcite ( see theorem 4.1 of @xcite for an explicit statement )
, there is a number @xmath274 such that the image under @xmath275 of every teichmller geodesic is an _
unparametrized @xmath276-quasi - geodesic in @xmath76_. this means that for every @xmath49 there is an increasing homeomorphism @xmath277 such that the curve @xmath278 is an @xmath276-quasi - geodesic in @xmath76 .
if @xmath262 then we have @xmath279 @xmath213 and the unparametrized @xmath276-quasi - geodesic @xmath280 converges as @xmath281 in @xmath282 to the point @xmath283 ( see @xcite ) .
in particular , for @xmath262 and every @xmath284 there is a number @xmath285 such that @xmath286 for all @xmath287 . since @xmath76 is a hyperbolic geodesic metric space , any finite subarc of an @xmath276-quasi - geodesic
is contained in a tubular neighborhood of a geodesic in @xmath76 of uniformly bounded radius .
this implies that there is no backtracking along an unparametrized @xmath276-quasi - geodesic : there is a constant @xmath96 only depending on @xmath276 and the hyperbolicity constant of @xmath76 such that @xmath288 for all @xmath289 and every @xmath276-quasi - geodesic @xmath290 . from the definition of the gromov distances
@xmath120 @xmath291 we obtain that there is a number @xmath292 such that for every @xmath276-quasi - geodesic ray @xmath290 with endpoint @xmath293 and every @xmath294 the gromov distances @xmath295 on @xmath123 satisfy @xmath296 on the @xmath295-ball of radius @xmath82 about @xmath186 .
let @xmath131 be as in inequality ( [ delta ] ) from section 2 and define @xmath297 .
by inequality ( [ deltacomparison ] ) from section 2 , for @xmath262 and @xmath269 we have @xmath298 on the @xmath299-ball @xmath300 . thus if for @xmath191 we choose @xmath301 sufficiently large that @xmath302 then for @xmath262 , for @xmath303 and for @xmath304 we have @xmath305 which shows the second part of the lemma . to show the first part of the lemma , for @xmath306 define @xmath307
. then the estimate ( [ betanull ] ) above shows that @xmath308 for every @xmath262 and every @xmath269 .
this completes the proof of the lemma . for a forward
recurrent point @xmath188 let @xmath309 be the borel subset of all recurrent projective measured geodesic laminations @xmath310 such that there is some @xmath311 with the following property .
the projection to @xmath58 of the orbit of @xmath44 under the teichmller geodesic flow contains @xmath312 in its @xmath187-limit set . by definition , the set @xmath313 is invariant under the action of @xmath7 .
moreover , every recurrent point @xmath310 is contained in one of the sets @xmath313 for some forward recurrent point @xmath188 .
note moreover that an orbit of @xmath42 in @xmath58 which is typical for the @xmath42-invariant lebesgue measure on @xmath58 is dense and hence for every forward recurrent point @xmath188 the set @xmath313 has full lebesgue measure .
write @xmath314 following @xcite , a _ borel covering relation _ for a borel subset @xmath315 of a metric space @xmath316 is a family @xmath317 of pairs @xmath318 where @xmath319 is a borel set , where @xmath320 and such that @xmath321 the covering relation @xmath317 is called _ fine _ at every point of @xmath315 if for every @xmath322 and every @xmath323 there is some @xmath324 with @xmath325 where @xmath326 denotes the open ball of radius @xmath82 about @xmath146 . for @xmath92 and the forward recurrent point @xmath188 with lift
@xmath327 there is by continuity of @xmath328 a compact neighborhood @xmath181 of @xmath329 in @xmath41 such that @xmath330 .
we call @xmath181 a _ @xmath94-admissible _ neighborhood of @xmath329 . for a number @xmath92 and such a @xmath94-admissible neighborhood @xmath181 of @xmath329
define @xmath331 we sometimes identify a pair @xmath332 with the set @xmath333 whenever the point @xmath186 has no importance .
let @xmath270 be as in lemma [ ballestimate ] .
we have .
[ fine ] let @xmath188 be a forward recurrent point and let @xmath327 be a lift of @xmath312 . then for every @xmath334 and every @xmath94-admissible compact neighborhood @xmath181 of @xmath329 the family @xmath335 is a borel covering relation for @xmath336 which is fine at every point of @xmath337 . using the above notations , let @xmath188 be a forward recurrent point and let @xmath329 be a lift of @xmath312 to @xmath41 .
let @xmath334 where @xmath270 is as in lemma [ ballestimate ] .
it clearly suffices to show the lemma for the covering relations @xmath335 where @xmath181 is a sufficiently small @xmath94-admissible neighborhood of @xmath329 . by relation ( [ delta ] ) in section 2 we infer that for every sufficiently small @xmath94-admissible neighborhood @xmath181 of @xmath329 we have @xmath338 in particular , for @xmath339 the set @xmath340 contains @xmath341 and is contained in @xmath342 . by the construction of the distances @xmath129 @xmath343 on @xmath123 it suffices to show that for every @xmath344 and every @xmath201 there is some @xmath345 such that the set @xmath333 contains @xmath248 and is contained in the open @xmath264-ball of radius @xmath193 about @xmath248 . for @xmath344 and @xmath191
let @xmath271 be as in the second part of lemma [ ballestimate ] .
choose some @xmath346 such that @xmath347 ; such a number exists by the definition of the set @xmath348 and by @xcite . by lemma [ ballestimate ]
we have @xmath349 . now if @xmath345 is such that @xmath350 then we obtain from @xmath94-admissibility of the set @xmath181 and equivariance under the action of the mapping class group that @xmath351 since @xmath191 was arbitrary , this shows the lemma .
the next proposition is the main technical result of this section . for its formulation ,
recall from @xcite the definition of a _
vitali relation _ for a finite borel measure on the borel subset @xmath337 of @xmath76 .
we show .
[ vitali ] let @xmath188 be a forward recurrent point for the teichmller geodesic flow .
then for every sufficiently small @xmath92 , every sufficiently small @xmath94-admissible compact neighborhood @xmath181 of @xmath329 and for every conformal density @xmath161 on @xmath37 which gives full measure to the set @xmath313 , the covering relation @xmath335 for @xmath337 is a vitali relation for the measure @xmath352 on @xmath125 .
the strategy of proof is to use the properties of the balls @xmath353 established in lemma [ ballestimate ] to gain enough control on @xmath210-volumes that the results of federer @xcite can be applied .
let @xmath188 be a forward recurrent point for @xmath42 and let @xmath327 be a lift of @xmath312 .
since no torsion element of @xmath7 fixes pointwise the teichmller geodesic defined by @xmath329 we may assume that the point @xmath354 is not fixed by any nontrivial element of the mapping class group .
by lemma [ fine ] and using the notations from this lemma , for every @xmath334 and every @xmath94-admissible compact neighborhood @xmath181 of @xmath329 the covering relation @xmath335 for @xmath355 is fine for the metric @xmath356 at every point of @xmath315 .
we first establish some geometric control on the covering relation @xmath335 for some particularly chosen small @xmath334 and a suitably chosen @xmath94-admissible neighborhood @xmath181 of @xmath329 .
for this let again @xmath91 be the distance on @xmath6 defined by the teichmller metric .
choose a number @xmath240 which is sufficiently small that the images under the action of the mapping class group of the closed @xmath91-ball @xmath357 of radius @xmath358 about @xmath359 are pairwise disjoint . by the estimate ( [ delta ] ) for the family of distance functions @xmath360 @xmath361 , by decreasing the size of the radius @xmath362
we may assume that @xmath363 recall from ( [ balldefinition ] ) above the definition of the closed balls @xmath364 @xmath57 .
by continuity of the projection @xmath365 there is an open neighborhood @xmath366 of @xmath329 in @xmath41 such that @xmath367 and that moreover the projection @xmath368 of @xmath366 to @xmath6 is contained in the open ball of radius @xmath362 about @xmath359 .
this implies in particular that @xmath369 for every nontrivial element @xmath345 .
since the horizontal measured geodesic lamination of @xmath329 is uniquely ergodic , lemma [ closed ] shows that there is a number @xmath266 such that @xmath370 since the map @xmath371 is continuous , there is an open neighborhood @xmath372 of @xmath329 such that @xmath373 . by the choice of @xmath366 we have @xmath374 for all @xmath375 . for all @xmath375
we also have @xmath376 by lemma [ ballestimate ] , there is a number @xmath377 such that for every @xmath269 we have @xmath378 now @xmath379 is an open neighborhood of @xmath329 in @xmath41 and therefore @xmath380 is an open neighborhood of @xmath329 in @xmath381 .
in particular , there is a number @xmath382 such that @xmath383 .
thus by lemma [ closed ] there is a number @xmath384 such that @xmath385 note that we have @xmath386 using once more continuity of the map @xmath387 , there is a compact neighborhood @xmath388 of @xmath329 such that @xmath389 in particular , @xmath181 is @xmath94-admissible . by inequality ( [ deltaxu ] ) for the dependence of the metrics @xmath264 on the points @xmath390
we then have @xmath391 by ( [ zwei ] ) above and continuity of the strong unstable foliation and of the map @xmath365 we may moreover assume that @xmath392 note that if @xmath393 then @xmath394 and hence if @xmath395 is such that @xmath396 then @xmath397 .
namely , the projective measured geodesic lamination @xmath398 of every @xmath399 is uniquely ergodic and therefore @xmath400 consists of a unique point .
however , by ( [ zwei ] ) above , the set @xmath380 contains such a point .
consequently , inequality ( [ deltaxu ] ) above shows that @xmath401 .
define @xmath402 by lemma [ fine ] , @xmath403 is a covering relation for the set @xmath404 which is fine at every point of @xmath315 . by the choice of the set @xmath405 , if @xmath406 , if @xmath407 and if @xmath294 are such that @xmath350 then @xmath345 is uniquely determined by @xmath408 . for @xmath409
define @xmath410 following @xcite , for @xmath409 define the _
@xmath228-enlargement _ of @xmath411 by @xmath412 where in this definition , the constant @xmath240 is chosen as in the beginning of this proof .
let @xmath161 be a conformal density of dimension @xmath155 which gives full measure to the set @xmath313 .
we may assume that the density is ergodic , i.e. that the @xmath7-invariant measure class it defines on @xmath37 is ergodic .
the measure @xmath210 induces a borel measure @xmath352 on the set @xmath413 .
recall from the beginning of this section that the measures @xmath414 @xmath157 define a family of @xmath7-invariant radon measures @xmath168 on strong unstable manifolds in @xmath41 .
these measures are invariant under holonomy along strong stable manifolds and they are quasi - invariant under the teichmller geodesic flow , with transformation @xmath415 .
for @xmath49 the measure @xmath168 on @xmath50 projects to a borel measure @xmath416 on @xmath315 .
for @xmath417 the measures @xmath418 are absolutely continuous , with continuous radon nikodym derivative depending continuously on @xmath419 .
by invariance of the measures @xmath168 under holonomy along strong stable manifolds and by the choice of the point @xmath329 and the number @xmath92 there is a number @xmath420 such that @xmath421 for all @xmath422 .
write @xmath423 ; we claim that there is a number @xmath112 such that @xmath424 for all @xmath409 . for this
let @xmath409 be arbitrary ; then there is some @xmath406 with @xmath425 and there is a number @xmath269 and some @xmath345 such that @xmath350 and that @xmath426 and @xmath427 . by equivariance under the action of the mapping class group and by the inclusion ( [ eins ] ) above , we have @xmath428 let @xmath429 be such that @xmath430 and that @xmath431 .
then there is a number @xmath432 $ ] , a point @xmath433 such that @xmath434 and some @xmath435 such that @xmath436 and that @xmath437 , @xmath438 . by equivariance under the action of @xmath7 and the inclusion ( [ eins ] ) above , we have @xmath439 and hence from the definition of the sets @xmath440 and the definition of the strong unstable manifolds we conclude that @xmath441 since the restriction of the map @xmath238 to @xmath189 is injective and the restriction of the map @xmath365 to @xmath442 is injective and since @xmath443 by assumption we obtain that @xmath444 . as a consequence ,
the distance in @xmath6 between the points @xmath445 and @xmath446 is at most @xmath447 and hence the distance between @xmath448 and @xmath449 is at most @xmath450 . on the other hand , since @xmath405 , for @xmath451
the distance in @xmath6 between @xmath452 and @xmath453 is not smaller than @xmath454 .
therefore we have @xmath455 and @xmath456 .
this shows that @xmath457 on the other hand , if @xmath458 , if @xmath459 and @xmath435 are such that @xmath460 and @xmath461 and if @xmath462 is such that @xmath463 and @xmath464 then @xmath465 . by the choice of the set @xmath181 , equivariance under the action of the mapping class group and the inclusion ( [ sixteen ] ) above
, we have @xmath466 and @xmath467 and hence @xmath468 .
in other words , there is some @xmath469 with @xmath470 . by the inclusions ( [ sixteen ] ) and ( [ drei ] ) and the following remark , since @xmath471 and @xmath472 we have @xmath473 and moreover @xmath474 from ( [ zweib ] ) above and invariance under the action of the mapping class group we obtain @xmath475 . the inclusion ( [ eins ] ) then yields that @xmath476 this shows that the @xmath228-enlargement @xmath477 of @xmath411 is contained in @xmath478 . since @xmath479 by assumption , by invariance under the action of the mapping class group the @xmath480-mass of @xmath478
is uniformly bounded .
therefore by the transformation rule for the measures @xmath481 under the action of the teichmller geodesic flow , the @xmath482-mass of @xmath477 is bounded from above by a fixed multiple of the @xmath482-mass of @xmath411 .
thus by the results of federer @xcite and by lemma [ fine ] , the covering relation @xmath335 is indeed a vitali relation for the measure @xmath483 and hence it is a vitali relation for the measure @xmath352 as well .
note that the same is true for the covering relation @xmath484 for every @xmath485 and every sufficiently small @xmath193-admissible neighborhood @xmath486 of @xmath329 .
using lemma [ ergodicity ] , lemma [ fine ] and proposition [ vitali ] we can now show .
[ confdim ] 1 .
a conformal density on @xmath37 which gives full measure to @xmath237 is of dimension at least @xmath487 , with equality if and only if it coincides with the lebesgue measure up to scale .
a conformal density which gives full measure to the set of returning projective measured geodesic laminations is of dimension @xmath487 and coincides with the lebesgue measure up to scale .
let @xmath161 be a conformal density of dimension @xmath155 which gives full measure to the set @xmath237 .
we may assume that the density is ergodic , i.e. that the @xmath7-invariant measure class it defines on @xmath37 is ergodic .
let moreover @xmath159 be the conformal density of dimension @xmath488 which defines the @xmath42-invariant probability measure on @xmath58 in the lebesgue measure class . we have to show that @xmath489 , with equality if and only if @xmath490 up to scale .
for this assume that @xmath491 .
the lebesgue measure @xmath20 on @xmath58 is of full support and ergodic under the teichmller flow and therefore the @xmath42-orbit of @xmath20-almost every @xmath180 is dense .
this implies that there is a recurrent point @xmath188 with the property that the measure @xmath492 gives full mass to @xmath313 .
recall that the conformal densities @xmath493 define families @xmath494 of radon measures on the strong unstable manifolds . for @xmath49
denote by @xmath495 the image under the map @xmath328 of the restriction of these measures to @xmath496 . since the conformal densities @xmath493 give full measure to the set @xmath237 , for every @xmath49 the measures @xmath497 on @xmath125 are of full support .
let @xmath327 be a lift of @xmath312 to @xmath41 . by proposition [ vitali ] , for sufficiently small @xmath92 and for a sufficiently small compact neighborhood @xmath181 of @xmath329 the covering relation
@xmath335 is a vitali relation for the measure @xmath498 on @xmath123 . using equivariance under the action of @xmath7 and the fact that @xmath499 is of full support ,
if the measures @xmath500 are singular then there for @xmath501-almost every @xmath502 there is a sequence @xmath503 such that for every @xmath206 the following holds .
1 . @xmath504 for some @xmath505 .
the @xmath506-mass and the @xmath507-mass of @xmath508 is bounded from above and below by a universal constant .
3 . the limit @xmath509 exists and equals zero . in particular ,
for every @xmath182 and all sufficiently large @xmath255 , say for all @xmath510 , we have @xmath511 . on the other hand , we also have @xmath512 for a universal constant @xmath112 and @xmath513 for a universal constant @xmath514 . if @xmath182 is sufficiently large that @xmath515 then for @xmath510 we obtain a contradiction .
in other words , if @xmath516 then the measures @xmath161 and @xmath159 are absolutely continuous .
moreover , they give full mass to the recurrent projective measured geodesic laminations .
then they define absolutely continuous @xmath42-invariant radon measures @xmath517 on @xmath58 which are ergodic by lemma [ ergodicity ] . as a consequence , the measures @xmath517 coincide up to scale and hence the measures @xmath493 coincide up to scale as well .
this shows the first part of the lemma .
to show the second part of the lemma , assume that @xmath518 and that the conformal density @xmath161 gives full measure to the subset of @xmath37 of returning points . by lemma [ ergodicity ] , @xmath161 gives full measure to the set @xmath189 of recurrent points .
since @xmath161 is ergodic , there is a forward recurrent quadratic differntial @xmath188 such that @xmath161 gives full measure to the set @xmath313 .
this implies that we can exchange the roles of @xmath159 and @xmath161 in the above argument and obtain that @xmath493 coincide up to scale .
the following proposition uses the results of minsky and weiss @xcite to show that up to scale , the lebesgue measure is the unique conformal density on @xmath37 which gives full measure to the set @xmath237 of filling projective measured geodesic laminations . [ confonfilling ]
let @xmath161 be a conformal density which gives full measure to @xmath237 .
then @xmath161 is the lebesgue measure up to scale .
we argue by contradiction and we assume that there is a conformal density @xmath161 which gives full measure to the set @xmath237 of all projective measured geodesic laminations whose support is minimal and fills up @xmath0 and which is singular to the lebesgue measure . without loss of generality
we can assume that @xmath161 is ergodic .
by lemma [ confdim ] , the dimension @xmath82 of @xmath161 is strictly bigger than @xmath519 and the @xmath210-measure of the set of returning points vanishes .
let @xmath47 be the locally finite borel measure on @xmath58 which can be written in the form @xmath520 where @xmath59 is the family of lebesgue measures on stable manifolds which transforms under the teichmller flow @xmath42 via @xmath521 .
the measure @xmath47 is quasi - invariant under the teichmller geodesic flow and transforms via @xmath522 since @xmath523 this implies that the measure @xmath47 is infinite .
moreover , it gives full mass to quadratic differential whose horizontal measured geodesic lamination is minimal and fills up @xmath0 .
the family @xmath59 of lebesgue measures on stable manifolds is invariant under the horocycle flow @xmath74 as defined in section 2 . by the explicit construction of the measures @xmath168
, this implies that the measure @xmath47 is invariant under @xmath74 .
however , following the reasoning of dani @xcite ( see the proof of corollary 2.6 of @xcite for a discussion in our context ) , this implies that the measure @xmath47 is necessarily _
finite_. namely , by the birkhoff ergodic theorem , applied to the horocycle flow @xmath74 and the locally finite @xmath74-invariant measure @xmath47 ( see theorem 2.3 of @xcite for the version of the birkhoff ergodic theorem for locally finite infinite measures needed here ) , for @xmath47-almost every @xmath180 and for every continuous positive function @xmath231 on @xmath58 with @xmath524 the limit @xmath525 exists , and the resulting function @xmath238 is @xmath74-invariant and @xmath47-integrable . on the other hand , consider the family of borel probability measures @xmath526 on @xmath58 where @xmath527 denotes the dirac mass at @xmath146 . by theorem h2 of @xcite ( more precisely , by the theorem in the appendix which is a slightly extended version of this result ) , for every @xmath191 there is a compact set @xmath528 such that for @xmath47-almost every @xmath180 , every weak limit @xmath529 of the measures @xmath530 as @xmath531 satisfies @xmath532 . since the function @xmath231 is _ positive _ , we have @xmath533 and therefore @xmath534 for @xmath47-almost every @xmath180 . but
this contradicts the fact that the measure @xmath47 is infinite and that @xmath238 is @xmath47-integrable and shows the proposition .
in section 3 we showed that conformal densities for the mapping class group which give full measure to the set of filling measured geodesic laminations coincide with the lebesgue measure up to scale .
conformal densities induce @xmath7-invariant radon measures on measured lamination space ( see the discussion in section 5 ) . to understand @xmath7-invariant radon measures on @xmath9 which are _ not _ of this form we use train tracks as the main technical tool .
in this section we summarize those properties of train tracks which are needed for our purpose . a _ maximal generic train track _ on the surface @xmath0 is an embedded 1-complex @xmath535 whose edges ( called _ branches _ ) are smooth arcs with well - defined tangent vectors at the endpoints . at any vertex ( called a
_ switch _ ) the incident edges are mutually tangent .
every switch is trivalent . through each switch
there is a path of class @xmath536 which is embedded in @xmath537 and contains the switch in its interior .
in particular , the branches which are incident on a fixed switch are divided into `` incoming '' and `` outgoing '' branches according to their inward pointing tangent at the switch .
the complementary regions of the train track are trigons , i.e. discs with three cusps at the boundary , or once punctured monogons , i.e. once punctured discs with one cusp at the boundary .
we always identify train tracks which are isotopic ( see @xcite for a comprehensive account on train tracks ) .
a maximal generic train track or a geodesic lamination @xmath134 is _ carried _ by a train track @xmath537 if there is a map @xmath538 of class @xmath536 which is homotopic to the identity and maps @xmath134 into @xmath537 in such a way that the restriction of the differential of @xmath238 to the tangent space of @xmath134 vanishes nowhere ; note that this makes sense since a train track has a tangent line everywhere .
we call the restriction of @xmath238 to @xmath134 a _ carrying map _ for @xmath134
. write @xmath539 if the train track or the geodesic lamination @xmath134 is carried by the train track @xmath537 . a _ transverse measure _ on a maximal generic train track @xmath537 is a nonnegative weight function @xmath46 on the branches of @xmath537 satisfying the _ switch condition _ : for every switch @xmath540 of @xmath537 , the sum of the weights over all incoming branches at @xmath540 is required to coincide with the sum of the weights over all outgoing branches at @xmath540 .
the train track is called _ recurrent _ if it admits a transverse measure which is positive on every branch .
we call such a transverse measure @xmath46 _ positive _ , and we write @xmath541 .
the space @xmath542 of all transverse measures on @xmath537 has the structure of a convex euclidean cone . via a carrying map
, a measured geodesic lamination carried by @xmath537 defines a transverse measure on @xmath537 , and every transverse measure arises in this way @xcite .
thus @xmath542 can naturally be identified with a subset of @xmath9 which is invariant under scaling .
a maximal generic train track @xmath537 is recurrent if and only if the subset @xmath542 of @xmath9 has nonempty interior .
a _ tangential measure _
@xmath46 for a maximal generic train track @xmath537 associates to every branch @xmath98 of @xmath537 a nonnegative weight @xmath543 such that for every complementary triangle with sides @xmath544 we have @xmath545 ( indices are taken modulo three ) .
the space @xmath546 of all tangential measures on @xmath537 has the structure of a convex euclidean cone .
the maximal generic train track @xmath537 is called _ transversely recurrent _ if it admits a tangential measure @xmath46 which is positive on every branch @xcite .
there is a one - to - one correspondence between the space of tangential measures on @xmath537 and the space of measured geodesic laminations which _ hit @xmath537 efficiently _ ( we refer to @xcite for an explanation of this terminology ) . with this identification ,
the pairing @xmath547 defined by @xmath548 is just the intersection form on @xmath9 @xcite .
a maximal generic train track @xmath537 is called _ complete _ if it is recurrent and transversely recurrent . in the sequel
we identify train tracks which are isotopic , and we denote by @xmath549 the set of isotopy classes of all complete train tracks on @xmath0 . a half - branch @xmath550 in a complete train track @xmath537 incident on a switch @xmath551 of @xmath537
is called _ large _ if every embedded arc of class @xmath536 containing @xmath551 in its interior passes through @xmath550 . a half - branch which is not large is called _ small_. a branch @xmath98 in a complete train track @xmath537 is called _ large _ if each of its two half - branches is large ; in this case @xmath98 is necessarily incident on two distinct switches , and it is large at both of them .
a branch is called _ small _ if each of its two half - branches is small .
a branch is called _ mixed _ if one of its half - branches is large and the other half - branch is small ( for all this , see @xcite p.118 ) .
there is a simple way to modify a complete train track @xmath537 to another complete train track .
namely , if @xmath552 is a large branch of @xmath537 then we can perform a right or left _ split _ of @xmath537 at @xmath552 as shown in the figure below . note that a right split at @xmath552 is uniquely determined by the orientation of @xmath0 and
does not depend on the orientation of @xmath552 .
using the labels in the figure , in the case of a right split we call the branches @xmath553 and @xmath29 _ winners _ of the split , and the branches @xmath554 are _ losers _ of the split .
if we perform a left split , then the branches @xmath554 are winners of the split , and the branches @xmath555 are losers of the split .
the split @xmath556 of a train track @xmath537 is carried by @xmath537 , and there is a natural choice of a carrying map which maps the switches of @xmath556 to the switches of @xmath537 .
there is a natural bijection of the set of branches of @xmath537 onto the set of branches of @xmath556 which maps the branch @xmath552 to the diagonal @xmath557 of the split .
the split of a complete train track is maximal , transversely recurrent and generic , but it may not be recurrent @xcite . for a complete train track @xmath558 denote by @xmath559 the convex set of all transverse measures on @xmath537 whose total weight ( i.e. the sum of the weights over all branches of @xmath560 equals one .
let moreover @xmath561 be the set of all area one quadratic differentials whose horizontal measured geodesic lamination is contained in @xmath562 and whose vertical measured geodesic lamination hits @xmath537 efficiently .
then @xmath563 is a closed subset of @xmath41 which however is not compact .
recall from @xcite that a _ vertex cycle _ for @xmath537 is a transverse measure on @xmath537 which spans an extreme ray in the convex euclidean cone @xmath542 . up to rescaling ,
such a vertex cycle is the counting measure of a simple closed curve which is carried by @xmath537 ( see @xcite for this fact ) . in the sequel we identify a vertex cycle for @xmath537 with this simple closed curve on @xmath0 . with this convention ,
the transverse measure on @xmath537 defined by a vertex cycle is integral .
define a map @xmath564 by associating to a complete train track @xmath537 one of its vertex cycles .
recall from section 2 the definition of the map @xmath565 .
we have .
[ upsilon ] there is a constant @xmath566 such that @xmath567 for every @xmath558 and every @xmath568 .
let @xmath558 and let @xmath568 .
then the transverse measure @xmath569 of the horizontal measured geodesic lamination of @xmath44 can be written in the form @xmath570 where @xmath571 is the transverse measure on @xmath537 defined by a vertex cycle and where @xmath572 . now
the number of vertex cycles for @xmath537 is bounded from above by a universal constant only depending on the topological type of the surface @xmath0 .
the weight disposed on a branch of @xmath537 by the counting measure of a vertex cycle does not exceed two @xcite .
this means that there is a number @xmath420 only depending on the topological type of @xmath0 and there is some @xmath206 such that @xmath573 .
the vertical measured geodesic lamination of @xmath44 defines a tangential measure @xmath186 on @xmath537 with @xmath574 .
this implies that the vertex cycle @xmath575 satisfies @xmath576 . on the other hand ,
the intersection number @xmath577 is bounded from above by a universal constant as well ( corollary 2.3 of @xcite ) .
moreover , the @xmath44-length of the simple closed curve @xmath82 is bounded from above by @xmath578 ( compare e.g. @xcite for this simple fact ) .
hence the @xmath44-length of @xmath82 is uniformly bounded and therefore lemma [ simple1 ] implies that the distance between @xmath104 and @xmath82 is bounded from above by a universal constant .
since on the other hand the distance in @xmath76 between any two vertex cycles of a complete train track on @xmath0 is also uniformly bounded @xcite , the lemma follows .
recall from @xcite the definition of a _ framing _ ( or _ marking _ ) for the surface @xmath0 .
such a framing @xmath238 consists of a pants decomposition @xmath579 for @xmath0 and a collection of @xmath80 essential simple closed _ spanning curves_. for each pants curve @xmath580
there is a unique such spanning curve which is disjoint from @xmath581 , which is not freely homotopic to any pants curve of @xmath579 and which has the minimal number of intersection points with @xmath29 among all simple closed curves with these properties .
any two such curves differ by a multiple dehn twist about @xmath29 .
there is a number @xmath582 and a hyperbolic metric @xmath117 such that the @xmath79-length of each curve from our framing @xmath238 is at most @xmath583 .
such a metric can be constructed as follows .
equip each pair of pants defined by @xmath579 with a hyperbolic metric such that the length of each boundary geodesic equals one . then glue these pair of pants in such a way that the spanning curves have the smallest possible length .
we call such a hyperbolic metric _ short for @xmath238_. by standard hyperbolic trigonometry , there is a number @xmath201 such that every hyperbolic metric which is short for some framing @xmath238 of @xmath0 is contained in the set @xmath584 of all metrics whose _ systole _ , i.e. the shortest length of a closed geodesic , is at least @xmath193 .
moreover , the diameter in @xmath6 of the set of all hyperbolic metrics which are short for a fixed framing @xmath238 is bounded from above by a universal constant .
the mapping class group @xmath7 naturally acts on the collection of all framings of @xmath0 . by equivariance and the fact that the number of orbits for the action of @xmath7 on @xmath549 is _ finite _
, there is a number @xmath182 and for every complete train track @xmath558 there is a framing @xmath238 of @xmath0 which consists of simple closed curves carried by @xmath537 and such that the total weight of the counting measures on @xmath537 defined by these curves does not exceed @xmath585 ( compare the discussion in @xcite ) .
we call such a framing _ short for @xmath537_. the intersection number @xmath586 between any two simple closed curves @xmath587 which are carried by some @xmath558 and which define counting measures on @xmath537 of total weight at most @xmath585 is bounded from above by a universal constant ( see corollary 2.3 of @xcite ) . in particular , for any two short framings @xmath588 for @xmath537 the distance in @xmath6 between any two hyperbolic metrics which are short for @xmath588
is uniformly bounded @xcite .
define a map @xmath589 by associating to a complete train track @xmath537 a hyperbolic metric @xmath590 which is short for a short framing for @xmath537 . by our above discussion
, there is a number @xmath591 only depending on the topological type of @xmath0 such that if @xmath592 is another choice of such a map then we have @xmath593 for every @xmath558 . in particular , the map @xmath594 is _
coarsely @xmath7-equivariant _ : for every @xmath558 and every @xmath345 we have @xmath595 .
the next observation will be useful in section 5 ( compare also @xcite for a related result ) .
for its formulation , let again @xmath108 be the canonical projection .
[ vertexcycle ] there is a number @xmath596 and for every @xmath191 there is a number @xmath597 with the following property .
let @xmath598 and assume that @xmath134 is carried by @xmath537 and that the distance in @xmath76 between any vertex cycle of @xmath134 and any vertex cycle of @xmath537 is at least @xmath599 .
let @xmath568 be a quadratic differential whose horizontal measured geodesic lamination @xmath600 is carried by @xmath134 .
if the total weight of the transverse measure on @xmath134 defined by @xmath600 is not smaller than @xmath193 , then @xmath601 . let @xmath78 be as in the definition of the map @xmath107 in section 2 .
let @xmath109 be as in lemma [ simple2 ] and let @xmath602 be such that for every @xmath143 the diameter in @xmath76 of the set of simple closed curves whose @xmath146-length is at most @xmath81 is bounded from above by @xmath603 .
let @xmath566 be as in lemma [ upsilon ] and let @xmath604 .
let @xmath605 be such that the distance in @xmath76 between any vertex cycle of @xmath537 and any vertex cycle of @xmath134 is at least @xmath599 .
let @xmath191 and let @xmath568 be such that the horizontal measured geodesic lamination @xmath600 of @xmath44 is carried by @xmath134 and is such that the total weight of the transverse measure on @xmath134 defined by @xmath600 is not smaller than @xmath193 .
we claim that the @xmath111-length of any essential simple closed curve on @xmath0 is at least @xmath606 .
namely , if @xmath269 is such that @xmath607 then @xmath608 .
thus if there is a simple closed curve @xmath29 on @xmath0 whose @xmath111-length is smaller than @xmath606 then the @xmath609-length of @xmath29 is at most @xmath81 ( see @xcite for this well known result of wolpert ) . by the choice of the constant @xmath603 , this means that @xmath610 and hence from lemma [ simple2 ] we conclude that @xmath611 . from lemma [ upsilon ]
we conclude that @xmath612 which contradicts our choice of @xmath599 .
in particular , we have @xmath613 .
now by lemma 3.3 of @xcite , there is a constant @xmath274 depending on @xmath193 such that for every @xmath49 with @xmath614 the singular euclidean metric defined by @xmath44 is @xmath276-bilipschitz equivalent to the hyperbolic metric @xmath111 .
thus for the proof of our lemma it suffices to show that for a quadratic differential @xmath568 as above the @xmath44-length of any short framing for @xmath537 is uniformly bounded .
let @xmath615 be the vertical measured geodesic lamination of @xmath44 .
then @xmath615 defines a tangential measure on @xmath537 .
we claim that the weight of @xmath615 disposed on any branch of @xmath537 is uniformly bounded . for this note
that the transverse measure on the train track @xmath134 defined by the horizontal measured geodesic lamination @xmath600 of @xmath44 can be reprensented in the form @xmath616 with vertex cycles @xmath571 for @xmath134 .
now a counting measure of a simple closed curve which is carried by @xmath134 is integral , and the total weight of the counting measure defined by a vertex cycle of @xmath134 is bounded from above by a universal constant @xmath182 .
moreover , the number of vertex cycles for @xmath134 is bounded from above by a universal constant @xmath617 . since by assumption
the total weight of @xmath600 on @xmath134 is at least @xmath193 , we have @xmath618 for at least one @xmath255 . by the choice of the constant @xmath619 , the distance in @xmath76 between any vertex cycle of @xmath134 and any vertex cycle for @xmath537 is at least @xmath599 . thus by lemma 4.9 of @xcite , the image of @xmath571 under a carrying map @xmath620 is a _
large subtrack _
@xmath187 of @xmath537 .
this means that this image is a train track on @xmath0 which is a subset of @xmath537 and whose complementary components are topological discs or once punctured topological discs .
thus each such complementary component is a polygon or a once punctured polygon which is a union of complementary components of @xmath537 .
the map @xmath621 is convex linear and therefore the @xmath600-weight of every branch of @xmath187 is bounded from below by @xmath622 .
in particular , if @xmath615 denotes the vertical measured geodesic lamination of @xmath44 then we obtain from the identity @xmath623 that the weight of every branch of @xmath187 with respect to the tangential measured defined by @xmath615 is bounded from above by @xmath624 . from the definition of a tangential measure for @xmath537 and the fact that @xmath187 is large we deduce that the total weight that @xmath615 disposes on the branches of @xmath537 is bounded from above by a constant only depending on @xmath193 .
namely , let @xmath625 be a complementary polygon of @xmath187 with more than 3 sides
. then the branches of @xmath537 which are contained in @xmath625 decompose @xmath625 into triangles .
there is at least one such triangle @xmath626 with two sides contained in the boundary of @xmath625 , i.e. with two sides contained in @xmath187 . by the definition of a tangential measure ,
the total weight disposed by @xmath615 on the third side of @xmath626 which is contained in the interior of @xmath625 is bounded from above by the sum of the total weights on the sides of @xmath626 contained in @xmath187 .
thus by the above consideration , the total weight disposed by @xmath615 on the boundary of @xmath626 is uniformly bounded .
since the number of complementary components of @xmath537 only depends on the topological type of @xmath0 , with a uniformly bounded number of steps we obtain inductively the above claim ( note that the argument is also valid for once punctured complementary polygons of @xmath187 ) .
since the total weight of the tangential measure on @xmath537 defined by @xmath615 is uniformly bounded , the intersection number @xmath627 for every curve from a short framing for @xmath537 is uniformly bounded as well .
moreover , by corollary 2.3 of @xcite there is a constant @xmath628 with the following property .
let @xmath558 and let @xmath29 be a curve contained in a short framing for @xmath537 .
then for every @xmath629 we have @xmath630 .
now for every quadratic differential @xmath51 with horizontal and vertical measured geodesic lamination @xmath631 the @xmath632-length of a simple closed curve @xmath29 is bounded from above by @xmath633 .
this shows that the @xmath44-length of every simple closed curve from a short framing of @xmath537 is uniformly bounded and completes the proof of the lemma .
every projective measured geodesic lamination @xmath185 determines an unstable manifold @xmath634 of all area one quadratic differentials whose vertical measured geodesic lamination is contained in the class @xmath186 .
this unstable manifold can naturally be identified with the set of all measured geodesic laminations @xmath635 such that @xmath635 and @xmath186 jointly fill up @xmath0 . by the hubbard - masur theorem , the unstable manifold @xmath636 projects homeomorphically onto @xmath6 and hence the teichmller metric defines a distance function @xmath91 on @xmath636 . for a train track @xmath558
let @xmath637 be the set of all projective measured geodesic laminations which hit @xmath537 efficiently .
note that if @xmath638 then @xmath639 . for @xmath640 and a number @xmath641 the train track @xmath537 is called _ @xmath642-tight _ if the diameter of @xmath643 with respect to the lift of the teichmller metric is at most @xmath644 .
the next corollary will be used in section 5 .
[ tight ] let @xmath596 be as in lemma [ vertexcycle ] .
then for every @xmath191 there is a number @xmath645 with the following property .
let @xmath646 and assume that the distance in @xmath76 between any vertex cycle of @xmath134 and any vertex cycle of @xmath19 as well as any vertex cycle of @xmath537 is at least @xmath599 .
let @xmath568 be a quadratic differential whose horizontal measured geodesic lamination @xmath600 is carried by @xmath19 .
let @xmath640 be the projective class of the vertical measured geodesic lamination of @xmath44 .
if the total weight of the transverse measure on @xmath19 defined by @xmath600 is not smaller than @xmath193 then @xmath134 is @xmath642-tight .
let @xmath647 be such that @xmath648 for every @xmath558 .
by possibly decreasing @xmath649 we may assume that @xmath650 is connected .
if we equip @xmath650 with the length metric induced by the finsler structure defining the teichmller metric then @xmath650 is a proper geodesic metric space on which the mapping class group @xmath7 acts properly discontinuously and cocompactly as a group of isometries .
the set @xmath549 is the set of vertices of the _ train track complex _
@xcite which is a connected metric graph on which the mapping class group acts properly and cocompactly as a group of isometries .
since the map @xmath594 is coarsely @xmath7-equivariant this means that there is some @xmath89 such that @xmath651 is an @xmath276-quasi - isometry .
let @xmath646 be as in the corollary , let @xmath191 and assume that there is some @xmath568 such that the horizontal measured geodesic lamination @xmath600 of @xmath44 is carried by @xmath19 and that the weight disposed on @xmath19 by @xmath600 is bounded from below by @xmath193 .
then the weight @xmath29 disposed on @xmath134 by @xmath600 is contained in the interval @xmath652 $ ] . by lemma [ vertexcycle ] , applied both to the train tracks @xmath653 and to the train tracks @xmath654 ( with the quadratic differential @xmath655 for @xmath656 ) the distance between @xmath657 and @xmath658 is bounded from above by a number @xmath659 only depending on @xmath193 .
then the distance between @xmath657 and @xmath658 in @xmath650 is uniformly bounded as well and hence the same is true for the distance between @xmath660 in the train track complex .
however there are only finitely many orbits under the action of the mapping class group of pairs @xmath638 whose distance in the train track complex is uniformly bounded .
thus by invariance under the mapping class group , there is a universal number @xmath131 with the following property .
if @xmath82 is any vertex cycle of @xmath134 , then the total weight disposed on @xmath537 by @xmath82 is at most @xmath141 .
thus the carrying map @xmath661 maps @xmath662 to a subset of @xmath542 consisting of transverse measures whose total weight is bounded from above by a universal constant . as a consequence
, there is a universal constant @xmath659 such that if @xmath663 is _ any _ quadratic differential with the property that the horizontal measured geodesic lamination @xmath664 of @xmath632 is carried by @xmath134 then the total weight which is disposed by @xmath664 on @xmath134 is bounded from below by @xmath228 . by lemma [ vertexcycle ] , the the set of all quadratic differentials
@xmath665 whose horizontal measured geodesic lamination is carried by @xmath134 projects to a ball of uniformly bounded radius about @xmath658 . together with the above observations this shows the corollary .
in this section we complete the proof of the theorem from the introduction .
we continue to use the assumptions and notations from section 3 .
recall first that for every point @xmath143 and every @xmath185 there is a unique quadratic differential @xmath666 of area one on the riemann surface @xmath146 whose horizontal measured geodesic lamination @xmath667 is contained in the class @xmath186 .
the assignment @xmath668 determines a homeomorphism @xmath669 by assigning to @xmath670 the measured geodesic lamination @xmath671 .
a conformal density @xmath156 on @xmath37 of dimension @xmath82 defines a radon measure @xmath672 on @xmath9 via @xmath673 where @xmath674 . by construction , this measure is quasi - invariant under the one - parameter group of translations @xmath675 on @xmath676 given by @xmath677 .
more precisely , we have @xmath678 .
the measure @xmath672 is moreover invariant under the action of the mapping class group @xmath7 .
namely , for @xmath185 and @xmath345 the lamination @xmath679 equals @xmath680 where @xmath681 is the cocycle defined in the beginning of section 3 . on the other hand
, we have @xmath682 by the definition of a conformal density , for @xmath210-almost every @xmath185 the radon nikodym derivative of the measure @xmath683 with respect to @xmath210 at the point @xmath684 equals @xmath685 and therefore the measure @xmath672 is indeed invariant under the action of @xmath7 . as a consequence ,
every conformal density on @xmath37 induces a @xmath7-invariant radon measure on @xmath676 which is quasi - invariant under the one - parameter group of translations @xmath675 .
now let @xmath19 be any ergodic @xmath7-invariant radon measure on @xmath9 .
then @xmath686 is a closed subgroup of @xmath687 @xcite .
the next lemma is an easy consequence of proposition [ confonfilling ] and sarig s cocycle reduction theorem @xcite . for its formulation ,
recall from section 3 the definition of the space @xmath237 of all projective measured geodesic laminations which fill up @xmath0 .
we call a measured geodesic lamination @xmath688 _ filling _ if its projectivization is contained in @xmath237
. the set of all filling measured geodesic laminations is invariant under the mapping class group .
[ groupnontrivial ] let @xmath19 be an ergodic @xmath7-invariant radon measure on @xmath9 which gives full mass to the filling measured geodesic laminations .
if @xmath689 then @xmath19 coincides with the lebesgue measure up to scale . define a measurable countable equivalence relation @xmath690 on @xmath37 by @xmath691 if and only if @xmath94 and @xmath186 are contained in the same orbit for the action of the mapping class group .
recall the definition of the cocycle @xmath692 . for a fixed point @xmath143
we obtain a real - valued cocycle for the action of @xmath7 on @xmath37 , again denoted by @xmath681 , via @xmath693 @xmath694 and @xmath345 ) . by the cocycle identity ( [ cocycle ] ) for @xmath681
we have @xmath695 , i.e. @xmath681 is indeed a cocycle which can be viewed as a cocycle on @xmath690 .
we also write @xmath696 instead of @xmath697 whenever @xmath698 ; note that this is only well defined if @xmath20 is not fixed by any element of @xmath7 , however this ambiguity will be of no importance in the sequel .
recall that the choice of a point @xmath143 determines a homeomorphism @xmath699 .
the cocycle @xmath681 then defines an equivalence relation @xmath700 on @xmath701 by @xmath702 let @xmath19 be an ergodic @xmath7-invariant radon measure on @xmath9 which gives full mass to the filling measured geodesic laminations . by the results in @xcite , if @xmath703 then @xmath19 is induced by a conformal density as in the beginning of this section .
in particular , by proposition [ confonfilling ] , in this case the measure @xmath19 equals the lebesgue measure up to scale .
thus for the proof of our lemma we are left with the case that @xmath704 for a number @xmath705 . by the cocycle reduction theorem of sarig ( theorem 2 of @xcite ) , in this case there is a borel function @xmath706 such that @xmath707 holds @xmath19-almost everywhere in @xmath700 . since @xmath112
we may assume without loss of generality that the function @xmath708 is _
bounded_. following @xcite , for @xmath709 define @xmath710 . by lemma 2 of @xcite , for a suitable choice of @xmath553 the measure @xmath711 is an @xmath712-invariant ergodic radon measure supported on @xmath713 .
we now follow ledrappier and sarig @xcite ( see also @xcite ) .
namely , since @xmath19 is invariant and ergodic under the action of @xmath7 and since the @xmath687-action on @xmath9 commutes with the @xmath7-action , for every @xmath714 the measure @xmath715 is also @xmath7-invariant and ergodic .
thus either @xmath715 and @xmath19 are singular or they coincide up to scale . as a consequence , there is some number @xmath716 such that @xmath717 . since @xmath718 and @xmath719 commute
, we also have @xmath720 .
consequently the measure @xmath721 is invariant under the translation @xmath722 . since moreover @xmath723 is supported in @xmath713 , it follows that @xmath724 with some measure @xmath47 on @xmath37 .
the measure @xmath47 is finite since @xmath725 is radon and the function @xmath708 is bounded .
the measure @xmath19 is @xmath7-invariant and therefore @xmath726 is invariant under @xmath727 . in particular , the measure class of @xmath47 is invariant under the action of @xmath7 . more precisely , we have @xmath728 for all @xmath345 and @xmath47-almost every @xmath185 ( see @xcite ) . as a consequence ,
if we define @xmath729 and @xmath730 for @xmath731 then @xmath156 defines a conformal density of dimension @xmath82 on @xmath37 .
note that the measure @xmath732 is finite since the function @xmath708 is bounded . by our assumption on the measure @xmath19 , the conformal density @xmath161 gives full measure to the @xmath7-invariant set @xmath237 of projective measured laminations which fill up @xmath0 and
hence we conclude from proposition [ confonfilling ] that @xmath19 equals the lebesgue measure @xmath20 up to scale .
however , the lebesgue measure is quasi - invariant under the translations @xmath733 which is a contradiction to the assumption that @xmath704 for some @xmath112 .
this shows the lemma .
the investigation of @xmath7-invariant ergodic measures @xmath19 on @xmath9 which give full measure to the filling laminations and satisfy @xmath734 is more difficult .
we begin with an observation which is similar to proposition [ confonfilling ] . for this call a measured geodesic lamination _
weakly recurrent _ if its projectivization is contained in the set @xmath189 .
[ fillingrec ] an @xmath7-invariant radon measure @xmath19 on @xmath9 which gives full measure to the filling measured geodesic laminations gives full measure to the recurrent measured geodesic laminations .
let @xmath19 be an @xmath7-invariant ergodic radon measure on @xmath9 which gives full measure to the filling measured geodesic laminations .
we use the measure @xmath19 to construct a locally finite borel measure @xmath47 on @xmath58 which is invariant under the horocycle flow .
namely , for every @xmath49 the assignment which associates to a quadratic differential @xmath632 contained in the unstable manifold @xmath735 its horizontal measured geodesic lamination is a homeomorphism of @xmath735 into @xmath9 .
thus by equivariance under the action of the mapping class group , the measure @xmath19 lifts to a @xmath7-invariant family @xmath736 of locally finite measures on unstable manifolds @xmath735 @xmath57 .
this family of measures then projects to a family @xmath737 of locally finite borel measures on the leaves of the unstable foliation on @xmath58 .
the family @xmath737 is invariant under holonomy along strong stable manifolds .
define @xmath738 where @xmath739 is a standard family of lebesgue measures on strong stable manifolds which is invariant under the horocycle flow @xmath74 .
then @xmath47 is a locally finite @xmath74-invariant borel measure on @xmath58 . for @xmath47-almost every point @xmath180 the horizontal and the vertical measured geodesic laminations of @xmath44 fill up @xmath0 . as in the proof of proposition [ confonfilling ] , dani s
argument together with the theorem from the appendix implies that @xmath47 is _
finite_. moreover , for every @xmath191 there is a compact subset @xmath740 of @xmath58 not depending on @xmath47 such that @xmath741 .
via replacing @xmath47 by @xmath742 we may assume that @xmath47 is a probability measure .
for @xmath743 define @xmath744 since @xmath745 for all @xmath746 , the borel probability measure @xmath747 is @xmath74-invariant and gives full measure to the points with filling horizontal measured geodesic laminations .
therefore we have @xmath748 for all @xmath743 , all @xmath191 .
this implies that there is a sequence @xmath749 such that the measures @xmath750 converge as @xmath751 weakly to a borel probability measure @xmath752 on @xmath58 which is invariant under both the horocycle flow and the teichmller geodesic flow . by the poincar recurrence theorem
, @xmath752 gives full measure to the forward _ recurrent _ quadratic differentials . as a consequence , @xmath47-almost
every @xmath180 contains a forward recurrent point in its @xmath187-limit set .
namely , for @xmath191 there is a compact subset @xmath753 of @xmath58 which consists of forward recurrent points and such that @xmath754 .
let @xmath755 be a family of open neighborhoods of @xmath753 such that @xmath756 for all @xmath599 and @xmath757 .
then for every @xmath596 there is some @xmath758 such that @xmath759 for all @xmath760 . for @xmath596 define @xmath761 for infinitely many @xmath762
; then @xmath763 for all @xmath599 .
we claim that @xmath764 for every @xmath599 .
namely , otherwise there is a number @xmath284 and there is a subset @xmath99 of @xmath58 with @xmath765 and such that @xmath766 for every @xmath767 and every @xmath768 .
then necessarily @xmath769 for all sufficiently large @xmath255 which is impossible . since the neighborhood @xmath770 of @xmath753 was arbitrary and
since @xmath47 is borel regular we conclude that @xmath771 .
thus the @xmath47-mass of all points @xmath180 which contain a point @xmath772 in its @xmath187-limit set is at least @xmath773 .
since @xmath753 consists of recurrent points and since @xmath191 was arbitrary we conclude that @xmath47-almost every @xmath180 contains a forward recurrent point in its @xmath187-limit set .
this shows the lemma . for the analysis of @xmath7-invariant radon measures
@xmath19 on @xmath9 with @xmath774 we use a construction reminiscent of symbolic dynamics where the markov shift is replaced by complete train tracks and their splits .
we next establish some technical preparations to achieve this goal . define a geodesic lamination @xmath186 on @xmath0 to be _ complete _ if @xmath186 is maximal and can be approximated in the hausdorff topology by simple closed geodesics .
the space @xmath775 of all complete geodesic laminations equipped with the hausdorff topology is a compact @xmath7-space @xcite .
a transversely recurrent generic train track is complete if and only if it carries a complete geodesic lamination .
every minimal geodesic lamination @xmath20 is a sublamination of a complete geodesic lamination .
if @xmath20 fills up @xmath0 then the number of complete geodesic laminations which contain @xmath20 as a sublamination is bounded from above by a universal constant ( for all this , see @xcite ) .
denote by @xmath776 the set of all complete geodesic laminations @xmath777 which contain a uniquely ergodic minimal component which fills up @xmath0 .
the mapping class group @xmath7 naturally acts on @xmath778 as a group of transformations .
there is a finite - to - one @xmath7-equivariant map @xmath779 which associates to every @xmath780 the unique projective measured geodesic lamination which is supported in @xmath20 .
the number of preimages in @xmath778 of a point in @xmath237 is bounded from above by a universal constant .
a _ full split _ of a complete train track @xmath537 is a complete train track @xmath134 which can be obtained from @xmath537 by splitting @xmath537 at each large branch precisely once .
full splitting sequence _ is a sequence @xmath781 such that for each @xmath255 , the train track @xmath782 can be obtained from @xmath783 by a full split . for a complete train track @xmath537 denote by @xmath784 the set of all complete geodesic laminations which are carried by @xmath537
then @xmath784 is a subset of @xmath775 which is both open and closed . if @xmath785 is an infinite full splitting sequence then @xmath786 consists of a unique point . for every complete train track @xmath537 and every complete geodesic lamination @xmath787
there is a unique full splitting sequence @xmath788 issuing from @xmath789 such that @xmath790 ( for all this , see @xcite ) . for @xmath558 let @xmath791 be the set of all measured geodesic laminations @xmath47
whose support is carried by @xmath537 and such that the total mass of the transverse measure on @xmath537 defined by @xmath47 equals one .
if the complete geodesic lamination @xmath780 is carried by @xmath537 then there is a unique measured geodesic lamination @xmath792 whose support is contained in @xmath20 .
there is a number @xmath420 only depending on the topological type of @xmath0 such that for all @xmath793 and all @xmath794 we have @xmath795 for some @xmath796 $ ] .
namely , @xmath797 is obtained from @xmath798 by a uniformly bounded number of splits . moreover ,
if @xmath799 is obtained from @xmath537 by a split at a large branch @xmath552 , with losing branches @xmath554 , and if @xmath47 is any transverse measure on @xmath19 , then @xmath47 projects to a transverse measure @xmath178 on @xmath537 with the following properties . using the natural identification of the branches of @xmath537 with the branches of @xmath19 , for every branch @xmath800 of @xmath537 ,
the @xmath190-weight of @xmath79 coincides with the @xmath47-weight of the branch @xmath79 in @xmath19 .
moreover , we have @xmath801 . as before ,
let @xmath802 be the set of all area one quadratic differentials whose horizontal measured geodesic lamination is contained in @xmath562 and whose vertical measured geodesic lamination hits @xmath537 efficiently . as in section 4 , call @xmath558 _
@xmath642-tight _ for a number @xmath641 and some @xmath640 if the diameter of @xmath803 with respect to the lift @xmath91 of the teichmller metric is at most @xmath644 . for
@xmath787 denote by @xmath788 the full splitting sequence issuing from @xmath789 and determined by @xmath20 .
let @xmath217 be the number of orbits of the action of @xmath7 on @xmath549 .
the next technical observation is the main ingredient for the completion of the proof of the theorem from the introduction .
it roughly says that controlled recurrence implies tightness .
[ tightcon ] let @xmath558 and let @xmath804 be the lift of a forward recurrent point @xmath188 .
let @xmath186 be the vertical projective measured geodesic lamination of @xmath329 .
there are numbers @xmath805 depending on @xmath329 and for every complete geodesic lamination @xmath806 with @xmath807 there is a sequence @xmath503 with the following properties . 1
. for each @xmath255 there are numbers @xmath808 , @xmath809 and there are numbers @xmath810,t\in [ t_i+k , t_i+k+ma]$ ] such that @xmath811 , @xmath812 and such that the train tracks @xmath813 are @xmath642-tight .
2 . for every @xmath255
there is some @xmath505 such that @xmath814 . by lemma [ simple2 ] and inequality
( [ lipschitz ] ) in section 2 , for all @xmath417 with @xmath815 ( where @xmath420 is as above ) , the distance in @xmath76 between @xmath104 and @xmath816 is bounded from above by a universal constant @xmath96 .
let moreover @xmath617 be the maximal distance in @xmath76 between any two vertex cycles of any complete train tracks @xmath817 on @xmath0 so that either @xmath818 or that @xmath819 can be obtained from @xmath820 by a full split . let @xmath821 be as in lemma [ upsilon ] and let @xmath596 be as in lemma [ vertexcycle ] . by assumption ,
the @xmath187-limit set of the @xmath42-orbit of @xmath312 contains @xmath312 . by lemma 2.2 and the results of @xcite we have @xmath822 @xmath213 .
thus we can find small open relative compact neighborhoods @xmath823 of @xmath329 in @xmath41 , numbers @xmath824 and mapping classes @xmath505 such that for all @xmath825 we have @xmath826 and @xmath827 let for the moment @xmath825 be an arbitrary quadratic differential with @xmath828 and denote by @xmath829 the horizontal measured geodesic lamination and the vertical measured geodesic lamination of @xmath44 , respectively .
let @xmath830 be a train track which carries @xmath600 and hits @xmath615 efficiently .
we assume that there is some @xmath831 $ ] with @xmath832 .
assume moreover that there is a complete geodesic lamination @xmath833 such that @xmath834 is the projective class of @xmath600 .
then @xmath20 determines a full splitting sequence @xmath835 issuing from @xmath836 . for @xmath837 and for @xmath838 as above we can find some @xmath839 $ ] such that @xmath840 , which is equivalent to saying that @xmath841 . by our choice of @xmath842
we have @xmath843 thus by lemma [ upsilon ] and our choice of @xmath844 , for all @xmath845 $ ] the distance in @xmath76 between any vertex cycle of @xmath846 and any vertex cycle of @xmath847 is at least @xmath599 . moreover , there is some @xmath848 $ ] such that @xmath849 and hence the total mass that the horizontal measured geodesic lamination @xmath850 disposes on the complete train track @xmath846 is bounded from below by @xmath851
. therefore corollary [ tight ] , applied to the train tracks @xmath852 , shows the existence of a universal constant @xmath641 such that the train track @xmath853 is @xmath854$]-tight where @xmath855 $ ] is the projective class of the vertical measured geodesic lamination of @xmath44 .
note that the number @xmath641 only depends on the point @xmath329 but not on @xmath825 or the train track @xmath830 with @xmath832 . for this number @xmath641 , choose a number @xmath856 such that @xmath857 and that @xmath858 for some @xmath859 ; such a number exists since the projection @xmath312 of @xmath329 to @xmath58 is forward recurrent .
choose an open neighborhood @xmath860 of @xmath329 such that @xmath861 for all @xmath862 and @xmath863 .
write @xmath864 and @xmath865 ; then we have @xmath866 for all @xmath862 and every @xmath867 .
define @xmath868 and note that @xmath585 only depends on @xmath329 . as above , let @xmath558 be a train track which carries the horizontal projective measured geodesic lamination of @xmath329 and such that the vertical projective measured geodesic lamination @xmath186 of @xmath329 hits @xmath537 efficiently .
let @xmath869 be such that @xmath807 .
let @xmath665 be such that the horizontal projective measured geodesic lamination of @xmath44 equals @xmath834 .
for the neighborhood @xmath870 of @xmath329 as above , the set @xmath871 is unbounded .
let @xmath872 be a sequence tending to infinity such that @xmath873 for some @xmath874 and every @xmath875 .
let @xmath788 be the full splitting sequence determined by @xmath876 and @xmath20 .
for @xmath875 we have @xmath873 .
there is a number @xmath831 $ ] and a number @xmath877 such that @xmath878 . using the above constants
@xmath838 , there are numbers @xmath879 and numbers @xmath880 $ ] such that @xmath881 @xmath882 . since there are only @xmath883 distinct orbits of complete train tracks under the action of the mapping class group , there are moreover numbers @xmath884 $ ] and @xmath885 $ ] and there is some @xmath345 such that @xmath886 . by the above consideration , the train tracks @xmath887 are @xmath642-tight .
then the sequence @xmath888 and the numbers @xmath889 have the required properties stated in the lemma with @xmath890 .
now we are ready to complete the main step in the proof of the theorem from the introduction .
[ fillingisleb ] an @xmath7-invariant radon measure on @xmath9 which gives full mass to the filling measured geodesic laminations coincides with the lebesgue measure up to scale . by lemma [ groupnontrivial ] and lemma [ fillingrec ]
, we only have to show that there is no @xmath7-invariant ergodic radon measure on @xmath9 with @xmath734 which gives full mass to the recurrent measured geodesic laminations . for this
we argue by contradiction and we assume that such a radon measure @xmath19 exists .
using once more the cocycle reduction theorem of sarig , there is a borel function @xmath706 such that the measure @xmath19 gives full mass to the graph @xmath891 of the function @xmath708 . in particular , @xmath19 projects to an @xmath7-invariant measure class @xmath892 on @xmath37 which gives full mass to the set @xmath189 of recurrent points . using the notations from proposition [ vitali ] and its proof ,
this means that there is a forward recurrent point @xmath188 such that the measure class @xmath892 on @xmath37 gives full measure to the set @xmath313 . to derive a contradiction we adapt the arguments of ledrappier and sarig @xcite to our situation .
denote again by @xmath776 the set of all complete geodesic laminations containig a minimal component which fill up @xmath0 and let @xmath893 be as before . every finite borel measure @xmath46 on @xmath37 which gives full mass to the returning projective measured geodesic laminations
induces a finite borel measure @xmath894 on @xmath778 .
namely , returning projective measured geodesic laminations are uniquely ergodic and hence for a borel subset @xmath315 of @xmath778 we can define @xmath895 let again @xmath188 be a forward recurrent quadratic differential such that the measure class @xmath892 gives full mass to @xmath313 .
let @xmath327 be a lift of @xmath312 .
we may assume that the vertical measured geodesic lamination @xmath896 of @xmath329 is uniquely ergodic and fills up @xmath0 .
we may moreover assume that there is a train track @xmath558 with @xmath897 ( as before , this can be achieved by possibly replacing @xmath329 by @xmath898 for some @xmath294 ) . by the considerations in section 4
we may assume that @xmath537 is @xmath899-tight for a number @xmath900 .
let @xmath901 be such that @xmath902 contains some density point of the locally finite measure @xmath903 on @xmath904 induced from @xmath19 whose horizontal measured geodesic lamination is carried by the train track @xmath537 ; such a number exists since by invariance under the mapping class group , the @xmath7-invariant measure class @xmath905 on @xmath37 is of full support and since moreover the set of all measured geodesic laminations carried by @xmath537 has non - empty interior .
let @xmath906 be the restriction of the measure @xmath903 to the set @xmath907 and denote by @xmath908 the projection of @xmath906 to @xmath37 .
since @xmath19 is a radon measure on @xmath9 by assumption , the measure @xmath908 is a locally finite borel measure on @xmath909 which gives full mass to @xmath313 .
since @xmath537 is @xmath899-tight for some @xmath900 , the set @xmath803 is relative compact and therefore the intersection of the support of @xmath906 with @xmath910 is compact as well .
since @xmath19 and hence @xmath906 is radon , the total @xmath908-mass of the set of all projective measured geodesic laminations which are carried by the train track @xmath537 is finite . by the above consideration , @xmath908 induces a finite nontrivial borel measure @xmath911 on @xmath784 which gives full mass to the set of complete geodesic laminations @xmath912 with @xmath913 .
similarly , for the constants @xmath914 as in lemma [ tightcon ] , let @xmath915 be the finite borel measure on @xmath784 which is induced from the restriction of @xmath903 to @xmath916 . since @xmath774 by assumption , the measures
@xmath917 are singular .
define a _ cylinder _ in @xmath784 to be a set of the form @xmath918 where @xmath830 is a complete train track which can be obtained from @xmath537 by a full splitting sequence .
a cylinder is a subset of @xmath775 which is both open and closed @xcite .
the intersection of two cylinders is again a cylinder .
since every point in @xmath784 is an intersection of countably many cylinders , the @xmath134-algebra on @xmath784 generated by cylinders is the usual borel @xmath134-algebra .
now @xmath919 are mutually singular borel measures on @xmath784 and hence there is a cylinder @xmath920 such that @xmath921 . by lemma [ tightcon ] , for @xmath911-almost every @xmath833 there is a sequence @xmath922 and there is some @xmath923 with the following properties . 1
. the train tracks @xmath924 are @xmath642-tight .
2 . @xmath925 .
3 . for @xmath926
there is some @xmath927 $ ] such that @xmath928 .
these properties imply the following .
let @xmath929 be such that the horizontal measured geodesic lamination @xmath600 of @xmath44 is carried by @xmath930 .
let @xmath931 be such that @xmath932 and let @xmath933 ; then the horizontal measured geodesic lamination of @xmath632 is carried by @xmath930 , and we have @xmath934 for some @xmath935 $ ] .
since @xmath936 it is now immediate from invariance of the measure @xmath19 under the action of @xmath7 , from the definitions of the measures @xmath919 and the fact that the action of @xmath7 on @xmath9 commutes with the action of the group of translations that @xmath937 . on the other hand , there is a countable partition of a subset of @xmath918 of full @xmath911-mass into cylinders @xmath938 with train tracks @xmath939 @xmath940 which can be obtained from @xmath134 by a full splitting sequence and which satisfy 1),2),3 ) above .
this partition can inductively be constructed as follows . beginning with the train track @xmath134
, there is a full splitting sequence of minimal length @xmath941 issuing from @xmath134 which connects @xmath134 to some train track @xmath942 with the above properties .
let @xmath943 be the collection of all train tracks which can be obtained from @xmath134 by a full splitting sequence of length @xmath944 and assume after reordering that we have @xmath945 .
repeat this construction simultaneously with the train tracks @xmath946 .
after countably many steps we obtain a partition of @xmath947-almost all of @xmath918 as required .
together we conclude that necessarily @xmath948 which contradicts our choice of @xmath134 . in other words ,
the case @xmath774 is impossible which completes the proof of our proposition .
* remark : * let @xmath188 be a forward recurrent point . the arguments in the proof of proposition [ fillingisleb ] can be used to construct a vitali relation for the lift to @xmath775 of any borel measure on @xmath313 . in other words , with some extra arguments , proposition [ vitali ] can be deduced from proposition [ fillingisleb ] and its proof .
however , we included proposition [ vitali ] in the present form since its basic idea is simpler and more geometric , moreover it is used in @xcite .
choose a complete hyperbolic metric on @xmath0 of finite volume .
let @xmath22 be a proper connected bordered subsurface of @xmath0 with geodesic boundary .
then @xmath22 has negative euler characteristic .
we allow that distinct boundary components of @xmath22 are defined by the same simple closed geodesic in @xmath0 .
denote by @xmath949 ( or @xmath950 ) the subgroup of @xmath7 of all elements which can be represented by a diffeomorphism fixing @xmath24 ( or @xmath22 ) pointwise .
the stabilizer @xmath951 of @xmath22 in @xmath7 contains a subgroup of finite index of the form @xmath952 where @xmath953 is the free abelian group of dehn twists about the geodesics in @xmath0 which define the boundary of @xmath22 .
let @xmath954 ( or @xmath955 ) be the surface of finite type which we obtain from @xmath22 ( or @xmath956 by collapsing each boundary circle to a puncture .
the space @xmath26 of all measured geodesic laminations on @xmath954 can be identified with the space of all measured geodesic laminations on @xmath0 whose support is contained in the interior of @xmath22 .
we say that a measured geodesic lamination @xmath957 _ fills _
@xmath22 if its support is minimal and intersects every simple closed geodesic contained in the interior of @xmath22 transversely .
there is a ray of lebesgue measures on @xmath26 which are invariant under the mapping class group @xmath958 of the surface @xmath954 and hence it is invariant under @xmath951 . for every @xmath959 ,
the image @xmath960 of @xmath22 under @xmath69 is a subsurface of @xmath0 which is distinct from @xmath22 .
the image @xmath961 under @xmath69 of a measured geodesic lamination @xmath635 on @xmath22 which fills @xmath22 is a measured geodesic lamination which fills @xmath960 and hence this image is not contained in @xmath26 . note that we have @xmath962 .
now let @xmath29 be any ( possibly trivial ) simple weighted geodesic multicurve on @xmath0 which is disjoint from the interior of @xmath22 .
then for every @xmath30 the union @xmath31 is a measured geodesic lamination on @xmath0 in a natural way which we denote by @xmath32 .
thus @xmath963 is naturally a closed subspace of @xmath9 .
this subspace can be equipped with a @xmath964-invariant ergodic radon measure @xmath965 induced by a measure @xmath908 from our ray of @xmath966-invariant lebesgue measures on @xmath26 . by invariance , we obtain a @xmath7-invariant ergodic wandering measure on @xmath9 by defining @xmath967 we call @xmath968 a _ standard subsurface measure _ of @xmath22 .
if the support of the weighted multi - curve @xmath29 contains every boundary component of @xmath22 then we call the resulting @xmath7-invariant measure @xmath968 on @xmath9 a _ special standard subsurface measure _ on @xmath9 .
recall from the introduction that a _ rational _ @xmath7-invariant measure on @xmath9 is a sum of weigthed dirac masses supported on the orbit of a simple weighted multi - curve .
such a rational measure is a special standard subsurface measure on @xmath9 ( for the empty subsurface ) .
we have . [ localfinite ]
a special standard subsurface measure on @xmath9 is locally finite .
let @xmath969 be any complete hyperbolic metric on @xmath0 of finite volume .
then for every measured geodesic lamination @xmath46 on @xmath0 the _ @xmath969-length _
@xmath970 of @xmath46 is defined . by definition , this length is the total mass of the measure on @xmath0 which is the product of the transverse measure for @xmath46 and the hyperbolic length element on the geodesics contained in the support of @xmath46 . for every _
compact _ subset @xmath181 of @xmath9 there is a number @xmath217 such that @xmath971 . to show the lemma observe first that for every weighted geodesic multi - curve @xmath29 on @xmath0 and every @xmath217 the set @xmath972 contains only finitely many images of @xmath29 under the action of the mapping class group .
namely , let @xmath420 be the minimal weight of a component of the support of @xmath29 .
then for every @xmath973 the @xmath969-length of the multi - curve @xmath974 is not smaller than @xmath553 times the maximal length of any closed geodesic on the hyperbolic surface @xmath975 which is freely homotopic to a component of @xmath974 .
however , there are only finitely many simple closed geodesics on @xmath0 whose @xmath969-length is at most @xmath976 and hence the intersection of @xmath972 with the @xmath7-orbit of @xmath29 is indeed finite .
in particular , a rational @xmath7-invariant measure on @xmath9 is radon .
now let @xmath22 be a bordered connected subsurface of @xmath0 of negative euler characteristic and geodesic boundary .
let @xmath29 be a simple weighted geodesic multi - curve which contains every boundary component of @xmath22 .
then the stabilizer @xmath977 of @xmath29 in @xmath7 contains the stabilizer @xmath978 of @xmath35 as a subgroup of finite index .
in particular , a @xmath979-invariant measure on @xmath26 in the lebesgue measure class induces an ergodic @xmath977-invariant radon measure on the space of laminations containing @xmath29 as a component . on the other hand , if @xmath980 _ does not _
stabilize @xmath29 then @xmath69 moves at least one component of @xmath29 away from @xmath29 .
therefore by the above consideration , there are only _ finitely many _ cosets in @xmath981 containing some representative @xmath69 such that @xmath982 .
this shows that a special subsurface measure on @xmath9 is radon .
finally we are able to complete the proof of the theorem from the introduction .
let @xmath19 be an @xmath7-invariant ergodic radon measure on the space @xmath9 of all measured geodesic laminations on @xmath0 . 1 .
if @xmath19 is non - wandering then @xmath19 is the lebesgue measure up to scale .
2 . if @xmath19 is wandering then either @xmath19 is rational or @xmath19 is a standard subsurface measure .
let @xmath19 be an ergodic @xmath7-invariant radon measure on @xmath9 . by proposition [ fillingisleb ] ,
if @xmath19 gives full mass to the measured geodesic laminations which fill up @xmath0 then @xmath19 coincides with the lebesgue measure up to scale .
thus by ergodicity and invariance we may assume that @xmath19 gives full mass to the measured geodesic laminations which do not fill up @xmath0 .
let @xmath20 be a density point for @xmath19 .
the support of @xmath20 is a union of components @xmath983 .
we assume that these components are ordered in such a way that there is some @xmath984 such that the components @xmath985 are minimal arational and that the components @xmath986 are simple closed curves . by ergodicity
, the same decomposition then holds for @xmath19-almost every @xmath987 . if @xmath988 then by ergodicity , @xmath19 is rational and there is nothing to show .
thus assume that @xmath989 .
then @xmath990 _ fills _ a subsurface @xmath22 of @xmath0 with @xmath599 connected components , each of which is of negative euler characteristic .
in particular , by ergodicity the measure @xmath19 is wandering and gives full mass to the set @xmath991 .
write @xmath992 . since @xmath20 is a density point for @xmath19 , the restriction of @xmath19 to @xmath963 does not vanish .
however , this restriction is a @xmath978-invariant radon measure on @xmath949 which gives full mass to the measured geodesic laminations filling up @xmath22 and therefore this restriction is an interior point of our cone of @xmath978-invariant lebesgue measure on @xmath963 . in other words
, @xmath19 is a standard subsurface measure .
this completes the proof of the theorem .
* remark : * in @xcite , lindenstrauss and mirzakhani obtain a stronger result .
they show that a locally finite standard subsurface measure on @xmath9 is special .
the purpose of this appendix is to present some results from the paper @xcite of minsky and weiss in the form needed in section 3 . as in section 2 , denote by @xmath58 the moduli space of area one holomorphic quadratic differentials on @xmath0 .
every @xmath180 defines an isometry class of a singular euclidean metric on @xmath0 .
the set @xmath250 of singular points for this metric coincides precisely with the set of zeros for @xmath44 .
we also assume that the differential has a simple pole at each of the punctures of @xmath0 and hence it can be viewed as a meromorphic quadratic differential on the compactified surface @xmath64 which we obtain by filling in the punctures in the standard way .
a _ saddle connection _ for @xmath44 is a path @xmath993 whose image in each chart is an euclidean straight line and which extends continuously to a path @xmath994\to \hat s$ ] mapping the endpoints to singularities or punctures .
a saddle connection does not have self - intersections .
two saddle connections @xmath995 are _ disjoint _ if @xmath996 the closure of any finite collection of pairwise disjoing saddle connections on @xmath0 is an embedded graph in @xmath0 . by proposition 4.7 of @xcite
( see also @xcite ) , the number of pairwise disjoint saddle connections for a quadratic differential @xmath180 is bounded from above by a universal constant @xmath997 only depending on the topology of @xmath0 .
recall that a _ tree _ is a graph without circuits . for @xmath191
let @xmath998 be the set of all quadratic differentials @xmath44 such that the collection of all saddle connections of @xmath44 of length at most @xmath193 is a tree .
we have .
thus let @xmath29 be any simple closed geodesic on @xmath0 for the @xmath44-metric .
then up to replacing @xmath29 by a freely homotopic simple closed curve of the same length we may assume that @xmath29 consists of a sequence of saddle connections for @xmath44 .
since the set of saddle connections of length at most @xmath193 does not contain a circuit , the curve @xmath29 contains at least one saddle connection of length at least @xmath193 .
but this just means that the @xmath44-length of @xmath29 is at least @xmath193 as claimed .
the following proposition is a modified version of theorem 6.3 of @xcite .
we use the notations from @xcite .
let @xmath1000 be the set of all saddle connections of the quadratic differential @xmath44 .
for @xmath1001 define @xmath1002 denote again by @xmath74 the horocycle flow on @xmath58 . for @xmath1003 and @xmath714 define @xmath1004 where @xmath1005 is the length of @xmath649 with respect to the singular euclidean metric defined by @xmath1006 . for @xmath1007 let @xmath1008 there are positive constants @xmath1009 depending only on @xmath0 with the following property .
let @xmath180 , let @xmath1010 be an interval and let @xmath1011 .
define @xmath1012 if @xmath1013 is an embedded tree with @xmath1014 edges then for any @xmath1015 we have : @xmath1016 let @xmath997 be such that for every @xmath180 the number of pairwise disjoint saddle connections of @xmath44 is bounded from above by @xmath1017 . by proposition 6.1 of @xcite there is a number @xmath1018 with the following property . if @xmath1003 is such that the closure @xmath1019 of the union of all simply connected components of @xmath1020 is all of @xmath0 then @xmath1021 .
let @xmath180 and let @xmath1022 be a union of pairwise disjoint saddle connections whose closure is an embedded graph in @xmath0 without circuits .
assume that @xmath99 consists of @xmath1014 segments .
we necessarily have @xmath1023 . for a number @xmath1024 to be determined later let @xmath1025 and let @xmath1026 for @xmath1027 define @xmath1028 we choose @xmath1024 in such a way that @xmath1029 . for @xmath1030 let @xmath1031 and let @xmath1032 .
then @xmath1033 for all @xmath1034 and hence @xmath1035 is the disjoint union of the measurable sets @xmath1036 .
thus there is some @xmath1037 for which @xmath1038 for this choice of @xmath585 define @xmath1039 and @xmath1040 .
note that we have @xmath1041 following the argument in section 6 of @xcite we only have to verify that @xmath1049 .
namely , let @xmath1050 and let @xmath1003 be such that @xmath1051 .
denote by @xmath1019 the closure of the union of the simply connected components of @xmath1020 . by proposition 6.1 of @xcite
we have @xmath1052 and hence since @xmath1053 and the graph defined by the saddle connections contained in @xmath99 does not have circuits , the boundary of @xmath1019 contains at least one saddle connection @xmath649 which is _ not _ contained in @xmath99 .
but this just means that @xmath1054 ( see claim 6.7 in @xcite ) .
this complete the proof of the proposition .
let @xmath44 be a quadratic differential with horizontal measured geodesic lamination which fills up @xmath0 .
then the horizontal saddle connections of @xmath44 form an embedded graph without circuits .
morever , the number of these saddle connections is bounded from above by a universal constant
. now if @xmath649 is any saddle connection whose length is constant along the horocycle flow then @xmath649 is horizontal .
but this just means that we can apply the above proposition as in the proof of theorem h2 of @xcite to obtain the theorem . r. canary , d. epstein , p. green , _ notes on notes of thurston _
, in `` analytical and geometric aspects of hyperbolic space '' , edited by d. epstein , london math .
lecture notes 111 , cambridge university press , cambridge 1987 .
u. hamenstdt , _
train tracks and the gromov boundary of the complex of curves _ , in `` spaces of kleinian groups '' ( y. minsky , m. sakuma , c. series , eds . ) , london math . soc .
notes 329 ( 2006 ) , 187207 . | let @xmath0 be an oriented surface of genus @xmath1 with @xmath2 punctures and @xmath3 .
we classify all radon measures on the space of measured geodesic laminations which are invariant under the action of the mapping class group of @xmath0 . |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.